<?page no="0"?> ISBN 978-3-8169-3544-5 MARC SENS (ED.) Ignition Systems for Gasoline Engines Knocking in Gasoline Engines For decades, scientists and engineers have been working to increase the efficiency of internal combustion engines. For spark-ignition engines, two technical questions in particular are always in focus: 1. How can the air/ fuel mixture be optimally ignited under all possible conditions? 2. How can undesirable but recurrent early and self-ignitions in the air/ fuel mixture be avoided? Against the background of the considerable efficiency increases currently being sought in the context of developments and the introduction of new fuels, such as hydrogen, methanol, ammonia and other hydrogen derivatives as well as biofuels, these questions are more in the focus than ever. In order to provide a perfect exchange platform for the community of combustion process and system developers from research and development, IAV has organized this combined conference, chaired by Marc Sens. The proceedings presented here represent the collection of all the topics presented at the event and are thus intended to serve as an inspiration and pool of ideas for all interested parties. MARC SENS (ED.) International Conference on Ignition Systems for Gasoline Engines International Conference on Knocking in Gasoline Engines <?page no="1"?> International Conference on Ignition Systems for Gasoline Engines - International Conference on Knocking in Gasoline Engines <?page no="3"?> Marc Sens (ED.) International Conference on Ignition Systems for Gasoline Engines - International Conference on Knocking in Gasoline Engines <?page no="4"?> DOI: https: / / doi.org/ 10.24053/ 9783816985440 © expert verlag 2022 ‒ ein Unternehmen der Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetztes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikro‐ verfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Alle Informationen in diesem Buch wurden mit großer Sorgfalt erstellt. Fehler können dennoch nicht völlig ausgeschlossen werden. Weder Verlag noch Autor: innen oder Herausgeber: innen übernehmen deshalb eine Gewährleistung für die Korrektheit des Inhaltes und haften nicht für fehlerhafte Angaben und deren Folgen. Diese Publikation enthält gegebenenfalls Links zu externen Inhalten Dritter, auf die weder Verlag noch Autor: innen oder Herausgeber: innen Einfluss haben. Für die Inhalte der verlinkten Seiten sind stets die jeweiligen Anbieter oder Betreibenden der Seiten verantwortlich. Internet: www.expertverlag.de eMail: info@verlag.expert printed in Germany ISBN 978-3-8169-3544-5 (Print) ISBN 978-3-8169-8544-0 (ePDF) ISBN 978-3-8169-0112-9 (ePub) Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. www.fsc.org MIX Papier aus verantwortungsvollen Quellen FSC ® C083411 ® www.fsc.org MIX Papier aus verantwortungsvollen Quellen FSC ® C083411 ® <?page no="5"?> 11 25 41 63 81 97 111 145 175 Inhalt Dr. Frank Altenschmidt, Dr. Eberhard Kraus; Mercedes-Benz AG, Germany Knock in SI-Engines - A continuing challenge for combustion system development Jan Reimer, Institut für Kolbenmaschinen, Karlsruher Institut für Technologie(KIT), Mitra Zabihigivi (KIT), Ina Volz (Mercedes Benz AG), Jürgen Pfeil (KIT), Frank Altenschmidt (Mercedes Benz AG), Thomas Koch (KIT) Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dongwon Jung (Hyundai Motor Company), Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Stability Limit in a Gasoline Engine . . . . . . . . . . . . . . . Tobias Michler, Olaf Toedter, Thomas Koch Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch Application of a time-resolved ignition spark measurement technique when using a power ignition system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dr.-Dipl.-Ing. Stephan Herbst Heraeus Deutschland GmbH & Co. KG Dipl.-Ing. Patrick Baake Heraeus Deutschland GmbH & Co. KG Dr.-Ing. Thomas Emmrich IAV GmbH Spark erosion tests on materials for spark plug electrodes . . . . . . . . . . . . . . . . . . . . Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang Advanced Ignition Strategies for Gasoline Engine Clean Combustion . . . . . . . . . . . R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma - Esgee Technologies, L. Raja - The University of Texas at Austin Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <?page no="6"?> 223 255 273 275 303 329 349 377 399 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi Innovative prechamber system with valve for future high efficiency engines . . . . Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . prechamber of a TJI engine by mean of detailed CFD simulations. . . . . . . . . . . . . . Tim Russwurm, Tobias Achenbach, Michael Wensing Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Antonino Vacca, Marco Chiodi, André Casal Kulzer, Michael Bargende, Sebastian Bucherer, Paul Rothe, Ivica Kraljevic, Hans-Peter Kollmeier, Albert Breuer, Ruhland Helmut Engines using 3D-CFD Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M. Balmelli, L. Merotto, P.Soltic Discharge in an Optical Accessible Pre-chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Marco Hess, Michael Grill, Michael Barende, André Casal Kulzer 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines . . . . . . . . . . . Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Inhalt Evaluation of the lean limit extension provided by H2 direct injection inside the Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi Alfio Siliato, Riccardo Sgarangella, Michela Fabbri, Leonardo Pulga, Claudio Forte, Gian Marco Bianchi <?page no="7"?> 421 439 461 477 495 525 537 553 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine . . . . . . . . . . . . . L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Michael Wörner, Michael Auerbach, Gregor Rottenkolber Esslingen University of Applied Sciences, Esslingen, Germany Injection during compression stroke for engine knock prevention . . . . . . . . . . . . . Sascha Holzberger, Maurice Kettner, Karlsruhe University of Applied Sciences Roland Kirchberger, Graz University of Technology Ivica Kraljevic, Florian Sobek, Fraunhofer Institute for Chemical Technology Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lukas Euchner, M.Sc, BMW Group; Laura Baumgartner, Dr.-Ing., BMW Group; Michael Wensing, Prof. Dr.-Ing., Friedrich-Alexander-Universität Erlangen-Nürnberg; Tim Russwurm, M. Sc., Friedrich-Alexander-Universität Erlangen-Nürnberg; Peter Janas, Dr.-Ing., Tenneco, Inc. Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio . . . . . . . . . . . . . . . . . . . . Matthias Biehl / Marc Benzinger Robert Bosch GmbH Holistic knock detection and control as the key to optimum ignition timing . . . . . 7 Inhalt M.Sc. Fabian Steeger, Dr.-Ing. Marco Günther, Dr.-Ing. Eike Stitterich, Prof. Dr.-Ing. Stefan Pischinger Nicolas Fajt, IFS - Institut für Fahrzeugtechnik, Universität Stuttgart Co-authors: M. Grill, M. Bargende Knock Probability Prediction and its Potential for a Knock Control Application . . Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <?page no="9"?> About Knock and Ignition Ignition System Basics Ignition System Basic II Active Pre Chamber I Active Pre Chamber II Knock Detection / Criterion / Control Pre Ignition / Combustion Phenomena Pre Chamber III Knock Detection / Pre Ignition II <?page no="11"?> Knock in SI-Engines - A continuing challenge for combustion system development Dr. Frank Altenschmidt, Dr. Eberhard Kraus; Mercedes-Benz AG, Germany Abstract: Since the invention of gasoline engines, the phenomenon of knocking combustion has been known. Particularly in the first half of the 20th century knock has been promoted by poor fuels, nowadays high power densities of engines and increasingly strict emission legislation, which prohibit substoichiometric operation, are mainly responsible for it. Alleviating measures, which lead to an increase in CO 2 -emissions, e. g. decrease of compression ratio, can’t be taken due to demanding consumption targets. Engine knock is primarily controlled by the thermal conditions in the combustion chamber. Acceleration of flame speed can be achieved by increasing the charge motion. However, this leads to higher peak temperatures in the combustion chamber, coupled with a higher heat loss to the combustion chamber walls. If the burning rate is too fast, this mechanism can even lead to a deterioration of the knock limit. In order to reduce the gas temperatures in particular during combustion, the use of cooled residual gas is advantageous. This enables significant earlier combustion, leading to comparable gas temperatures and wall heat losses at the knock limit. From this it follows there is a thermal limit for a given combustion chamber configuration that cannot be exceeded without further measures such as cooling or water injection. Stuttgart, September 2022 Dr. Frank Altenschmidt, Dr. Eberhard Kraus 1 Introduction Ever since Christian Reithmann was granted a patent for a four-stroke gasoline engine in 1860, it has been impossible to imagine our daily lives without this type of engine. There is probably no other engine that, after 160 years, can be found in every corner of the world and performs its service reliably under such climatically diverse conditions. However, the demands on reliability, performance, consumption and emissions have changed fundamentally. In the first decades after the invention of the internal combus‐ tion engine, the focus of development was on improving mechanical durability. When this reached an acceptable level, more efforts were made to increase the performance of the engines, which also brought the phenomenon of knocking combustion with <?page no="12"?> it. Even though at the beginning of the twentieth century the measuring techniques were far from being available as they are today, the cause of engine knock could be determined. In his textbook “Motorwagen und Fahrzeugmaschinen für flüssige Kraftstoffe” from 1925, Dr.-techn. Arnold Heller [1] describes the process of combustion including the emergence of knocking working cycles. Since pure gasoline was in short supply, especially during the Great Depression after World War I, engine damage due to knocking combustion occurred more frequently because of the poor fuel quality of substitute fuels. The investigations at this time by Thomas Midgley (inventor of tetraethyl lead) and others at General Motors and the findings derived from it led Arnold Heller to the statement that “knocking has completely lost its dangers in practice”. Even though technical progress has been unmistakable since that time, the problem of knocking combustion still represents a limit to the use of high compression ratios at optimum centre of combustion at full load. Particularly in turbocharged series engines, the aim is to use the highest possible compression ratio to achieve good part-load efficiency. However, if this is chosen too high, the losses at knock-limited load points can very quickly cancel out the advantages at part load. Figure 1 shows that very high losses occur in the case of late centres of combustion, which are much higher than profits in the low percentage range due to a high compression ratio. At the same time, late centres of combustion lead to high exhaust temperatures, which are counterproductive with respect to maximum turbine and catalytic converter temperatures. In the past, therefore, the fuel/ air mixture was adjusted substoichiomet‐ ric to reduce process temperatures, which is no longer permissible today [2]. Figure 1: Correlation between COC and IMEP in full load operation Figure 2 shows a typical engine map with areas of stoichiometric and substoichiometric operation. The area, where enrichment is necessary, is small compared to the whole map. It follows that measures and technologies that enable stoichiometric operation 12 Frank Altenschmidt, Eberhard Kraus <?page no="13"?> over the entire map may only degrade the other operating range to a very small extent, if at all. Therefore, for example, lowering the compression ratio to improve the knock limit is no option in present times. In contrast, increasing the charge motion to accelerate combustion or using cooled residual gas to improve the centre of combustion and efficiency and thus lower the exhaust gas temperature are interesting measures. Figure 2: Engine map with λ=1 operation limit This article examines the influence of charge motion, cooled residual gas or a combi‐ nation of both on the working process of the gasoline engine. 2 Theoretical Examination 2.1 Fundamentals With increasing supercharging, knocking combustion is the main obstacle for thermo‐ dynamically optimal engine operation. A large number of research projects already initiated on the subject of knocking (e. g. [3], [4]) bear witness to the fact that this topic has not lost its topicality and will not lose it any time soon. Therefore, precise knowledge of the processes in the combustion chamber that lead to knock is vital for further improvement of combustion efficiency especially at full load. Up to the 80s of the last century, two different theories were developed to explain the origin of engine knock. The detonation theory describes the onset of knocking combustion as an acceleration of the primary flame [5, 6]. The largely recognized self-ignition theory assumes secondary ignition ahead of the primary flame, which together determine the further combustion process [7, 8]. A characteristic feature of the chemical reactions which lead to eventual auto ignition is the ignition delay time. This time interval is strongly dependent on pressure, temperature and the mixture composition [9]. In a variety of publications, it was attempted to capture these very complex interactions with knock criteria [e. g. 3, 10, 11, 12, 13]. However, there is no model 13 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="14"?> which can adequately capture the local mixture states and thermal conditions in the combustion chamber sufficiently to predict engine knock under all circumstances. This shows the needs for further research work regarding improvement of engine knock behaviour. 2.2 Impact of water injection and cooled EGR on the thermodynamic cycle In order to be able to use stoichiometric mixture composition throughout the entire engine operating map, the gas temperature upstream of the turbocharger turbine and the catalytic converter must not be too high. There are several possibilities to achieve this, all of which in principle rely on the same mechanism. This mechanism is shown in Figure 3 by means of water injection and cooled exhaust gas recirculation. In case of water injection a liquid medium is injected into the inlet port, usually during the intake stroke. As shown in [14], evaporation takes place during late compression and during the combustion phase. The evaporation cools the gas and increases the mass and heat capacity of the cylinder load. Both lead to an improvement in the knock limit, which additionally lowers the exhaust gas temperature. Figure 3: Effect chain of water injection and colled EGR on the SI-engine working cycle When using cooled residual gas, it is mixed with fresh air during the intake stroke, in‐ creasing mass and heat capacity of the cylinder load. Since this is an inert gas, it doesn't participate in the combustion process. However, the released heat is distributed over the now larger mixture mass in the cylinder, resulting in lower process temperatures and thus a better knock limit. With increasing residual gas fraction in the combustion chamber the oxygen concentration and thereby the flame speed decreases. The question arises whether this should be compensated by an increase of the charge motion generated by the intake 14 Frank Altenschmidt, Eberhard Kraus <?page no="15"?> port and what influence this has on the combustion cycle as well as the knock limit. These questions will be examined hereinafter in more detail. 3 Experimental Results 3.1 Experimental setup A research single-cylinder engine was used for the experiments whose combustion chamber geometry is derived from the Mercedes-Benz 6-cylinder engine M256. Three cylinder heads with different intake ports were used. Figure 4 shows the tumble curves, which were determined by CFD-simulations. The base configuration Tz 1 already has a high tumble at the level of the M256 production engine. Starting from this, the tumble levels are increased by 25% in two steps. With each variant a series with increasing load were measured both with and without cooled EGR, where an HD-AGR system is used. Because the three intake ports have different flow resistances due to the different charge motion levels, all tests were carried out with a constant differential pressure of 200 mbar between the intake and exhaust port to keep the internal residual gas rate almost constant. All series were measured at knock limit and an inlet port temperature of 45°C. Figure 4: Comparison of the used tumble levels 3.2 Influence of charge motion on the knock limit Figure 5 shows curves of 50% mass fraction burned (MFB 50%) for all investigated tumble levels at three different engine speeds and increasing load. 15 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="16"?> Figure 5: Influence of tumble level on the knock limit for different engine speeds at λ=1 At 2000 rpm the behavior is as expected. The increased charge motion leads to an improved knock limit by 4°CA, with no further improvement with the highest tumble level Tz 1 +50%. At 4000 rpm there is still an improvement of the knock limit with variant Tz 1 +25% compared to the base port Tz 1 , but the variant Tz 1 +50% falls back to the level of the base port and is the worst variant at 5000 rpm. In the following the unexpected results are analyzed in more detail using optical measurements and combustion analysis. Figure 6 shows the distribution of knock events for all intake port variants at engine speed 5000 rpm and IMEP=21 bar. The first two variants Tz 1 and Tz 1 +25% have most knock events on the exhaust side, whereas the variant Tz 1 +50% on the opposite intake side. Figure 6: Distribution of knock events at n=5000 rpm, IMEP=21 bar and λ=1 16 Frank Altenschmidt, Eberhard Kraus <?page no="17"?> Figure 7: Cylinder pressure, burn rate and gas temperature at n=5000 rpm, IMEP=18 bar and λ=1 for two different tumble levels From this one could deduce that there is a hot spot on the exhaust side, e. g. at the exhaust valves, for the two variants with lower charge motion level. But this would also have a negative influence on the variant with the highest charge motion. To further investigate this, the flame propagation was visualized for the intake port Tz 1 +25% at a slightly reduced load using high-speed endoscopy. The analysis showed a fast flame movement towards intake side of most cycles, which explains the knock events at the exhaust side. Thus the intake port Tz 1 +50% has not only the highest tumble level, but also leads to a different flame propagation. Figure 7 shows cylinder pressure, burn rate and gas temperature for the intake ports Tz 1 +25% und Tz 1 +50% at 5000 rpm and IMEP=21 bar. Due to a later MFB 50% the peak pressure of variant Tz 1 +50% is lower than of variant Tz 1 +25%, but the peak temperatures are almost identical due to the higher flame speed of variant Tz 1 +50%. Since the two cylinder heads differ only in the inlet channels and knocking is significantly influenced by the gas temperatures, the results shown here could indicate a thermal limit of the combustion chamber. This will be the focus of chapter 3.4. 3.3 Influence of cooled EGR on knock limit, exhaust gas temperature Due to the high power density of modern gasoline engines, sufficient cooling in the vehicle is definite a challenge. If cooled residual gas is used, it increases the need for additional cooling. Therefore, the maximum residual gas rate is restricted in real vehicle 17 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="18"?> operation. Since this is not a problem on test benches, the investigations were carried out with the maximum possible residual gas rates. Figure 8 shows the change of MFB 50% with increasing residual gas rate (EGR) applied for different speeds at a load of IMEP=18 bar. For all variants, an improvement of the MFB 50% can be observed with increasing residual gas rate, but its amount depends significantly on the speed. In particular, the behaviour of the variant Tz 1 +50% with the worst knock limit (see Figure 5) is remarkable, which shows the biggest improvement in MFB 50% upwards n=4000 rpm compared to the other two variants. Figure 8: Effect of EGR on the MFB 50% for different speeds, IMEP=18 bar and λ=1 Figure 9: Cylinder pressure, burn rate and gas temperature at n=5000 rpm, IMEP=18 bar and λ=1 for different residual gas contents, tumble level Tz 1 +50% 18 Frank Altenschmidt, Eberhard Kraus <?page no="19"?> Figure 9 shows for four different residual gas rates the cylinder pressure, heat release rate and mass mean temperature curve at n=5000 rpm and IMEP=18 bar for intake port Tz 1 +50%. Although the peak pressure in the combustion chamber increases significantly with rising residual gas mass and earlier MFB 50%, the maximum mass mean temperature decreases and accordingly the heat flow into the combustion chamber wall and its temperature. This significant improvement in knock limit is also evident in the case of increasing load with and without residual gas, which is illustrated in Figure 10. The dependency of the MFB 50% on the load for the best variant Tz 1 +25% and the worst variant Tz 1 +50% are compared with and without residual gas for the speeds n=2000, 4000 & 5000 rpm. While at 2000 rpm MFB 50% and therefore the knock limit are almost the same, the situation changes significantly at 4000 rpm. In both cases without residual gas (as shown in Figure 5), the intake port Tz 1 +50% has a significantly worse knock limit than the port Tz 1 +25%. After residual gas is added, the knock limit of the intake port Tz 1 +50% catches up and is comparable to the other intake port. Figure 10: MFB 50% at increasing loads with and without cooled EGR, λ=1 Since the variant Tz 1 +50% has the greatest changes of the knock limit, in Figure 11 the pressure and heat release rate as well as the mass mean temperature and the wall heat flow at n=5000 rpm and IMEP=20 bar are displayed. Due to the higher cylinder mass and the significant earlier MFB 50% the peak pressure in the case with 15% residual gas is higher. Surprisingly, because of the earlier knock limit the burning rate is almost as fast as without EGR. The mass mean temperatures is lower due to the increased cylinder mass, despite of the approximately 12°CA better MFB 50%. Although there are considerable differences in the mixture composition and the combustion process, the total heat loss (Qw) to the walls towards the end of the expansion stroke is almost the same. 19 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="20"?> Figure 11: Combustion analysis for n=5000 rpm, IMEP=20 bar, λ=1, EGR=0 & 15% In most papers on knock criterions [e.g. 3, 10], the temperature in the unburnt zone plays a key role. In addition, high charge motion levels are considered advantageous for good knock limits. However, the results presented her suggest the benefits are not unlimited. 3.4 Existence of a thermal limit As shown in Figure 7, the higher flame speed of the intake port Tz 1 +50% results in a comparable maximum mass mean temperature despite the significantly later knock limit spark advance. For a better understanding concerning the thermal gas conditions in the combustion chamber, the results from a two-zone calculation for the load variation presented in Figure 10 are analysed. Figure 12 shows the maximum temperature in the burnt zone, Figure 13 in the unburnt zone. As expected, the maximum temperature in the burnt zone can be significantly reduced with the addition of cooled residual gas. However, it is striking that at n=2000 rpm the temperatures in the case with no EGR are almost identical and at higher speeds slightly lower for the maximum tumble stage. On the contrary, the values for all speeds with cooled residual gas lie very close together. 20 Frank Altenschmidt, Eberhard Kraus <?page no="21"?> Figure 12: Maximum temperature in burnt zone with and without cooled EGR, λ=1 Considering the temperatures in the unburnt zone, which are used in all common knock models, there are comparable values for all variants with and without added residual gas within the model accuracy. This is particularly remarkable at 5000 rpm for the Tz 1 +50% variant, as here at the highest load the MFB 50% is approximately 12°CA different from the other variants. Figure 13: Maximum temperature in unburnt zone with and without cooled EGR, λ=1 Since the examined cylinder heads differ only in the intake ports and not in the combustion chamber geometry or in the water jacket, it is shown that a combustion chamber configuration has a thermal limit. If additional cooling measures are no option, an improvement in the knock limit spark advance can only be achieved with significant lower process temperatures, whereby the thermal limit here assumed is not changed. The use of cooled residual gas is therefore a very effective method to improve the knock limit. A thoroughly designed EGR-system is one of the few technologies which not only contributes to an increase in thermodynamic efficiency in full load operation and enables stoichiometric fuel-air mixture in the entire engine operation map, but also improves the fuel consumption in part load. 21 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="22"?> 4 Summary and outlook The presented studies of three different charge motion levels have shown that the overall system “combustion chamber” must be carefully designed. The occurrence of engine knock is determined by a variety of parameters that influence each other. In addition to a good mixture formation, which can be achieved by an adequate charge motion level, the thermal condition during the combustion process is crucial. Lower charge motion causes longer combustion duration and lower heat losses to the combustion chamber walls. Higher charge motion with faster combustion speed at the same MFB 50% lead to higher thermal losses to the combustion chamber walls compared to lower charge motions. If the MFB 50% is earlier, the thermal losses increase even further. By adding cooled residual gas, the mean gas temperature during combustion can be significantly reduced. The results presented here show that the maximum gas temperatures in the unburned zone at knock limit spark advance for all tested charge motion levels are approximately at the same height, regardless of the charge motion and amount of residual gas. It can be concluded that there is an individual thermal limit for a given combustion chamber configuration. To shift this limit towards a better knock limit, the process temperature during combustion must be lowered. Since the maximum possible residual gas rate is restricted due to the cooling capacity in a vehicle, additional measures must be taken. This could be the use of water injection, improved charge air cooling or combustion chamber cooling. Even if it was not explicitly shown in this article, the use of cooled residual gas represents a very effective, if not even the most effective measure to improve the knock limit compared to the other mentioned technologies. 5 References Heller, A.: Motorenwagen und Fahrzeugmaschinen für flüssigen Brennstoff, Erster Band: Motoren und Zubehör, Springer Verlag 1925 https: / / ec.europa.eu/ germany/ news/ 20181218-co2-grenzwerte-autos_de Worret, R.: FVV-Vorhaben Nr. 700 “Klopfkriterium”, IfKM Universität Karlsruhe, 2001 Rothe, M.: FVV-Vorhaben Nr. 816 “Extremklopfer”, IfKM Universität Karlsruhe, 2005 Curry, S.: A-Three-Dimensional Study of Flame Propagation in a Spark Ignition Engine, SAE Trans. 71, 1963 Sokolik, A.: Self-Ignition, Flame and Detonation in Gases, Israel Program for Scientific Transla‐ tion Ltd., Jerusalem, 1963 Ganser, J.: Untersuchungen zum Einfluss der Brennraumströmung auf die klopfende Verbren‐ nung, Dissertation RWTH Aachen, 1994 Stiebels, B.: Flammenausbreitung bei klopfender Verbrennung, Dissertation RWTH Aachen, 1997 Warnatz, J. et al.: Verbrennung, Springer Verlag 2. Auflage, 1996 22 Frank Altenschmidt, Eberhard Kraus <?page no="23"?> Franzke, D. E.: Beitrag zur Ermittlung eines Klopfkriteriums der ottomotorischen Verbrennung und Vorausberechnung der Klopfgrenze, Lehrstuhl für Verbrennungskraftmaschinen und Kraftfahrzeuge, TU München, 1981 Binder, S.: Implementierung und Verifikation verschiedener Klopfkriterien auf Basis vorhand‐ ener Indizierdaten, Diplomarbeit, IVK Universität Stuttgart, 2003 Marchi, A. et al.: Neuer Ansatz zur Berechnung des Klopfens bei Ottomotoren, Ricardo Deutsch‐ land GmbH, MTZ 03/ 2015 Mally, M. et al.: Klopfen bei Volllast-Abgasrückführung, RWTH Aachen, MTZ 02/ 2018 Altenschmidt, F. et al.: Water-Port-Injection at SI-Engines: Effects and Challenges, 17th Confer‐ ence “The Working Process of the Internal Combustion Engine”, Graz, 2019 23 Knock in SI-Engines - A continuing challenge for combustion system development <?page no="25"?> Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell Jan Reimer, Institut für Kolbenmaschinen, Karlsruher Institut für Technologie(KIT), Mitra Zabihigivi (KIT), Ina Volz (Mercedes Benz AG), Jürgen Pfeil (KIT), Frank Altenschmidt (Mercedes Benz AG), Thomas Koch (KIT) Abstract: This paper investigates the effect of lubricant oil and the additives on low-speed pre-ignition (LSPI). The abnormal combustion phenomenon of LSPI is still an important issue in the development of modern internal combustion en‐ gines. Despite numerous studies to understand the potential sources for low-speed pre-ignition, this phenomenon has not yet been fundamentally understood. Therefore, this paper provides a test procedure to investigate the necessary thermodynamic conditions leading to LSPI. A major goal of this study was to reduce the complexity in a gasoline engine arising from the interaction of diverse processes typically taking place in an engine. Therefore, the experiments were conducted by using a newly developed constant volume combustion chamber as a centerpiece of a completely new testbed. Additionally, a device to generate a defined small amount of tempered oil droplets in the size range of microns into the combustion cell was designed and used for this study. This experimental approach focuses on lubricant oils as a potential trigger for pre-ignition. Therefore, three oils with different contents of Calcium and Magnesium were used to study ignition delay times and self-ignition temperatures of oil under various engine-like pressure conditions. Furthermore, to separate possible evaporation effects of the liquid oil droplets, the different oil samples were also admixed to the base fuel. The test results showed that in the Calcium range of 0.025% to 0.075%, the Calcium detergents impact LSPI activity as promoters. In addition, from the length of the ignition delay time for oil droplets in the hot air, it could be concluded that an oil droplet in the engine combustion chamber cannot initiate a pre-ignition in the same working cycle. However, an oil droplet, deposited on a hot component in the engine may lead to a pre-ignition in a further working cycle. <?page no="26"?> 1 Introduction In general, modern turbocharged gasoline engines with direct injection are increasingly showing anomalies during the combustion process. These anomalies may be caused by Low-speed pre-ignition (LSPI), which is characterized by an ignition before the actual spark ignition timing. Low-speed pre-ignition (LSPI) is an undesirable combustion phenomenon and it potentially leads to extreme knocking events, which can cause serious engine damage. Consequently, exploring possible sources and the mechanism of this type of abnormal combustion in internal combustion engines plays a significant role in the development of modern gasoline engines. Since pre-ignition seems to be a serious challenge for improving the performance of gasoline engines, numerous researches have been conducted to clarify the mechanism of this phenomenon and to propose solutions to eliminate LSPI. Over the last several years exploring a proper solution for LSPI has been a serious technical challenge for automotive companies. Due to this fact, there have been major efforts to establish a systematic methodology for exploring and evaluating potential sources of LSPI events. Because of the risk of engine damage after a pre-ignition event, investigations to reduce LSPI frequency have attracted considerable attention. However, despite several high-quality studies looking to understand the mechanism behind the LSPI and searching for the main origin of pre-ignition events, this phenomenon needs still more researches in order to explain the mechanism and the probable causes. While some early publications focused on engine design and operation to reduce LSPI frequency, recent publications have brought other probable causes of pre-ignition into focus because LSPI should be explained by a combination of engine-related, oil-related, and fuel-related mechanisms interacting in an engine [1-3]. Jatana et al. [2] , for instance, have performed engine testing with three single component fuels in order to understand the effect of fuel properties on LSPI. The aim of this study was to investigate the dependence of fuel distillation on Low-speed pre-ignition (LSPI). On the other hand, in several works lubricant oil has been identified as the main origin of pre-ignition [4-8]. These studies have attributed LSPI events in gasoline engines to the auto-ignition of lubricant oil droplets in the combustion chamber. Considering the results of these publications, one of the most probable explanation for the occurrence of pre-ignition in a gasoline engine is that the oil droplets, released from the cylinder liner, can become a source of LSPI [5, 6, 9]. Furthermore, in several studies, composition of lubricant oil and the oil additives are believed to be the major contributing factors in the occurrence of LSPI events in the engine [7, 10]. Morikawa et al. [11] indicated that the oil properties will affect the tendency of LSPI. Three different oil properties such as cetan number, distillation characteristics, and Calcium additive were examined in this work as critical points in pre-ignition. Dahnz et al. [6] carried out experimental investigations using a test engine accompanied by numerical simulations to assess possible causes for pre-ignition. The results have identified that oil dilution greatly affects the amount of released oil droplets and hence the pre-ignition frequency. Okada et al. [12] have applied visualization approach to 26 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="27"?> perform optical investigations and to clarify the mechanism of pre-ignition in the engine. Ritchie et al. [13] presented a statistical analysis by assessing data collected from six different engines. Furthermore, tests were conducted on various lubricant components to determine their effects on LSPI. The effects of Calcium and Magnesium detergents on LSPI were investigated. They found out that in contrast to the effect of Calcium, an increase in Magnesium concentrations has no effect on the occurrence of LSPI. Hence, LSPI results for mixtures of Calcium and Magnesium are directly proportional to the amount of Calcium. Takeuchi [14] has performed several engine tests with a prototype turbocharged DI-SI engine to investigate the influence of the oil additives on pre-ignition. This study confirmed that the type of oil and additives have an intensive impact on LSPI. Results indicated that Calcium detergent has a contributory effect on a pre-ignition event and higher Calcium content can benefit the pre-ignition rate. In addition, several other studies indicated that increasing levels of Calcium lead to an increase in the LSPI rate [4, 15]. However, a close examination of the impact of Calcium additives is lacking and this point requires further in-depth investigations to be sufficiently clarified. Exploring and understanding the dependence of pre-ignition on Calcium detergents is a major goal of this paper. Therefore, a fundamental evaluation of the impact of Calcium on LSPI activity has been undertaken in this work. The vast majority of publications describing lubricant impacts on LSPI only consider engine tests to assess the impact of different factors on this phenomenon [5-7, 11, 16]. Most of these investigations have been carried out inside an engine, where many other different factors can influence the mechanism of pre-ignition. For this reason, a detailed investigation of LSPI by engine tests is not possible due to the complexity coming from several processes taking place simultaneously in the engine. Therefore, the results of experiments depend strongly on engine variables. Once these several influencing variables are reduced, it is possible to get a more accurate understanding of the process. Therefore, the investigations are to be carried out with some simplifications through delimiting the influencing variables of the engine. Consequently, in this study the investigations are not conducted in a combustion engine, but in a constant volume combustion cell. It was decided to utilize a combustion cell, which is newly designed and developed at KIT. The detailed characteristics of combustion chamber will be explained in the following sections of this paper. The experiments and results presented in this paper are to be understood as a following project to the previous paper presented in 5 th International Conference of Knocking in Gasoline Engines [17] and also as a preliminary step to the upcoming experiments in the future. While the general features of the testbed and the measure‐ ment system were described in the previous publication comprehensively, this paper focuses on the experimental methodology and results. In addition, some newly added components and features are also presented in this paper. 27 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="28"?> 2 Testbed development and setup This paper is a follow-up study and the second part of a paper series presenting the further developments of the testbed, methodology, and results of the project. In the first part of this series of papers [17] the general requirements, challenges, and features of the testbed were discussed. Additionally, a detailed description of the design and dimensioning of the combustion chamber, as well as the measurement setup, were presented. The current paper provides first a general overview of the combustion chamber system. It will be explained, how the combustion chamber has been further developed and which other new components have been added to the system. Then in the following sections, it focuses on test procedure and results. Following the aim of studying the conditions leading to pre-ignition, a combustion chamber was developed to be utilized for the experiments. In order to reduce the system complexity, the engine was replaced with a combustion chamber, since the complexity of the engine makes detailed analysis difficult. This provides a great possibility to perform fundamental experiments to clarify the necessary conditions for the occurrence of pre-ignition accurately. Consequently, a constant volume combustion chamber was manufactured for optical investigations of pre-ignition at high pressures and high temperatures. Different parts of the system are described in Fig. 1. The system consists of the pre-conditioning chamber, the combustion chamber with optical accesses, fuel injector, and a newly developed oil dosing device. In addition, the fuel and air management systems serve to supply air and fuel during experiments. The fuel management system can provide an injection pressure of 200 bar using an accumulator serving as a pressure reservoir. Furthermore, several valves and sensors are utilized to implement the experiments. The positions of different pressure sensors (blue) and temperature sensors (green) are described in Fig. 1. High-pressure cut-off valves are utilized to control the flow through the system and also to prevent the damaging of other components. At the beginning of each experiment, air flows from the air management system through valve V1 into the pre-conditioning chamber. After some procedures in the pre-conditioning chamber, which are explained in detail in section 3, air or the gaseous mixture flows through valve V2.1 into the combustion chamber. Then, combustion takes place in the combustion chamber. The combustion chamber has been developed for a maximum gas pressure of 350 bar and a maximum gas temperature of 500 °C. The combination of high pressures and high temperatures has been the main challenge for the design, dimensioning, and safety of the combustion chamber. In order to provide the same preliminary conditions for all experiments, it must be possible to release the burned gas from the chamber completely and to flush it. Therefore, after the combustion is done completely, valve V2.2 opens and makes it possible for the air in the pre-conditioning chamber to release into the environment. Valve V3 has a similar function and serves the burned gases to exit from the combustion chamber. Applying a switching valve makes it possible to flush the system by air or nitrogen. 28 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="29"?> In addition to the cut-off valves, there is also a pressure control valve, which serves as an element for precise adjustment of the pressure in the system. Furthermore, since pressure peaks of higher than 350 bar can occur during the experiments, a pressure relief valve is utilized as a security measure in the system. The relief valve opens when the pressure in the combustion chamber exceeds 350 bar. Although the relief valve is used to prevent pressure peaks over 350 bar in the combustion chamber, the dimensioning of the chamber is based on much higher pressures in the order of magnitude of 1000 bar. Controlling and adjusting of the different actuators and valves as well as the record‐ ing of the pressure and temperature values are performed by a modular CompactRIO System from National Instruments. The pressures inside the two chambers are recorded by an indicating system with a frequency of 100 kHz. Air Management System P amb., T amb. Fuel Injector Oil dosing device Cut-off valve V2.1 P amb., T amb. P Oil T Oil P pre-conditioning cham. T pre-conditioning cham. Fuel Management System Cut-off valve V1 Pre-conditioning Chamber Cut-off valve V2.2 Cut-off valve V3 Combustion Chamber T Combustion cham., Wall P Combustion cham. T HE1 , T HE2 , T HE3 , T HE4 T Combustion cham., Gas P Fuel T Fuel Fig.1: Overview of the developed Testbed [17] The combustion chamber with its four accesses is shown in Fig. 2. The design and po‐ sitioning of the different accesses in the chamber required precise calculations in order to satisfy the demanding requirements. The combustion chamber was conditioned by regulated wall heating, depending on the temperature requirement of each experiment. For the reason of reproducibility, a homogeneous temperature field in the gas phase inside the combustion chamber was indispensable. 29 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="30"?> The combustion chamber was designed with a cylindrical inner geometry with two optical accesses on opposite sides with an optical clear width of 18 mm in diameter. Furthermore, two cylindrical accesses were positioned at a 90° angle to the main cylinder, which provide the possibility to mount the newly developed dosing device for hot oil and also a glow plug, while using the other two optical accesses. Positioning of these different accesses in the combustion chamber provided great opportunities to perform a variety of optical investigations like “High speed Shadowgraphy”. 2nd access (Glow plug) 1st access (Optical access) 4th access (Oil dosing device) 3rd access (Optical access) Fig. 2: Overview of the combustion chamber with its four accesses One of the main focuses of this work is the fundamental investigation of the influence of lubricating oil on the pre-ignition. This was done by monitoring oil droplets in a hot environment. For this reason, an oil dosing device was designed and developed with a micro hole of 100 micrometers in diameter, which makes it possible to disperse the hot oil droplets up to 245 °C in the combustion chamber very finely. Fig. 3 gives a general overview of the developed oil dosing device. 30 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="31"?> Total length of oil dosing device: 415 mm 104 mm 70 mm Fig. 3: Oil dosing device and positioning of micro hole 3 Experimental Methodology Studying the effects of lubricant oil on LSPI requires a systematic methodology to be followed. After designing and developing the testbed, as detailed in the previous section, various experiments were performed in the combustion chamber to investigate the conditions, which can cause the pre-ignition. Since the dilution of the oil by the fuel can change lubricant oil droplet properties, it was necessary to investigate the effect of lubricating oils, with and without fuel dilution, as a possible source of pre-ignition. Consequently, in this work two different testing procedures have been used to evaluate the effect of different oils on LSPI. The experiments were performed in two parts: • Investigating the ignition caused by oil droplets in hot air • Investigating the ignition caused by homogeneous oil-fuel-mixtures in hot air For the optical investigations, a Phantom v1612 color high speed camera was combined with an optical zoom lens. An event trigger is sent to the optical recording system, when the oil dosing device injects a small amount of oil droplets into the combustion chamber and the image recording starts. The event trigger was advantageous for evaluating the results because it can be applied as a time reference for two measurement systems. 31 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="32"?> 3.1 Investigating the ignition caused by oil droplets in hot air A general overview of the testbed is shown in Fig.4. In the first step, air flows from the air management system through a cut-off valve into the pre-conditioning chamber. Before entering the combustion chamber, the temperature conditioning of air is done in the pre-conditioning chamber at the adjusted pressure. A mass flow regulator is used for controlling the mass of air that is allowed to pass through the cut-off valve and enter the pre-conditioning chamber. The gas in the pre-conditioning chamber is heated at about 100 °C. After a defined and reproducible conditioning time the hot air flows through another cut-off valve into the combustion chamber. The combustion chamber is already heated up to a certain wall temperature. All the timeframes, temperatures and pressures can be varied for different experiments, if necessary. These are, of course, kept constant by the testbed automation system for the same experiments and the comparison of the different oils in order to achieve good reproducibility and comparability. After the overflow into the combustion chamber and after a certain time, a small amount of oil droplets is injected into the hot air by the dosing device. If the combustion chamber is heated up to the ignition temperature, the self ignition of the oil droplets occurs. The whole process is recorded by the high-speed camera to monitor if an oil ignition occurs in the chamber and if so, the whole process can be analyzed. LED Oil dosing device High-speed Camera Combustion Chamber Fig. 4: Core of the combustion system 32 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="33"?> All experiments presented in this study involved the use of the same three oils (A, B, and C) with different physical and chemical properties. The specification of test oils is given in Table 1. Since one of the main focuses of this research was the fundamental investigation of the influence of Calcium content on the pre-ignition, the applied oils each contain a different Calcium and Magnesium content. Based on the understanding from [13], Magnesium concentrations in lubricant oil has no effect on the occurrence of LSPI. Therefore, the results of the experiments are directly related to the Calcium concentration. A B C Description Ca 750ppm/ Mg 250ppm Ca 500ppm/ Mg 500ppm Ca 250ppm/ Mg 750ppm Calcium Detergent 0.6 0.4 0.2 Magnesium Detergent 0.33 0.67 1 PAO 6 99.07 98.93 98.8 Ash 0.336 0.359 0.379 Ca (%) 0.075 0.05 0.025 Mg (%) 0.025 0.05 0.075 Tab. 1: Specification of test Oils 3.2 Investigating the ignition caused by homogeneous oil-fuel-mixtures in hot air In the second part of the experiments, the ignition of homogeneously spread, pre-evaporated fuel-oil-mixtures in hot air was investigated, since fuel blending into the oil can lead to changes in properties. For all experiments pure iso-octane was used as the reference fuel. Oils A, B, and C described in Table 1 were used again for these experiments. Like in the experiments in the first part, air flows first in the pre-conditioning chamber. As the pressure in the pre-conditioning chamber reaches the defined value with a pre-defined mass of air, the fuel-oil mixture is injected into the air by an HDEV 5.2 injector. Then, the conditioning of the air-fuel-oil mixture (evaporation and homogenization) takes place in the pre-conditioning chamber, which is heated at about 100 °C wall temperature. To create a homogeneous mixture, there is a magnetic stirrer inside the chamber. For the reason of reproducibility of experiments, homogeneity of the gaseous mixture is necessary. After a certain mixingand conditioning time, the gaseous mixture flows through the cut-off valve into the combustion chamber. The combustion chamber is already heated up to a certain temperature on the wall. Once the valve gets closed, the pre-conditioning chamber gets flushed and the pressure inside the chamber drops. After a distinct time after overflow, a self-ignition may happen in the gaseous mixture. Otherwise, if there is no 33 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="34"?> ignition, a glow plug is utilized to make sure that the mixture is combustible. As the glow plug control system is activated, the glow plug begins to heat up. If the mixture is combustible, it ignites immediately after the temperature increases locally by the glow plug. It is a metal plug with a temperature sensor and all tests are conducted with the same plug. 4 Results and Discussion In this section, the measurement results are presented and discussed. The study examined the conditions under which the lubricant oil initiates combustion in the combustion chamber. Investigations were performed with different oils, containing different Calcium content. As a first step, the influence of the chamber pressure and chamber temperature on the ignition delay was investigated. Fig. 5 and Fig. 6 illustrate the variation of the ignition delay time of oil droplets with different pressures and temperatures in the combustion chamber. According to Fig. 5, at a constant chamber pressure, the ignition delay time of oil droplets decreases as the chamber temperature increases. Furthermore, according to the obtained results shown in Fig.6, at a constant chamber temperature, an increase in chamber pressure leads to a decrease in the ignition delay time of oil droplets. However, the influence of chamber temperature is stronger compared to the influence of chamber pressure. 300 310 320 330 340 350 360 0 20 40 60 80 100 120 140 P_cham.= 20 bar P_cham.= 30 bar P_cham.= 47.5 bar P_cham.= 50 bar Ignition delay [ms] Chamber Temperature [°C] Fig. 5: Variation of ignition delay time of oil droplets in air with different chamber temperatures 34 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="35"?> 20 30 40 50 60 0 20 40 60 80 100 120 140 T_cham.= 310 °C T_cham.= 325 °C T_cham.= 335 °C T_cham.= 345 °C Ignition delay [ms] Chamber Pressure [bar] Fig. 6: Variation of ignition delay time of oil droplets in air with different chamber pressures In addition, the effect of Calcium content on LSPI events was investigated. For this reason, the investigations were performed with test oils containing 0.025%, 0.05%, and 0.075% Ca, as described in Table 1. The data obtained from experiments are shown in Fig. 7. The chamber temperature was varied from 280 to 320 °C. The chamber pressure was varied from 25 to 50 bar. Droplet self ignition occurred in higher chamber temperatures as the chamber pressure decreases. In addition, oil C shows higher ignition temperatures compared to oil A, while oil C has lower Calcium content than oil A. As a result, in this Calcium range, oils with higher Calcium content ignite at lower chamber temperatures and Calcium has a contributory effect on pre-ignition events. 35 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="36"?> 25 30 35 40 45 50 280 290 300 310 320 Chamber Temperature [°C] Chamber Pressure [bar] Oil A, 0.075% Calcium Oil B, 0.05% Calcium Oil C, 0.025% Calcium Fig. 7: Measurement results for self-ignition of different oils in air In the next step, investigations were continued with mixing different oils in the reference fuel in order to understand the chemical effects of oil and additives in fuel-oil-mixture. That means, 5 and 10% vol. of oil was mixed into the reference fuel and the mixture was injected into the air through the fuel injector in the pre-conditioning chamber. Then, this homogenous and heated mixture flows into the combustion chamber, which is heated up to a certain wall temperature. After the overflow, a self ignition may happen in the gaseous mixture, if the temperature in combuction chamber is high enough. The obtained results are described in Fig. 8. As this diagram indicates, admixture of oil in the base fuel leads to an increase in the ignition temperature of the mixture. As shown in this diagram, the ignition temperature of iso-octane is lower than the ignition temperature of a fuel-oil mixture containing 95% vol. iso-octane and 5% vol. Oil A. If 10% vol. oil is added to the fuel, the ignition temperature will be even higher because the oil content is higher. If another oil with different Calcium content is added to reference fuel, the ignition temperatures are also different. As illustrated in Fig. 8, adding oils with higher Calcium content to fuel decreases the ignition temperature of the mixture. The results of these experiments confirmed the same tendency, as in the previous results shown in Fig. 7, since Calcium acts as a promoter for LSPI. Nevertheless, it must be mentioned that even with the addition of oil A with the highest Calcium content into the fuel, the ignition temperature was still higher than the pure iso-octane. 36 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="37"?> 0 5 10 15 20 25 30 200 220 240 260 280 300 320 340 360 Chamber Temperature [°C] Chamber Pressure [bar] iso-octan iso-octan+5% Oil A (0.075% Ca) iso-octan+10% Oil A (0.075% Ca) iso-octan+5% Oil C (0.025% Ca) Fig. 8: Measurement results for self-ignition caused by fuel-oil-mixture To explain the reasons for this phenomenon, the results were compared with recent publications and literature. Tanaka [18] has done several pieces of researches regarding oil droplet ignition. He heated the commercial lubricating oil droplets (0W-20) and determined the auto-ignition temperature and activation energy. He then calculated the activation energy for fuel, oil, and the fuel-oil-mixture. As shown in Fig. 9, the activation energy of pure oil is about 30% higher than for iso-octane. For this reason, oil ignition may happen at higher temperatures than fuel. Based on the ignition delay of oil droplets, obtained in the current study, there is a possibility that the oil droplet released from the cylinder liner, does not ignite imme‐ diately in the same cycle after entering the combustion chamber, but it accumulates on different parts of the cylinder because the droplet temperature is still low. Then the droplet can be heated during combustion and it can be the ignition source in the next engine cycle. 37 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="38"?> 0 20 40 60 80 100 5000 6000 7000 8000 9000 Activation Energy x10^3 [J/ mol] Proportion of Fuel [%] Fig. 9: Activation energy of the mixture of iso-octane and lubricating oil [18] 5 Summary and Outlook This study focused on pre-ignition behavior and the effects of engine oil additives on LSPI. In the first part of this series of papers [17], general requirements, challenges, and features of the developed testbed were explained. In the current paper, the approach was further followed and experiments were conducted on various lubricant oils to determine the effect of Calcium content on LSPI. The effects of lubricating oil are studied and the following conclusions were drawn from the study: 1. Experiments at a constant chamber temperature indicated that an increase in chamber pressure leads to a decrease in the ignition delay time of the lubricant oil. In the same manner, based on similar experiments at constant chamber pressure, the ignition delay time of oil droplets decreases as the chamber temperature increases. 2. The ignition temperature of admixture of oil in the base fuel (iso-octane) leads to an increase in the ignition temperature of the homogenous gaseous mixture. 3. The influence of Calcium contents in the oil on the pre-ignition could be confirmed as expected from the literature. Investigations were performed with test oils con‐ taining 0.025%, 0.05%, and 0.075% Ca. In this Calcium range, the Calcium detergents impact LSPI activity as promoters and pre-ignition tendencies correspond to the level of Calcium content. 4. From the length of the ignition delay for oil droplets in air, it could be concluded that an oil droplet in the engine combustion chamber cannot initiate a pre-ignition in the same working cycle. An oil droplet, deposited on a hot component in the engine (e.g. piston rings or exhaust valve), can lead to a pre-ignition in a further working cycle, if it is kept permanently above its activation temperature. 5. The next step would therefore be to check whether an oil droplet on a hot surface can lead to pre-ignition and whether the surface materials result in catalytic effects. 38 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="39"?> The results of this study demonstrate that lubricant oil formulation plays a significant role in LSPI activity. Based on these results even relatively simple formulation changes can impact the LSPI rate of the engine significantly. Therefore, the obtained results are important to develop a balanced lubricant oil to meet all performance expectations of the engine including LSPI. Acknowledgements The authors express their special thanks to Daimler AG for funding this research project. Also, we wish to thank Infineum UK Ltd. for providing oil samples for the project. References [1] B. Tormos et al., “Experimental assessment of ignition characteristics of lubricating oil sprays related to low-speed pre-ignition (LSPI),” International Journal of Engine Research, 146808742110132, 2021, doi: 10.1177/ 14680874211013268. [2] G. S. Jatana, D. A. Splitter, B. Kaul, and J. P. Szybist, “Fuel property effects on low-speed pre-ignition,” Fuel, vol. 230, pp. 474-482, 2018, doi: 10.1016/ j.fuel.2018.05.060. [3] T. Hülser, G. Grünefeld, T. Brands, M. Günther, and S. Pischinger, “Optical Investigation on the Origin of Pre-Ignition in a Highly Boosted SI Engine Using Bio-Fuels,” in SAE Technical Paper Series, 2013. [4] K. A. Fletcher, L. Dingwell, K. Yang, W. Y. Lam, and J. P. Styer, “Engine Oil Additive Impacts on Low Speed Pre-Ignition,” SAE Int. J. Fuels Lubr., vol. 9, no. 3, pp. 612-620, 2016, doi: 10.4271/ 2016-01-2277. [5] K. Fujimoto, M. Yamashita, S. Hirano, K. Kato, I. Watanabe, and K. Ito, “Engine Oil Development for Preventing Pre-Ignition in Turbocharged Gasoline Engine,” SAE Int. J. Fuels Lubr., vol. 7, no. 3, pp. 869-874, 2014, doi: 10.4271/ 2014-01-2785. [6] C. Dahnz, K.-M. Han, U. Spicher, M. Magar, R. Schiessl, and U. Maas, “Investigations on Pre-Ignition in Highly Supercharged SI Engines,” SAE Int. J. Engines, vol. 3, no. 1, pp. 214-224, 2010, doi: 10.4271/ 2010-01-0355. [7] M. Ohtomo, H. Miyagawa, M. Koike, N. Yokoo, and K. Nakata, “Pre-Ignition of Gasoline-Air Mixture Triggered by a Lubricant Oil Droplet,” SAE Int. J. Fuels Lubr., vol. 7, no. 3, pp. 673-682, 2014, doi: 10.4271/ 2014-01-2627. [8] M. Ohtomo, T. Suzuoki, H. Miyagawa, M. Koike, N. Yokoo, and K. Nakata, “Fundamental analysis on auto-ignition condition of a lubricant oil droplet for understanding a mechanism of low-speed pre-ignition in highly charged spark-ignition engines,” International Journal of Engine Research, vol. 20, no. 3, pp. 292-303, 2019, doi: 10.1177/ 1468087417751240. [9] S. Hirano, M. Yamashita, K. Fujimoto, and K. Kato, “Investigation of Engine Oil Effect on Abnormal Combustion in Turbocharged Direct Injection - Spark Ignition Engines (Part 2),” in SAE Technical Paper Series, 2013. 39 Basic investigations on the cause of initial pre-ignition in a constant volume combustion cell <?page no="40"?> [10] B. Tormos, J. M. Garcia-Oliver, S. Bastidas, B. Domínguez, F. Oliva, and D. Cárdenas, “Investigation on low-speed pre-ignition from the quantification and identification of engine oil droplets release under ambient pressure conditions,” Measurement, vol. 163, p. 107961, 2020, doi: 10.1016/ j.measurement.2020.107961. [11] K. Morikawa, Y. Moriyoshi, T. Kuboyama, T. Yamada, and M. Suzuki, “Investigation of Lubricating Oil Properties Effect on Low Speed Pre-Ignition,” in SAE Technical Paper Series, 2015. [12] Y. Okada, S. Miyashita, Y. Izumi, and Y. Hayakawa, “Study of Low-Speed Pre-Ignition in Boosted Spark Ignition Engine,” SAE Int. J. Engines, vol. 7, no. 2, pp. 584-594, 2014, doi: 10.4271/ 2014-01-1218. [13] A. Ritchie, D. Boese, and A. W. Young, “Controlling Low-Speed Pre-Ignition in Modern Automotive Equipment Part 3: Identification of Key Additive Component Types and Other Lubricant Composition Effects on Low-Speed Pre-Ignition,” SAE Int. J. Engines, vol. 9, no. 2, pp. 832-840, 2016, doi: 10.4271/ 2016-01-0717. [14] K. Takeuchi, K. Fujimoto, S. Hirano, and M. Yamashita, “Investigation of Engine Oil Effect on Abnormal Combustion in Turbocharged Direct Injection - Spark Ignition Engines,” SAE Int. J. Fuels Lubr., vol. 5, no. 3, pp. 1017-1024, 2012, doi: 10.4271/ 2012-01-1615. [15] M. C. Kocsis, T. Briggs, and G. Anderson, “The Impact of Lubricant Volatility, Viscosity and Detergent Chemistry on Low Speed Pre-Ignition Behavior,” SAE Int. J. Engines, vol. 10, no. 3, pp. 1019-1035, 2017, doi: 10.4271/ 2017-01-0685. [16] A. Zahdeh et al., “Fundamental Approach to Investigate Pre-Ignition in Boosted SI Engines,” SAE Int. J. Engines, vol. 4, no. 1, pp. 246-273, 2011, doi: 10.4271/ 2011-01-0340. [17] I. Volz, J. Pfeil, T. Koch, and F. Altenschmidt, “Investigating the Cause of Initial Pre-ignition - A New Approach,” in Knocking in Gasoline Engines, M. Günther and M. Sens, Eds., Cham: Springer International Publishing, 2018, pp. 37-54. [18] J. Tanaka, “Behaviour of Lubricating Oil Droplets in Cylinder and Fuel Ingredients on Pre-Ignition of Supercharged SI Engine,” in 17th Conference “The Working Process of the Internal Combustion Engine”. 40 Jan Reimer, Jürgen Pfeil, Ina Volz, Frank Altenschmidt, Thomas Koch <?page no="41"?> An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Stability Limit in a Gasoline Engine Dongwon Jung (Hyundai Motor Company), Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee Abstract: With a substantial increase of hybrid electric vehicles (HEV), further improvement of the thermal efficiency is necessitated, in particular for spark-ig‐ nition(SI) engines. Lean SI operation can offer a considerable improvement of the thermal efficiency relative to that of traditional stoichiometric SI operation. However, the combustion becomes unstable with an increase of air dilution, which leads to unacceptable lean operation. For improving the stability of lean operation, this work examines the use of multiple spark discharges by varying the time interval between spark discharges and the number of spark discharges in a four-cylinder engine. Twin ignition coils are applied to each cylinder and connected to a spark discharge timing controller capable of transmitting signals to each coil, respectively. Also, a 48-volt inductive ignition system is used to implement various multiple spark discharge strategies, and its result is compared to that of a regular 12-volt inductive ignition system. Three operating conditions are considered for stoichiometric and lean operation. First, using the 12-volt ignition system, the baseline lean operating points were established with the single-stage spark discharge by a conventional single ignition coil as a base case and the twin ignition coils as a case without the multiple spark discharges. Then, the effects of two-stage spark discharge were examined by varying the time interval between spark discharges. Compared to the sin‐ gle-stage spark discharge, the two-stage spark discharge is effective at shifting the lean-stability limit to leaner operation, even for the same input energy to primary coil. However, the time interval between spark discharges should be adjusted because excessive time interval leads to unstable lean operation. Based on the optimized time interval, the 48-volt inductive ignition system was used to increase the number of spark discharges with the higher spark discharge energy. The lean-stability limit could be extended further with the increase of the number of spark discharges. However, for the higher engine speed and load, the spark blowout occurs at the end of the multiple spark discharges due <?page no="42"?> to too strong in-cylinder flow, which leads to unstable lean operation despite the increased number of spark discharges. This indicates that the number of spark discharges should be also adjusted with the time interval between spark discharges, depending on the operating conditions. 1 Introduction Due to ever-stricter regulations for exhaust gas emissions from vehicles [1,2], the development of internal combustion engines with higher efficiency is necessitated [3]. In particular, a considerable efficiency improvement is required for gasoline spark-ig‐ nition (SI) engine because of its widespread use for automotive applications, including hybrid electric vehicles (HEV) [4]. With the increase of HEV since the mid-2000’s, SI engines for HEV have been developed mainly for improving the thermal efficiency. This happens because the maximum thermal efficiencies are typically observed in the main operating points of HEV [5]. As a result, SI engines with the maximum brake thermal efficiency (BTE) more than 40% have been mass produced since the late 2020s [6]. For additional improvement of BTE, various technologies have been demonstrated such as turbocharging, downsizing and cooled exhaust gas recirculation (EGR) [7,8]. Ultimately, all of these technologies are the methods to overcome the limitations of stoichiometric operation which is a dominating combustion mode for SI engines. Although stoichiometric operation allows the application of a three-way catalyst for cost-effective reduction of nitrogen oxides (NOx), oxidation of carbon monoxide (CO) and hydrocarbon (HC) [9], there are several reasons why it limits the thermal efficiency, especially for low and intermediate loads. First, high combustion temperatures lead to both high heat-transfer losses and unfavorable thermodynamic properties of the combustion products [10]. Second, the required intake throttling results in pumping losses [11]. A third factor is the inability of stoichiometric combustion to fully complete near TDC due to dissociation of CO 2 in the hot O 2 -depleted gases [12]. Lean operation can offer a considerable thermal efficiency improvement of SI engines by reducing all of these effects [13]. Furthermore, the benefits of lean operation are more pronounced for the leaner operation. To better understand the underlying mechanisms for the thermal efficiency improvement observed in the lean operation, Fig. 1a presents an example of heat balance, which demonstrates how BTE (i.e. brake work) varies with an excess-air ratio (l). As can be seen, the BTE increases with an increase of l in the 1.0 to 2.4 range. Of the observed increase BTE for the data from lean operation with l = 2.4, 91% is mainly attributed to reduced cooling losses resulting from the lower combustion temperatures that leads to reduced heat-transfer and higher specific heat ratio of the combustion products. The remaining 9% of the relative increase is attributed to reduced pumping losses for the de-throttled operation. However, the combustion is getting inefficient with an increase of air dilution, which leads to an increase of unburned fuel losses [14]. Furthermore, the engine operation becomes unstable with an increase of coefficient of variation (COV) of indicated mean effective pressure (IMEP), as shown in lower Fig. 1b. 42 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="43"?> This happens because of a reduction of flame speeds [15]. The reduction of flame speeds becomes a particular problem from the perspective of ignition and effective flame spread throughout the in-cylinder charge. Fig. 1: Effect of excess-air ratio (l) on heat balance and COV of IMEP To increase the flame speed for stable lean operation, various ignition systems have been studied previously [16,17]. In this work, a 48-volt inductive ignition system is applied to implement the multiple spark discharges for lean operation in a four-cylinder engine. The main objective of this work is to experimentally investigate the best opti‐ mized multiple spark discharge strategy for extending the lean-stability limits. First, the engine facility and the ignition system are described. Then, the data acquisition and analysis are explained. The results are divided into three main parts: 1. For the single-stage spark discharge, the baseline lean operating points are estab‐ lished by a conventional single ignition coil as a base case, and its lean-stability limit is compared to that of the twin ignition coils as a case without the multiple spark discharges. 2. Based on the 12-volt ignition system, the two-stage spark discharge is implemen‐ ted by varying the time interval between spark discharges (∆t int ) as a case of the multiple spark discharges. The effect of two-stage spark discharge is compared to that of single-stage spark discharge from singleand twin ignition coils. 43 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="44"?> 3. To increase the performance of the multiple spark discharges, the 48-volt ignition system is used. Using the 48-volt ignitions system, the effects of the number of spark discharges and the interval between spark discharges are investigated for extending lean-stability limits further. 2 Experimental Setup 2.1 Engine Facility All experiments are conducted in a four-cylinder direct-injection engine with AC dynamometer. Specifications of the engine are listed in Table 1. Displacement (four-cylinder) 2497.5 cm 3 Bore 88.5 mm Stroke 101.5 mm Connecting Rod Length 154.8 mm Compression Ratio 13 : 1 Tab. 1: Engine specifications A piezoelectric pressure transducer (Kistler 6065 A) is mounted in the cylinder head to measure the in-cylinder pressure. Single-injection strategy is only applied using Ron 92 gasoline. In order to accurately determine the amount of fuel supplied to the engine, Coriolis flowmeter is used, which can measure the fueling rate from 0.2 to 125 kg/ h with the measurement uncertainty lower than 0.1 2%. To obtain the lean operating point where the intake air is required more than that of the unthrottled operation, a 48-volt electric supercharger is used for intake-pressure boosting. An exhaust gas analyzer (Horiba MEXA-ONE) is used to measure the exhaust gas emissions (NOx, THC, CO, CO 2 , and O 2 ) as well as the air-fuel ratio. The value of air-fuel ratio is employed to yield excess-air ratio (l) presented here. 2.2 Data Acquisition and Analysis For each operating point, the engine is allowed to run for several minutes until all measured parameters are stable, at which point data are acquired. Fuel-flow rate and thermocouple readouts are averaged over one minute. The in-cylinder pressure, the primary and the secondary current and voltage for the spark discharge are acquired for 300 consecutive cycles using 0.1°CA resolution. In this work, the crank angles (CA) is referenced as after top dead center of the combustion stroke, aTDC. The heat-release rate (HRR) is computed from the in-cylinder pressure for each individual cycle using a constant specific-heat ratio (g = 1.33) following [10]. For computing the combustion phasing metrics like the 50% mass fraction burned (MFB50), 44 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="45"?> the HRR is integrated over the crank-angle range for which HRR is positive, after the firing of the spark. For all data presented, the computation of MFB is performed in a conventional way, whereby the integral of HRR is scaled so that it rises from 0 to 100% for every cycle, irrespective of the actual combustion efficiency. To quantify combustion instability, COV of IMEP is used, and the operation is considered as unstable when COV of IMEP > 3%. 2.3 Ignition System To implement various multiple spark discharge strategies, an inductive ignition system is fabricated. Figure 2 shows a part of circuit diagram of the inductive ignition system for each cylinder. Fig. 2: Part of a circuit diagram of the inductive ignition system for each cylinder Two ignition coils are combined into a single high tension cord as a set, and each set is connected to a spark plug featuring a laser iridium ground electrode. The spark gap width is adjusted to 1.1 mm prior to the engine tests. Furthermore, the spark plug is mounted centrally in the cylinder head and adjusted so as to orient the ground electrode to the rear of the engine. In addition to 12-volt ignition coil (150mA), 48-volt ignition coil (200mA) is fabricated to implement the more powerful multiple spark discharges. Each ignition coil is charged by 12-volt and 48-volt battery, respectively. To maintain the battery voltage even for the use of a large amount of current by consecutive multiple spark discharges, a programmable DC power supply is connected to the battery. An ignition-timing controller is connected to each ignition coil to transmit the ignition signal, respectively. After a first ignition signal is transmitted to one of the ignition coils of each set, the remaining ignition signals are alternately transmitted in the time domain, in accordance with the time interval programmed. Depending on the ignition signal, the number of spark discharges is determined with the specific time interval, and various multiple spark discharges can be established, eventually. The primary current 45 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="46"?> between the power supply and the ignition coil is measured by Tektronix TCP303 current probe with Tektronix TCPA300 amplifier. For measuring the secondary voltage and current (i.e. spark discharge voltage and current), Tektronix P6015A high-voltage probe and Tektronix TCPA312 current probe are used at the high tension cord. To better understand how the multiple spark discharge strategy is implemented by this inductive ignition system, Fig. 3 presents an example of the electric characteristics of the eight-stage spark discharge with the time interval between spark discharges (∆t int ) of 0.2 ms for lean operation (l = 1.7) at 1200RPM. First, the ignition-timing controller transmits four ignition signals to each ignition coil, respectively, at 0.2 ms intervals. The charging time (dwell time) of the primary coil is set to 0.9 ms for the first spark discharge and set to 0.2 ms for the rest of three spark discharges. Fig. 3: An example of traces of (a) primary current, (b) discharge current, (c) discharge voltage, (d) discharge energy release rate and (e) integrated discharge energy for eight-stage spark discharge by 48-volt ignitions system 46 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="47"?> During the charging time, the primary coil of ignition coil is charged, and the primary current is increased, as shown in Fig. 3a. Then, the spark discharge occurs at the end of the charging time alternately, and this leads to the serrate shape of discharge current (Fig. 3b). As a result, the discharge energy release rate which is calculated by multiplying the discharge current and the discharge voltage (Fig. 3c) is changed, as Fig. 3d shows. Finally, to present the degree of a spark discharge energy transferred to the in-cylinder gas, the discharge energy release rate is integrated in Fig. 3e over the time range for which the discharge energy release rate is positive. (Note that “discharge duration” is defined as the time or crank angle range for which the discharge energy release rate is positive.) In the following, the maximum integrated discharge energy will be referred to as “discharge energy”. 2.4 Test Rig for Spark Discharge To aid the interpretation of the experiment results in the engine, a test rig is fabricated to acquire the high-speed imaging of the spark-channel. Figure 4 shows a rendering of cross section of test rig with a spark plug. Fig. 4: Rendering of cross-section of test rig for spark discharge As marked with a blue triangle, the air can be injected toward the spark plug electrode by varying the flow velocity. Side-view imaging is performed with a Phantom VEO 1310S camera and a camera lens with a focal length of 105 mm. Fully open, the f-stop of this lens is 1.4. The large aperture in combination with a very short distance between the lens and the side window provides a high light-collection efficiency. Therefore, the presented high-speed plasma images of the spark-channel are from 47 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="48"?> the side-view. Figure 5 presents the change of a relative length of spark-channel (upper) with its spark-channel images (lower) when the eight-stage spark discharge with ∆t int = 0.2 ms occurs. (The lightning bolt indicates each spark timing of eight spark discharge.) The red circles on the graph corresponds to the image, following the number. The rig test is performed at atmosphere with the air flow velocity of 30m/ s. Three phases of breakdown are expected, following the definitions in Ref. [18]: primary streamers, secondary streamers, and transient arc / spark. Unfortunately, primary streamer breakdown across the electrode gap could not be visualized in this setup due to its weak luminosity and short time duration. This is the phase where high-energy (<10 eV) electrons are produced, and therefore the highest density of active species that can accelerate combustion. Fig. 5: An example of a relative length of spark-channel (upper) and its spark-channel images (lower) for eight-stage spark discharge with ∆t int = 0.2 ms Secondary streamer breakdown and spark breakdown produced enough luminosity to be observed (#1). Just after the occurrence of spark breakdown channel in the spark-plug gap, the spark-channel length starts to increase linearly, because a comparatively strong air flow during spark discharge shifts ions in the spark plasma, which leads to strong stretching of the spark-channel. However, as shown in the graph, the spark-channel length increases only to a certain degree (#2), and then decreases abruptly (#3). From the comparison of the image for #2 and #3, it is clear that the existing spark-channel disappears and a short spark-channel appears newly. This happens because the impedance of the spark-channel increases with the degree of spark-channel stretch, so a new spark-channel is formed repeatedly to flow the remaining primary current easily. Interestingly, if the spark-channel can be maintained regardless of spark-channel length, the discharge current flows only through the existing spark-channel without the occurrence of spark breakdown. It can be identified by comparing images #4 (just before the spark discharge) 48 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="49"?> and #5 (just after the spark discharge). As shown in images, only the color of the spark-channel of image #5 is brighter than that of #4 without the additional spark breakdown when the spark discharge occurs. It should be noted that multiple spark breakdowns do not occur even for the multiple spark discharges when a circuit is already formed across the electrode gap by the spark-channel. 3 Results All data were acquired at the engine speed of 1200 rpm and load of 120 Nm. 3.1 Lean Operation using 12-volt Ignition System 3.1.1 Effects of Spark Discharge Energy Before implementing the multiple spark discharges using twin ignition coils, there is a need to identify the sole effect of the spark discharge energy on the lean-stability limits. First, for the single-stage spark discharge, the baseline lean operating points are established by a conventional single ignition coil as a base case, and its lean-stability limit is compared to that of the twin ignition coils as a case without the multiple spark discharges. Figure 6 shows a comparison of spark discharge characteristics between the single ignition coil and the twin ignition coils for lean operation at l = 1.7. Fig. 6: Comparison of a) discharge current, b) discharge energy release rate and c) integrated discharge energy for lean operation at l = 1.7 with single-stage spark discharge from single ignition coil and twin ignition coils 49 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="50"?> As can be expected, the higher discharge current (Fig. 6a) and discharge energy release rate (Fig. 6b) is observed for the twin ignition coils, which leads to the higher discharge energy and the longer discharge duration. As a result, the discharge energy from the twin ignition coils becomes 290mJ, which is 2.4 times higher than that from the single ignition coil. The resulting changes in COV of IMEP with l are compared in Fig. 7 for the two data sets. As can be seen, COV of IMEP increases with increasing l to the point of being unacceptable (COV of IMEP > 3%) regardless of the number of ignition coils. Nonetheless, it can be noted that the increased spark discharge energy by the twin ignition coils is more effective at shifting the lean-stability limits to higher l. Fig. 7: Comparison of COV of IMEP as a function of excess-air ratio (l) for lean operation with single ignition coil and twin ignition coils 3.1.2 Effects of Two-Stage Spark Discharge To serve as a fundament for the following sections about the multiple spark discharges, the two-stage spark discharge is implemented by varying the time interval between spark discharges (∆t int ) as a case of the multiple spark discharges. Figure 8 shows a comparison of spark discharge characteristics for the two-stage spark discharge when varying ∆t int as 0.0, 0.1, 0.3 and 0.6 ms. (The result for ∆t int of 0.0 ms is identical to that of the single-stage spark discharge from twin ignition coils in Fig. 6.) As shown in Fig. 8a for two-stage spark discharge, the second peak of discharge current is appeared according to ∆t int , and it leads to the second peak of discharge energy release rate (Fig. 8b). It should be noted that the discharge duration is almost same regardless of ∆t int , but the discharge energy is somewhat less for the two-stage spark discharge, as shown in Fig. 8c. The resulting changes in COV of IMEP with l are compared in Fig. 9 for the four data sets. As can be seen, the lean-stability limit is extended with the increase of ∆t int from 0.0 to 0.3 ms. However, for ∆t int > 0.3 ms, COV of IMEP starts to increase with the increase of ∆t int . 50 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="51"?> Fig. 8: Comparison of a) discharge current, b) discharge energy release rate and c) integrated discharge energy for lean operation at l = 1.7 with the time interval between spark discharges (∆t int ) of 0.0, 0.1, 0.3 and 0.6 ms Fig. 9: COV of IMEP as a function of excess-air ratio (l) for lean operation with the time interval between spark discharges (∆t int ) of 0.0, 0.1, 0.3 and 0.6 ms It can be clearly identified in Fig. 9 for the case of ∆t int = 0.6 ms. This indicates that for the multiple spark discharges, it is critically important to set the optimal ∆t int to extend the lean-stability limit for stable lean operation. 51 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="52"?> 3.1.3 Effects of the Number of Spark Discharges Based on the above results for the two-stage spark discharge, increasing the number spark discharges is considered to extend the lean-stability limit further. Figure 10 shows a comparison of spark discharge characteristics between the twoand eight-stage spark discharge by the 12-volt ignition system. Fig. 10: Comparison of a) discharge current, b) discharge energy release rate and c) integrated discharge energy for lean operation at l = 1.7 with twoand eight-stage spark discharge by 12-volt ignition system For the eight-stage spark discharge, ∆t int is set to 0.2 ms because the lean operation at l = 1.7 is too unstable for the eight-stage spark discharge with ∆t int = 0.3 ms. This means that ∆t int should be changed, depending on the number of spark discharge. As Figs. 10a and 10b show, eight peaks of the discharge current and the discharge energy release rate are appeared at interval of 0.2 ms, which leads to the increase of the discharge duration. However, as Fig. 10c shows, the discharge energy is almost same, despite the relatively large difference of the discharge duration. The resulting changes in COV of IMEP with l are compared in Fig. 11 for the two data sets. Contrary to expectation, the lean-stability limit cannot be extended further by the eight-stage spark discharge. Rather, more unstable lean operations are observed at the same operating point. It should be noted that for the eight-stage spark discharge, the discharge current of the multiple spark discharges needs to be higher during the discharge duration to have a strong influence on the lean-stability limits. 52 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="53"?> Fig. 11: Comparison of COV of IMEP as a function of excess-air ratio () for lean operation with twoand eight-stage spark discharge by 12-volt ignition system 3.2 Lean Operation using 48-volt Ignition System 3.2.1 Comparison of 12-volt and 48-volt Ignition System As a method for improving the stability of lean operation, this work fabricates a 48-volt inductive ignition system to effectively implement the multiple spark discharges by varying the time interval between spark discharges and the number of spark discharges. First, the result of the 48-volt ignition system is compared to that of the 12-volt ignition system. Figure 12 presents an example of the primary current from the two-ignition coils when the eight-stage spark discharge is implemented by the 12-volt (upper) and 48-volt ignition system (lower). Fig. 12: An example of primary current for eight-stage spark discharge by a) 12-volt and b) 48-volt ignition system 53 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="54"?> Each trace of primary current from ignition coil 1 and 2 is indicated as solid line and dotted line, respectively. The time interval between spark discharges (∆t int ) is set to 0.2 ms for both ignition systems. As explained in the section for “Ignition System”, the twin ignition coils are applied to each cylinder and connected to a spark discharge timing controller capable of transmitting signals to each coil, respectively. To charge the primary current up to 12 ampere just before the first spark discharge, the charging time (dwell time) of the primary coil is set to 9.0 ms for the 12-volt ignition system and 0.9 ms for the 48-volt ignition system. The charging time for the rest of three spark discharges is set to 0.2 ms due to ∆t int . It should be noted that for the 48-volt ignition system, the charging time can be significantly reduced from 9.0 ms to 0.9 ms to reach the same amount of the primary current. In the same manner, the more primary current can be charged during the same charging time of 0.2 ms. Figure 13 shows a comparison of spark discharge characteristics between the 12and 48-volt ignition system for the eight-stage spark discharge. Fig. 13: Comparison of a) discharge current, b) discharge energy release rate and c) integrated discharge energy for lean operation at l = 1.7 with eight-stage spark discharge by 12-volt and 48-volt ignition system Clearly, the higher discharge current and the discharge energy release rate are observed during the spark discharge, which leads to the more discharge energy even for the shorter discharge duration. To facilitate the comparison, the resulting changes in COV 54 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="55"?> of IMEP with l for the eight-stage spark discharge by the 48-volt ignition system are plotted in Fig. 11, which is shown in Fig. 14. Fig. 14: Comparison of COV of IMEP as a function of excess-air ratio (l) for lean operation with twoand eight-stage spark discharge by 12-volt and 48-volt ignition system Because of the ability of the 48-volt ignition system, the eight-stage spark discharge strategy allows more stable lean operation, with COV of IMEP less than 3% up to l = 2.1. The ability to operate stably at such lean conditions may be attributed to a more stable flame initiation offered by supplying the higher spark discharge energy continuously to the flame kernel. 3.2.2 Effects of the Number of Spark Discharges Based on the 48-volt ignition system, this section examines the effects of the number of spark discharges on the lean-stability limit. Figure 15 shows the discharge energy release rate (solid line) and the integrated discharge energy (dotted line) of a) two-, b) four-, c) sixand d) eight-stage spark discharge for the lean operation at l = 1.7 with the 48-volt ignition system. 55 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="56"?> Fig. 15: Discharge energy release rate (solid line) and integrated discharge energy (dotted line) of a) two-, b) four-, c) sixand d) eight-stage spark discharge for lean operation at l = 1.7 with 48-volt ignition system All ∆t int is set to 0.2 ms. The ten-stage spark discharge cannot be implemented in this work because of the risk of coil explosion caused by overheating. As can be expected from Fig. 10, both the discharge duration and the discharge energy are increased with the number of spark discharges. However, it should be noted that the level of discharge energy release rate is getting lower during the multiple spark discharges. The resulting changes in COV of IMEP with l are compared in Fig. 16 for the four data sets. 56 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="57"?> Fig. 16: Comparison of COV of IMEP as a function of excess-air ratio (l) for lean operation with two-, four-, sixand eight-stage spark discharge by 48-volt ignition system As can be seen, the lean-stability limit can be extended with the increase of the number of spark discharges. Eventually, best results are achieved by the eight-stage spark discharge, which allows stable operation up to l = 2.1. However, it should be noted that the degree of the extension of lean-stability limit is not proportional to the number of spark discharges. 3.2.3 Effects of Time Interval between Spark Discharges To examine the effects of the number of spark discharges on the lean-stability limit, the eight-stage spark discharge is implemented by varying ∆t int . Figure 17 shows the discharge energy release rate (solid line) and the integrated discharge energy (dotted line) of the eight-stage spark discharge with ∆t int of a) 0.1 ms, b) 0.15 ms, c) 0.2 ms and d) 0.3 ms for the lean operation at l = 1.7 with 48-volt ignition system. As might be expected, both the discharge duration and the discharge energy are increased with the increase of ∆t int . Instead, the lower level of discharge energy release rate is observed for the longer ∆t int . 57 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="58"?> Fig. 17: Discharge energy release rate (solid line) and integrated discharge energy (dotted line) of eight-stage spark discharge with ∆t int of a) 0.1 ms, b) 0.15 ms, c) 0.2 ms and d) 0.3 ms for lean operation at l = 1.7 with 48-volt ignition system The resulting changes in COV of IMEP with l are compared in Fig. 18 for the four data sets. As can be seen, by increasing ∆t int from 0.0 ms to 0.2 ms, the lean-stability limit can be extended with a substantial reduction of COV of IMEP. However, COV of IMEP starts to increase again by increasing ∆t int from 0.2 ms to 0.3 ms, and then the stable lean operation cannot be achieved at l = 2.1. Based on these observations, it can be concluded that of critical importance to realizing improved combustion stability for stable lean operation is to set optimal combination of the time interval between spark discharges and the number of spark discharges. 58 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="59"?> Fig. 18: Comparison of COV of IMEP as a function of excess-air ratio (l) for lean operation with eight-stage spark discharge with ∆t int = 0.1, 0.15, 0.2 and 0.3 ms 4 Conclusions This work experimentally investigates the use of multiple spark discharges by varying the time interval between spark discharges and the number of spark discharges for extending the lean-stability limits in a four-cylinder engine. The following conclusions are offered: 1. The results for the single-stage spark discharge show that the more extension of lean-stability limit can be achieved by the two ignition coils. This happens because a combination of the higher energy and the longer discharge duration, compared to single ignition coil. Especially, the discharge energy is an important parameter impacting the lean operation. 2. For the two-stage spark discharge, the lean-stability limit can be extended with the increase of the time interval between spark discharges (∆t int ) from 0.0 to 0.3 ms. (∆t int = 0.0ms is the case of single-stage spark discharge.) However, for ∆t int > 0.3 ms, COV of IMEP starts to increase with the increase of ∆t int . This indicates that for the multiple spark discharges, it is critically important to set the optimal ∆t int for stable lean operation. 3. Compared to the two-stage spark discharge, the eight-stage spark discharge is not very effective for lean operation when using the 12-volt ignition system. In contrast, best results are achieved with the eight-stage spark discharge by the 48-volt ignition system, which allows stable operation up to l = 2.1. This happens because the more primary current can be charged during the same charging time, which leads to the higher discharge energy. 4. For the 48-volt ignition system, the lean-stability limit can be extended with the increase of the number of spark discharges. The most stable lean operation can 59 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="60"?> be achieved by the eight-stage spark discharge with ∆t int = 0.2 ms. However, the degree of the extension of lean-stability limit is not proportional to the number of spark discharges. 5. For the eight-stage spark discharge, the lean-stability limit can be extended by increasing ∆t int from 0.0 ms to 0.2 ms. However, COV of IMEP starts to increase again by increasing ∆t int from 0.2 ms to 0.3 ms, and then the stable lean operation cannot be achieved at l = 2.1. Based on these observations, it can be concluded that of critical importance to realizing improved combustion stability for stable lean operation is to set optimal combination of the time interval between spark discharges and the number of spark discharges. References [1] Environmental Protection Agency, “Regulations for Emissions from Vehicles and Engines,” https: / / www.epa.gov/ regulations-emissions-vehicles-andengines, Accessed 3 May 2021. [2] California Environmental Protection Agency Air Resources Board, “Low-Emission Vehicle Regulations & Test Procedures,” https: / / ww2.arb.ca.gov/ our-work/ programs/ advanced-clean -carsprogram/ lev-program/ low-emission-vehicle-regulations-test, Accessed 3 May 2021. [3] Joshi, A., “Review of Vehicle Engine Efficiency and Emissions,” SAE Int. J. Advances & Curr. Prac. in Mobility 2(5): 2479-2507, 2020, https: / / doi.org/ 10.4271/ 2020-01-0352. [4] Hwang, K., Hwang, I., Lee, H., Park, H. et al., “Development of New High-Efficiency Kappa 1.6 L GDI Engine,” SAE Technical Paper 2016-01-0667, 2016, https: / / doi.org/ 10.4271/ 2016-01 -0667. [5] Lee, B., Oh, H., Han, S., Woo, S. et al., “Development of High Efficiency Gasoline Engine with Thermal Efficiency over 42%,” SAE Technical Paper 2017-01-2229, 2017, https: / / doi.org/ 10.42 71/ 2017-01-2229. [6] Matsuo, S., Ikeda, E., Ito, Y. and Nishiura, H., “The New Toyota Inline 4 Cylinder 1.8L ESTEC 2ZR-FXE Gasoline Engine for Hybrid Car,” SAE Technical Paper 2016-01-0684, 2016, https: / / doi.org/ 10.4271/ 2016-01-0684. [7] Han, D., Han, S., Han, B. and Kim, W., “ Development of 2.0L Turbocharged DISI Engine for Downsizing Application,” SAE Technical Paper 2007-01-0259, 2007, https: / / doi.org/ 10.4271/ 2 007-01-0259. [8] Jung, D. and Lee, S., “An investigation on the potential of dedicated exhaust gas recirculation for improving thermal efficiency of stoichiometric and lean spark ignition engine operation,” Appl. Energy 228(15): 1754-66, 2017, https: / / doi.org/ 10.1016/ j.apenergy.2018.07.066. [9] Dale, J.D., Checkel, M.D. and Smy, P.R., “Application of high energy ignition systems to engines”, Prog. Energy Combust Sci. 23(5-6): 379-98, 1997, https: / / doi: 10.1016/ S0360-1285(97 )00011-7. [10] Zhi, W., Hui, L. and Reitz, R., “Knocking combustion in spark-ignition engines,” Prog. Energy Combust Sci. 61: 78-112, 2017, https: / / doi.org/ 10.1016/ j.pecs.2017.03.004. [11] Heywood J., “Internal Combustion Engine Fundamentals,” New York: McGraw-Hill, 1988. 60 Dongwon Jung, Kiseon Sim, Jinyoung Jung, Wongyu Kim, Yousang Son, Sungwook Lee <?page no="61"?> [12] Newhall, H.K., “Kinetics of Engine-generated Nitrogen Oxides and Carbon Monoxide”, Proc. Comb. Inst. 12(1): 603-613, 1969, https: / / doi: 10.1016/ S0082-0784(69)80441-8. [13] Jung, D., Sasaki, K., Sugata, K., Matsuda, M. et al., “Combined effects of spark discharge pattern and tumble level on cycle-to-cycle variations of combustion at lean limits of SI engine operation,” SAE Technical Paper 2017-01-0677, 2017, http: / / dx.doi.org/ 10.4271/ 2017-01-0677. [14] Jung, D., Sasaki, K. and Iida, N., “Effects of Increased Spark Discharge Energy and Enhanced In-Cylinder Turbulence Level on Lean Limits and Cycle-to-Cycle Variations of Combustion for SI Engine Operation,” Appl. Energy 205(1): 1467-77, 2017, https: / / doi.org/ 10.1016/ j.apener gy.2017.08.043. [15] Kuo, T.K., “What Causes Slower Flame Propagation in the Lean-Combustion Engine? ,” J. Eng. Gas Turbines Power 112(3): 348-56, 1990, https: / / doi.org/ 10.1115/ 1.2906502. [16] Ignition Systems for Gasoline Engines, eds. Günther, M. and Sens, M., Springer, Cham, Switzerland, 2017, ISBN 978-3-319-45503-7. [17] Ignition Systems for Gasoline Engines, eds. Günther, M. and Sens, M., expert verlag, Tubingen, Germany, 2018, ISBN 978-3-8169-3449-3. [18] Marode, E., “The mechanism of spark breakdown in air at atmospheric pressure between a positive point and a plane”, J. Appl. Phys 46(5): 2005-15, 1975, https: / / doi.org/ 10.1063/ 1.321882. 61 An Investigation of Multiple Spark Discharge Strategy using 48-volt Ignition System for Extending Lean-Sta‐ bility Limit in a Gasoline Engine <?page no="63"?> Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System Tobias Michler, Olaf Toedter, Thomas Koch Abstract: In this work, the influence of pressure on the electrical parameters of a standard automotive ignition system and the characteristic spectra of the ignition spark were shown. The spectrum of the breakdown shows the strongest change, from a relatively discrete N 2 band spectrum to a bright continuum spectrum with intense NII line radiation in it. In the arc discharge, the metal lines from the electrode material dominate the shape of the spectrum. These lines can be used to determine the excitation temperature and to estimate the intensity with which the electrode material is released by the thermionic effects of the discharge. Lastly, the shape of the glow discharges is not affected by the pressure, only their intensity decreases. Because of the differences in the spectra, several methods are needed to determine different temperatures or other plasma characteristics. 1 Introduction 1.1 Motivation The phases of the ignition spark were first described by the observations of Maly [1-8]. He divided the spark into the three phases of the breakdown, the arc and the glow discharge, which differ of their plasma-physical mechanism of electron supply. These different mechanisms resulting in different inflammation and erosion behavior of the phases, which are summarized in Fig. 1. Inflammation probability Wear Breakdown Arc discharge Glow discharge Fig. 1: Connection between the inflammation probability and the potential of electrode wear of the three ignition spark phases. <?page no="64"?> In addition to the phases, the knowledge of the electrical structure of an ignition system is necessary. Therefore, Fig. 2 shows as an example the schematic of an electrical circuit of a standard automotive ignition system, consisting of inductors, capacitors and resistors. The location of these resistances affects the discharge characteristics of the remaining elements by increasing or decreasing their time constants, depending on whether the element is an inductance or capacitance. A closer look to the capacitances of the system, the spark plug comes more into the focus. It can be divided into a capacitance in front of the suppressor resistance C SP1 and behind it C SP2 , resulting in different discharge behavior of these two capacitances [9]. Especially the second one is interesting, because it is not inhibiting by the resistance and can discharge extremely fast. L Sec C Coil C Cable C SP1 C SP2 R SP R Coil Secondary coil Spark plug Primary coil Trigger L prim R prim R SP C SP2 C SP1 U Bat 12 V Ignition cable Fig. 2: Schematic of the electric circuit of an ignition system (inspired by [10]) Therefore, the connection between ignition system design, the ignition phases, the erosion and inflammation behavior need further observation. In this work, a study on the influence of the pressure on the spectrum of the ignition phases is done. 1.2 Plasma-physical Basics of Spark Ignition In the following subsection, the plasma physical basics of the three phases are roughly outlined to illustrate some special features of the measurements. 1.2.1 Breakdown The breakdown process begins with the avalanche or townsend mechanism described by Townsend [11-13]. It begins with a start electron supplied by radiation, field emission or similar. This electron is accelerated by the electric field and accumulates 64 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="65"?> energy. If the energy is too low for ionization, the electron-particle collisions are elastic and there is no energy loss for the electrons. Once the energy exceeds the ionization energy of the electron collision partner, the particle is ionized and the colliding electron loses its energy by inelastic collisions. Afterwards, the two electrons move further in the direction of the anode and get accelerated, and the process repeats. The resulting ions move towards the cathode and release new electrons at this electrode by secondary electron emission. These electrons starting new avalanches until a conductive channel is formed. This process requires more than one avalanche to form this channel. As soon as the electron number reaches a critical value of 10 8 [14], the space-charge cannot be neglected compared to the external field, which marks the transition to the so-called streamer mechanism described by Loeb, Meek and Raether [15-17]. The fast electrons leave the slow ions behind, forming a negatively charged streamer head and positively charged streamer tail. Increased recombination processes occur between these two regions. In addition, ionization and recombination are enhanced in front of the streamer head due to the high field strengths. This leads to radiation, which can also ionize the gas again even at a greater distance. Therefore, the streamer can move very fast forward to the anode. The time to form a conductive channel with this process is in the range of ten ns. Moreover, the streamer can form a conductive channel with only a single avalanche. 1.2.2 Arc Discharge The arc discharge is characterized by its high current densities, which can only be delivered by thermionic-field emission [18]. In this process, the electrode material melts and evaporates in a very localized area of some tens µm in each plasma channel [19]. This area is called cathode spot. In addition, atomic oxygen produces in the plasma column form volatile metal oxides with the electrode material, transporting them into the gas [10]. There, the metal can be ionized and form a strong electric field to the cathode in which electrons can be released from the cathode by field emission. These processes are very efficient in supplying electrons and therefore require a low burning voltage of the plasma [20]. Furthermore, the metal atoms can be detected with a spectrograph and are thus a direct indicator of the arc discharge. 1.2.3 Glow Discharge In contrast to the arc discharge, in the glow discharges electrons supply happens in the gas takes place by electron collisions. The spatial structure of the glow discharge is the most complex of the three ignition phases and can be divided into three relevant areas [21, 22]. In the negative glow area, high ionization lead to high inhomogeneous electrical field strength near the cathode, which results in a high cathode fall and burning voltage of the discharge [23]. This area is by far the brightest of the glow discharge. Next is the area of the positive column, which is significantly darker. In this area, electrons are accelerated by the constant homogenous field, resulting in uniform radiation. Near the 65 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="66"?> anode, electrons accumulate, flow off into it and form the region of the so-called anode glow, which is slightly brighter than the radiation of the positive column. 2 Experimental Setup and Methodology The ignition coil used in these experiments is a commercial cassette coil system for passenger cars. It has a secondary energy of 90 mJ and an initial spark current of 150 mA at atmospheric pressure. A commercial tip-to-tip j-gap spark plug with an IrRh10 alloy at the center electrode and PtRh30 at the ground electrode is used. The electrode diameters and spacing are 0.6 mm and 0.9 mm, respectively. The suppressor resistance is 4.2 kΩ and the total capacitance of the spark plug is 12 pF. Due to the geometry of the spark plug, this capacitance can be divided into the two capacitances C SP1 and C SP2 with 4 and 8 pf, respectively. Ignition coil and spark plug are connected by a high-voltage cable without resistance. The spark plug is fitted into a pressure chamber filled with synthetic air (79 vol.-% N 2 and 21 vol.-% O 2 ) 2.1 Spectroscopic and Electrical setup 2.1.1 Spectroscopic Setup A scheme of the experimental electrical and spectroscopic setup used is shown in Fig. 3. The core element is the Acton SP2556 grating spectrograph from Princeton Instruments. Its focal length is 500 mm and has an adjustable inlet slit width between 10 and 3000 µm. In addition, it has three different gratings with 150, 600 and 1200 lines/ mm mounted on a turret, which can be easily switched between. Mounted on the spectrograph is a PI-MAX 2 ICCD camera from Princeton Instru‐ ments with a multichannel plate (MCP) amplifier. The chip used has a dimension of 1024x256 pixels in direction of the wavelength and axial to the spark plug gap, respectively. Therefore, an analysis of the plasma over the spark plug gap is possible. The ignition spark uses 58 pixels in the height, which results in a resolution of 16 µm/ pixel. During the investigations, 20 additional pixels are evaluated because the glow of the glow discharge protrudes beyond the center electrode. Three combinations of slit width and gratings are used in this work, resulting in different instrumental broadenings of the system. These broadenings were meas‐ ured using a LOT-Oriel LSP035 mercury-argon arc lamp. For high resolution, the 1200 lines/ mm grating in combination with a 10 µm slit is used. This results in an instrumental broadening of less than 0.1 nm. A medium resolution was achieved with the 600 lines/ mm grating and a slit width of 200 µm, resulting in an instrumental broadening of 0.7 nm. To detect a wide range of wavelengths, the 150 lines/ mm with the 200 µm slit width is used. The determined instrumental broadening was 2.7 nm. 2.1.2 Electrical Setup For measuring the high-voltage a Tektronix P6015A is used. In addition, two Pear‐ son 2877 current monitors in the high-voltage and ground path are used to measure the 66 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="67"?> currents. A LeCroy Waverunner 6030A oscilloscope with adjustable input impedance is used to acquire data from the devices. The P6015A is connected via its compensation box to an electric circuit that cuts off the high-voltage peaks of the breakdown to pre‐ tend the A/ D-converter of the oscilloscope from being overdriven at higher resolutions. This circuit is then connected to the high impedance input of the oscilloscope. The two current clamps are connected to the oscilloscope with a 50 Ω input. Battery DC 12 V High-voltage probe Charging Signal Oscilloscope iCCDcamera Spectrograph Spark plug Ignition coil PC Spectrograph Function generator CCD Gate Trigger Lens Trigger Oscilloscope PTG Pressure chamber Current clamps Fig. 3: Spectroscopic and electrical measurement setup (based on [24, 25]) 2.2 Methodology 2.2.1 Methodology of Data Collection The gating of the camera is recorded with the oscilloscope. Therefore, the point in time of the ignition spark and the acquired spectrum can be matched with this signal. This technique allows to differentiate between the different spark phases under consideration of the measured voltage and currents. To increase the accuracy of the following post processing, 50 spectra are acquired and then averaged. The determination of the three phases had to be performed with different exposure times and gains due to their different time intervals and intensities. In the arc and glow phase of the spark, the gain was in maximum setting and the exposure time was 67 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="68"?> (1) (2) adjusted in the range of 100 to 200 µs to sustain acceptable signal-to-noise ratios of the signals. The short duration of the breakdown resulted in short exposure times of 2 µs. 2.2.2 Determination of Electrical Energy With the voltage curve of the ignition spark, the supplied energy to the different spark phases can be determined. Fig. 4 shows the voltage profile of 200 sparks. For clarification of the relevant voltage levels, the breakdown voltage is cut off. The two voltage levels with a difference of approximately 300 V are clearly visible. It is also noticeable, that the arc discharge is not found over the entire duration of the ignition spark and is found at higher currents. 0 -500 -1000 -1500 -2000 Voltage / V 0 200 400 600 800 1000 1200 1400 1600 Time / µs Voltage level of the arc discharge Voltage level of the glow discharge Fig. 4: Persistent plot of 200 ignition sparks at a pressure of 6 bar under an atmosphere of synthetic air The calculation of the electrical energy of each phase can be done with the following two equations (1) and (2). E BD = 12 C SP 2 × U BD 2 E D = u D (t) − R SP × i D (t) × i D (t)dt In these equations, E is the electrical energy of the breakdown (BD) and the inductive discharge (D). The other parameters are the capacitance C SP 2 of the spark plug, the breakdown voltage U BD , the discharge current and voltage i D (t) and u D (t), and the spark plug resistance R SP . 68 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="69"?> 3 Results 3.1 Pressure Influence on Electrical Parameters First, the electrical behavior of the ignition system used must be evaluated under different pressures. Starting with the behavior of the breakdown voltage U BD in the upper graph of Fig. 5. In addition, the static breakdown voltage U Stat , calculated by paschen’s law, and the ratio of both U BD / U Stat are shown here. At low pressures below five bar, the measured breakdown voltage is much higher compared to the static voltage. This is can be explained by the absence of a start electron for the avalanche process in the dark and shielded pressure chamber [26]. With increasing pressure, this start electron is more likely supplied by field emission from the electrodes [14], therefore the ratio decreases. In addition, the voltage slope of charging the secondary capacitances decreases also due to electrical behavior of the ignition coil, which gives the discharge more time to form a conductive channel. The increase in breakdown voltage results in an increase in the energy stored by the capacitances and therefore decrease the energy stored in the inductance. As a consequence, the initial current of the inductive discharge also decreases, which is shown in the bottom left graph of Fig. 5. In the case of the ignition coil used, the current decreases in two steps. Between atmospheric pressure and eight bar absolute pressure, the current is reduced from -150 to -140 mA. In the second step, this decrease is much stronger and reduces this initial current to approximately -100 mA. Finally, these effects also the influence the duration of the glow and arc discharge, which is shown in Fig. 5 (bottom right). The area between both discharges shows the duration of the respective discharge. The top line shows the total duration of the ignition spark. As the pressure increases, the spark shortens due to the lower energy in the inductance. However, an increase in the arc duration can be observed. The higher pressures support higher charge carrier densities, which can lead to arc discharges. On the other hand, the reduction of initial current can lead to the opposite effect and reduces the arcs duration slightly, which explains the maximum between six and eight bar and the subsequent reduction. 69 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="70"?> 0 2 4 6 8 10 12 14 16 -0.16 -0.15 -0.14 -0.13 -0.12 -0.11 -0.10 0 2 4 6 8 10 12 14 16 0 5 10 15 20 25 30 U BD U BD / U Stat U Stat Pressure / bar Voltage / kV 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Voltage ratio / - Initial current / A Pressure / bar 0 2 4 6 8 10 12 14 16 0 200 400 600 800 1000 1200 1400 1600 1800 Phase duration / µs Pressure / bar Arc Glow Fig. 5: Effect of the pressure on the measured breakdown voltage (U BD ), the static breakdown voltage (U Stat ) and their ratio on top, the initial current of the inductance on the left, and the duration of arc and glow discharge on the right 3.2 Spectra of the Breakdown At atmospheric pressure, the spectrum is shaped by the nitrogen bands of the first positive system (FPS) between 600 and 770 nm, the second positive system (SPS) between 300 and 430 nm, as well as the first negative system (FNS) with its transition at 391.4 nm, as shown in the upper graph of Fig. 6. In the at the right edge of the spectrum, the triple transition of the atomic oxygen at 777.2-777.5 nm is also visible. This shape of the spectrum is preserved up to a pressure of four bar, above which the spectrum suddenly changes to continuum radiation with intense atomic nitrogen radiation. Three examples of this continuum are shown in the bottom graph of Fig. 6. The change in the shape of the spectrum indicates a change of the physical mechanism of the breakdown. At low pressures, the breakdown appears to be initiated by the townsend mechanism, in which the main reason for the formation of a conductive channel is electron collisions. On the other hand, the continuum radiation appears to be a mixture of different physical mechanisms such as, photo-ionization, electron-ion interaction (free-free) and recombination (electron-ion interaction, free-bound) [27-29]. These mechanisms occur in the process of the streamer formation described in chapter 1.2.1. 70 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="71"?> 300 400 700 800 0E+0 2E+9 5E+9 7E+9 300 400 500 600 700 800 0E+0 1E+12 2E+12 3E+12 Intensity / a. u. W avelength / nm nitrogen bands of second positive system O I First negative system ∆v = 0 v' = 0 nitrogen bands of first positive system Intensity / a. u. W avelength / nm 5 bar x3 10 bar 15 bar N II N I O I N II N II N II Fig. 6: Spectrum of the breakdown at atmospheric pressure (upper graph) and higher pressures (lower graph) at a temperature of 20 °C and synthetic air As the pressure increases, the electrical energy stored in the capacitances of the ignition system also increases, which is shown in the left graph of Fig. 7. For calculation of the energy, the capacitor C SP2 was used. The resulting energy with using this capacitance is in the range of 0.2 to 4 mJ. This energy range is slightly underestimated because it is assumed, that the energy is only supplied by this one capacitance. In reality, the other capacitances of the ignition system discharge slightly during this phase and therefore add some of their energy to the breakdown. The right graph of Fig. 7 shows the intensity of the continuum radiation as a function of the pressure. Since the continuum appears above pressures of four bar, the graph starts at this pressure. At first, the intensity increases nearly linear with the pressure until eight bar. Afterwards it has a degressive tendency. However, in this pressure range, the supplied electrical energy increases almost linear, which means there is a gap between supplied energy and radiation yield. The energy from this difference can probably be found in dissociation processes or gas heating. 71 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="72"?> 0 2 4 6 8 10 12 14 16 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0 2 4 6 8 10 12 14 16 1E+9 1E+10 Pressure / bar Intensity of continuum radiation / a. u. Electrical supplied energy / mJ Pressure / bar Fig. 7: Electric supplied energy (left) and intensity of the continuum radiation (right) 3.3 Spectra of the Arc Discharge 300 400 700 800 0E+0 1E+9 2E+9 3E+9 4E+9 5E+9 300 400 700 800 0.0E+0 5.0E+8 1.0E+9 1.5E+9 2.0E+9 Intensity / a. u. W avelength / nm pseudo continuum of nitrogen bands + Metal lines O I SPS ∆v = 0 v' = 0 5 bar 10 bar 15 bar Intensity / a. u. W avelength / nm 350 360 370 Intensity / a. u. Wavelength / nm High resolution: Rh / Ir radiation Fig. 8: Spectrum of the arc discharge at atmospheric pressure (upper graph) and higher pressures (lower graph) at a temperature of 20 °C and synthetic air. The small graph shows a high-resolution spectrum between 345 and 375 nm 72 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="73"?> (3) (4) Fig. 8 shows the spectra of an arc discharge at atmospheric pressure in the upper and at higher pressures at the lower graph. All spectra are obtained from the near cathode region of the spark plug gap. At atmospheric pressure, the species of N 2 are overlaid by metal lines of iridium and rhodium from the center electrode material. Due to the low resolution used to obtain the spectrum, the different lines are not resolved and form a so-called pseudo-continuum, which is shown resolved in the small lower graph. With increasing pressure, the radiation from the nitrogen molecules is masked by the metal lines and can no longer be detected. The radiation of the metal lines is a direct result of the melting and evaporation of the electrode material due to the thermionic-field emission of the arc discharge. The intensity of the metal radiation is proportional to the particle density of the observed material, assuming a constant excitation temperature. The formulaic connection between the intensity ε ji and the particle density n 0 is described by equation (3) ε ji = ℎ × c o 4 × π × λ ji × A ji × n j with the Boltzmann distribution equation (4) n j = g j U (T ) × n 0 × e − Ej kB × Texc The equations contain the Planck constant ℎ, the speed of light in vacuum c o , the wavelength of the transition λ ji , the transition probability A ji , the occupation density of the particle n j , the statistic weight g j , state sum U (T ), the energy of the upper state E j , the Boltzmann constant k B and the excitation temperature T exc . Therefore, the intensity of a line is connected to the material evaporated by the cathode spot of the arc discharge. The more electrode material there is in the gas, the greater the probability that some of the released atoms will leave the electrode and thus wear it out. The left graphic of Fig. 9 shows the supplied electrical energy, power and exposure time as a function of the pressure. Up to a pressure of three bar, the energy and power of the arc discharge increase. After that, both remain at a constant level until the pressure reaches nine bar, at which point they decrease again. This is explained by the reduction of the initial current, which makes it harder to reach the necessary current density for an arc discharge. On the right graph of Fig. 9 the intensity of the 5d 7 6s( 5 F)6p à 5d 8 ( 3 F)6s iridium transition in dependence of the pressure is shown. The intensity of this line is not to mixed up with the intensity of the pseudo-continuum from Fig. 8. The rise of the intensity at a pressure of seven bar is obviously. Even with decrease of the electrical supplied power and energy, the intensity remains at a high level or is even increases with pressure. A reason for this behavior could be the contraction of the plasma channel and the cathode spot area due to 73 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="74"?> the higher charge carrier particle density which can be achieved at these pressures. The decrease between 13 and 15 bar correlates with the decrease of supplied energy, power and the reduction of the initial current of the discharge. 0 2 4 6 8 10 12 14 16 0 10 20 30 40 50 60 70 Mean supplied power / W mean supplied power mean supplied energy exposure time Pressure / bar 0 1 2 3 4 5 6 Mean supplied energy / mJ 20 40 60 80 100 Exposure time / µs 0 2 4 6 8 10 12 14 16 0.0E+0 5.0E+7 1.0E+8 1.5E+8 Pressure / bar Intensity / a. u. Fig. 9: Influence of the pressure on the supplied power, energy and exposure time (left), and the intensity of ir-radiation 3.4 Spectra of the Glow Discharge The spectrum of the glow discharge has only minor changes with increasing pressure, as shown in Fig. 10. Like in the breakdown, the second positive system majorly shapes the spectrum. In the negative glow of this discharge, strong emissions of the FNS Δv = 0 are observed. In the remaining part of the plasma column this emission is not very bright. The strong presence of the FNS in the negative glow is a typical indicator for this area of the glow discharge, due to the high ionization processes. 300 400 700 800 0.0E+0 5.0E+8 1.0E+9 1.5E+9 2.0E+9 2.5E+9 3.0E+9 Intensity / a. u. W avelength / nm 1 bar 5 bar 10 bar 15 bar SPS FPS O I v' = 0 ∆v = 0 FNS ∆v = 0 v' = 0 Fig. 10: Spectrum of the glow discharge in the negative glow at different pressures and a temperature of 20 °C with synthetic air 74 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="75"?> In the left graph of Fig. 11, the electrical supplied power and energy are shown. With increasing pressure, both decrease slightly. At seven bar, the exposure time was doubled to sustain the signal-to-noise ratio of the measurements, so the supplied energy increases at this pressure. The dependence of the intensity across the entire gap as a function of the pressure is shown in the right graph of Fig. 11. It shows a decreasing trend until a pressure of seven bar. From this point on, the intensity does remains on a constant level. This constant level could be a remnant of the data acquisition process, in which measurements below a certain level were not recorded. Remarkable is the difference between one and two bar. At one bar, the discharge is significantly brighter and varies much less than at two bar. This can be explained by the spatial propagation in the radial direction of the discharge. At one bar, the discharge propagates the whole cathode, therefore its intensity remains very constant. With increasing pressure, the discharge contracts and cannot propagate the entire cathode anymore. Therefore, it is possible, that the discharge misses the optical path of the lens and the slit of the spectrograph. This results in lower intensities and the higher fluctuation of it. 2 4 6 8 10 12 14 5 10 15 20 25 P el E el Pressure / bar Supplied power / W Increase exposure time 0 2 4 6 8 10 Supplied energy / mJ 2 4 6 8 10 12 14 0E+0 1E+8 2E+8 3E+8 4E+8 5E+8 6E+8 7E+8 Pressure / bar Intensity / a. u. 25-75 % Percentile 1.5 IQR Median Mean Outlier Fig. 11: Electrical supplied power and energy, and Intensity of the N 2 SPS Δv = 0 band as a function of gas pressure. The boxes show the 25 and 75 percentiles, the circle is the median, the line in the boxes is the arithmetic mean and the whiskers show the standard deviation 4 Discussion The spectra of the three ignition phases differ significantly from each other. Therefore, it is necessary to obtain different methods of data acquisition to each phase and to apply different methods to determine plasma characteristics such as temperatures or particle densities. 75 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="76"?> The breakdown is the brightest phase of the ignition spark. Up to a pressure of three bar, N 2 band spectra can be observed, which can be used to determine rotational and vibrational temperature by comparing measured with simulated spectra. Above four bar, the continuum radiation could be used to determine particle densities and electron temperature. For these values, the absolute radiance of the plasma has to be obtained, which is not possible with the measurement setup used. Another approach is the determination of excitation temperature via the NII ions in the range of 500 nm using the line ratio method, which was performed, for example, by [30, 31]. In addition, the NII lines at 500 nm can be used to obtain the electron density due to stark broadening. This method is for example used in [32]. Due to the thermionic-field emission and the volatile oxides in the arc discharge, only metal lines could be observed. This makes it difficult to determine a temperature from the gas in this phase. However, the metal lines can be used to determine an excitation temperature of these metal lines by using the boltzmann-plot method. The prerequisite for this method is, that the metal lines used are carefully selected and the resolution has to be as high as possible to obtain accurate measurement results [33]. Nevertheless, the metal intensities can be used to estimate the intensity and erosion behavior of the observed arc discharge. In the glow discharge, the N 2 bands of the second positive system and the first negative system are easily measurable. Therefore, they can be used to determine the rotational and vibrational temperature of this discharge. Moreover, by calculating the rate coefficients, the electron temperature can be determined by the line ratio method using the first negative system and the second positive system, as shown in [34]. The spatial area used to calculate these temperatures has to be chosen carefully because the temperatures depend strongly on the axial height of the discharge [9, 24, 25, 35]. 5 Conclusion In this work, we showed a measurement setup and methods for determining the influence of the pressure on the spectra of the ignition phases. The spectra showed significant influence from the pressure and the observed ignition phase. Also, the results show, that the ignition spark cannot be described by a single plasma-physical parameter. Rather, it is necessary to apply different methods to determine characteristic plasma-physical quantities of the phase. Furthermore, it was shown that the measure‐ ment setup used is suitable to estimate the intensity of arc discharge in terms of the release of electrode material into the gas. In the next steps, different calculation methods will be applied to determine the characteristic plasma parameters for each phase. 76 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="77"?> References [1] H. Albrecht, W. H. Bloss, W. Herden, R. Maly, B. Saggau, and E. Wagner, “New Aspects on Spark Ignition,” in SAE Technical Paper Series, 1977, pp. 1-11. [2] H. Albrecht, R. Maly, W. H. Bloss, W. Herden, B. Saggau, and E. Wagner, “Neue Ergebnisse über die Entflammung durch den elektrischen Funken,” 4 th Satusseminar Kraftfahrzeug- und Straßenverkehrstechnik, 1977, 1977. [3] H. Albrecht, R. Maly, B. Saggau, E. Wagner, and W. Herden, “Neue Erkenntnisse über elektrische Zündfunken und ihre Eignung zur Entflammung brennbarer Gemische,” Automo‐ bild-Industrie, vol. 22, no. 4, pp. 45-50, 1977. [4] W. Herden, R. Maly, B. Saggau, and E. Wagner, “Neue Erkenntnisse über elektrische Zündfunken und ihre Eignung zur Entflammung brennbarer Gemische -2. Teil,” Automo‐ bild-Industrie, vol. 23, no. 2, pp. 15-22, 1978. [5] R. Maly et al., “Die drei Phasen einer elektrischen Zündung und ihre Auswirkungen auf die Entflammungseinleitung,” 5 th Statusseminar Kraftfahrzeuge und Straßenverkehr, 1978, 1978. [6] R. Maly, “Spark Ignition: Its Physics and Effect on the Internal Combustion Engine,” in Fuel Economy: in Road Vehicles Powered by Spark Ignition Engines, New York: Springer US, 1984, pp. 91-148. [7] R. Maly and M. Vogel, “Initiation and propagation of flame fronts in lean CH4-air mixtures by the three modes of the ignition spark,” Symposium (International) on Combustion, vol. 17, no. 1, pp. 821-831, 1979, doi: 10.1016/ S0082-0784(79)80079-X. [8] R. R. Maly, “Die Zukunft der Funkenzündung,” MTZ - Motortechnische Zeitschrift, vol. 59, no. 7, pp. XIX-XXIV, 1998, doi: 10.1007/ BF03226465. [9] T. Michler, W. Kim, O. Toedter, T. Koch, and C. Bae, “Influence of the Electrical Parameters of the Ignition System on the Phases of Spark Ignition,” in Ignition Systems for Gasoline Engines: 4 th International Conference, December 6-7, 2018, Berlin, Germany. Ed.: M. Günther, 2018, pp. 222-238. [10] J. Rager, Funkenerosion an Zündkerzenelektroden. Aachen: Shaker, 2006. [11] J. S. E. Townsend, Electrons in gases: London : Hutchinson’s scientific and technical publications, 1947 [12] J. S. E. Townsend, Electricity in Gases: Clarendon Press, 1915. [13] J. S. E. Townsend and H. T. Tizard, “The Motion of Electrons in Gases,” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 88, no. 604, pp. 336-347, 1913, doi: 10.1098/ rspa.1913.0034. [14] A. Küchler, Hochspannungstechnik: Grundlagen,Technologie, Anwendungen, 3 rd ed.: Springer-Verlag, 2009. [15] J. M. Meek and J. D. Craggs, Eds., Electrical breakdown of gases. Oxford: Clarendon Press, 1953. [16] L. B. Loeb, Basic processes of gaseous electronics. Berkeley, Calif. [u. a.]: Univ. of California Pr, 1955. [17] H. Raether, Electron avalanches and breakdown in gases. London: Butterworths, 1964. [18] W. Ramberg, “Über den Mechanismus des elektrischen Lichtbogens,” Ann. Phys., vol. 404, no. 3, pp. 319-352, 1932, doi: 10.1002/ andp.19324040305. 77 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="78"?> [19] B. Jüttner, “Cathode spots of electric arcs,” J. Phys. D: Appl. Phys., vol. 34, no. 17, R103-R123, 2001. [20] A. Güntherschulze, “Zusammenhang zwischen Stromdichte und Kathodenfall der Glim‐ mentladung bei Verwendung einer Schutzringkathode und Korrektion der Temperaturer‐ höhung des Gases,” Zeitschrift für Physik, vol. 49, no. 5, pp. 358-379, 1928, doi: 10.1007/ BF01337924. [21] A. Fridman and L. A. Kennedy, Plasma Physics and Engineering, Second Edition, 2 nd ed. Hoboken: CRC Press, 2011. [22] J. P. Raizer, Gas discharge physics. Berlin: Springer, 1991. [23] A. Güntherschulze, “Die Abhängigkeit des normalen Kathodenfalles der Glimmentladung von der Gasdichte,” Zeitschrift für Physik, vol. 49, no. 7, pp. 473-479, 1928, doi: 10.1007/ BF01333631. [24] T. Michler, O. Toedter, and T. Koch, “Spatial and time resolved determination of the vibrational temperature in ignition sparks by variation of the dwell time,” SN Applied Sciences, vol. 2, no. 7, p. 1311, 2020, doi: 10.1007/ s42452-020-3104-6. [25] T. Michler, O. Toedter, and T. Koch, “Measurement of temporal and spatial resolved rotational temperature in ignition sparks at atmospheric pressure,” Automot. Engine Technol., vol. 5, 1-2, pp. 57-70, 2020, doi: 10.1007/ s41104-020-00059-w. [26] K. B. Yahia, “Hybridmodellierung der Anfangsphase der Zündung einer Gasentladung,” Dissertation, Universität Karlsruhe, Karlsruhe, 2002. [27] S. Park, W. Choe, H. Kim, and J. Y. Park, “Continuum emission-based electron diagnos‐ tics for atmospheric pressure plasmas and characteristics of nanosecond-pulsed argon plasma jets,” Plasma Sources Science and Technology, vol. 24, no. 3, p. 34003, 2015, doi: 10.1088/ 0963-0252/ 24/ 3/ 034003. [28] S. Park, W. Choe, S. Youn Moon, and J. Park, “Electron density and temperature measure‐ ment by continuum radiation emitted from weakly ionized atmospheric pressure plasmas,” Appl. Phys. Lett., vol. 104, no. 8, 084103-1 - 084103-5, 2014, doi: 10.1063/ 1.4866804. [29] K. T. A. L. Burm, “Continuum radiation in a high pressure argon-mercury lamp,” Zeitschrift für Physik, vol. 13, no. 3, pp. 387-394, 2004, doi: 10.1088/ 0963-0252/ 13/ 3/ 004. [30] N. Minesi, P. Mariotto, G.-D. Stancu, and C. O. Laux, “Ionization Mechanism in a Thermal Spark Discharge,” in AIAA Scitech 2021 Forum, VIRTUAL EVENT, 01112021. [31] N. Minesi, S. Stepanyan, P. Mariotto, G. D. Stancu, and C. O. Laux, “Fully ionized nanosecond discharges in air: the thermal spark,” Plasma Sources Science and Technology, vol. 29, no. 8, 2020, doi: 10.1088/ 1361-6595/ ab94d3. [32] A. M. EL Sherbini, A. M. Aboulfotouh, and C. G. Parigger, “Electron number density measurements using laser-induced breakdown spectroscopy of ionized nitrogen spectral lines,” Spectrochimica Acta Part B: Atomic Spectroscopy, vol. 125, pp. 152-158, 2016, doi: 10.1016/ j.sab.2016.10.003. [33] A. Mašláni, V. Sember, T. Stehrer, and H. Pauser, “Measurement of Temperature in the Steam Arcjet During Plasma Arc Cutting,” Plasma Chemistry and Plasma Processing, vol. 33, no. 3, pp. 593-604, 2013, doi: 10.1007/ s11090-013-9443-y. 78 Tobias Michler, Olaf Toedter, Thomas Koch <?page no="79"?> [34] A. Zerrouki, H. Motomura, Y. Ikeda, M. Jinno, and M. Yousfi, “Optical emission spectroscopy characterizations of micro-air plasma used for simulation of cell membrane poration,” Plasma Physics and Controlled Fusion, vol. 58, no. 7, p. 75006, 2016, doi: 10.1088/ 0741-3335/ 58/ 7/ 075006. [35] D. Staack, B. Farouk, A. F. Gutsol, and A. Fridman, “Spatially Resolved Temperature Measurements of Atmospheric-Pressure Normal Glow Microplasmas in Air,” IEEE Trans. Plasma Sci., vol. 35, no. 5, pp. 1448-1455, 2007, doi: 10.1109/ TPS.2007.904959. 79 Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parameters of a TCI-Ignition System <?page no="81"?> Application of a time-resolved ignition spark measurement technique when using a power ignition system Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch Abstract: Current tendencies for increasing the efficiency of gasoline engines and thus achieving the legal CO 2 requirements include the use of diluted mixtures, lean mixtures or the use of pre-chamber spark plugs and combustion processes adapted to them. In order to reliably ignite lean, diluted mixtures, higher ignition energies are required - especially for highly charged engines. With conventional TCI ignition systems, however, these energies are inevitably accompanied by increased spark plug wear. It is necessary to adapt current ignition systems to higher ignition energies without increased spark plug wear at the same time. The objective of the study is to analyze the influence of the secondary current on the spark by using a special optical test rig. For this purpose, the emitted light emission of the different spark phases is observed (with special resolution) while varying the secondary current of the ignition system. The ignition system used is a BorgWarner system that allows the secondary current and the spark duration to be controlled by means of two combined ignition coils. Through regulating the secondary current strength and duration, higher energy can be delivered to the gas, resulting in a reliable inflammation. Furthermore, a limited secondary current minimizes spark plug wear. In order to assess these effects, the intensity of the metal lines in the radiation spectrum are considered as an indicator of spark plug wear. 1 Introduction 1.1 Motivation One possibility for improving the efficiency of internal combustion engines is the application of new combustion processes. For spark-ignited engines (SI-engines), lean or diluted mixtures [1-3] as well as pre-chamber spark plugs have the potential to increase the efficiency [4, 5]. In case of pre-chamber spark plugs, the high amount of remaining residual gas due to poor flushing increases the energetic requirement on the ignition system [6]. With leaner mixtures or mixtures diluted by exhaust gas <?page no="82"?> recirculation (EGR), the necessary ignition energy increases as well [7-9], which has a negative effect on the service life of spark plugs using a TCI system. 1.2 Spark Phases The ignition spark was decisively studied by Maly in the 1970`s and 80`s [10-17]. His studies showed, that three different plasmas, the breakdown, the arc and the glow discharge occur during an ignition spark. The phases differ in their physical mechanism of electron supply, resulting in different behavior of inflammation probability and spark plug wear, as shown in Fig. 1. Inflammation probability Wear Breakdown Arc discharge Glow discharge Fig. 1: Connection between the inflammation probability of the ignition phases (left, inspired by [15]) and the connection between of the phases and the spark plug wear (right) 2 Theoretical Background and Measurement Setup 2.1 Plasma-Physical Basics of Spark Ignition Although the ignition spark can be divided into three phases, only the physical basics of the arc and glow discharge are discussed below due to their importance in this work. 2.1.1 Arc Discharge The arc discharge is the most erosive of the three ignition phases. Its process of electron supply bases on thermionic-field emission, whereby the electrode material of the cathode is melted and vaporized [18] in a very local spot in the scale of some tens µm [19]. Moreover, the atomic oxygen produced in the plasma column of the arc discharge form volatile metal oxides and transports particles of the electrode material into the gas [20]. The metal ions accumulate near the cathode and form a strong inhomogeneous electrical field, which releases electrons by field emission out of the cathode. Due to the efficiency of these processes, the burning voltage of the arc discharge is very low [20] and in the range of the work function of the electrode material used. 2.1.2 Glow Discharge In the case of the glow discharge, the process of electron supply is carried out by the collision of electrons with gas particles, which are ionized as a result, making this phase nearly non-erosive compared to the arc discharge. This process of electron supply is very inefficient, resulting in high burning voltages of this discharge [21]. A feature of 82 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="83"?> this phase is the strong radiation of the nitrogen band emission, which can serve as an indicator for this type of discharge. 2.2 Principle of the Power-Ignition System A first prototype of the so-called Coupled Multi-Charge (CMC) ignition system has been presented in previous publications [22, 23]. Meanwhile a further developed system designed for mass production is available. The CMC-system is using a microcontroller to perform a closed-loop control of the primary and secondary current. With this, the system is capable to supply a continuous and adjustable current to the spark plug. The parameters for the secondary current level and burning duration are adjusted by the Engine Control Unit (ECU) via a communication interface with a digital 5 V signal, similar to the interface described in [24]. The technical key data of the CMC-system are summarized in the following Tab. 1. Maximum Output Power 150 W Maximum Secondary voltage (at 1 MΩ / / 25 pF) 42 kV Adjustable Secondary Current 35 mA - 150 mA Adjustable Burning Duration 100 µs - 4500 µs Tab. 1: Technical data of the CMC-system at a nominal supply voltage of 14 V A principal schematic of the CMC-system is shown in Fig. 2: Two identical transformers T1 (L 1 and L 2 ) and T2 (L 3 and L 4 ) are connected to one spark plug and are decoupled from each other via the high-voltage diodes D 1 and D 2 . The diodes are connected to ground via the shunt R s , which is used to measure the secondary current I s . The primary circuit is designed with the five semiconductor switches S 1 to S 5 and are interconnected as shown in Fig. 2, shunt R p is used to monitor the primary current I p . U B R p S 4 S 1 S 2 S 3 S 5 L 1 L 2 L 3 L 4 R s D 1 D 2 T1 T2 T2 T1 Fig. 2: Schematic of the CMC-ignition system with the primary and secondary circuit. On the right-hand side a fully assembled CMC-coil 83 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="84"?> Via the switches on the primary side the secondary current can be controlled and adjusted to the required level. The different ignition states are shown in Fig. 3 and the primary and secondary current data in Fig. 4. For each ignition cycle, the system is controlled in the following way: 1. Initial charge: Both inductances L 1 and L 3 are connected in series and are charged at the same time to the wanted energy level, which is needed to maintain the required breakdown voltage. The length of the initial charge is controlled by the ECU via the Dwell-Time (t dwell ). 2. Initial breakdown: Both inductances L 1 and L 3 are switched off at the same time initiating the high voltage breakdown at the spark plug. This state is maintained for ~100 µs to guarantee the growth of a stable ignition spark. 3. Extend ignition burning duration (CMC-Mode): Transformer T1 recharges and T2 discharges or vice versa, by doing this a continuous secondary current is supplied to the ignition spark. The toggling of the two transformer stages is continued as long as desired by the ECU (burning duration: t burn ). 4. Freewheeling: The freewheeling mode disconnects the primary circuit from the battery to limit the primary current of the ongoing charge of either transformer T1 or T2. With this the secondary current can be controlled to the adjusted level by the ECU - meaning the freewheeling state is acting like a step-down-converter. In Fig. 4 the freewheeling mode is always active when the primary current falls to zero, which occurs frequently for a secondary current set point of 50 mA. U B L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s I s U B L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s I s L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s I s L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s I s U B L 1 R p I p L 3 S 1 S 2 S 3 S 4 S 5 L 2 L 4 D 1 D 2 R s I s 1. 2. 3. 3. 4. 4. Fig. 3: Different states of the CMC-system: 1. Initial Charge, 2. Initial Breakdown, 3. Extend ignition burning duration, 4. Freewheeling. The different currents in each state are highlighted. Bold-faced red: Primary current, Bold-faced green: Secondary current 84 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="85"?> During the described standard operation, the CMC-system is monitoring the burning voltage at the spark plug via the voltage at the high-voltage diodes D 1 / D 2 . If the voltage across the diodes is too high, which occurs, if the spark is extinguished via high turbulences during the combustion process, the system starts to recharge both transformers and initiates a new ignition spark. time / ms -2 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 -20 -10 0 10 20 30 40 -1 0 1 2 3 4 5 3. 2. 1. t Dwell t Burn I s / A I p / A Fig. 4: Primary and secondary current data measured at U bat = 14 V and 1000 V Zener-Load Combustion data using the CMC or a similar high-power ignition system to operate a gasoline engine were published in [22, 23, 25]. 2.3 Measurement Setup The measurement setup is shown in Fig. 5 [26] and can be divided into the electrical and spectroscopic setup. The ignition system used is the power ignition system already presented in the previous chapter in combination with a commercial spark plug. Its resistance and electrode gap are 4.5 kΩ and 0.9 mm, respectively. The electrode material is a common iridium-rhodium (IrRh) alloy. 85 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="86"?> Battery DC 12 V High-voltage probe Charging Signal Oscilloscope iCCDcamera Spectrograph Spark plug Ignition coil PC Spectrograph Function generator CCD Gate Trigger Lens Trigger Oscilloscope PTG Pressure chamber Current clamp Fig. 5: Schematic measurement setup (based on [26-28]) 2.3.1 Electrical Measurement Setup For data acquisition. a LeCroy Waverunnter 6030A digital storage oscilloscope with adjustable input impedances was used and operated with a sampling frequency of 10 MHz. The high-voltage was acquired using the Tektronix P6015A high-voltage probe, which was high impedance coupled with the oscilloscope. Because of the high resolution needed to distinguish between the arc and glow discharge voltage levels, the oscilloscopes A/ D-converter has been protected from overdrive by an electrical circuit, that eliminates the high-voltage peaks of the breakdown. A Pearson 2877 current monitor in the high voltage path of ignition system was used to measure the currents in this path. The current clamp was connected to a 50 Ω input coupling of the scope, reducing the sensitivity of the clamp by half. 2.3.2 Spectroscopic Measurement Setup The spectrograph used was an Acton SP2556 grating spectrograph from Princeton Instruments with a focal length of 500 mm and an adjustable entrance slit. Moreover, it has three different gratings mounted on a turret. However, only the 600 and 1200 lines/ mm gratings are used in this study. A PI-MAX 2 ICCD camera from Princeton Instruments with a multi-channel plate (MCP) intensifier was used to record the spectra. The dimensions of the CCD are 1024x256 pixels in wavelength and axial direction of the ignition spark. For the spark 86 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="87"?> plug used, only 51 pixels in height are needed to completely cover the area between the electrodes. In this study, two settings of the entrance slit and the gratings are used. The high-resolution setting with the 1200 line/ mm grating and a slit width of 10 µm was used to obtain spectra of the metal lines of the arc discharge. This resolution has an instrumental broadening of approximately 0.1 nm, which is enough to distinguish most of the discrete metal lines. For the nitrogen bands of the glow discharge, a lower resolution with the 600 lines/ mm and a slit width of 200 µm was enough to obtain the spectra. 3 Results All results were obtained at room temperature and ambient pressure, respectively 10 bar absolute pressure in a pressure chamber. The higher pressure is used to represent more engine-like conditions at ignition timing. At all times, a slight airflow is guaranteed to flush the chamber due to ozone generation caused by the discharges and therefore to ensure consistent test conditions. To determine if the discharge respectively the observed part of the discharge is in arc or glow discharge mode the metallic lines are observed via spectroscopically measurements simultaneously to the electrical measurements [29]. 3.1 Standard Pulse Mode at Ambient Pressure and Usual Pressure For comparative purposes, the standard pulse mode is examined. Fig. 6 displays the persistent plot of 100 consecutive sparks at ambient pressure. The voltage curve shows a characteristic transition area in which the spark changes from arc discharge to glow discharge. As described in 2.1 the spark phases differ in their burning voltages because of the different efficiencies of their electron supply mechanisms. This means if the voltage increases abruptly, the spark phase changes from arc discharge to glow discharge and vice versa. At the given conditions this transition area occurs between 100 and 200 µs after breakdown, respectively at around 250 - 150 mA secondary current. The small voltage rise at the end of the glow discharge phase shows the point, at which the current becomes too low to obtain the discharge. The system tries to remain the current by raising the voltage but the energy of the inductances is discharged too far already and therefore the discharge ends. 87 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="88"?> Fig. 6: Secondary voltage and secondary current of the spark at 1 bar abs in SP mode Once the pressure is raised to 10 bar absolute, different voltage and current curves are observed. The transition area moves to around 300-400 µs after the breakdown, 100 - 50 mA secondary current respectively. In contrast to 1 bar absolute the glow discharge shows a quasi-stationary burning voltage before the characteristic voltage rise at the end of the discharge. Fig. 7: Secondary voltage and secondary current of the spark at 10 bar abs in SP mode These observations can be explained by the higher density of electrical charge carriers at higher pressures. Due to this higher density, the probability for thermionic-field 88 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="89"?> emission is increased by a higher intensity of ion-bombardment on the cathode [20, 29], resulting in increased cathode-spot heating. Therefore, arc discharges are characterized by higher current densities compared to glow discharges. In case of spark ignition at 10 bar absolute pressure, the current density of glow discharges is somewhat closer to that of the arc discharge. A higher pressure therefore results in a longer arc discharge, increasing the wear on the sparkplug. 3.2 Electrical Results CMC Mode For this test, the CMC mode was applied with 100 mA secondary current and a burning duration of 2 milliseconds. Fig. 8 presents the obtained voltage and current curves from 100 sparks. It is clearly visible how the system switches between the two coils, resulting in a small rise in burning voltage at every alternation. This voltage increase is the consequence of the increased ohmic loss across the sparkplug resistor due to the higher secondary current at the alternation point. The frequency of the alternation/ pulses depends on the secondary current level. If the current drops below a certain threshold, the ignition system switches to the other coil. At ambient pressure all pulses are found to be in the glow discharge phase. Before the first pulse, the secondary current is pulled from breakdown level to the target level and therefore an arc discharge in this phase is inevitable. Fig. 8: Secondary voltage and secondary current of the spark at 1 bar abs in CMC mode with 100 mA secondary current If the pressure is raised to 10 bar absolute, the discharge varies between arc discharge and glow discharge from pulse to pulse. The persistent plot shows an occurrence distribution of the discharge modes for all pulses. In contrast to 1 bar, there is a high 89 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="90"?> probability for the discharge to be in arc discharge mode, resulting in higher spark plug erosion. Fig. 8: Secondary voltage and secondary current of the spark at 10 bar abs in CMC mode with 100 mA secondary current In comparison to the standard pulse mode, the CMC mode constantly supplies energy and therefore the current to the discharge through the alternating of the coils. For this reason, the arc discharge can be maintained for as long as demanded. However, in order to minimize the wear of the spark plug and still ensure the ignition of the mixture, it is beneficial to reduce the secondary current, so that the current density is below that of the arc discharge. Although, the glow discharge itself has a low probability of inflammation, as showed in Fig. 1, it can feed the plasma core created by the breakdown and arc discharge and thus ensure safe ignition of diluted or lean mixtures. 3.3 Electrical Results with Limited Secondary Current For reducing the secondary current of the CMC mode to 40 mA, the 5 V trigger signal for the ignition system is changed according to the manufacturers specification. The voltage and current curves measured are shown in Fig. 9. At 1 bar absolute all the pulses are in glow discharge mode, as also seen with 100 mA secondary current. In this operating point the secondary current curve shows, that the system needs around 500 µs to regulate the current to the demanded 40 mA. 90 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="91"?> Fig. 9: Secondary voltage and secondary current of the spark at 1 bar abs in CMC mode with 40 mA secondary current When repeating the test at 10 bar absolute pressure, the discharge varies between arc discharge and glow discharge from pulse to pulse. In difference to the test with 100 mA, the probability for the discharge to be in arc discharge mode is overall significantly lower. At this operating point; especially the first pulses, in which the current is still being regulated downward, show a higher occurrence of the arc discharge mode. Fig. 10: Secondary voltage and secondary current of the spark at 10 bar abs in CMC mode with 40 mA secondary current 91 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="92"?> The reduction of the secondary current results in minimizing the probability of arcing. In order to make a comparison regarding erosion, the metal lines have to be examined. 3.4 Spectral Results to Correlate Erosion Effects To evaluate the metallic lines the spectroscopical measurements of 50 sparks are averaged and then processed. The observed wavelengths are those of iridium and rhodium. The result of the processing is a non-dimensional number of intensity, which is proportional to the absolute radiation of these lines and the standard deviation of the measurements. Fig. 11 shows the results for the different modes and phases in the modes at 10 bar absolute pressure. CMC40 has no measurement for the eighth pulse because this pulse was in glow discharge mode mostly (see Fig. 10) and therefore not measurable in this configuration. For SP mode the index 1 designates a phase of time after the arc discharge at ambient pressure. The first phase at CMC100 and CMC40 (index 0) has a high intensity of metallic lines because as mentioned previously in this phase the secondary current is still regulated downward and arc discharge is inevitable. Because this phase is unaffected by the current regulation to the target value the intensities are to be considered the same within the scope of measurement accuracy. When regarding the actual pulses of the ignition system differences between the setpoints for 100 and 40 mA appear. In CMC100 mode the first pulse has the lowest intensity of metallic lines, the following second and eighth pulse have higher intensity again. This correlates to the electrical measurements seen in Fig. 8. The first pulse has the lowest secondary current and therefore the least erosive arc. After the first pulse the intensity raises and stays almost the same as seen when comparing pulse 2 and pulse 8. This observation also matches the electrical measurements as the secondary current of those pulses is higher. The seemingly higher intensity in the eighth pulse may be due to the observation window within the pulse, which shifts slightly from pulse to pulse as the pulse lengths of the previous pulses vary somewhat. Therefore, it is possible that more intense parts of the pulse switching are recorded in individual recordings. In CMC40 mode this behavior changes since the current is regulated further downward. The first pulse has a slightly higher intensity compared to the first pulse of CMC100 mode. This is also reflected in the electrical measurements, as the first pulse in CMC100 has a lower secondary current than in CMC40. Thereafter a contrary trend is visible as pulse 2 shows a lower intensity than pulse 1. This coincides with the electrical measurement in Fig. 10, which shows a decreasing secondary current over time and even a significantly lower probability for the discharge to change into an arc discharge. The referenced SP mode also shows this strong correlation between spectroscopic measurement and electrical measurement. Fig. 7 displays a nearly linear curve for the secondary current over time which matches with decreasing intensity of metallic lines over time (phase 0 & 1). 92 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="93"?> (1) (2) (3) 0 1 2 8 0 1 2 0 1 CMC100 CMC40 SP 0E+0 2E+7 4E+7 6E+7 8E+7 Intensity / au Fig. 11: Metallic lines of the spectroscopical measurements at 10 bar absolute pressure during the discharges CMC an SP mode. The indices represent the pulses of the ignition system (0 is the phase before the first pulse) 4 Discussion As demonstrated by the experiments, there is a strong correlation between the intensity of the arc phases and the intensity of the emission signals from the metallic lines. The metallic lines result from vaporized and exited atomic emission signals after relaxing of the metallic atoms. Depending on external circumstances like flow, turbulence and pressure. As evaluated in [26, 30] the metal line emission intensity ε ji is proportional to the particle density n j and this particle density is proportional to the metallic erosion Err j caused by plasma effects of the electrodes. ε ji = ℎ * c o 4 * π * λ ji * A ji * n j with the boltzmann distribution equation (2) n j = g j U (T ) * n 0 * e − Ej kB * Texc The equations contain the planck constant ℎ, the speed of light in vacuum c o , the wavelength of the transition λ ji , the transition probability A ji , the occupation density of the particle n j , the statistic weight g j , state sum U (T ), the energy of the upper state E j , the Boltzmann constant k B and the excitation energy T exc . ε ji ∼ n j ∼ Err j Each ignitions spark has to pass all three spark phases (breakdown - arc phase - glow phase). The only possibility to combine a maximum of ignitability of a spark ignition 93 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="94"?> with a minimum of erosion is the combination of a high energy breakdown with a minimum arc phase. Therefore, the presented ignition system gives the possibility to limit the secondary current resulting in a minimum of arc phases or arc phases with minimum energy. The major parameter to switch between the arc phase and the glow phase (with only limited ignition probability - waiting for λ = 1 mixtures) as described by Richardson and Dushman [31, 32] is the current density in the plasma channel. The conditions of this switch depend on several combustion parameters like gas mixture, pressure and temperature. With no dedicated value to be adjusted in general, this current limitation stays to be an application parameter for each engine with such an ignition system. The presented results show, that the arc phase cannot be suppressed completely un‐ der all conditions but limited to an acceptable level. The intensity of the visible emission of the metallic lines is reduced significantly even if the electrical measurements show, that there are still arc phases but with lower energy and lower probability. 5 Conclusion The combination of high-speed electrical measurements with spatial and time-resolved spectroscopically measurements allow to evaluate the erosion potential of a dedicated ignition system (spark plug, connector and ignition coil) even in a pressure chamber lab environment. The demonstrated correlation between the based on electrical measurement identifiable arc phase probabilities and intensities and the measurable metallic emissions gives a possibility to cover an erosion risk in development phase. Potential further development could go into two directions: a. Integration of flow behavior and its influence on these physical processes by combining the presented experimental setup with a flow chamber setup b. Evaluation of a predictable erosion model based on in line measurements on an engine 1 References [1] S. Potteau, P. Lutz, S. Leroux, S. Moroz, and E. Tomas, “Cooled EGR for a Turbo SI Engine to Reduce Knocking and Fuel Consumption,” in SAE Technical Paper Series, 2007. [2] T. Alger and B. Mangold, “Dedicated EGR: A New Concept in High Efficiency Engines,” SAE Int. J. Engines, vol. 2, no. 1, pp. 620-631, 2009, doi: 10.4271/ 2009-01-0694. [3] M. Akif Ceviz, A. K. Sen, A. K. Küleri, and İ. Volkan Öner, “Engine performance, exhaust emissions, and cyclic variations in a lean-burn SI engine fueled by gasoline-hydrogen blends,” Applied Thermal Engineering, vol. 36, no. 7, pp. 314-324, 2012, doi: 10.1016/ j.applth‐ ermaleng.2011.10.039. [4] S. Szwaja, A. Jamrozik, and W. Tutak, “A two-stage combustion system for burning lean gasoline mixtures in a stationary spark ignited engine,” Applied Energy, vol. 105, 1-4, pp. 271-281, 2013, doi: 10.1016/ j.apenergy.2012.12.080. 94 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="95"?> [5] E. Toulson, H. J. Schock, and W. P. Attard, “A Review of Pre-Chamber Initiated Jet Ignition Combustion Systems,” in SAE Technical Paper Series, 2010. [6] P. Janas and W. Niessner, “Towards a Thermally Robust Automotive Pre-Chamber Spark Plug for Turbocharged Direct Injection Gasoline Engines,” in Ignition Systems for Gasoline Engines: 4 th International Conference, December 6-7, 2018, Berlin, Germany. Ed.: M. Günther, Berlin, 2018. [7] M. Zhou, G. Li, J. Liang, H. Ding, and Z. Zhang, “Effect of ignition energy on the uncertainty in the determination of laminar flame speed using outwardly propagating spherical flames,” Proceedings of the Combustion Institute, vol. 37, no. 2, pp. 1615-1622, 2019, doi: 10.1016/ j.proci.2018.07.084. [8] J. D. Dale, M. D. Checkel, and P. R. Smy, “Application of high energy ignition systems to engines,” Progress in Energy and Combustion Science, vol. 23, 5-6, pp. 379-398, 1997, doi: 10.1016/ S0360-1285(97)00011-7. [9] W. Chen et al., “Impact of Ignition Energy Phasing and Spark Gap on Combustion in a Homogenous Direct Injection Gasoline SI Engine Near the EGR Limit,” in SAE Technical Paper Series, 2013. [10] H. Albrecht, W. H. Bloss, W. Herden, R. Maly, B. Saggau, and E. Wagner, “New Aspects on Spark Ignition,” in SAE Technical Paper Series, 1977, pp. 1-11. [11] H. Albrecht, R. Maly, W. H. Bloss, W. Herden, B. Saggau, and E. Wagner, “Neue Ergebnisse über die Entflammung durch den elektrischen Funken,” 4th Satusseminar Kraftfahrzeug- und Straßenverkehrstechnik, 1977, 1977. [12] H. Albrecht, R. Maly, B. Saggau, E. Wagner, and W. Herden, “Neue Erkenntnisse über elektrische Zündfunken und ihre Eignung zur Entflammung brennbarer Gemische,” Automo‐ bild-Industrie, vol. 22, no. 4, pp. 45-50, 1977. [13] W. Herden, R. Maly, B. Saggau, and E. Wagner, “Neue Erkenntnisse über elektrische Zündfunken und ihre Eignung zur Entflammung brennbarer Gemische -2. Teil,” Automo‐ bild-Industrie, vol. 23, no. 2, pp. 15-22, 1978. [14] R. Maly et al., “Die drei Phasen einer elektrischen Zündung und ihre Auswirkungen auf die Entflammungseinleitung,” 5th Statusseminar Kraftfahrzeuge und Straßenverkehr, 1978, 1978. [15] R. Maly, “Spark Ignition: Its Physics and Effect on the Internal Combustion Engine,” in Fuel Economy: in Road Vehicles Powered by Spark Ignition Engines, New York: Springer US, 1984, pp. 91-148. [16] R. Maly and M. Vogel, “Initiation and propagation of flame fronts in lean CH4-air mixtures by the three modes of the ignition spark,” Symposium (International) on Combustion, vol. 17, no. 1, pp. 821-831, 1979, doi: 10.1016/ S0082-0784(79)80079-X. [17] R. R. Maly, “Die Zukunft der Funkenzündung,” MTZ - Motortechnische Zeitschrift, vol. 59, no. 7, pp. XIX-XXIV, 1998, doi: 10.1007/ BF03226465. [18] W. Ramberg, “Über den Mechanismus des elektrischen Lichtbogens,” Ann. Phys., vol. 404, no. 3, pp. 319-352, 1932, doi: 10.1002/ andp.19324040305. [19] B. Jüttner, “Cathode spots of electric arcs,” J. Phys. D: Appl. Phys., vol. 34, no. 17, R103-R123, 2001. [Online]. Available: http: / / stacks.iop.org/ 0022-3727/ 34/ i=17/ a=202 [20] J. Rager, Funkenerosion an Zündkerzenelektroden. Aachen: Shaker, 2006. 95 Application of a time-resolved ignition spark measurement technique when using a power ignition system <?page no="96"?> [21] A. Fridman and L. A. Kennedy, Plasma Physics and Engineering, Second Edition, 2 nd ed. Hoboken: CRC Press, 2011. [22] P. Weyand, F. Lorenz, S. Schilling, B. Geringer, J. Graf, and T. Hofherr, “Adaptive continuous spark ignition as enabler for high dilution EGR operation,” in Ignition Systems for Gasoline Engines, Berlin, 2014. [23] F. Lorenz, P. Weyand, E. Jacque, and S. Schilling, “Advanced Ignition Systems: Technical Possibilities and Limitations,” in The Spark Ignition Engine of the Future, Strasbourg, 2013. [24] J. Graf and T. Lauer, “Innovative Zündung,” Forschungsvereinigung Verbrennungskraftma‐ schinen (FVV), 1008, 2011. [25] M. Schenk, F. X. Schauer, C. Sauer, G. Weber, J. Hahn, and C. Schwarz, “Challenges to the Ignition System of Future Gasoline Engines - An Application Oriented Systems Comparison,” in Ignition Systems for Gasoline Engines, Berlin, 2016, pp. 3-25. [26] T. Michler, O. Toedter, and T. Koch, “Influence of the Pressure on Spatial and Temporal Resolved Plasma Physical Parame-ters of a TCI-Ignition System,” in 5 th International Confer‐ ence on Ignition Systems for Gasoline Engines, Berlin, 2021. [27] T. Michler, O. Toedter, and T. Koch, “Measurement of temporal and spatial resolved rotational temperature in ignition sparks at atmospheric pressure,” Automot. Engine Technol., vol. 5, 1-2, pp. 57-70, 2020, doi: 10.1007/ s41104-020-00059-w. [28] T. Michler, O. Toedter, and T. Koch, “Spatial and time resolved determination of the vibrational temperature in ignition sparks by variation of the dwell time,” SN Applied Sciences, vol. 2, no. 7, p. 1311, 2020, doi: 10.1007/ s42452-020-3104-6. [29] T. Michler, W. Kim, O. Toedter, T. Koch, and C. Bae, “Influence of the Electrical Parameters of the Ignition System on the Phases of Spark Ignition,” in Ignition Systems for Gasoline Engines: 4 th International Conference, December 6-7, 2018, Berlin, Germany. Ed.: M. Günther, Berlin, 2018, pp. 222-238. [30] S. Baus, Funkenerosionsmodell von Nickelbasislegierungen. Aachen: Shaker, 2011. [31] O. W. Richardson, “On the Negative Radiation from Hot Platinum,” in Mathematical Proceedings of the Cambridge Philosophical Society, Cambridge: UK: Cambridge University Press, 1901, pp. 286-295. [32] S. Dushman, “Electron Emission from Metals as a Function of Temperature,” Phys. Rev., vol. 21, no. 6, pp. 623-636, 1923, doi: 10.1103/ PhysRev.21.623. 96 Moritz Grüninger, Tobias Michler, Frank Lorenz (BorgWarner), Olaf Toedter, Thomas Koch <?page no="97"?> Spark erosion tests on materials for spark plug electrodes Dr.-Dipl.-Ing. Stephan Herbst Heraeus Deutschland GmbH & Co. KG Dipl.-Ing. Patrick Baake Heraeus Deutschland GmbH & Co. KG Dr.-Ing. Thomas Emmrich IAV GmbH Abstract: The technical challenges of combustion engines these days lie in the reduction of emission values, as well as resource conservation and the search for affordable, functioning materials and manufacturing technologies. In the case of spark plugs, it is the electrodes based on precious metals (e.g., palladium [Pd], platinum [Pt], iridium [Ir], rhodium [Rh] and ruthenium [Ru]) which, due to the volatile and rapidly increasing precious metal prices, make an economic calculation of the manufacturing costs and the pricing of the spark plug considerably more difficult. Since these materials consist primarily of costly platinum and iridium alloys, cheaper alternatives are more in demand than ever. A procedure for the development, evaluation and preselection of materials for spark plug electrodes is presented in the article. This considers the wear caused by spark erosion, which has a decisive influence on the wear on spark plug electrodes. Work results on alloy materials from endurance runs on the IAV ignition test bench are presented. The knowledge gained from the experiment shows, under the selected test conditions, an unexpected result that the conventional IrRh-based alloys have less resistance to erosion than the alloys with ruthenium. In order to confirm this statement, engine tests under more realistic application conditions are essential. The results from the engine run could deviate from the erosion test, since the changed test boundary conditions (fuel-air mixture, temperature) have corrosive influences and thus also influence the wear behaviour of the spark plug electrode. 1 Introduction Today, in addition to technical challenges such as engine downsizing with higher charge densities as well as higher pressures and temperatures in the cylinder, ecological and economic requirements such as lower emission values, fuel consumption and a lower cost structure in production also form the framework conditions to produce spark plugs at competitive prices. <?page no="98"?> The test procedures available to put a spark plug through its paces are also time-consuming and cost intensive. Both the stationary engine test and the field test in a car consume a lot of fuel and emit a high level of emissions over the test period. These expensive tests are not the most appropriate means of testing new materials. If the materials to be tested are still unknown for the application and it cannot even be estimated yet whether they can fulfil the required properties, the costs of these tests are out of proportion to the yield. The precious metals used as electrode materials have the most price-intensive impact on spark plug production, even if they are the smallest components. Over the years, platinum and iridium based alloys have proven themselves as electrode materials. Their advantage of very good oxidation and erosion resistance, as well as thermal and electrical conductivity, is offset by their disadvantage of high and volatile material prices. Due to the longer service life of the spark plug and the associated lower maintenance costs for the engine compared to spark plugs with nickel-based electrodes, engine manufacturers prefer to use spark plugs with expensive precious metal alloys. This is the motivation of this paper to present a cost-effective test method for the pre-selection of materials for spark plug electrodes and to use this method to compare standard alloys with less costly precious metal-based materials. The prices for rhodium and iridium have strongly increased in the last 2 years, as shown in Fig. 1. This was influenced by a higher demand from automotive as well as further new technologies and applications, e.g., fuel cells. The resulting shortage on the market drives up the price significantly and reliable cost and price planning cannot be guaranteed. The precious metal price for ruthenium is on a significantly lower level than for rhodium and iridium. Ruthenium also has the advantage of having similar material properties to iridium, but much lower density (see Table 1). Ruthenium Iridium Rhodium Platinum density [g/ m³] 12,4 22,6 12,5 21,5 melting point [°C] 2334 2466 1964 1772 electrical conductivity [m/ Ω*mm²] 13,7 19,7 21,1 9,7 thermal conductivity [W/ m*K] 117 147 150 72 Tabelle 1: Overview of material properties [3] 98 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="99"?> Fig. 1: Overview of precious metal price history from 2017 to 2021 [1] The significantly lower density is associated with a lower weight with the same volume compared to iridium. This point is included as an advantage in the cost calculation of the spark plug. The properties shown in Table 1 in connection with a low precious metal price were the basis for the decision to test ruthenium materials against standard alloys made of iridium-rhodium. 2 Methodology 2.1 General boundary conditions In the combustion chamber of the internal combustion engine, the spark plug is exposed to pressure, temperature and charge movement influences. Because the ignition spark occurs before combustion, the boundary conditions for spark erosion are only considered until then. Of course, the spark plug is exposed to the effects of the combustion of the working gas. These are regularly separated from one another in time and are, therefore, irrelevant for the ignition spark and the resulting spark erosion. The ignition point on the engine varies depending on the load point and speed, which is why the influencing variables pressure, temperature and charge movement can have different relationships to one another. The type of engine and the fuel used also play a role. In modern, supercharged car engines, maximum pressures of 40 … 50 bar at the point of ignition are reached at temperatures of 400 … 450 °C. The maximum flow speeds 99 Spark erosion tests on materials for spark plug electrodes <?page no="100"?> between the electrodes of the spark plug are 10 … 30 m/ s. Gas engines for stationary applications are usually derived from diesel engines. Pressures of 60 … 80 bar at up to 600 °C are achieved here at the point of ignition. 2.2 Description of the IAV ignition test bench The IAV ignition test bench combines the properties of a pressure chamber with the properties of a flow tube in one test bench. The structure is schematically as follows: Fig. 2: Schematic structure of the IAV ignition test bench The test bench is used to examine ignition systems or components under conditions close to the engine. It consists of a test chamber which has 4 orthogonal entrances. Depending on the test requirements, these can be provided with observation accesses (quartz glass) or sample holders (e.g., for spark plugs). These are closed with blind lids, if not in use. This test chamber is connected to a circulation via a closed pipe system, which is used exclusively to generate the movement of the charge. By adjusting the speed of the fan, the flow rate through the chamber can be influenced in a defined manner between 0 and 30 m/ s. In order to simulate the conditions at the point of ignition in the engine, the interior of the entire system can be pressurized with up to 40 bar via a gas cylinder. All non-explosive gases and gas mixtures are suitable as gases. Usually nitrogen is used because of its inert behavior and its high proportion in air. It is possible to influence the gas temperature in the range from 30 °C to approx. 150 °C using heating jackets along the pipeline system. 100 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="101"?> All common discharge ignition systems can be controlled by means of the open engine control unit FI2re (IAV). A high-resolution photo camera for recording integral spark images or a high-speed camera for high-resolution recordings is available for optical examinations. Cur‐ rent/ voltage curves of the primary and secondary circuit of the ignition system can be recorded with up to 100 MS/ s. In Fig. 3 the ignition test bench is shown in the configuration with a high-resolution photo camera. Fig. 3: Setup of the ignition test bench for investigations of the spark deflection 2.3 Sample holder Heraeus manufactures precious metal alloys as semi-finished products for spark plug production. In order to avoid the manufacture of special spark plugs, special sample holders were developed for examining material samples. These material samples with a diameter of 0.7 mm (common pin diameter on spark plugs for car applications) are clamped in brass clamps and mounted in the sample holder. Due to the clamped fastening of the samples, it is possible to adjust the distance between the electrodes by moving them axially. At the current stage, these sample holders are being produced using a 3D printing process. In this way, customer requirements regarding the electrode arrangement can be fulfilled quickly and inexpensively. The plastic used here also serves as an insulator to resist the breakdown voltages that occur. Concerns about mechanical strength could be dispelled in the pressure and temperature range used. These sample holders are then inserted, fastened and contacted into the entrances to the test chamber. A completed sample holder is shown in Fig. 4. 101 Spark erosion tests on materials for spark plug electrodes <?page no="102"?> Fig. 4: complete sample holder 2.4 Test boundary conditions For the present endurance run, average values should be selected that are as represen‐ tative as possible for the number of ignition events and for the boundary conditions of gas density and flow velocity. The following assumptions were made for this purpose: • Assumed vehicle mileage: 25,000 km • Average speed: 50 km/ h • Average engine speed: 2600 min -1 With these assumptions, there are 39 million ignition events. These events are counted using a pulse counter on the primary side of the ignition coil. The chamber pressure was set at 10 bar at 30 °C. The gas density, which is assumed to be constant for comparable ignition conditions, was used as the reference variable for the transformation of the test bench conditions to engine conditions. This results in typical boundary conditions at the time of the ignition event of a gasoline engine. In modern gasoline engines, the level of charge movement due to the tumble flow is 10 … 30 m/ s. Therefore, the flow velocity of the gas mass in the test chamber was 102 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="103"?> fixed at 20 m/ s. This parameter has a very strong influence on the wear behaviour of the electrodes, because the movement of the charge after the first breakdown spark leads to spark deflection or spark extinction with subsequent spark breakdowns until the discharge of the ignition coil. In order to get close to the conditions in the real engine, air was used as the working gas in the present application. 2.5 Experimental setup The test bench was provided with a metal housing for the endurance tests. This contributes to occupational health and safety, but also to electromagnetic compatibility (EMC). In principle, 4 material pairings can be tested at the same time. The solution shown in Fig. 5 with 3 sample holders = 3 material pairings and a visual access allows a visual check of the function and, thus, offers an additional monitoring option. Fig. 5: Test bench setup with 3 sample holders and optical access The activation of the active ignition coils takes place in a manner similar to that of a real internal combustion engine, but here with a view to better distributing the charging current of the ignition coils over a work cycle. The voltage probe, which is shown in Fig. 5: Test bench setup with 3 sample holders and optical access, is only required for measuring the breakdown-voltage and is removed for the actual endurance run. The ignition energy is provided by series ignition coils from VW with approx. 90 mJ each. To limit the ignition current (EMC), a spark plug connector with 5kΩ impedance and high-voltage ignition cables from Beru with 1kΩ are used. 103 Spark erosion tests on materials for spark plug electrodes <?page no="104"?> Fig. 6: Adapted ignition module with ignition cable, spark plug connector and fan cooling 2.6 Carrying out the experiment Before the material samples are installed in the brass clamp, the initial mass is determined on a precision balance. Subsequently, a photographic recording is made by means of a reflected light microscope of the sample face (facing the spark) and the two adjacent jacket sides for each sample holder with the material sample (see Fig. 7 left illustration). The installation in the sample holder takes place. An electrode gap of 0.7 mm is set by axially shifting the material samples in the brass clamp. The experiment starts with a measurement of the breakdown voltage for each sample holder = one pair of electrodes. For this purpose, the voltage in the secondary circuit is measured using a probe from PinTEC and recorded with 25 MS/ s (see Fig. 5 for setup). After predetermined intervals, the endurance test is interrupted for interim analyses: Secondary voltage measurements and photographic recordings are carried out as described above for better documentation of the course of wear (see Fig. 7 middle illustrations). In order to prevent inadmissibly high electrode wear = impermissible enlargement of the electrode gap, the electrode spacing resulting from the continuous run is determined with a feeler gauge with an accuracy of ± 0.05 mm and adjusted again to 0.7 mm, if necessary. At the end of the endurance run, the secondary voltage measurement, the photo‐ graphic recordings and the determination of the electrode gap are followed by a final mass determination of the material samples. 104 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="105"?> Fig. 7: Documentation of the visual wear of the top and side view 2.7 Evaluation of the results and outlook The reactions of the material surfaces to the spark erosion are shown by means of continuous logging and documentation (see Fig. 7). From this, the suitability of certain alloys can be deduced, because the spark introduces heat and energy into the material, which leaves traces of the individual alloy. The removed volume can be deduced from the difference in mass via the density of the alloy. This parameter gives better information about the state of wear than the measurement of the electrode distance. In relation to the number of ignition events, wear indicators (see Fig. 8) can be derived from this, which are decisive for the alloy and allow a comparison of different alloy materials and their suitability as electrode material. If the second electrode partner is included in the analysis, the measured wear between the electrodes can also be taken into account. The influence of the charge movement results in an asymmetrical wear behavior on the electrodes (see Fig. 7), which is the cause of the discrepancy between the measured and calculated change in the electrode spacing (Fig. 9). The change in the measured electrode gap is smaller than the volume erosion suggests. This results in a smaller increase in the breakdown voltage. This effect has a positive influence on the function of the ignition system: The electrode gap is one of the main factors influencing the breakdown voltage [2]. For the transferability of the results obtained to a real distance covered with the vehicle or engine running time, comparative measurements on the engine test bench are useful. A recorder measurement, e.g. from an RDE trip, can be used as the basis for this. This covers representative inner-city operation, cross-country and motorway journeys. 105 Spark erosion tests on materials for spark plug electrodes <?page no="106"?> Fig. 8: Determination of the wear indicators Fig. 9: Comparison of measured and recalculated wear values 106 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="107"?> 3 Evaluation of the examined alloy with regard to its suitability as an electrode material 3.1 Material selection Platinum and iridium alloys are the common spark plug alloys. Due to the price difference, a decision was made to concentrate on a replacement of iridium alloys. Various contents of iridium [Ir], ruthenium [Ru] and rhodium [Rh] were chosen in the tested replacement alloys. Two comparison materials were chosen. This was pure iridium and an iridium-rhodium alloy usually used in spark plugs (Table 2). Table 2: Listing of tested alloys The ruthenium content was increased, and the iridium content decreased. However, to control the oxidation behaviour small amounts of rhodium where added. 3.2 Results spark erosion test The spark erosion test of the alloys was made with alike materials on each electrode (e.g., pure Ir against pure Ir). Therefore, a test result of each alloy is obtained for the ground and middle electrode. The volume loss for the full test length (39 mio sparks) is the defined value to characterise the developed alloys. The results are shown in Fig. 10 for each electrode. 107 Spark erosion tests on materials for spark plug electrodes <?page no="108"?> Fig. 10: results spark erosion test ruthenium alloys In Fig. 10 can be seen that the volume loss is always higher at the ground electrode than at the middle electrode. The common IrRh alloy had the highest wear at both, the ground and middle electrode. Around 0.06 mm³ wear were seen for pure iridium, the IrRu alloy and the RuIrRh alloy 1 at the ground electrode. The wear at the ground electrode reduces for the RuIr alloy, RuRh alloy and RuIrRh alloy 2, respectively. At the centre electrode, the pure iridium has the second-best result. The spark erosion getting better in the developed alloys starting with RuIr followed by IrRu, RuIrRh alloy 1, RuRh and RuIrRh alloy 2, respectively. It shows that the spark erosion at the middle electrode is the lowest with pure Ir and RuIrRh alloy 2, respectively. This result raises the two questions on why the Rh content has a minor influence on the spark erosion and whether the ground electrode has less wear than the centre electrode. In the spark erosion tests conducted, an electrode erosion was assumed to take place in the cold moment before fuel ignition. Additionally, an oxidation test was conducted after the results were being obtained to see if the engine heat and air atmosphere has a further influence on volume loss. 3.3 Oxidation test description An oxidation test was conducted to see whether heat and air atmosphere has an additional influence on the volume loss of precious metals, or not. For the test 100 mm lengths of 0.7 mm diameter wire were selected of each alloy. The wires were measured in weight, length and diameter to record the starting conditions. The furnace was able to circulate with the ambient oxygen. However, the air was not actively moved inside 108 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="109"?> the furnace. The materials were heated up to 1100 °C temperature and hold for 30 hours. The heating ramp and cooling ramp not considered in the loss calculation. The weight, length and diameter measurement after the heat treatment shall show a difference to the beginning. Oxidation will first increase all the three values. However, if there is prolonging heat Ir and Ru will generate volatile oxides which evaporate. The weight and volume will shrink. 3.4 Results of oxidation and spark erosion compared A volume loss of the oxidation test and of the spark erosion test is not comparable. Therefore, all values were normalised to the value of pure iridium. Pure iridium was chosen to be value 1. Fig. 11 shows the results for each electrode side. Fig. 11: spark erosion to oxidation loss (pure Ir = 1; oxidation at 1100 °C / 30 h; spark erosion after 39 mio sparks) In Fig. 11 it can be seen that the alloys with a high iridium content show a low loss of oxidation. The alloys with high ruthenium content have high oxidation loss. The addition of rhodium did not change this behaviour. However, the RuIrRh alloy 2 and the IrRu alloy are a potential candidate for further tests on the ground electrode due to the low spark erosion. The RuIrRh alloy 2 has the potential for further centre electrode testing. Currently, all alloys with high Ru content are cheaper than the standard IrRh alloys on the market. Although, this can change any time being. Therefore, the RuRh alloy might also be interesting for further applications, like engines running at lower temperature. The investigations show optimistic results with regard to the substitution of iri‐ dium-based alloys for ruthenium-based alloys, which currently offer a cost advantage. On the basis of the investigations presented, cost-intensive engine investigations can be better justified. The engine test must then provide practical evidence that ruthe‐ nium-based alloys exhibit wear behavior that is comparable to that of iridium-based alloys. 109 Spark erosion tests on materials for spark plug electrodes <?page no="110"?> 4 Summary A short summary in bullet points is given below: • Spark plugs with precious metal electrodes are in greater demand due to their better wear properties. • Constantly rising precious metal prices are causing strong cost pressure in spark plug production, which is driving the search for cheaper solutions for the substitution of rhodium and iridium. However, the prices are volatile. • Practical tests are necessary to evaluate new alloys regarding their suitability as electrode material. Engine tests are expensive and time consuming. The erosion test presented on the IAV ignition test bench allows the alloys to be assessed and pre-selected under comparable conditions close to the engine. This takes place under time and economically favorable conditions. • The evaluation of the wear is carried out on the basis of the volume loss. • The comparison of the wear levels from the erosion test and the real engine test with regard to the engine running time / distance covered is still pending. • The chosen ruthenium alloys show a lower spark erosion then the standard iridium alloys. • Oxidation testing at elevated temperature reveals a lower wear with iridium alloys than ruthenium alloys. • A mixture of ruthenium, iridium and rhodium indicates the preferred alloy for both the ground and centre electrode. 5 Bibliography [ 1 ] https: / / www.heraeus.com/ de/ hpm/ pm_prices/ prices/ prices.html . (5 th of october 2021) [ 2 ] Paschen, F.: Über die zum Funkenübergang in Luft, Wasserstoff und Kohlensäure bei verschiedenen Drücken erforderliche Potentialdifferenz. Dissertation Universität Straßburg (1889) [ 3 ] Periodensystem der Elemente: https: / / www.chemie.de/ lexikon/ Periodensystem.html 110 Stephan Herbst, Patrick Baake, Thomas Emmrich <?page no="111"?> Advanced Ignition Strategies for Gasoline Engine Clean Combustion Ming Zheng 1 , Guangyun Chen 2 , Jimi Tjong 1 , Liguang Li 3 , Xiao Yu 1 , Linyan Wang 1 1 2 3 Department of Mechanical Automotive & Materials Engineering, University of Windsor, ON, Canada Zhuzhou Torch Spark Plug Co., LTD, Hunan, China. School of Automotive Engineering, Tongji University, Shanghai, China Abstract: Future transportation requires a variety of power sources, using sustainable and renewable energy sources. Spark ignition engines remain to be one of the most important power units for automotive applications, especially for passenger vehicles. Further improvement in engine efficiency is crucial to meet the mandatory CO 2 emissions regulations in the future. A charge dilution strategy incorporated with the strong in-cylinder flow can improve engine efficiency and emission performances. However, stable flame kernel initiation receives severe challenges under such conditions. In this paper, constant volume combustion chambers mimicking engine tumble flow were employed to investigate the impact of discharge parameters on flame kernel initiation. Transient current surge strategy can deliver up to 18 J of spark energy within 20 μs, and shows best ignition capability under quiescent conditions, because of the intensive thermal expansion caused by the high temperature plasma, however, the ignition capability is significantly mitigated under strong flow conditions, because of the short discharge duration. On the other hand, ignition capability of glow phase boost strategies relies on the flow, which can stretch the plasma and enlarge the ignition volume. Both discharge current amplitude and discharge duration are critical to generate a self-sustained flame kernel rapidly under extreme lean conditions. The multi-core ignition concept with discharge energy distribution strategy was further studied in detail on both single cylinder research engine and 4-cylinder di‐ rect-injection turbocharged gasoline production engine. Single cylinder research engine results indicated that multi-core ignition strategies are more effective to convert spark energy into a bigger initial flame kernel volume, providing the spark energy for each spark gap is sufficient to generate a stable flame kernel. A long <?page no="112"?> discharge duration of 7 ms, realized by a dual-coil ignition system, outperforms multi-core strategy with traditional discharge characteristics under extreme early spark timing, where each flame kernel initiation demands much more ignition energy and longer duration. On the production engine, ignition performances of the multi-core ignition system were tested within a wide range of engine loads, and results show that the multi-core ignition system has a major advantage under harsh operating conditions such as engine idling and lean burn conditions. 1 Introduction Presently, internal combustion engines (ICEs) power over 97% of automotive vehicles in the world, as shown in Figure 1. New ICEs will continuously power the majority of automotive vehicles including hybrids for the foreseeable future. The greenhouse gas (GHG) emission challenge can only be met optimally with an appropriate technology mix, adapted to each respective application [1]. Electrical energy, hydrogen or E-fuels, renewable fuels, will respectively be able to meet different mobility and transport requirements optimally and CO 2 -neutrally [2]. The high energy density of hydrocarbon fuels makes them the most suitable fuel for heavy-duty applications, such as long-haul airplanes and heavy-duty freight trucks [3]. Renewable natural gas and renewable hy‐ drogen are considered to have zero carbon footprint, allowing zero-carbon powertrain using internal combustion engines. Fisher-Tropsch products can further increase the energy density of such kinds of fuel (E-fuels). Due to their high energy density, E-fuels as chemical energy carriers are not only preferred for vehicle use, but also for energy transport and storage. Furthermore, the location and time of the manufacturing process are independent of the use, which offers decisive advantages [2]. Figure 1: ICEV and BEV stock ratio. 112 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="113"?> To meet CO 2 emission regulations, the efficiency of gasoline engines needs to be further improved, together with the adoption of low-carbon renewable fuels. Worldwide re‐ search programs are targeting to raise the ICE brake thermal efficiency (BTE) to 55~60% for heavy-duty engines, and 40~45% for light-duty engines, while lowering the exhaust pollutants by 70~90% from current standards. Engine technology innovations and combustion control advances are still critical to further enhance engine efficiency and reduce tailpipe emissions. For gasoline engines, charge dilution strategies, including lean burn and exhaust gas recirculation, increase the fuel efficiency primarily due to (1) reduction in pumping loss at partial loads, (2) mitigation in combustion knock to allow better combustion phasing, (3) decrease in heat loss because of lower combustion temperature [4-6]. Furthermore, an adequately lean and/ or diluted cylinder charge potentially allows the use of a higher compression ratio for additional improvements in thermal efficiency. However, a diluted cylinder charge presents challenges to combustion stability, because of the prolonged ignition delay and reduced burning rate of the mixture. The combustion phasing needs to be controlled within a crank angle window, in order to achieve optimal thermal efficiency. However, the slow flame kernel growth and burn rate of a highly diluted cylinder charge make it challenging to achieve the optimal combustion phasing, as well as maintain the combustion stability. The degree of dilution is generally limited by late combustion phasing, severe cyclic variation, and declined mixture ignitability. Charge stratification and in-cylinder flow intensification have shown promise to secure ignition and accelerate flame propagation. The enhancement of the ignition source becomes important under such conditions because it can extend the tolerance of worsened mixture ignitability and thus allows higher levels of mixture dilution and flow intensity. The enhancement of the conventional single-spot ignition sparkplugs normally seeks a high discharge power [7-11] or a prolonged duration of spark glow [12,13]. Advanced ignition techniques such as non-equilibrium plasma discharge can achieve volume-type ignition through the discharge of transient plasma [14-16] or radio-frequency corona [17]. Recent research results from the author’s lab show that the performance of various types of ignition strategies is highly sensitive to background conditions, such as flow intensity, background density, and chemical reactivity of the combustible mixture. In this paper, the authors perform a comparative study of ignition strategies performance under quiescent/ flow conditions. Results show that transient high energy strategy shows better ignition performance under quiescent and low flow intensity conditions, while prolonged discharge duration and glow phase current boost strategies show strong ignition capability under intensified flow conditions. Especially under intensi‐ fied flow conditions, the discharge duration plays a critical role to sustain the initial flame kernel. The multi-core igniter system was further tested on multiple engine research platforms as a drop-in technology to improve the ignition and combustion under diluted conditions. Engine tests were carried out on multiple engine operation 113 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="114"?> conditions, from engine idling (0.7 bar BMEP) to high engine load (18 bar BMEP), with the main focus on lean-burn combustion under low to medium engine load. 2 Experimental Setups 2.1 Optical constant volume combustion vessel system A schematic diagram of the combustion chamber platform is illustrated in Figure 2. The constant volume optical chamber has a working volume of 0.2L with two quartz windows (Ø62mm viewport). High-speed shadowgraph imaging has been conducted to visualize the ignition flame kernel development and flame propagation processes under both quiescent and swirl flow conditions. The shadowgraph imaging setup includes two identical parabolic mirrors with a diameter of 6 inches and a focal length of 48 inches, a cold white LED light source, and a 0.4 mm pinhole. Images are recorded by a Phantom v7.3 digital high-speed camera. Figure 2: Ignition research platform with FPGA-synchronized electrical and optical measurements. The spark plug is installed on the top of the chamber. To generate a flow field mimicking the tumble flow in spark ignition (SI) engines, three gas channels are arranged at one side of the chamber, along the tangential direction, as shown in Figure 3. With this design, a global swirl flow field is generated inside the chamber. In this swirl flow field, the peripheral region of the chamber had the highest swirl ratio, while the turbulence intensity gradually decayed along the radial direction of the flow field. An Environics 4040 gas divider is employed to provide accurate access air 114 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="115"?> ratio control on the air-fuel mixture. The combustion chamber is fitted with dynamic pressure transducers for combustion pressure measurement. The combustion pressure is recorded by a data acquisition system which is externally triggered by the spark energizing command signal. All the control signals, including high-speed camera, oscilloscope, transistor-transistor logic (TTL) command are synchronized and adjusted by the multi-task field-programmable gate array (FPGA) system. Figure 3: Constant volume combustion chamber configuration and swirl flow field. In order to investigate the impact of discharge duration and discharge current ampli‐ tude independently, a dual-coil ignition system is connected in parallel with a discharge current management module. The discharge current management module alone can provide flexible adjustment over both discharge current amplitude and discharge duration. However, the low supply voltage (up to 2000 V) is not sufficient to reestablish the plasma channel if blown out by the strong cross flow. A dual coil ignition system is then introduced to provide a baseline discharge process, which has flexible control over discharge duration with a relatively low discharge current level. In the dual-coil ignition system, two identical inductive ignition coils are connected in parallel to a single spark plug. The internal resistance of the spark plug is 1.5kΩ and the spark gap size is 0.86 mm. The schematic setup of the ignition system with the current 115 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="116"?> management module is shown in Figure 4. The transformer and AC-to-DC converter (output 200~2000V DC) in the current management module are applied to charge a 3-µF capacitor. The ignition coils are primarily applied to establish and maintain the spark channel. Subsequently, the energy stored in the module capacitor is released to the spark channel to modulate the discharge current profile. The discharge current amplitude is controlled by the voltage across the capacitor as well as the resistance in the discharge circuit, while the discharge timing and duration are manipulated by a high voltage MOSFET, which is controlled by the TTL signal generated by the RT-FPGA system. During each test, the plasma channel is initiated by the dual-coil ignition system. Once the plasma channel is established across the spark gap, the spark current management module is activated to boost the discharge current to desired levels. Figure 3: Spark ignition system using multi-coil and current management module. The electric circuit demonstrated in Figure 5 is used to generate a high transient current spark. An inductive ignition coil system is used to generate a breakdown event, and is connected in parallel with a high voltage energy storage capacitor. The energy storage capacitor is pre-charged by an external power supply to store a large amount of electric energy. Because of the nature of capacitive discharge, the discharge process is fast, with a surge discharge current high enough to generate a micro explosion. To achieve appreciable thermal impact, the energy stored in the high voltage capacitor is usually much higher than the energy stored in the ignition coil. The energy storage capacitor used in this study varies from 3 μF up to 18 μF. The spark discharge process relies on the ignition coil to initiate the breakdown event initially. Once a plasma channel is formed, the energy storage in the capacitors starts to release to the spark gap. The spark discharge process of the transient high-current is finished within tens of microseconds. 116 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="117"?> Figure 5: Spark ignition system using high transient current strategy. 2.2 Single-cylinder research engine The ignition systems were evaluated on a single-cylinder research engine platform, which was operated in homogeneous charge spark ignition mode with port fuel injection strategy. The engine had been extensively modified and instrumented for spark ignition operation. The piston geometry was altered to form a shallow bowl combustion chamber on 80% of the piston surface with a compression ratio of 9.2: 1. A port fuel injector was integrated into the intake manifold. The research engine was coupled to a General Electric® 26G215 direct current (DC) dynamometer for load and speed control, and torque measurements. The engine was instrumented with numerous thermocouples and pressure transducers for detailed combustion research. The spark plug, used as an ignition source was mounted between the intake and exhaust valves on the flat profile cylinder head as the ignition source. The relatively centered location of the spark plug was preferred to reduce the flame propagating distance in the relatively large engine cylinder. In-house designed ignition systems were used for spark-ignition and discharge duration control. The specifications of the engine test platform are summarized in Table 1 and the schematic diagram is presented in Figure 6(a). The position and orientation of the spark plug, as well as the piston profile, are shown in Figure 6(b). Dry intake air was supplied to the engine from an oil-free air compressor which was equipped with drier and filter systems. The air pressure was controlled using an electronic pressure regulator. The mass air flow (MAF) rate was measured using a ROOTS 5M175 flow meter installed downstream of the pressure regulator. A throttle valve was installed upstream of the intake manifold. Air flow rate was adjusted using a combination of throttle valve position and upstream air pressure from the pressure regulator. To avoid any pressure fluctuations created by the displacement of engine valves, an intake surge tank was set up in between the air flow meter and the throttle valve. The engine coolant circulation and temperature control were realized using the FEV coolant conditioning unit and the coolant temperature was maintained at 80 °C 117 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="118"?> throughout the engine operation. Gasoline fuel was injected at a pressure of 4.5 bar using a port fuel injector at 10 after the beginning of intake stroke. The injection was commanded well before the compression stroke to maximize mixing between the fuel and the intake air. The fuel flow rate is measured using an Ono Sokki FP-213 piston-type flow meter. A Kistler piezoelectric pressure transducer was used to transmit the in-cylinder pressure readings at a resolution of 0.1 crank angle degrees, synchronized to a crank-mounted rotary encoder and camshaft sensor. A piezo-resistive absolute pressure transducer was employed to measure the intake manifold pressure for pegging the cylinder pressure. Spark discharge voltage and current are sampled using Tektronix P6015 high voltage probe and a Pearson 411 current probe respectively and the profiles are recorded using scope automotive oscilloscope. Engine parameter Value Displacement 0.84 L Bore 102 mm Stroke 105 mm Connecting rod length 165 mm Compression ratio 9.2: 1 Tabelle 1: Single-cylinder research engine specifications. (a) 118 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="119"?> (b) Figure 6. (a) Single-cylinder engine setup and (b) spark plug location. The fuel injection and ignition events were deterministically controlled using National Instruments RT-FPGA based engine control system. The injection and spark discharge timings are resolved on the crank angle basis, while the injection and discharge durations are controlled in the time domain. Pressure data for 200 consecutive engine cycles were recorded for each test point. The engine research platform was instrumented with sampling points for detailed emission measurements and analysis. The intake and exhaust gases were sampled simultaneously to feed to the emission analyzers, made by California Analytical Instruments (CAI). The emission analyzers included the flame ionization detector (FID) for HC measurement, nondispersive infrared (NDIR) sensor for CO and CO 2 analysis, paramagnetic oxygen sensor for intake and exhaust oxygen measurement, and heated chemiluminescence detector (HCLD) for NOx measurements. The inlet (O 2 and CO 2 ), as well as the exhaust (NOx, HC, CO, CO 2 , and O 2 ) concentrations, were all measured through CAI emission analyzers during the engine operation. 2.3 Multiple-cylinder production engine An inline 4-cylinder production engine coupled to an AC motoring dynamometer was used for the experiment. The 4-stroke, 2.3-liter engine system setup is shown in Figure 7, and specifications are given in Table 2. The in-cylinder pressure was measured using a Kistler pressure transducer instrumented inside the cylinder head. The pressure measurement was paired with an AVL encoder to acquire the in-cylinder pressure at a resolution of 0.2 crank angle degrees. Each cylinder was instrumented with a gasoline direct injection injector, wherein E10 gasoline fuel was supplied. Both injection duration and injection pressure were adjusted by the powertrain control module (PCM) under various engine operating conditions. 119 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="120"?> Engine Type Inline 4-cylinder GDI Displacement 2.3 L Bore [mm] 87.55 mm Stroke [mm] 94 mm Compression Ratio 9.5: 1 Tabelle 2: Multi-cylinder engine specifications. Figure 7: Schematic of the multi-cylinder engine setup. The PCM allowed for independent control over the engine parameters, including, throttle position, fuel injection pattern, boosting, and spark timing. The exhaust emissions were analyzed using a Horiba MEXA 7100 emission analyzer system. A wide-band exhaust oxygen sensor was paired with a closed-loop feedback control system to maintain the excess air ratio according to test demand. The engine oil and coolant were controlled by an external heat exchanger to maintain a temperature of 90.5 °C. The lean burn strategy was realized by opening up the throttle body to allow more air into the cylinder. Numerous operating conditions were extensively tested and compared with standard plug-on-coil ignition systems. A range of engine loads and speeds were operated in single-cylinder and multi-cylinder operation, as summarized in Figure 8. 120 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="121"?> Figure 8: Summary of multi-cylinder engine tests with the 3-core ignition system in cylinder #1 or fitted to all four cylinders. 3 Results and Discussions 3.1 Impact of plasma stretch on spark energy distribution The impacts of plasma stretch on the spark discharge energy distribution are investi‐ gated under controlled flow conditions with N 2 as the background gas. The discharge voltage and current are measured simultaneously with high-speed imaging of the spark plasma. The current boost strategy is adopted to study the impacts of discharge current (60 mA~2.8 A) on plasma stretch under flow conditions. Figure 9 compares the discharge current and voltage between quiescent and flow conditions. Under flow conditions, an increase in discharge voltage is observed compared with quiescent conditions, resulting in a drop in discharge current amplitude. Figure 9: Effects of air flow on the spark discharge energy. (current level 300 mA, spark gap size 1.3 mm) 121 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="122"?> The calculated discharge energies are also demonstrated in Figure 9, and results show that the discharge energy is elevated from 58 mJ to 118 mJ with the same ignition strategy, when the flow velocity increases from 0 m/ s to 5 m/ s. The discharge energy with different flow velocities and discharge current levels are summarized in Figure 10. It is observed that the increase of flow speed can increase the spark energy, and this impact is amplified with the increase of discharge current. Under quiescent conditions, the impacts of discharge current on the spark energy are very limited. The discharge duration is fixed at 1.5 ms for all cases, the discharge energy under quiescent condition changes from 50 mJ to 75 mJ when discharge current increases from 120 mA to 2800 mA. Higher discharge current amplitude would trigger the plasma to transfer from glow phase to arc phase, which significantly decreases the plasma resistance, resulting in less energy distributed to the spark gap [18]. Under flow conditions, the spark energy can be increased by around 9 times compared with that of quiescent conditions when discharge current is above 1200 mA. Lower discharge level would limit the spark energy, because of the more frequent restrike events. The spark energy is likely to be enhanced by two factors. First, the increased discharge current amplitude can significantly decrease the tendency of plasma blow-off and restrike events [19]. Second is the longer plasma stretch leads to an increase in spark gap resistance, leading to higher energy distribution to the spark gap during the discharge process. This impact will be discussed further in detail. Figure 10: Discharge energy using different current levels under different cross-flow velocities. 122 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="123"?> Under flow conditions, the resistance of the plasma channel increases with extended plasma length. Consequently, the spark voltage and power increase, resulting in higher spark energy. Experimental results with 300 mA discharge current level are selected as an example to demonstrate the plasma stretch impact on plasma resistance and discharge voltage, as shown in Figure 11. It is observed that the plasma can be stretched to approximately 12 mm, under a flow velocity of 5 m/ s while the plasma resistance increases from 200 Ω to 3 kΩ. The 2-D plasma stretch length is calculated via direct imaging, and the results are compared with the resistance change during the discharge process, as shown in Figure 12. It is observed that the evolution of plasma length on the timeline shares a similar trend with that of the plasma resistance. The plasma resistance per unit length can then be calculated. The results show that the plasma resistance per unit length remains similar under both quiescent and flow conditions. The result indicates that the longer plasma channel contributes to higher spark energy, because of the increase in spark plasma resistance. During the discharge process, the energy is distributed to three main resistances in the secondary winding, including secondary coil resistance, spark plug resistance, and plasma channel resistance. The increase in plasma resistance can enhance the spark energy because of the higher discharge voltage, and this impact can be magnified when discharge current is further increased. Figure 11: Plasma patterns with correlated spark plasma resistance and voltage under flow conditions. (current level 300 mA, spark gap size 1.3 mm) 123 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="124"?> Figure 12: Comparison of plasma resistance and plasma length. The current management module is adopted to study the impacts of discharge current (60 mA~2.8 A) on plasma stretch under flow conditions. Further experimental work has been taken out to investigate the impacts of boundary conditions on plasma channel resistance, and preliminary results are shown in Figure 13. It is shown that discharge current is the main parameter governing the plasma resistance per unit length, while the plasma resistance per unit length is less sensitive to the flow. Similar trends are observed under both atmospheric and pressurized conditions. 124 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="125"?> Figure 13: Ratio between plasma resistance and plasma length. 3.2 Improvement of ignition flame kernel development A range of advanced ignition technologies has been developed in the clean combustion engine laboratory (CCEL) in the past decade [19]. Three strategies are chosen to investigate the impact of discharge current amplitude and discharge duration on the flame kernel formation process. Typical discharge waveforms are demonstrated in Figure 14. The energy of the transient high current strategy is modulated between 2~18 J of energy per spark event by adjusting the capacitor voltage as mentioned in the previous section. Figure 14: Discharge current demonstration of advanced ignition technology developments at CCEL. 125 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="126"?> A dual coil ignition system is first used to provide flexible discharge duration adjust‐ ment. The discharge current amplitude can be increased by enhanced primary voltage or prolonged charging duration. However, the maximum discharge current amplitude is generally less than 200 mA because of the hardware limitations. To acquire a higher spark glow current level, the boosted-current ignition system is deployed to adjust the discharge current up to 3 A. The enhancement of spark glow current is considered an effective way to secure the ignition kernel development in high-speed flow. Another technique named high transient current is to deploy transient high ignition energy using direct capacitor discharge. The peak transient current and capacitive discharge energy can reach up to 1700A and 18J, respectively, with a discharge duration of around 20 μs. 3.2.1 Improvement of ignition flame kernel development under quiescent conditions Figure 15 shows the discharge current profile and the total spark gap energy under quiescent conditions. The discharge current level ranges from 80mA to 2.5A. However, under quiescent conditions, the discharge energy didn’t increase monotonically with the increase of discharge current amplitude. The discharge energy first drops with the increase of the discharge current amplitude, then gradually increases to merely 143mJ. The major reason for this phenomenon is the significant drop in discharge voltage when the discharge current is boosted [18]. Figure 15: Discharge current, voltage and energy using boosted current ignition strategy. 126 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="127"?> Figure 16 and Figure 17 show the flame kernel propagation process under various discharge current amplitudes. Both flame kernel initiation and flame propagation are not sensitive to the increase in discharge current under quiescent conditions. The lowest discharge current case (80 mA) even has a slightly faster flame propagation speed because of the more wrinkled flame front generated by the less stable glow phase. Under quiescent conditions, the minimum ignition energy is relatively small, which makes the flame kernel initiation less challenging regardless of the spark energy level used in the present test. After the initiation of the flame kernel, the plasma channel is contained within the burned region, and the change in discharge current has a limited impact on the flame front. The relatively low spark power during the glow phase, compared with that of breakdown, also has limited impacts on thermal expansion of the flame kernel. Figure 16: Flame kernel development using boosted current strategy with different current levels. 127 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="128"?> Figure 17: Flame area calculation using high transient current with different discharge energies. Figure 18: Flame kernel development using high transient current with different discharge energies. Transient high current strategy was also tested under the same boundary condition to investigate the impact of spark energy on flame propagation process. As shown by the shadowgraph images in Figure 18, the brightness of the initial spark indicates the transient high energy delivered to the spark gap. The intensive thermal expansion 128 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="129"?> caused by the capacitive discharge generates high local turbulence in the vicinity of the spark gap, resulting in flame front wrinkles. The flame propagation becomes faster with the increased discharge energy, contributed by both enhanced thermal expansion and the increased wrinkled flame front area. Figure 19 shows the impact of the spark energy on the flame area, and it is observed that the flame propagation is more sensitive to the change in ignition energy when spark energy is beyond 5.7 J. No significant difference is observed between 0.7 J and 3.2 J cases, indicating the importance of initial thermal expansion on the later flame propagation process. Figure 20 compares the flame area of both discharge boost and transient high current strategy, and it is obvious that the initial thermal expansion is important for flame propagation acceleration under quiescent conditions. The results suggest that transient high energy/ power strategies are more preferable under quiescent or weak flow conditions, where the wrinkled flame front is mainly formed by the plasma channel. Figure 19: Flame area calculation using high transient current with different discharge energies. Figure 20: Flame area comparison using boosted current and high transient current strategies. 129 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="130"?> 3.2.2 Improvement of ignition flame kernel development in high-speed flow In this section, the ignition performance of transient high current and glow current boost strategies are investigated under flow conditions using methane/ air mixture at lambda 1.9 and 2 bar abs. background pressure. The average flow velocity is controlled at approximately 40 m/ s. Under such conditions, the ignition process becomes very challenging, and traditional ignition methods can not generate a self-sustained flame kernel. Under such conditions, various ignition strategies were tested to investigate their potential to secure initial flame kernel. Figure 21 shows the flame propagation using the high transient current strategy with different discharge energies, ranging from 2J to 18J. With a very short discharge duration, the flame detaches from the spark plug at approximately 1ms after the discharge event. Subsequently, flame propagates along with the swirl flow field. Figure 21: Flame kernel images for micro-explosion with different discharge energies. Misfire event is observed for the 2J ignition energy case. Because of the short discharge process (~20 μs), a much higher amount of ignition energy is needed to generate a self-sustained flame kernel. Higher ignition energy also leads to faster flame propagation, mainly because of the much bigger initial flame kernel. Four ignition strategies are used to demonstrate the impact of ignition energy and discharge duration on flame kernel initiation, as shown in Figure 22, including a low current discharge (90mA) but with extremely long discharge duration (~30ms); two boosted current discharge (~400mA) and (~2000mA) with the same shorter discharge duration (~3 ms); and a high transient current discharge with an high transient current 130 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="131"?> (~1000A) and high discharge energy (~7J). The flame kernel development of the four spark strategies is recorded by the highspeed shadowgraph imaging system and the images are shown in Figure 23. The high transient current and the boosted current with 2A are initiated with brighter sparks because of the stronger plasma channel. The transient high current discharge release high energy in a very short period form of a “micro explosion”. This high transient energy generates the largest flame kernel at the initial state within 40μs after the breakdown event. However, due to the very limited discharge duration, this initial advantage is not carried to the later phase of flame kernel development. As the flame further propagates, the boosted current of 2A with 3ms discharge duration is observed to perform the fastest flame propagation speed. For the long discharge duration with 90 mA discharge current amplitude, the flame kernel generated can only be secured when discharge duration is longer than 30 ms. The flame areas and pressure measurements of the four experiments are shown in Figure 24. The flame area comparison mainly concentrates on the flame kernel initiation process, while the in-chamber pressure provides overall ignition results. The case with 2000 mA discharge current and 3 ms discharge duration case demonstrates a strong and reliable flame kernel initiation process, leading to the fasted flame development and combustion process. The longest discharge duration case, on the other hand, shows the slowest compression process, because the flame kernels generated by the weak plasma channel also don’t sustain properly under these extreme conditions. The 400 mA discharge current case (spark energy 0.45 J) and the 7J transient high current case behaves similarly, indicating an ignition strategy with moderate discharge current and sufficient discharge duration is more cost-effective under flow conditions. Figure 22: Current waveforms for boosted current strategies with various discharge current levels and durations. 131 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="132"?> Figure 23: Comparison of flame kernel images for different ignition strategies. Figure 24: Comparison of pressure during combustion for different ignition strategies. 132 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="133"?> 3.3 Effect of advanced ignition strategies on lean/ diluted gasoline combustion - evaluation on single-cylinder research engine 3.3.1 Effect of multi-core on lean-burn control compared with dual coil offset strategy The efficacy of the ignition system depends extensively on the energy scheduling and background conditions. Two strategies of enhancing the ignition energy, 3-core ignition and dual coil offset (DCO) discharge, are investigated on a single-cylinder research engine at low load (IMEP = 3 bar) operation under lean conditions (λ=1.6). The ignition under low load is a challenge because of adverse background conditions (lower background temperature and pressure). The impact is amplified under lean conditions as the spark timing is required to be advanced into the compression stroke to maintain the optimum combustion phasing. As the spark timing is advanced into the compression stroke, the reduction in-cylinder pressure and temperature at the spark timing is further compounded. Consequently, the density of the background gas mixture in the vicinity of the spark gap is reduced. The spark plasma interacts with a lower number of fuel molecules. As a result, flame kernel development and early flame propagation are slowed. Higher ignition energy can aid in the formation and propagation of flame kernels under such adverse conditions. Although both systems use multiple ignition coils to enhance the ignition energy, the scheduling and spatial distribution of the energy are fundamentally different. DCO provides a temporal distribution of spark energy while the size of the resulting ignition volume remains similar. 3-core ignition on the other hand uses multiple independent spark channels to distribute the spark energy spatially. The interaction of the resulting flame kernels forms an overall larger ignition volume while the discharge duration remains relatively short and constrained. The discharge waveforms of both the ignition strategies are presented in Figure 25. Both the ignition strategies are compared in terms of combustion stability, phasing, and ignition delay during the spark timing sweep including early, lowest COV of IMEP for the sweep (MBT), and late spark timings. The load is benchmarked at stoichiometric conditions and thereafter, the throttle opening is increased to realize the desired excess air dilution conditions, while fuel delivery amount remains the same. The overall results of spark timing sweep using both ignition systems at an excess air dilution of λ=1.6 are presented in Figure 26(a). The pressure and AHRR traces of an example case at spark timing of 290 °CA with the two ignition strategies are shown in Figure 26(b). The impact of enhancing the ignition energy by either of the ignition strategies is evident from the combustion stability and phasing. Prolonging the discharge duration using DCO or a larger flame kernel of 3-core ignition stabilizes the combustion near the spark advance limit. Moreover, the spark advance limit is extended at a higher discharge duration of DCO discharge. However, this is not the case with 3-core ignition. The spark advance limit of 3-core ignition 290°CA, beyond this spark timing, the combustion becomes unstable indicated by the sharp rise in COV of IMEP. DCO ignition is highly unstable at the discharge duration of 3 ms. As the discharge duration is increased, the combustion becomes progressively stable, and the spark advance limit is extended to 275°CA at 7 ms 133 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="134"?> Figure 25: (a) Discharge current and voltage for DCO ignition at different discharge durations (b) discharge voltage for three cores of the 3-core ignition strategy. spark duration. Using DCO ignition, combustion phasing is only influenced near the spark advance limit. 3-core ignition has advanced combustion phasing throughout the spark timing sweep. Considering the ignition delay remains the same for both ignition strategies and all discharge durations, 3-core ignition accelerates the flame propagation as compared to DCO ignition. As the spark timing is retarded from the spark advance limit towards the MBT timing, the impact of DCO ignition starts to diminish and 3-core ignition proves to be relatively more effective near the MBT timings. Hence, under extremely lean conditions, the advantages of DCO ignition are majorly observable near the spark advance limit, where the ignition encounters maximum chal‐ lenge from the background conditions, whereas 3-core ignition is highly effective near MBT timings. DCO ignition tends to stabilize the combustion with minimal impact on the flame propagation and combustion phasing. A longer spark discharge can increase the probability of availability of spark (and hence, ignition) during the favorable background conditions, thereby reducing the variability in the ignition process. Therefore, DCO strategy doesn’t “speed up” or “displace” the ignition window resulting in minimal to no impact on the combustion phasing and ignition delay. A bigger ignition volume of 3-core ignition, on the other hand, can speed up the initial flame propagation in addition to the 134 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="135"?> stabilization of the ignition process, consequently advancing the combustion phasing. Moreover, the impact of larger ignition volume on the flame propagation can be noticed near MBT timings as well. However, as the spark timing is advanced towards the spark advance limit, a longer discharge duration starts to dominate the impact of a bigger ignition volume. Figure 26: Effect of DCO and 3-core ignition on lean burn engine operation at l = 1.6 3.3.2 Three-core performance analysis on Four-cylinder turbocharged production engine A multi-cylinder engine (refer to Section 2.1) was used to further investigate the efficacy of the 3-core ignition system. A wide range of engine loads from 0.7 bar BMEP (idling) to 18 bar BMEP were compared amongst ignition systems and tests from stoichiometric to lean conditions. A spark sweep was conducted to observe the spark advance for maximum brake torque (MBT) amongst all conditions. A summary of the stability of engine load at MBT conditions is shown in Figure 27. The data is shown for cylinder 1 in all cases. Generally, combustion at stoichiometric operation is readily achievable with rea‐ sonable stability as shown in Figure 27(a). Minimal differences are realized between 135 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="136"?> ignition systems except for low load operation. At engine loads of 2.62 bar BMEP and greater, the environment for combustion is ideal and combustion initiation is not a concern. Under these circumstances, engine performances are not sensitive to the change in ignition strategies. For baseline engine load from 0.7 bar to 5 bar cases, the combustion was tested to their lean limits. This limiting factor occurs as combustion stability and control become challenging, as shown in Figure 27(b). At 7 and 9 bar BMEP, the turbocharger cannot provide enough airflow to further lean out the mixture and therefore the highest lambda achieved was presented. Higher engine speeds are needed to further increase the excess air ratio under this load. However, doing so will change the boundary conditions such as in-cylinder flow patterns and background temperature. Nonetheless, it is clear a higher lambda can be maintained with lower cycle-to-cycle variations with increased engine load. Moreover, the 3-core ignition system provides improved stability that is increasingly apparent when ignition becomes unstable. Figure 27: Summary of engine loads with standard (1-core) and distributed (3-core) ignition systems operating at (a) stoichiometric mixture and (b) lean limits - the operating lambda conditions are labelled above the bars. 136 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="137"?> Low load engine operation is an ongoing challenge for spark ignition systems, even at stoichiometric conditions. Because of the low engine load, both background pressure and temperature are much lower than other engine operating conditions, such as idling conditions. Both factors provide a challenging environment for the ignition process and idling normally has high cycle to cycle variations. The main factor leading to combustion instabilities at low loads is the inability to secure a strong flame kernel at the spark timing. The minimum ignition energy (MIE) for a fixed spark gap distance typically decreases with the increase of the background density for air-fuel mixture. The plasma channel can energize more fuel molecules when the density around the spark gap is increased. With decreased background pressure comes lower background density and the less air-fuel mixture is energized by the plasma channel, resulting in less heat generated by the initial flame kernel. Under these circumstances, higher ignition energy is needed to sustain the initial flame kernel to develop and propagate. Therefore, at later spark timings the background pressure and temperature are greater and thus flame kernel initiation is more readily achieved, as shown in Figure 28. On the other hand, the increased flame initiation capabilities of the 3-core ignition system are best demonstrated, as both combustion stability and combustion phasing can be tightly controlled throughout the ignition timing sweep test. Figure 28: Idling spark timing and combustion phasing. Improving the ignition capabilities at low loads appreciably improves the cycle-to-cycle cylinder pressure steadiness, as shown in Figure 29. Achieving early combustion phasing improves the work output and effective efficiency through higher in-cylinder pressure and significantly fewer partial-burn cases. This test scenario is only to demonstrate the ignition capability under such extreme conditions. In practice, the strong ignition capability of such ignition systems would allow the engine to idle under the same engine load with less fuel consumption because of the higher thermal efficiency gained at favorable combustion phasing. 137 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="138"?> Figure 29: In-cylinder 300 cycle continuous pressure traces at best stability points under low load operation using (a) 3-core and (b) 1-core ignition system. The complete test summary of spark timing and lambda sweeps at BMEP of 3.2 bar and 5.5 bar is shown in Figure 30. For standard ignition systems (Fig 30(a)), operation up to lambda 1.3 and 1.4 realize reasonably engine stability at 3.2 and 5.5 bar IMEP, respectively. On the other hand, a minor impact is observed with the 3-core ignition system (Fig 30(b)), even up to lambda 1.5 and 1.6, respectively. With the 3-core ignition system, a distributed ignition energy can influence more fuel-air mixture to ignite and provide a larger initial flame kernel. Engine efficiency is strongly correlated to the combustion phasing (CA50) of the combustion cycle. As instabilities arise, controlling the combustion phasing effectively to reach optimum timings (365~370°CA) becomes increasingly difficult. The 3-core ignition system can span a complete range of combustion phasing nearing TDC even under heavy diluted conditions relative to SI combustion operation. While standard ignition can reach the earliest CA50 of 375°CA at lambda 1.5, the 3-core ignition can achieve combustion phasing near TDC all the while at low engine instability. A comparison of the cycle-by-cycle engine load at the lean limits is shown in Figure 31. Previous combustion timing analysis shows that a majority of instabilities arise in the initiation process from the start of ignition to 5% energy release (e.g. start of combustion) in the cycle, also known as the ignition delay period [20]. The length and consistency of the ignition delay are highly dependent on the flame kernel initiation process. Likewise, the interaction between the plasma channel and the unburned air-fuel mixture is the key factor that affects the start of combustion. Even with enhanced discharge current or prolonged discharge duration, the ignition capability of various ignition strategies will be compromised if the plasma channel has limited contact with the air-fuel mixture, such as quiescent conditions. 138 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="139"?> Figure 30: A summary of engine operation at (a) 3.2 bar IMEP (2.62 bar BMEP full-engine) and (b) 5.5 bar IMEP (5 bar BMEP full-engine) operating with standard 1-core and 3-core ignition. The results shown are with respect to cylinder 1 for all cases. 139 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="140"?> Figure 31: Continuous cycle to cycle IMEP (scatter plot) and its cyclic deviation from the mean (line plot) The 3-core ignition strategy is generating multiple flame kernels simultaneously. The three initial flame kernels can either develop independently or interact with each other to form a stronger initial flame kernel. Both possibilities can speed up the flame kernel development process, contributing to a shortened ignition delay. Especially under lean burn conditions, a shorter ignition delay provides a spark timing window closer to compression TDC, which is beneficial to the ignition process because of the higher in-cylinder pressure and temperature. As a result, a shorter and more consistent ignition delay is realized with the 3-core system, as shown in Figure 32. The figure presents the ignition delay of the current cycle concerning the previous to demonstrate a cycle-to-cycle comparison amongst systems. Stoichiometric operation is added for reference to observe the ideal operation and span of deviations. Figure 32: The cycle-by-cycle ignition delay period representation at lean limits under standard (1-core) and distributed (3-core) ignition systems 140 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="141"?> At the highest engine load-tested, similar combustion characteristics were observed, as shown in Figure 33. The concern for electric insulation capability of 3-core sparkplugs under high breakdown voltage was finally verified via production engine test. Both stoichiometric mixture and high in-cylinder pressure/ temperature further reduce the performance gap between traditional igntion system and 3-core igntion system. An advanced combustion phasing is still observed when using 3-core igntion system, indicating the system still functions under high load conditions. Lean burn operation under high load conditions with the help of e-boost strategies will be performed in the near future to extend the engine operational lean limit, in order to further improve engine efficiencies. Figure 33: High load operation under standard (1-core) and 3-core ignition systems Conclusions In this paper, the impact of flow on spark energy is first characterized, revealing that the stretch of the plasma is the main reason for discharge energy increase under flow conditions. Then, the efficacy of advanced ignition technologies was investigated under quiescent and flow conditions, demonstrating their suitability under various background conditions. Finally, both single cylinder research engine and 4-cylinder turbocharged GDI production engines were used to further investigate the performance of dual coil and 3-core igntion system under lean conditions. Conclusions can be drawn as below. 1. Under flow conditions, the stretch of the plasma channel can increase its total resistance, resulting in an increase in total discharge energy; under higher discharge current amplitude, the spark energy can be amplified as much as 9 times under 20 m/ s cross flow speed compared with quiescent conditions. 2. The plasma resistance per unit length is mainly governed by the discharge current, and is less sensitive to background pressure and flow speed. 141 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="142"?> 3. Transient ignition strategies, such as transient high current have better ignition performance under quiescent conditions, because of the initial thermal expansion and locally generated turbulence; the performance of the discharge current boost strategy relies on the flow to stretch the plasma channel to increase both ignition volume and ignition energy; both discharge current and discharge duration are important to initiate the flame kernel under extreme conditions. 4. Long discharge duration by dual-coil strategy can not shorten the ignition delay, but can stabilize the combustion process at extreme early spark timings; 3-core ignition system can provide robust ignition control under a wide range of engine loads, while only showing major ignition benefits under lean limit conditions, where ignition receives major challenges. References 1 O’connor, J., M. Broz, D. Ruth et al. Optimization of advanced combustion strategy towards 55% BTE for the Volvo Supertruck program. SAE International Journal of Engines, 10(3): 1217-1227, 2017 2 Position Paper: Technical, regulatory and social challenges for realising CO2-neutral drive technology for cars and commercial vehicles during the coming decades, International Association of Sustainable Drivetrain and Vehicle Technology Research, June, 2021 3 Yu, X., Yang, Z., Yu, S., Ives, M., and Zheng, M., 2017, “Discharge characteristics of current boosted spark events under flow conditions.” ASME 2017 Intern. Combust. Engine Div. Fall Tech. Conf. ICEF 2017. DOI: 10.1115/ ICEF2017-3657. 4 Takahashi, D., Nakata, K., Yoshihara, Y., Ohta, Y. et al., “Combustion Development to Achieve Engine Thermal Efficiency of 40% for Hybrid Vehicles,” SAE Technical Paper 2015-01-1254, 2015, doi: 10.4271/ 2015-01-1254. 5 Zheng, M., Tan, Y., Mulenga, M., and Wang, M., “Thermal Efficiency Analyses of Die‐ sel Low Temperature Combustion Cycles,” SAE Technical Paper 2007-01-4019, 2007, doi: 10.4271/ 2007-01-4019. 6 Gukelberger, R., Alger, T., Gingrich, J., and Mangold, B., “Impact of Operating Parame‐ ters on Ignition System Energy Consumption,” SAE Technical Paper 2014-01-1233, 2014, doi: 10.4271/ 2014-01-1233 7 Hampe, C., Kubach, H., Spicher, U. et al. “Investigations of Ignition Processes Using High Frequency Ignition,” SAE Tech Paper, 2013-01-1633, 2013; doi: 10.4271/ 2013-01-1633. 8 Shiraishi, T., Urushihara, T., Gundersen, M., “A Trial of Ignition Innovation of Gasoline Engine by Nanosecond Pulsed Low Temperature Plasma Ignition,” Journal of Physics D: Applied Physics, 2009, 42: 135208; doi: 10.1088/ 00223727/ 42/ 13/ 135208. 9 Sjöberg, M., Zeng, W., Singleton, D., Sanders, J. et al. “Combined Effects of Multi-pulse Transient Plasma Ignition and Intake Heating on Lean Limits of Well-mixed E85 DISI Engine Operation,” SAE Int J Engines, 2014, 7(4): 1781-1801; doi: 10.4271/ 2014-01-2615. 10 Rohwein, G. J., “An Efficient Power-Enhanced Ignition System,” IEEE Transactions on Plasma Science, 25(2): 306-310, 1997, doi: 10.1109/ 27.602504. 142 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="143"?> 11 Rohwein, G. and Camilli, L., “Automotive Ignition Transfer Efficiency,” SAE Technical Paper 2002-01-2839, 2002, doi: 10.4271/ 2002-01-2839. 12 Yoshida, K., Shoji, H., and Tanaka, H., “Performance of Newly Developed Plasma Jet Igniter,” SAE Technical Paper 1999-01-3327, 1999, doi: 10.4271/ 1999-01-3327. 13 Dale, J.D., Checkel, M.D., Smy, P.R., “Application of High Energy Ignition Systems to Engines,” Progress in Energy and Combustion Science, 23(5-6): 379-398, 1997, doi: 10.1016/ S0360-1285(97)00011-7. 14 Heise, V., Farah, P., Husted, H., and Wolf, E., “High Frequency Ignition System for Gasoline Direct Injection Engines,” SAE Technical Paper 2011-01-1223, 2011, doi: 10.4271/ 2011-01-1223. 15 Alger, T., Gingrich, J., Mangold, B., and Roberts, C., “A Continuous Discharge Ignition System for EGR Limit Extension in SI Engines,” SAE Int. J. Engines 4(1): 677-692, 2011, doi: 10.4271/ 2011-01-0661. 16 Hall, M., Matthews, R., and Ezekoye, O., “Railplug Ignition Operating Characteristics and Performance: A Review,” SAE Technical Paper 2007-01-1832, 2007, doi: 10.4271/ 2007-01-1832. 17 Suess, M, Guenthner, M, Schenk, M, et al., “Investigation of the Potential of Corona Ignition to Control Gasoline Homogeneous Charge Compression Ignition Combustion,” P I MECH ENG D-J AUT, 2011, 226(2): 275-286, doi: 10.1177/ 0954407011416905. 18 Yu X., Zhu H., Zheng M., and Li L., Electrical waveform measurement of spark energy and its effect on lean burn SI engine combustion, SAE Int. J. Adv. & Curr. Prac. in Mobility 2(1): 292-300, 2020. 19 Zheng M., Chen G., Tjong J., Li L., Yu S., Yu X., Yang Z. Spark-based advanced ignition control for future diluted gasoline engines. 4th International Conference on Ignition Systems for Gasoline Engines, December 6-7, 2018, Berlin, Germany. 20 Yu X., Zhang X., Zhang T. et al. Combustion Stability Improvement via Multiple Ignition Sites on a Production Engine. SAE Paper 2020-01-1115, 2020. Definitions/ Abbreviations ICEs Internal combustion engines GHG Greenhouse gas BTE Brake thermal efficiency BMEP Brake mean effective pressure SI Spark ignition TTL Transistor-transistor logic FPGA Field-programmable gate array DC Direct current AC Alternating current EGR Exhaust Gas Recirculation 143 Advanced Ignition Strategies for Gasoline Engine Clean Combustion <?page no="144"?> MAF Mass air flow CAI California analytical instruments FID Flame ionization detector NDIR Nondispersive infrared HCLD Heated chemiluminescence detector PCM Powertrain control module DCO Dual coil offset CA Crank angle IMEP Indicated mean effective pressure MFB Mass Fraction Burnt CA50 Crank Angle of 50% MFB TDC Top dead center Contact information Prof. Ming Zheng Department of Mechanical, Automotive & Material Engineering, University of Wind‐ sor, Canada Email: mzheng@uwindsor.ca 144 Ming Zheng, Guangyun Chen, Jimi Tjong, Liguang Li, Xiao Yu, Linyan Wang <?page no="145"?> Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma - Esgee Technologies, L. Raja - The University of Texas at Austin Abstract: In SI engines, the initial stage of flame kernel formation plays an important role in determining the overall engine efficiency and in reducing the cycle-to-cycle variability. This paper presents a multiphysics computational model to simulate a flame kernel triggered by a spark channel, in a premixed fuel air mixture. The model resolves the breakdown physics that is responsible for generating the spark, and the finite rate chemical kinetics responsible for generating the flame kernel, in a single coupled framework. The spark gap breakdown is modelled using an ignition circuit to accurately capture the elec‐ tromagnetic energy deposited into the gas. As the spark channel stabilizes and stretches under the influence of a cross-flow, the computational model captures the transition of this spark channel into a self-sustained combustion kernel in a single coupled simulation. Key mechanisms via which the spark channel ignites the mixture are identified and studied in detail. A three-dimensional spark ignition and subsequent flame propagation in a crossflow is simulated for a standard spark plug geometry. The flame kernel is shown to exhibit symmetric flame kernel formation around the electrodes at ignition. Later stage of the flame propagation shows the effect of the cross-flow on the flame kernel growth. 1 Introduction In recent decades, as a measure to combat global climate change and promote energy efficiency, there has been a collective thrust by many countries to increase the efficiency of fossil fuel-based automobiles. This has increased the pressure on original equipment manufacturers (OEM) all over the world to produce increasingly fuel-efficient internal combustion engines. Although active research is being carried out in novel combustion techniques such as HCCI [1] and low temperature plasma ignition [2], spark ignited (SI) engines are still widely used for most light-duty automobile applications. To improve the efficiency of SI engines, several solutions have been proposed predominantly with the goal of igniting the mixture under leaner conditions. These include high emission gas recirculation (EGR) combustion <?page no="146"?> [3] igniting at higher compression ratios with diluted mixtures, supercharging [4] and gasoline direct injection [5]. An important factor that affects the stability of the flame in the lean burn limit is the ignition source. Hence, there has been an increased interest towards understanding the dynamics associated with the development of the initial flame kernel and to predict ignition in spark-discharge ignition systems under different operating/ design conditions. Different combustion regimes can be expected in the vicinity of the spark depending on the engine load and speed [6]. Hence, it is important to understand the parameters that govern the initial flame kernel as it significantly influences the thermal efficiency and cyclic variability of the engine. Since the spark channel controls the energy deposited into the fuel-air mixture, manipulating this energy deposition profile has yielded positive results. One of the key parameters that determines the initial flame kernel is the spatial extent of the initial spark channel [7]. It has been found that increasing the gas flow velocity within the spark gap reduces initial combustion fluctuations by shortening the initial combustion period. The increased flow velocity stretches the spark channel and thereby increases the total volume within the initial fuel-air mixture effectively activated by the high-temperature spark kernel. However, it is also found that a large gas velocity can cause the spark to blowout or restrike in which case the desired volumetric activation effect is not achieved [7]. The restrike behavior and the maximum spark channel length before restrike is also found to be strongly correlated to the total current that is supplied to the spark gap [7], which is effectively governed by an external ignition circuit. Several experimental studies have been carried out to classify the initial spark channel formed under different ignition environments [9] [10]. The development of efficient and robust spark ignition system requires a way to effectively explore a highly non-linear design space with several design parameters. In this regard, numerical simulations present a cost-effective option. Numerical simulations have played a key role in elucidating the physics underlying the interaction between the spark channel and the initial flame kernel. Due to a large scale disparity between the spark channel and overall engine dimensions, the Eulerian framework used to perform engine simulations is typically not used to model the initial spark [11]. Instead a Lagrangian framework is employed where the spark is discretized using a fixed number of particles. These particles are placed over the spatial region composed by the initial spark channel and are passively convected by the gas flow. Furthermore, for each particle, mass conservation is accounted for by computing the chemical reaction source terms and energy conservation by solving for the energy deposited by the electrical circuit. The first method that belongs to this class of computational models is the Discrete Particle Ignition Kernel (DPIK) model [12] [13] [14]. In this method the discrete Lagrangian particles are placed on the surface of a sphere that surrounds the spark plug. The particles are only allowed to expand in the radially outward direction with a velocity that depends on the laminar flame speed, local turbulence and energy deposited by the external electrical circuit. An improvement over this method was the arc and kernel tracking ignition model (ATKIM) 146 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="147"?> where the local flow and air-fuel ratio is considered [15] [16]. The Lagrangian particles are placed inside a cylinder surrounding the initial spark channel and are evolved by solving mass and energy equations that account for chemical reactions, energy transferred from the external circuit and wall losses. The newest addition to this family of methods is the Spark-CIMM method which is similar to the ATKIM method but is specialized for stratified-charge direct-injection engines [17]. Although these models provide significant insights into the interaction dynamics between the spark channel and the initial flame kernel, they are inherently limited by the Lagrangian approximation and the number of such points used to discretize the spark channel. More recently, with significant advances in computing power available to perform full-scale engine simulations, several Eulerian-based methods have been proposed to simulate the spark channel [20, 21, 22, 23]. In this class of methods, the spark channel is represented by a fixed collection of computational cells that are assigned an external energy source term that is directly added to the fluid energy equation. The amount of energy imparted to these cells is determined using an energy deposition model [21] and their approximate spatial distribution is obtained from experiments [20]. However, unlike the Lagrangian based methods, the conventional Eulerian approach predetermines the shape of the spark channel which is more or less fixed during ignition. This poses significant challenges when using this approach to model complex spark-plug geometries or cross-flow assisted ignition. To overcome the limitations posed by the Lagrangian approach, the computational model presented in this paper models the spark using a Eulerian framework. However, unlike the conventional Eulerian methods, the spark channel and shape are not predetermined but instead computed by consistently coupling the Maxwell’s equations that describe the electromagnetic behavior of the spark with the Navier-Stokes equa‐ tions that describe its fluid-dynamic behavior. In doing so this framework accurately resolves the spark channel and its volumetric extent. Further, by solving the Maxwell’s equations in conjunction with a multi-species reactive Navier-Stokes formulation, the overall goal of this paper is to consistently resolve spark-initiated combustion. 2 Model description This study employs VizSpark® [16], a multi-physics thermal plasma modeling tool, to simulate spark ignited combustion. This high-fidelity computational tool has been utilized previously in the context of modelling arc formation, stretch and re-strike phe‐ nomena in wide variety of applications such as in spark-plugs, electric vehicle relays and circuit breakers [18] [19]. The VizSpark® computational framework is designed to simultaneously model chemically reacting fluid flow coupled with electromagnetics, surface ablation physics and external circuit dynamics in a fully coupled manner. The thermal plasma model has been validated on a component basis to test the individual flow (viscous, inviscid), gas property thermodynamic and transport generation, and 147 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="148"?> (1) (2) electromagnetic physics. Validation of the full arc physics has been performed for a spark channel in crossflow and for a stationary free-burning arc. 2.1 Arc Model The arc model is obtained by coupling the Navier-Stokes equations that describe the flow / transport phemomena within the arc with the Maxwells equations that describe its electromagnetic response. Further, since the spark is generated in an environment that contains multiple chemically reacting species, the reactive Navier-Stokes equa‐ tions are employed to accurately resolve the combustion physics. The sections that follow outline the key components of the VizSpark® computational model and the manner in which these components are coupled to obtain a multi-physics model for spark ignited combustion. 2.1.1 Navier-Stokes Equations The compressible Navier-Stokes equations are solved to describe the mass averaged flow velocities, the total pressure, mass density, and temperature of the multi-species fluid system. These governing equations are written in vector form as ∂U f low ∂t + ∇ ⋅ F inv + F vis = S f low Here, the vector of conserved variables U f low are listed below as U f low = ρ ρu ρv ρw ρE mix Where ρ is the total fluid density (summed over the densities of all the individual species), u, v and w are the x , y and z components of velocity, and E mix is the mixture total specific energy (sum of kinetic and specific internal energy) per unit volume of the fluid. Gas properties such as ρ and E (specific enthalpy) are obtained as a function of gas static pressure P temperature T and the densities of the individual species. The flow convective/ inviscid flux term F inv is given as 148 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="149"?> (3) (4) F inv = ρux + ρvy + ρwz ρu 2 + P x + ρuvy + ρuwz ρvux + ρv 2 + P y + ρvwz ρwux + ρwvy + ρw 2 + P z ρ E + P ux + vy + wz and the flow diffusive/ viscous flux term F vis is written as F vis = 0 τ xx τ xy τ xz uτ xx + vτ xy + wτ xz + κ ∂T ∂x x + 0 τ yx τ yy τ yz uτ yx + vτ yy + wτ yz + κ ∂T ∂y y + 0 τ zx τ zy τ zz uτ zx + vτ zy + wτ zz + κ ∂T ∂y z The energy diffusive flux consists of energy diffusion due to viscosity and heat conduction represented by Fourier’s law. The viscous stresses are given by τ x i x j = μ ∂v i ∂x j + ∂v j ∂x i + λδ ij ∇ ⋅ V where μ is the coefficient of viscosity and λ is the bulk coefficient of viscosity. Stokes hypothesis is used to express the bulk viscosity term as λ = − 23 μ. The last term in the coupled Navier-Stokes system of equations is the source term S f low given by 149 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="150"?> (5) (6) (7) S f low = 0 J⃗ × B⃗ x J⃗ × B⃗ y J⃗ × B⃗ z J⃗ ⋅ E⃗ − Q˙ rad Here, Joule heating due to the electric field is J ⋅ E , where J is the current density and E the electric field. Lorentz effects on the arc are modelled by J × B forcing terms, where B is the local magnetic field. Cooling of the arc due to radiation is treated using a source term Q˙ rad and a net emission coefficient model. 2.1.2 Finite-rate chemistry equations Individual convection-diffusion-reaction equations are used to model the density of each species in the multi-species mixture. These equations are given by ∂ρY i ∂t + ∇ ⋅ ρY i V − D i ρ ∇Y i = P i Here, ρ represents the total fluid density, Y i the mass fraction of species i, V the convective fluid velocity, D i , P i the mass averaged diffusion coefficient and net production rate of species i. The diffusion coefficients are obtained using the Hirschfielder-Curtiss approximation [24]. Since this approximation does not naturally ensure mass conservation, this constraint is imposed by introducing a correction velocity V c . The correction velocity is given by V c = k = 1 N sp D k ∇ Y k The correction velocity is added to the convective velocity to ensure that when the individual species equations are summed up over all the constitutive species, the fluid continuity equation is recovered. The production terms are computed using the mass fraction of the constitutive species and the fluid temperature that’s obtained from the Navier-Stokes system. The multi-species system, is in turn coupled to the Navier-Stokes system through the internal energy term that appears in the Naver-Stokes energy equation. The mixture total specific internal energy is given by 150 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="151"?> (8) (9) (10) (11) (12) (13) E mix = e mix + V ⋅ V 2 Here e mix is the mixture specific internal energy given by e mix = ℎ mix − p ρ And ℎ mix = ∑ i Y i ℎ i is the mixture specific enthalpy, obtained by mass averaging the individual species enthalpies that are obtained by solving the convection-diffusion-re‐ action system. 2.1.3 Electromagnetic Equations The electromagnetic fields are obtained by solving the current continuity equation ∇ ⋅ J = 0 Where the current density J , is expressed using Ohm’s Law J = σ T , P E If time-dependent magnetic fields can be neglected, then the electric field can be expressed approximately as the gradient of a scalar potential function ϕ. E = − ∇ϕ Equations (10), (11) and (12) in turn lead to the governing equation in the form of a second-order elliptic equation for electrostatic potential ϕ ∇ ⋅ σ ∇ ϕ = 0 For spark plug applications, the currents and current densities are low enough that we do not solve for self-induced magnetic fields in this study. The electromagnetic system is coupled to the flow system through the electrical conductivity σ T , P which is a function of the mixture temperature T and total pressure P . 2.1.4 Immersed Boundary for Electrodes The solid electrodes are modeled using an immersed boundary formulation [19]. The idea behind an immersed boundary method is to identify or tag cells or regions of the mesh as solid, and then to apply forcing terms to the governing equations in those cells such that they behave as solids. The same governing equations for fluid flow 151 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="152"?> (14) and electric field are solved in the arc domain and in the parts of the mesh that are tracked as immersed boundary electrodes. The presence of the solid material is modeled by applying forcing terms on the momentum equations, essentially forcing the fluid velocity in the immersed object regions to the desired velocity [20]. Material properties of the solid are imposed in the regions of the immersed anode and cathodes so that the governing electric field and heat transfer equations are consistent with a solid. 2.1.4.1 Governing Equations in Immersed Electrodes In the immersed electrode regions, the Navier-Stokes equations are modified with the addition of forcing terms that act to force the fluid velocity in the gas to go to zero. For the case of zero velocity, one can effectively ignore the momentum equations. Diffusive heat transfer within the immersed objects and between the gas-solid interface is accounted for by the energy equation of the Navier-Stokes equations. The fifth row corresponding to the energy equation in the Navier-Stokes system reduces to the unsteady heat coduction equation ∂ρC p ∂t + ∇ ⋅ κ ∇ T = 0 To correctly model heat transfer within the immersed object, the effective density, thermal conductivity and specific heat of the immersed object cells are overridden with those of the metal being modeled (in this study copper is used). For obtaining the electric fields, no modification to the governing equation for current continuity is necessary. The only requirement is to set the electrical conduc‐ tivity of the immersed object electrodes to that of the electrode metal. Consequently, electric current can flow seamlessly through the electrodes, through the electrode-gas interface, and through the arc channel. 2.1.5 Multi-timescale chemistry model The spark-ignited combustion problem is a multi-timescale problem with physics occurring at widely disparate timescales. The plasma generation, which is a process involving the acceleration of electrons by the electric field leading to the breakdown and ionization of the mixture, occurs on the nanosecond timescale. On the other hand, the combustion kernel formation, initiated by energy deposited into the mixture by the hot plasma channel, occurs on the microsecond timescale. Simultaneous resolution of both the plasma and combustion chemistry would limit the simulation to a nanosecond timestep and would make it virtually impossible to resolve flame kernel evolution that takes place on the timescale of a few hundred microseconds. To get around this issue it is assumed that on the microsecond timescales, the plasma is in a state of chemical and thermal equilibrium. This is a valid assumption since it takes few microseconds for the electrons to thermalize with the heavy species and the ionization rates to saturate. The equilibrium plasma composition is computed using a Gibbs free energy 152 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="153"?> minimization approach. This method generates, at a given pressure and temperature, a plasma composition that minimizes the Gibbs free energy of the system, which is then taken to be the equilibrium plasma composition. This contains information such as the ionization fraction that are taken into account in determining transport properties such as electrical conductivity. 2.1.6 Gas Properties Closure of the governing equations requires specification of gas thermodynamic properties: the density and specific heats of the mixture; as well as transport properties: viscosity, thermal conductivity, and electrical conductivity. The approach followed in this work divides the constitutive properties into two categories. First, there are properties that are determined by the fast timescale plasma chemistry and second, are those that depend on the slow timescale combustion chemistry. The only property that belongs to the first category is the electrical conductivity which is determined using the electron concentration obtained using the Gibbs free energy minimization approach. All the other gas properties are determined using a mixture averaged formulation that uses the individual species densities obtained from equation (6) and a hard sphere collision model to determine the mass diffusion, viscosity and thermal conductivity coefficients. A mass-averaged ideal gas equation is used to determine the mixture density given the total pressure and mixture temperature. 2.1.7 Radiation Model Arc radiation cooling is modelled using a net emission coefficient model, which acts as a cooling volumetric source term. Net emission coefficients are tabulated as a function of temperature and assumed to vary linearly with pressure. 2.1.8 Modelling Circuits The deposition of electrical energy into the spark-gap occurs through an external circuit attached to the spark-plug. Typical inductive-based ignition circuits contain both primary and secondary sides. A schematic of an inductive based ignition circuit is shown in Figure 1. The primary side contains a battery (power source) and switches S1 and S2 determine whether the primary coil is connected to the battery. L P and L S denote primary and secondary inductances along with their leakage values. Parasitic spark plug capacitance and resistances are denoted by C S and R G respectively. The voltage across the spark gap is given as V SG . 153 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="154"?> (15) (16) (17) Figure 1: Representative circuit diagram of an inductive ignition circuit A simplified version of this circuit could be obtained by ignoring the primary side (Figure 2). This simplification is a good approximation to the full circuit model and captures important circuit transients observed in single-pulse inductive ignition systems. Figure 2: Ignition circuit considering only the secondary side. The circuit equations for the secondary side are given as: L s di L dt + i L R Ls + R SG i SG + V SG = 0 ∫ 0 t i c dt C s − V SG − R SG i SG = 0 i L = i C + i SG 154 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="155"?> (18) i SG = − σ ∇φdS The inputs into this model are effective secondary inductance (L s ), coil resistance (R L , spark plug resistance (R SG ), parasitic capacitance (C s , and initial coil energy. The spark gap current is calculated by integrating current density at the circuit connection boundary (Equation 18). Knowledge of stored energy in the coil could be used to estimate the initial current flowing through the inductor. Before gas breakdown, no current flows through the spark-gap, and the capacitor charges up. As a result of this, the voltage across the spark gap increases until gas breakdown occurs, and the spark plug begins to conduct current. Finally, as the inductor loses energy, the arc begins to weaken and eventually quenches. 2.1.9 Modelling of Gas Breakdown Electrical breakdown of a gas occurs when enough of the free electrons in the gas gain sufficient energy from an applied electric field such that they are able to create and maintain a conductive plasma channel through ionizing collisions with neutral particles. The breakdown phase occurs over very short time scales (tens of nanoseconds) and the plasma during this transition phase is in a state of thermal and chemical non-equilibrium. To accurately capture the physics of plasma breakdown; one would ideally use a non-equilibrium model that treats the gas temperature and the electron temperatures separately and models the individual species along with their finite rate kinetics such as electron impact reactions. In an equilibrium model, it is assumed that all particles have the same temperature and that the chemical composition, thermodynamic and transport properties are in thermal and chemical equilibrium allowing them to be tabulated as functions of temperature. The breakdown event is due to the tight coupling between Joule heating which depends on the (temperature dependent) electrical conductivity and the gas temperature (which increases due to Joule heating). The electric field and the small but finite amount of electrical conductivity at low gas temperatures leads to Joule heating of the gas. As the gas temperature increases, so does the electrical conductivity leading to more Joule heating. Once the gas temperature reaches its breakdown threshold, the conductivity explosively increases and the gas breaks down. This leads to the formation of a conductive plasma medium with high temperatures, high electrical conductivities and low electric fields. The production/ destruction and motion of the electrons in the gas discharge can be approximated using a temperature dependent electrical conductivity. Electrical conductivity in the gas is due almost entirely to electron motion, so one can think of the electrical conductivity as a stand-in for the electron physics and a determining factor for predicting current densities in the arc. The electrical conductivity for most gases is a highly non-linear function of the gas temperature (see Figure 3). Below the 155 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="156"?> threshold value, conductivities are very low and the gas behaves as a dielectric. Above the threshold, the gas has essentially broken down and is a conductive plasma. Figure 3: Electrical conductivity of air at 1 atmosphere as a function of temperature. At low gas temperatures, the dependence of the electrical conductivity on the gas temperature is not sufficient to fully model breakdown as it does not capture the non-equilibrium nature of the electrons at high electric fields and low temperatures. At low gas temperatures and high applied electric fields, the electrons will be efficiently heated and can have a different, much higher temperature then the surrounding “cold” gas. These “hot” electrons provide the gas an effectively higher electrical conductivity than the single (gas) temperature equilibrium thermal conductivity in Figure 3 would suggest. Thermal arc models can account for the non-equilibrium electron temperature dependence of electrical conductivity either by the inclusion of a full two-temperature model for electrons and heavy species, or by incorporating a model that adds the electric-field dependence of the electrical conductivity when the gas temperatures are low. 2.1.9.1 Two-Parameter Breakdown Model In this work, we incorporate the electric field/ electron temperature dependence of electrical conductivity using a simplified two parameter model. The model is a Heaviside step function that modifies the electrical conductivity in the Joule heating source term when a breakdown electric field threshold value is reached. A breakdown reduced electric field En bd for the gas is specified. The reduced electric field is defined as the electric field strength divided by the total gas particle number 156 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="157"?> (19) density (or equivalently the total pressure). Breakdown electric field values are known for many common gas mixtures from Townsend gas breakdown theory [22]. The Joule heating term, which is of the form σ E is modified by adding and electric-field dependent contribution σ e − f ield to the electrical conductivity whenever the reduced electric in a region exceeds the specified breakdown electric field. S joule = σ + σ e − f ield E if , En > En bd σ E otℎerwise The chief advantage of the two-parameter breakdown model is its simplicity. The breakdown electrical conductivity however, is an unknown calibration parameter and needs to be adjusted to match experiments or estimated using a full non-equilibrium breakdown model to determine the electrical conductivity as a function of the electric fields in the gas. In this work, we used 125 Td (Townsend) as the breakdown reduced electric field, En br eakdown which is approximately the breakdown reduced electric field of air. The electrical conductivity σ e − f ield term is a parameter and values ranging from 0.001 to 0.1 give reasonable results. In the simulations in this work, a breakdown σ e − f ield value of 0.0009 was used. 2.2 Simulation Configuration 2.2.1 Geometry and mesh Figure 4: Geometry and numerical mesh for spark plug configuration The computational domain to resolve the pin-to-pin spark and the subsequent com‐ bustion event is shown in Figure 4. A cylindrical pin anode 0.5 mm in diameter and 157 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="158"?> 0.7mm in length is placed 0.7 mm away from a cylindrical cathode. The cathode is 0.5 mm in diameter and is 0.7 mm long. The entire domain is 20 mm in the dimension perpendicular to the gap (x) and 10 mm along the direction parallel to the interelectrode spark gap (y). The domain is meshed using ~90,000 mixed (quad and tri) mesh elements, with fine quad elements employed in the vicinity of the spark gap to capture the spark channel during the breakdown stage. 2.2.2 Initial and boundary conditions A premixed methane-air mixture is considered for the simulation of spark ignited combustion. The methane combustion chemistry is modelled using the GRI 3.0 mechanism [25] that contains 53 species and 325 reactions. The domain is initialized using a premixed methane-air mixture at a stochiometric ratio of 1. This corresponds to the mass fractions of methane, oxygen and nitrogen being set to 0.05, 0.2 and 0.75 respectively. The flow is set to an initial pressure of 3 bar and an initial temperature of 700K with the intention to mimic the pre-ignition conditions in a typical IC engine. Since one of the main objectives of this simulation is to demonstrate a flame kernel generated in the presence of a cross-flow, the entire domain is initialized with a constant cross flow velocity of 15 m/ s in the x direction (going left to right in Figure 4). The left boundary in Figure 4 is assigned an inflow boundary condition with a velocity of 15 m/ s in the horizontal direction. The inflow boundary is designed to transport uncombusted fuel into the domain, this is achieved by setting the inflow pressure to 3bar, temperature to 700 K and the fuel air mass fraction to the same value as the initial conditions. All the other fluid boundaries (top, bottom and right) are set to a non-reflecting outflow boundary condition. The surfaces of the electrodes are assigned a no-slip adiabatic wall boundary condition. The electromagnetic equations are set up by assigning the cathode base as the boundary connected to the external circuit. The current density is computed as an integrated quantity over this surface and solved in a coupled manner with the circuit equations. All the remaining surfaces of the cathode are assigned a continuous boundary condition effectively letting the conductive channel pick the path of least resistance through the simulation domain. A simplified secondary ignition circuit as shown in Figure 2 is considered for this work. The circuit parameters are as follows Electrical Circuit Parameters Value Secondary inductance (L s 1 H Secondary coil resistance (R L ) 5000 Ohms Spark-plug resistance (R SG ) 10000 Ohms Parasitic spark-plug capacitance (C s ) 50 pF Initial Coil Energy 100 mJ Tab. 1: Circuit parameters 158 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="159"?> 2.3 Results The dynamics of spark ignited combustion can be divided into two stages. First is the breakdown stage, characterized by intense gas heating due to the electromagnetic energy deposited via the external circuit. This generates conductive spark channels that sustain high current densities. As the spark gap starts to sustain current, the effective voltage drop across the gap reduces. This results in the saturation of the rapid gas heating that was seen in the early stages of the breakdown process and as a result the temperature and current densities within the channel attain quasi-steady values. The second stage is characterized by the ignition of the mixture, triggered by the energy deposited into the gas by the hot conductive channel. Compared the breakdown stage that takes approximately few microseconds, the combustion and flame front propagation is relatively slower, taking place over several hundreds of microseconds. In the sections that follow, results pertaining to the spark formation and breakdown stage are presented first and are followed by the results of the combustion stage. 2.3.1 Breakdown Figure 5: Temporal trace of voltage across the spark gap 159 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="160"?> Figure 6: Temporal trace of current across the spark gap The waveforms of the voltage and current across the spark gap are shown in Figure 5 and Figure 6 respectively. The breakdown process is triggered by charging up the inductor with a finite amount of energy (100 mJ Table 1). In the initial stages of the breakdown transient, the spark-gap does not conduct any current, and as a result the energy stored in the inductor goes into charging up the capacitor C s (Figure 2). As the capacitor charges up, the voltage across the spark gap V SG ) begins to rise. This is shown in Figure 5, between 0-2 microseconds, the voltage across the spark gap rises linearly from 0V to -15000V. At -15000V the arc voltage reaches the breakdown threshold for the gas. This triggers the spark-gap breakdown, leading to the formation of conducting channels that sustain large current densities. Figure 6 shows a sharp rise in current at the time instant when the gas breakdown (2 microseconds) occurs. As current begins to flow through the gap the voltage across the gap drops to around 50-100 V. The voltage drop leads to a corresponding drop in the current, the current falls from around 1.6 A at breakdown to around 0.4 A post-breakdown. The highly non-linear nature of spark-gap breakdown, particularly the almost instantaneous rise in current and drop in voltage at the breakdown threshold, illustrates the importance of modelling the external circuit to resolve this phenomenon in a physically consistent manner. Over the next few microseconds, the arc stretches due to cross-flow, thereby increas‐ ing its resistance. The arc voltage shows an increase due to this rise in arc resistivity, and reaches around 500 V at 25 microseconds. Over the next few hundred microseconds, the arc voltage gradually begins to rise and the current through the spark channel slowly begins to fall. This quasi-steady state corresponds to the case where the spark channel temperature attains a steady value and the channel begins to stretch under 160 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="161"?> Figure 7: Temperature disctibution in the spark gap at 2 and 3 microseconds the influence of the cross-flow. It is this spark channel and the high temperatures that it sustains over relatively long timescales (hundreds of microseconds) that eventually ignites the mixture. Figure 7 shows the temperature distribution in the spark gap at time instants just before and after breakdown. Shown on the left is the temperature at 2 microseconds prior to breakdown. It is observed that the spark gap is more or less uniformly heated by the externally applied electric field and there is no clear indication of a current conducting channel. However, at 3 microseconds, which corresponds to a time instant just after breakdown, a single high temperature current conduction channel is formed. Figure 8: Current density distribution in the spark gap at 2 and 3 microseconds This can also be seen in Figure 8, that illustrates the current density transients before and after breakdown. Before breakdown (Figure 8, left) the current density distribution 161 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="162"?> across the spark gap is diffuse. Peak values are attained at the electrode corners but there is no clear indication of a current carrying channel. However, once breakdown occurs (Figure 8, right) a single current conduction channel can be observed on the left side of the spark gap. Since maximum electric field amplification occurs at the electrode corners, the arc root at the cathode and anode are observed to preferentially attach to these corners. Another important point to note is that although the breakdown timescales are an order of magnitude smaller than the flame kernel propagation timescales, some amount of combustion does occur during breakdown. Shown in Figure 9 is the spatial distribution of the methane mass fraction at 2 and 3 microseconds. At 2 microseconds which corresponds to a time instant of incipient breakdown, it is observed that the relatively higher value of joule heating between the electrodes initiates combustion reactions. The methane mass fraction is observed to be the lowest around the electrode corners which are regions where maximum joule heating is obtained. Post-breakdown (Figure 9, right) it is seen that the all the methane in the interelectrode region is completely consumed and a cylindrical flame kernel, with a diameter approximately equal to the spark gap diameter, is obtained. The nearly complete combustion that occurs within the spark gap during the breakdown process indicates the importance of modelling the electromagnetics and finite rate chemical kinetics in a fully-coupled manner. Although the time taken for the entire mixture in the domain to ignite is much larger than the electromagnetic timescales, there is significant overlap between combustion and joule heating over smaller length scales. Accurately simulating this overlap could be potentially crucial for capturing phenomena such as misfires and blow-outs that occur in the lean burn limit. Figure 9: Methane mass fraction in the spark gap at 2 and 3 microseconds 162 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="163"?> 2.3.2 Spark initiated ignition The spark gap gas breakdown is followed by a dwell state where the current and voltage across the gap attain quasi-steady values and the peak temperature within the spark channel begins to saturate. Figure 10: Temperature distribution in the spark gap and 10 and 20 microseconds Although the peak temperature and the electromagnetic power transmitted to the gas begin to attain steady values, the shape of the spark channel and effectively the regions of energy deposition dynamically vary in response to the flow field. Figure 10 shows the temperature in the spark channel and its vicinity at 10 and 20 microseconds. On the timescale of tens of microseconds, the spark channel begins to stretch due to the applied flow field. It is seen that the spark channel, which, as shown in Figure 8, is formed towards the left side of the spark gap, begins to move towards the right due to the 15 m/ s cross flow (directed from left to right). The 10-microsecond snapshot shows the spark channel with the roots still attached to their original locations, at the left corners of the cathode and anode. The spark channel however, demonstrates a significant amount of stretching in response to the flow field. Compared to the 3-microsecond snapshot where the spark channel shows a convex profile pointed away from the spark gap, at 10 microseconds, it shows a concave profile pointing into the spark gap. At 20 microseconds (Figure 10, right) the cross flow begins to affect the arc root locations as well. As the spark stretches it becomes ineffective from an energy standpoint to sustain such a long spark channel. As a result, the arc root locations also begin to move towards the right, following the motion of the bulk of the channel. It can be seen in Figure 10 that the anode root has completely moved from the left corner at 10 microseconds to the right corner at 20 microseconds. The cathode root is shown in transition, attached to the mid-point of the cathode face at 20 microseconds. 163 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="164"?> Figure 11: X velocity within the spark gap at 10 and 20 microseconds Figure 11 shows the X component of velocity (component in the direction of cross-flow) within the spark gap at 10 and 20 microseconds. In 2D simulations it is observed that as the spark channel stretches, it entrains fluid from the side opposite to the stretch direction. This entrainment effect effectively increases the velocity within the spark gap, and consequently the spark stretch rate, which is directly proportional to this velocity. It is observed that the average velocity preceding the spark increases from around 150 m/ s at 10 microseconds to nearly 250 m/ s at 20 microseconds. It is important to note here that this phenomenon is a consequence of the 2-dimensional nature of the problem and this makes it all the more important to carry out 3D simulations of spark-ignited combustion to understand the importance of spark gap entrainment. Figure 12 shows the mass fraction of methane and water in the spark gap at 10 and 20 microseconds. The mass fraction of water serves as a good indicator of the flame front location since it is produced at the highest rate at the flame front. It can be seen that in addition to the spark channel, the flame kernel shape is also significantly affected by the cross flow. The top panel is Figure 12 indicates a significant increase in the water concentration on the right side of the spark gap from 10 to 20 microseconds. This indicates that as the spark channel, that deposits energy into the gas and initiates combustion, moves to the right side due to the cross flow, the spark ignited flame follows this energy deposition zone. This is also seen in the bottom panel of Figure 12 that shows the methane mass fraction within the spark gap. It can be observed that more methane is consumed in the direction of cross flow, again consistent with the location of the spark channel. This dynamics also indicates the nature of energy deposition by the spark channel. The high temperature regions with large values of Joule heating are typically constrained to the diameter of the spark channel. As shown in the temperature profile in Figure 10 the effective volume of the spark is a small fraction of the total volume of the flame kernel. The spark channel can also be seen in Figure 12 where the 164 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="165"?> Figure 12: Water (top) and Methane (bottom) mass fraction in the spark gap at 10 and 20 microseconds water concentration is very low in the high temperature channel due to the initiation of reverse reactions. As the flame kernel grows volumetrically, due to the expansion of the flame front, it is observed that the total volume of the spark channel relative to the volume of the ignition kernel actually decreases with time. However, it is important to maintain this energy deposition until the flame attains self-sustained combustion. This again emphasizes the importance of the external ignition circuit that controls the duration of the dwell phase. 165 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="166"?> Figure 13: Temperature distribution in the simulation domain at 50 and 150 microseconds The final phase of the transient is characterized by the quenching of the spark channel. During the initial stages of the spark transient preceding breakdown, the capacitor is charged up to a finite amount of energy. As the capacitor begins to discharge this energy through the spark gap, the current through the gap begins to drop as shown in Figure 6. Figure 14 shows the current density in the domain at 50 and 150 microseconds. At 50 microseconds a clear spark channel with high current densities can be observed, however at 150 microseconds as the current through the spark gap goes to zero, the current density drops by three orders of magnitude. This results in a corresponding drop in the joule heating and the effect of this can be seen on the temperature distribution in the simulation domain. At 50 microseconds (Figure 13) when the joule heating in the channel is high, a clear high temperature spark channel is observed. The high temperature regions correspond to the high current density regions in Figure 14. However, at 150 microseconds, the drop in current density results in the quenching of the spark channel. The overall temperature profile also shows a clear indication of a spatially growing flame kernel indicating that the combustion has reached a self-sustained state. This effectively means that, for the circuit parameters chosen in Table 1, the total time over which electromagnetic energy was deposited into the spark gap was sufficient to trigger self-sustained combustion of the fuel-air mixture. 166 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="167"?> Figure 14: Current density distribution in the simulation domain at 50 and 150 microseconds 2.3.3 Three-dimensional simulations Preliminary results from a 3D simulation of spark ignited combustion were also performed. A standard J-type spark plug with an inter-electrode gap of 0.86 mm is studied. A cuboidal flow field around the spark plug is simulated for a cross flow velocity of 25 m/ s. The inlet boundary conditions are specified at the y-z plane at x = -5 mm, upstream of the spark plug centerline. The outlet plane is set at 10 mm downstream of spark plug centerline. The lateral x-z and x-y plane boundaries are set to symmetric boundary conditions to ensure that the cross flow remains normal to the spark plug. The spark plug electrodes are set at no-slip boundary condition. The base of the J-electrode (blue in Figure 15, bottom image) is set to ground while the base of the cylindrical electrode is connected to an external circuit. The powered electrode is connected to a circuit shown in Figure 2. The initial ambient pressure is set at 4 bar, while the flow field is initialized at u = -25 m/ s (in x-direction). The entire flow volume is filled with the methane air mixture with an air-fuel equivalence ratio of 1.6. Figure 14 (TOP) shows a cross-sectional view of the underlying mesh as observed along the y axis, normal to the x-z plane. The inter-electrode gap is refined to have a maximum cell size of 100 µm. A cuboidal refinement zone in the direction of the cross-flow is set with minimum mesh size of ~200 µm. The size of this zone is sufficient to capture the maximum spark stretch during the course of the simulation. 167 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="168"?> Figure 15: (TOP) Schematic of 3D simulation domain illustrating the direction and magnitude of the cross flow. (BOTTOM) Cross-sectional view of the mesh refinement around the spark gap 2.3.3.1 Initial stage of ignition Upon switching on the circuit, the voltage difference across the electrodes starts to build up, which increases the electric field in the inter-electrode gap. This sudden build-up of energy at the electrodes gets transferred to the gaseous media in the gap, which results in the pressure wave moving from the powered electrode towards the ground. Figure 16(a) shows the pressure wave which starts at the cathode and progresses towards the anode (ground) prior to the formation of the spark channel. The snapshot showed in Figure 16(a) refers to the 2µs after the simulation is started. It could be noted that the peak pressure in the wave reaches up to 11 bar in the wake of the incipient shock wave. Simultaneously, the temperature field exhibits local maxima near the edge of the cathode electrode. The maximum tempearture is found to be ~2800K which is higher than the auto-ignition tempearture of methane-air mixture. This suggests that combustion reactions at the cathode precede the spark formation. Figure 16(b) shows the pressure field at the 4 µs, when it has reflected off the opposing face on the ground electrode. It could be seen that as the pressure wave reflects, it spreads radially outward and becomes weaker as it progresses. The maximum pressure in this case, lies in the corner between the ceramic around the cathode. The thermal field at t = 4µs 168 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="169"?> exhibits a spherical flame kernel wrapped symmetrically around the electode. It should be noted that planar field does not show the spark channel, which is expected to have temperature larger than 10,000K. Since the spark channel lies at the outer edge of the electrode, the planar field shown here does not intersect with the spark channel. Figure 16: Pressure and Temperature field at a. t = 2 µs b. t = 4 µs Fig 16(b) also shows the spark channel (shown in red) which forms in the inter-electrode gap close to the left corner of the ground electrode. The spark channel in this representation is defined the by regions in the flow field which has electric conductivity higher than the threshold 5 S/ m. It should be noted that the initial spark channel formation seems to be unaffected by the cross-flow direction, as it forms on the upstream side of the interelectrode gap. The arc root is found to anchor near the sharp corner of the ground electrode. The presence of the methane-air flame could be marked by the presence of com‐ bustion products such as water or carbon dioxide concentration. In this work, we specify the flame surface by the water vapor mass fraction greater than 1%. Figure 16 shows the flame surface and the velocity field in the central x-z plane at t=4 us. In the inter-electrode a symmetric flame ball has already formed, which is shown from a different perspective angle in Figure 17. It is noted that the flame formation occurs in the wake of the reflected pressure wave. The strength of the pressure wave and 169 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="170"?> the combustion in the flame kernel generates a radially outward flow, as is evident in the vector field. Hence, the flame growth around the electrodes also happens in a symmetric fashion. The fast-propagating pressure wave nullifies the effect of the cross flow on the initial flame propagation. Figure 17: Flame surface, spark and velocity field at t = 4 µs 2.3.3.2 Flame propagation Previous experimental studies have differentiated between the stages of flame prop‐ agation during the spark ignition. It has been reported that while the initial stage dominated by a pressure wave sustains for a few microseconds, the later stage of the flame propagation is characterized by a slow and steady propagation of flame kernel into the fuel air mixture. At this stage, the flame propagates slowly, and as expected, is dominated by the cross-flow which shapes the flame kernel. Figure 18 shows the spark, the flame surface and the velocity field at t = 20 µs, 40 µs and t = 80 µs respectively. It is noted that the spark channel stretches as a result of its interaction with the cross flow. However, the arc roots on the electrode surface remain at nearly the same location. The spark stretch occurs in the central section of the spark channel which moves along the crossflow. The flame kernel at t=20 µs, appears to grow in the direction of the cross-flow. As the time progresses, the effect of the flow advection is apparent as the flame moves to the left as shown in the Figure 17 170 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="171"?> (b) and (c). The left front of the flame kernel moves at an average speed of ~94 m/ s and ~37 m/ s over the time duration between the 20-40 µs and 40-80 µs respectively. The slow down of the flame front over time is typical of the transition between the first stage of spark ignition vs the self-sustained propagation of the flame front. Figure 18: Flame, spark (red) and planar velocity field at a. t = 20 µs, b. t = 40 µs and c. t = 80 µs Figure 18 shows the growth of the flame kernel over time. The volume of the flame kernel is calculated using the volume enclosed by the 1% water vapor mass fraction isosurface. At time t = 10 µs, we see a clear change in the rate of flame kernel growth. At the initial stages of flame propagation, the rate of flame kernel volume growth is 2.2 mm 3 / µs, while this rate decreases to 0.29 mm 3 / µs at the self-sustained flame propagation stage. 171 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="172"?> 3 Conclusion In summary, a multiphysics framework for simulating spark-initiated combustion events has been developed and utilized to demonstrate successful ignition in a premixed methane-air mixture. The results obtained using this framework provide key insights into the dynamics of spark-initiated combustion. A coupled external circuit system was employed to resolve spark-gap breakdown and the formation of conducting spark channels. The circuit coupling was crucial for accurately resolving the sharp rise in the current and a sharp drop in the voltage across the spark gap. Moreover, it also ensures that the spark energy that eventually ignites the mixture is delivered to the gas in a physically realistic manner. The spark channel was initially formed with the roots attached to left corners of the cathode and anode. However due to the 15 m/ s cross flow velocity the spark channel was observed to translate and attach to the right side of the spark gap (right corners of electrodes). It is also important to note that the within the spatial confines of the gap the spark formation and combustion take place on similar timescales. Post-breakdown, the spark channel is formed inside the flame kernel, further corroborating the approach of simulating the electromagnetics and the finite rate kinetics in a coupled manner in one single computational model. The attachment of the spark channel to right side of the electrode and subsequent stretching under the influence of the cross flow, activates a large volume of the fuel air mixture leading to the formation of a stretched flame kernel. It was observed that although the volume over which electromagnetic energy was deposited into the gas was a small fraction of the total volume of the flame kernel, it was crucial to maintain this power until the flame kernel attained self-sustained propagation. The three-dimensional simulation provided interesting insights into the initial phase of the flame ignition and the flame propagation under cross flow. It was noted that the pressure wave travels from the cathode to the anode. Upon reflection from the anode surface, the pressure wave progresses radially outward while decreasing in the pressure wave strength. The generation of the flame kernel occurs in a symmetric fashion, however with time the flame kernel propagates faster along the cross flow. This leads to the movement of the flame kernel in the direction of the crossflow. The flame kernel growth could be distinguished as initial fast propagation followed by slow sustained propagation. 4 References [1] Thring, R.H., 1989. “Homogeneous-charge compression-ignition (HCCI) engines” (No. 892068). SAE Technical paper. [2] Ju, Y., & Sun, W. (2015). “Plasma assisted combustion: Dynamics and chemistry”. Progress in Energy and Combustion Science, 48, 21-83. [3] Jackson, N. S., Stokes, J., Lake, T. H., Sapsford, S. M., Heikal, M., & Denbraft, I. (1996). “Un‐ derstanding the CCVS stratified EGR combustion system” (No. 960837). SAE Technical Paper. 172 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="173"?> [4] Guzzella, L., Wenger, U., & Martin, R. (2000). “IC-engine downsizing and pressure-wave supercharging for fuel economy” (No. 2000-01-1019). SAE Technical Paper. [5] Iwamoto, Y., Noma, K., Nakayama, O., Yamauchi, T., & Ando, H. (1997). “Development of gasoline direct injection engine” (No. 970541). SAE technical paper. [6] Peters, N. (1988, January). “Laminar flamelet concepts in turbulent combustion” Symposium (International) on Combustion (Vol. 21, No. 1, pp. 1231-1250). Elsevier. [7] Shiraishi, T., Teraji, A., & Moriyoshi, Y. (2016). “The effects of ignition environment and discharge waveform characteristics on spark channel formation and relationship between the discharge parameters and the EGR combustion limit.” SAE International Journal of Engines, 9(1), 171-178. [8] Kono, M., Kumagai, S., & Sakai, T. (1977, January). “The optimum condition for ignition of gases by composite sparks.” In Symposium (International) on Combustion (Vol. 16, No. 1, pp. 757-766). Elsevier. [9] Pashley, N., Stone, R., & Roberts, G. (2000). “Ignition system measurement techniques and correlations for breakdown and arc voltages and currents” (No. 2000-01-0245). SAE Technical Paper. [10] Lee, M. J., Hall, M., Ezekoye, O. A., & Matthews, R. D. (2005). “Voltage, and energy deposition characteristics of spark ignition systems” (No. 2005-01-0231). SAE Technical Paper. [11] Thiele, M., Selle, S., Riedel, U., Warnatz, J., & Maas, U. (2000). “Numerical simulation of spark ignition including ionization.” Proceedings of the combustion institute, 28(1), 1177-1185. [12] Tan, Z., & Reitz, R. D. (2003). “Modeling ignition and combustion in spark-ignition engines using a level set method” (No. 2003-01-0722). SAE Technical Paper. [13] Dahms, R., Fansler, T. D., Drake, M. C., Kuo, T. W., Lippert, A. M., & Peters, N. (2009). “Modeling ignition phenomena in spray-guided spark-ignited engines” Proceedings of the Combustion Institute, 32(2), 2743-2750. [14] Falfari, S., & Bianchi, G. M. (2007). “Development of an ignition model for SI engines simulation” (No. 2007-01-0148). SAE Technical Paper. [15] Duclos, J. M., & Colin, O. (2001). “Arc and Kernel Tracking Ignition Model for 3D Spark Ignition Engine Calculations”, 5th Int. In Symp. on Diagnostics and Modeling of Combustion in Internal Combustion Engines, COMODIA. [16] Duclos, J. P., Zolver, M., & Baritaud, T. (1999). “3D modeling of combustion for DI-SI engines.” Oil & Gas Science and Technology, 54(2), 259-264. [17] Dahms, R. N., Drake, M. C., Fansler, T. D., Kuo, T. W., & Peters, N. (2011). “Understanding ignition processes in spray-guided gasoline engines using high-speed imaging and the ex‐ tended spark-ignition model SparkCIMM. Part A: Spark channel processes and the turbulent flame front propagation.” Combustion and flame, 158(11), 2229-2244. [18] Karpatne, A., Breden, D., & Raja, L. (2018). “Simulation of Arc Quenching in Hermetically Sealed Electric Vehicle Relays” (No. 2018-01-0765). SAE Technical Paper. [19] Breden, D. P., Karpatne, A., & Raja, L. (2018). “Modelling of Electrode Erosion for Prediction of Spark Plug Lifetime” (No. 2018-01-0175). SAE Technical Paper. 173 Full-fidelity numerical modelling of spark ignition and subsequent flame kernel evolution <?page no="174"?> [20] Zhang, A., Scarcelli, R., Lee, S. Y., Wallner, T., & Naber, J. (2016). “Numerical Investigation of Spark Ignition Events in Lean and Dilute Methane/ Air Mixtures Using a Detailed Energy Deposition Model” (No. 2016-01-0609). SAE Technical Paper. [21] Yang, X., Solomon, A., & Kuo, T. W. (2012). “Ignition and combustion simulations of spray-guided SIDI engine using Arrhenius combustion with spark-energy deposition model” (No. 2012-01-0147). SAE Technical Paper. [22] Scarcelli, R., Matthias, N., & Wallner, T. (2013). “Numerical investigation of combustion in a lean burn gasoline engine” (No. 2013-24-0029). SAE Technical Paper. [23] Givler, S. D., Raju, M., Pomraning, E., Senecal, P. K., Salman, N., & Reese, R. (2013). “Gasoline combustion modeling of direct and port-fuel injected engines using a reduced chemical mechanism” (No. 2013-01-1098). SAE Technical Paper. [24] Poinsot, T., & Veynante, D. (2005). Theoretical and numerical combustion. RT Edwards, [25] Bowman, C. T., Frenklach, M., Gardiner, W. R., & Smith, G. (1999). “The GRI 3.0 chemical kinetic mechanism.” University of California: Berkeley, CA, 77. 174 R. Ranjan, A. Karpatne, V. Subramaniamm, S. Thiruppathiraj, D. Breden, A. Sharma <?page no="175"?> Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine Tim Franken 1 , Krishna P. Shrestha 1 , Lars Seidel 2 and Fabian Mauß 1 1 2 Brandenburg University of Technology Cottbus-Senftenberg, Siemens-Halske-Ring 8, D-03046 Cottbus LOGE Deutschland GmbH, Querstraße 48, D-03046 Cottbus Abstract: The climate protection plan of the European Union requires a signifi‐ cant reduction of CO 2 emissions from the transportation sector by 2030. Today ethanol is already blended by 10vol-% in gasoline and further increase of the ethanol content to 20vol-% is discussed. During the ethanol production process, distillation and molecular sieving is required to remove the water concentration to achieve high-purity ethanol. However, hydrous ethanol can be beneficial to suppress knock of spark ignition engines. The hygroscopic nature of ethanol can allow to increase the water content in gasoline - water emulsions even more, without adding additional surfactants, and improve the thermal efficiency by optimized combustion phasing, while keeping the system complexity low. Hence, the effect of gasoline - ethanol - water mixtures on the auto-ignition in a single-cylinder spark ignition engine is investigated by using multi-dimensional simulation and detailed chemistry. The gasoline - ethanol mixtures are defined to keep the Research Octane Number constant, while the Motored Octane Number is decreasing. In total five surrogates are defined and investigated: E10 (10vol-% ethanol-in-gasoline), E20, E30, E70 and E100. The water content is determined according to experimentally defined ternary diagrams that evaluated stable gasoline - ethanol - water emulsion at different gasoline - ethanol blending ratios. The auto-ignition modes of the surrogates are analyzed using the diagram, which determines if hotspots are within harmless deflagration or harmful developing detonation regime. The strongest auto-ignition is observed for the E10 surrogate, while increasing ethanol content reduces the surrogate reactivity and increases the resonance parameter . No auto-ignition of the unburnt mixture is observed for the E70 and E100 surrogates. The addition of hydrous ethanol decreased the excitation time of the surrogates, especially at low ethanol content, wherefor the reactivity parameter is significantly increased. The hotspots for E10, E20 and E30 surrogates with hydrous ethanol are found within the developing detonation <?page no="176"?> regime, while hotspots of the E70 surrogate with hydrous ethanol are found in the transition regime. For the hydrous E100 surrogate no auto-ignition is predicted because of reduced temperature of the unburnt mixture due to water vaporization, which outweighs the increased reactivity due to water vapor addition. 1 Introduction The climate protection plan of the European Union “The European Green Deal” [1] requires a 13% reduction of green-house-gas (GHG) intensity from the transportation sector by 2030. Thereby, the utilization of advanced biofuels shall be increased to 2.2% at the same time. Today, bioethanol is already blended by 10vol-% in gasoline, and further increase of the ethanol content to 20vol-% is discussed. The advantage of blending ethanol with gasoline is the reduced C: H and C: O ratio, higher vaporization enthalpy, increased research octane number (RON) and motored octane number (MON) [2], and renewable production pathways via 1 st generation and especially 2 nd generation feedstocks [3,4]. The consideration of feedstocks is important, e.g. Lark et al. [5] stated a 24% increase of the GHG life cycle emissions of corn ethanol compared to baseline gasoline because of intensified domestic land use, which is opposing findings of Lewandrowski et al. [6], who projected 39-43% lower GHG emissions of corn ethanol compared to gasoline. During the ethanol production process, distillation and molecular sieving is required to remove the water concentration to achieve high-purity ethanol [7]. However, hydrous ethanol could be beneficial to suppress knock of spark ignition (SI) engines to improve thermal efficiency. Further, the hygroscopic nature of ethanol can allow to increase the water content in gasoline - ethanol - water emulsions, even more, to improve the thermal efficiency while keeping the system complexity low. Many experiments of ethanol and gasoline - ethanol blends in SI engines have been conducted in the past to determine the advantages and disadvantages of its combustion. Nakata et al. [8] investigated ethanol and gasoline-ethanol blends in a SI engine. They showed an increase of engine torque and thermal efficiency with addition of ethanol because the spark timing could be advanced due to its higher knock suppression. Further, they concluded that the lower combustion temperature of ethanol reduced the heat losses and decreased the engine-out NO x emissions. Koc et al. [9] investigated the performance of gasoline, 50vol-% ethanol-in-gasoline (E50), and 85vol-% ethanol-in-gasoline (E85). They concluded that ethanol reduces the knock occurrence in the SI engine, wherefore torque was increased compared to gasoline. Further, the brake-specific fuel consumption (BSFC) is increased because of the lower heating value of ethanol. Turner et al. [10] investigated various blend ratios of bioethanol and gasoline in a direct injection engine. Their findings show a reduced combustion duration with higher ethanol content in the blend due to higher laminar flame velocities. 176 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="177"?> Further, combustion efficiency is increased for ethanol containing blends because of improved evaporation due to the reduced fraction of heavy-boiling components in the blend. El-Faroug et al. [11] analyzed recent publications on hydrous ethanol combustion in spark-ignition engines. They concluded that the thermal efficiency of SI engines can be improved by combustion of hydrous ethanol because of reduced knock tendency and increased burn speed compared to gasoline. At small water contents, they reported faster combustion due to improvement of chain reactions. In contrast, at a higher water content of up to 30%, the engine could be operated at maximum break torque (MBT) without knocking combustion, and thermal efficiency can be improved. Costa et al. [12] compared the combustion performance and emissions of gasoline - ethanol and gasoline - hydrous ethanol blends for various operating conditions. They found that a maximum thermal efficiency improvement of 14% points is achieved with hydrous ethanol. On the other hand, the fuel consumption was increased by 54% for hydrous ethanol blends. More insight of the combustion characteristics of different RON / MON fuels is obtained by multi-dimensional computational simulation. Yue et al. [13] applied three-dimensional (3D) computational fluid dynamics (CFD) to study the effect of fuel properties on the knock propensity in SI engines. They stated an increase of heat of vaporization (HoV) by 50% reduces the end-gas temperature before spark timing by 5K. When the laminar flame speed increases by 50%, they found that the knock limit spark advance (KLSA) must be reduced by 0.77 - 5.6°CA to avoid knocking combustion. Boldaji et al. [14] utilized 3D CFD to investigate the performance of wet ethanol in a Thermally Stratified Compression Ignition engine. The high HoV of wet ethanol was used to control the thermal stratification of the engine. With increasing water content, the combustion duration is increased, and heat release rate is reduced. Fagundez et al. [15] utilized a cylinder pressure analysis in conjunction with detailed chemistry simulation to investigate the effect of different ethanol - water blends on knock occurrence. The results indicate that the fuel mixture reactivity increases with increasing water content, while the laminar burning velocity is reduced. The cooling effect of the water addition outweight the increased reactivity, wherefore the risk for auto-ignition is reduced. The highest efficiency gains were observed for E70 and E80 blends with water addition since they could be operated at advanced spark timings. Anderson et al. [2] discussed the sensitivity of the RON / MON values towards the test conditions, and reported larger deviations for predicted RON / MON for ethanol enriched blends, which leaves the question of accuracy of RON / MON regarding evaluation of knock tendency of fuels. Bates et al. [16] utilized the Bradley detonation diagram to investigate the auto-ignition modes of different fuels, and determine if the hotspots are within harmless deflagration or harmful developing detonation regime. Their results indicated lower reaction front velocities for ethanol than toluene and Primary Reference Fuels (PRF) below 950K hotspot temperature, while at higher temperature above 950K the reaction front velocity became faster than PRF. The reason for this was found in longer ignition delay times for ethanol at low and intermediate 177 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="178"?> temperatures. Cho et al. [17] characterized the knock tendency of different ethanol reference fuels with constant RON at 100, while MON decreases with higher ethanol content in the blend. Similarly, their results indicated that the reaction front velocity is decreased with higher ethanol content due to the increased ignition delay times, wherefore blends with high ethanol content show a reduced knock tendency. Building up on the work of Netzer et al. [18], this work extends the research on auto-ignition properties of gasoline - ethanol - water mixtures by utilizing 3D CFD simulation and detailed chemistry. The gasoline - ethanol blends are defined such that RON is constant at 96.7, while MON decreases with increasing ethanol content. Finally, the auto-ignition modes of the predicted hotspots are characterized using the detonation diagram introduced by Bradley et al. [19,20]. The results provide insights into the effect of anhydrous ethanol and hydrous ethanol addition on the auto-ignition under engine-relevant conditions. 2 Methodology The investigation within this paper is based on 3D CFD simulation using the Rey‐ nolds-Averaged Navier Stokes Equations (RANS) in Converge v2.4. The simulation framework includes a spray model to describe the liquid phase penetration and vaporization of multi-component fuel mixtures. The combustion is modelled using the G-Equation model to predict turbulent flame propagation and the Well-Stirred-Reactor (WSR) model to predict auto-ignition in the unburnt mixture, and emission formation in the unburnt and burnt mixture. The low temperature and high temperature chemistry governing the combustion processes is based on the Ethanol-Toluene-Reference-Fuel (ETRF) reaction mechanism developed by Seidel [21]. The simulation methodology is introduced by Netzer et al. [18,22-24] to investigate the performance of different gasoline surrogates. The finite volume method is applied to discretize the computational domain using a cartesian grid with a base grid size of 4mm for x, y, z directions. The combustion chamber grid size is further refined to obtain a grid size of 1mm. During the direct fuel injection, the grid size around the spray plumes is refined to 0.25mm grid size to account for the higher velocity gradients. At spark timing and during the early flame propagation, the grid around the spark plug electrode is refined to 0.125mm grid size to resolve the flame propagation, which is governed by the turbulent flame speed. Additionally, adaptive grid refinement is used to refine the combustion chamber grid size to 0.5mm for regions with high velocity gradients, temperature, and CH 2 O concentration. 2.1 Spray Model The spray model contains several sub-models to describe the physical processes of liquid phase penetration, droplet break-up, droplet-wall interaction, and vaporization. The sub-models and respective literature sources are summarized in Table 1. 178 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="179"?> Model Description Evaporation Model Frossling with Boiling Model [25] Wall Film Model O’Rourke Model [26] Collision Model NTC Model [27] Initial Droplet Size Distribution Rosin-Rammler Distribution [28] Break-Up Model Kelvin-Helmholtz Model [29] Table 1: Converge v2.4 spray model setup The characteristics of the non-reactive sprays of gasoline (RON98) and water were investigated in the research project “Water Injection I + II” funded by the “For‐ schungsgemeinschaft für Verbrennungskraftmaschinen (FVV)” [30]. The results of the spray investigation are used to calibrate the Kelvin-Helmholtz Break-up model and Rosin-Rammler initial droplet size distribution function. The final model parameters are summarized in Table 2. Parameter Value Break-up Size Constant (B 0 ) 0.61 Break-up Time Constant (B 1 ) 5.0 Fraction of Parent Parcels (s) 0.1 Break-up Velocity Constant (C 1 ) 0.188 Rosin Rammler Constant (q rr ) 5.5 (gasoline), 1.5 (water) Table 2: Kelvin-Helmholtz break-up model and Rosin Rammler initial droplet size distribution function The surrogate composition in Table 3 represents the liquid thermophysical properties of the RON95E10 from the experiment. The liquid thermophysical properties of the individual species are taken from the NIST database [32]. I-C8H18 N-C7H16 C7H8 C2H5OH 0.402 0.128 0.365 0.105 Table 3: RON95E10 surrogate composition in mass fraction used for liquid physical properties calcula‐ tion The comparison of the measured and predicted liquid penetration length for gasoline and water are outlined in Figure 1. The experiments are conducted within a high-pres‐ sure spray chamber at 3.6bar chamber pressure and 303K chamber temperature. The 179 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="180"?> liquids are injected at a pump pressure of 150bar for 5000μs injection duration. During the initial phase between 0 and 1000μs, the gasoline fuel shows a higher penetration compared to water. After 1000μs the liquid water penetration surpasses gasoline, which shows a slower penetration rate in the late injection phase. Similar behavior is also reported by Aleiferis et al. [33], who attributed the slower penetration at the beginning of injection to the lower vapour pressure and potentially less cavitation in the nozzle. The higher density can explain the acceleration of the water spray around 1000μs and worse atomization, wherefore water can sustain its momentum longer. The parameters s and q rr in Table 2 are adjusted to fit the liquid penetration of the 3D CFD simulation to the experiments. Overall, the simulation predicts the early phase of the injection (up to 1000μs) closely, while in the late phase, it shows a tendency to overpredict the liquid water penetration and underpredict the liquid gasoline penetration. Figure 1: Spray penetration for RON95E10 and water injection with 150 bar pressure at 3.6bar spray chamber pressure and 303K spray chamber temperature The Sauter Mean Diameter (SMD) comparison in Figure 2 shows a difference between the experiments and 3D CFD simulations. However, it must be noted that this is just a qualitative comparison since the SMD of RON98 is shown for the experiments. Further, the injection pressures used for the investigations are different. Nevertheless, the comparison of gasoline and water shows that water tends to form larger SMD compared to gasoline due to worse droplet atomization [33]. 180 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="181"?> Figure 2: Sauter mean diameter for gasoline and water spray. For the experiments RON98 sauter mean diameter is measured, while for 3D CFD simulation the RON95E10 is shown The initial droplet size distributions are determined by the Rosin Rammler equation [28] using the parameter q rr in Table 2 (see Figure 3). The droplet size distributions mimic the droplet characteristics at the injector nozzle outlet, which are governed by in-nozzle flow and caviations effects [34]. The water injection shows a broader droplet size distribution, which is related to the higher liquid density compared to gasoline (E10). Kooij et al. [35] investigated the droplet size distrbutions of different ethanol-water mixtures. They stated that the droplet size distribution is mainly dependent on surface tension, density, and mass flow, and less on viscosity. The ethanol (E100) droplet size distribution is calculated to be close to gasoline and shows slightly larger droplet diameters because of the higher liquid density. This is inline with the observations of Aleiferis et al. [33], who reported a slightly larger SMD for ethanol sprays at 20°C temperature compared to gasoline. Figure 3: Initial droplet size distribution at the injector nozzle outlet, for E10, E100 and water sprays 181 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="182"?> The value of the q rr parameter for the ethanol spray is determined based on a linear correlation function in equation (1), which is based on the liquid density of the fluid: (1) q rr = �q rr,g -q rr,w ρ g -ρ w � ∙ρ+ �q rr,g - �q rr,g -q rr,w ρ g -ρ w � ∙ρ g � The correlation function utilizes the input parameters stated in Table 4. The function is used to determine the initial droplet size distributions for the gasoline - ethanol - water mixtures investigated in this work. Parameter Unit Value q rr,g - 5.5 q rr,w - 1.5 ρ g kg/ m³ 764 ρ w kg/ m³ 1000 Table 4: Rosin Rammler distribution function parameters. The subscript g and w represents gasoline and water respectively 2.2 Combustion Model The combustion model describes the initialization of the flame propagation, the turbulent flame propagation, and the chemistry in the unburnt and burnt mixture. The sub-models for the combustion are summarized in Table 5. Model Description Flame Propagation G-Equation with CEQ solver in the reaction zone [36,37] Laminar Flame Speed Look-up Table [24] Turbulent Flame Speed Peters Model [36] Unburnt / Burnt Zone SAGE with detailed chemistry [38] Reaction Mechanism 211 species and 2360 reactions [21] Spark Model Initialize Passive Scalar of G-Equation Model Table 5: Converge v2.4 combustion model setup The G-Equation model introduces the non-reactive scalar G, to distinguish the mixture into three parts: the unburnt mixture (G<0), the burnt mixture (G>0), and the flame front (G=0). A transport equation for the non-reactive scalar G and a transport equation 182 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="183"?> (2) for its variance , facilitate the propagation of the flame front. The transport equation includes a term for the turbulent flame propagation s T , which is calculated using the equation (2) by Peters [36]: S T = S L + u ′ − a 4 b 32 2b 1 Da + a 4 b 32 2b 1 Da 2 + a 4 b 32 Da 1/ 2 The laminar flame speed in equation (2) is provided by a subroutine in the user-defined function of the combustion model, which reads the data from a look-up table. The laminar flame speed look-up tables are calculated a-priori using the software LOGEresearch [39]. The modelling approach for the flame speed is used in previous work for 0D and 3D simulation methods by Netzer et al. [18] and Franken et al. [40]. The model parameters of the Peters turbulent flame propagation model are outlined in Table 6. The model parameter is adjusted in this work to fit the experimental heat release rate and cylinder pressure. Parameter Value Parameter (a 4 ) 0.78 Parameter (b 1 ) 3.0 Parameter (b 3 ) 1.0 Table 6: Turbulent flame propgation model by Peters [36] 3 Experiment The experiments are conducted on a single-cylinder research engine at the Technical University (TU) of Berlin. The single cylinder engine is built up for fundamental investigations of port and direct water injection in the FVV project “Water Injection in SI engines” [30]. The engine specifications are depicted in Table 7. Parameter Unit Value Bore mm 71.9 Stroke mm 82.0 Rod mm 137.0 Compression Ratio - 10.75 Number of Valves - 4 Table 7: Single-cylinder research engine specifications 183 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="184"?> The engine experiments are conducted using a centrally mounted direct fuel injection. More details on the experimental campaign can also be found in the work of Kauf et al. [41] and Gern et al. [42]. The investigations are performed for a RON95E10 gasoline fuel from the tank station. The engine is equipped with lowand high-pressure sensors to measure the cylinder and manifold pressures of 250 consecutive cycles. The simplified testbench scheme and positions of the pressure (p), temperature (T ), air-fuel ratio (λ), mass flow (m˙ ) sensors and emission analyzers are shown in Figure 4. The intake system consists of a compressor unit to set the boost pressure, a Sensyflow for air mass flow measurement, a throttle, a mixing tank, and two intake runners. The exhaust system consists of two exhaust runners: an exhaust orifice to control the back pressure, a diffuser, and exhaust ventilation. Figure 4: Single cylinder research engine testbench scheme from TU Berlin [30]. Engine-out emission concentration measurements used for model training and vali‐ dation entail carbon monoxide (CO) and carbon dioxide (CO 2 ) measurements using a nondispersive infrared system with an accuracy ±2% of the measurement range. Nitrogen oxide (NO x ) measurements were based on a chemiluminescence detector with an accuracy of ±2% of the measurement range with no separate nitric monoxide (NO) reporting. Unburnt hydrocarbons (uHC) emissions are measured on a wet basis by a flame ionization detector with an accuracy of ±1% of the measurement range in C 1 -equivalent. All other exhaust gas analyzers work on a dry basis [41]. 4 Fuel Definition The gasoline surrogates are composed of four components, and the software LOGEtable v2.1 [43] is used to calculate the mixture composition. The properties of the neat components are outlined in Table 8. 184 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="185"?> C8H18 C7H16 C7H8 C2H5OH H2O RON 100 0 120 109 - MON 100 0 109 91 - Density at 20°C [kg/ l] 0.69 0.684 0.867 0.789 1.0 LHV [MJ/ kg] 44.4 44.6 40.6 28.9 - Dynamic Viscosity at 20°C [cP] 0.46 0.38 0.54 1.04 0.85 HoV at 20°C [kJ/ kg] 307.0 364.0 412.0 919.0 2440.0 Table 8: Properties of the neat fuels [21,32] The n-heptane and iso-octane fractions are mixed to match the RON, and the toluene fraction represents the aromatic content of real gasoline. The Modified linear by Volume Method from Morgan et al. [44] is used to determine the TRF (Toluene Reference Fuel) mixture. The composition of the ethanol - TRF (Ethanol Toluene Reference Fuel - ETRF) mixture is calculated by the assumption of linear blending on a molar basis according to Anderson et al. [45]. The equations are implemented and validated in LOGEtable v2.1 based on the work of Seidel [21]. 4.1 Gasoline - Ethanol Mixtures The liquid volume fraction of ethanol is increased from 10vol-% to 70vol-% in the ETRF mixture (see Table 8). Figure 5: Liquid volume fraction of gasoline - ethanol mixtures. The constraint of the investigation is to keep the RON constant, while MON will decrease due to the higher n-heptane and ethanol content in the blend (see Table 9). Beyond E70, the RON could not be kept constant, wherefore no additional blends are investigated. The physical properties of the liquids (viscosity, density, heat of vaporization (HoV), thermal conductivity, and surface tension) are determined by a 185 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="186"?> linear blending on a mass basis, utilizing the mixture compositions in Figure 5. Hence, the mixture composition is the same in the liquid phase as in the gas phase. E10 E20 E30 E70 E100 RON 96.7 96.7 96.7 96.7 109 MON 87.4 85.6 84.2 80.2 91 Density at 20°C [kg/ l] 0.76 0.76 0.77 0.77 0.79 LHV [MJ/ kg] 41.5 40.0 38.4 32.3 28.9 Dynamic Viscosity at 20°C [cP] 0.54 0.60 0.65 0.87 1.04 HoV at 20°C [kJ/ kg] 417.0 476.0 535.0 765.0 919.0 C HO 6.3 11.7 0.21 5.4 10.6 0.38 4.7 9.6 0.5 2.8 7.3 0.85 2.0 6.0 1.0 Table 9: Properties of gasoline - ethanol mixtures (E10 = 10% liquid volume fraction of ethanol in ETRF mixture). 4.2 Gasoline - Ethanol - Water Mixtures The liquid volume fraction of water is determined based on the experimental investi‐ gation of gasoline - ethanol - water ternary blends by Letcher et al. [46], Kyriakides et al. [47], and Karaosmanoglu et al. [48]. According to the ternary mixture diagrams, water is mixed on a molar basis into the ETRF blends in Figure 5 to obtain a stable gasoline - ethanol - water emulsion. The TRF composition in the mole fraction for the respective E10, E20, E30, and E70 surrogates is kept the same, while the ethanol mole fraction is reduced to account for the added water in the mole fraction. The resulting blends are outlined in Figure 6. Figure 6: Liquid volume fraction of gasoline - ethanol - water mixtures. 186 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="187"?> The thermophysical properties of the liquids (viscosity, density, HoV, thermal conduc‐ tivity, and surface tension) are determined by a linear blending on a mass basis (see Table 10). The mixture composition is the same in the liquid phase as in the gas phase. E10H E20H E30H E70H E100H Ethanol Content [vol-%] 9.2 18.8 28.3 65.0 76.4 Water Content [vol-%] 0.3 0.5 0.8 3.0 23.6 Density at 20°C [kg/ l] 0.77 0.77 0.77 0.78 0.85 Dynamic Viscosity at 20°C [cP] 0.54 0.59 0.65 0.85 0.99 HoV at 20°C [kJ/ kg] 420.0 482.0 544.0 815.0 1350.0 Table 10: Properties of gasoline - ethanol - water mixtures (E10H = equivalent to E10 with water in the mixture) 5 Results and Discussion The simulation methodology is tested and validated for eight operating points with different engine speeds and torque measured at the single-cylinder engine testbench at TU Berlin. The experimental data was already used before to validate 0D combustion engine models for auto-ignition and emission prediction [49]. The measured cylinder pressure is processed by a thermodynamic analysis in LOGEengine v3.2 [50], to determine the apparent heat release rate, the cylinder temperature, and internally recirculated burnt gas fraction. Finally, the validated model is applied to investigate the effect of gasoline - ethanol - water mixtures on auto-ignition at 1500rpm and 15bar indicated mean effective pressure (IMEP). 5.1 Model Training and Validation The finally obtained model parameters after model training are outlined in chapter 2. One set of parameters is applied for the spray and combustion model to predict the experimental data. The laminar flame speed look-up table is generated for the RON95E10 base fuel (see Table 5) and applied to each operating point. Only the spark timing is delayed for each operating point individually to match the start of combustion. The 3D CFD simulation results of one consecutive run of the eight operating points are presented in Figure 7. The figure shows the cylinder pressure and normalized burn rate during the high-pressure phase of the engine cycle. The operating points at 2000rpm, 10bar and 20bar IMEP and 2500rpm, 15bar IMEP tend to underpredict the maximum normalized burn rate. For the operating points at 1500rpm - 2500rpm, 15bar IMEP, the cylinder pressure in the expansion stroke is overpredicted, which could be due to underpredicted heat transfer or too high fuel mass injected. Comparing the results to 187 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="188"?> the cycle-to-cycle variation of the engine, as shown by Gern et al. [42], to the deviations of the simulation, they are within the range of the cyclic variations. The operating point at 1500rpm, and 15bar IMEP show the strongest auto-ignition in the end gas, which is indicated by the second peak in the normalized burn rate in the expansion stroke at 25°CA aTDC. The pressure rise is induced by the strong heat release in the unburnt mixture. This operating point is evaluated experimentally to be at the knock limit. The 3D CFD simulation can predict the occurrence of strong auto-ignition in the unburnt gas and evaluates the auto-ignition to be within the transition regime according to the Diagram [23,51]. Figure 7: Cylinder pressure and normalized burn rate for RON95E10 and different operating conditions 188 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="189"?> The engine-out emissions are shown in Figure 8 for the eight operating points. No additional model training is performed to improve the engine-out emission prediction. The simulation predicts generally lower CO 2 emissions but can capture the trend of the experiments. The differences in the CO 2 emissions are within ±5%, which is higher than the accuracy of the measurement device (see section 3). However, the summation of slight differences in fuel C: H: O ratio, fresh gas composition, and exhaust gas composition can lead to differences in the CO 2 prediction. The simulation well captures the trend of the NOx emissions, but the absolute concentrations are overpredicted by 20% to 70%. The source for deviation could be related to variations in the air-fuel ratio, which impact the formation of emissions, as it is discussed by Esposito et al. [52]. The predicted CO emissions are generally higher than the experiments, besides operating point 1 at 1500rpm, 15bar IMEP, and show a deviation of 20% to 80%. Regarding the unburnt hydrocarbon (uHC) emissions, the simulation cannot predict the experimental levels correctly. Including the amount of uHC that can result from the unburnt fuel in the wall film, the simulated uHC increases, but the experimental levels are still underpredicted. uHC emissions from crevices can probably contribute to the high levels observed in the experiments [52], which are underpredicted by the simulation. Further investigations are needed to improve the uHC emissions. Figure 8: Engine-out emissions for CO 2 , NO x , uHC and CO for RON95E10 and different operating conditions 189 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="190"?> 5.2 Gasoline - Ethanol Mixtures The predicted mean cylinder pressure for the different gasoline - ethanol mixtures from Table 9 is shown in Figure 9. The E70 and E100 surrogates show the largest increase in the maximum cylinder pressure, and the location of the maximum cylinder pressure is advanced. The E20 and E30 surrogates show only a slight increase in the maximum cylinder pressure compared to the E10 surrogate. Figure 9: Mean cylinder pressure for different gasoline - ethanol mixtures The results indicate advanced combustion and shorter burn duration with increasing ethanol content in the surrogate. These findings are confirmed by the apparent heat release rate shown in Figure 10. The E20 and E30 surrogates show a higher maximum heat release rate than E10. For the E70 and E100 surrogates, the combustion is further advanced, achieving a higher maximum heat release rate and shorter burn duration than E10, E20 and E30 surrogates. However, the E100 surrogate shows a lower maximum heat release rate than the E70 surrogate. The peak in heat release rate at 25°CA aTDC for the E10 surrogate due to auto-ig‐ nition of the unburnt mixture is significantly reduced with increasing ethanol content in the surrogate. 190 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="191"?> Figure 10: Mean apparent heat release rate for different gasoline - ethanol mixtures The faster combustion with increasing ethanol content can be explained by the higher flame speed, which is outlined in Figure 11. The laminar flame speed calculations are conducted using LOGEresearch [39] software package for the surrogates defined in Table 9. The pressure is set to 30bar and the temperature to 600K, at equivalence ratios between 0.5 to 1.5. The reaction mechanism of Seidel [21] is used for flame calculations. The E70 and E100 surrogates show the highest flame speeds compared to E30, E20 and E10 at lean, stoichiometric, and especially rich conditions. The reaction mechanism predictions are in line with the findings reported in the literature [53, 54] for blends of ethanol and hydrocarbon fuels. Figure 11: Laminar flame speed calculated at 30 bar and 600 K for different gasoline - ethanol mixtures Though E70 and E100 surrogates show similar flame speeds, the maximum apparent heat release rate of E100 is lower compared to E70 (see Figure 10). The analysis of the mixture formation of the different surrogates reveals a more inhomogeneous mixture for the E100 surrogate shortly before spark ignition. This observation is outlined in Figure 12 by comparing the equivalence ratio probability density function (pdf) for the different surrogates at -20°CA aTDC. The E20, E30, and E70 surrogates show the 191 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="192"?> maximum of the pdf at slight lean mixtures in comparison to E10, which also results in an overall lean mixture (see Figure 16). The E100 surrogate shows a larger variance of the pdf, with more lean and rich regions, while overall, the mixture is slightly rich as shown in Figure 16. During combustion, the lean regions inhibit local flame propagation for the E100 surrogate, wherefore, the maximum apparent heat release rate is reduced compared to E70. Figure 12: Equivalence ratio probability density function at -20 °CA aTDC, for different gasoline - ethanol mixtures The worse mixture formation of the E100 surrogate can also be attributed to the injection rate applied to the 3D CFD simulation. The different surrogates are injected at the same injection pressure as shown in Figure 13. The injection rate is determined experimentally for gasoline fuel (RON95E10) and is applied for all the different surrogates. The injection mass increases for surrogates with high ethanol content to account for the different heating values (see Table 9). By keeping the injection pressure constant, the injection duration is increased for surrogates with higher ethanol content. Figure 13: Injection pressure for different gasoline - ethanol mixtures 192 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="193"?> The increased injection mass for surrogates with higher ethanol content leads to a later end of injection, as shown in Figure 14. Since the spark timing is kept constant, the time for mixture formation is shortened for the E100 surrogate. Further, the higher HoV for surrogates with high ethanol content (see Table 9) increases the amount of energy extracted from the surrounding cylinder gas. The gas temperature is reduced as observed in Figure 15. During injection, this effect can also prolong the vaporization of surrogates with high ethanol content, and it can inhibit the mixture formation. This can be seen for the E100 surrogate, which shows a higher mean cylinder temperature unexpectedly compared to the E70 surrogate in Figure 15. Figure 14: Injection mass for different gasoline - ethanol mixtures Figure 15: Mean cylinder temperature for different gasoline - ethanol mixtures The higher injection mass for increasing ethanol content results in a higher specific fuel consumption as shown in Figure 16. For the E70 surrogate, the fuel consumption is increased by 24%, while for E100 surrogate, the fuel consumption is increased by 52%. However, the faster combustion with higher ethanol content led to an increase of the indicated mean effective pressure (IMEP) and indicated efficiency (η ieff ) for the E20, E30, and E70 surrogates compared to E10. The highest indicated efficiency is obtained 193 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="194"?> for the E70 surrogate, which is 1.5% points higher compared to E10. The IMEP and indicated efficiency of the E100 surrogate is on the same level as the E10 surrogate, because of the combustion inhibition due to inhomogeneous mixture. Figure 16: IMEP, equivalence ratio, ISFC and indicated efficiency for different gasoline - ethanol mixtures The lower carbon content and slightly lean mixtures of the surrogates with higher ethanol content (see Table 9) leads to a lower CO 2 mass fraction in the cylinder during combustion, as it is shown in Figure 17. The higher oxygen content with a higher ethanol liquid volume fraction in the surrogate reduces the CO mass fraction for the E20, E30, and E70 surrogates, as is also shown by Iodice et al. [55]. However, for the E100 surrogate the CO mass fraction is increased because of the slightly rich mixture and slower CO oxidation rate during the expansion stroke. The maximum OH mass fraction is increased for E20, E30 and E70 surrogates, which increases the reactivity of the air - fuel mixture and promotes faster burn rates. For the E100 surrogate, the maximum OH mass fraction is reduced because of the worse mixture formation and inhibited combustion. The CH 2 O mass fraction is reduced for the surrogates with higher ethanol content, which indicates a reduced reactivity of the unburnt mixture and less tendency for auto-ignition, as was also seen in the apparent heat release rates in Figure 10. The maximum NO mass fraction shows a trade-off, where the E20 and E30 surrogates predict higher values compared to E70 and E100. For the E100 surrogate, the NO mass fraction is the lowest due to the reduced maximum heat release rate and lower adiabatic flame temperature of neat ethanol. The NO oxidation slows down during the expansion stroke, and the E70 and E100 surrogates show the lowest NO oxidation rates. The calculated NO 2 mass fraction is considerably lower compared to NO. However, with increasing ethanol content in the surrogate, the NO 2 mass fraction during combustion 194 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="195"?> increases. After a short oxidation period during the expansion stroke, the formation of NO 2 continues to increase towards the end. Figure 17: Crank-angle resolved engine-out emissions for different gasoline - ethanol mixtures The increased maximum NO mass fraction for the E20, E30 and E70 surrogates can be explained by the higher mean cylinder temperatures between 10°CA to 20°CA aTDC, as it can be seen in Figure 18. The higher temperatures are the result of advanced combustion. During the expansion stroke, the NO oxidation rate is reduced, especially 195 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="196"?> for surrogates with high ethanol content, wherefore, the NO mass fraction remains at a higher value. Figure 18: Mean cylinder temperature for different gasoline - ethanol mixtures The tendency for auto-ignition is reduced for surrogates with higher ethanol content, as indicated by the CH 2 O mass fraction in Figure 17, even though the RON is kept constant and the MON is decreasing (see Table 9). This trend is analyzed more in detail by visualizing the local CH 2 O, OH mass fraction, and temperature during the combustion in Figure 19. The CH 2 O mass fraction is shown at the time step before auto-ignition of the unburnt mixture, while the OH mass fraction and temperature are shown at the time step when auto-ignition in the end gas occurs. The higher formation of CH 2 O in the unburnt mixture indicates for active low-temperature chemistry, especially for the E10, E20, and E30 surrogates. The highest local CH 2 O mass fractions for the E10, E20, and E30 surrogates are found close to the exhaust valves, while no significant CH 2 O mass fraction can be observed for the E70 and E100 surrogates. The reason is the reduced reactivity for surrogates with high ethanol content, and the fundamental aspects are described by Cheng et al. [56] and Sarathy et al. [57]. At auto-ignition, the OH mass fraction and the temperature in the unburnt mixture increase, and a reaction front propagation is initiated. 196 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="197"?> Figure 19: Local CH 2 O and OH mass fraction, and temperature for different gasoline - ethanol mixtures The rapid auto-ignition of the unburnt mixture introduces high pressure gradients and velocity magnitudes for the E10 surrogate, while only minor changes are observed for the E20 surrogate (see Figure 20). Though low-temperature chemistry is kicked in for the E30 surrogate (see Figure 19), it does not result in a significant increase in pressure gradients and velocity due to auto-ignition within the cylinder 197 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="198"?> Figure 20: Local normalized squared pressure ratio and velocity magnitude for different gasoline - ethanol mixtures. 198 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="199"?> (3) (4) (5) The gas acoustic and reactivity of the hotspots in the unburnt mixture are evaluated using the Bradley detonation diagram [19,20]. The Bradley detonation diagram was introduced by Netzer et al. [24] for knock prediction in RANS simulation, and further validated by LES simulations [22]. Bates et al. [58] defined different auto-ignition regimes within the Bradley detonation diagram, that can occur during engine com‐ bustion. The method introduces the dimensionless resonance parameter , which is determined as the ratio of the sound of speed a of the acoustic wave introduced by the auto-ignition, and the reaction front velocity u a : ξ = a u a The reaction front velocity is calculated by post-processing the growth rate of hotspot radius along the highest velocity gradient Δr (see Figure 19 and Figure 20) in relation to a characteristic timescale of the auto-ignition [24]. The timescale is determined based on the change of ignition delay time Δτ I DT at the local conditions of the hotspot: u a = Δr Δτ I DT Additionally, the reactivity of the hotspot is evaluated by the dimensionless reactivity parameter ε in equation (5), which is the ratio of the acoustic timescale and an excitation time τ e . The acoustic time scale describes the time needed by the acoustic wave to travel through the hotspot of size r 0 with the speed of sound a. The excitation time describes how fast the hotspot can release energy to drive the acoustic wave. ε = r 0 a ⋅ τ e For the E10, E20 and E30 the hotspots are evaluated to be within the subsonic regime, which is characterized by s L < u a < a, where s L corresponds to the laminar flame speed of the mixture. The E20, and E30 surrogates show a slower reaction front velocity than E10. The reactivity parameter is within the same range for all three surrogates. Pan et al. [59] showed similar results for E10, E20 and E30 fuels at 40atm pressure. 199 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="200"?> Figure 21: Bradley detonation diagram for different gasoline - ethanol mixtures. ξ u and ξ l are determined for syngas mixture by Bradley et al. [19] The auto-ignition characteristics of the different surrogates are further studied within a homogeneous 0D reactor. The initial pressure and temperature for the reactor are extracted from the 3D CFD simulation for the hotspot of the E10 surrogate, and the mixture composition corresponds to the unburnt mixture composition of the respective surrogate. The homogeneous reactor simulation results in Figure 22 show a delayed ignition with higher ethanol content in the surrogate, which is indicated by the later rise of CH 2 O and OH mass fraction and heat release and temperature. Hence, the reduced auto-ignition tendency for surrogates with high ethanol content is not only attributed to HoV (see Table 9) and subsequently reduced gas temperature (see Figure 15) but also due to slower chemistry. As discussed by Cheng et al. [56], higher ethanol content in the gasoline blend decreases the mixture reactivity at low temperatures, while it can increase reactivity at high temperatures. They concluded that ethanol consumes O 2 and OH radicals and forms more stable HO 2 and aldehydes. Further, they showed that the increased reactivity with ethanol addition is moved to higher temperatures with increasing pressure. 200 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="201"?> Figure 22: Constant volume reactor simulation results at 70bar pressure, 900K temperature and =1.0 for different gasoline - ethanol mixtures 5.3 Gasoline - Ethanol Mixtures at Knock Limit To evaluate the potential of the reduced reactivity of surrogates with high ethanol content to improve thermal efficiency, the combustion is advanced for E20, E30, E70 and E100. For the E20, E30, and E70 surrogates, the spark timing was advanced by 2°CA, and for the E100 surrogate, the spark timing was advanced by 6°CA without observing auto-ignition within the developing detonation regime of the ξ − ε diagram (see Table 11). Beyond 6°CA spark timing advance, auto-ignition is observed for the E100 surrogate, which is explained by the higher local temperatures in the unburnt mixture and following the increased reactivity of ethanol [56]. E10 E20 E30 E70 E100 ΔST [°CA] 0.0 -2.0 -2.0 -2.0 -6.0 Table 11: Spark timing adjustment for gasoline - ethanol mixtures 201 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="202"?> The cylinder pressure of the different surrogates is depicted in Figure 23. The maximum cylinder pressure increases, while the highest maximum cylinder pressure is observed for the E100 surrogate due to the more advanced combustion. Figure 23: Cylinder pressure for different gasoline - ethanol mixtures and at KLSA. * Spark timing advanced by 2°CA. *** Spark timing advanced by 6°CA Figure 24: Apparent heat release rate for different gasoline - ethanol mixtures and at KLSA. * Spark timing advanced by 2°CA. *** Spark timing advanced by 6°CA The corresponding apparent heat release rate in Figure 24 shows the advanced combustion for the E20, E30, E70, and E100 surrogates. Similarly to the base spark timing the E100 surrogate does not achieve a higher maximum heat release rate than 202 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="203"?> the E70 surrogate. The heat release rate peak in the expansion stroke at 20°CA to 30°CA aTDC is more pronounced for the E20 and E30 surrogates compared to the base spark timing. No distinct heat release rate peak in the expansion stroke can be observed for the E70 and E100 surrogates. The advanced combustion phasing of the E20, E30, E70, and E100 surrogates leads to a better transfer of the chemical energy into piston work, which is indicated by a higher IMEP and η ieff as shown in Figure 25. The cases with advanced spark timing are marked as knock limit spark advance (KLSA) within the figure. Since the injected fuel mass is not adjusted, the equivalence ratio is not changed, and the ISFC is reduced since the power output is increased. Figure 25: IMEP, equivalence ratio, ISFC and indicated efficiency for different gasoline - ethanol mixtures at KLSA The advanced spark timing for the E20, E30, E70, and E100 surrogates promotes the auto-ignition of the unburnt mixture due to higher temperature and pressure. As it can be seen in Figure 26, the formation of CH 2 O is increased for all surrogates, but different characteristics are evident. While the E10, E20 and E30 surrogates show a formation of CH 2 O in the unburnt mixture along the cylinder walls due to a more active low-temperature chemistry, however E70 and E100 surrogates show two distinct local regions within the cylinder where more CH 2 O is formed. For E70 and E100 surrogates, the local higher temperature at those two points kicks in the high-temperature chemistry, wherefore auto-ignition of the unburnt mixture occurs. After the unburnt mixture ignites, significantly more mass is burned for the E10, E20, and E30 surrogates, as seen in the larger area of OH mass fraction in Figure 26. 203 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="204"?> Figure 26: Local CH 2 O and OH mass fraction, and temperature for different gasoline - ethanol mixtures at KLSA Figure 27 shows that the stronger auto-ignition for the E20 and E30 surrogates induces higher pressure gradients and velocity magnitudes. However, the time duration in which the auto-ignited mixture burns is longer for the E20 and E30 surrogates compared to E10, as can be seen in Figure 26. The crank angle numbers in the figure indicate the overall time step before auto-ignition and overall time step after auto-ignition. Hence, the velocity magnitudes decrease with higher ethanol content since the hotspots are expanding slower. Further, the E70 and E100 surrogates show no distinct pressure gradient and velocity magnitudes due to auto-ignition. As shown in Figure 26 the hotspots are much smaller, which indicates less mass to be burnt and less energy is released that can be transferred into the increase of pressure and gas velocity. 204 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="205"?> Figure 27: Local normalized squared pressure ratio and velocity magnitude for different gasoline - ethanol mixtures at KLSA The Bradley detonation diagram in Figure 28 shows the hotspots for the E10, E20, E30, E70, and E100 surrogates within the subsonic and transition regime. Further, multiple hotspots occur for the E20, E30, and E100 surrogates, where the subsequent hotspot is induced at the respective opposite site of the combustion chamber. In Figure 26, it is seen that the CH 2 O formation differs for surrogates with high ethanol content, 205 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="206"?> and a difference can also be observed for these hotspots in the ξ-ɛ diagram. While the hotspots for the E10, E20 and E30 surrogates are found within the subsonic regime (s L < u a < a, ɛ < 8), the hotspots for E70 and E100 are within the transition regime at high resonance parameters and high reactivity parameters (ξ > ξ u , ɛ > 8). The span of the transition regime above the detonation peninsula is shown by Bates et al. [58] and Lechner et al. [60] for different hydrocarbon fuels. These auto-ignitions show a relatively slow reaction front propagation velocity u a , while the excitation time τ e is decreased compared to surrogates with less ethanol content (see Figure 29). The second hotspot for E100 at ɛ ≈ 25 was actually evaluated at much higher reactivity parameters, but the ξ-ɛ diagram is just shown until ɛ=25 because validation, results beyond that value were not reported (see Bates et al. [58] and Lechner et al. [60]). Figure 28: Bradley detonation diagram for different gasoline - ethanol mixtures at KLSA. ξ u and ξ l are determined for syngas mixture by Bradley et al. [19] Figure 29: Excitation time (τ e ) and reaction fron velocity (u a ) for different gasoline - ethanol mixtures at KLSA 206 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="207"?> 5.4 Gasoline - Ethanol - Water Mixtures The injected mass of the gasoline - ethanol - water mixtures from Table 10 is increased to keep the energy of the injected fuels constant (Figure 30). Since the injection pressure is kept the same as for the E10, E20, E30, E70, and E100 surrogates, it is required to extend the injection duration (see Figure 31). Figure 30: Injection mass for different gasoline - ethanol - water mixtures Figure 31: Injection pressure for different gasoline - ethanol - water mixtures The 3D CFD simulation results for the cylinder pressure are shown in Figure 32 for the different gasoline - ethanol - water mixtures at KLSA. The addition of hydrous ethanol shows an increased maximum cylinder pressure for the E10H, E20H, E30H, and E70H surrogates. The E100H shows a considerably decreased maximum cylinder pressure compared to the E100 surrogate. 207 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="208"?> Figure 32: Cylinder pressure for different gasoline - ethanol - water mixtures at KLSA The apparent rate of heat release in Figure 33 shows that the combustion is faster with the addition of hydrous ethanol for the E10H, E20H, and E30H surrogates, while for the E70H surrogate, the maximum apparent heat release rate is increased. Further, the heat release rate peak in the expansion stroke due to auto-ignition of the unburnt mixture becomes more pronounced for the E10H, E20H, and E30H surrogates. The E100H surrogate shows a clear delay in the apparent heat release rate compared to the E100 surrogate. Figure 33: Apparent heat release rate for different gasoline - ethanol - water mixtures at KLSA The addition of hydrous ethanol reduces the laminar flame speed at 30bar pressure and 600K temperature for an equivalence ratio between 0.5 to 1.5 (see Figure 34). Thereby, the one-dimensional flame simulation neglects the effect of vaporization and only considers changes of gas phase thermodynamics, third-body efficiency and chemistry. The simulation results indicate that the most substantial reduction in laminar flame speed is found for the surrogates with the highest water content. As it is discussed by van Treek et al. [61], water addition reduces the laminar flame speed of ethanol/ air mixtures by participating in termolecular and bimolecular reactions and altering the thermodynamic properties of the mixture. 208 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="209"?> Figure 34: Laminar flame speed calculated at 30 bar and 600 K for different gasoline - ethanol - water mixtures The advanced combustion due to hydrous ethanol addition cannot be explained by changes in the flame speed but from changes in the local mixture formation. The probability density function of the equivalence ratio in Figure 34 indicates an increase of the maximum value at ϕ=1.0 to 1.1 for the E10H, E20H, E30H and E70H surrogates at -20°CA aTDC. The shift to a slightly rich mixture enhances the local flame propagation, wherefore combustion is advanced. Figure 35: Equivalence ratio distribution at -20 °CA aTDC, for different gasoline - ethanol - water mixtures However, since significantly more water mass is injected and vaporized for the E100H surrogate compared to the other surrogates (see Figure 36), the reduced cylinder temperature effect (see Figure 37) due to water vaporization and slower laminar flame speed (see Figure 34) delays the combustion significantly. 209 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="210"?> Figure 36: Probability density function of vaporized water mass fraction at -20°CA aTDC, for different gasoline - ethanol - water mixtures at KLSA Figure 37: Mean cylinder temperature for different gasoline - ethanol - water mixtures Because of the advanced combustion for the E10H, E20H, E30H and E70H surrogates, the IMEP is increased with hydrous ethanol addition. The E100H surrogate shows no significant change compared to the E100 surrogate. The indicated efficiency increases between 0.2% to 0.5% points for the E10H, E20H, E30H, and E70H surrogates compared to the surrogates without the water content. 210 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="211"?> Figure 38: IMEP, equivalence ratio, ISFC and indicated efficiency for different gasoline - ethanol - water mixtures at KLSA The enhanced auto-ignition due to hydrous ethanol addition is investigated more in detail by the local CH 2 O, OH mass fraction, and temperature in the cylinder in Figure 39. The CH 2 O mass fraction in the unburnt mixture is increased for the E10H, E20H, E30H and E70H surrogates at earlier crank angles, wherefore auto-ignition occurs 0.5°CA to 2.0°CA earlier compared to the surrogates without hydrous ethanol addition (see Figure 36). For the E100H surrogate, no auto-ignition can be observed due to the reduced gas temperature because of water vaporization, which is indicated by the lower local temperatures in Figure 38 compared to the E100 surrogate in Figure 26. Similar to the E70H surrogate, small hotspots close to the intake valves can be found for the E10H, E20H, and E30H surrogates. 211 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="212"?> Figure 39: Local CH 2 O and OH mass fraction, and temperature for different gasoline - ethanol - water mixtures at KLSA The auto-ignition for E10H, E20H, E30H, and E70H surrogates induce higher pressure gradients, as shown in Figure 40, compared to the surrogates without hydrous ethanol addition (see Figure 27). At the same time, the velocity magnitudes of the E10H, E20H, and E30H surrogates are not significantly increased. Only for the E70H surrogate, a slight increase in the velocity magnitudes is observed close to the intake valves. 212 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="213"?> Figure 40: Local normalized squared pressure ratio and velocity magnitude for different gasoline - ethanol - water mixtures at KLSA The evaluation of the local hotspots by the Bradley detonation diagram shows the increasing reactivity of the E10H, E20H, E30H, and E70H by hydrous ethanol addition. The excitation time is significantly reduced for some of the hotspots of the different surrogates (see Figure 42), wherefore the reactivity parameter is increased up to ɛ ≈ 25. The reactivity parameters are predicted to be even higher, but since validation data is missing for the Bradley detonation diagram beyond ɛ=25, the plot was limited to this value. The hotspots for the surrogates with a higher ethanol content show higher 213 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="214"?> resonance parameters, wherefore they are close to the upper limit of the detonation peninsula. Besides the increased reactivity of the hotspots, multiple hotspots occur, as can be seen in Figure 39. The additional hotspots for the E10H, E20H, and E30H surrogates are induced close to the intake valves by the pressure gradient from the preceeded hotspot close to the exhaust valves. The amount of mass burnt by those hotspots is lower compared to the hotspot relative to the exhaust valves, and they are within the subsonic (s L < u a < a, ɛ < 8) or transition regime ( ξ l <ξ<ξ u , ε>8 ) of the ξ − ε diagram. For the E70H surrogate, the first hotspot occurs close to the intake valves, while the second hotspot appears close to the exhaust valves. Both hotspots can be found within the developing detonation regime ( ξ l <ξ<ξ u , ε>8 ) . Figure 41: Detonation diagram for different gasoline - ethanol - water mixtures at KLSA. ξ u and ξ l are determined for syngas mixture by Bradley et al. [19]. The full symbols indicate the surrogates without water. The hollow symbols or crosses indicate the surrogates with water Figure 42: Excitation time (τ e ) and reaction front velocity (u) for different gasoline - ethanol - water mixtures at KLSA 214 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="215"?> The effect of hydrous ethanol addition is further investigated within a homogeneous 0D reactor at 70bar pressure, 900K temperature and composition of the unburnt mixture, which corresponds to the local conditions of the hotspot in the engine. The addition of water vapor promotes the formation of CH 2 O and OH and enhances the auto-ignition of the mixture as shown in Figure 43. Gong et al. [62] investigated the thermal, third-body efficiency and chemical effects of water vapor addition for methane/ n-heptane mixtures. For high-pressure and low-temperature conditions, they concluded that water acts as an efficient collision partner in the reactions H 2 O 2 (+M) = 2OH (+M) and H + O 2 (+M) = HO 2 (+M) and other third-body reactions, which enhance the system reactivity. The thermal effect of water vapor addition has a delaying effect on the ignition delay time. At higher temperatures (T=1200K), the chemical effect of water vapor enhances the OH formation via the reactions H 2 + OH <-> H + H 2 O and O + H 2 O <-> 2OH. On the other hand, Netzer et al. [63] investigated the thermodynamic and chemical effect of water injection on the suppression of engine knock. Thereby, they showed the impact of different water - gasoline ratios on the ignition delay time between 700K to 1000K at 70bar pressure. With increasing water vapor content, the ignition delay time increases at low and intermediate temperatures, while at high temperatures (T>950K) no significant change of ignition delay time was observed. Figure 43: Homogeneous constant volume reactor simulation results at 70bar pressure, 900K temper‐ ature and ϕ = 1.0 for different gasoline - ethanol - water mixtures 215 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="216"?> The discrepancies in the effect of water addition can be explained by how the gasoline - ethanol - water mixtures are prepared. The ignition delay times of three different mixtures for the E70 surrogate are computed by performing homogeneous 0D reactor simulations at 70bar pressure and ϕ=1.0. The mixtures are outlined in Table 12. The E70 and E70H mixtures are the same as the ones outlined in Table 9 and Table 10, while the third mixture, “E70 + 3vol-% Water” is prepared according to the method of Netzer et al. [63]. The water content in mole fraction is kept constant. For the third mixture the water mole fraction is mixed in the total composition, wherefor all other species mole fractions are decreased. For the E70H surrogate, only the ethanol mole fraction is changed by water addition. The results in Figure 44 show the increased reactivity of the E70H surrogate at low and intermediate temperatures. In contrast, the “E70 + 3vol-% Water” mixture shows no difference compared to the E70 surrogate, which corresponds to the observations from Netzer et al. [63]. C8H18 C7H16 C7H8 C2H5OH H2O E70 5.1% 11.3% 3.6% 84.6% 0% E70H 5.1% 11.3% 3.6% 73.6% 11% E70 + 3vol% Wa‐ ter 4.6% 10.0% 3.2% 75.3% 11% Table 12: Species mole fraction for the E70, E70H and E70 + 3vol-% water surrogates. Figure 44: Homogeneous constant volume reactor simulation results of ignition delay time for different temperatures at 70bar pressure and ϕ = 1.0 for E70, E70H and E70 + 3vol-% water surrogates 216 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="217"?> 6 Conclusions This study investigates the influence of different gasoline - ethanol - water surrogates on the occurrence of auto-ignition within a single-cylinder spark ignition engine. The gasoline - ethanol mixtures are defined so that the RON is constant while the MON decreases with increasing ethanol content. Water is mixed to the surrogate according to the maximum possible mole fraction at which a stable gasoline - ethanol - water emulsion can be formed. The investigation is conducted by using a 3D CFD RANS simulation with detailed chemistry. The model incorporates a lagrangian multi-component spray model to predict the liquid droplet penetration, break-up, and vaporization of the gasoline - ethanol - water surrogates. The flame propagation is calculated using the G-Equation model with pre-compiled laminar flame speed look-up tables for the surrogates. The auto-ignition of the unburnt mixture is predicted by using a Well-Stirred Reactor model with a detailed ETRF reaction mechanism. The conclusions can be summarized as follows: 1. Increasing the ethanol content from 10vol-% to 100vol-% increases the HoV of the surrogate by 120% and reduces the end compression temperature by up to 22K. The laminar flame speed is increased, and hence the combustion is advanced for increasing ethanol content in the surrogate. 2. Increasing the ethanol content in the surrogate reduces the tendency for auto-ig‐ nition of the unburnt mixture because of lower gas temperature and reduced reactivity at low and intermediate temperatures. The strongest auto-ignition at base spark timing is found for the E10 surrogate, followed by E20 and E30. No auto-ignition of the unburnt mixture is found for the E70 and E100 surrogates. 3. The spark timing of the E20, E30, and E70 surrogates can be advanced by 2°CA, and the spark timing of the E100 surrogate can be advanced by 6°CA without observing a strong auto-ignition of the unburnt mixture. As a result, the indicated efficiency can be improved by 1-2% points for the E20, E30, E70, and E100 surrogates compared to E10. The highest indicated efficiency is obtained for the E70 surrogate. 4. The characteristic of auto-ignition is different for the investigated surrogates: a. The auto-ignitions for the E10, E20, and E30 surrogates are found in the subsonic regime of the Bradley detonation diagram, with active low-temper‐ ature chemistry highlighted by the higher CH 2 O mass fraction in the unburnt mixture. b. The auto-ignitions for the E70 and E100 surrogates are found in the transition regime of the Bradley detonation diagram, with lower excitation times. 5. The E10, E20, E30, and E70 surrogates show a reduced auto-ignition tendency, wherefore they are not following the trend of the linear molar blending rule for RON and MON, which determines a constant or increased auto-ignition tendency for increasing ethanol content in the surrogate. 6. Increasing the hydrous ethanol content increases the surrogate's HoV, and reduces the end compression temperature by up to 70K. The laminar flame speed is 217 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="218"?> decreased; however, the combustion is advanced for E10H, E20H, E30H, and E70H surrogates, which is attributed to the improved mixture formation. 7. Increasing the hydrous ethanol content in the surrogate increases the tendency for auto-ignition of the unburnt mixture of the E10H, E20H, E30H, and E70H surrogates because of increased reactivity at low and intermediate temperatures. The strongest auto-ignition at advanced spark timing is found for the E10H, E20H, and E30H surrogates, followed by the E70H surrogate. No auto-ignition is found for the E100H surrogate. 8. The characteristic of auto-ignition is different for increased hydrous ethanol content in the surrogate, showing a strongly reduced excitation time, while the reaction front velocity is not significantly affected. Within the Bradley detonation diagram, those hotspots can be found within the developing detonation ξ < ξ u and transition regime ξ > ξ u at high values. 7 References [1] E. Commission, European Green Deal, (n.d.). https: / / ec.europa.eu/ info/ strategy/ priorities-2 019-2024/ european-green-deal_de. [2] J.E. Anderson, T.G. Leone, M.H. Shelby, T.J. Wallington, J.J. Bizub, M. Foster, M.G. Lynskey, D. Polovina, Octane numbers of ethanol-gasoline blends: measurements and novel estimation method from molar composition, 2012. https: / / doi.org/ https: / / doi.org/ 10.4271/ 2012-01-1274. [3] D.F. Correa, H.L. Beyer, J.E. Fargione, J.D. Hill, H.P. Possingham, S.R. Thomas-Hall, P.M. Schenk, Towards the implementation of sustainable biofuel production systems, Renew. Sustain. Energy Rev. 107 (2019) 250-263. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.rser.2019.0 3.005. [4] H. Zabed, J.N. Sahu, A. Suely, A.N. Boyce, G. Faruq, Bioethanol production from renewable sources: Current perspectives and technological progress, Renew. Sustain. Energy Rev. 71 (2017) 475-501. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.rser.2016.12.076. [5] T.J. Lark, N.P. Hendricks, A. Smith, N. Pates, S. A. Spawn-Lee, M. Bougie, E.G. Booth, C.J. Kucharik, H.K. Gibbs, Environmental outcomes of the US Renewable Fuel Standard, Proc. Natl. Acad. Sci. 119 (2022) e2101084119. https: / / doi.org/ https: / / doi.org/ 10.1073/ pnas.2101084 119. [6] J. Lewandrowski, J. Rosenfeld, D. Pape, T. Hendrickson, K. Jaglo, K. Moffroid, The greenhouse gas benefits of corn ethanol--assessing recent evidence, Biofuels. (2019). https: / / doi.org/ http s: / / doi.org/ 10.1080/ 17597269.2018.1546488. [7] H.A. Saffy, W.F. Northrop, D.B. Kittelson, A.M. Boies, Energy, carbon dioxide and water use implications of hydrous ethanol production, Energy Convers. Manag. 105 (2015) 900-907. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.enconman.2015.08.039. [8] K. Nakata, S. Utsumi, A. Ota, K. Kawatake, T. Kawai, T. Tsunooka, The effect of ethanol fuel on a spark ignition engine, 2006. https: / / doi.org/ https: / / doi.org/ 10.4271/ 2006-01-3380. 218 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="219"?> [9] M. Koç, Y. Sekmen, T. Topgül, H.S. Yücesu, The effects of ethanol--unleaded gasoline blends on engine performance and exhaust emissions in a spark-ignition engine, Renew. Energy. 34 (2009) 2101-2106. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.renene.2009.01.018. [10] D. Turner, H. Xu, R.F. Cracknell, V. Natarajan, X. Chen, Combustion performance of bio-ethanol at various blend ratios in a gasoline direct injection engine, Fuel. 90 (2011) 1999-2006. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2010.12.025. [11] M.O. El-Faroug, F. Yan, M. Luo, R. Fiifi Turkson, Spark ignition engine combustion, performance and emission products from hydrous ethanol and its blends with gasoline, Energies. 9 (2016) 984. https: / / doi.org/ https: / / doi.org/ 10.3390/ en9120984. [12] R.C. Costa, J.R. Sodré, Hydrous ethanol vs. gasoline-ethanol blend: Engine performance and emissions, Fuel. 89 (2010) 287-293. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2009.06.017. [13] Z. Yue, S. Som, Fuel property effects on knock propensity and thermal efficiency in a direct-injection spark-ignition engine, Appl. Energy. 281 (2021) 114221. https: / / doi.org/ https : / / doi.org/ 10.1016/ j.apenergy.2019.114221. [14] M. Rahimi Boldaji, B. Gainey, B. Lawler, Thermally stratified compression ignition enabled by wet ethanol with a split injection strategy: A CFD simulation study, Appl. Energy. 235 (2019) 813-826. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.apenergy.2018.11.009. [15] J.L.S. Fagundez, R.L. Sari, A. Garcia, F.M. Pereira, M.E.S. Martins, N.P.G. Salau, A chemical kinetics based investigation on laminar burning velocity and knock occurrence in a spark-ig‐ nition engine fueled with ethanol--water blends, Fuel. 280 (2020) 118587. https: / / doi.org/ htt ps: / / doi.org/ 10.1016/ j.fuel.2020.118587. [16] L. Bates, D. Bradley, Deflagrative, auto-ignitive, and detonative propagation regimes in engines, Combust. Flame. 175 (2017) 118-122. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.comb ustflame.2016.05.023. [17] J. Cho, H.H. Song, Understanding the effect of inhomogeneous fuel--air mixing on knocking characteristics of various ethanol reference fuels with RON 100 using rapid compression machine, Proc. Combust. Inst. 37 (2019) 4911-4919. https: / / doi.org/ https: / / doi.org/ 10.1016/ j. proci.2018.07.062. [18] C. Netzer, L. Seidel, F. Ravet, F. Mauß, Impact of the surrogate formulation on 3D CFD engine knock prediction using detailed chemistry, Fuel. 254 (2019) 115678. https: / / doi.org/ 10 .1016/ j.fuel.2019.115678. [19] D. Bradley, C. Morley, X.J. Gu, D.R. Emerson, Amplified pressure waves during autoignition: Relevance to CAI engines, SAE Tech. Pap. (2002). https: / / doi.org/ 10.4271/ 2002-01-2868. [20] X.J. Gu, D.R. Emerson, D. Bradley, Modes of reaction front propagation from hot spots, Combust. Flame. 133 (2003) 63-74. https: / / doi.org/ 10.1016/ S0010-2180(02)00541-2. [21] L. Seidel, Development and reduction of a multicomponent reference fuel for gasoline, BTU Cottbus - Senftenberg, 2017. https: / / opus4.kobv.de/ opus4-btu/ frontdoor/ index/ index/ year/ 2 017/ docId/ 4269. [22] C. Netzer, L. Seidel, F. Ravet, F. Mauss, Assessment of the validity of RANS knock prediction using the resonance theory, Int. J. Engine Res. 21 (2020) 610-621. https: / / doi.org/ 10.1177/ 146 8087419846032. 219 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="220"?> [23] C. Netzer, Simulation and assessment of engine knock events, BTU Cottbus - Senftenberg, 2019. https: / / opus4.kobv.de/ opus4-btu/ frontdoor/ index/ index/ year/ 2019/ docId/ 4877. [24] C. Netzer, L. Seidel, M. Pasternak, H. Lehtiniemi, C. Perlman, F. Ravet, F. Mauss, Three-di‐ mensional computational fluid dynamics engine knock prediction and evaluation based on detailed chemistry and detonation theory, Int. J. Engine Res. 19 (2018) 33-44. https: / / doi.org / 10.1177/ 1468087417740271. [25] A.A. Amsden, P.J. O’Rourke, T.D. Butler, KIVA-II: A computer program for chemically reactive flows with sprays, 1989. [26] P.J. O’Rourke, A.A. Amsden, A spray/ wall interaction submodel for the KIVA-3 wall film model, SAE Trans. (2000) 281-298. https: / / doi.org/ https: / / doi.org/ 10.4271/ 2000-01-0271. [27] D.P. Schmidt, C.J. Rutland, A new droplet collision algorithm, J. Comput. Phys. 164 (2000) 62-80. https: / / doi.org/ https: / / doi.org/ 10.1006/ jcph.2000.6568. [28] P.A. Vesilind, The Rosin-Rammler particle size distribution, Resour. Recover. Conserv. 5 (1980) 275-277. https: / / doi.org/ https: / / doi.org/ 10.1016/ 0304-3967(80)90007-4. [29] R.D. Reitz, Mechanism of breakup of round liquid jets, Encycl. Fluid Mech. 10 (1986). [30] M.S. Gern, G.M. Kauf, A. Vacca, T. Franken, A.C. Kulzer, Ganzheitliche methode zur bewertung der wassereinspritzung im ottomotor, MTZ-Motortechnische Zeitschrift. 80 (2019) 124-129. https: / / doi.org/ https: / / doi.org/ 10.1007/ s35146-019-0070-x. [31] Converge, CONVERGE Manual v2.4, (2017) 1008.[32] E.W. Lemmon, I.H. Bell, M.L. Huber, M.O. McLinden, NIST Standard Reference Database 23: Reference Fluid Thermodynamic and Transport Properties-REFPROP, Version 10.0, National Institute of Standards and Technology, Stand. Ref. Data Program, Gaithersbg. (2018). [33] P.G. Aleiferis, J. Shukla, M. Brewer, R.F. Cracknell, Spray development of iso-octane, ethanol, hydrous ethanol and water from a multi-hole injector under ultra cold fuel temperature conditions, Fuel. 303 (2021) 120983. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2021.120983. [34] C.K. Mohapatra, D.P. Schmidt, B.A. Sforozo, K.E. Matusik, Z. Yue, C.F. Powell, S. Som, B. Mohan, H.G. Im, J. Badra, others, Collaborative investigation of the internal flow and near-nozzle flow of an eight-hole gasoline injector (Engine Combustion Network Spray G), Int. J. Engine Res. (2020) 1468087420918449. https: / / doi.org/ https: / / doi.org/ 10.1177/ 14680874 20918449. [35] S. Kooij, R. Sijs, M.M. Denn, E. Villermaux, D. Bonn, What determines the drop size in sprays? , Phys. Rev. X. 8 (2018) 31019. https: / / doi.org/ https: / / doi.org/ 10.1103/ PhysRevX.8.031 019. [36] N. Peters, Turbulent combustion, Cambridge university press, 2000. [37] S.B. Pope, CEQ: A Fortran library to compute equilibrium compositions using Gibbs function continuation, (2003). [38] P.K. Senecal, E. Pomraning, K.J. Richards, T.E. Briggs, C.Y. Choi, R.M. McDavid, M. A. Patterson, Multi-dimensional modeling of direct-injection diesel spray liquid length and flame lift-off length using cfd and parallel detailed chemistry, SAE Tech. Pap. (2003). https: / / doi.org/ 10.4271/ 2003-01-1043. [39] LOGEresearch: combustion and chemical kinetics simulation software., (2021). 220 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="221"?> [40] T. Franken, L. Seidel, A. Matrisciano, F. Mauss, A.C. Kulzer, F. Schuerg, Analysis of the Water Addition Efficiency on Knock Suppression for Different Octane Ratings, SAE Tech. Pap. 2020-April (2020) 2531-2552. https: / / doi.org/ 10.4271/ 2020-01-0551. [41] M. Kauf, M. Gern, S. Seefeldt, Evaluation of Water Injection Strategies for NOx Reduction and Charge Cooling in SI Engines, JSAE/ SAE Powertrains, Fuels Lubr. (2019). https: / / doi.org / https: / / doi.org/ 10.4271/ 2019-01-2164. [42] M. Sophie, G. Technische, U. Berlin, A. Vacca, Experimental Analysis of the Influence of Water Injection Strategies on DISI Engine Particle Emissions, (2019) 1-9. https: / / doi.org/ 10. 4271/ 2019-24-0101.Abstract. [43] LOGEsoft, Manual: LOGEtable v2.0, (2020). www.logesoft.com. [44] N. Morgan, A. Smallbone, A. Bhave, M. Kraft, R. Cracknell, G. Kalghatgi, Mapping surrogate gasoline compositions into RON/ MON space, Combust. Flame. 157 (2010) 1122-1131. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.combustflame.2010.02.003. [45] J.E. Anderson, U. Kramer, S. A. Mueller, T.J. Wallington, Octane numbers of ethanoland methanolgasoline blends estimated from molar concentrations, Energy \& Fuels. 24 (2010) 6576-6585. https: / / doi.org/ https: / / doi.org/ 10.1021/ ef101125c. [46] T.M. Letcher, C. Heyward, S. Wootton, B. Shuttleworth, Ternary phase diagrams for gasoline-water-alcohol mixtures, Fuel. 65 (1986) 891-894. https: / / doi.org/ https: / / doi.org/ 10.1 016/ 0016-2361(86)90192-4. [47] A. Kyriakides, V. Dimas, E. Lymperopoulou, D. Karonis, E. Lois, Evaluation of gasoline-etha‐ nol-water ternary mixtures used as a fuel for an Otto engine, Fuel. 108 (2013) 208-215. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2013.02.035. [48] F. Karaosmanoǧlu, A. I\cs\iǧ\igür, H.A. Aksoy, Effects of a New Blending Agent on Ethanol-Gasoline Fuels, Energy & Fuels. 10 (1996) 816-820. https: / / doi.org/ https: / / doi.org/ 10 .1021/ ef950131z. [49] T. Franken, F. Mauss, L. Seidel, M.S. Gern, M. Kauf, A. Matrisciano, A.C. Kulzer, Gasoline engine performance simulation of water injection and low-pressure exhaust gas recirculation using tabulated chemistry, Int. J. Engine Res. 21 (2020) 1857-1877. https: / / doi.org/ 10.1177/ 14 68087420933124. [50] LOGEsoft, Manual: LOGE engine v3.0, (2015). www.logesoft.com.[51] G.T. Kalghatgi, D. Bradley, Pre-ignition and “super-knock” in turbo-charged spark-ignition engines, Int. J. Engine Res. 13 (2012) 399-414. https: / / doi.org/ 10.1177/ 1468087411431890. 52] S. Esposito, M. Mally, L. Cai, H. Pitsch, S. Pischinger, Validation of a RANS 3D-CFD gaseous emission model with space-, species-, and cycle-resolved measurements from an SI DI engine, Energies. 13 (2020) 4287. https: / / doi.org/ https: / / doi.org/ 10.3390/ en13174287. [53] Q. Yang, Z. Liu, X. Hou, X. He, M. Sjöberg, D. Vuilleumier, C. Liu, F. Liu, Measurements of laminar flame speeds and flame instability analysis of E30-air premixed flames at elevated temperatures and pressures, Fuel. 259 (2020) 116223. https: / / doi.org/ https: / / doi.org/ 10.1016/ j .fuel.2019.116223. [54] J.P.J. Van Lipzig, E.J.K. Nilsson, L.P.H. De Goey, A.A. Konnov, Laminar burning velocities of n-heptane, iso-octane, ethanol and their binary and tertiary mixtures, Fuel. 90 (2011) 2773-2781. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2011.04.029. 221 Effect of Gasoline - Ethanol - Water Mixtures on Auto-Ignition in a Spark Ignition Engine <?page no="222"?> [55] P. Iodice, M. Cardone, Ethanol/ gasoline blends as alternative fuel in last generation spark-ignition engines: a review on CO and HC engine out emissions, Energies. 14 (2021) 4034. https: / / doi.org/ https: / / doi.org/ 10.3390/ en14134034. [56] S. Cheng, D. Kang, A. Fridlyand, S.S. Goldsborough, C. Saggese, S. Wagnon, M.J. McNenly, M. Mehl, W.J. Pitz, D. Vuilleumier, Autoignition behavior of gasoline/ ethanol blends at engine-relevant conditions, Combust. Flame. 216 (2020) 369-384. https: / / doi.org/ 10.1016/ j.co mbustflame.2020.02.032. [57] S.M. Sarathy, P. Oßwald, N. Hansen, K. Kohse-Höinghaus, Alcohol combustion chemistry, Prog. Energy Combust. Sci. 44 (2014) 40-102. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.pecs.2 014.04.003. [58] L. Bates, D. Bradley, G. Paczko, N. Peters, Engine hot spots: Modes of auto-ignition and reaction propagation, Combust. Flame. 166 (2016) 80-85. https: / / doi.org/ https: / / doi.org/ 10.1 016/ j.combustflame.2016.01.002. [59] J. Pan, Z. Hu, Z. Pan, G. Shu, H. Wei, T. Li, C. Liu, Auto-ignition and knocking characteristics of gasoline/ ethanol blends in confined space with turbulence, Fuel. 294 (2021) 120559. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.fuel.2021.120559. [60] R. Lechner, A. Hornung, M. Brautsch, Prediction of the knock propensity of biogenous fuel gases: Application of the detonation theory to syngas blends, Fuel. 267 (2020) 117243. https: / / doi.org/ 10.1016/ j.fuel.2020.117243. [61] L. Van Treek, M. Lubrano, L. Seidel, F. Mauss, A.A. Konnov, Experimental and modelling study of laminar burning velocity of aqueous ethanol, Fuel. 257 (2019) 116069. https: / / doi.or g/ 10.1016/ j.fuel.2019.116069. [62] Z. Gong, L. Feng, W. Qu, L. Li, L. Wei, Auto-ignition characteristics of methane/ n-heptane mixtures under carbon dioxide and water dilution conditions, Appl. Energy. 278 (2020) 115639. https: / / doi.org/ https: / / doi.org/ 10.1016/ j.apenergy.2020.115639. [63] C. Netzer, T. Franken, F. Mauss, L. Seidel, H. Lehtiniemi, A.C. Kulzer, Numerical analysis of the impact of water injection on combustion and thermodynamics in a gasoline engine using detailed chemistry, SAE Int. J. Engines. 11 (2018) 1151-1166. https: / / doi.org/ https: / / doi.org/ 1 0.4271/ 2018-01-0200. 222 Tim Franken, Krishna P. Shrestha, Lars Seidel and Fabian Mauß <?page no="223"?> Innovative prechamber system with valve for future high efficiency engines Dimitrios Karageorgiou 1 , Li Cao 1 , Durgada Sankesh 2 , Patrick Gastaldi 1 , Matej Myslivecek 3 , Vianney Rabhi 4 1 2 3 4 Aramco Overseas Company, 232 Avenue Bonaparte, Rueil-Malmaison, 92500, France EMC France, 4 Allée de la rhubarbe, Achères, 78260, France Apside, 58 avenue du Général Leclerc, 92100 Boulogne-Billancourt Valvijet SAS, France Abstract: Continuous efforts to develop clean and efficient internal combustion engines have led to further development of various solutions towards improving diluted combustion in spark ignition engines. Pre-chamber ignition is considered as an effective enabling technology to extend the dilution operating limit (lean/ EGR), by providing spatially distributed flame jet ignitions. Compared to other pre-chamber ignition systems, the innovative Valvijet TM pre-chamber concept introduces a unique feature of a pressure driven valve that can separate the pre-chamber from the main chamber, hence allowing more favourable and independent mixture preparation inside the pre-chamber. The preliminary single cylinder engine testing results demonstrate that the Valvijet TM engine is capable of achieving the ultra-lean operation up to λ=2.5 with excellent combustion stability at the part load of 2000 rpm 7.2 bar IMEPg, whilst also maintaining a very good combustion stability in the ultra-lean limit up to l~1.8 at the low load of 1500 rpm 3 bar IMEPg. To better understand the potential benefits of this pre-chamber valve, presented in this concept, the full-cycle engine simulation coupled with Fluid-Structure Interaction (FSI) model is performed in this study. The combustion model and jet behaviour are validated against both engine testing and RCM optical measurement under the part load condition. The detailed analysis is then carried out to highlight the pre-chamber operation with assistance of valve (Valvijet TM ) and without valve (conventional active pre-chamber), in both ultra-lean and high level of EGR operations. The simulation study shows that an ISFC benefit of up to 5% at ultra-lean condition λ~2.2 and 3.5% gain at 40% EGR operating condition can be expected from the Valvijet TM , as compared to without valve operation. The main benefits of using the valve can be attributed to the combined effects of the ability to generate significantly higher levels of turbulence in pre-chamber due <?page no="224"?> to closed volume injection, and the ability to produce stronger jets due to more favourable mixture preparation in the pre-chamber, especially with high level of EGR. In addition, the pre-chamber valve helps to adjust the ignition energy more effectively, hence providing additional flexibility to optimize the overall system. 1 Introduction The transportation sector currently accounts for almost one quarter of the total global energy-related CO 2 emissions. With the ever-increasing demand for mobility, it is projected that at continued current rates of growth, the greenhouse gas emissions (GHG) from the transport sector could increase by up to 50% by 2035 [1]. To meet the global climate objectives, sustainable transport solutions must be implemented which requires diverse policy and technological adaptations [2]. In this context, it is essential to further develop clean and efficient internal combustion engines (ICEs), as they are about to represent an important part of the fleet mix, for the following decades [3]. In the recent years, the technology trend to push the efficiency of spark ignition engines has been mainly to operate under diluted (lean/ EGR) conditions in conjunction with higher compression ratios. The benefits achievable can be attributed to: i] reduced pumping losses, especially at low and partial loads, ii] increased mixture specific heat ratio, iii] favorable combustion phasing and iv] lower total heat losses. The extent of dilution that is possible is limited by mixture ignitability and combustion stability. To this end, the pre-chamber combustion system constitutes one of the technology bricks which can achieve these efficiency benefits while maintaining robust operation. During the last decades, extensive research on pre-chamber technology has introduced a number of concepts with different approaches on introducing an ignitable mixture into the pre-chamber [4]. The ‘passive PC’ concept is the simplest system in which the PC (pre-chamber) is filled with the charge from the MC (main chamber) during the compression stroke. This system offers advantages over standard spark ignition engines; however, the benefits are limited. The use of passive PC in the passenger car segment has been mostly in high performance applications such as motorsports. The ‘active PC’ concept is an improvement over the passive system, in which a secondary injector is used to introduce a small quantity of fuel to improve the ignitability of the PC charge. The secondary injection not only helps to produce much stronger reactive jets resulting in faster combustion, but also offers additional potential for a wider range of conditions where thermal efficiency can be further pushed, such as high dilution operation at low and partial loads. In terms of the injected fuel, several options have been investigated in the past: from liquid fuel injection [5], to air and fuel injection [6, 7, 8]. Air injection offers an additional advantage to scavenge the residuals from the PC and thereby improves the ignitability of the PC charge, at the cost of extra parasitic work. The purpose of the current research is to present an initial assessment of the Valvijet TM pre-chamber concept [9], which introduces a unique feature of a valve that 224 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="225"?> can separate the PC from the MC region, allowing more favorable mixture preparation inside the PC. 2 Description of the innovative pre-chamber system with valve Valvijet TM is actually an active PC concept, which uses a secondary high flow rate injector to provide a gaseous, pre-mixed air-fuel (A/ F) mixture to the PC volume. A small compressor supplies air to a compact external mixer, and maintains the required system injection pressure as well. The homogenous A/ F mixture is well prepared in the mixer, before it is fed into the PC through gaseous injection. The key element of the Valvijet TM concept relies on a unique and simple pre-chamber check valve, which is capable of separating the PC from the MC. This valve is primarily driven by the pressure difference between the PC and MC. The valve remains in a closed position when the MC pressure is slightly higher than the PC pressure. Additional magnetic force helps to ensure the valve closure, only when the pressure difference between the two chambers is very small. A magnet integrated on the upper part of the PC generates the magnetic field. During the compression stroke, the pressure difference between the two chambers leads to the valve closure to disconnect both chambers as illustrated in Figure 1. When the spark plug initiates the combustion in the PC, the PC pressure builds up rapidly up to the point when it exceeds the MC pressure. Then the valve starts to move towards its fully open position, which connects both chambers in the same way as any other pre-chamber system. The specific design of the PC includes a small ‘pneumatic damping chamber’ (DC) at the tip, which ensures no damage from hard collision on the valve seat. The authors fully acknowledge that ensuring the robustness of the PC valve is a key to the success of this concept. Significant amount of design and mechanical development efforts have been made but this is out of the scope of this study. Figure 1 - Operating principle of Valvijet TM system. Valve in closed position (left) during compression stroke (PC pressure < MC pressure) and Valve in open position (right) after PC spark ignition (PC pressure > MC pressure). [1. MC, 2. PC, 3. Valve, 4. Air-fuel injection, 5. Jets, 6. DC, 7. Spark plug] 225 Innovative prechamber system with valve for future high efficiency engines <?page no="226"?> The flexibility of disconnecting and connecting PC and MC with assistance of a simple valve offers a few advantages, which are summarized as follows: 1. When PC and MC are disconnected by this pressure driven valve in the compres‐ sion stroke, it enables a favorable mixture preparation inside PC, independent of MC condition. On the contrary, liquid fuel delivery into a conventional active pre-chamber with high EGR seems to be less effective for A/ F mixing due to less oxygen availability. Separate air and fuel gaseous injection can improve this by excessive over-fueling to scavenge the open pre-chamber, but at the expense of high parasitic losses. 2. This closed-volume (PC with closed valve) mixture injection provides more flexibility in adjusting the PC ignition energy (PC fueling) to achieve repeatable combustion at ultra-lean or at high EGR limits. 3. To achieve similar PC mixture quality, mixture injection into PC with closed valve requires less fueling than active open pre-chamber. This attribute is due to the fact that the excessive over-fueling (or air) to purge the over-diluted charge, trapped in PC, is inevitably needed for an active open pre-chamber, to improve mixture quality. This indicates Valvijet TM can utilize lower gaseous injection pressure for the same injector configuration, as compared to the active open pre-chamber also using gaseous injection. Figure 2 - Exemplary implementation of Valvijet TM system in an engine. [1. PC, 2. Valve, 3. Mixer, 4. Air-fuel injection, 5. Return line, 6. Spark plug] 226 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="227"?> To better illustrate the system (Figure 2) and its operating conditions, the pressure traces of PC and MC are shown in Figure 3, for an operating point achieved in a Rapid Compression Machine (RCM). This is an operating point without combustion in the MC, as only air is introduced in the MC. Five different phases of the PC combustion can be identified: • Phase 1, with the air/ fuel mixture injection (around 49 bars for this experiment); as the flow remains sonic at the injector nozzle, the pressure in the PC changes linearly versus time. • Phase 2, just after ignition by a conventional spark plug. The early flame devel‐ opment can be enhanced by the turbulence generated by the injection. The time interval between the end of injection and the spark timing is a key parameter to be optimized for the favourable turbulence level. • Phase 3, a very quick closed volume combustion occurs in the PC, before the PC valve opens; the partial closed volume combustion leads to a very fast pressure rise as the PC valve starts to open. After the peak pressure, hot burnt gases exit into the MC. • Phase 4 represents a small back flow from the MC as, during Phase 5, the remaining burnt gases are cooled down due to wall heat losses. Figure 3 - Test on a Rapid Compression Machine (RCM), equivalent conditions for an operating point of 2000 rpm 12 bar IMEP, PC_ φ =1.2, air in MC 227 Innovative prechamber system with valve for future high efficiency engines <?page no="228"?> 3 System operation and initial SCE test results 3.1 Experimental and engine setup The tests were carried out on a single cylinder engine equipped with the Valvijet TM pre-chamber system. PC pre-mixed A/ F mixture was supplied by a hydraulically controlled poppet valve gas injector, whereas the MC fuel was supplied in PFI config‐ uration. In order to accurately control of A/ F ratio inside PC, two Coriolis flow meters were used to measure air mass flow and fuel mass flow separately, which are then fed to an A/ F ratio control system dedicated for PC fueling metering as illustrated in Figure 4. The MC equivalence ratio is varied according to the set point in the test, whilst the PC equivalence ratio is kept constant at slightly rich condition with PC_ φ=1.1. To avoid any fuel condensation, the mixture temperature was maintained at 100°C. The injection pressure for PC was set at 49 bar for this preliminary testing. The PC pressure was measured by AVL GH15DK pressure transducer. A second piezoelectric pressure transducer (Kistler 6041B) installed on the cylinder head was used to measure the in-cylinder pressure in the main chamber. Figure 4 - Description of the PC air and fuel mixing system The SCE has a diesel engine flat cylinder head slightly modified to integrate the Val‐ vijet TM design. The valvetrain arrangement and the water jackets remained unchanged. To better assess the effect of the jet induced turbulence, low tumble and high flow rate intake ports are designed for no swirl charge motion. The main engine specification is listed in Table 1. 228 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="229"?> Bore (mm) 85 Stroke (mm) 88 Compression Ratio 12.8 Cylinder head Flat (diesel design) Piston Open bowl PC volume 0.7 cc PC jet hole size (diameter) 6 x φ1.0 mm Main Chamber Injection system Port Fuel Injection (PFI) Pre Chamber Injection system Gaseous mixture injection Fuel E10 gasoline (RON 95) Fuel Lower Heating Value (LHV) 42.3 MJ/ kg Table 1: Specifications of the single-cylinder engine 3.2 Single Cylinder Engine test results The preliminary SCE testing focused on the ability to operate the Valvijet engine at ultra-lean condition. The combustion stability and the engine-out emissions at ultra-lean limit were assessed at both operating points of 2000 rpm 7.2 bar IMEP g and 1500 rpm 3 bar IMEP g . The combustion system optimization, especially PC jet hole design, piston design, CR and charge motion optimization, have not been included in the first testing phase. Nevertheless, the obtained results are mainly used to verify the proof of concept of valve-based prechamber as well as to demonstrate some promising potential. The engine was able to operate at the part load point (7.2 bar IMEP g and 2000 rpm) with ultra-lean combustion up to MC_φ=0.4 (λ=2.5) with very good combustion stability (COV IMEP ≤ 1%). The lean limit was attained at MC_φ=0.35 (λ~2.9) with COV IMEP = 6%. All achieved operating points from λ=2 to λ=2.5 also show robust operation across the spark advance sweep (Figure 5). In Figure 6 (left), 2 operating points are presented as examples of a cycle-to-cycle variation on a sample of 100 cycles, at MC_φ=0.4 (λ=2.5), without detecting any misfire and with very few slow burn cycles. At the lean limit of MC_φ=0.35 (λ~2.9) no misfire was detected, which shows a promising potential for improvement of combustion stability with further optimization of the system (Figure 6 - right). 229 Innovative prechamber system with valve for future high efficiency engines <?page no="230"?> Figure 5 - COVIMEP vs. MC_ φ (left) and COVIMEP vs. spark advance for different MC_ φ, showing robust operation window (ref. operating point: 7.2 bar IMEPg at 2000 RPM) Figure 6 - Example of IMEPg distribution (left): MC_φ=0.4 (λ=2.5) at SA 21° (COVIMEP = 1,7%) and SA 23° (COVIMEP = 1%) - no misfire detected Lean stability limit (right): MC_φ=0.35 (λ=2.9) at SA 22° (COVIMEP = 5,4%) - no misfire detected. (ref. operating point: 7.2 bar IMEPg at 2000 RPM) At the low load operating point of 3 bar IMEP g and 1500 rpm the engine achieves stable combustion as shown in Figure 7. Combustion stability (COV IMEP ≤ 3%) across a wide range of SA was achieved from MC_φ=0.65 (λ~1.5) to MC_φ=0.55 (λ~1.8). 230 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="231"?> Figure 7 - COV IMEP vs. MC_ φ (left) and COV IMEP vs. spark advance for different MC_ φ (right). (Ref. operating point: 3 bar IMEP g at 1500 rpm) As discussed earlier, one of the advantages of a PC system with a valve is additional flexibility in adjusting the PC ignition energy to achieve repeatable combustion according to the MC dilution level. This phenomenon can be seen in Figure 8, which highlights the impact of the mixture flow in the PC. The mixture flow is an indicator of the ignition energy introduced in the PC. The increased ignition energy provided by the PC system in the case of MC_φ=0.40 operating point enables a faster combustion and therefore achieves a stable engine operation even at such lean conditions. Figure 8 - Plots showing PC and MC median pressure curves for MC_ φ=0.6 and 0.4. MC individual cycles are also shown (ref. operating point: 7.2 bar IMEPg and 2000 rpm) The evolution of the ISFC g (including PC and MC fuel consumption) related to the MC equivalence ratio and spark advance for the operating point of 7.2 bar IMEP g and 231 Innovative prechamber system with valve for future high efficiency engines <?page no="232"?> 2000 rpm is shown in Figure 9. The lowest value of 198 g/ kWh corresponding to an efficiency (ITE) of 43% is obtained at MC_φ=0.45. ISFC g remains relatively stable when SA is varied. The lean operation at low load of 3 bar IMEP g and 1500 rpm achieves an ISFC g of 224 g/ kWh at MC_φ=0.55 ( Figure 9). Pumping work has been evaluated at 3 bar IMEP g and the ISFC n (exhaust back-pressure close to ambient) is presented. The reduced pumping work at leaner mixtures enables to obtain a relatively low ISFC n of around 272 g/ kWh at MC_φ=0.55 (λ~1.8). It must be noted that at this stage of the preliminary testing activity the combustion system is not fully optimized. Further optimization of the system would possibly allow to reach λ=2 and beyond, which would bring additional benefits and reduction in NO x emissions. Figure 9 - Above: Plot showing ISFC g (PC+MC) vs. MC_λ. Ref. operating points: 7.2 bar IMEP g and 2000 rpm (left), 3 bar IMEP g and 1500 rpm (right). Below: ISFC g vs. SA (ref. operating point: 7.2 bar IMEP g and 2000 rpm) 232 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="233"?> The combustion phasing across the MC_φ sweep is presented in the Figure 10. At 7.2 bar 2000 rpm, an increase of the combustion duration is observed up to MC_φ=0.45 (λ~2.2). At MC_φ=0.40 (λ~2.5), close to the lean limit, the combustion duration shortens. This is due to increased PC ignition energy that is necessary to maintain good combustion stability. As a result the combustion efficiency is also improved. At 3 bar 1500 rpm, the increasing trend in the combustion duration with λ can be observed. Figure 10 - Combustion phasing comparison. Ref. operating points: 7.2 bar IMEPg and 2000 rpm (left), 3 bar IMEPg and 1500 rpm (right) The evolution of the combustion effciency and the uHC and CO emissions is shown in Figure 11 and Figure 12, respectively. The general trend observed at 7.2 bar IMEP g and 2000 rpm shows a decrease in combustion efficiency with increasing lambda, therefore increasing uHC and CO due to lower combustion temperature. The increase in combustion efficiency and decreasing trend in the uHC emissions beyond MC_λ=2 can be explained by the fact that the PC ignition energy used is higher and therefore the jets are much stronger. As a result, the MC combustion is more complete. As expected, the ability to operate the engine in ultra-lean conditions allows to drastically reduce the NO x emissions. At the best ISFC g points the NO x levels are lower than 0.5 g/ kWh, as illustrated in Figure 12. Similar trends in the evolution of emissions are observed for the low load case of 3 bar and 1500 rpm, as shown in Figure 12. Combustion efficiency decreases with increasing CO and uHC, as expected. The NO x levels show a steady decreasing trend, reaching levels below 1 g/ kWh for MC φ=0.55 (λ~1.8). 233 Innovative prechamber system with valve for future high efficiency engines <?page no="234"?> Figure 11 - Combustion efficiency comparison: ref. operating points: 7.2 bar IMEPg and 2000 rpm (left), 3 bar IMEPg and 1500 rpm (right) Figure 12 - Plot showing engine-out emissions for different MC_φ. Ref. operating points: 7.2 bar IMEPg and 2000 rpm (left), 3 bar IMEPg and 1500 rpm (right) • In summary, the preliminary tests at part load (7.2 bar, 2000 rpm) showed that the engine is capable to operate on a wide range of A/ F ratios up to λ=2.5 with good combustion stability (COV IMEP < 3%). The best ISFC g achieved is 198 g/ kWh at MC_φ=0.45 (λ~2.2) while maintaining the NO x emissions below 0.5 g/ kWh. • At low load conditions of 3 bar 1500 rpm, good combustion stability (COV IMEP < 3%) was achieved for MC_φ=0.65 to MC_φ=0.55 (λ~1.55 to λ~1.8). The best ISFC n achieved is 272 g/ kWh at MC_φ=0.55 (λ~1.8) with NO x emissions below 1 g/ kWh. Future work will focus on the further optimization of the system. 234 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="235"?> 4 Validation of modeling 4.1 CFD tools In addition to the conventional active pre-chamber system, a particular pre-chamber valve driven by pressure force can be used not only to control the pre-chamber ignition energy and jet penetration, but also to allow a favorable mixture preparation inside pre-chamber, independent of the mixture condition in main chamber. To a better understanding of the role of this valve in controlling turbulent jet ignition, a commercial CFD software CONVERGE is used for supporting this new concept evaluation [10]. A rigid-body Fluid Structure Interaction (FSI) model coupled with CFD simulation is used to predict the valve motion as shown in Figure 13. In the FSI model, at each computational time step, the valve kinematic motion is calculated based on all forces acting on the valve body, including pressure force, shear force, inertial force and any other external force, such as magnetic force. Figure 13 - Valve lift in relation to pressure difference between PC and MC Due to the added complexity of the PC valve and the uncertainties on the residuals trapped in both PC and MC chambers, a full-cycle simulation is performed to simulate the gas exchange in exhaust and intake strokes, the gas exchange between PC and MC with PC valve intermittent open/ close depending on the pressure difference, the gaseous mixture injection inside PC, and the subsequent ignition and combustion 235 Innovative prechamber system with valve for future high efficiency engines <?page no="236"?> processes. The base mesh size is 1 mm and refined down to 0.25 mm according to AMR based on gradients of temperature and velocity inside MC to ensure sufficient mesh resolution to resolve the flame front. The fixed mesh size of 0.25 mm is applied for the whole PC, except of the spark plug gap region with 0.125 mm further refined mesh size. Additional fixed geometry based mesh refinement is set for the key locations such as valve seat, nozzle hole (8-10 layers of mesh across the jet hole) and injector valve seat region with 0.03 mm of extremely fine mesh to well capture the choked flow. Turbulence is modeled using two-equation k-ε RNG model and a PRF surrogate fuel mechanism containing 48 species and 152 reactions [11] is used to simulate E10 gasoline pump fuel with RON 95. Two widely adopted flamelet based combustion models, namely G-equation model and Extended Coherent Flamelet Model (ECFM), and one multi-zone well-stirred reactor model (SAGE model in Converge) were attempted to simulate pre-chamber combustion [10, 12, 13]. There is still a lack of fundamental understanding of the effect of the boundary layer in a such small pre-chamber volume with very high surface/ volume ratio, ignition mechanism with small Damköhler number (ratio of flow timescale to chemical reaction timescale) and thickening flame behavior under highly turbulent flow (in a thin reaction zone with high Ka number) as well. Each combustion model has its own strengths and limitations. ECFM combustion model shows an advantage in simulating both pre-mixed and diffusion combustion expected the combustion inside active pre-chamber with liquid fuel injection, but it would be difficult for ECFM to capture jet ignition, in which chemical radicals in the hot jet and the hot jet re-ignition play an important role. G-equation model shows the capability to account for the effect of both large and small-scale turbulence on combustion, which might be a dominant factor in the combustion inside PC. Both flamelet based combustion models require accurate laminar flame speed tabulation and additional model to address flame quenching. In addition, there is a detailed chemistry based SAGE model, which treats each computation cells as a well-stirred reactor, using multi-zone grouping to reduce the computational cost. SAGE model can naturally consider the effect of boundary condition (flame quenching inside jet hole) on reaction rate. Although a certain level of the turbulent effect on combustion is considered through turbulent mixing, the uncertainties associated with SAGE model lie in that there is no explicit sub-grid Turbulence Chemistry Interaction (TCI). Therefore it is still necessary to further develop more predictive physical and chemical models dedicated for the pre-chamber combustion in future. For engineering application purpose, the SAGE model is chosen for the current simulation. A User Defined Function (UDF) is used to adjust the reaction multiplier based on flow regions. The reaction multiplier for PC needs to be calibrated against the test data, whilst the reaction multiplier is always maintained as 1 for MC to avoid unrealistic engine knocking incurred by increasing reaction multiplier. The combustion model calibration and assessment are arranged as followings: 236 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="237"?> 1. The reaction multiplier for PC only and Schmidt number are calibrated first against the SCE test data at IMEP g 7.2 bar, 2000 rpm with MC_φ=0.65 and PC_φ=1.1 2. The combustion model predictivity is assessed at equivalence ratio sweep inside MC (MC_φ=0.65, 0.50 and 0.45) and spark timing sweep (varied from -24 to -20 aTDC, in 2 CA deg. step) 3. Correlation with RCM measurement to understand hot jet behavior in terms of OH penetration and ignition delay 4.2 SCE data validation A testing point of IMEP 7.2 bar at 2000rpm from the preliminary SCE testing is used to calibrate the combustion model for MC equivalence ratio at 0.65 and PC equivalence ratio being optimized at 1.1. When PC reaction multiplier is increased from 1 to 1.4 and Schmidt number from 0.78 to 0.81, the simulation shows a very good match to the experiment with MC equivalence ratio 0.65, in terms of the combustion phasing, peak pressure and ignition delay in both PC and MC as shown in Figure 14. The simulation predicts reasonably well in the PC combustion, as the PC ignition energy increases with leaner MC mixture. As the MC equivalence ratio is leaner to 0.50 and 0.45, the predicted MC combustion phasing with MC_φ=0.50 and MC_φ=0.45 are delayed by 1 CAD and 2 CAD respectively, as compared to the experiment. Using the same model constants, the CFD model coupled with FSI method is able to capture the combustion phasing and ignition delay in both PC and MC under different lean conditions in a satisfactory way. It is worth mentioning that the discrepancy between the simulation and experiment increases slightly as the MC equivalence ratio approaches the lean limit. Similarly, the combustion model is also assessed with the spark timing sweep at the reference load point as shown in Figure 15. In general, the simulation matches reasonably well with the test data with large (early spark timing at -24 deg. aTDC) and without the overlap (late spark timing at -20 deg. aTDC) between EOI (at -20 deg. aTDC) and spark timing. It is unable to capture the sensitivity of the intermediate overlap between EOI and spark timing, due to oversimplifying the injector valve lift in the closing slope profile. At this stage, optimizing the overlap between EOI and spark timing should be done at the SCE testing, rather than by CFD simulation. 237 Innovative prechamber system with valve for future high efficiency engines <?page no="238"?> Figure 14 - The predicted and measured pressure profiles at different MC_φ 238 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="239"?> Figure 15 - The predicted and measured pressure profiles at different spark timing 4.3 RCM Correlation To further understand the behavior of the flame jets and to capture the underlying phenomena, CFD simulations were performed to correlate with RCM measurements based on OH chemiluminescence technique. High-speed schlieren based measurements have been performed previously to characterise the cold-jets and jet penetration. More details related to test setup etc. can be found here [14]. 239 Innovative prechamber system with valve for future high efficiency engines <?page no="240"?> Figure 16 - CFD simulation approach for high-pressure valve based active pre-chamber combustion system The above Figure 16 shows the modelling approach of valve based active pre-chamber combustion system. PC injection is simulated using a simplified rate of injection profile based on test measurements. The reference testing condition chosen is 12 bar IMEP at 2000 RPM with MC equivalence ratio of 0.65. The equivalence ratio of PC injection mixture is 1.2. The MC and PC pressure curves are shown in Figure 17. The pressure difference between PC and MC drives the motion of the valve. In the FSI modelling approach, the main resisting forces acting on the valve (i.e. inertial and magnetic) are taken into consideration. The side forces on the valve and the effect of friction cannot be accounted in the model and therefore are omitted. This limitation brings in some uncertainties in predicting the opening profile of the valve. As it can be observed in Figure 17, the closed volume phase of PC combustion is very well captured. When the valve starts to open, combustion pressure gradient in PC and the peak pressure is under-predicted. This can be linked to the valve profile itself, i.e. if the valve opens slower, the pressure rise would be faster due to ‘more’ fixed volume combustion. It must be noted that, the time when the hot jets start to exit the PC closely coincides with the valve reaching its maximum lift. The valve remains at its maximum lift for considerable amount of time before it starts to close. It can be seen from the plot that during the period of hot jet flow, the valve is almost fully opened. Therefore, it can be said that during the period of hot jet flow, the uncertainty associated with the valve is less (at least when compared to the phase of cold jet flow). As a result, the start of MC combustion is relatively well captured. The ignition delay is roughly about 1.5 CAD. Almost half of the injected mixture is pushed out of the PC before the hot jets are seen. 240 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="241"?> Figure 17 - Comparison of pressure curves in PC and MC between CFD simulation and RCM measure‐ ment For this test case, the peak ΔP is about 55 bar and the peak velocity inside the nozzles from CFD is in the range of 450 m/ s. A key feature of the valve based pre-chamber system is its ability to generate high velocity flame jets. As the PC region is isolated from the MC, the initial pressure in PC before injection is around 1-2 bar. High-pressure gaseous injection into a low-pressure region generates very high levels of turbulence. The spark timing is set just before end of injection such that there is some overlap between injection and ignition events. This helps to boost the start of PC combustion and subsequently helps to burn the PC fuel at a faster rate. 241 Innovative prechamber system with valve for future high efficiency engines <?page no="242"?> Figure 18 - OH evolution - RCM measurement (left), CFD simulation (right). (t=0 represents the start of hot jet.) Figure 19 - Comparison of axial OH penetration between RCM measurement and CFD simulation. (Time=0 represents the start of hot jet.) The evolution of OH axial penetration in the MC can be observed in Figure 19, in which the average penetration of the six jets is shown. As can be seen, the axial penetration as predicted by CFD is only slightly slower than experiments. This is mainly due to: i) lower peak ΔP (pressure difference b/ w PC and MC), ii) lower rate of pressure drop inside the PC during the phase of hot-jet flow (Figure 17). As discussed earlier, the 242 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="243"?> difficulty in matching the peak pressure is partly linked to the opening profile of the valve, which is not yet validated from measurements. Another feature that is well captured in CFD is the unequal OH distribution between the six flame jets, especially at the very beginning, which is a consequence of high turbulent flow in the PC. The side position of the injector in the PC also plays an important role in flow dispersion, which affects the flame propagation inside the PC. The impingement of the flame on the piston wall can also be observed from both RCM images and CFD (t = 0.375 ms). In summary, the simulation approach implemented, coupled with FSI modeling for valve motion, has satisfactorily reproduced the overall jet growth and hence the subsequent start of MC ignition is well captured. The radial spread of the flame jet is found to be slightly slower, particularly near the nozzle region, however, it must be noted that RANS modelling approach has limitations in predicting turbulent mixing for very high velocity flows. The measurement of the valve lift profile will help to better correlate the PC pressure gradient. Future work will focus on addressing these challenges. 4.4 Discussion on PC ignition energy One additional advantage of Valvijet TM system is that PC fueling can be adjusted quite easily, when A/ F mixture is injected into a closed-volume PC, rather than into an open-chamber PC as seen in the conventional active pre-chamber. The preliminary SCE test has shown that the PC ignition energy indicated by PC fueling increases as the MC mixture is leaner as shown in Fig. 6. Increasing PC ignition energy leads to higher jet enthalpy and longer jet penetration, hence shorter MC ignition time and faster flame propagation. This might help explaining why the combustion stability remains very good (COV_IMEP < 1%), even at the lean condition of MC_φ=0.40, by increasing the PC fueling by 75% in the SCE test, as compared to the MC_φ=0.6 case. Figure 20 - PC fueling (injected mixture) at different MC_φ 243 Innovative prechamber system with valve for future high efficiency engines <?page no="244"?> To understand the PC ignition energy trade-off between MC combustion enhancement and ignition success probability, the CFD simulation based study on the effect of PC fueling is performed at the lean condition of MC_φ =0.45, in which the calibrated combustion model had shown a good predictive capability. Figure 21 shows the pressure traces with varied PC fueling. As PC fueling increases from 2% to 3% of MC fueling, higher pressure differential between PC and MC increases jet momentum. As a result, the combined effect of more hot gas volume ignition and higher jet induced turbulence contribute to faster flame propagation inside MC as evidently observed in Figure 22. It is also evident that smaller jet hole size is beneficial to faster MC combustion due to the same reason . Smaller jet hole may be susceptible to PC flame quenching due to excessive wall heat loss and higher flame stretch rate through the orifice. It is interesting to note that the ignition of one hot jet becomes not sustainable, when the jet velocity increases by further increasing PC fueling from 3% to 3.4% as shown in Figure 23. At the lean limit condition, further increase PC fueling increases jet flame quenching probability. As the hot jet penetrates into MC, the jet surface contains many small size vortices. These vortices actually help mixing the hot jet with the cold fresh charge in MC during the jet penetrating process. If the turbulent jet is too strong with relatively high velocity, the excessively fast mixing between the hot jet and cold fresh charge might cause the hot jet temperature drop too rapidly, hence it is not able to ignite the mixture in MC. This observation is consistent with the experimental results of non-ignition case . The system has more flexibility to control the PC ignition energy, which helps to maintain very good combustion stability at MC_φ=0.40. As discussed on the trade-off of PC ignition energy, the required PC ignition energy needs to be optimized to further improve thermal efficiency in the next engine testing campaign. Figure 21 - Comparison of PC and MC pressure curves between with-valve and no-valve cases. (ref. operating point: 7.2 bar IMEPg 2000 RPM) Pressure profiles with different PC fueling at MC_φ=0.45 244 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="245"?> Figure 22 - Flame development with different PC fuelings (iso-temperature=1800K) Figure 23 - Jet velocity at -14° aTDC 5 Assessment of the valve based system In this section, we present the first results of our CFD comparative study between Valvijet TM system (with-valve) and conventional active pre-chamber (no-valve) system. The objective is to highlight how closed volume injection and combustion using a valve has the potential to further improve the performance of active pre-chamber system. 245 Innovative prechamber system with valve for future high efficiency engines <?page no="246"?> At first, the major benefits with-valve operation are explained, which can be attributed to two main factors: i] the ability to generate significantly higher levels of turbulence in PC due to closed volume injection into a low pressure region, ii] the ability to produce stronger jets due to favorable mixture preparation in the PC, especially at high MC dilution conditions. To this purpose, for a direct comparison we have used the same PC geometrical design for no-valve case with the valve removed. The reference operating condition is 7.2 bar IMEP g at 2000 RPM with MC equivalence ratio of 0.45. The operating parameters of PC injection such as injection timing, duration, injector lift and pressure is kept the same as with-valve case. In other words, PC fueling as well as MC fueling is maintained the same for with and without the valve cases. Next, the benefits in ISFC achievable because of these advantages are assessed. To this end, in order to obtain similar work, only the injector lift is varied to increase PC fueling when needed while keeping the same rest of the injection parameters. This comparison is performed at validated MC lean conditions (φ = 0.65, 0.50, 0.45). A case of high EGR (40%) in the MC is also presented by extending the same CFD modeling approach as used for lean cases. • Comparison of injection-induced turbulence Figure 24 - Comparison of PC turbulence between with-valve and no-valve cases, for same PC fueling (ref. operating point: 7.2 bar IMEP g 2000 RPM) At the start of PC injection in the case of with-valve, the pressure in PC is close to intake pressure (1-2 bar), as can be seen from Figure 24. High-pressure gas injection into a low-pressure region (closed PC) creates significantly higher levels of turbulence. Whereas in the no-valve case, as the PC pressure is at the same level as MC pressure (open PC), the injection momentum is dampened and therefore turbulence levels are not as high as with-valve case. The mean TKE in the PC close to spark timing is almost 246 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="247"?> double, as can be observed from Figure 24. The distribution of TKE is shown in Figure 25. Figure 25 - Comparison of PC TKE, φ and temperature distribution between with-valve (left) and no-valve (right) cases before ignition, for same PC fueling. (ref. operating point: 7.2 bar IMEP g 2000 RPM, MC_φ = 0.45) • Comparison of mixture preparation The valve isolates PC and MC regions, which helps to prepare the PC mixture independently from MC conditions. As a result, the PC can be operated with almost no dilution when MC is operated at high levels of dilution (lean/ EGR). The penetration of the injected gaseous mixture inside the PC in the case of with-valve is much higher because of lower density in the PC at the start of injection. Therefore, the injected fuel in the PC in the case of no-valve does not mix completely as compared to with-valve case. This can be seen from the equivalence ratio distribution in Figure 25. The equivalence ratio in the PC close to the spark plug region with no-valve case is about 0.83. This diluted mixture causes slower combustion in the PC, as is evident from the pressure gradient in Figure 26. Secondly, some of the diluted mixture is pushed out into the MC as ‘cold jet’, whereas in the case of with-valve, the ‘cold-jet’ is always close to the injected air-fuel ratio, which burns first causing rapid MC ignition, as seen from the SCE test results. In the case of the no-valve system the diluted ‘cold jet’ causes slower flame propagation in the MC and, in the worst-case, partial burn or misfire may occur, as shown in Figure 26. The stronger ‘cold-jet’ for with-valve case can be seen from the iso-contours of fuel 247 Innovative prechamber system with valve for future high efficiency engines <?page no="248"?> mass fraction in Figure 26. The ‘cold jets’ play a significant role in initiating MC mixture combustion by enriching the mixture locally, especially at high lean/ EGR conditions. It must be noted that the rapid expansion of the injected gas (choked flow) in the PC in the case of with-valve system results in much lower gas temperature as shown in Figure 25; however, turbulence and dilution effect is found to be more dominant. The rate of enthalpy flow of the jets (function of mass flow rate and temperature) as shown in Figure 26 clearly indicates that the with-valve system produces much stronger jets, as a direct result of faster combustion (due to injection-induced turbulence) and with little dilution in the PC. Figure 26 - Comparison of ‘cold jets’ and ‘flame jets’ between with-valve and no-valve cases, for same PC fueling (ref. operating point: 7.2 bar IMEP g 2000 RPM, MC_φ = 0.45) • ISFC comparison One way to reduce PC dilution for the no-valve case is to over-fuel the PC, i.e. by continuously injecting more fuel such that some of the fuel is pushed out into the MC, as shown in Figure 27. This not only reduces the PC dilution but also produces richer cold jets, similar to the with-valve case, which finally helps to initiate MC combustion, as can be seen from the pressure curves in Figure 27. The increased fueling also increases turbulence inside the PC. 248 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="249"?> Figure 27 - Effect of increased PC fueling on mixture preparation for no-valve cases. (φ mean corresponds to top region of PC) (ref. operating point: 7.2 bar IMEP g 2000 RPM, MC_φ = 0.45) CFD simulations show that PC fueling must be increased by at least 1 to 3.5 times more as compared to with-valve case, as illustrated in Figure 28. This translates into about 5.3% (at MC_φ=0.45) and about 3.5% (at MC_EGR=40%) benefit in ISFC for the validated reference condition of 7.2 bar IMEP g and 2000 RPM. In addition, MC combustion is also faster as can be seen from the pressure curves in Figure 29, which is an indicator of very good combustion stability. The ISFC benefit obtained at higher equivalence ratios (MC_φ = 0.65, 0.50) is in the range of 1.0% to 1.5%. At these equivalence ratios for the reference operating point, over-fueling is not necessary and hence the benefit mainly comes from increased 249 Innovative prechamber system with valve for future high efficiency engines <?page no="250"?> turbulence. It is worth mentioning that PC dilution can also be reduced for no-valve case at high EGR conditions by injecting only air to scavenge the residual high EGR mixture into the MC. Figure 28 - Comparison of PC ignition energy (fuel mass) between with-valve and no-valve cases. (ref. operating point: 7.2 bar IMEP g 2000 RPM) 250 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="251"?> Figure 29 - Comparison of PC and MC pressure curves between with-valve and no-valve cases. (ref. operating point: 7.2 bar IMEPg 2000 RPM) In summary, high dilution conditions (lean/ EGR) can be more efficiently operated with Valvijet TM system due to the ability to produce stronger turbulent jets, as a result of separating the PC and MC regions by using a valve in the PC. Based on validated CFD simulations, a benefit of up to 5% in ISFC is seen at ultra-lean MC condition of φ=0.45 (λ=2.2) as compared to without valve operation. CFD study also shows the benefit in ISFC at ultra-high EGR operation of 40%. Future studies will focus on pushing the EGR limit and to assess EGR vs. lean operation at low load conditions. 6 Conclusions/ Summary This article introduces an innovative active pre-chamber combustion system in which a valve is introduced in the PC to separate the PC and MC regions. This helps to optimize favorable thermodynamic and mixture conditions in the PC, independent of the MC conditions. Preliminary SCE test results for a wide range of MC lean conditions at a reference operating point of 7.2 bar IMEP g and 2000 RPM are presented. In addition, CFD simulation study has been performed to validate our modeling approach and finally assess the system operation with and without the valve. The key highlights are summarized below: • Lean operation of up to φ = 0.40 (λ = 2.5) is achieved with excellent combustion stability (COV of IMEP g ) of 1.0%. • Our CFD modeling approach is able to satisfactorily reproduce: 251 Innovative prechamber system with valve for future high efficiency engines <?page no="252"?> • Valve motion based on FSI (fluid-structure interaction) modeling • Combustion phasing in both PC and MC regions for different air-fuel ratios and spark timing sweep. • The jet behavior and start of MC ignition based on OH measurements. • An ISFC benefit of up to 5% at φ = 0.45 (λ = 2.2) and 3.5% at 40% EGR can be achieved as compared to without valve operation, based on CFD simulations. The main benefits of using the valve can be attributed to two main factors i.e. the ability to generate significantly higher levels of turbulence in PC due to closed volume injection into a low pressure region, and ability to produce stronger jets due to favorable mixture preparation in the PC, especially at high MC dilution conditions. In addition, the valve helps to adjust the PC ignition energy more effectively, thus providing additional flexibility to optimize the overall system. Next series of SCE tests will focus on EGR operation at low load and high load (knock-limited) with higher compression ratio, to assess the potential of Valvijet TM system. This will also help us to gain further understanding of PC combustion system and improve our predictive CFD modeling capability, especially at dilution limits for future work. 7 References [1] R. Sims, R. Schaeffer, F. Creutzig et al., 'Climate Change 2014: Mitigation of Climate. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change,’ Cambridge University Press, Cambridge, 2014. [2] A. Milovanoff, I. Posen and H. MacLean, ‘Electrification of light-duty vehicle fleet alone will not meet mitigation targets,’ Nature Climate Change, pp. 1102-1107, 2020. [3] ‘Global EV Outlook 2021,’ International Energy Agency, 2021. [4] E. Toulson, H. Schock and W. Attard, ‘A Review of Pre-Chamber Initiated Jet Ignition Combustion Systems,’ in SAE Powertrains Fuels & Lubricants Meeting, 2010. [5] M. Bunce and H. Blaxill, ‘Development of Both Active and Passive Pre-Chamber Jet Ignition Multi-Cylinder Demonstrator Engines,’ in Aachen Colloquium Sustainable Mobility, Aachen, 2019. [6] T. Russwurm, M. Schumacher and M. Wensing, ‘Active Fuelling of a Passenger Car Sized Pre-Chamber Ignition System with Gaseous Components of Gasoline,’ in SAE Powertrains, Fuels & Lubricants Meeting, 2020. [7] C. Müller, B. Morcinkowski, C. Schernus et al., ‘Development of a Pre-chamber for Spark Ignition Engines in Vehicle Applications,’ in Ignition Systems for Gasoline Engines, 4th International Conference, Berlin, 2018. [8] M. Sens, E. Binder, P.-B. Reinicke et al., ‘Pre-Chamber Ignition and Promising Complemen‐ tary Technologies,’ in Aachen Colloquium Sustainable Mobility, Aachen, 2018. [9] V. Rabhi, ‘PRECHAMBRE D'ALLUMAGE A CLAPET’. France Patent WO2018/ 130772A1, 19 07 2018. 252 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="253"?> [10] K. J. Richards, P. K. Senecal and E. Pomraning, CONVERGE 3.0 Manual, Convergent Science, Madison, 2020. [11] Y.-D. Liu, M. Jia, M.-Z. Xie and B. Pang, ‘Enhancement on a Skeletal Kinetic Model for Primary Reference Fuel Oxidation by Using a Semidecoupling Methodology,’ Energy Fuels, no. 12, pp. 7069-7083, 2021. [12] J. Kim, R. Sarcelli, S. Som, Shah, A., M. Biruduganti and D. Longman, Assessment of Turbulent Combustion Models for Simulating Pre-Chamber Ignition in a Natural Gas Engine, American Society of Mechanical Engineers, 2019. [13] A. Zhang, X. Yu, N. Engineer, Y. Zhang and Y. Pei, Numerical Investigation of Prechamber Jet Combustion, Americal Society of Mechanical Engineers , 2020. [14] Dr. P. Gastaldi, Dr. L. Cao, D. Karageorgiou, et al., ‘A new concept of active pre-chamber for low CO2 SI engines: contribution to combustion understanding based on CFD and visu‐ alization techniques’,’ in THIESEL, Thermo and fluid dynamic processes in direct injection engines, Valencia, 2020. [15] S. Biswas, S. Tanvir, H. Wang and L. Qiao, ‘On Ignition Mechanisms of Premixed CH 4 / Air and H 2 / Air Using a Hot Turbulent Jet Generated by Pre-Chamber Combustion,’ Appl. Therm. Eng, no. 106, pp. 925-937, 2016. Appendix Abbreviations A/ F Air to Fuel ratio CAD Crank Angle Degree DC Damping Chamber ECFM Extended Coherent Flamelet Model EGR Exhaust Gas Recirculation EOI End of Injection FSI Fluid Structure Interaction IMEP Indicated Mean Effective Pressure ITE Indicated Thermal Efficiency ISFC Indicated Specific Fuel Consumption MC Main Chamber PC Pre-Chamber RANS Reynolds-averaged Navier-Stokes RCM Rapid Compression Machine RON Research Octane Number 253 Innovative prechamber system with valve for future high efficiency engines <?page no="254"?> SA Spark Advance SCE Single Eylinder Engine TCI Turbulence Chemistry Interaction TDC Top Dead Center a after Symbols λ lambda (actual/ stoichiometric air-fuel ratio) φ equivalence ratio Subscripts g gross n net 254 Dimitrios Karageorgiou, Li Cao, Durgada Sankesh, Patrick Gastaldi, Matej Myslivecek, Vianney Rabhi <?page no="255"?> Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin Abstract: Towards carbon neutral future, we must consider every kind of possible efficient powertrain. Internal combustion engine obviously keep taking an important role. Diluted gasoline engine have been developed continuously as one of key technology to reduce CO2 emission. In order to achieve higher thermal efficiency, increasing dilution rate is one possibility. Pre-chamber system has been considered as one solution to achieve stable combustion under highly diluted condition because it can increase combustion speed significantly. Passive pre-chamber has simple structure, therefore it is easy to introduce. Whereas, controlling lambda separately between pre-chamber and main chamber is not possible. On the other hand, active pre-chamber can enable that and has potential to control mixture distribution. However, injecting liquid fuel into pre-chamber is problematic regarding fuel impingement and consequently difficult to obtain optimized mixture distribution for ignition. In this study, fuel and air mixture in‐ jection into pre-chamber was tested to overcome these issues. We investigated the potential of the system and the effect of pre-chamber configuration by conducting 3D-CFD analysis of cold flow and single cylinder experiment for assessment of combustion behaviour. The results show that the required configuration is different for air dilution and exhaust gas dilution because purging exhaust gas from pre-chamber is important for EGR combustion. Combustion stability is ensured at lambda 2.8 for air dilution and at 41% EGR rate for exhaust gas dilution. 1 introduction Requirement of CO2 emission reduction is increasing for automotive industry. Electri‐ fication is rapidly advancing, however, diversity of powertrain considering different usage and location should be respected to achieve carbon neutral society. Therefore improving combustion efficiency of internal combustion engine is important. Diluted burn with air or EGR offer significant improvements in thermal efficiency. In addition, engine out NOx emission is also reduced by diluted burn. With standard spark plug, the limit of air fuel ratio and EGR ratio should be around 2.0 and 30% because combustion <?page no="256"?> get unstable in highly diluted condition[1][2]. Passive pre-chamber can improve thermal efficiency thanks to faster combustion by flame jets propagation [3]. On the other hand, in order to push the dilution limit up, providing adequate mixture around ignition point is mandatory for early flame kernel growth, but difficult at the same time with passive pre-chamber. Therefore, Active pre-chamber has been developed as the solution. Active pre-chamber system which inject fuel into pre-chamber can achieve stable combustion under up to λ2.2[4]. The system was also tested for EGR diluted combustion. In passive pre-chamber, EGR rate around ignition point gets higher than main chamber because of not complete residual gas scavenging [reference]. The active pre-chamber is able to extend EGR limit and improve thermal efficiency significantly because fuel/ air mixture injection improves the pre-chamber scavenging. Hence active pre-chamber is considered as key technology for gasoline dilution engine. In this study, we applied the active pre-chamber system to a single cylinder module which is similar to our base design and we tried to find what effect each design parameter has and consequently optimise the pre-chamber specification. Both calculation and experimental approaches have been adopted for practical and efficient development. 2 development condition and pre assessment 2.1 Structure of active pre-chamber with mixture injection system Fig.1 shows the active pre-chamber system structure. The pre-chamber is located in the centre of the cylinder. Air is continuously supplied at regulated pressure before air/ fuel injector. An injector inject fuel before air/ fuel injector, then air/ fuel injector inject the mixture into the pre-chamber. Direct injector provides fuel for main combustion chamber. Pre-chamber design parameters to be analysed in this study are chosen as shown in the figure. Fig. 1: Overview of the active pre-chamber system and design parameter 256 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="257"?> 2.2 Calculation Simulation modelling of combustion process with pre-chamber ignition has been studied in order to analyse and understand phenomena in combustion chamber and for practical design development [5][6]. In this study, 3D CFD was used for analysis of cold flow in both main chamber and pre-chamber. Regarding combustion regime, lean mixture with pre-chamber ignition is expected to generate broken reaction and thin reaction flame due to slow laminar flame speed and strong turbulence caused by jets [5]. Nowadays modelling this complex interaction between turbulence and flame structure is one of the main research topics related to combustion simulation development, and significant progress can be expected in the next future. But for this study, measurement method was chosen to assess combustion behaviour. 2.3 Test engine Tab.1 shows the specification of single cylinder test engine. Cylinder charge motion is variable by tumble flap. External super charger provides sufficient boost pressure to reach high lambda and EGR ratio. Together with the boost pressure, exhaust back pressure is also regulated by a flap to simulate a real turbocharger. The back pressure is set to 50 mbar higher than intake pressure, which is also necessary to reach high EGR ratio. The engine was operated under 2000rpm engine speed and 8bar IMEP. MFB50 is targeted to 8deg.ATDC as long as knocking does not prevent. In order to study the effect of design parameters and consequently optimise them, pre-chambers with specifications listed in tab.2 were manufactured and tested. Bore (mm) 83.1 Stroke (mm) 92.0 Stroke bore ratio 1.1 Displacement (cm3) 498 Compression ratio 13 : 1 Air charging system Super charged Fuel injection pressure for main/ pre-chamber (bar) 200 / 9 Air supply pressure for pre-chamber (bar) 6.4, (7.4) Tab. 1: Engine specification 257 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="258"?> Volume [cm3] Lateral hole diameter [mm] A/ V [cm-1] hole number (L + C) Central hole diameter [mm] Swirl angle [deg.] Inclination angle [deg.] Base V0 1.6 1.30 0.058 7 + 0 - 15 70 V with const. LHD V1 2.0 1.30 0.046 7 + 0 - 15 70 V2 2.5 1.30 0.037 7 + 0 - 15 70 V with const. A/ V V3 2.0 1.45 0.058 7 + 0 - 15 70 V4 2.5 1.63 0.058 7 + 0 - 15 70 A/ V with const. V V5 1.6 1.16 0.046 7 + 0 - 15 70 V6 1.6 1.04 0.037 7 + 0 - 15 70 CHD with const. LHD V7 1.6 1.30 0.063 7 + 1 1 15 70 V8 1.6 1.30 0.078 7 + 1 2 15 70 CHD with const. A/ V V9 1.6 1.24 0.058 7 + 1 1 15 70 V10 1.6 1.06 0.058 7 + 1 2 15 70 HN with const. A/ V V11 1.6 1.54 0.058 5 + 0 - 15 70 V12 1.6 1.40 0.058 6 + 0 - 15 70 Swirl angle V13 1.6 1.30 0.058 7 + 0 - 10 70 V14 1.6 1.30 0.058 7 + 0 - 20 70 Inclination angle V15 1.6 1.30 0.058 7 + 0 - 15 60 V16 1.6 1.30 0.058 7 + 0 - 15 78 Tab. 2: Pre-chamber variant for measurement 2.4 Injection strategy The pre-chamber system has two injectors, one injects fuel and the other injects air/ fuel mixture. Injection timing, duration and pressure need to be calibrated to operate engine. Injection pressure of each injector was fixed as shown in tab.1. As a result of preliminary study, injection strategy was defined as described below: • Air/ fuel injector needs to be closed by when cylinder pressure (cylinder pressure or pre-chamber pressure) exceeds injection pressure to avoid injector backflow. • Air/ fuel injector opens as late as possible in order to keep the mixture in the pre-chamber. • Fuel should be injected while air/ fuel injector is opening to get good mixing 258 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="259"?> • Fuel injected before mixture injector should be increased as lambda increases in air lean operation. Whereas it is not necessary for EGR operation because gas from main chamber is at lambda=1 • Longer Air/ mixture injection for EGR operation to purge Pre-chamber is required. Fig.2 shows CFD result of injection strategy assessment for EGR operation. EGR rate of intake gas is set as 45% but EGR rate in pre-chamber is more than 60% even after intake stroke. Mixture injection purge pre-chamber and EGR rate decrease to around 5% at injection end timing. Then back flow to pre-chamber happen in the latter part of compression stroke and increase EGR rate to around 30% at ignition timing. Active pre-chamber contribute to better ignition for EGR operation in this way. Fig. 2: injection strategy assessment with 3D-CFD 2.5 Ignitability assessment Flame regimes inside pre-chamber was assessed in order to ensure ignition and flame propagation by Borghi-peters diagram. As shown in fig.3, according to the publication of J.Benajes, et al, jet from pre-chamber is located in the broken reaction zones, whereas conventional spark ignition combustion is located in the thin reaction zones or corrugated flamelets[5]. The cold flow simulation results are analysed from -40 crank angle degree to +40 crank angle degree aroud TDC. Mixture conditions and turbulence in spark gap, analytical calculations of laminar flame speed and flame thickness by Gulders equation are used to be located in the diagram. The red and black curves represent, for EGR and lean case respectively, the average flame regime evolution 259 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="260"?> within 10 mm diameter spherical volume around ignition gap. Flame regimes inside pre-chamber are similar to conventional SI stoichiometric combustion. Therefore, ignitability may not be critical issue for the pre-chamber system. Fig. 3: combustion regime with Borghi-Peters diagram 3 Results and discussions 3.1 Base design pre-chamber Fig.4 shows a performance comparison of standard spark-plug and V0 pre-chamber variant as base of the design variation without tumble flap condition. The combustion stability limit here is set to COV = 3%. Pre-chamber can extend both lean and EGR limit thanks to fast and stable combustion. The change in COV at lambda = 2.3, which can be seen in the red curve, is due to a change of the injection strategy in the pre-chamber. More fuel is injected into the pre-chamber, resulting in a richer mixture with better ignition conditions. As a result, the mixture can be further enleaned globaly, yet with increasing particulate emissions. This adjustment of the injection strategy was carried out in the following measurements for each pre-chamber variant. 260 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="261"?> Fig. 4: Performance comparison of standard plug and base design pre-chamber V0 3.2 Charge motion Charge motion of recent engine has been enhanced to achieve fast combustion. Especially for diluted burn, sufficient turbulence energy is necessary to compensate low laminar flame speed. On the other hand, intake port design which enables strong charge motion causes a decrease of flow coefficient and production difficulty. Pre-chamber also enables fast combustion, therefore strong charge motion may not be necessary. To clarify this point, performance comparison between with and without tumble flap was carried out. As CFD results in fig.5 and 6 show, the tumble flap increases velocity and TKE in the main chamber, but TKE inside the pre-chamber is not so affected, so the tumble flap does not influence flow inside pre-chamber. In addition, high tumble lead to better mixture homogenization as shown in fig.7. Fig.8 shows the engine dyno result with and without tumble flap. For the lean operation, combustion is slightly more stable with higher tumble condition below lambda=2, but there is no big difference. For EGR operation, higher tumble can make combustion faster and results in more stable combustion and lower fuel consumption. Thus strong turbulence and better homogenization thanks to high tumble seem to be necessary even for pre-chamber combustion with EGR operation. Therefore high tumble was applied to the following measurement. 261 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="262"?> Fig. 5: TKE in main combustion chamber and pre-chamber Fig. 6: comparison of velocity magnitude at -60deg.ATDC Fig. 7: comparison of lambda at -30deg.ATDC for EGR operation 262 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="263"?> Fig. 8: Performance comparison of with and without tumble flap 3.3 Lean operation 3.3.1 Interaction between Pre-chamber volume and hole diameter Fig. 9 shows comprehensive investigation results of the influence of Pre-chamber volume and lateral hole diameter as most important design parameter. From the comparison of pre-chamber volume with constant lateral hole diameter, Jet from pre-chamber with small volume seems not to make enough penetration into main chamber. As a result, the combustion efficiency is lower with small volume. Whereas, the result of comparison of pre-chamber volume with constant A/ V does not make significant difference on COV, fuel consumption and combustion efficiency in spite that combustion energy in pre-chamber is assumed higher with larger volume by increase of NOx. This result indicates that A/ V is key factor to gain high combustion efficiency. The result of A/ V variation with constant volume shows smaller A/ V can achieve lower fuel consumption because of higher combustion efficiency keeping NOx amount. The smaller A/ V results the better performance in this measurement, however too small A/ V is expected to cause too much pressure loss between pre-chamber and main chamber, and worse performance at last. To summarize the above, appropriate A/ V to gain jet penetration with minimum pre-chamber volume is necessary for low fuel consumption and emission of NOx and HC. 263 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="264"?> Fig. 9: Performance comparison of Pre-chamber volume and lateral hole diameter 3.3.2 The effects of central hole Expected main benefit of the central hole is to help scavenge residual gas from pre-chamber. Fig.10 shows the results of the measurement of central hole variant with constant A/ V. COV and THC emission get worse with larger central hole bore. Larger central hole cause less intense jet from lateral holes to main chamber as assumed with longer burn duration. In addition, a higher flow velocity is generated with the inflow through the central hole in the spark plug gap. This can have a negative influence on the flame kernel growth and can also be the cause of the results. Thus Central hole is not needed as far as residual gas in the pre-chamber is not too much for ignition. 264 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="265"?> Fig.10: Performance comparison of central hole to Performance 3.3.3 The effects of Lateral hole number The measurement for lateral hole number effect was conducted with constant A/ V as shown in Fig.11. Fewer hole number tend to obtain lower COV and fuel consumption. A Jet from one hole is assumed to be strong and reach near cylinder liner with fewer holes variant. As a result, THC emission decrease as shown in the graph. However, burn duration gets longer probably because of big clearance between each jet. From this result, it is expected that large hole number show better performance with smaller A/ V.. In general, optimization of combination between hole number and A/ V would need to be done considering this interaction. Fig. 11: Performance comparison of Lateral hole number 265 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="266"?> 3.3.4 The effects of Swirl angle Fig.12 shows result of effect of swirl angle. Biggest swirl angle seems to have slightly better performance. Shorter duration of first combustion phase and less THC concentration indicate that bigger swirl angle has good effect to ignition condition in the pre-chamber. Fig. 12: effect of swirl angle to Performance 3.3.5 The effects of Inclination angle No significant difference is found as shown in fig.13. In general, too narrow or wide jet can cause increase of wall heat loss and the angle should be designed considering interaction with piston shape. With the piston shape in this measurement, 60 to 70 degree seems to make little difference. 266 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="267"?> Fig. 13: effect of inclination angle to Performance 3.3.6 Best variant and spark timing optimization for fuel consumption Through measurement of all variations, V6 (variant of smallest A/ V with constant volume), has the most stable combustion and the lowest fuel consumption. Fig.14 shows the comparison of standard plug, base design V0 and V6. The advantage get bigger by optimizing injection setting and spark timing with a bit higher NOx emission. For further optimization, smaller A/ V and bigger swirl angle might lead to better fuel consumption according to the results above. Fig. 14: performance comparison of base design and best design 267 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="268"?> 3.4 EGR operation Fig. 15 shows COV comparison for all design variants. Tendency for almost all variant is that combustion gets unstable at around EGR 40%. From this result, ignition and flame propagation inside pre-chamber are assumed to limit maximum EGR rate. Variants with small swirl angle, big central hole and small A/ V have worse ignition stability. Big center hole has bad effect for air lean operation as well. But small swirl angle and small A/ V don’t have bad effect in lean condition. The reason of this different tendency is considered as below. Fig. 15: COV comparison for all design variant of EGR operation 3.4.1 A/ V Fig. 16 shows the result of A/ V variation with constant volume. Smallest A/ V shows worst performance while it was the best for lean operation. THC emission and burn duration get worse with the smallest A/ V. from this fact, the smallest A/ V may leads worse jet penetration because of big pressure drop through small holes. 3.4.2 Swirl angle The smallest swirl angle resulted in poor combustion stability as shown in Fig. 17. According to 3D-CFD result, TKE inside pre-chamber at around ignition timing is much larger. Fig.18 shows, gas flow through small swirl angle into pre-chamber seems to cause turbulence at the center of the pre-chamber. Therefore stability of combustion in early phase seems to be worse additionally in difficult condition with high EGR ratio. Hence swirl angle contribute to ignition and early combustion propagation. 268 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="269"?> Fig. 16: effect of A/ V to performance for EGR operation Fig. 17: effect of Swirl angle to performance for EGR operation 269 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="270"?> Fig. 18: TKE comparison at -30deg.ATDC for swirl angle variation 3.4.3 Air/ fuel mixture injection pressure Air/ fuel mixture injection pressure can be controlled by adjusting air supply pressure. Air is supplied continuously at regulated pressure. Purging residual gas from pre-cham‐ ber is expected to be enhanced by increasing air/ fuel mixture injection pressure. This comparison was done with same injection setting. As shown in Fig. 19, Thanks to additional 1 bar higher pressure, COV keep under 3% up to around 41.5% EGR, gaining 1.5pt. Burn duration got faster as well. 3D-CFD was conducted to analyse this result as shown in Fig. 20. By increasing purge pressure, both gas flow from injector to pre-chamber and from pre-chamber to main chamber increase. As a result, residual gas concentration in pre-chamber decreases, and keep low by 3pt at -30deg.ATDC as ignition timing. For further improvement, not only increasing purge pressure, but also delay or extend injection duration would be important. However, required effort to provide higher purge pressure must be compromising with the thermodynamic advandage. Fig. 19: effect of Air/ fuel mixture injection pressure to performance for EGR operation 270 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="271"?> Fig. 20: comparison of behaviour of gas interaction between pre-chamber and main chamber, residual gas distribution at -30deg.ATDC 4 Conclusions In order to obtain knowledge to optimize specification and settings of active pre-cham‐ ber with mixture injection system, pragmatic investigation was carried out. Single cylinder engine measurement with pre-chamber design variation was conducted to obtain engine performance results and cold flow CFD supported pre assessment and analysis of measurement results. The conclusions can be summarized as below: • High tumble is important for fast combustion even for engine with pre-chamber. • A/ V (ratio of hole area to pre-chamber volume) and hole number are important factors for combustion efficiency and stability. • Pre-chamber volume should be as small as possible considering NOx emission. • Purging residual gas from pre-chamber is more sensitive with EGR combustion because of ignitability, therefore suitable design and setting is different from lean combustion. • High air pressure supplied to pre-chamber injector is effective to extend EGR limit. As shown above, active pre-chamber with mixture system is one of key technologies for future diluted combustion engine. To get the best out of the system, further understanding and optimization are necessary. In addition, engine specification such as compression ratio can be adjusted with the system to achieve even higher thermal efficiency. 271 Development of Active Pre-chamber with Mixture Injection System for Diluted Gasoline Engine <?page no="272"?> References [1] P. Luszcz, et al, “Homogeneous Lean Burn Engine Combustion System Development Concept Study”; 18th Stuttgart International Symposium, 13-14 th March 2018 [2] T. Tomoda, “Approach of Gasoline Engine Combustion Development in Toyota and Future Outlook”, Journal of the Combustion Society of Japan Vol.60 No.191 (2018) 27-34 [3] Marc Sens, et al, “Pre-Chamber Ignition as a Key Technology for Highly Efficient SI Engines - New Approaches and Operating Strategies”, 39. International Wiener Motor symposium 2018, [4] Marc Sens, et al, “Pre-Chamber Ignition and Promising Complementary Technologies”, 27th Aachen Colloquium Automobile and Engine Technology 2018 [5] J.Benajes, et al, “Computational assessment towards understanding the energy conversion and combustion process of lean mixtures in passive pre-chamber ignited engines”, Applied Thermal Engineering, 178 (2020), 115501 [6] O.Benoit, et al, “LES Simulations of turbulent jet ignition in a constant volume vessel pre-chamber - chamber assembly”, Large-Eddy Simulation for Energy Conversion in electric and combustion Engines, 17-18 th June 2021 [7] A. Bianco, et al, “Modelling of combustion and knock onset risk in a high-performance turbulent jet ignition engine”, Transportation Engineering 2 (2020) 100037 272 Sho Tomita, Yann Drouvin, Michael Günther, Mario Medicke, Lorenz von Römer, Ronny Trettin <?page no="273"?> Evaluation of the lean limit extension provided by H2 direct injection inside the prechamber of a TJI engine by mean of detailed CFD simulations. Alfio Siliato a* , Riccardo Sgarangella a , Michela Fabbri a , Leonardo Pulga a , Claudio Forte a , Gian Marco Bianchi b a b * NAIS s.r.l., Bologna (Italy) Department of Industrial Engineering, University of Bologna contact: alfio.siliato@naisengineering.com Abstract: To cope with ever more stringent legislations in terms of CO2 and pollutant emissions, now and in the next years, the ICE development community must develop new technologies to reduce fuel consumption while maintaining the highest level of performance. Recently, one of the most advanced ignition systems, the prechamber turbulent jet ignition, derived from the F1 technology was introduced to the market, paving the way for more developments that could affect not only, but also high-per‐ formance passenger-cars. The full potential of this technology, however, is still to be unleashed, and a lot of research can be done to this aim. In this context, the present work proposes an ignition concept for GDI engines with a fueled prechamber, using ethanol as main fuel, and hydrogen injected directly inside the prechamber body. This combination would allow to run effectively using biofuels at leaner conditions, in compliance with the decarbonization trend, while assuring the combustion stability and power output required by the automakers also in difficult operating conditions, such as strong part load and catalyst heating conditions. On the other hand, this concept would allow to leverage the growing diffusion of hydrogen as an energy carrier for transportation, without the need for a complete redesign of the vehicle, considering the reduced amount of mass that should be introduced inside the prechamber with respect to the quantity that would be required to run the entire engine. After introducing the concept, in the present work, by mean of 3D-CFD simula‐ tions with detailed chemistry and a-priori combustion analysis, the benefits of this technology are presented in terms of lean limit extension and stability improvement in a part load operating condition, traditionally difficult to perform efficiently with a prechamber engine. The results are finally compared with other configurations (traditional spark ignition, passive prechamber) to underline the benefits that would come by developing further this concept. <?page no="275"?> Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model Tim Russwurm, Tobias Achenbach, Michael Wensing Abstract: Pre-Chamber Ignition Systems are a way to enable homogeneous lean or diluted combustion. Both strategies can significantly increase the efficiency of SI engines. The spark-initiated combustion inside the pre-chamber generates hot gases, which penetrate the cylinder fast and ignite the diluted charge on multiple sites. Active fueled pre-chamber ignition systems can influence directly on the composition in the pre-chamber to ensure good ignitability and sufficient ignition energy. A key aspect for passenger-car applications is the usage of volatile fuel components to scavenge the pre-chamber. These gases are very suitable as a pre-chamber fuel due to the facilitated mixture preparation and the higher, easier to dose total volume compared to liquid fuel. To understand more about the ignition processes, engine measurements were performed using an active pre-chamber with integrated pressure transducer. The scavenging is adapted in timing, pressure and composition to meet the different requirements of lean and EGR-diluted combustion. Whereas in lean operation the scavenging contains a high proportion of fuel in combination with low scavenging pressure and early timing, high amounts of exhaust gases in the cylinder require high scavenging pressure with near-stoichiometric composition. For both strategies different fuel systems are used to provide the appropriate pre-chamber scavenging. These engine measurements are used to calibrate a 1D-simulation in GT-Power to gain deeper insight into the combustion process. Vibe combustion models in the pre-chamber and heat transfer coefficients calculated from the engine measurements are used to match up with the measured pressure curves. Special focus is on the right representation of the pressure peak inside the pre-chamber after ignition. With this model, the air/ fuel-ratios inside the pre-chamber at ignition timing can be calculated. This is a decisive parameter in the tuning of the demanding firing process. <?page no="276"?> 1 Introduction An increase in the thermal efficiency of an ideal gasoline engine can be achieved by lifting the compression ratio and the isentropic coefficient of the used working gas. To make use of the resulting positive effects, the engine has to be able to deal with the difficulties coming along with these measures. High compression leads to an enhanced tendency to knock, while an increase of the isentropic exponent is usually achieved by running the engine highly diluted via enleanment or EGR, what results in a weak flammability of the mixture. The engine should be able to run the process as near as possible to the idealised otto-cycle, defined by compression ratio and isentropic coefficient. Concerning this, the typical drawbacks are uncomplete and non-isochoric combustion, wall heat losses, pumping losses and a decrease in the isentropic coeffi‐ cient with rising temperature. Enabling the engine to run on high isentropic coefficients not just increases the ideal thermal efficiency, but also leads to reduced wall heat and pumping losses because of downscalded combustion temperatures and dethrottling of the engine. On the other hand it gets more difficult to achieve isochoric and complete combustion with diluted mixtures. Using a pre-chamber ignition system which can provide a large amount of energy by igniting a small dose of fuel under ideal conditions in a separate chamber allows reliable, complete and fast combustion of diluted mixtures in the main chamber. Ignition in the main cylinder is achieved by the hot jets exiting the pre-chamber. Fast combustion is paralleled by igniting the main chamber charge before knocking occurs, allowing elevated compression ratios even under high engine loads. Stadler, Sauerland et al. [1] report of 60% faster combustion with pre-chamber systems compared to conventional SI engines. Besides mixture enleanment, EGR or high compression, pre-chamber ignition also can help to facilitate other strategies of operation facing reduced ignitability of the main chamber charge like water-injection, miller valve timing, or extended use of intercoolers. Nevertheless, pre-chamber ignition comes along with some drawbacks mainly resulting from an increase of wall heat transfer in the main chamber, enlarged heat transfer surface and uHC emissions. These drawbacks can be reduced by proper geometrical design and an optimized operation strategy. In order to utilize the pre-chamber ignition in a useful way, a multitude of variants are subject of current research projects. A distinction is made between pre-chamber systems with active fuel enrichment and passive systems without a separate purging option. Passive pre-chambers are easier and less expensive to integrate in existing engine architectures, but clearly feature a less extended lean-burn limit as shown in [2] and inferior scavenging, reducing the feasible EGR rate [3]. This especially becomes a hindrance facing the fact that high dilution has the potential to compensate the intensified wall heat losses of a pre-chamber ignition system. The in [3] observed scavenging and heat loss disadvantages linked to passive pre-chambers lead to no efficiency gains under part load compared to a well-designed SI engine. To retain the performance drawbacks of passive pre-chambers under part load conditions, concepts exists which use a second spark plug in the main chamber 276 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="277"?> allowing to operate in a more conventional SI mode. This concept for example was chosen by the car manufacturer Maserati for their 2020 introduced new supercar engine [4]. Under part load the combustion is controlled by varying the time offset between firing the spark plug in the main chamber first and the pre-chamber spark plug in a second step. This strategy especially allows for catalyst heating when the engine is cold. Passive pre-chamber systems can unfold their full potential only under high load. For active pre-chambers different approaches are considered distinguished mainly by liquid gasoline or gas scavenging. Both variants were considered in [5], resulting in a misfire limit of λ = 3, NO x raw emissions of less than 0.1 g/ kWh and reduced fuel consumption of up to 10% [5]. The thermodynamic performance of both variants turned out to be comparable, but focus of the study was on the application in a hybrid power train targeting the best point of the engine. Under part load the realizable air-fuel equivalence ratio is typically reached below λ = 2 [6,7,8], what elucidates the necessity to further investigate the performance of pre-chamber ignition systems at part load to also improve performance for conventional applications. As highlighted by Serrano, Zaccardi et al. [9] a crucial condition for the success of running a gasoline engine in the scope of λ >= 2 is a homogenous mixture, as too lean areas can stop the flame propagation. Therefore, the usage of high pressure port injection coupled with induced charge motion via the inlet valves or tumble flaps etc. can offer advantages regarding the results achieved in [9]. To further improve the EGR tolerance of pre-chamber ignition systems studies have been undertaken using an additional air supply to the pre-chamber, as described in [8,10]. This system is called Dual Mode Turbulent Jet Ignition and allows to exhale residual gas remaining in the pre-chamber. Residuals lead to a significantly lower dilution tolerance compared to lean operation. Intensive scavenging of the pre-chamber therefore enables the engine to reach dilution rates around 40%. Following [8,10], this results in similar good thermal efficiencies as achieved with high amounts of excess air, but still offers the advantages of using a three-way catalyst. Apart from the challenges concerning the stability of combustion under high dilution, also the necessary ability to allow late ignition under low load in order to heat the catalyst when running a cold engine constitutes a key challenge for practical applicability. [11,12,13] could demonstrate that it is possible to operate a pre-chamber ignited engine in the whole map of operation, while being especially able to perform catalyst heating strategies. Dimensioning of pre-chambers is settled in an area of conflict between a fast combustion in the main chamber, providing a near stoichiometric, homogenous mixture in the pre-chamber and the increasing wall heat losses due to heavy turbulence induced by the jets and larger surface area. Thus Yu, Zhang et al. [3] figured out that pre-chamber designs which could achieve the fastest combustions don’t provide the best efficiencies. An increase in releasable energy usually goes along with an increase in the pre-chamber volume and fluid wall interaction as shown in [14]. Already Gussak et al. [15,16] reported that pre-chamber volumes of 2-3% of the main chamber volume and a transfer port cross-section area of 0.03… 0.04 cm²/ cm 3 pre-chamber volume show 277 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="278"?> the best results. Most of the currently tested prechambers are still in this scope. Yu, Zhang et al. [3] claim that they were able to minimize additional heat losses, while maintaining fast burn rates by increasing the number of nozzles keeping the total nozzle area unchanged. Also Atis, Chowdhury et al. [8] confirm the general observation that smaller orifices provide shorter combustion durations. The influence of orifice diameters has been already categorised by Yamaguchi et al. [17]. Following [17] the orifices should be large enough to avoid quenching in order to achieve thermal and chemical ignition in the main-chamber. Tudela, Barrowet al. [18] confirm this result through illustrating the negative effects on the main chamber combustion resulting from quenching and long reignition times as a result of too strong jets. Jet penetration length and jet velocity generally increase with smaller orifice diameters [12]. The intensity of the generated jets influences the kind of fluid-wall interaction in the main chamber. For example impingement with the cylinder walls results in large scale vortices which accelerate the combustion of the main chamber charge at later times, but also contribute to high wall heat losses. Moreover it was found that local flame-flame interactions close to the nozzle and the lower main chamber wall exist and can increase the propagation speed up to six times compared to planar flames [19]. The quality of the jets furthermore strongly depends on the mixture in the pre-cham‐ ber itself [20]. As stated in [20], the pre-chamber charge should burn at stoichiometric conditions to supply plenty active radicals, whereas rich pre-chamber charges lead to a significant amount of low-carbon species being rejected to the main chamber. Serrano, Zaccardi et al. [9] point out the impact of the main injection mode in combination with the in-cylinder charge motion on the mixture homogenization in the pre-chamber. Results of their CFD-Simulations reveal that an essential part of the fuel inside the pre-chamber derives its origin from the main injection. Summarizing the influences on the pre-chamber mixture before combustion Serrano, Zaccardi et al. [9] list three points: • the transport mechanisms during the flow of the air-gasoline mixture from the main chamber into the pre-chamber • the timing of the main injection affects the fuel distribution around the pre-cham‐ ber orifices • the quantity of gasoline injected into the pre-chamber Thereby the injection of a small amount of fuel into the pre-chamber remains the main challenge. Stadler, Wessoly et al. [12] point out the recognised relevance of the interaction between intake air motion and the mixture formation in the pre-chamber on the emission of particles. Silva, Sanal et al. [21] focus attention on the occurring backflow into the pre-chamber early in the main combustion. This backflow goes along with another combustion and the trapping of partially burned species. Biswas and Qiao [22] conducted experiments igniting ultra-lean premixed H 2 / air mixtures. Besides testing straight and convergent nozzles, also converging-diverging nozzles were used. Inspecting supersonic jets, they observed Mach diamonds and zones of 278 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="279"?> high temperature downstream the shocks which are supposed to further reduce the misfire limit of ultra-lean main chamber charges. Pan, He et al. [23] investigated the influence of temperature conditions on the pre-chamber performance for a stoichiometric pre-chamber and a lean main chamber mixture. Performing 2D DNS simulations, they identified different types of flame propagation depending on the temperature conditions. In the range of 900 K up to 1000 K the pressure rise in the prechamber induces a flow of unburned mixture entering the main-chamber prior to the flame front. Such generated areas of strong turbulence in the main-chamber are subsequently reached by the flame front and the main combustion starts. The authors refer to this mechanism as a normal turbulent jet flame propagation. For further increase in the temperature of up to 1100 K a pre-flame heat release can be spotted in the jet orifices. Higher temperature levels in the scope of up to 1200 K lead to autoignition and supersonic deflagration inside the pre-chamber. This results in faster flame propagation, greater heat release rate and larger flame thickness. The flame development in the main chamber related to occurring autoigniton processes is described as spherical flame propagation. The focus in this work lies on the performance of a gasoline engine running on lean (up to λ = 2) and stoichiometric EGR diluted mixtures. The ignition system is a self-developed active pre-chamber ignition system. For the purpose of enriching the pre-chamber with fuel, volatile components of gasoline are used. This gasoline-va‐ pour-air-mixture is mainly composed of butane and pentane with a maximum concen‐ tration of approx. 60 wt. % at room temperatures [24]. This highly fuelrich mixture is very suitable for pre-chamber enrichment in lean engine operation. The authors showed in previous publications that stable combustion is possible up to λ = 2 in part load operation. For EGR-diluted combustion processes, this work uses a new developed fuel system to scavenge the pre-chamber with a near stoichiometric mixture of gasoline vapour and air at an elevated scavenging pressure of 9 bar. These engine test results are used to calibrate a 1D Simulation in GT Power in order to gain more detailed information about the thermodynamic behaviour of the engine running with this pre-chamber system. 2 Methodology 2.1 Test engine setup As test engine a common 1.8 L four-cylinder DI-engine with turbocharger and variable intake and exhaust valve phasing was used. The engine was operated with 100 bar injection pressure and 95 °C ± 1 °C coolant temperature. The oil temperature is limited to 95 °C ± 2 °C by the internal oil-water heat exchanger of the engine. To reduce the samples of prototypes only one cylinder is fired, while the other three are closed by a sealing plate in the intake. The engine has a displacement of 450 cm 3 per cylinder and a compression ratio of 9.6. The direct fuel injector is placed in a lateral position. The operation point is kept steady over all measurements to 1500 rpm 279 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="280"?> and 4.5 bar indicated mean effective pressure (IMEP). This point was identified as a medium load point with high relevance to WLTP fuel consumption earlier [24]. The engine is equipped with a low-pressure sensor in the intake and tempered high pressure sensors Kistler 6041B in the cylinders. Both, the pressure in the pre-chamber and in the cylinder, were measured with a resolution of 0.36° CA and averaged over 200 cycles. All energetic calculations include the fuel mass flow of the combustion chamber and the pre-chamber, whereby both are measured by coriolis mass flow meters. The amount of fuel injected to the main chamber is controlled by varying the beginning of the injection event. Average mass flowrate and end of injection (270° CA BTDCF) are held constant. The engine speed is prescribed by the coupled electric machine. Ignition timing is set such that the anchor angle (MFB 50%) is at 8° CA ATDCF. The amount of mass injected directly into the pre-chamber is determined such that the engine runs as stable as possible. 2.2 Pre-chamber fuelling system This work investigates both lean and EGR-diluted combustion at low engine load with active pre-chamber ignition. The authors showed in an earlier publication a pre-chamber fuelling system that supplies a fuel rich gasoline-vapour-air-mixture at max. 3 bar to the pre-chamber. As this system is very suitable for pre-chamber enrichment in lean operation, the same fuel system is used in this work. As EGR-diluted combustion with an active pre-chamber has different requirements to the pre-chamber scavenging gas, an entirely new system was developed for this operation. In Figure 1 the working ranges of the two fuel systems are illustrated. Both fuel systems together can provide a wide range of scavenging gas for the active pre-chamber. For lean operation of the engine the pre-chamber is scavenged with a rich mixture, which can only be achieved under moderate pressures (steam pressure). The scavenging takes place during the intake stroke in order to achieve a dilution of the rich pre-chamber charge in the compression stroke with the lean main chamber charge. For EGR-diluted operation a different approach is followed. To prevent burned residuals inside the pre-chamber hampering combustion, the scavenging is set late in the compression stroke. This comes along with the necessity of higher scavenging pressures. Higher pressure levels of the scavenging gas limits the possible amount of hydrocarbons in the mixture (steam pressure). The low share of hydrocarbons is not a problem, as the EGR-diluted mixtures being stoichiometric anyway demand for near stoichiometric scavenging. 280 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="281"?> Fig. 1: comparison of the working ranges of the two fuel systems Details on the fuel system for lean operation can be found in [25]. The working principle of the new laboratory fuel system for EGR-diluted combustion is shown in Figure 2. In an additional fuel tank air is conveyed through the liquid fuel and thereby enriched with volatile components of gasoline. This fuel tank is filled to approx. 50% with fuel and is continuously flushed with the liquid fuel for injection in the main combustion chamber. Since the tank contains under 1 L of volume, it is ensured that fuel with new light volatiles can be continuously introduced. The entire tank is brought to the desired pre-chamber scavenging pressure of approximately 7 to 10 bar via the compressed air supply. At this increased pressure the partial pressure of the fuel vapour in the gas phase decreases, resulting in a lower HC concentration. Further dilution of this mixture, especially at low purge pressure, can be achieved via an additional bypass valve. The supply of compressed air is provided externally in this laboratory system. To allow liquid fuel to flow continuously through the pressure tank, an additional fuel pump is installed to overcome the gas pressure in the fuel tank. The entire laboratory system is controlled via an industrial logic controller to ensure constant pressure and HC-concentration in the pre-chamber fuelling line. Several additional valves are installed to ensure laboratory safety, as this system handles an ignitable fuel vapour air mixture at elevated pressure. In order to intervene as little as possible in the engine's system, the fuel in this laboratory system is again expanded to approx. 5 bar before it is fed to the high-pressure pump at the cylinder head. This step is not necessary in a vehicle because this pump can be adapted to the increased pre-feed pressure. Likewise, the pre-feed pump from the tank can be upgraded to provide the required fuel pressure to fuel the evaporation tank. 281 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="282"?> Fig. 2: schematic diagram of the new fuel system In this investigation, the fuel system is set to 9 bar gauge pressure that results in a HC-concentration of 4.9 ± 0.2 wt.% in the gas phase with the used summer gasoline at 20 °C laboratory temperature. The used HC-sensor is butane calibrated and is designed to work in near ambient pressure conditions. Therefore, a separate gas line with reduced pressure is integrated in the fuel system to measure the HC-concentration. An HC-concentration of 5% equals λ = 1.2, so scavenging gas is in the slight lean area. 2.3 GT-Power Model The data achieved by the real engine tests is used to calibrate a GT-Power model. Important characteristics affecting the process inside the cylinder of the engine are mainly • initial state of the air • flow inside the cylinder • mass flow rates of air and fuel • uncomplete combustion • substance properties • combustion rates • leakage • wall heat losses • flow between main-and pre-chamber • gas exchange 282 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="283"?> Especially non isochoric, uncomplete combustion, wall heat losses and losses due to gas exchange are typically the major drawbacks causing the real process to differ significantly from an ideal otto-cycle. Despite most of the upper points are accessible for predictive simulation, the prognosis of combustion rates and wall heat losses without elaborate 3D-CFD simulations is difficult due to the complex ignition mechanism in a chamber engine. Besides the influence of the increased turbulence on the combustion rates in the main-chamber, also multiple ignition sites and reigniton processes of quenched jets affect the combustion. This makes prediction of combustion and wall heat losses difficult. Established non predictive models like the approach of Vibe, de‐ scribing combustion rates and the approach of Woschni, determining Nußeltnumbers inside the cylinder, were developed with focus on conventional engines, making the application in the context of pre-chamber ignition challenging. Furthermore, non-sto‐ ichiometric and EGR-diluted mixtures exacerbate this difficulty. To overcome these problems, the simulation is split up into a predictor and corrector run of the GT-model. The predictor simulation is performed with assumptions for combustion efficiencies in the preand main-chamber. The heat losses are calculated with the WoschniGT approach. The WoschniGT model is an improved version of Woschnis approach to better match up with EGR-operation during gas-exchange. Next a correction of the combustion efficiencies and the wall-heat losses is conducted. The simulation is then run again on these corrected values, to obtain the final results. The following remarks point out, what are considered to be the key aspects for a 0D/ 1D simulation under the focus of pre-camber ignition. Besides prescribing the correct ambient temperature and pressure, the provision of a correct initial air state at IVC is achieved by setting the temperature after the intercooler to the known value from the engine test. The intercooling process is complex to simulate and of no further interest for our investigations. Therefore, the correct temperature after the intercooler is achieved by setting the heat transfer coefficient inside the intercooler to an unrealistic high value. Prescribing the wall temperature equal to the desired air temperature, this procedure assures the air to adopt the correct temperature after the intercooler. Following the intercooler the air runs through the throttle body and the intake manifold. The pressure inside the manifold is known from the engine tests. In contrast to the control strategy of a real engine, the throttle body is not used to control the IMEP, but to hit the measured pressure inside the intake manifold in order to provide a correct initial state of the air. Giving a wide berth to global variables like the IMEP within the control strategy of the model, avoids compensation of errors at points far off their actual origins. Passing the intake manifold, the air flows through the intake ports and into the main chamber. The wall temperature of the intake ports is not known from measurements. The intake ports being already a part of the cylinder head have a high temperature and therefore an essential influence on the air condition. Using a further PID controller, the wall temperature of the intake ports is adjusted such that the average airflow after the intercooler provides the required amount of air to match up the measured air-fuel-ratio. 283 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="284"?> The total air mass is calculated from the exhaust lambda, the overall fuel mass flow to the engine and the air fuel ratio corresponding to the used gasoline. All these values are known in advance. To obtain the amount of air entering the main chamber via the intake ports, the air coming along with pre-chamber scavenging has to be subtracted from this overall air mass. Using two more PID-controllers, the average mass flow rates through the main chamber and pre-chamber injectors are assured to be in accordance with the measurements. The combustion rate in the main chamber is provided by the pressure analysis software and prescribed to the model as an array. Because the pre-chamber is an open system, it is difficult to gain correct combustion rates out of the measured pressure curves. Therefore Vibe curves are used to shape the progression of combustion inside the pre-chamber. The ignition timing is known and used as the starting point of combustion (Figure 3). The end of the pre-combustion is set to the crank angle at which the pre-chamber runs into under pressure for the first time after the peak pressure. This effect results from the inertia of the overpassing jets and can be seen in Figure 3 in form of the undercut of the pre-chamber pressure with respect to the main-chamber pressure. This point is chosen because it approximately represents the point of flow reverse from outflow to inflow. Several trials showed that the position of the pressure peak coincides best with the measured peak, when the anchor angle is set in the middle between the crank angle of the pressure peak and the position of the most steep pressure rise before the peak (Figure 3). With these three positions known, a Vibe parameter can be calculated. Fig. 3: characteristic points for vibe-curves 284 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="285"?> Based on a Vibe curve with 99.9% combustion efficiency, a vibe parameter of m=2.3 leads to an anchor angle in the middle between combustion start and end. The com‐ bustion efficiency is adjusted later on in GT-Power by scaling the y-axis maintaining the principle form of the 99.9% curve. Lowering the vibe parameter leads to anchor angles before the midpoint crank angle and vice versa. Depending on the relative position of the calculated anchor angle to the midpoint, the Vibe parameter is adjusted such that the position of the Vibe-curves anchor angle coincides with the calculated anchor angle (comp. Figure 4). Fig. 4: determining the vibe-parameter m 2.4 Pre-calculation of wall heat losses and combustion efficiencies For both combustion rates (main-chamber and pre-chamber) the combustion efficien‐ cies are not known. Calculation of these is performed in a second step basing on the results of the pre-simulation. For the purposes of pre-simulation the burned fuel fractions are set to 1.0 in the main chamber and to 0.9 in the pre-chamber. Based on the gas-exchange calculation of the pre-simulation and an array of compressibility factors depending on current temperature, the temperature in the main-chamber is calculated with the gas law and the pressure from engine measurements. These temperatures are considered to represent the real temperatures inside the cylinder and can be used to set up the balance of internal energy. Neglecting the change of kinetic energy inside the cylinder, the balance of energy for the main-chamber reads 285 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="286"?> dU dt = Q˙ wall + Q˙ comb + P piston + U˙ netto + pV˙ netto + E˙ kin, netto The temporal change of internal energy is a result of heat transfer, combustion, piston-work, transport of internal and kinetic energy with mass flow between the two chambers and the volume work done on the interface between mainand pre-chamber. With the heat release from combustion, this leads to the wall heat losses. The energy resulting from combustion Q comb depends on the known injected fuel mass and the unknown burned fuel fraction. To obtain the burned fuel fraction and therewith the wall heat loss, a balance of energy for the coolant in dependency of the burned fuel fraction is set up. The energy transferred to the coolant and the average wall heat losses are calculated together within a for-loop. After every iteration the burned fuel fraction is reduced by 0.001. The iteration stops, when the calculated cooling water loss and the measured cooling water loss are in accordance. The energy transferred to the coolant is approximately calculated with P coolant = P wall, MC + P wall, P C + P ports + P exℎaustmanif old + P f riction The power transferred to the coolant consists of the energy transfer from the cylinder walls P wall,MC , the pre-chamber wall P wall,PC , the ports P ports and the mechanical losses P friction . The test engine is equipped with an integrated exhaust manifold whose contribution is considered with P exhaust manifold . Following [7], the pre-chamber wall heat loss is estimated to be in the scope of 5% from the wall heat loss of the main chamber for stoichiometric conditions. The energy transfer from the ports and the exhaust manifolds to the coolant is taken from the pre-simulation results. Based on the so obtained wall heat loss, an array of heat transfer coefficients for the main chamber is calculated. With this array of heat transfer coefficients a correction function for the coefficients calculated with the approach of Woschni is introduced. This correction function serves as the array of heat transfer multipliers for GT-Power. The correction is limited to the range of closed valves. During open valves the values from the WoschniGT model are used. Figure 5 gives an overview of the procedure. The pressure peaks of the pre-chamber are adjusted to the measured pressure peak by reducing the amount of fuel burned during the pre-combustion. This is done by dividing the percentage of burned fuel during the pre-combustion by another correction factor. This factor is calculated based on the peak pressures p peak,pre-sim and p peak,measured occurring during pre-combustion. cor . = p peak, pr e − sim p peak, measur ed 2 . 4 Further critical points which have to be considered are: • heat capacity of the coolant • surface area - volume-relation of the engine 286 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="287"?> • average wall temperature of the cylinder Fig. 5: calculation of wall heat loss Figure 6 shows an example for the difference between the result obtained from pre-simulation and the final result. Fig. 6: comparison of preand corrector-simulation 3 Results 3.1 Variation of air-fuel equivalence ratio (λ) The engines performance was investigated for different air-fuel equivalence ratios reaching from λ = 1.1 up to the misfire limit of λ = 2. The average mass flow rates are shown in Figure 7a) and b). The air-fuel equivalence ratio is lifted by dethrottling the engine (Figure 7b)) and reducing the fuel mass flow of the main injector (Figure 7a). Based on the coefficient of variation (Figure 7d), the pre-chamber scavenging is adjusted to obtain an as stable as possible engine performance. With the aim of igniting lean main-chamber charges, the used scavenge gas is a very rich mixture and contains 43% hydrocarbons and 57% air. With increasing lean conditions in the main chamber, the scavenging is increased to achieve sufficient enrichment of the pre-chamber charge (Figure 7a and b). Figure 7c) shows the ignition timing and the anchor angle with changing air-fuel equivalence ratios. 287 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="288"?> Fig. 7: a) and b) average mass flow rates, c) ignition timing and anchor angle, d) coefficient of variation Figure 8 shows the development of the lambda values inside the pre-chamber during the engine process. The scavenging leads to values below 0.2, followed by a dilution with the main-chamber charge in the compression stroke. During the pre-combustion the air-fuel equivalence ratio drops and begins to increase again with backflow from the main-chamber. The achieved mixtures inside the pre-chamber are rich, what is beneficial at very lean main-chamber mixtures. The jets transporting some fuel to the lean main-chamber during pre-combustion is expected to improve ignition conditions inside the main-chamber. This is at the expense of slowed down pre-combustion and adverse for richer main-chamber charges in the scope of λ = 1. 288 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="289"?> Fig.8: λ inside the pre-chamber Figure 9 a) sums up the air-fuel equivalence ratios inside the pre-chamber at ignition time depending on the global air-fuel equivalence ratio. For global air-fuel equivalence ratios from λ = 1.4 to λ =2 the mixtures inside the pre-chamber are in the scope of λ = 0.7-0.8. The richer operation points 3 and 5 show a drop at ignition time near to λ = 0.6. Comparing this with the 10-90 combustion durations of the pre-chamber shown in Figure 9 b), this explains the slowed down pre-combustion for these two points. The combustion duration inside the main-chamber reaches a minimum value for λ = 1.4 and shows an increase for larger and smaller air-fuel equivalence ratios of the main-chamber charge. As an increase in combustion duration is generally expected for globally lean mixtures, one would expect a further decrease for the points 5 and 3, which still show values above λ = 1. The greater amount of fuel injected to the main-chamber at near stoichiometric mixtures, together with the rich jets from the pre-chamber (λ = 0.6) leads to locally rich conditions near the pre-chamber. As noticed in [26], this results in ignition occurring in the outer areas of the cylinder and combustion duration increases. 289 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="290"?> Fig. 9: a) λ inside PC at combustion start, b) 10-90 combustion durations The time offset between ignition inside the pre-chamber and the start of combustion in the main-chamber is quite long for the points 3 and 5. This can be seen at the very early ignition timing of the pre-chamber, while the main-chamber combustion still starts near TDCF (Figure 10 a)). This delay may also be caused by the supposedly rich conditions in the center of the cylinder, causing a long time span to achieve a lean enough mixture for ignition in the outer area of the cylinder. This is further enhanced by less scavenging at richer main-chamber conditions, leading to lower pre-chamber over-pressures resulting in slower jets. Figure 10 b) shows the indicated efficiencies in comparison to the combustion efficiencies. Despite continuously decreasing combustion efficiencies with increasing global λ, the indicated efficiencies keep rising until λ = 1.8. This is mainly a consequence of reduced wall heat losses, high isentropic coefficients as well as dethrottling of the engine. Besides the dilution of the mixture with excess air, the isentropic coefficients also rises because of the lower combustion temperatures. Fig. 10: a) ignition timing pre-chamber and combustion start main-chamber b) combustionand indicated-efficiencies 290 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="291"?> Figure 11 visualizes the difference between a wall heat loss curve, calculated with the approach of WoschniGT, and a corrected wall heat loss. Fig. 11: comparison of corrected wall heat loss and WoschniGT wall heat loss The wall heat loss during compression and the peak-value are significantly higher, compared to the estimation in the pre-simulation. During expansion the wall heat loss is smaller. Apart from that, this figure illustrates that the pre-chamber ignition shifts the peak towards TDFC. This is critical because energy lost near TDFC has a higher potential to do work on the piston than energy lost in the later phase of expansion. Another characteristic feature is the rise in the wall heat loss during pre-combustion as a result of turbulence induced by the overpassing jets. Lean mixtures provide the potential to compensate the high wall heat losses coming along with pre-chamber ignition. This is illustrated in Figure 12 giving a comparison between two wall heat loss curves, covering the global air-fuel equivalence ratios λ = 1.1 and λ = 2. The reduction with increasing λ is a result of decreasing combustion temperatures. Not only the higher mass of diluted mixtures causes this behaviour, but also the continuously dropping combustion efficiencies resulting from hampered combustion (compare Figure 10 b). 291 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="292"?> Fig. 12: comparsion of wall heat losses for different λ The fluid wall interaction induced by the jets can also be seen in the measured combustion rates. After ignition of the pre-chamber, the pressure analysis software calculates negative values (Figure 13). In processes with long ignition delays, like in naturally aspired diesel engines, similar undercuts can be observed as a result of high injection pressures and evaporation of the fuel. Fig. 13: visible wall heat loss in measured combustion rate The decreasing total average wall heat losses with increasing global air-fuel equiva‐ lence ratio are shown in Figure 14a). The average wall heat loss inside the pre-chamber remains more or less constant due to the very similar mixture conditions as shown in Figure 9 a). This leads to an increasing share on the main-chamber wall heat loss with increasing global lambda (Figure 14b). 292 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="293"?> Fig. 14: a) average wall heat loss MC b) share of PC wall heat loss on MC wall heat loss Summarizing, the wall heat losses in the main chamber show a characteristic lift during pre-combustion. After that, the heat transfer rises steep with starting main-combustion for moderate dilution. This behaviour is likely to be a consequence of the ignition at multiple sites at the same time and higher turbulence. Erstwhile investigations showed that the combustion starts from the outer areas of the cylinder for low global air-fuel equivalence ratios and from the center of the cylinder for lean mixtures. This means that the temperature rise at the multiple ignition sites in richer mixtures is very close to the cylinder walls, what may explain the abrupt rise in heat transfer. Caused by the reduced temperature level and ignition occurring in the center of the main-chamber (compare [26]), the characteristic of the highly diluted mixtures is much less distinct. Fig. 15: comparison of measured and simulated pressure curves 293 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="294"?> Figure 15 shows a comparison between the measured and simulated pressure curves for mainand pre-chamber. The first two pressure curves are the results for a global λ = 1.1. The two following curves represent the pressures for a global lambda of λ = 2. The stronger pre-chamber activity for the lean point can be seen clearly on the higher pressure rise in the chamber during pre-combustion. 3.2 EGR In a further test the effects of internal EGR were investigated. The phasing of the camshafts was limited to 30°. The pre-chamber was scavenged with a gas consisting of 5% hydrocarbons and 95% air. As described in 2.2, scavenging was conducted with pressures of 6 to 7 bar during the compression stroke. Figure 16 a) and b) sum up the mass flow rates. Figure 16 c) shows the ignition timing and the anchor angles. In Figure 16 d) the coefficients of variation are displayed. Fig. 16: a) and b) average mass flow rates, c) ignition timing and anchor angle, d) coefficient of variation The development of the air-fuel equivalence ratio inside the pre-chamber during the engine process is shown in Figure 17. Due to the very lean scavenging gas, the air-fuel equivalence ratio reaches values above λ = 1.2 at early compression. After that, enrichment with the main-chamber charge during the remaining compression stroke can be observed. With starting pre-combustion the air-fuel equivalence ratio drops 294 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="295"?> below λ = 1. Backflow from the main-chamber then lifts the air-fuel equivalence ratio again near to λ = 1. The achieved mixtures inside the pre-chamber at ignition time are lean, which is advantageous for the ignition conditions in the main-chamber at global stoichiometric mixtures. This is again because of locally richer conditions in the center of the cylinder. Fig. 17: λ inside the pre-chamber The resulting air-fuel equivalence ratios inside the pre-chamber at ignition time are shown in Figure 18 a). The values remain almost constant for all considered EGR-rates. The combustion duration in the mainand pre-chamber continuously increases as a result of hampered combustion due to burned residuals (comp Figure 18 b)). In contrast to the combustion durations of the lean operation points, the combustion durations of the EGR diluted points show an increase with higher dilution and are generally longer. 295 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="296"?> Fig. 18: a) λ inside pre-chamber at ignition time b) 10-90 combustion durations of preand main-com‐ bustion Figures 19 b) shows the almost linear rise of burned residuals inside the pre-chamber with increasing EGR-rates. The indicated and combustion efficiencies are shown in Figure 19 a). Besides two exceptions (2 and 1), the combustion occurs to be almost complete for all considered EGR-rates. The gain in efficiency is again a result of reduced wall heat losses, dethrottling and an increase in the isentropic coefficient due to the reduced temperature level. Fig. 19: a) combustion and indicated efficiencies b) exhaust gas mass fraction PC at ignition time Figure 20 shows a comparison between a WoschniGT curve and a corrected wall heat loss for an amount of 19% mass fraction of burned residuals at combustion start in the main-chamber. 296 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="297"?> Fig. 20: comparison of corrected wall heat loss and WoschniGT wall heat loss Compared to the Woschni estimated wall heat loss of the pre-simulation, the character‐ istic features for the shown EGR rate are again an increase during pre-combustion due to the turbulence of the overpassing jets, a steep rise with starting main combustion and a smaller wall heat loss during expansion. The abrupt rise in wall heat loss with starting main combustion may be again explained with the combustion starting near-wall from the outer areas of the cylinder at stoichiometric conditions. The development of the wall heat losses is visualized in Figure 21. As a result of the hot temperatures inside the cylinder, the wall heat losses are very high for small shares of residual gas and decrease as expected for higher shares of residual gas. Fig. 21: comparison of wall heat losses for different EGR-rates 297 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="298"?> For all EGR-rates the wall heat losses are larger than the largest average wall heat loss of the air-diluted main-chamber charges. Compared to lean operation, this results mainly from hotter combustion temperatures due to less dilution and a lifted temperature level due to the hot residual gases. Fig. 22: a) average wall heat loss MC b) share of PC wall heat loss on MC wall heat loss Figure 23 shows a comparison between the measured and simulated pressure curves. The curves show the obtained results for a minimum EGR-rate of 11% and a maximum EGR-rate of 29%. Due to the long combustion durations inside the pre-chamber for high EGR-rates, the pre-chamber overpressure keeps reducing with higher dilution. At 29% EGR-rate only a small lift of the chamber pressure during pre-combustion remains. As the combustion is very stable, this is not an issue for the combustion process. Fig. 23: comparison of measured and simulated pressure curves 298 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="299"?> 4 Conclusion The authors show an active scavenged pre-chamber ignition system that can be operated both in lean operation up to λ = 2.0 and with high EGR dilution up to 28%. This is achieved by using volatile components of gasoline. For this purpose two different laboratory fuel systems which provide the appropriate scavenging pressure and HC concentration depending on the operating strategy are used. These measurements are evaluated with an advanced 1D-simulation. The representation of the heat transfer to the cylinder wall with the WoschniGT model is improved via a crank angle resolved correction factor for the heat transfer coefficient. The combustion rate inside the pre-chamber is modelled with a separate vibe curve. With this modelling, a good representation of the measurements in the 1D simulation is achieved, in which both the energy and mass flows as well as the pressure curves are reproduced accurately. The crank-angle resolved calculation of the air-fuel-ratio λ inside the pre-chamber shows that in lean operation the pre-chamber is flushed with a rich scavenging gas that is diluted with excess air from the main chamber during the compression stroke, leading to a λ = 0.8 at ignition timing. The mixture conditions inside the pre-chamber at ignition timing for EGR-diluted operation are in the scope of λ = 1.1. These values were not known during the measurement and are thus not predefined. The scavenging of the pre-chamber was adjusted to minimize the fuel consumption as long as the combustion stability is sufficient. The wall heat losses in the main-chamber are strongly influenced by the turbulent jets and multiple ignition sites. The main characteristics for moderate dilution turned out to be a lift in the heat transfer during pre-combustion and a steep rise with starting main-combustion due to multiple ignition sites occurring near the cylinder walls. With increasing dilution the wall heat losses could be significantly reduced. In this investigation, the lean combustion showed higher values of the indicated efficiency up to 36.5% whereas the EGR dilution only achived 33.3%. However, this engine setup did not reach the EGR-limit, so further improvements for example via external EGR are possible. 5 Acknowledgements The authors gratefully acknowledge funding of this work by the Bavarian Research Foundation in the project AZ1378-19 “LEANition” as well as the Erlangen Graduate School in Advanced Optical Technologies (SAOT). Special thanks are due to Simon Stecker and Johannes Hebrank for their assistance during the measurements. 6 Literature [1] Stadler, H. Sauerland, M. Härtl, G. Wachtmeister, The Potential of Gasoline Fueled Pre Chamber Ignition Combined with Elevated Compression Ratio, in SAE Technical Paper, 2020. 299 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="300"?> [2] F. Bozza, V. De Bellis, D. Tufano, E. Malfi, C. Müller, K. Habermann, 1D Numerical and Experimental Investigations of an Ultralean Pre-Chamber Engine, in: SAE Int. J. Engines, 2020. [3] X. Yu, A. Zhang, A. Baur, N. Engineer, The Impact of Pre-Chamber Design on Part Load Efficiency and Emissions of a Miller Cycle Light Duty Gasoline Engine, in: SAE Technical Paper, 2021. [4] Maserati: Benzinverbrennungsmotor mit einer Verbrennungsvorkammer und zwei Zünd‐ kerzen, in: Europäisches Patentamt, EP 3 453 856 B1, 2017 [5] C. Müller, S. Pischinger, S. Tews, A. Müller, K. Habermann, Analysis of experimental results with an active pre-chamber ultra-lean burn SI engine, in: International Journal of Engine Research, 2020. [6] J. Hua, L. Zhou, Q. Gao, Z. Feng, H. Wei, Influence of pre-chamber structure and injection parameters on engine performance and combustion characteristics in a turbulent jet ignition (TJI) engine, in: Fuel, 2021. [7] J. Benajes, R. Novella, J. Gomez-Soriano, I. Barbery, C. Libert, F. Rampanarivo M. Dabiri, Computational assessment towards understanding the energy conversion and combustion process of lean mixtures in passive pre-chamber ignited engines, in: Applied Thermal Engineering, 2020. [8] C. Atis, S. Sabah Chowdhury, Y. Ayele, T. Stuecken, H. Schock, A. K. Voice Ultra-Lean and High EGR Operation of Dual Mode, Turbulent Jet Ignition (DM-TJI) Engine with Active Pre-Chamber Scavenging, in: SAE Technical Paper, 2020. [9] D. Serrano, J. Zaccardi, C. Müller, C. Libert, K. Habermann, Ultra-Lean Pre-Chamber Gasoline Engine for Future Hybrid Powertrains, in: SAE Int. J. Adv. & Curr. Prac. in Mobility, 2020. [10] C. Atis, H Schock, Comparison of Excess Air (Lean) vs EGR Diluted Operation in a Pre-Chamber Air/ Fuel Scavenged Dual Mode, Turbulent Jet Ignition Engine at High Dilution Rate (40%), in: SAE Int. J. Adv. & Curr. Prac. in Mobility, 2021. [11] M. Bunce, A. Cairns, H. Blaxill, The use of active jet ignition to overcome traditional challenges of pre-chamber combustors under low load conditions, in: International Journal of Engine Research, 2020. [12] A. Stadler, M. Wessoly, S. Blochum, M. Härtl, G. Wachtmeister, Gasoline Fueled Pre-Chamber Ignition System for a Light-Duty Passenger Car Engine with Extended Lean Limit, in: SAE Int. J. Engines, 2019. [13] A. Cooper, A. Harrington, M. Bassett, S. Reader, M. Bunce, Application of the Passive MAHLE Jet Ignition System and Synergies with Miller Cycle and Exhaust Gas Recirculation, in: SAE Technical Paper, 2020. [14] M. H. B. Sandoval, C. E. C. Alvarez, V. Roso, N. D. S. A. Santos, The influence of volume variation in a homogeneous prechamber ignition system in combustion characteristics and exhaust emissions, in: Journal of the Brazilian Society of Mechanical Sciences and Engineering, 2020. [15] L. Gussak, M. Turkish, D. Siegla, High Chemical Activity of Incomplete Combustion Products and a Method of Prechamber Torch Ignition for Avalanche Activation of Combustion in Internal Combustion Engines, in: SAE Technical Paper, 1975 300 Tim Russwurm, Tobias Achenbach, Michael Wensing <?page no="301"?> [16] L. Gussak, V. Karpov, Y. Tikhonov, The Application of Lag-Process in Prechamber Engines, in: SAE Technical Paper, 1979. [17] S. Yamaguchi, N. Ohiwa, T. Hasegawa, Ignition and burning process in a divided chamber bomb, in: Combustion and Flame, 1985. [18] W. V. Tudela, C. Barro, K. Boulouchos, Investigations on spark pre-chamber ignition and subsequent turbulent jet main chamber ignition in a novel optically accessible test rig, in: International Journal of Engine Research, 2021. [19] S. Benekos, C. E. Frouzakis, K. G. Giannakopoulos, C. Altantzis, K. Boulouchos, A 2-D DNS study of the effects of nozzle geometry, ignition kernel placement and initial turbulence on pre-chamber ignition, in: Combustion and Flame, 2021. [20] W. Tang, M. Sarathy, Investigate chemical effects of pre-chamber combustion products on main chamber ignition performance under an ultra-lean condition, in: SAE Technical Paper, 2020. [21] M. Silva, S. Sanal, P. Hlaing, E. Cenker, B. Johansson, H. G. Im, Effects of Geometry on Passive Pre-Chamber Combustion Characteristics, in: SAE Technical Paper, 2020. [22] S. Biswas, L. Qiao, Prechamber hot jet ignition of ultra-lean H2/ air mixtures: effect of supersonic jets and combustion instability, in: SAE Int. J. Eng., 2016. [23] J. Pan, Y. He, T. Li, H. Wei, Effect of Temperature Conditions on Flame Evolutions of Turbulent Jet Ignition, in: Energies, 2021 [24] T. Russwurm, M. Schumacher, M. Wensing Active Fuelling of a Passenger Car Sized Pre-Chamber Ignition System with Gaseous Components of Gasoline, in: SAE Technical Paper, 2020. [25] M. Schumacher, M. Wensing, A Gasoline Fuelled Pre-Chamber Ignition System for Homo‐ geneous Lean Combustion Processes, in: SAE International Powertrains, Fuels & Lubricants Meeting, SAE International, 2016 [26] T. Russwurm, M. Wensing, L. Euchner, Flame Luminesce in an Optically Accessible Engine with an Active Fuelled Pre-Chamber Ignition System, in: Springer, 21. Internationales Stuttgarter Symposium, 2021 The Authors: Tim Russwurm, M.Sc. Professur für Fluidsystemtechnik, Friedrich-Alexander-Universität Erlangen-Nürn‐ berg Tobias Achenbach, M.Sc. Professur für Fluidsystemtechnik, Friedrich-Alexander-Universität Erlangen-Nürn‐ berg Prof. Dr. Ing. Michael Wensing Professur für Fluidsystemtechnik, Friedrich-Alexander-Universität Erlangen-Nürn‐ berg 301 Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model <?page no="303"?> Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations Antonino Vacca, Marco Chiodi, André Casal Kulzer, Michael Bargende 1 Sebastian Bucherer, Paul Rothe, Ivica Kraljevic, Hans-Peter Kollmeier 2 Albert Breuer, Ruhland Helmut 3 1 FKFS - Forschungsinstitut für Kraftfahrwesen und Fahrzeugmotoren Stuttgart 2 Fraunhofer ICT 3 Ford Werke GmbH Abstract: Internal combustion engines operating with gas could represent an immediate and cost-effective solution in cutting CO 2 emissions thanks to the higher hydrogen to carbon ratio compared to commercial fuels and exploiting the higher knock resistance of methane compared to gasoline. Already successfully used in stationary gas engines, the active pre-chamber ignition for on-road internal combustion engines allows decoupling the mixture realized in the main combustion chamber from the one produced at the electrode. This makes the engine capable of operating in extremely lean conditions and thus increasing its efficiency, while maintaining adequate combustion characteristics, such as low cycle-to-cycle variation and sufficiently short combustion. In addition to this, the adoption of a very lean mixture allows keeping the engine-out NOx under control. Therefore, the design of the pre-chamber, the position of the electrode with respect to the pre-chamber injector and the charge motion generated in the pre-chamber itself, thanks to an ad hoc design of the pre-chamber holes, assume a fundamental role in the new engine development. In this paper, different active pre-chamber designs are studied by means of 3D-CFD Simulation. A single-cylinder engine is reproduced in the calculation environment of the 3D-CFD-Tool QuickSim, allowing multi-cycle simulations and including the test bench peripheries. A conventional piezo-actuated hollow-cone gasoline injector is used to dose the amount of methane in the pre-chamber. Since the pre-chamber could represent a source of NOx and in order to improve cold start operations, the prototype active pre-chamber can be conditioned through a dedicated derivation of the engine water jacket. The 3D-knock model based on detailed chemistry calculation supports the evaluation of the pre-chamber at full load. The performances of the different active pre-chambers are tested both at low <?page no="304"?> load for lean operation and at high load for stoichiometric condition. Following the performance prediction of the first manufactured geometry, the engine can operate ultra-lean up to λ ~ 1.8, with indicated efficiency higher than 42%. Keywords: 3D-CFD, CNG lean combustion, active pre-chamber. 1. Introduction The transportation sector is geared toward reducing greenhouse gasses (GHGs) in the atmosphere. Within this frame, different fuels are considered as possible alternatives to traditional gasolines to accomplish the emission targets. Among these, methane the major component of natural gas, can play a crucial role in achieving GHG reduction goals as an immediate solution, since it has long been employed in internal combus‐ tion engines (ICEs), taking advantage of an already partially existing infrastructure network. Methane’s high hydrogen-carbon ratio leads to a CO 2 reduction of 24% comparing to burning gasoline, just considering the fuel replacement, even from fossil origin. In addition, the high RON (Research Octane Number) of natural gas (RON-Methane ≥120 vs. RON-Gasoline = 95-100), produces higher knock resistance, allowing the increase of the compression ratio (ε) and therefore, of the engine efficiency. Furthermore, methane presents favorable ignition characteristics compared to gasoline, which allows the adoption of diluted-combustion process, either using exhaust gas recirculation or lean operation, which enable a further increase in efficiency [1] [2]. According to these, a reduction of 30% CO 2 can be achieved using CNG (compressed natural gas) instead of gasoline [2]. Hydrogen can also be added to methane which leads to further advantages in terms of CO 2 and emission reduction [3]. Considering these aspects, CNG plays an ecologic and economic long-term perspective in the automotive sector. With the aim of further increasing ICEs indicated efficiency, one interesting sol‐ ution is the adoption of pre-chamber ignition system. The principle of operation of a pre-chamber is based on the fact that the internal geometry and holes of the pre-chamber are designed in such a way as to induce a strong turbulence at the spark plug electrode, making combustion very rapid and in particular shortening the time required to progress from 50% to 90% burned fuel mass. High pressure-gradients, localized near top dead center, result in very efficient combustion that at the same time avoid anomalous forms of combustion such as knocking. Reducing the completion time of the combustion process makes the expansion stroke more effective and reduces heat loss through the wall [4] [5]. In addition, pre-chamber enables the control of the local mixture close to the spark plug, decoupling the mixture formation from the main combustion chamber. Looking for further rise of the indicated efficiency, it can be worth to operate the engine with lean mixture, especially at operating points where the throttle is partially closed. Here with the help of a fully variable valve system, the engine can be operated with mixture at lambda greater than 1.7 and throttle valve fully open, reducing pumping losses. The homogenization of the mixture in the pre-chamber 304 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="305"?> is managed through the insertion of an injector in the pre-chamber it-self building a so-called active pre-chamber (ACP) or scavenged pre-chamber. Following the previous considerations, an engine working with methane and featur‐ ing high efficiency requires a specific design, for exploiting CNG advantages. In order to realize such an engine, a consortium made by Ford-Werke GmbH, Fraunhofer-Institut für Chemische Technologie (ICT), Forschungsinstitut für Kraftfahrwesen und Fahr‐ zeugmotoren Stuttgart (FKFS), Rosswag GmbH and BRIGHT Testing GmbH (financed by BMWI) decided to run a 36 months project (Nr. 19I20014E), called MethMag (Methan Mager Motor). The aim of the project is the development of a new prototype 3-cylinder engine derived from the combustion optimization of a single-cylinder engine. In the previous works of the authors considering the methane engine with active pre-chamber [6] [7], the steps towards the new methane engine design have been explained, with focus on the cylinder head concept as well as on the injector and pre-chamber position. The valve strategies have been also discussed resulting in a reduced valve overlap (no need for high residual gas rates) and high boost pressure, realizing de-throttling operation with lean mixture. At high loads, a reduction of the intake valve stroke has been considered to generate a light-millerization concept. Table 1 resumes the main characteristics of the new single-cylinder engine operating with methane injection. The pre-chamber is placed exactly in the middle of the combustion chamber to enhance its scavenging at low loads. For the methane injection in the pre-chamber, an outward-opening piezo-actuated injector for minimal dosing of the methane is chosen. This solution can support the injection of small fuel quantity (below 10% of the total injected fuel) in the pre-chamber as required for operating the engine at low load and in condition of low scavenging of the pre-chamber. In addition, a prototype solenoid-actuated injector is placed below the intake ports realizing a direct injection at 16 bar injection pressure . Single-Cylinder Engine Injection Active pre-chamber (16 bar) and direct injection (16 bar) Valve Overlap small (high boost pressure for lean operation and de-throttling) Intake Valve Strategy High lift to reduce the boost pressure at low load and light millerization at high load Mixture Stoichiometric / lean (λ > 1.7) Bore [mm] 84 Stroke [mm] 89.8 Compression Ratio [-] 15.1: 1 Max. Boost Pressure [bar] 2.2 bar 305 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="306"?> Max. Cyl. Pressure [bar] 180 bar Max. Power 40 kW @ 5500 rpm Table 1. Features of the new designed single-cylinder engine for methane injection and active pre-chamber. In the new engine design the direct injector is placed under the tumble inlet port. As mentioned, the active pre-chamber is fitted in a central position with the pre-chamber injector and the spark plug perpendicular to the crankshaft axis. This cylinder head layout can be seen in Figure 1. The central position of the pre-chamber brings advantages, especially when running the engine in very lean operation. Due to the tight packaging between the active pre-chamber and the exhaust valve seat rings a central position for the direct injector was not possible. The central injector would result in an insufficient coolant jacket in a thermally high loaded area and was ruled out despite showing a better engine efficiency in early simulations of the concept. Figure 1. Positioning of direct injector concept, below intake port The single-cylinder engine represents the basis for the thermodynamic development of the engine concept with the final objective to build up a 3-cylinder natural gas engine (1.5 l displacement) that can operate with a very lean mixture ( λ >> 1) over a wide range of the engine map, which could be suitable for use in a light duty commercial vehicle. The after-treatment system should be limited to a conventional three-way catalyst (avoiding the implementation of SCR, therefore the raw NOx emissions of the engine must be very low). The engine should provide 110 kW power at 5500 rpm with a maximum peak pressure of 140 bar (for the thermodynamic optimization of the combustion process the single-cylinder engine is built to manage a peak pressure of 180 bar). 306 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="307"?> In this context, 3D-CFD and CHT simulations are used for a virtual engine develop‐ ment to study different design solutions. The most promising geometry is then selected and realized for a single-cylinder engine. 2. Virtual Engine Design The numerical investigations within the present project are carried out with the fast response 3D-CFD tool QuickSim, developed at IFS/ FKFS Stuttgart, which is specifically designed for ICEs simulation. The 3D-CFD tool QuickSim offers the possibility to reduce the time expense for a simulation (2 h for an operating cycle of a full engine with a 12 cores CPU) by using ICE-adapted and improved computational models, coarser meshes, with respect to traditional 3D-CFD approaches, without sacrificing the quality of the results. The fast calculation frame allows to extend the simulation domain up to a full engine, but also the investigation of multiple successive engine cycles in a reasonable time frame. The fluid domain calculated in every 3D-CFD simulations is reported in Figure 2, where the engine and the test bench peripheries are as well depictured. Figure 2. Single-cylinder engine test bench modelled in the 3D-CFD environment of QuickSim. More information about the models implemented in QuickSim is resumed in Table 2. The software provides the possibility to deliver the results of one-engine-cycle to commercial cold-flow software in order to design the engine cooling system with traditional approach, as realized in the current analysis. Simulation methodology Multiphase RANS Multi-phase flow Euler / Lagrange Turbulence k-ε Combustion - Flame Propagation Two-zone Weller-model adapted to GDI-com‐ bustion for spark ignition flame propagation. 307 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="308"?> Time step Δt Const. = 0.5 degree crank angle (°CA) Fuel model Detailed chemistry solved separately in Can‐ tera (laminar flame speed and auto-ignition) Table 2. Overview of QuickSim models 2.1 Approach and Modelling of Knock In the 3D-CFD tool QuickSim, the fuel model and knock model are linked through the calculation of the laminar flame speed (LFS) and of the auto-ignition delay (IDT). These properties are preliminarily evaluated using detailed chemical kinetics calculations performed with a tool developed in Cantera. In the tool a wide range of lambda, temperature, pressure, residual gas rate and composition are considered. The adopted chemical mechanism for the calculation of LFS and IDT was developed by Lawrence Livermore National Laboratory (LVLL), which includes 324 species and 5739 reactions, and it has been extensively validated. Figure 3 reports the calculations of the LFS and IDT for a conventional gasoline RON 95 ethanol 10% (E10) in comparison with the one of methane, for a certain pressure, temperature and residual gas rate in the combustion chamber. The LFS and IDT for different parameter combinations are tabulated and called through look-up tables in the 3D-CFD tool QuickSim. Methane presents slightly lower LFS with respect to gasoline, but higher IDT with respect to the ideal gasoline vaporization (IDT is calculated for the gasoline considering a full evaporation with ideal homogenization), thus making methane actually more resistant to knock compared to gasoline. In reality, considering real engine conditions, the higher knock resistance of methane due to its chemical composition reduces with respect to gasoline since the lack of the evaporation process and the consequent loss in charge temperature reduction. Figure 3. Laminar flame speed and ignition delay time calculated through detailed chemistry and included in the 3D-CFD tool QuickSim Particularly the evaluation of knock considers the following parameters in addition to the fuel composition: 308 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="309"?> - Pressure of unburned gas - Temperature of unburned gas - Residual gas content and its composition (reactive or inert) - Local air to fuel mixture (lambda) - Charge motion - Presence of water (from external water injection or humidity of the air) These parameters are considered regarding their spatial and time evolution (as knock integral). The presented knock model was validated for several engine applications [10] [11] [12] and it is used to evaluate different design concepts of the new methane engine. As explained in [6], the measurements of a reference engine with gasoline injection and conventional spark plug (compression ratio 12.5: 1) were used to calibrate the knock model of the virtual test bench and then used to test the knock resistance of the methane engine with pre-chamber. The reference engine (gasoline injection and conventional spark plug) is simulated at the same conditions as the engine at the test bench, whose ignition timing has been found at the knock limit. The knock model is calibrated so that no knocking combustion is induced with this combustion calibration. Figure 4 shows the different knocking behavior of the engine once in operation with gasoline and once in operation with methane and using an active pre-chamber, for the operating point 2000 rpm full load. Both the simulations have been run at the same spark advance, which is specified considering the firing top dead center (FTDC) at 720 °CA (crank angle). The main color scale represents the combustion progression, where red is the burned mass and blue represents the mass not yet crossed by the flame front. Purple regions identify the areas where the mass in self-ignition condition is located. The mass in self-ignition condition is the amount of fresh charge, that is not yet crossed by the flame front, but has conditions (pressure, temperature, composition, time-accumulation) prone to self-ignite. The percentage of this mass compared to the rest of the end gas, and the distance of the flame front to this mass are fundamental for distinguishing between normal and knocking combustion. In this way, the knock model takes into account a spatial and temporal resolution, also considering temperature, pressure, lambda, residual gas mass and fuel composition. 309 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="310"?> Figure 4. Comparison of the reference engine operated with E10-gasoline and conventional spark plug (left) and with methane and active pre-chamber (right) at the same ignition timing, at 2000 rpm full load. Again with reference to Figure 4, the pre-chamber combustion (right side) is faster, in comparison with the conventional spark plug (left side), moving the MFB50 (50% mass fraction burned) almost 4 °CA earlier. This can be assessed since the detailed chemistry calculations implemented in the 3D-CFD model (see again Figure 3), show that the flame speed of methane is comparable to that of an E10 (ethanol 10%) fuel, at most slightly lower. For this reason, the faster flame speed can be addressed to the pre-chamber ignition. Then, a sweep of spark advanced, using the calibrated knock model could predict the new ignition point limit. The comparison between the pressure curves of the measured reference engine (gasoline injection) and the respective 3D-CFD simulation are reported in Figure 5 (black and grey curves). In addition, Figure 5 depicts in blue the new pressure curve obtained equipping the engine with pre-chamber and methane injection, at limit of knock for the load point 2000 rpm full load. With the use of a pre-chamber and methane injection, it is possible to shift the center of combustion from 30 °CA a.TDC (after top dead center) to 10 °CA a.TDC without promoting knocking combustion. The calibration of the knock model is therefore realized considering the measure‐ ments led on the reference gasoline engine and then the results have been extended through the modelling of the virtual engine equipped with pre-chamber and methane injection. Starting from the virtual engine a new geometry is designed by ICT and simulated by FKFS and, after several optimization loops, a design freeze for the first prototype hardware batch is determined. 310 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="311"?> Pressure [bar] 0 10 20 30 40 50 60 70 80 90 100 110 120 Crank angle [deg] -90 -45 FTDC 45 90 Pressure in Cylinder Pressure in Pre-chamber Measurements Ref. Engine Simulation Ref. Engine Simulation Ref. Engine Geometry Methane Injection & Pre-Chamber Inl.Valve Exh.Valve. Ign & Comb. Figure 5. Comparison of the pressure curves of the reference engine (measurements and simulation) and the pressure curve obtained by replacing gasoline injection with methane and conventional spark plug with pre-chamber ignition system, at 2000 rpm full load. The operating points (OP) examined through 3D-CFD simulations are listed below: - OP1 1500 rpm 1.5 bar IMEP - OP2 1500 rpm 4.0 bar IMEP - OP3 2000 rpm 10 bar IMEP - OP4 2000 rpm full load - OP5 5500 rpm full load These operating points are simulated with a conventional spark plug and gasoline in‐ jection as well as with the new engine geometry described in Table 1 (for further details about the new engine design refer to [6] [7]). For OP1, OP2 and OP3, pre-chamber injection is used to run the engine as lean as possible. λ = 1.8 represents the maximum lean limit, without affecting combustion stability, for OP2. OP4 and OP5, full load points, are run at stoichiometric conditions with both gasoline and methane. 3. Active pre-chamber The decision to place the active pre-chamber in the middle of the combustion chamber is based on previous experience with passive pre-chamber development [4] [5] and grants the advantages mentioned before. The cleaning of the pre-chamber volume 311 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="312"?> from residual gas as well as the enhancing of the mixture formation at the electrode is supported by the central position. Based on this, multiple concepts regarding the position of the spark plug and the pre-chamber injector were designed and simulated. As mentioned before, for the methane injection in the pre-chamber, a gasoline piezo-actuated injector with an outward-opening nozzle was chosen due to the self-reinforcing sealing of the injector needle during the combustion events and therefore high-pressure phases in the pre-chamber. The choice of an outward opening-nozzle is ideal to avoid stagnation of fuel close to an eventual backflow. Furthermore, the pre-chamber volume houses a high-pressure indicating sensor as well as a thermocouple, for research purposes. In the first pre-chamber design concept, the injector was placed in the top right corner of the pre-chamber volume, with the spark plug next to it, as shown in Figure 6 (the right side of Figure 6 corresponds to the engine intake side). This arrangement of spark plug and injector results in a quite large inner diameter at the top part of the pre-chamber with the electrode being far away from the centre of the pre-chamber volume. In order to fit in the housing of the cylinder head, the injector and the spark plug have to be placed high from the cylinder head roof, leading to a long, bottleneck shaped design of the pre-chamber volume. These geometrical restrictions lead to very poor design flexibility of the pre-chamber volume for this design concept. In addition, 3D-CFD simulations have shown poor mixture homogenization in the pre-chamber and particularly the absence of turbulence at the electrode , therefore this design was discarded. 45° cone Spark plug injector Figure 6. Central pre-chamber section, first internal-flow-design concept with direct connection be‐ tween injector and electrode For the following variants of pre-chamber geometries, a different design approach for the positioning of the spark plug as well as the injector was taken. To enable 312 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="313"?> a high freedom of design and a more central position of the electrode the injector was pulled back, allowing the spark plug to move to a more central position. The injector is only connected by a small channel to the pre-chamber volume to avoid the aligned arrangement of spark plug and injector next to each other in the roof of the pre-chamber, as it can be seen in the section of Figure 7. This change of arrangement leads to a good flexibility, especially around the upper part of the pre-chamber volume by lowering the minimum required inner diameter. Moreover, the electrode (on the left side of Figure 7) was moved closer to the main combustion chamber resulting in a shorter and overall more compact pre-chamber design, in view to increase the turbulence at the electrode. 45° cone Spark plug injector Radial gap cap Figure 7. Central pre-chamber section, second internal-flow-design concept, using a connection channel between injector and electrode, with 45 ° external housing conus. The pre-chamber is mounted directly in the pre-chamber coolant jacket resulting in the need for sealing combustion gas as well as coolant fluid. To have an efficient cooling of the thermally high loaded part of the pre-chamber, it is required to have the sealing close to the cylinder head roof. This requirement results in the need of a combined sealing for coolant fluid and combustion gas. In a structural simulation, a solely axial sealing with a copper ring showed high tensile tensions around the exhaust seats - hence a 45 ° cone design for the pre-chamber sealing was chosen. The axial position of the 45 ° cone towards the cylinder head roof is limited by the required minimal wall thickness to the exhaust seat rings. The structural stiffness of the pre-chamber, the 313 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="314"?> minimal angle overlap between pre-chamber and cylinder head cone, as well as the stud bolt preload were optimized in an iterative process. In order to ensure that the aluminium cylinder head roof, above the pre-chamber cap and underneath the 45 ° cone does not produce a thermal overload, a small radial gap of around 0.5 mm was implemented. 3.1 Pre-Chamber Cooling The use of additive manufacturing of the cylinder head as well as the pre-chamber housing enabled the full split of the cylinder head coolant jacket and the pre-chamber coolant jacket. Therefore, the pre-chamber housing could be thermally conditioned to the desired temperature level, depending on the engine load point, without interfering with the cylinder head coolant system. Split-cooling is implemented only for the single-cylinder research engine, while the 3-cylinder engine is operated with a single cooling circuit. Thus, the filigree channels (the one cooling the pre-chamber central geometry, as it will be described in Figure 8) are not implemented in the 3-cylinder engine. The split-cooling is used to investigate the influence of the wall temperature of the pre-chamber on the combustion process and on the emission formation. In the next paragraph, conjugate heat transfer simulation (CHT), considering the pre-chamber geometry are carried out, using a coupled simulation of ANSYS Forté and ANSYS Fluent. The reactive 3D-CFD simulation of the gas flow in Forté provides the convection boundary conditions for the calculation of the wall temperatures in Fluent. Here, in turn, the heat fluxes in the solid parts are coupled with the 3D-CFD simulation of the cooling water jackets. The resulting wall temperatures are then coupled back to the gas phase. This iterative procedure is carried out until the heat balance over the cylinder head has converged. CHT simulations were considered for the feasibility analysis and validation of the component strength. These calculations were led using the boundary conditions of the gas phase calculated from the 3D-CFD simulation, at the rated power operating point (OP5, 5500 rpm full load) to consider the highest thermal conditions for design purposes. Bearing in mind the second pre-chamber design (injector and pre-chamber volume linked through a small channel), the coolant jacket consists of two symmetrical channels as can be seen in Figure 8, which can cool down particularly the inner wall of the pre-chamber volume and at the same time reduce the peak temperature at the cap. In Figure 8, the pre-chamber is sketched through its fluid domain. The circular coolant channel around the pre-chamber volume is split in two symmetrical channels to give the pre-chamber housing more structural stiffness and to reduce the deformation in the critical sealing cone area. The cross-section area of the coolant channels in the pre-chamber is 2.8 mm² for each channel. 314 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="315"?> Figure 8. Streamlines of the pre-chamber coolant jacket around the pre-chamber inner surface, in case of 45 ° housing conus. Additionally, the lower temperature fluctuations at the sealing surface reduce the relative movement between the pre-chamber and the cylinder head. The cylinder head coolant jacket is flowed through at 50 l/ min at an inlet temperature of 103 °C. The pre-chamber coolant jacket, on the other hand, is flowed through at 6 l/ min at an inlet temperature of only 20 °C. The single-cylinder engine test bench with a dedicated cooling system can ensure an inlet temperature of 20 °C without generating local boiling conditions in the filigree channels, since the presence of bubbles could occlude the channels. As mentioned before, for the 3-cylinder engine, the pre-chamber has to be modified since the absence of a dedicated cooling system. The possibility to cool down separately the pre-chamber is just used in the single-cylinder engine for research purposes. Moreover, the adoption of 20 °C coolant temperature in the pre-chamber represents the worst-case scenario for the structural simulation of the pre-chamber, since the high thermal gradient of the coolant channel and pre-chamber wall in the pre-chamber housing induces high tensile stress. Furthermore, the system pressure in all cooling circuits is 3 bar. In the next paragraph, a comparison between the wall temperatures of different pre-chamber designs is discussed. In this comparison, the internal geometry of the pre-chamber is mostly unchanged (injector and pre-chamber volume linked through a small channel) while the housing conus has been modified from 45 ° to 20 °. Figure 9 shows the temperature distribution of the wall in the pre-chamber area resulting from a CHT simulation, in case 45 ° housing conus. 315 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="316"?> Figure 9. Cross Section of pre-chamber with 45 ° housing conus, showing the temperature distribution of the wall. In the section, it is visible the indicating pressure sensor (not to be confused with the injection channel). The circulating cooling in the central area of the pre-chamber volume efficiently dissipates the heat, thus reducing both the wall temperature in this area and the peak temperature at the cap by almost 300 °C. At the same time, the wall temperature at the combustion chamber roof around the pre-chamber decreases because the cooling channels largely prevent the heat flow over the sealing surface. As consequence, the undercut at the cap can be reduced in a new variant, which increases the combustion chamber roof temperature again, but further reduces the cap temperature. Based on this simulation, the sealing cone of the pre-chamber has been reduced from 45 ° to 20 °, which has several advantages. Firstly, the sealing surface reaches closer to the cylinder head roof and therefore is enlarged, improving heat transfer to the cylinder head, particularly in the lower part of the pre-chamber. Secondly, the cross-section of the additional cooling channels can be increased, which improves the flow, and thus reduces the risk of local boiling. Moreover, the available space for different pre-chamber geometries was enlarged giving more design freedom for future designs. Figure 10 shows the new coolant jacket obtained and the 20 ° housing conus. 316 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="317"?> 20° cone Spark plug injector Reduced radial gap cap Figure 10. Central pre-chamber section, second internal-flow-design concept using a connection channel between injector and electrode, with 20 ° external housing conus. Due to the described design modifications, enlarging the package space of the coolant channels in the pre-chamber, the cross-section area of each coolant channel was raised from 2.8 mm² to 4.8 mm². The most important aspect is that, due to the improved flow, the maximum temperature at the coolant jacket can be lowered to 122 °C, which is below the boiling temperature of water at 3 bar and thus opens up the possibility of raising the cooling water inlet temperature. Figure 11 depictures the new streamlines of the pre-chamber with 20 ° external housing conus. In addition, due to packaging constraints, the bleeding channel previously placed on the exhaust side of the engine, in the last geometry has been moved towards the intake side, as can be seen from Figure 11 in comparison to Figure 8. 317 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="318"?> Figure 11. Streamlines of the pre-chamber coolant jacket around the pre-chamber inner surface, in case of 20 ° housing conus. Considering again the temperature of the wall, the peak temperature at the pre-cham‐ ber cap can be lowered by more than 200 °C, as shown in Figure 12, considering the 20 ° conus pre-chamber. Figure 12. Cross Section of pre-chamber with 20 ° housing conus, showing the temperature distribution of the wall. In the section, it is visible the indicating pressure sensor (not to be confused with the injection channel). 318 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="319"?> 4. Thermodynamic Investigation and Results In the current section, the potentials of the 3D-CFD tool QuickSim are exploited to test different pre-chamber geometries, considering the effect on the internal gas flow field. The aim is to study different configurations of the pre-chamber holes, position of the electrode and of the scavenging channel (channel that links the injector tip with the pre-chamber volume). These tests have been always run parallel on two operating points: OP2 representing the case with very lean operation (λ = 1.8) at low load, where additionally different injection strategy in the pre-chamber have been tested and OP4, full load point, operating at λ = 1 where no injection in the pre-chamber is needed and therefore it is operated as a passive pre-chamber. The most suitable engine design resulting from this analysis, is then tested on the other operating points mentioned before. For the internal flow analysis the pre-chamber housing with 20 ° conus is considered, since it provides the highest fluid design freedom. For each simulation the complete test bench is reproduced, including the single-cyl‐ inder engine peripheries (up to the pressure vessels) and multi-cycles are considered, since the pre-chamber behavior depends on the residual gas mass that the pre-chamber is able to scavenge before ignition point. The correct amount of residual gas and the interdependent knocking behavior, can be predicted just by reproducing faithfully the exhaust back pressure wave of the single-cylinder engine, through a full engine multi-cycle 3D-CFD simulation. The pre-chamber position and the internal geometry arrangement have been discussed in [6], here instead the analysis focuses on the pre-chamber holes and their influence on the pre-chamber mixture formation and on the turbulence. In the next subsections, an overview of the different operating points investigated through 3D-CFD simulation is provided, by highlighting the differences between the results obtained with the reference gasoline engine compared to the new engine geometry, operating with methane and active pre-chamber. Furthermore, considering OP2 a detailed discussion on the different pre-chamber designs is provided. 4.1 Operating Point 1 - 1500 rpm, 1.5 bar IMEP OP1 represents a challenge for the combustion stability and the optimization of the fluid exchange process. As a solution to reduce pumping losses, the new engine adopts a reduction of the intake valve maximum lift of 4.5 mm, allowing the complete de-throttling of the engine in combination with very lean combustion (as mentioned before the engine is designed to host a fully variable valve train). By exploiting the injection in the pre-chamber it is possible to reach almost stoichiometric conditions in the pre-chamber and slightly rich conditions at the electrode. The global lambda reached is 1.65 with a stable combustion, which lasts 32 °CA (time from 10% to 90% mass fraction burned). These results are obtained with an asymmetric design of the holes in the pre-chamber cap as explained in the next paragraph. 319 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="320"?> 4.2 Operating Point 2 - 1500 rpm, 4 bar IMEP OP2 is used to test different pre-chamber arrangements in case of very lean operations. At first, an optimization of the intake valve profile is conducted to find-out the most suitable de-throttling strategy. The intake valve stroke is therefore extremely reduced, with a maximum lift of 6 mm as discussed in [6] and showed in Figure 13. Figure 13. Valve profile and timing for OP2 and injection strategy for very lean operation. With this adjustment, the engine can run with wide open throttle and at a global lambda of 1.8, keeping the center of combustion at 4 °CA a.TDC and a combustion duration of 28°CA (time from 10% to 90% mass fraction burned). A good combustion stability could be achieved thanks to the optimization of the injection strategy in the pre-chamber (again showed in Figure 13) and the turbulence at the electrode generated by the holes. Experiences in previous works of the authors showed that pre-chamber effectiveness at low load grows when the ignition point is set not too far from TDC. The reason is essentially that the scavenging of the pre-chamber from residual gas is realized in the last part of the compression stroke, therefore too early spark advance could reduce the speed of ignition by higher rate of residual gas. In the case of an engine, running with very lean mixture the objective is to realize a homogeneous mixture as much as possible at stoichiometric condition in the pre-chamber. With this, the first part of the combustion process results quite effective being affected by the high turbulence of the pre-chamber, but then the jets have to propagate within a very lean combustion chamber. In this case, is then worth igniting the pre-chamber a bit earlier when the residual gases are slightly higher but with the aim of keeping the second part of the combustion (between 50% and 90% mass fraction burned) shorter. In this case, the 320 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="321"?> engine has a spark advance of 22 °CA before TDC with 12 °CA time to move from 10% to 50% mass fraction burned and 16 °CA to complete the combustion from 50% to 90% mass fraction burned. These results can be obtained by the optimization of the gas flow in the pre-chamber and of the injection strategy. Figure 14 to Figure 21 show the results for different pre-chamber holes at ignition point (spark advance = 22 °CA before TDC). In some cases, there are slight differences in the external shape of the pre-chamber cap or in the injector tip, since some boundary constraints change during the project, but they have no impact on the gas flow study coming from different geometries of the holes. The comparison considers the same injection strategy, namely one short injection in the pre-chamber and two in the main combustion chamber (see again Figure 13). The overall methane injected in the pre-chamber is about 8% of the total injected gas. The injection in the pre-chamber is suitable to help cleaning the pre-chamber from residual gas. The first injection in the main combustion chamber supports the turbulence in the cylinder and the second one corrects the mixture stoichiometry at the end of the compression stroke in the pre-chamber. The starting point is represented by the pre-chamber layout with injector and pre-chamber volume linked by a channel, 6 holes cap and an external cone angle of 20 °. In the further variants, only the pre-chamber holes are modified. As can be seen from Figure 14 and Figure 15, the pre-chamber H01 (layout of the holes 1) features a symmetric hole design. While the combustion chamber has a global lambda of 1.8, at the electrode there is a stratification of the charge still globally lean around λ = 1.2. The majority of the turbulence is directed towards the middle of the pre-chamber, without generating any structured movement within the pre-chamber volume. Figure 16 and Figure 17 (pre-chamber H02) show the first attempt to have a certain asymmetry in the pre-chamber holes, keeping the same hole angle (110 °) of pre-chamber H01. In fact, the holes corresponding to the exhaust side (left of Figiure 16 and Figure 17) have been shifted 0.5 mm upward the central point. In this case (variant H02) the holes produce a stratification of the charge parallel to the spark plug middle axis and direct the turbulence exactly at the electrode. For this arrangement of the holes, the mixture at the electrode is still too lean and it is not stratified towards the direction of the flame propagation. Figure 18 and Figure 19 depicture the pre-chamber variant H03 that has mirrored holes layout compared to the variant H02, previously described. In this case the holes produce a small tumble around the electrode, which is the design target, but they generate again a stratification of the charge parallel to the spark plug middle axis and in addition, a very reach mixture at the electrode. 321 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="322"?> Fig. 14 Pre-chamber H01 mixture formation at ignition point. Fig. 15. Pre-chamber H01 velocity vectors at ignition point. Fig. 16 Pre-chamber H02 mixture formation at ignition point. Fig. 17. Pre-chamber H02 velocity vectors at ignition point. Finally, the last cap geometry is realized combining the advantages of the asymmetry of the holes found for the variant H03 with smaller hole angles (100 °) to direct the jets less towards the piston. As results, variant H04 is generated and investigated in Figure 20 and Figure 21. In this case, a structured tumble is present at the electrode and the turbulence in the injector side of the pre-chamber (intake side of the engine) supports the homogenization of the mixture, producing at the electrode a slightly stratification of the charge, in the direction of the flame propagation, with an almost stoichiometric 322 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="323"?> volume surrounding the electrode (the position of the lateral counter electrode has been changed in variant H04 because of packaging constraints). As consequence, the configuration H04 has been chosen and successfully tested also for OP4. Therefore, it has been tested for all the operating points and represent the first manufactured active pre-chamber for the methane engine. Fig. 18 Pre-chamber H03 mixture formation at ignition point. Fig. 19. Pre-chamber H03 velocity vectors at ignition point. Fig. 20 Pre-chamber H04 mixture formation at ignition point. Fig. 21. Pre-chamber H04 velocity vectors at ignition point. 323 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="324"?> 4.3 Operating Point 3 - 2000 rpm, 10 bar IMEP OP3 can be considered as a load point close to cruise mode condition of such an engine application and therefore is interesting from the point of view of the indicated efficiency. In fact, at OP3 the engine is operated with a boost pressure of 1.85 bar (Abs.), with pre-chamber injection and global lambda of 1.5. The higher the power demand the more difficult to obtain a stable combustion with lean conditions. In this case with pre-chamber H04 a center of combustion of 8 °CA a.FTDC can be reached with a combustion duration of 35 °CA (time from 10% to 90% mass fraction burned). 4.4 Operating Point 4 - 2000 rpm, 23.5 bar IMEP OP4 represents a full load operating point with high tendency to knocking combustion especially considering the reference gasoline engine (already adopting a millerization concept). The new methane engine exploits the high resistance of methane to the onset of knock, in combination with passive pre-chamber ignition which brings the advantages to short the combustion duration and speed up the flame front propagation towards the unburned gas. At OP4 the engine is operated at global stoichiometric conditions with a boost pressure of 2.2 bar (Abs.). The intake valve lift can be increased with respect to the gasoline reference engine, still adopting a millerization. The increased intake valve stroke supports the rise of the in-cylinder turbulence, improving the mixture formation. The peak pressure can be raised up to 132 bar, before meeting any knocking combustion, prediction realized through the knock model described in the previous paragraph. In this case, the engine is operated with a center of combustion of 16 °CA a.TDC, almost 15 °CA earlier with respect to the gasoline reference engine. 4.5 Operating Point 5 - 5500 rpm, 19.5 bar IMEP The rated power operating point can be run both for the reference gasoline engine and the new geometry with methane injection, at stoichiometric conditions. At OP5 the consideration realized on the valve timing described for OP4 are still valid. Even at OP5, the higher knock resistance of the new engine geometry appears clear. Following the knock model, the engine can be operated almost at the optimal center of combustion (9 °CA a.TDC), 10 °CA earlier than the reference gasoline engine. The new engine adopts a reduced boost pressure with respect to the reference engine (1.7 bar Abs.), thanks to the higher intake valve lift. Notable is the increase in the indicated efficiency as it will be shown later. OP5 can be obtained with 145 bar peak pressure and still being within the knock limit condition. The produced indicated power is 44.5 kW, fulfilling the project target (10% indicated power increase with respect to the reference gasoline engine). 4.6 Conclusion and Outlook As discussed before, the new engine geometry with active pre-chamber, higher com‐ pression ratio (15: 1 instead of 12.5: 1) and methane injection presents several advantages with respect to the reference gasoline engine in terms of combustion and fluid exchange 324 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="325"?> process. The lean operation (thanks to de-throttling and pre-chamber injection and combustion) and the higher compression ratio led the new engine to reduce the specific fuel consumption at low loads, with a relative short and stable combustion process. At high load, the new engine exploits the higher methane resistance to knock, combining a good mixture formation with a very fast combustion, thanks to the adoption of the pre-chamber H04 supporting the turbulence at the spark ignition electrode. Figure 22 summarizes the raise of the indicated efficiency obtained in the simulated load points, moving from the reference gasoline engine to the new engine geometry and methane injection. For the low loads and the high loads, the increase of the efficiency ranges from 3%to 7%-points, while efficiency improvement is more moderate at medium load. Further refinements of the pre-chamber geometry and of the injection strategy are ongoing revealing some additional margins of improvement in terms of indicated efficiency. lambda [-] IMEP [bar] 2 4 6 8 10 12 14 16 18 20 22 24 Engine speed [rpm] 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 λ >> 1 λ = 1 = 35 % → 42 % = 36.5 % → 40.1 % = 40.7 % → 41.1 % = 37.5 % → 40.4 % = 27.5 % → 33.6 % Figure 22. Operating map of the reference engine (gasoline), with focus on the load point investigated through 3D-CFD simulations and the possible rise of the indicated efficiency realized by the new geometry with active pre-chamber and methane injection. For the next pre-chamber iteration, the pre-chamber is shortened by 0.75 mm while the sealing cone remains untouched, so the penetration of the pre-chamber in the main combustion chamber is reduced as shown in Figure 23. 325 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="326"?> 20° cone Spark plug injector Cap/ corpus contact surface Smaller cap Figure 23. Pre-chamber design with external housing conus of 20 ° and smaller cap for improved cap peak temperature. Moreover, the outer geometry of the cap is redesigned to reduce its mass, and therefore the thermal capacity. This change of design reduced the volume of the cap by around 30%, keeping the connected surface area of cap and corpus constant to enable a high thermal heat transfer from the cap to the pre-chamber housing. In future studies, the coolant inlet temperature of the pre-chamber coolant jacket will also be raised to 103 °C. The single-cylinder engine with the new geometry is currently in the commissioning phase. After the validation of the virtual development performed in the first part of the project, the 3D-CFD investigations will be further carried out on the new 3-cylinder engine, with the goal of the full engine simulation for further optimization and refinement of the overall engine thermodynamics and efficiency. References [1] Kramer, U., Lorenz, T., Hofmann, C., Ruhland, H. et al., «, "Methane Number Effect on the Efficiency of a Downsized, Dedicated, High Performance Compressed Natural Gas (CNG) Direct Injection Engine," SAE Technical Paper 2017-01-0776, 2017». [2] Breuer, M., Bartsch, G., Friedfeldt, R., et al., «3- Cylinder SI Engine with fully variable Valve Train UpValve on Intake and Exhaust Side». 326 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="327"?> [3] Soltic, P., Hilfiker, T., Hutter, R., Haenggi, S., «Experimental comparison of efficiency and emission levels of four-cylinder lean-burn passenger car-sized CNG engines with different ignition concepts. Combustion Engines. 2018, 176(1), 27-35.DOI: 10.19206/ CE-2019-104». [4] Schmid, H., Kollmeier, H.-P., Kraljevic, I., Gottwald, T. et al., LPG and Prechamber as Enabler for Highly Performant and Efficient Combustion Processes Under Stoichiometric Conditions,” SAE Technical Paper 2021-24-0032, 2021, doi: 10.4271/ 2021-24-0032. [5] Vacca, A., Rossi, E., Cupo, F., Chiodi, M. et al., “Virtual Development of a Single-Cylinder Engine for High Efficiency by the Adoption of eFuels, Methanol, Pre-Chamber and Milleri‐ zation,” SAE Technical Paper 2022-37-0018, 2022, doi: 10.4271/ 2022-37-0018. [6] Vacca, A. et al. (2022)., Virtual Development of a New 3-Cylinder Natural Gas Engine with Active Pre-chambe, In: Bargende, M., Reuss, HC., Wagner, A. (eds) 22. Internationales Stuttgarter Symposium. Proceedings. Springer Vieweg, Wiesbaden. https: / / doi.org/ 10.1007/ 9 78-3-658-37009-1_31. [7] Bucherer, S., Rothe, P., Kraljevic, I., Kollmeier, H.-P. et al., “Design of an Additive Manufac‐ tured Natural Gas Engine with Thermally Conditioned Active Prechamber,” SAE Technical Paper 2022-37-0001, 2022, doi: 10.4271/ 2022-37-0001. [8] Chiodi, M., An Innovative 3D-CFD-Approach towards Virtual Development of Internal Combustion Engines", PhD thesis, University of Stuttgart, 2010. [9] Vacca, A., Cupo, F., Chiodi, M., Bargende, M. et al., “The Virtual Engine Develop‐ ment for Enhancing the Compression Ratio of DISI-Engines Combining Water Injec‐ tion, Turbulence Increase and Miller Strategy,” SAE Technical Paper 2020-37-0010, 2020, doi: 10.4271/ 2020-37-0010. [10] Cupo, F., "Modeling of Real Fuels and Knock Occurrence for an Effective 3D-CFD Virtual Engine Development,", Ph.D. thesis, University of Stuttgart, 2021. [11] Vacca, A., Cupo, F., Chiodi, M., Bargende, M. et al., “The Virtual Engine Develop‐ ment for Enhancing the Compression Ratio of DISI-Engines Combining Water Injec‐ tion, Turbulence Increase and Miller Strategy,”, SAE Technical Paper 2020-37-0010, 2020, doi: 10.4271/ 2020-37-0010. [12] Jonas Villforth, et al., Methods for the Evaluation of eFuel Potentials on the Combustion and Emission Behavior of DISI Engines, 9th Int. Symposium on Development Methodology, 2021 Wiesbaden, Germany. [13] F. Berndt, «Ottomotorische Magerbrennverfahren: NOx- und partikelarme Verbrennung durch neue Zünd- und Einspritzkonzepte, TU Braunschweig: Dissertation, 2015». [14] Weber, C., Friedfeldt, R., Ruhland, H. et al., Downsizing und hohe Leistung mit zukünftigen Kraftstoffen und Emissionslimits. MTZ Motortech Z 82, 72-77 (2021). https: / / doi.org/ 10.100 7/ s35146-021-0669-6. [15] Villforth, J., Kulzer, A.C., Deeg, H.-P., Vacca, A. et al., Methods to Investigate the Importance of eFuel Properties for Enhanced Emission and Mixture Formation, SAE Technical Paper 2021-24-0017, 2021, doi: 10.4271/ 2021-24-0017. [16] Seboldt, D., Untersuchungen zum Potenzial der CNG-Direkteinblasung zur Reduktion von HC-Emissionen in Gasmotoren, PhD thesis, University of Stuttgart, 2017, https: / / doi.org/ 10. 1007/ 978-3-658-17906-9. 327 Study of Different Active Pre-chamber Ignition Layouts for Lean Operating Gas Engines using 3D-CFD Simulations <?page no="328"?> [17] Cupo, F., Chiodi, M., Bargende, M., Koch, D. et al. "Virtual Investigation of Real Fuels by Means of 3D-CFD Engine Simulations, SAE Technical Paper 2019-24-0090, 2019, doi: 10.4271/ 2019-24-0090. 328 A. Vacca, M. Chiodi, A. C. Kulzer, M. Bargende, S. Bucherer, P. Rothe, I. Kraljevic, H. Kollmeier, A. Breuer, H. Ruhland <?page no="329"?> Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber M. Balmelli a *, L. Merotto a , P.Soltic a a Empa, Swiss Federal Laboratories for Materials Science and Technology, Überland‐ strasse 129, 8600 Dübendorf, Switzerland Abstract: Several investigations have shown that the use of short-duration ignition pulses at a high repetition frequency, like Nanosecond Repetitively Pulsed Discharge (NRPD), leads to very favorable conditions for ignition and combustion. However, the detailed mechanisms of a short-pulsed discharge on the early stage of gas mixture ignition in engine-relevant conditions are still largely unclear. In this article, the plasma energy deposition efficiency is analyzed for different sparkplugs' electrode geometries in a constant volume ignition cell and in an optical pre-chamber. Moreover, the spark kernel and flame expansion are experimentally investigated for different air to fuel ratios and different initial turbulence levels in the optical pre-chamber using Schlieren imaging. The results of NRPD are compared with a classical inductive ignition system. An approach to analyzing the energy delivered during the discharge and its influence on plasma to early flame transition is discussed. A pulse generator is used to create the transient plasma necessary for NRPD ignition. The pulse generator produces 50 ns pulses at 10 kHz repetition frequency and can give maximal gap voltage up to 20kV. Different numbers of pulses allow the change of the energy deposited during the discharge. Plasma expansion and early flame kernel development are analyzed for different air to fuel ratios and for laminar or turbulent initial conditions in the pre-chamber. The plasma generation efficiency at 10 kHz pulse repetition frequency is higher when the surface gap sparkplug is used because the gap voltage decrease rate after a breakdown is slower, and the average breakdown voltage in the train of pulses has a higher value. The evolution of the flame position using NRPD shows two different regimes. The first one is where the expansion is not affected by the cell condition (AFR, turbulence level and number of pulses), which lasts approximately 1.5 ms. In this first phase, the expansion rate with NRPD is much higher when compared with <?page no="330"?> the inductive ignition system. The second regime is the region where the AFR and turbulence have an impact in propagation speed. In this region, the influence of longer NRPD is only apparent when the flame speed is low (low AFR and low turbulence) and has the effect of increasing the flame velocity. Keywords: Nanosecond repetitively pulsed ignition, non-thermal plasma, meth‐ ane-air mixtures, Schlieren, spectroscopy, pre-chamber. 1 Introduction It is known that ignition has major influences on the later stages of combustion: it gives rise to cyclic variability in engines, and it limits the range of operation both for air dilution or exhaust gas dilution [1]. More robust ignition compared to classical ignition system is achieved also in diluted conditions if the ignition energy is supplied in the shortest possible time [2,3]. Under engine conditions and nanosecond voltage pulses, the gas breaks down according to the streamer mechanisms [4]. Transient plasma (plasma in the transition from streamer to spark) exhibits the advantage of high electron temperature, which leads to more efficient production of active species [5]. A high amplitude pulse is needed to supply the minimum ignition energy with a single nanosecond discharge. This results in the requirement of complex high voltage insulation and careful shielding to cope with electromagnetic interferences. A similar performance at medium voltage can be achieved by using multiple pulses at high repetition frequency. Such an ignition system is called Nanosecond Repetitively Pulsed Discharge (NRPD) [6]. Tests performed using single-pulse discharges have shown that breakdown at much higher voltages than static ones is possible with nanosecond pulses. The higher the breakdown voltage, the higher the energy deposition to the plasma. The ideal pulse shape for ignition is achieved with the fastest possible pulse rise rate to incept the breakdown at higher voltage levels. The duration should be short (ten to twenty nanoseconds) to avoid the transition to an arc discharge. [4] Fast breakdown formative times under nanosecond voltage pulses have highlighted seed electron provision's important role. Evidence for a field-assisted emission from the electrode surface of seed electrons with a pressure-dependent onset field is reported in [7], which could allow for the creation of multiple streamer channels between the electrodes. [7] M. Balmelli, Y. Lu, R. Farber, L. Merotto, P. Soltic, D. Bleiner, J. Biela, C.M. Franck, Breakdown of Synthetic Air Under Nanosecond Pulsed Voltages in Quasi-Uniform Electric Fields, IEEE Access. 10 (2022) 53454-53467. https: / / doi.org/ 10.1109/ ACCESS.2 022.3175460 Although several investigations were devoted to NRPD ignition, most of the avail‐ able literature is limited to low pressure conditions and/ or large electrode gaps [7-9], 330 M. Balmelli, L. Merotto, P.Soltic <?page no="331"?> and the actual mechanisms involved in NRPD-based ignition in conditions relevant to engine applications are still largely unclear. To achieve a better insight on the mechanisms involved in energy deposition and flame kernel initiation when NRPD is applied in density and gap conditions relevant for engine applications, two different strategies are used in this work. First, the energy deposition mechanism is investigated comparing the classical shaped sparkplug with a surface electrode sparkplug. This allows obtaining new insight on the electrodes' geometry effect on energy deposition and on ignition efficiency. Second, a sparkplug with J-shape electrodes is used to ignite methane/ air mixtures in an optical pre-chamber setup, and the ignition and first phases of flame kernel evolution is investigated using Schlieren visualization. An algorithm for the analysis of the flame kernel evolution is developed that allows recognizing two different phases in NRPD-based ignition. Based on this analysis, a possible explanation for the NRPD ignition behavior is discussed. The combined approaches presented in this work provide a better understanding of the potential of NRPD for ignition, thus offering important guidelines for the optimization of NRPD-based ignition of gas mixtures in challenging conditions. 2 Materials and methods 2.1 Experimental setup 1: constant volume cell A constant volume cell introduced in [10] is used in conjunction with a commercial nanosecond pulse generator (FID 15-10NK) coupled to automotive sparkplugs to characterize repetitive pulsed discharge at 10 kHz repetition frequency. Table 1 lists the main specifications of the used pulse generator. Maximum amplitude 15 kV (@ 100 Ohm) Rise time (from-to % of maximum) 3-5 ns (10-50%) Pulse duration (at % of maximum) 50 ns (50%) Maximum energy in the pulse 80 mJ Maximum Pulse Repetition Frequency 10 kHz High-voltage Polarity positive Tabelle 1: Pulse generator specifications The description of the experimental setup used for nanosecond pulsed discharge is explained in detail in [4]. Hereafter are summarized the key features. First, the sparkplug is connected to the pulse generators through a coaxial cable with an impedance of 75 Ohms. Next, the center electrode is connected to the center core of the 331 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="332"?> coaxial cable, while the ground electrode is connected with the shield of the coaxial cable through the ignition cell. Two different sparkplugs are used in this investigation. One is a fiber-optics-equip‐ ped spark plug with a J-shaped ground electrode and a gap distance of 0.4 mm ( J-gap sparkplug). The other is a commercially available surface gap sparkplug (NGK 2522), with a gap distance of 0.75mm. Surface gap sparkplug seems advantageous to create large plasma surfaces, as shown in [11]. Figure 1 shows the sparkplugs' electrodes. Figure 1: Sparkplugs electrodes shapes used in the present work. On the left the J-gap sparkplug, on the right the surface gap is shown 2.2 Experimental setup 2: Optical Pre-chamber (OPC) A constant volume combustion chamber, the Optical Pre-Chamber (OPC, Figure 2), is used to investigate the plasma to flame transition under different operating conditions. The test rig contains two volumes, which are connected through a single nozzle. The smaller of the two volumes (pre-chamber) contains an ignition source (spark plug), which allows the ignition of the mixture. Both chambers have windows to allow optical access into the respective volumes, which enable optical measurements during the combustion event, as shown in Figure 2. The operation and control sequences of the OPC allow the accurate setting of conditions in both chambers before ignition (pressure, temperature, air to fuel ratio). First, the fuel is injected into the main chamber. Then a purge valve is opened to allow the pre-chamber air out while allowing the air-fuel mixture from the main chamber to flow through the nozzle into the pre-chamber. A graphical sequence of the processes is shown in Figure 3. The flow through the purge valve creates a dynamic air movement. The dwell time between the closing of the purge valve and the ignition event affects how much turbulence is dissipated. Dwell time variations are used as an indicator of laminar or turbulent conditions in the chamber [12]. 332 M. Balmelli, L. Merotto, P.Soltic <?page no="333"?> Figure 2: Overview of the Optical Pre-Chamber (OPC) setup. Figure 3: Sequence of the filling process. (1) The main chamber is filled with air, (2) the fuel is injected into the main chamber, (3) a purge valve is opened, (4) the air-fuel mixture flows The temperature at intake and pressure before ignition (12.4 bar and 363 K) are selected to correspond to a density of 11.6 kg/ m 3 (density condition of an engine having a compression ratio of 10 and pressure 1 bar at the beginning of compression). Three different air-fuel equivalence ratios (λ) were considered, namely λ =1, λ =1.4, and λ =1.8. Laminar or turbulent initial conditions in the pre-chamber were obtained by setting the dwell time to 0.5 and 0.005 s, respectively. In the OPC NRPD, ignition is compared against a classical inductive ignition system. The complete analysis of these results is reported in [13]. In this work, the effect of NRPD on plasma expansion to flame development is discussed. 2.3 Current and voltage measurement The electrical characterization of nanosecond pulsed discharge for the used setup is explained in detail in [4]. When a pulse produced from the pulse generator arrives at the anode, it is partially reflected and transmitted depending on the ratio between cable impedance and air-plasma resistance. The transmission line circuit equations describe for the current setup the voltage and current during the discharge. For a purely resistive load between the electrodes (R), the gap voltage (U gap ) and total current I tot can be reconstructed by measuring the incident (I pulse ) and reflected (I r ) pulse current and knowing the coaxial cable impedance (Z ) according to Equations (1), (2), and (3): 333 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="334"?> (1) (2) (3) (4) U gap = U pulse − U r I tot = I pulse + I r U pulse = Z I pulse ; U r = Z I r The energy contained in one pulse can be calculated by knowing the pulse current waveform according to Equation (4). E pulse = t 0 t end I pulse (t) 2 Z dt To measure the incident and reflected waveforms avoiding superposition, a 30 m long cable is used. The shielding is removed from a small section in the middle of the 30 m long coaxial cable. A shielded current monitor (Pearson current monitor model 6585 with a rise time of 1.5 ns) is placed around the exposed cable section to measure the current. In this way, the incident and reflected pulses are divided by 30 m of cable (ca. 150ns). Since the time delay between the incident and reflected pulse is longer than twice the pulse length, I pulse and I r can be measured without any superposition. The cable shielding is reconstructed around the current monitor. 2.4 Schlieren visualization technique In order to visualize the ignition and flame propagation in the pre-chamber a sin‐ gle-pass Schlieren arrangement is setup as shown in Figure 4. Figure 4: Simultaneous Schlieren setup for full pre-chamber at 10kFPS and close-up near the spark plug at 100kFPS. The system is equipped with a FOSP (fiber-optic spark plug) for ignition with NRPD systems 334 M. Balmelli, L. Merotto, P.Soltic <?page no="335"?> 2.5 Flame front measurement In the pre-chamber, a Matlab routine is used to detect the flame front position. The flame front is chosen as the point on the flame contour which is furthest from the central electrode of the sparkplug. Figure 5 shows the flame front detection at four different time instants (0.4, 0.6, 1.5, and 11 ms) of an experiment in lean condition (λ=1.8) where ten pulses are applied at 12.4 bar and 90°C in laminar condition. Figure 5: Flame front position measurement The red surface in Figure 5 represents the flame contour detected by the algorithm. For further post-processing, two flame positions are used. The first is the maximal radial distance on the detected contour from the central sparkplug's electrode; the yellow lines represent this distance in the four subplots in Figure 6. The second is the furthest flame position in the horizontal direction and is represented as blue vertical lines in Figure 6. This two-positions strategy was selected because following the flame in the vertical position resulted in incomplete tracking due to the sparkplug's ground electrode presence hindering the view of the kernel development. The proposed post-processing routine allows for the accurate tracking of the plasma and flame position during the early ignition phases. 3 Results 3.1 Voltage variation during multiple high voltage nanosecond pulses By assuming a purely resistive load between the electrodes and neglecting the sparkplug natural capacitance, the voltage and current at the gap depend solely on the ratio between coaxial cable impedance and air-plasma resistance (R) according to Equations(5) and (6). 335 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="336"?> (5) (6) (7) U gap = U pulse 2R Z + R I tot = U pulse Z 2Z Z + R When the nanosecond pulse amplitude is high enough, there is a transition from streamer to spark. During the spark phase, the current rises to several amperes. The electrical resistance between the electrodes, therefore, varies over several orders of magnitudes. Before the breakdown, the air-gas mixture act as a good insulator. The resistance between the electrodes can be assumed to have infinity impedance in comparison to the coaxial cable impedance. The generated transient plasma with high electron density is a good conductor. The 75 Ohm cable impedance is negligible compared to the air-gas resistance, whereas it is the only factor to consider when a spark is present. Two extreme regimes are recognizable from equations (5) and (6), namely a doubling of the pulse voltage when air is between the electrodes, and a limitation of the transmitted current when the electron density is high. The cumulative energy per pulse deposition to the plasma is calculated according to Equation (7). E plasma (t) = t 0 t end I tot (t)U gap (t)dt Figure 6 shows the evolution of the voltage between the sparkplug electrodes for two different sparkplugs. Ten pulses at 10 kHz were applied in both experiments. In the experiment depicted on the left side, a surface gap sparkplug is mounted in the constant volume cell, filled with air and methane (λ=1.6) at 10 bar and ambient temperature. On the right, the J-gap sparkplug is mounted on the OPC at 12.4 bar and λ of 1.8. The pressure in the constant volume cell is chosen to replicate the density condition of the OPC at ambient temperature. The gap voltage for the first, second, third, and ninth pulses is depicted in the upper subplots in different colors. The maximal voltages that would appear across the gap if no discharge were present is shown in black for the same pulses. The cumulative energy deposited to the plasma is plotted for the same pulses with the same color scheme in the bottom subplots. In the experiment with the surface gap, the first pulse was of lower amplitude; this frequently occurs due to the internal circuit of the pulse generator. Nevertheless, the first breakdown voltage and time are similar in the two experiments at ca. 20 kV and 30 ns, respectively, after the start of the pulse. The only difference is that for the J-gap sparkplug experiment, the gas breaks down with the first pulse while for the surface gap experiment with the second one. Similar energy deposition in the two cases is achieved with the first discharge (18 mJ and 22 mJ, respectively). The energy deposition 336 M. Balmelli, L. Merotto, P.Soltic <?page no="337"?> (8) and maximal gap voltage per pulse decrease rapidly for the J-gap sparkplug. After one to two pulses, a quasi-stationary energy per discharge of less than 10 mJ is reached. For the surface sparkplug, the maximal gap voltage decreases as well but at a much slower pace. In fact, due to the longer discharge time available (the breakdown happens earlier) and the still relatively high maximal gap voltage of 16 kV, the energy deposit to the plasma increases for the third pulse up to 28 mJ. Afterward, the breakdown voltage decreases, as does the plasma's energy input back to a quasi-stationary value of ca. 20 mJ. -10 0 10 20 30 40 50 60 70 Time [ns] 0 5 10 15 20 Gap Voltage [kV] -10 0 10 20 30 40 50 60 70 Time [ns] 0 10 20 30 Plasma Energy [mj] Pulse 1 Pulse 2 Pulse 3 Pulse 9 Figure 6: Different pulses voltages waveforms in a ten pulses NRPD at 10 kHz (left: surface gap sparkplug; right: J-gap sparkplug) The energy deposition to the plasma is the highest when the cable impedance links current and gap voltage, according to Equation (8). When this condition is fulfilled, the plasma impedance equals the cable one; there is no reflection, and therefore no energy waste. U gap, P max = I gap, P max Z Assuming that the impedance varies linearly from infinity to zero during the transition from streamer to spark, a higher plasma energy is achieved with a lower voltage decrease rate. Slower voltage decrease rates after the breakdown are visible for the surface sparkplug. The efficiency of the NRPD discharge can be defined as the ratio of the plasma to pulse energy. The efficiency of discharge for the surface gap sparkplug is 40%, while the one for the J-gap sparkplug is around 19%. Despite these findings, the J-gap sparkplug is used for the OPC experiments because it allows for a better view of the ignition process. Namely, the surface sparkplug would prevent the Schlieren visualization of the plasma and flame interaction in the early 337 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="338"?> ignition stages due to its geometry. Furthermore, the J-gap sparkplug allows further plasma analysis thanks to spectral techniques; these results are presented in [13]. 3.2 Ignition success and early flame development Schlieren images of inductive and NRPD (50 pulses at 10 kHz) at laminar and turbulent conditions are shown in Figure 7 and Figure 8 at four different times after discharge (from 1 to 4 ms). The sequence shows the evolution of a very lean case (λ = 1.8) at 12.4 bar. Figure 7: NRPD vs. inductive ignition in laminar and turbulent conditions, 1 and 2 ms after ignition Figure 8: NRPD vs. inductive ignition in laminar and turbulent conditions, 3 and 4 ms after ignition 338 M. Balmelli, L. Merotto, P.Soltic <?page no="339"?> For the NRPD ignition, a bright spark is visible at all depicted time steps. Ten pulses each ms for the first 5 ms are applied, and the pulse position is synchronized within a few hundred ns with the camera framerate. It can be seen that for NRPD, the flame quickly develops for both initial laminar and turbulent conditions. Instead, the inductive case is much slower, and after a few ms, it quenches before reaching the end of the pre-chamber (showed in [13]). Figure 9 shows the percentage of successful ignition for at λ equal to 1.8, 12.4 bar, and 90 °C. The ignition is successful if the flame can reach the end of the pre-chamber. Each experimental condition is repeated twenty times. Figure 9: Percentage of successful ignition for inductive and NRPD at different pulse numbers and 10 kHz repetition rate The green bars represent the success rate for inductive ignition, while the red ones are used for NRPD ignition. It can be seen that the percentage of successful ignition increases when NRPD is used. Five pulses are already enough for the turbulent case to ignite in all the repetitions, while ten pulses are needed for the laminar one. For higher pulse number, ignition is always successful. The effect of different pulse numbers can be investigated using the flame position development in time. 3.3 Flame position analysis Figure 10 shows the average flame front distance from the central electrode over 20 repetitions against time. The laminar cases are depicted on the left-hand side, and the turbulent ones on the right. From top to bottom, the air dilution increases: on top, the stoichiometric cases are shown, in the middle the lambda is equal to 1.4, and on the bottom it is equal to 1.8. In each plot, different colors are used for the different ignition strategies. The dotted line represents the inductive ignition system. The colored full lines (red to green) represent NRPD ignition with varying numbers of pulses (5, 10, 20, 339 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="340"?> and 50). In the bottom subplots, the black dots on the NRPD curves represent the time corresponding to the last delivered nanosecond pulse. 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Inductive 5 Pulses 10 Pulses 20 Pulses 50 Pulses 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Figure 10: Flame front distance from the sparkplug's central electrode In all the cases, NRPD ignition leads to faster propagation; the difference is larger for laminar cases and when the air to fuel ratio is higher. The difference between different pulse numbers is visible only at lean conditions, and it is more pronounced for higher air to fuel ratios. Two different flame propagation phases seem to be present. During the first 1.5 ms, the expansion for NRPD is not dependent on the gas composition nor the initial turbulence level. In this phase, NRPD exhibit a higher flame-plasma expansion rate when compared to inductive ignition; this higher expansion rate is also kept during the second phase. The second phase starts after 2 ms and is only visible where the flame speeds are expected to be low (high AFR or low turbulence). In this phase, higher pulse numbers are advantageous to achieve a higher flame speed. The black dots in the bottom subplots show when the NRPD ignition ends. As expected, the flame position starts to deviate from where the ignition is still active only after the black dots. 340 M. Balmelli, L. Merotto, P.Soltic <?page no="341"?> In order to overcome the limit of the ground electrode impeding the visualization of the kernel development in the vertical direction, it is interesting to examine the kernel development considering the velocity only in the “undisturbed" horizontal direction. Figure 11 shows the horizontal flame front distance for λ equal to 1.4 and 1.8 and for laminar and turbulent initial conditions. 0 1 2 3 time [s] 10 -3 0 2 4 6 Flame y distance [mm] Inductive 5 Pulses 10 Pulses 20 Pulses 50 Pulses 0 1 2 3 time [s] 10 -3 0 2 4 6 Flame y distance [mm] 0 1 2 3 time [s] 10 -3 0 2 4 6 Flame y distance [mm] 0 1 2 3 time [s] 10 -3 0 2 4 6 Flame y distance [mm] Figure 11: Lateral flame front distance from the sparkplug's central electrode The maximal distance of the visible flame front in the lateral direction is ca. 6 mm. Afterward, it reaches the limit of the window. The flame expansion for inductive is slower (blue points) while no constant trend is recognizable when a different number of pulses are applied, suggesting that different pulse sequences do not change the initial ignition phases. In order to examine the flame kernel development, two benchmarks were selected, namely 5 mm radial distance, and 10 mm radial distance. The distances are measured from the central electrode as illustrated in Figure 5. 341 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="342"?> Figuer 12 shows the time needed by the flame to reach 5 mm distance from the central electrode (radial distance) for initial laminar conditions on the top, and for initial turbulent conditions on the bottom, for the different ignition strategies investigated. Inductive5 pulses 10 pulses 20 pulses 50 pulses 0 0.002 0.004 0.006 0.008 0.01 time to 5 mm Laminar Inductive5 pulses 10 pulses 20 pulses 50 pulses 0 0.002 0.004 0.006 0.008 0.01 time to 5 mm Turbulent Figure 12: Time needed to reach 5 mm The time needed to reach the 5 mm distance for pulsed discharge is, on average, always less than 1 ms for all NRPD cases. Therefore, no appreciable difference is recorded between different NRPD pulse patterns at 10 kHz. The time to reach the 5 mm mark is affected by the AFR for the inductive case: the higher the air dilution, the longer the time. NRPD expansion to 5 mm is not affected by the AFR. Figure 13 shows the time the flame needs to reach the 10 mm radial distance from the central electrode for the different ignition strategies in laminar (top) and turbulent (bottom) conditions. 342 M. Balmelli, L. Merotto, P.Soltic <?page no="343"?> Inductive5 pulses 10 pulses 20 pulses 50 pulses 0 0.002 0.004 0.006 0.008 0.01 time to 10 mm Laminar Inductive5 pulses 10 pulses 20 pulses 50 pulses 0 0.002 0.004 0.006 0.008 0.01 time to 10 mm Turbulent Figure 13: Time needed to reach 10 mm The flame reaches the 10 mm distance faster when a higher number of pulses is used. The effect is more pronounced for the leaner condition. Distances above 10 mm are reached more quickly the closer the AFR gets to stoichiometric. Figure 14 shows the flame distance as a function of time for different λ values. The experiments are grouped for the different turbulence initial conditions (laminar and turbulent) and different ignition strategies. 343 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="344"?> 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Laminar Inductive = 1 = 1.4 = 1.8 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Laminar 10 Pulses 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Laminar 50 Pulses 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Turbulent Inductive 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Turbulent 10 Pulses 0 2 4 6 time [s] 10 -3 0 5 10 15 Flame distance [mm] Turbulent 50 Pulses Figure 14: Flame front distance from central sparkplug's electrode lambda variation For the inductive case (laminar and turbulent conditions), the flame distance starts to vary in dependence of AFR already after 0.5 ms. For NRPD, the effect on lambdas appears later, at ca. 1.5 ms. Higher pulse numbers reduce the propagation speed loss due to dilution. 4 Discussion Higher energy deposition to the plasma and higher energy deposition efficiency is reached with NRPD when a surface gap sparkplug is used compared to a classical J-type sparkplug. Two factors are responsible for higher efficiency. First, the breakdown voltage decreases for subsequent pulses at 10 kHz repetition frequency. The reduced local density reduction due to the temperature increases for the previous pulse or the presence of ionized molecules, excited species, and free electrons can explain this effect. This breakdown voltage decrease is less pronounced for the surface gap sparkplug, thus resulting in more energy delivered to the plasma. Second, the voltage decrease rate after the breakdown is lower for the surface gap sparkplug during the nanosecond pulse. Both observations can be explained by the quantity of plasma created with the 344 M. Balmelli, L. Merotto, P.Soltic <?page no="345"?> two discharges. The J gap sparkplug creates a small plasma volume between the central electrode and the ground one; all the energy is dumped in a relatively small channel, which rapidly transits to a highly conducting channels. The high currents heat up the channels, which break down more easily with the following pulse due to lower local density. This explanation also aligns with the results reported in [7]. Suppose the seed electrons necessary for breakdown are provided through field emission from the cathode surface. In that case, the more uniform electric field present for the surface gap sparkplug will allow for a larger active area, leading to a higher number of simultaneously propagating streamers and a larger plasma volume. [7] M. Balmelli, Y. Lu, R. Farber, L. Merotto, P. Soltic, D. Bleiner, J. Biela, C.M. Franck, Breakdown of Synthetic Air Under Nanosecond Pulsed Voltages in Quasi-Uniform Electric Fields, IEEE Access. 10 (2022) 53454-53467. https: / / doi.org/ 10.1109/ ACCESS.2 022.3175460. In the OPC, a more robust ignition is achieved with NRPD in comparison to classical inductive ignition system. The difference between different pulse numbers is visible only at lean conditions, and it is more pronounced for higher air to fuel ratios. This can be explained by the more energy required to ignite leaner conditions. Two regions of the ignition event are detected using NRPD, a first one lasting one to two ms, which is not affected by the OPC conditions (initial turbulence level and AFR) nor the number of pulses. The second region starts after 2 ms after the start of ignition and lasts till the end of the flame propagation. In this second phase, a higher number of pulses is advantageous to increase the flame propagation speed when the flame is usually slow (high AFR and low initial turbulence conditions). It is known that the heat addition of the ignition increases the flame speed during the early kernel expansion [14]. Pulsed discharge seems to enhance this effect significantly. In fact, the increases in ignition success rate at high AFR with NRPD could probably be explained by the increase in ignition radius using pulsed discharge [15]. The increase in flame speed when NRPD is applied after the flame has reached the critical radius is currently under investigation. A possible explanation could be the effect of the shockwave created by the nanosecond pulsed discharge [16]. This effect could enhance the turbulence level and hence the flame speed. 5 Conclusions This investigation of NRPD has shown that the sparkplug geometry has a significant impact on the efficiency of the discharge. Besides the sparkplug shape, the electrical circuit active during the discharge could be modified to increase the energy deposition by varying the sparkplug natural capacitance or the inductive behavior. Two discharge regimes are recognized. A possible way to exploit these regimes could be to use different repetition rates depending on the moment: high repetition rates 345 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="346"?> in the beginning to create a large kernel, and slower ones afterward to exploit the propagation speed enhancement and reduce the ignition energy input. Acknowledgements This study was conducted within the REAL project (Research on alternative combus‐ tion concepts for efficient gas engines) and financially supported by the Swiss Federal Office of Energy, contract number SI/ 501755-01. References [1] J.B. Heywood, Internal combustion engine fundamentals, McGraw-Hill, New York, 1988. [2] R. Maly, B. Saggau, E. Wagner, G. Ziegler, Prospects of Ignition Enhancement, in: 1983: p. 830478. https: / / doi.org/ 10.4271/ 830478. [3] J.C. Hilliard, G.S. Springer, eds., Fuel Economy, Springer US, Boston, MA, 1984. https: / / doi.o rg/ 10.1007/ 978-1-4899-2277-9. [4] M. Balmelli, R. Farber, L. Merotto, P. Soltic, D. Bleiner, C.M. Franck, J. Biela, Experimental Analysis of Breakdown With Nanosecond Pulses for Spark-Ignition Engines, IEEE Access. 9 (2021) 100050-100062. https: / / doi.org/ 10.1109/ ACCESS.2021.3095664. [5] S.V. Pancheshnyi, D.A. Lacoste, A. Bourdon, C.O. Laux, Ignition of Propane&ndash; Air Mixtures by a Repetitively Pulsed Nanosecond Discharge, IEEE Trans. Plasma Sci. 34 (2006) 2478-2487. https: / / doi.org/ 10.1109/ TPS.2006.876421. [6] S.M. Starikovskaia, Plasma assisted ignition and combustion, J. Phys. Appl. Phys. 39 (2006) R265-R299. https: / / doi.org/ 10.1088/ 0022-3727/ 39/ 16/ R01. [7] S. Barbosa, G. Pilla, D.A. Lacoste, P. Scouflaire, S. Ducruix, C.O. Laux, D. Veynante, Influence of nanosecond repetitively pulsed discharges on the stability of a swirled propane/ air burner representative of an aeronautical combustor, Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 373 (2015) 20140335. https: / / doi.org/ 10.1098/ rsta.2014.0335. [8] J.K. Lefkowitz, P. Guo, T. Ombrello, S.H. Won, C.A. Stevens, J.L. Hoke, F. Schauer, Y. Ju, Schlieren imaging and pulsed detonation engine testing of ignition by a nanosecond repetitively pulsed discharge, Combust. Flame. 162 (2015) 2496-2507. https: / / doi.org/ 10.1016 / j.combustflame.2015.02.019. [9] S. Lovascio, T. Ombrello, J. Hayashi, S. Stepanyan, D. Xu, G.D. Stancu, C.O. Laux, Effects of pulsation frequency and energy deposition on ignition using nanosecond repetitively pulsed discharges, Proc. Combust. Inst. 36 (2017) 4079-4086. https: / / doi.org/ 10.1016/ j.proci.2016.07. 065. [10] T. Kammermann, L. Merotto, D. Bleiner, P. Soltic, Spark-induced breakdown spectroscopy for fuel-air equivalence ratio measurements at internal combustion engine-relevant condi‐ tions, Spectrochim. Acta Part B At. Spectrosc. 155 (2019) 79-89. https: / / doi.org/ 10.1016/ j.sab .2019.03.006. 346 M. Balmelli, L. Merotto, P.Soltic <?page no="347"?> [11] D. Singleton, S.J. Pendleton, M. A. Gundersen, The role of non-thermal transient plasma for enhanced flame ignition in C 2 H 4 -air, J. Phys. Appl. Phys. 44 (2011) 022001. https: / / doi .org/ 10.1088/ 0022-3727/ 44/ 2/ 022001. [12] W. Vera-Tudela, et al., Investigations on spark pre-chamber ignition and subsequent turbulent jet main chamber ignition in a novel optically accessible test rig, Int. J. Engine Res. IJER-21-0092. (2021). [13] L. Merotto, M. Balmelli, W. Vera-Tudela, P. Soltic, Comparison of Ignition and Early Flame Propagation in Methane/ Air Mixtures using Nanosecond Repetitively Pulsed Discharge and Inductive Ignition in a Pre-Chamber Setup under Engine Relevant Conditions, Submitt. Publ. (2021). [14] K. Boulouchos, T. Steiner, P. Dimopoulos, Investigation of Flame Speed Models for the Flame Growth Period During Premixed Engine Combustion, in: 1994: p. 940476. https: / / doi.org/ 10. 4271/ 940476. [15] R. Maly, Ignition model for spark discharges and the early phase of flame front growth, Symp. Int. Combust. 18 (1981) 1747-1754. https: / / doi.org/ 10.1016/ S0082-0784(81)80179-8. [16] D.A. Xu, D.A. Lacoste, C.O. Laux, Schlieren Imaging of Shock-Wave Formation Induced by Ultrafast Heating of a Nanosecond Repetitively Pulsed Discharge in Air, IEEE Trans. Plasma Sci. 42 (2014) 2350-2351. https: / / doi.org/ 10.1109/ TPS.2014.2311328. 347 Plasma to Early Flame Kernel Transition under Nanosecond Repetitively Pulsed Discharge in an Optical Accessible Pre-chamber <?page no="349"?> A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD Dáire James Corrigan 1 , Sebastiano Breda 2 , Luca Arrizza 1 , Roberto Mariconti 1 , Stefano Fontanesi 3 1 2 3 Ferrari S.p.A., R&D CFD S.r.l., Università di Modena e Reggio Emilia Abstract: Knock has been studied by engine researchers for over one hundred years. It remains a key limitation on engine efficiency today. Designing engines for improved knocking tendencies, combined with calibrating and controlling precisely an engine at its knock limit, is fundamental for minimizing CO 2 output for hydrocarbon fueled combustion engines and to minimize energy consumption of carbon free energy carriers. High bandwidth cylinder pressure sensors have been the standard development instrumentation for knock for a number of decades. Knock is typically quantified by reduction of high pass filtered signal content from such sensors to a single value per cycle, indicative of the amplitude or signal energy. Manufacturers have experience on what is an acceptable level of such signals for their applications and calibrate accordingly. It has recently been shown however, that much more information can be extracted from a single cylinder pressure sensor when combined with acoustic modeling of the combustion chamber using Computational Fluid Dynamics (CFD). Among other things, the importance of the sensor position relative to the knocking location can be better understood and some estimation of the knock onset region can already be made. In the current paper, the combined experimental/ CFD approach has been extended to two cylinder pressure sensors. One of these is a conventional flush-mounted research sensor, requiring a dedicated indicating bore, whilst the second is integrated in an instrumented spark plug. This approach was chosen because it does not require any further modifications to an engine in comparison to what is already typical in a development and calibration environment. It is shown that the combined methodology allows a more robust quantification of knocking <?page no="350"?> Equation 1 intensity and more precise estimation of potential knocking locations to be made, without having to move to a more invasive instrumentation setup or expensive and potentially delicate optical measurements. 1 Background The 2015 Paris agreement set the target of limiting global warming to a maximum of 2 °C in comparison to pre-industrial levels and ideally to within 1.5 °C [1]. A key part in achieving this goal is a reduction of CO 2 emissions. Road transport contributes 11.9% to total global greenhouse gases [2], and the majority of these vehicles at present are powered by the Internal Combustion Engine (ICE). ICE powered vehicles running on fossil fuel contribute to global warming. Whilst it is foreseen that the market share of the ICE in road transport will diminish as more fully electric vehicles are purchased, a large number of ICE powered vehicles will likely still be on the road in 2050. Many of these will be part of hybridized, electrified powertrains. Assuming operation on a fossil derived fuel, the main way an ICE can contribute to reduced CO 2 output is through high efficiency. It is also possible to render an ICE powered vehicle carbon neutral from a Tank To Wheel (TTW) perspective by using non-fossil derived fuels. The energy expenditure involved in creation of such fuels is high however, hence high engine efficiency is once again key in keeping this to a minimum. With this in mind, it is important that researchers and Original Equipment Manu‐ facturers (OEMs) continue to investigate methods for making the ICE more efficient. 1.1 Limits on SI engine efficiency The majority of Light Duty Vehicles (LDVs) are powered by Spark Ignition (SI) engines. The idealized thermodynamic cycle which describes SI engine operation is the Otto cycle, as described in Equation 1, where η is the cycle efficiency, γ is the ratio of specific heats of the working fluid and CR is the compression ratio. η = 1 CR γ − 1 It is clear that to maximize efficiency, a high compression ratio is required. For an SI engine, the main limit on the highest compression ratio that can be tolerated is knock. 1.2 Knock Description Knock has been studied intensely by engine researchers for over one hundred years [3]. It was originally the name given to the unpleasant noise made by an engine under certain operating conditions [4]. It was hypothesized at an early stage that the cause of the phenomenon was autoignition of the trapped charge in the end-gas region [5]. 350 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="351"?> There was some disagreement on this over the years [6] but this is widely accepted today. Autoignition may result in heat release rates much higher than those associated with flame-front propagation [7]. If the local heat release results in sufficiently rapid gas expansion, a pressure wave is created [8]. This pressure wave will travel across the combustion chamber, reflecting on solid boundaries. Standing waves will therefore result. These have characteristic frequencies, which were explained by Draper [9] [10] in the 1930s as acoustic modes of the combustion chamber. The wave action will also result in structural vibrations of the engine [11], which can lead to audible noise. Strong knocking is also associated with increased heat transfer [12] and higher mechnical loading. It may therefore result in structural damage to the engine if it is not avoided, or at least controled. 1.3 Knock Avoidance Techniques There are two main ways of avoiding knock. One is to reduce the risk of autoignition. From an engine design perspective, the main way of achieving this is by limiting the charge temperature for a given pressure. Charge cooling and Miller [13] cycles are examples of techniques frequently applied to this end. A more knock resistant fuel can also be used. A fuel’s knock resistance is typically characterized by its octane values. These are based on engine testing approaches developed in the 1920s in the case of the RON test [14] and the 1930s for the MON test [15] by the Cooperative Fuels Research (CFR) group. Such results may be difficult to transfer to modern engine types however, although a number of methods of doing such have been suggested over the years [16], [17]. Another approach is to use a Rapid Compression Machine (RCM) [18]. Here, the time until autoignition can be measured in a more direct method over a range of pressures, temperatures, equivalence ratios and other variables of interest. Livengood and Wu [19] suggested a method of comparing such results to engine test data in their seminal paper. This was based on the idea of a critical species concentration. In order to arrive at this concentration, a number of individual reaction steps must take place. Each of these has a characteristic rate, which depends on temperature, pressure, species concentration and other variables. A Negative Temperature Coefficient (NTC) region may exist for some hydrocarbons, particularly paraffins (alkanes). An integration technique can be applied to account for the varying conditions in an engine and this approach is still utilised today. Figure 1 shows the outcome of IDT measurements for Unleaded Gasoline (ULG) with a Research Octane Number (RON) of 95 based on fit terms from [20]. These are compared to estimated IDT maps based on the commonly used Douaud and Eyzat method. 351 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="352"?> Figure 1: Ignition Delay Times for RON 95 ULG, based on Yates et al. compared to estimated times from Douaud and Eyzat method Autoignition will not have time to occur if all the air/ fuel mixture is consumed first by flame-front propagation. Acceleration of the flamefront is normally performed by increased turbulence generated by techniques such as use of high tumble ports [21] or high squish pistons [22]. Due to the complexities of modern engine geometries and the in-cylinder flowfields and mixture distributions, the flame-front will not propagate 352 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="353"?> in a perfectly uniform manner across the chamber. Knock is therefore more likely to occur in regions where the flame-front arrives late. These are often regions of negative flame curvature [23]. A knock resistant engine design is one where the flame-front consumes the charge rapidly, particularly in areas at higher risk of autoignition due to local temperatures. If this is done well, it can be expected that the autoignition onset will be distributed randomly around the periphery of the chamber. If, on the other hand, autoignition events occur primarily in a single region, this implies room for optimization such as improved cooling or more appropriate mixture formation and charge motion. To understand how to improve an engines knock limit, it is therefore useful to know where knock is occurring. 1.4 Typical Knock Instrumentation Knock instrumentation can be divided into three main categories: • Fundamental research: pressure indication complimented by optical techniques • Development: pressure indication only • Production: accelerometers or ion-current sensing Whilst the optical techniques used in research settings are excellent for giving clear information on knock locations, they are typically more expensive and often more delicate than pressure indication. Production sensors must be inexpensive and robust but generally are relied upon only to give binary feedback of knock occurrence. Development pressure indication sensors, such as the model shown in Figure 2, are robust and have high bandwidth. They have been used to triangulate knock in the past, but this technique requires at least three sensors. Such an installation may be prohibitively invasive, or even unfeasible, for a multicylinder development engine. Such development engines used for calibration purposes typically run with one dedicated indicating bore per cylinder. An alternative is to use an indicating spark plug, such as that shown in Figure 3. Here a miniature sensor is installed inside the spark plug itself. Figure 2: A typical development pressure transducer - Reference [27] 353 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="354"?> Figure 3: An indicating spark plug - Reference [28] CFD has recently been applied to better understand pressure wave behavior in both simplified [25] and complex [26] combustion chamber geometries. The insight given by such modeling allows extraction of further information from the pressure sensor signal. The current study aims to see what improvements in knock understanding can be gained through use of just two sensors (a flush-mounted development grade sensor [27] in a dedicated indicating bore and a miniaturised sensor installed in a spark plug [28]) complemented by CFD. Such an approach could easily be applied in any typical test cell environment with minimal additional costs or engine modifications. 2 CFD Activity 2.1 Engine geometry and mesh The engine considered in this investigation was the Ferrari F154FA, used in the SF90 Stradale road car. This is a turbocharged engine with a centrally mounted direct injector and a compression ratio of 9.5: 1. Further details of the engine are shown in Table 1 and a mesh of the combustion chamber in Figure 4. It can be seen that the chamber features a pronounced pent-roof and an offset piston bowl. The spark plug is also offset towards the exhaust side of the chamber. Figure 4: Combustion chamber mesh of test engine 354 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="355"?> Name F154FA Type V8 Turbo Bore (mm) 88 Swept Volume 3990 cc Injection System Direct, Central Compression Ratio 9.5: 1 Maximum Speed (rpm) 8000 Specific Output (CV/ L) 195 Table 1: Details of test engine 2.2 Computational methodology CFD models were constructed using SIMCENTER STAR-CCM+ version 2020.3, as in previous activity [25], [26]. The sudden heat release of an autoignition event was represented by superimposing a 1000 W heat source on a spherical surface, obtained by subtracting a 2.5 mm diameter solid sphere from the fluid volume. This sphere was translated to simulate different cases. The heat source starts with a 0.03 ms delay from the beginning of the simulation and the duration of the source is fixed to 0.1 ms, as shown in Figure 5. Figure 5: Heat source power profile 355 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="356"?> A target mesh size of 1 mm was used, coherent with the output of mesh sensitivity sweeps in the previous study. A timestep of 0.5 µs was applied, as in previous activity. This corresponds to a Courant number of 0.2 - significantly better than that required for the CFL (Courant Friedrichs Lewy) condition. The working fluid was set as air at standard atmospheric conditions, again taking inspiration from the historic Draper studies [9], [10]. Expected acoustic velocities, and hence resonant frequencies, for knocking combustion in an engine will be approximately three times higher. Further details of the calculation setup can be found in [26]. 2.3 Virtual Sensors 61 pressure measurement locations in a plane orthogonal to the cylinder axis were implemented in the model. This was with a view to post-processing, which was performed using MathWorks MATLAB R2020b. These “virtual sensors” were on a single plane offset into the chamber by 1 mm from the cylinder head face (z = -1 mm). The layout of sensors is shown in Figure 6. There are five groups of twelve, each group on a diameter concentric to the cylinder axis. A final 61st sensor was at the centre of the chamber. This group of sensors was used, among other things, to extract mode shapes from the simulation model. Figure 6: Virtual sensor layout 356 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="357"?> A further two sensors were added corresponding to the experimental transducer locations. One of these sensors is flush mounted, hence there was no need to model complex geometry in front of the sensor element. This sensor is located between the intake and exhaust valves near the periphery of the chamber. The second sensor is integrated into the spark plug. It is therefore much closer to the center of the chamber and also features a narrow passage (indicating bore) between the combustion chamber and the measuring element. The impact of this detailed geometry was examined in CFD. Further details of the sensors are given in Table 2. Sensor Type: Kistler 6045A Kistler 6113C Diaphragm Location: Flush Recessed Sensor Position: Periphery, transverse plane Mid chamber, symmetry plane, exhaust side Natural frequency: > 80 kHz > 120 kHz Table 2: Details of pressure sensor characteristics 2.4 Base Case The piston height was provisionally set at a position corresponding to 25° ATDC (After Top Dead Centre) - shortly after knock was seen to occur in the experimental dataset. Taking 0° as the symmetry plane on the exhaust side of the chamber and positive angles corresponding to clockwise locations relative to this, the heat source was provisionally placed at 45° and a radial distance of 37 mm from the cylinder axis. This position approximately corresponds to the exhaust valve pocket region - considered prone to knock in a previous study [26]. Data from the 61 sensors were interpreted in two manners. Firstly, Continuous Wavelet Transforms (CWT) were performed. The Morse wavelet was used with a time-bandwidth product of 120 and 12 voices per octave (semitone resolution). The resulting amplitudes in each time-frequency bin were averaged across the 61 sensors to give a space averaged time-frequency plot. This is shown in Figure 7. The effect of the heat source ignition can be seen as a short term event with a broad frequency spectrum. Thereafter, horizontal bands can be seen. These are the modal frequencies of the chamber driven by standing wave behavior. 357 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="358"?> Figure 7: Mean of 61 CWT Scalograms, base case Fast Frequency Transform (FFT) analysis was also applied for each sensor. A peak finding algorithm was used to extract key frequencies from the mean Power Spectral Density (PSD), as shown in Figure 8. 358 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="359"?> Figure 8: Mean Power Spectral Density with peak recognition for mean of all sensors, base case Band Pass Filtering (BPF) was then applied to the data from all sensors in narrow bands corresponding to the identified frequency peaks. In this way, individual modes could be isolated. The time sample corresponding to the maximum standard deviation of pressure across the chamber was automatically selected to highlight the modal shapes for all modes of interest. This is shown in Figure 9. The nomenclature of Draper is employed with mode names 'Fxy' where 'x' is the number of nodal planes and 'y' is the number of nodal cylinders (excluding the cylinder of the combustion chamber itself). 359 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="360"?> Figure 9: First six modal pressure distributions, base case A cylindrical combustion chamber has a unique acoustic diameter. This means that the modes that are excited do not depend on the heat source angular location and simply orient themselves according to the initial stimulus. This is not the case for some modern combustion chambers, such as that under study in the present work. As has previously been described [26], this effectively has two acoustic diameters - one corresponding to the symmetry plane and another orthogonal to this, both of which are perpendicular to the cylinder axis. This can clearly be seen in Figure 9 where there are two F10 modes - one at 2455 Hz and one at 2727 Hz. 2.5 Piston height sensitivity The onset of knock in the current dataset was found to occur shortly after TDC. The piston is therefore descending. The most appropriate piston position to use for a stationary mesh calculation is not immediately apparent, as pressure oscillations will continue for some milliseconds after the autoignition event. A piston height sweep was performed. It became apparent that the characteristic mode frequencies were sensitive to piston height when the latter was close to TDC. This is because features such as squish area become increasingly relevant as the cylinder volume is reduced. For low piston positions, the dominant acoustic diameter is that of the bore. Figure 10 shows how the modal frequencies change with piston height. Note how the relationship between the two F10 peaks depends on this variable. The piston height for the remainder of the calculations was therefore chosen to match the F10 modes frequency ratio to that seen in the experimental data. This corresponds to around 20° CA ATDC. 360 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="361"?> Figure 10: Mean PSD response to piston position 2.6 Sensitivity to non-ideal sensor installation The effect of the “non-ideal” pressure transducer of the indicating spark plug was studied in some detail in CFD. Firstly, a pressure measurement was implemented in the model at the spark plug, but flush with the combustion chamber and without any transducer bore. Secondly, a simplified cylindrical bore was introduced but the measuring element was retained as flush with the chamber. Thirdly the measurement point was moved to the closed end of the cylindrical transducer bore. Lastly, the detailed geometry of the transducer passage was included, as supplied by the sensor manufacturer. The impact of each of these steps on the measurement of the incident wave is shown in Figure 11 and on the PSD of the entire measurement duration in Figure 12. It can be seen in Figure 11 that measuring the pressure at the end of the transducer bore has the effect of amplifying the measurement, even from the initial incident wavefront. A characteristic frequency associated with the passage is evident in Figure 12 and is somewhat higher for the real geometry than for a cylindrical approximation. It is interesting to note however, that up until around 10 kHz, the frequency spectrum is very similar for all variants. Whether or not the transducer is mounted flush or 361 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="362"?> with a small indicating bore is therefore largely irrelevant for the estimation of mode strengths up until around F50. This is an important result. Figure 11: Effect of passage geometry on pressure measurement at spark plug location 362 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="363"?> Figure 12: Effect of passage geometry on PSD of pressure data at spark plug location 2.7 Autoignition location sensitivity Having set the piston height and investigated the influence of the transducer bore, the significance of autoignition location on the pressure measurements was then studied in CFD. The autoignition location was swept in steps of 15° at various radial distances from the center of the chamber: 44 mm, 37 mm, 30 mm and 23 mm. Figure 13 shows the results for the 44 mm sweep. Results were qualititavely similar for the other distances. 363 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="364"?> Figure 13: Mean PSD response to angular heat source location, radial distance 44 mm Note that the relative strength of the two F10 modes depends on the angular location of the heat source. The flush mounted transducer is on an antinode for F10a and a node for F10b. The opposite is true for the plug mounted transducer. Figure 14 shows how the strengths of these modes vary at the experimental measurement locations. They are mutually 90° out of phase. The mean mode strength at the flush mounted transducer is higher, as it is closer to the periphery where this mode results in the highest amplitude oscillations. In fact the ratio of the modes at the measurement locations over the sweeps was found to be very consistent for all radial distances: 2.97 ± 2%. 364 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="365"?> Equation 2 Figure 14: Relative strengths of the F10a mode measured at the flush mounted transducer and F10b measured at the plug transducer over the angular sweep at 44 mm radial distance This fact allows a knock intensity index to be created which is largely independent of the heat source angular location. The index is shown in Equation 2 K I F 10 = P SD F 10aF lusℎ + 3P SD F 10bP lug The ratio of the two modes at the measurement locations varies strongly with heat source angular location but is almost independent of radial location, as shown in Figure 15. This diagram allows four potential angular locations of a knocking heat source to be determined by comparing with the experimental ratio of F10a as determined by the flush mounted transducer data to F10b from the spark plug transducer. 365 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="366"?> Figure 15: Ratio of F10a at the flush mounted transducer location to F10b at the spark plug transducer location 3 Experimental activity 3.1 Instrumentation Layout and Measurement Conditions The measurement chain used for this work was as follows. Along with the already described sensors, a Kistler 5064E charge amplifier was used with 200 kHz bandwidth. Data were acquired at 0.1 deg crank angle resolution by an Indimaster 672 acquisition system. At 3000 rpm, this corresponds to a sampling rate of 180 kS/ s or a timestep of 5.6 ms. The analogue input of the acquisition system features 100 kHz anti-aliasing filters. An AVL 365C pulse-multiplier was applied to the 60-2 engine encoder used by the Engine Control Unit (ECU) to generate the reference angle base. The engine was operated at 3000 rpm and 26 bar of Indicated Mean Effective Pressure (IMEP) with stoichiometric mixture of 98 RON ULG. Ignition timing was selected for stronger knocking than applied in the production calibration in order to generate an adequate number of knocking cycles for analysis. 366 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="367"?> 3.2 Experimental Data Post Processing 3.2.1 Acoustic Velocity Compensation Post processing of the experimental data was also carried out in MATLAB. The geometry of the combustion chamber in a running engine of course varies with piston position. The acoustic velocity is also variable due to changing gas properties. This was estimated using a single zone thermodynamic model based on the First Law of Thermodynamics. The ratio of the estimated acoustic velocity at each instant to that of air at standard atmospheric conditions was used to compensate the measurement data. Put simply, if the estimated acoustic velocity at a given timestep was three times that of air at reference conditions, the time base was locally stretched by the same factor. The adjusted data were then re-sampled at a constant rate for ease of processing. In such a manner, experimental data from the fired engine and CFD data from the stationary model with fixed gas properties could be directly compared. Figure 16 shows an example of the effect of this process on a single cycle of data. The impact of spatially varying acoustic properties was not considered. It is assumed that this effect is rather small shortly after knock has occurred, as a large majority of the mixture has already been burned and the remainder will be shortly afterwards. This will be commented upon later in this work. Figure 16: Acoustic velocity compensation of experimental data 367 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="368"?> 3.2.2 Average Power Spectrum Analysis of Experimental Data Figure 17 shows the average of 500 cycles PSD analysis through application of FFTs for both sensors. The two distinct F10 mode frequencies can be clearly seen and are well aligned to the CFD, although note that this is not surprising as the piston height was chosen to match the ratio between these modal frequencies. As could be expected by the sensors’ locations relative to the predicted modal pressure distributions, the flush mounted transducer measures only the lower frequency F10a mode whilst the indicating spark plug registers activity only for the higher frequency F10b. The experimental F20 and F00 peaks are slightly higher than predicted by CFD but the error is of the order of 100 Hz. F30 appears well predicted as does F11. The relative strength of F10b and F10a modes was determined on a cyclic basis for the entire dataset of 500 cycles with a view to knock localisation using the relationship shown in Figure 15. Figure 17: Average FFT spectral analysis of acoustically compensated experimental data and compar‐ ison to frequency peaks predicted by CFD analysis 3.2.3 Multiple Autoignition Event Screening The CFD analysis is based on the idea that a single autoignition event is responsible for the modal pressure oscillations in the chamber. It may be the case however, that 368 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="369"?> multiple autoignition events occur during a single cycle. The data were therefore analyzed with a view to identifying cycles with multiple knocking events and removing them from the dataset. Figure 18: Example of fourth order fit quality of pressure decay used to determine whether a single knock event exists The method chosen to do this was based on the idea that pressure oscillations should continuously diminish after a knocking event. A fourth order polynomial function was found to fit the cumulative absolute Band Pass Filtered (BPF) pressure decay very well. The BPF frequency band chosen was 2-12 kHz to avoid the influence of the transducer bore. The fitting was applied starting from the end of the relevant portion of the cycle and gradually increasing the number of samples until the start of the analyzed section. For a single knock event, the correlation coefficient value should be close to the maximum just after the start of knock. Such is the case for Cycle 4 and the flush mounted transducer in Figure 18. For the plug transducer on the other hand, the correlation coefficient curve shows a higher fit value in the middle of the analysed section - a multiple autoignition event cannot be excluded for this cycle. Around 20% of cycles exhibited evidence of multiple autoignition events on at least one sensor. 369 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="370"?> 3.2.4 Start of Knock The criterion for Start of Knock (SoK) in the retained cycles was defined as the first timestep at which the BPF pressure exceeded 1/ 3 of the peak value of that cycle. This was at around 8° ATDC for both sensors on average with a standard deviation of 1.6° CA. The mean Mass Fraction Burned (MFB) at SoK was 69% for both sensors. Assuming a density ratio of unburnt to burnt gas of 4: 1 as suggested by Heywood [29], the Volume Fraction Burned (VFB) can be estimated. VFB at SoK was determined as 90% for both sensors. As a first estimate of radial distance of the knock occurence, a cylindrical burnt zone was assumed to define a Radial Fraction Burned (RFB). This gives an estimated radial distance at SoK of 42 mm. Obviously the real flamefront can be far from a cylindrical surface, but as a first assumption, it would appear reasonable to assume that knocking events in the present dataset occur close to the bore wall. Figure 19: Mass, Volume and Radius Fraction Burned at Start of Knock 3.2.5 Final autoignition location estimation Based on the above analysis, it should be possible to narrow down an autoignition location for a cycle with a single knock event to one of four possibilities. The relative strength of the F10 modes gives four potential directions. The location is highly likely 370 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="371"?> to be in the periphery of the chamber based on Figure 19. The last step is to look at expected relative wave arrival time to each sensor for each of these locations and compare it to the experimental relative wave arrival time. This was performed for all cycles in the dataset where no multiple knocks were identified. Figure 20 shows an example of the process for Cycle 1. Figure 20: Knock location identification, Cycle 1 This process was performed for all single-knocking cycles. The median location error was 8 mm. Cycles with greater than 10 mm of location error were excluded from the final output. Remaining cycles had an estimated location error of below 5 mm. These are shown in Figure 21 together with the knocking intensity. Equation 2 was used to define Knock Intensity due to its low position sensitivity. It can be seen that knocking cycles are distributed around the combustion chamber but there are some patterns. For example, the fewest cycles are in the lower right sector and what cycles there are are weak. The strongest cycles are in the upper two sectors. Some strong cycles are seen towards the exhaust valve pockets, as expected from previous studies. There are also some strong cycles on the upper intake side. The bias of cycles towards the upper side of Figure 21 has a potential explanation. This is the inter-bore area between cylinders 371 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="372"?> 8 and 7 whereas the lower side is the end of the engine and hence does not share its cooling with any other cylinders. Figure 21: Knock location estimation and knock intensity. Left side of diagram is the exhaust side and the right part is the intake. The upper part is the inter-bore region whereas the lower part is the end of the engine. 4 Conclusions A new approach to understanding knock in internal combustion engines has been presented. This is based on twin indicating sensors combined with CFD based acoustic analysis of the combustion chamber. As one of these sensors was an indicating spark plug, only a single dedicated indication bore is required. This is commonly already present in development and calibration engines. The CFD model is also simple to implement. The following observations were made: • CFD acoustic modelling of combustion chambers allows for better understanding of high frequency data from indicating sensors. • The necessary transducer bore of the spark plug sensor has little influence on the evaluation of mode strengths up until around 30 kHz on the fired engine but introduces a small delay on the signal. • Use of twin sensors also allows a more robust estimation of knock intensity to be made, which is almost independent of knock location. A new knock index was proposed based on relative F10 mode strengths from both sensors. 372 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="373"?> • A relationship between relative mode strength and autoignition source can be established. This gives four possible knock locations per cycle. • A new method for identifying multiple knocking events in a single cycle has been proposed. • Comparisons of wave arrival time at each sensor together with frequency analysis can then be used to determine the final knock location. • Results suggested the interbore area of the chamber was more prone to knock, implying that an improvement in detailed cooling system design could bring knock and hence efficiency benefits. 5 Future Activity The proposed methodology can be considered an alternative to optical spark plug based methods of knock location determination such as AVL Visio-knock. A comparison will therefore be made in the future of the knock locations determined by the new method and that from optical systems. Acknowledgement Roger Leutwyler and Andrea Bertola of Kistler for supplying detailed transducer geometry used in the CFD investigation. 6 Bibliography [1] United Nations Framework Convention on Climate Change, “Report of the Conference of the Parties on its twenty-first session, held in Paris from 30 November to 13 December 2015},” Paris, 2016. [2] Our World In Data, “Global greenhouse gas emissions by sector,” [Online]. Available: https: / / ourworldindata.org/ emissions-by-sector. [Zugriff am 02 04 2022]. [3] D. J. Corrigan and S. Fontanesi, “Knock: A Century of Research,” SAE International Journal of Engines, vol. 15, no. 1, pp. 57-127, July 2021. [4] T. Midgely and T. A. Boyd, “Methods of Measuring Detonation in Engines,” in Pre-1964 SAE Technical Papers, SAE International, 1922. [5] H. R. Ricardo, “Recent Research Work on the Internal-Combustion Engine,” in Pre-1964 SAE Technical Papers, SAE International, 1922. [6] S. Curry, “The Relationship Between Flame Propagation and Pressure Development During Knocking Combustion,” in Pre-1964 SAE Technical Papers, SAE International, 1963. [7] Y. B. Zeldovich, “Regime Classification of an Exothermic Reaction with Nonuniform Initial Conditions,” Combustion and Flame, vol. 39, no. 2, pp. 211-214, 1980. [8] D. Bradley, C. Morley, X. J. Gu and D. R. Emerson, “Amplified Pressure Waves During Autoignition: Relevance to CAI Engines,” in SAE Powertrain & Fluid Systems Conference & Exhibition, 2002. 373 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="374"?> [9] C. S. Draper, “The Physical Effects of Detonation in a Closed Cylindrical Chamber,” in NACA Technical Reports, National Advisory Committee for Aeronautics, 1935, pp. NACA-TR-493. [10] C. S. Draper, “Pressure Waves Accompanying Detonation in the Internal Combustion Engine,” in Power Plants and Propellers Session, Sixth Annual Meeting, I. Ae. S., Institute of Aeronautical Sciences, 1938. [11] T. Priede and R. K. Dutkiewicz, “The Effect of Normal Combustion and Knock on Gasoline Engine Noise,” in SAE Noise and Vibration Conference and Exposition, SAE International, 1989. [12] W. Lee and H. J. Schaefer, “Analysis of Local Pressures, Surface Temperatures and Engine Damages under Knock Conditions,” in SAE International Congress and Exposition, SAE International, 1960. [13] R. Miller, “Supercharging and internal cooling cycle for high output,” in Transactions of the American Society of Mechanical Engineers, ASME, 1947. [14] ASTM International, Standard Test Method for Research Octane Number of Spark-Ignition Engine Fuel, ASTM International, 2019. [15] ASTM International, Standard Test Method for Motor Octane Number of Spark-Ignition Engine Fuel, ASTM International, 2019. [16] C. B. Veal, “C.F.R. Committee Report on 1934 Detonation Road Tests,” in Pre-1964 SAE Technical Papers, SAE International, 1935, p. 350094. [17] G. Kalghatgi, “Fuel Anti-Knock Quality - Part I. Engine Studies,” in SAE 2002 World Congress & Exhibition, SAE International, 2002. [18] C. F. Taylor, E. S. Taylor, J. C. Livengood, W. A. Russell and W. A. Leary, “Ignition of Fuels by Rapid Compression,” in Pre-1964 SAE Technical Papers, SAE International, 1950. [19] J. Livengood and P. Wu, “Correlation of Autoignition Phenomena in Internal Combustion Engines and Rapid Compression Machines,” Symposium (International) on Combustion, vol. 5, no. 1, pp. 347-356, 1955. [20] A. D. B. Yates, A. Swarts and C. L. Viljoen, “Correlating Auto-Ignition Delays And Knock-Limited Spark-Advance Data For Different Types Of Fuel,” in 2005 SAE Brasil Fuels & Lubricants Meeting, 2005. [21] P. O. Witze, J. K. Martin and C. Borgnakke, “Measurements and Predictions of the Precombustion Fluid Motion and Combustion Rates in a Spark Ignition Engine,” in SAE Technical Papers, SAE International, 1983. [22] I. Nagayama, Y. Araki and Y. Iioka, “Effects of Swirl and Squish on S.I. Engine Combustion and Emission,” in SAE Technical Papers, SAE International, 1977. [23] F. Catapano, M. Costa, G. Marseglia, P. Sementa, U. Sorge and B. M. Vaglieco, “Experimental and Numerical Investigation in a Turbocharged {GDI} Engine Under Knock Condition by Means of Conventional and Non-Conventional Methods,” SAE International Journal of Engines, vol. 8, no. 2, pp. 437-446, 2015. [24] P. M. Liiva, J. N. Valentine, J. M. Cobb and W. P. Acker, “Use of Multiple Pressure Transducers to Find In-Cylinder Knock Location,” in International Fuels & Lubricants Meeting & Exposition, SAE International, 1992, p. 922368. 374 Dáire James Corrigan, Sebastiano Breda, Luca Arrizza, Roberto Mariconti, Stefano Fontanesi <?page no="375"?> [25] D. J. Corrigan, S. Breda and S. Fontanesi, “A Simple CFD Model for Knocking Cylinder Pressure Data Interpretation: Part 1,” in 15th International Conference on Engines and Vehicles, Capri, 2021. [26] D. J. Corrigan, S. Breda, S. Fontanesi and F. S. Mortellaro, “Knocking Cylinder Pressure Data Interpretation for Modern High-Performance Engines—A Computational Fluid Dynamics Informed Approach,” SAE International Journal of Engines, vol. 16, no. 2, 2022. [27] Kistler, “High temperature pressure sensor for combustion engine measurements: 6045A,” [Online]. Available: https: / / www.kistler.com/ files/ document/ 000-618e.pdf ? callee=frontend. [Zugriff am 2 July 2022]. [28] Kistler, “M10x1 Measuring spark plug with integrated 3 mm cylinder pressure sensor - Type 6113C,” [Online]. Available: https: / / www.kistler.com/ files/ document/ 003-281e.pdf ? callee=fr ontend. [Zugriff am 2 July 2022]. [29] J. B. Heywood, Internal Combustion Engine Fundamentals, Singapore: McGraw-Hill, 1988. 375 A new low-cost method for knocking analysis: twin indicating sensors combined with combustion chamber modelling in CFD <?page no="377"?> 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines Marco Hess 1 , Michael Grill 2 , Michael Barende 1 , André Casal Kulzer 1 1 2 Institute of Automotive Engineering (IFS), University of Stuttgart, Pfaffenwaldring 12, 70569 Stuttgart, Germany Research Institute of Automotive Engineering and Vehicle Engines Stuttgart (FKFS), Pfaffenwaldring 12, 70569 Stuttgart, Germany Abstract: SI engines are typically designed to operate at the knock boundary, as engine knock limits their efficiency and thus further reduction of CO 2 emissions. Therefore, when performing an engine process simulation, an accurate knock model is needed to account for this phenomenon. Auto-ignitions precede knock‐ ing events, however not every auto-ignition results in engine knock. Therefore, besides predicting when the unburnt mixture auto-ignites, the knock model subsequently needs to evaluate the occurring auto-ignition. Having extensively validated the prediction of the auto-ignition onset in previous publications, within this publication, the authors focus on the knock criterion to evaluate the auto-ignition. The introduced Hess criterion calculates a dimensionless parameter Π at the time of the occurring auto-ignition and second parameter I k,SOC considering the state of the pre-reactions within the unburnt mixture at the start of combustion. The di‐ mensionsless parameter Π considers different values influencing the amplitude of pressure oscillations due to an auto-ignition. By evaluating these two parameters, the knock criterion specifies whether the knock boundary of an operating point is reached or not. In addition to a detailed explanation of the Hess criterion and the validation of its predicted knock boundary, this publication provides insights into its practical application. For the validation of the knock criterion, many different operating points that include not only changing operating conditions, but also different engine configurations as well as different fuels are investigated. Despite the extensively varied operating conditions, such as different injection strategies, engine speeds and loads, intake air temperatures, cooling temperatures or compression ratios, the criterion only has to be calibrated once per engine. Using this knock criterion, the knock model is able to predict the knock boundary within engine simulations very accurately. After calibrating the knock model to <?page no="378"?> an engine, the center of combustion of an operating point at the knock boundary can be predicted with small standard deviations down to 1.25 °CA. Consequently, the introduced knock criterion helps to improve the development of SI engines in 0D/ 1D simulation. 1 Introduction To meet increasingly stringent emission standards and reduce fuel consumption, the efficiency of internal combustion engines is steadily improved. Higher compression ratios, downsizing and downspeeding as well as shifting the load point of the spark ignition (SI) engine through hybridization contribute to increase its efficiency. Unfortu‐ nately, all of these concepts also increase the knocking tendency of the SI engine. Since such knocking phenomena can damage the engine and even lead to its failure, knocking restricts the operation of SI engines by limiting the spark advance [1, 2]. Cost-efficient 0D/ 1D engine simulations are widely used in engine development as they provide a good trade-off between low computation time and high prediction quality [3, 4]. For these reasons, a 0D/ 1D knock model that predicts the knock boundary very accurately is crucial for engine development. Such a knock model predicts the auto-ignition onset of the unburnt mixture and evaluates the auto-ignition to predict the knock boundary of an operating point in 0D/ 1D engine simulations. This evaluation of the auto-ignition is performed by the knock criterion, which makes it high accuracy very important. In this publication, the authors present their recently developed and improved Hess criterion. After a detailed explanation of the Hess criterion, the authors show its validation at a broad variation of operating conditions - inter alia four different gasoline fuels. Subsequently, the accuracy of the knock model using the Hess criterion is compared to the knock model using different available knock criteria. Finally, some insights into the practical application of the knock model are given, showing the high accuracy and power of the Hess criterion. 2 0D/ 1D Knock Model In general, the input of the 0D/ 1D knock model is determined using pressure trace analysis or engine simulation. For higher accuracy, a distinction is generally made between burnt and unburnt zone. The model input includes the trace of the pressure and the temperature of the unburnt zone as well as stationary boundary conditions such as the air-fuel equivalence ratio, the exhaust gas recirculation and fuel properties. At first, in order to be able to use the Hess criterion (or most other knock criteria) the auto-ignition of the unburnt mixture must be modeled separately. For this purpose, there are various auto-ignition models available. The fundamentals of the auto-ignition model used in this publication were developed and explained in detail by Fandakov [5-7]. Since the authors have already extensively validated and explained this accurate auto-ignition model in [8, 9], only the fundamentals of this 378 Marco Hess, Michael Grill, Michael Barende <?page no="379"?> model are explained in the following. The used auto-ignition model accounts for the possible two-stage auto-ignition of gasoline fuels by calculating two Livengood-Wu integrals I k1 and I k2 [10]. Each integral is calculated via the ignition delay times of the respective stage. The first integral represents the first stage of the auto-ignition with its associated low-temperature heat release. Following this low-temperature heat release, the calculation of the second integral begins, which finally calculates the actual auto-ignition of the unburnt mixture. Following the modeled onset of the auto-ignition, the auto-ignition is evaluated using a knock criterion. For this, the knock criterion defines the knock boundary of the investigated operating point. Various knock criteria have already been developed using different approaches. One of these approaches assumes that knock cannot occur after a constant value of mass fraction burnt (MFB) [11-19]. This assumption implies that an auto-ignition after this point would not lead to engine knock, because the remaining unburnt mass fraction was too low. A more recent, commonly used knock criterion was developed by Fandakov [6, 7, 20]. This criterion takes the thermal boundary layer that developes at the cooler cylinder walls into account. It assumes that knocking events occur in late stages of the combustion, so that at the beginning of the auto-ignition a significant fraction of the remaining unburnt mass is within the cooler thermal boundary layer influencing the knocking behavior. The following chapter extensively explaines the Hess criterion, recently developed by the authors [9, 21], before its accuracy is compared with both of the mentioned knock criteria. 3 Hess Criterion The Hess criterion was developed by the authors in [9, 21] and presented as “Pi criterion”. This knock criterion evaluates the auto-ignition by evaluating the physical state at the time of auto-ignition. This offers the advantage that the knock criterion can be linked to any auto-ignition model in order to subsequently evaluate the auto-ignition. For the development of a new knock criterion, the authors take up theoretical considerations of Kleinschmidt [2]. Kleinschmidt formulated different parameters influencing the amplitude of pressure oscillations due to auto-ignition and derived two dimensionless parameters from them - the so-called “auto-ignition mach number” and the “dimensionless temperature gradient”. Ultimately, according to Kleinschmidt, the amplitude of the pressure oscillations is a function of these two dimensionless parameters multiplied by the cylinder pressure at the time of auto-ignition. Since the amplitude of the pressure oscillation of an auto-ignition is used for classification into “knocking” and “non-knocking”, it seems physically most reasonable to the authors to develop a knock criterion based on the influencing variables of this amplitude. Although in their original form [2] neither of these two parameters is suitable as a knock criterion of a 0D/ 1D knock model, the authors were nevertheless able to pick 379 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="380"?> (1) (2) up the physical correlations in order to develop a suitable 0D/ 1D knock criterion after several steps [9, 21]. The developed Hess criterion only considers the correlations of the variables of the parameter “auto-ignition mach number” (equation (1)). Furthermore, this equation is significantly changed by not taking into account the fuel-describing parameters A, β and M B . Since the resulting term ex p −1/ T 1 is approximately 1 due to the high temperatures within a working cycle, it can be neglected. Multiplied by the cylinder pressure, this finally gives equation (2). Besides the variables taken into account, also the unit of the resulting parameter Π 0 differs significantly from Kleinschmidt’s “auto-ignition mach number”. Furthermore, the variables in Kleinschmidt’s parameter (equation (1)) represent the physical and chemical state within an endgas pocket in the unburnt mixture. Since a 0D/ 1D knock model does not have any information about such endgas pockets, the variables in equation (2) correspond to the physical state of the entire unburnt mixture at the respective time. Π SZ = Aρ 1 ex p −β / T 1 V 11/ 3 M B a s1 Π 0 = p ub ρ ub V ub 1/ 3 κ ub R ub T ub However, further steps are required for practical application in a 0D/ 1D knock model. The necessity of these steps is clearly shown in Figure 1 and Figure 2. The values shown in these figures are determined using pressure trace analysis of the available measurement data and each dot represents a different operating point at the knock boundary (here: 4-10% knock frequency). The investigated data contain 116 operating points, covering the following wide-ranging variations of operating conditions affecting the knocking behavior of the engine: • four gasoline fuels - RON95E10 with RON=96.5, MON=85.2 - RON98 with RON=98.2, MON=88.4 - RON92M20 with RON=100.2, MON=87.1 - RON92E20 with RON=100.7, MON=87.6 • two compression ratios: 10.76 and 11.8 • various injection strategies (central and lateral direct injection, split injection, port fuel injection) with varying starts of injection • varying charge motion port • six engine speeds: 1500, 2000, 2500, 3000, 3500 and 4000 1/ min • three engine loads: about 12, 16 and 20 bar IMEP • three coolant and oil temperature: 75, 90 and 105 °C • three intake air temperatures: 25, 35 and 45 °C 380 Marco Hess, Michael Grill, Michael Barende <?page no="381"?> From the upper diagram in Figure 1, which shows the value of the parameter Π 0 at the time of auto-ignition (Π 0,AI ), two findings become clear. First, it can be clearly seen that the value of the parameter Π 0,AI depends on the engine load, represented by the indicated mean effective pressure (IMEP), of the respective operating point. In addition, it can be seen that the values of the parameter Π 0,AI have a relatively large scatter range even at the same engine load. Fig. 1: Π0 at auto-ignition onset (AI) and at start of compression (start) The lower diagram in Figure 1 shows the values of the parameter Π 0 at the start of compression (Π 0,start ). Analogous to the upper diagram, it can again be seen that the values of Π 0 depend on the engine load of the operating point. In addition to the different cylinder pressure p ub , this load dependence is mainly due to the different density ρ ub of the quantitatively controlled SI engine. On the other hand, compared to the upper diagram, it can be seen that the scatter range of the parameter Π 0,start is significantly smaller than the scatter range of the parameter Π 0,AI at same engine loads. Thus, it can be concluded that the different scatter range is not directly caused by stationary boundary conditions 381 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="382"?> (3) (such as engine speed, compression ratio, fuel, EGR, air-fuel ratio, intake air temperature, …), otherwise the scatter range of both parameters would be the same. For a practical application of Π 0 as a knock criterion, especially the large scatter range of the parameter Π 0 at the same x-value is a hindrance, and the load dependence is also unwanted. For this reason, the parameter Π 0,AI is first divided by Π 0,start , resulting in the dimensionless parameter Π: Π = Π 0, AI Π 0, start This dimensionless parameter Π is shown in Figure 2 for the different operating points and their engine loads. As can be clearly seen, the influence of the engine load could be significantly reduced thanks to this division, thus solving the first problem. However, as expected, the problem of the large scatter range of the parameter Π at the same x-value remains. In the final step, it is therefore important to find a variable for the x-values for which the Π values scatter significantly less at the same x-value. Only this important step of a good correlation enables the application as a 0D/ 1D knock criterion. 12 13 14 15 16 17 18 19 20 21 Indicated mean effective pressure [bar] 20 30 40 50 60 70 80 [-] Fig. 2: Dimensionless parameter Π As already mentioned, the wide scatter range of Π is not directly caused by stationary boundary conditions (such as engine speed, compression ratio, fuel, EGR, air-fuel ratio, intake air temperature, …). Nevertheless, these boundary conditions obviously have a major influence on the knock behavior of an engine. Among other things, their influence can be seen in the state of the pre-reactions of the unburnt mixture at the start of combustion. Since these pre-reactions in the unburnt mixture lead to its auto-ignition, their state at the start of combustion is an significant factor influencing the knock behavior, as the authors have shown in detail in [21, 22]. The “state of the pre-reactions at the start of combustion” I k,SOC can be quantitatively determined from 382 Marco Hess, Michael Grill, Michael Barende <?page no="383"?> the value of the Livengood-Wu integral I k . Only this consideration of the pre-reaction state at the start of combustion enables the knock criterion to be used in engine simulations. The correlation between the parameter Π and the state of the pre-reactions at the start of combustion I k,SOC is shown in Figure 3 for the different operating points with 4-10% knock frequency. As can be clearly seen, by taking I k,SOC into account the scatter range of Π shown in Figure 2 now can be accurately described by a regression curve. Such a precise regression curve is immensely important, since it ultimately forms the knock criterion. This regression curve (curve in Figure 3) now models the knock boundary of all operating points and can be used as a knock criterion in the engine simulation. If a simulated operating point is below this curve of the knock criterion, this means that the knock boundary of this operating point has not yet been reached according to the Hess criterion and the spark timing can be advanced. As soon as, due to the advanced spark timing, the operating point is above the curve, its knock boundary is exceeded according to the Hess criterion - which in this case means that its knock frequency is higher than 4-10%. Thus, in simple terms, the parameter Π as a function of I k,SOC can be interpreted as a measure of the knock frequency of the operating point. 0 0.5 1 1.5 2 2.5 3 I k,SOC [-] 20 30 40 50 60 70 80 [-] Fig. 3: Hess criterion As a possible equation of the regression curve shown in Figure 3, either a linear equation or the natural logarithmic function shown (equation (4) and (5)) can be used. The shape of the regression curve results from the calibration of the criterion, which is described in more detail in the following chapter. If both mentioned shapes of the regression curve describe the values of the investigated operating points with approximately equally high accuracy, the linear equation is recommended due to its higher robustness. This equation is also recommended if only a few operating points are available for the calibration of the criterion. 383 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="384"?> (4) (5) Π KB = m • I k, SOC + c Π KB = m • ln I k, SOC + c It is necessary to mention that in this publication the definition of “state of the pre-reactions at the start of combustion” has been changed in comparison with previous publications of the authors [9, 21, 22]. Due to the new definition of the “state of the pre-reactions at the start of combustion” I k,SOC , the cubic equation mentioned in [9, 21, 22] is no longer considered as a possible shape of the regression curve of the knock criterion. Although in [9, 21, 22] the authors defined the value of the “state of the pre-reactions at the start of combustion” as the sum of the two Livengood-Wu integrals at the time of 2% mass fraction burnt, this definition could be improved. From now on, the “state of the pre-reactions at the start of combustion” I k,SOC is no longer defined as the sum of the two Livengood-Wu integrals, but as the value of the high-temperature Livengood-Wu integral I k2* . At this point, it must be mentioned again that the auto-ignition onset is still calcu‐ lated by the used two-stage auto-ignition model. For this purpose, the low-temperature Livengood-Wu integral I k1 is calculated from the start of compression and only after the low-temperature ignition has taken place (I k1 =1), the calculation of the second integral I k2 begins with the starting value I k2 =0.3 [5-8]. Independent of this modeling of the auto-ignition onset (and without any effects on the auto-ignition modeling), the high-temperature Livengood-Wu integral I k2* is calculated in the knock criterion from the start of compression. This integral differs from I k2 only by the start of its calculation and its starting value (I k2* =0) and is therefore referred to as I k2* to avoid confusion. For clarification, Figure 4 shows the different integrals schematically to model the two-stage auto-ignition of a gasoline fuel. Due to the different definitions of I k1 , I k2 and I k2* , high values of I k,SOC can occur in some cases, as shown in Figure 3, but these should not be misinterpreted. For the two-stage auto-igniting gasoline fuel, such high values do not mean that the auto-ignition has already occurred at the start of combustion. The auto-ignition still takes place as soon as the integral I k2 calculated by the auto-ignition model reaches the value 1, regardless of the value of the integral I k2* . If the investigated fuel shows a single-stage auto-ignition behavior, such as hydrogen or methane, the auto-ignition is modelled solely by a single Livengood-Wu integral. In this case, the integral I k2* used by the knock criterion (to quantify the “state of the pre-reactions at the start of combustion”) corresponds to the integral used to model the auto-ignition. Moreover, in this new improved definition of the “state of the pre-reactions at the start of combustion”, the integral I k2* is evaluated at 10% mass fraction burnt, in contrast to the evaluation at 2% mass fraction burnt given in [9, 21, 22]. This additional change helps to increase the robustness of the developed knock criterion. 384 Marco Hess, Michael Grill, Michael Barende <?page no="385"?> Fig. 4: Calculation of Ik2* separated from the auto-ignition model for a two-stage auto-igniting gasoline fuel 4 Validation of the Hess Criterion 4.1 Application of the Knock Criterion To validate the accuracy of a knock criterion, the center of combustion (MFB50) at the knock boundary predicted by the knock model in performed 0D/ 1D engine simulations is compared with the MFB50 at the actual knock boundary of the investigated measurement data. Therefore, at first, it is important to define the term “knock boundary”, as there are various definitions. In this publication, the knock boundary of an operating point is defined as its center of combustion at which 4 to 10% of the working cycles knock. To classify whether a working cycle is a knocking one or not, the knock peak-to-peak (KPP) value (calculated at the test bench using a real-time capable weighted moving average filter) is used. A working cycle is defined as knocking if its KPP value is above the respective KPP threshold , which is equal to the current engine speed divided by 1000 and has the unit bar [6-8]. The measurement data used for the validation were obtained in the wake of experimental investigations performed on a single-cylinder research engine with a displacement of 399 cm 3 , a stroke of 90.5 mm and a bore of 75 mm [6-9, 21]. Two different pistons enable two different compression ratios. It features external boosting, different injection strategies and a tumble generation device. Furthermore, for the purpose of realistic boundary conditions of the measurement campaign, the engine is equipped with an exhaust throttle to simulate the backpressure of the turbine. In boosted operation, the average exhaust backpressure is controlled to match the intake pressure. For validation, the measurement data are first investigated to determine the MFB50 at the actual knock boundary by performing a two-zone pressure trace analysis that 385 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="386"?> considers the burnt and unburnt zone. The measured air and fuel mass flows are used to estimate the cylinder mass at the beginning of the high-pressure phase. The internal exhaust gas mass fraction completes the cylinder mass since no external exhaust gas recirculation is investigated in the available measurement data. In this work, the internal exhaust gas mass fraction is estimated using a three-pressure analysis. To reduce errors, the cylinder mass at the start of the calculation is iteratively adjusted (100% iteration) to correspond to the measured pressure profile [21]. Subsequently, 0D/ 1D engine simulations of these different operating points are performed with an appropriate two-zone SI combustion modeling approach. In this work, the Entrainment model [23] is used to gain the needed input for the 0D/ 1D knock model. Of course, all simulation models were calibrated to reproduce the actual engine behavior as accurately as possible, resulting in precise traces of temperature, pressure and heat release rate. Also the mass flows in the engine simulations are the same mass flows as in the pressure trace analysis after the 100% iteration to match the boundary conditions of the high-pressure phase as precisely as possible. Before using and validating the Hess criterion, it needs to be calibrated. However, before calibrating the knock criterion, the preceding auto-ignition model should be calibrated. Since the auto-ignition onset is modeled by calculating ignition delay times that depend exponentially on the temperature, even small changes in temperature have large effects on the modeled auto-ignition onset. Therefore, the used auto-ignition model is calibrate by slightly adjusting the input temperature. This slight adjustment does not affect any models other than the auto-ignition model and can be considered as the temperature of a hotspot in the unburnt zone [6, 7]. There are two different methods to calibrate the auto-ignition model. After performing engine simulations of different operating points at the measured knock boundary, one of the two calibration methods can be carried out. In the fastest method, the input temperature of the auto-ignition model is adjusted so that the modeled auto-ignition of these operating points occurs at realistic values of mass fraction burnt. The more time-consuming but much more accurate method is described in [8, 22]. In this method, the modeled auto-ignition onset is directly compared with the actual auto-ignition onset of the measurement data. This method provides the most accurate boundary conditions at the time of auto-ignition and thus enables the most valid validation of the knock criterion. Therefore, this method of accurate calibration of the auto-ignition model is chosen for the validation of the the Hess criterion. Finally, the Hess criterion can be calibrated. For this, engine simulations of different operating points at the measured knock boundary need to be performed with the previously calibrated auto-ignition model. Subsequently, the values of Π and I k,SOC of these simulated operating points are needed to calibrate the Hess criterion. Only two different operating points are needed for this calibration. To ensure a good accuracy of the calibration, the operating points should be chosen so that they have significantly different values of I k,SOC . Such different values often occur at operating points with clearly different auto-ignition onsets regarding the mass fraction burnt (e.g. two 386 Marco Hess, Michael Grill, Michael Barende <?page no="387"?> operating points with clearly different intake air temperatures, engine speeds or EGR). However, obviously, the more operating points available for such a calibration, the better results can theoretically be achieved. As explained in the previous chapter, the Hess criterion can be calibrated as a linear equation or as a natural logarithmic function, depending on which regression curve fits best. Finally, the knock criterion is calibrated and can be used to predict the knock boundary in engine simulations. 4.2 Calibration and Validation To validate the Hess criterion, all operating points with available measurement data are investigated. The provided data contain 116 operating points at the knock boundary (4 to 10% knock frequency), covering wide-ranging variations of operating conditions (as stated in chapter 3) affecting the knocking behavior of the engine. As mentioned before, the auto-ignition model is calibrated using the time-consuming but accurate method described in detail in [8, 22]. This calibration results in the input temperature of the model (the temperature of the unburnt zone) being adjusted by the following cylinder pressure dependent factor: T ub = T ub • 1 + 0 . 085 • p cyl − 6 / 100 . This adjustment does not affect any models other than the auto-ignition model. The cylinder pressure has the unit bar and this adjustment is only made above 20 bar cylinder pressure. Therefore, this is a slight ad‐ justment of the unburnt temperature of maximum 4% or about 20 K, which is a realistic value of temperature fluctuations in the unburnt mixture [24, 25]. Subsequently, the Hess criterion is calibrated using all 116 operating points at the knock boundary, so that the following validation shows the potential accuracy of the Hess criterion. Here, as shown in Figure 5, the knock boundary is calculated by the Hess criterion according to the following linear equation: Π KB = 53 − 10 . 3 • I k, SOC . The linear equation is chosen, as it fits the data points better and has a higher robustness than a natural logarithmic function. The differences between the values in Figure 3 and Figure 5 are because the values of Figure 3 are gained via pressure trace analysis, whereas the values in Figure 5 are gained via engine simulations. Therefore, as described earlier, it is important to calibrate the knock criterion using the results of the engine simulations, since the knock criterion predicts the knock boundary in performed engine simulations. 387 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="388"?> 0 0.5 1 1.5 2 2.5 3 I k,SOC [-] 20 30 40 50 60 70 80 [-] Fig. 5: Calibration of the Hess criterion 0 10 20 30 40 50 MFB50 KB,Meas. [°CA a. FTDC] 0 10 20 30 40 50 MFB50 KB,Model [°CA a. FTDC] RMSE=1.25 °CA Fig. 6: Validation of the Hess criterion Finally, Figure 6 shows the results of the validation of the Hess criterion. While the x-values correspond to the MFB50 at the actual knock boundary of the measurement data, the y-values represent the MFB50 at the knock boundary predicted by the knock model. This figure contains all 116 operating points with 4 to 10% knock frequency, covering the wide-ranging variations of operating conditions mentioned in chapter 3. 388 Marco Hess, Michael Grill, Michael Barende <?page no="389"?> It is clearly shown, that the knock model using the Hess criterion is able to predict the knock boundary very accurately for different variations of operating conditions. Since the accuracy of the predicted knock boundary does not depend on any of the investigated operating conditions, no specific operating condition is highlighted in color. The different operating points are scattered around the angle bisector with about the same accuracy with a very small root-mean-square error of 1.25 °CA. Having shown that the knock model, using the Hess criterion to predict the knock boundary, predicts the knock boundary very accurately, it is important to determine the actual contribution of the Hess criterion itself to this high accuracy. For this, the shown results are compared to the knock boundary predicted by the knock model using different knock criteria. 4.3 Comparison to other Knock Criteria To compare the accuracy of the Hess criterion to different available, commonly used knock criteria, the explained 0D/ 1D engine simulations are performed with identical settings, while the respective used knock criterion predicts the knock boundary of the investigated operating point. In addition, the applied auto-ignition model is of course calibrated identically, so that the performed simulations differ only in the used knock criterion. Thus, in this comparison, all boundary conditions are identical and the reason for the different accurately predicted knock boundaries is solely due to the applied knock criterion. This allows a direct comparison of the accuracy of the different knock criteria. The two knock criteria to compare the Hess criterion with are explained in chapter 2. In comparison to the validation of the Hess criterion (Figure 6), both other knock criteria are iteratively calibrated to achieve best results to show their potential accuracy. Resulting in the calibration parameter of the Fandakov-criterion [6, 7, 20] being calibrated to x ub,bl =0.13, while the “constant MFB criterion” is calibrated to auto-ignitions at 94% mass fraction burnt. The results of the validation of these two knock criteria are shown in Figure 7 that contains the same 116 operating points as the validation of the Hess criterion in Figure 6. Figure 7 clearly shows that the Hess criterion is able to predict the knock boundary significantly better than both of the other two, commonly used knock criteria. This can be easily seen as the knock boundaries predicted by the Hess criterion are equally, closely scattered around the actual knock boundaries of the measurement data (angle bisector). In comparison, the validation of the other two knock criteria shows a clear trend in the predicted knock boundaries, as both of these criteria predict too early MFB50 KB at early MFB50 KB and too late MFB50 KB at late MFB50 KB . Quantitatively, both of these knock criteria predict the knock boundary with a significantly higher root-mean-square error of 5.20 °CA and 5.91 °CA, respectively. This means, the Hess criterion is able to predict the knock boundary of the investigated operating points about 4 to 5 times more accurately than the commonly used Fandakov-criterion and the “constant MFB criterion”. Finally, this shows how significant the actual contribution of the Hess criterion itself is to the high accuracy of the knock model shown in Figure 6. 389 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="390"?> Fig. 7: Validation of Fandakov criterion (left) and knock criterion with constant MFB at auto-ignition (right) The investigated 116 operating points at the knock boundary cover extensively varied operating conditions mentioned in chapter 3. As the authors showed in [21, 22], this huge variation of operating conditions results, inter alia, in auto-ignitions at different values of mass fraction burnt covering about 60% MFB to 95% MFB. Therefore, obviously, the “constant MFB criterion” is not suitable for such a wide-ranging variation of operating conditions. The Fandakov-criterion is also not suitable for such broadly varied operating conditions, as such early auto-ignitions exceed the limits of the physical plausibility of Fandakov’s criterion [21]. In contrast, the Hess criterion is very well suited even for such extensively varied operating conditions. This is because the Hess criterion was developed based on physical variables influencing the pressure oscillations due to an auto-ignition (Π) as well as the pre-reactions of the unburnt mixture at combustion start (I k,SOC ) that have a significant influence on the knocking behavior. This profound physical basis enables the Hess criterion to predict the knock boundary very accurately even for extensively varied operating conditions. 4.4 Further Validation at Different Application Cases In addition to the shown validation of the potential accuracy of the Hess criterion, the criterion is validated at different application cases. At first, the Hess criterion is intentionally miscalibrated to imitate a possible application, when not many or even no measurement data are available to calibrate the knock criterion. Even though only two operating points are needed to calibrate the Hess criterion, obviously, the more operating points available for such a calibration, the better the knock boundary can theoretically be predicted. If only few measurement data are available, it could happen that the knock criterion is not perfectly calibrated, due to the small amount of calibration points. Or even worse, if there are no measurement data of the investigated engine available, the user obviously cannot calibrate the criterion based on measurement data and needs to utilize experience values. This is a valid fact for all available knock criteria and not specific for the Hess criterion. 390 Marco Hess, Michael Grill, Michael Barende <?page no="391"?> To validate the accuracy of the Hess criterion in such application case, the same 116 operating points (4 to 10% knock frequency) as in Figure 6 are investigated, where the Hess criterion is correctly calibrated. Moreover, the explained 0D/ 1D engine simu‐ lations are performed with identical settings and an identically calibrated auto-ignition model so that the performed simulations differ only in the calibration of the knock criterion. This allows a direct investigation of the influence of a possible miscalibration of the criterion due to missing measurement data. For this, Figure 8presents the knock boundary predicted by the intentionally miscalibrated Hess criterion. This miscalibration represents a clearly bad case of calibrating the knock criterion, as shown in the upper diagram of Figure 9. Even though this clear miscalibration of the Hess criterion, the criterion still predicts the knock boundary relatively good with a mean deviation of 2 °CA. Of course, compared to the accuracy of the correctly calibrated Hess criterion (Figure 6), the error of the knock boundary predicted by the intentionally miscalibrated Hess criterion is 60% higher. However, compared to the other available, commonly used knock criteria that were iteratively calibrated to achieve their highest possible accuracy (Figure 7), the knock boundary is still predicted 2.5 to almost 3 times more accurately. Not to mention that in this imitated application case of few or even no available measurement data, both other criteria could not have been calibrated iteratively to achieve their best results and the accuracy advantage of the Hess criterion would again be much higher than 2.5 times. 0 10 20 30 40 50 MFB50 KB,Meas. [°CA a. FTDC] 0 10 20 30 40 50 MFB50 KB,Model [°CA a. FTDC] RMSE=2.00 °CA Fig. 8: Predicted knock boundary by intentionally miscalibrated Hess criterion 391 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="392"?> This means that the accuracy of the Hess criterion is so high, that even intentionally clearly miscalibrated, the results are still much better compared to the other commonly used criteria that were iteratively calibrated to achieve best results. Thus, the Hess criterion is very well suited to predict the knock boundary even if only few or no measurement data are available for its calibration. However, since the shown case imitates an intentionally clearly bad calibration of the criterion, such constant value of Π should of course not be used as calibration when no measurement data are available, but a linear equation with a slope. Furthermore, different companies or research institutes do not have a uniform definition of the “knock boundary”. Nevertheless, it is important that a knock criterion is able to predict the knock boundary accurately, independent on its exact definition. For this validation, in the following, the knock boundary is not defined as 4 to 10% knock frequency anymore, but as exactly 2% and 4% knock frequency, respectively. Of course, also higher knock frequencies would be possible, but since more operating points with lower knock frequencies are available for the authors, these knock frequencies allow a better validation. Besides, the knock boundary is usually defined as rather low than high knock frequencies. To apply these different definitions of the knock boundary to the engine simulations, the Hess criterion solely needs to be calibrated at operating points with the respective knock frequency. The calibration of the Hess criterion is presented in the middle and lower diagram of Figure 9. Compared to the knock boundary at 4 to 10% knock frequency (upper diagram), there are less operating points available for the calibration, however still enoguh, as there are more than two operating points at the respective knock boundaries. The accuracy of the Hess criterion for different definitions of the knock boundary is shown in Figure 10. Analogous to the previous figures, both diagrams compare the center of combustion at the knock boundary predicted by the knock model to the actual knock boundary of the measurement data according to its respective definitions (2% and 4% knock frequency). Even though less contained operating points compared to the previous figures, both diagrams still cover many of the mentioned varied operating conditions. Again, after the respective calibration, the Hess criterion very accurately predicts the knock boundaries with low root-mean-square errors and predicted knock boundaries well scattered around the actual measured knock boundaries (angle bisector). This validation is very important as it allows the Hess criterion to be broadly applied, independent on the exact definition of the knock boundary, which is not uniformly defined by different companies or research institutes. Rather, it is important that all operating points used for the calibration represent the same knock boundary at approximately the same knock frequency. 392 Marco Hess, Michael Grill, Michael Barende <?page no="393"?> Fig. 9: Hess criterion intentionally miscalibrated (upper). Hess criterion calibrated to 2% knock frequency (middle) and 4% knock frequency (lower) 393 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="394"?> Fig. 10: Validation of the Hess criterion calibrated to different definitions of the “knock boundary” at 2% (left) and 4% knock frequency (right) 5 Summary and Conclusions This publication extensively explains the Hess criterion and its physical background. This knock criterion evaluates the occurring auto-ignition of the unburnt mixture and thus can be applied subsequently to any separate auto-ignition model. For a profound evaluation, the Hess criterion was developed based on physical variables influencing the pressure oscillations due to an auto-ignition (Π) as well as the pre-reactions of the unburnt mixture at combustion start (I k,SOC ) that have a significant influence on the knocking behavior. Afterwards, the practical application of the criterion and insights into its calibration are described in detail. After its calibration, the criterion defines threshold values of Π that depend on I k,SOC . If the value of Π of an operating point is above its respective threshold value, the knock boundary is reached or exceeded. The accuracy of the Hess criterion is validated against 116 different operating points at the knock boundary (4 to 10% knock frequency) covering wide-ranging variations of operating conditions. It is shown that the knock model, using the Hess criterion, is able to predict the knock boundary very accurately with a low root-mean-sqaure error of 1.25 °CA. Moreover, the predicted knock boundaries for these operating points are almost perfectly scattered around the actual knock boundaries of the measurement data. To determine the actual contribution of the Hess criterion itself to this high accuracy, these results are compared to results of the knock model using other knock criteria to predict the knock boundary. In comparison to these other two available, commonly used knock criteria, the Hess criterion predicts the knock boundary about 4 to 5 times more accurately, showing the huge contribution of the knock criterion to the prediction of the knock boundaries. Thus, the Hess criterion is a significant 394 Marco Hess, Michael Grill, Michael Barende <?page no="395"?> improvement compared to the other commonly used knock criteria. This high accuracy is because of its profound physical basis that enables the Hess criterion to cover and account for the infuences of the investigated broad variations of operating conditions on engine knock. Moreover, further validations at different application cases are performed. At first, the Hess criterion is intentionally miscalibrated to imitate a possible application, when not many or even no measurement data are available to calibrate the knock criterion. Though intentionaly very badly calibrated, the Hess criterion still predicts the knock boundary 2.5 to 3 times more accurately than the other two commonly used knock criteria even though they are perfectly calibrated. This again shows the high accuracy and power of the Hess criterion. Furthermore, as different companies or research institutes do not have a uniform definition of the “knock boundary”, it is important that a knock criterion is able to predict the knock boundary accurately, independent on its exact definition. After being calibrated, the Hess criterion again predicts the knock boundaries very accurately for different definitions of the “knock boundary” (at 2% and 4% knock frequency, respectively). This validation proofs that the Hess criterion can be broadly applied, independent on the exact definition of the knock boundary. Thus, due to its profound physical basis, the Hess criterion enables the 0D/ 1D knock model to predict the knock boundary for extensively varied operating conditions very accurately with deviations down to 1.25 °CA. Moreover, as the investigated operating conditions cover inter alia different engine configurations and four different gasoline fuels, the Hess criterion helps to improve the development of SI engines in 0D/ 1D simulation for future applications. References [1] Heywood, J.B., Internal Combustion Engine Fundamentals, (McGraw-Hill, Inc., 1988), ISBN: 0-07-028637-X. [2] Kleinschmidt, W., “Selbstzündung im Klopfgrenzbereich von Serienmotoren”. In: Klopfrege‐ lung für Ottomotoren II, 2006, 1-21. ISBN: 3816926746. [3] Schmid, A., Grill, M., Berner, H.J., and Bargende, M., “Transient Simulation with Scaveng‐ ing in the Turbo Spark-Ignition Engine,” MTZ Worldwide 71(11): 10-15, 2010, doi: 10.1007/ BF03227995. [4] Fandakov, A., Grill, M., Bargende, M., and Casal Kulzer, A., “Investigation of Thermodynamic and Chemical Influences on Knock for the Working Process Calculation,” in 17 th Stuttgart International Symposium (Springer Vieweg, 2017), doi: 10.1007/ 978-3-658-16988-6_13. [5] Fandakov, A., Grill, M., Bargende, M., and Casal Kulzer, A., “Two-Stage Ignition Occurrence in the End Gas and Modeling Its Influence on Engine Knock,” SAE International Journal of Engines 10(4): 2109-2128, 2017, doi: 10.4271/ 2017-24-0001. [6] Cai, L., Fandakov, A., Mally, M., Ramalingam, A. et al.,“Knock with EGR at Full Load,” Final Report on FVV Project 6301, H1144, Research Association for Combustion Engines e. V. (FVV), Frankfurt am Main, 2017. 395 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="396"?> [7] Fandakov, A., “A Phenomenological Knock Model for the Development of Future Engine Con‐ cepts,” Wissenschaftliche Reihe Fahrzeugtechnik Universität Stuttgart, (Wiesbaden, Springer Vieweg, 2019), doi: 10.1007/ 978-3-658-24875-8. [8] Hess, M., Grill, M., Bargende, M., and Casal Kulzer, A., “Knock Model Covering Thermody‐ namic and Chemical Influences on the Two-Stage Auto-Ignition of Gasoline Fuels,” SAE Technical Paper 2021-01-0381, 2021, doi: 10.4271/ 2021-01-0381. [9] Hess, M., Grill, M., Bargende, M., and Casal Kulzer, A., “Two-Stage 0D/ 1D Knock Model to Predict the Knock Boundary of SI Engines,” in 21 st Stuttgart International Symposium (Springer Vieweg, 2021), doi: 10.1007/ 978-3-658-33466-6_37. [10] Livengood, J.C. and Wu, P.C., “Correlation of Autoignition Phenomena in Internal Com‐ bustion Engines and Rapid Compression Machines,” Symp. Int. Combustion 5: 347-356, 1955, doi: 10.1016/ S0082-0784(55)80047-1. [11] Chen, L., Li, T., Yin, T., and Zheng, B., “A Predictive Model for Knock Onset in Spark-Ig‐ nition Engines with Cooled EGR,” Energy Conversion and Management 87: 946-955, 2014, doi: 10.1016/ j.enconman.2014.08.002. [12] Douaud, A. and Eyzat, P., “Four-Octane-Number Method for Predicting the Anti-Knock Behavior of Fuels and Engines,” SAE Technical Paper 780080, 1978, doi: 10.4271/ 780080. [13] Worret, R., Bernhardt, S., Schwarz, F., and Spicher, U., “Application of Different Cylinder Pressure Based Knock Detection Methods in Spark Ignition Engines,” SAE Technical Paper 2002-01-1668, 2002, doi: 10.4271/ 2002-01-1668. [14] Hoepke, B., Jannsen, S., Kasseris, E., and Cheng, W., “EGR Effects on Boosted SI Engine Operation and Knock Integral Correlation,” SAE Int. J. Eng. 5(2): 547-559, 2012, doi: 10.4271/ 2012-01-0707. [15] Schmid, A., Grill, M., Berner, H.-J., and Bargende, M., “Ein neuer Ansatz zur Vorhersage des ottomotorischen Klopfens,” Ottomotorisches Klopfen-irreguläre Verbrennung, 2010, 256-277, ISBN: 3816930476. [16] Franzke, D.E., “Beitrag zur Ermittlung eines Klopfkriteriums der ottomotorischen Verbren‐ nung und zur Vorausberechnung der Klopfgrenze,” Ph.D. thesis, Technical University of Munich, 1981. [17] Burluka, A.A., Liu, K., Sheppard, C.G.W., Smallbone, A.J. et al., “The Influence of Simulated Residual and NO Concentrations on Knock Onset for PRFs and Gasolines,” SAE Technical Paper 2004-01-2998, 2004, doi: 10.4271/ 2004-01-2998. [18] Elmqvist, C., Lindström, F., Ångström, H., Grandin, B. et al., “Optimizing Engine Concepts by Using a Simple Model for Knock Prediction,” SAE Technical Paper 2003-01-3123, 2003, doi: 10.4271/ 2003-01-3123. [19] Wayne, W.S., Clark, N. N., and Atkinson, C.M., “Numerical Prediction of Knock in a Bi-Fuel Engine,” SAE Technical Paper 982533, 1998, doi: 10.4271/ 982533. [20] Fandakov, A., Grill, M., Bargende, M., and Casal Kulzer, A., “A Two-Stage Knock Model for the Development of Future SI Engine Concepts,” SAE Technical Paper 2018-01-0855, 2018, doi: 10.4271/ 2018-01-0855. 396 Marco Hess, Michael Grill, Michael Barende <?page no="397"?> [21] Hess, M., Grill, M., Bargende, M., and Casal Kulzer, A., “New Criteria for 0D/ 1D Knock Models to Predict the Knock Boundary for Different Gasoline Fuels,” SAE Technical Paper 2021-01-0377, 2021, doi: 10.4271/ 2021-01-0377. [22] Blomberg, M., Hess, M., Hesse, R., Morsch, P., “Engine Knock Model”, Final Report on FVV Project 1313, Research Association for Combustion Engines e. V. (FVV), Frankfurt am Main, 2021 [23] Grill, M., Billinger, T., and Bargende, M., “Quasi- Dimensional Modeling of Spark Ignition Engine Combustion with Variable Valve Train,” SAE Technical Paper 2006-01-1107, 2006, doi: 10.4271/ 2006-01-1107. [24] Schießl, R., Maas, U.: “Analysis of Endgas Temperature Inhomogeneities in an SI En‐ gine by Laser-Induced Fluorescence,” Combustion and Flame 133: 19-27, 2003, doi: 10.1016/ S0010-2180(02)00538-2. [25] Schießl, R., Schubert, A., Maas, U.: “Temperature Fluctuations in the Unburned Mixture: Indirect Visualisation Based on LIF and Numerical Simulations,” SAE Technical Paper 2006-01-3338, 2006, doi: 10.4271/ 2006-01-3338. Acknowledgments The presented investigations were performed at the Institute of Automotive Engineer‐ ing (IFS) of the University of Stuttgart based on the research project “Engine Knock Model” assigned by the Research Association for Combustion Engines eV (FVV). The project was financially supported by the Federal Ministry for Economic Affairs and Energy (BMWi) via the German Federation of Industrial Research Associations eV (AiF) (IGF-Nr. 19787 N). The authors would like to thank the working group that accompanied the research work and all the companies involved for their support, the colleagues from the Institute for Combustion Technology and the Physico-Chemical Fundamentals of Combustion of the RWTH Aachen University for the collaboration and especially the Institute for Combustion Engines of the RWTH Aachen University for providing the investigated measurement data, as well as the FVV, AiF and BMWi for granting the financing. 397 0D/ 1D Knock Criterion to Predict the Knock Boundary of SI Engines <?page no="399"?> Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function Pierpaolo Napolitano b , Irina Jimenez a , Benjamín Pla a , Carlo Beatrice b a b CMT-Motores Termicos, Universitat Politècnica de València, Camino de Vera s/ n, E-46022 Valencia, Spain. Consiglio Nazionale delle Ricerche, Piazzale Aldo Moro,7, Rome RM, Italy. Abstract: Increasing demands on higher performance and lower fuel consump‐ tion and emissions have lead the path for internal combustion engine develop‐ ment; this race is nowadays directly related of CO2 emissions reduction. In spark-ignited (SI) engines, knock is one of the major barriers to achieve high thermal efficiency at high loads. The knocking risk is even higher in heavy-duty (HD) engines due to the size of the cylinders and to the low rotation speed, with potential impact on engine durability, which in turns is a mandatory feature for their applications. This paper proposes a knock detection strategy based on the combination of knock sensors and combustion modeling applied to a HD natural gas (NG) engine. The aim is to have a reliable, economic and computationally efficient algorithm to be implemented directly on the engine ECU. The method proposed has been applied to an extensive set of experimental data acquired on a SI NG heavy-duty engine. The results of the proposed knock estimation method are benchmarkt with those based on in-cylinder pressure analysis using piezoelectric transducers. The extension of the method based on in-cylinder pressure to a high displacement heavy-duty NG engine not only rep‐ resents an innovation, but improves the knock recognition based on in-cylinder pressure compared with conventional methods as MAPO or IMAP. Besides, the development of an alternative method based on knock sensor signal, allows to obtain a higher or equal sensitivity compared to the traditional MAPO method based on in-cylinder pressure, with the advantage of only using knock sensors. 1 Introduction In recent years, diverse legislative measures have been implemented by international regulatory agencies in order to increase the development of alternative fuels in <?page no="400"?> transport systems [1]. One of the most widely investigated alternative fuels found in literature is Natural Gas (NG) [2], which is a gas mixture consisting primary of methane, with smaller percentages of other gases such as ethane, propane, and butane. Nowadays, considering the environmental sustainability targets imposed by several worldwide governments (e.g. fit for 55 in UE), instead to NG, renewable methane, like bio-methane or the so-called “e-methane” is rapidly growing their attention considering the great potential in CO2 footprint in the whole supply chain (Well-To-Wheel balance) [4]. The recent sensibility in the reduction of CO2 emissions from thermal system has pushed the research to the use of alternative low carbon content fuels, and they diffusion to any possible combustion system [3]. The sector of heavy duty (HD) engine for road and off-road application was not excluded from this phenomenon [5]. Since methane-based fuels, like NG or bio-methane, have a higher octane number than gasoline, it is possible to work with higher compression ratio in spark ignition (SI) engines [6][7]. Nevertheless, using a gas instead of a liquid fuel involves the displacement of some air by NG, then leading to a reduction in the engine power output in port fuel injection cases [8]. In order to overcome this problem, two solutions can be found in literature: on the one hand, increasing the compression ratio, or in the other hand using lean combustion [9][10]. However, the compression ratio increase is limited by knock phenomena in SI engines due to higher combustion pressures and temperatures [7][10], and as regards lean combustion, this has an operation limit, i.e over lean the mixture may lead to instability and misfire [11][12]. Knock is an abnormal combustion phenomenon in SI engines, related with the uncontrolled combustion of the end gas [13]. When knock occurs, a rapid combustion is observed due to the high local pressure, which produces shock waves that heavily excite the in-cylinder resonant modes [14][15]. The engine exposition to knock during several cycles may lead to piston rings braking, piston melting, engine efficiency decreases and engine damage in general reducing the durability [16]. By the way, in any HD engine application, energy efficiency and high durability are mandatory features for the competition on the market. On this way, knock recognition techniques are important in order to achieve high thermal efficiency and long engine life cycle. These methods can be mainly classified in two principal groups: direct and indirect knock recognition methods [16]. The first group is based on the in-cylinder pressure measurement, which is directly influenced by the phenomena [14][17][18]. The second group is based on indirect measurements such as cylinder block vibration [19][20]. Despite methods based on in-cylinder pressure measurement show higher reliability and accuracy, their application in mass production engines is limited by sensors durability and cost [21][22]. Classical knock recognition techniques are based on a fixed threshold, as the Maximum Amplitude of Pressure Oscillations (MAPO), which consist on comparing the absolute value of the band-pass filtered pressure signal with a predefined threshold [23]. Some authors developed knock metrics for knocking recognition [24][25], or 400 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="401"?> classification models as is shown in [26], where a machine learning algorithm is presented. Recently, knock recognition methods had been developed with the aim of being able to recognize knocking events from combustion without the need of a fixed threshold. For example, in [18] the band-pass in-cylinder pressure is compared in two windows locations: at the main combustion process location and at the end of combustion. The comparison between the signal amplitudes at both locations allows to identify low intensity knocking cycles. Additionally, a knocking threshold based on Mass Fraction Burned (MFB) evolution is presented in [14], where knocking cycles are differentiated from normal combustion using a resonance index which is compared with the expected resonance index produced by a constant volume combustion of the remaining fuel. Although these methods show good results and are able to recognize knock, even with low intensity, they are in-cylinder pressure based, which makes its application expensive in production engines [27]. Regarding knock recognition methods based on vibration signal, in recent years several indexes based on Fast Fourier Transform (FFT) [28], Empirical Mode Decom‐ position (EMD) [29] or Walvelet [19] analysis have been developed. The main problem with these approaches is the need to set a threshold in order to distinguish knock from normal combustion. The objective of the present work is to develop an improved understanding of the information contained in the knock sensor signal by analyzing two different sensor locations. The novelty of this work is to extend the method presented in [14] for a light-duty SI engine based on in-cylinder pressure signal, to a HD SI engine basing the method on knock sensor signals. This work is organized as follows, first the experimental set-up and tests performed are presented. Then, a frequency analysis of both knock sensor signal positions is performed and compared with in-cylinder pressure signal information. After, the knock recognition method is presented, where the MFB model and the knock recognition procedure are introduced. Then, results and discussions about the MFB model and the knock recognition are presented. The final section highlights the main contribution of the work and proposes future work for control applications. 2 Experimental set-up and test methodology Experimental tests for calibration, illustration and validation proposes were carried out in a HD NG SI engine. The engine was coupled with a variable frequency fast response dynamometer (AVL Dynodur), able to perform both steady state and dynamic tests. The engine was full instrumented and monitored, all low acquisition frequency measures were made by a National Instruments DAQ device. Instead, high sampling rate indicated signals were acquired and recorded by means of an AVL IndiSmart indicating system coupled with the AVL Indicom software. The main specifications of the engine are collected in Table 1. 401 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="402"?> Displaced volume 5883 cc Stroke 120 mm Bore 102 mm Compression ratio 10.3: 1 Number of cylinders 6 Valves per cylinder 2 Injection system Multy point. Port Fuel Injection (PFI) Fuel Methane Rated power 117kW Max torque 630Nm Table 1: Engine main specifications In order to analyze the information contained in the knock sensor signal and evaluate the knock recognition method, two cylinders were equipped with in-cylinder pressure sensors (cylinders 1 and 3) as is indicated in Figure 1. Two measuring spark plugs with miniature piezo-electric pressure transducer from Kistler, type 611xC were employed. In addition, signals from three knock sensors, piezo-quartz accelerometer with integrated discharge resistor (4.8 M Ω), with a nominal sensitivity of 30 mV/ g, max mechanical vibrations 70 g peak , were installed in different positions of the engine block. In particular, two of them in correspondence with the cylinders 1 and 3 location (Ks A and B) and a third located between both cylinders (Ks C). - --- A ------- C ------- B ----- --- 1 -------- 2 -------- 3 --------- 4 ------- 5 -------- 6 Figure 1: In-cylinder and knock sensors configuration. Knock sensors A and B correspond to a case where six sensors are required for knock recognition, i.e. one for each cylinder. On the other hand, knock sensor C represents the case where two knock sensors are necessary to evaluate knock in all six cylinders, one near cylinder 2 as is shown in Figure 1 and a second one near cylinder 5. 402 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="403"?> The sampling frequency for knock and pressure sensors was of 0.05 CAD. During the experiments, the engine was tested at different steady operating conditions by keeping the speed and load constant while the Spark Advance (SA) was modified. For each of the thirteen testing points, ranging between 1100 and 1800 rpm, and 200 and 600 Nm, the SA was progressively advanced from the reference value with 2 CAD steps until a maximum advance of 6 CAD. For the analysis of the data the instantaneous engine speed fluctuations during tests are negligible. 3 Frequency analysis of knock sensors signals In this section, the frequency content of signals from knock sensors A, B and C are compared. For this analysis, the content of the pressure sensor in the chamber of cylinder 1 will be compared with the knock sensors already mentioned. In order to estimate the resonance frequencies during knocking and no-knocking cycles, the approach presented by Draper [30] is used. In [30], the wave equation was solved with Bessel functions for a cylindrical geometry, showing that the characteristic frequency generated in the combustion chamber can be expressed as: (1) where the axial modes g are neglected near the TDC because the height is too low (h < D), D is the bore of the cylinder, a s the speed of sound and B (i,j) are the Bessel constants related with the radial modes, i and j represent the number of circumferential pressure modes and number of radial pressure modes. The speed of sound can be calculated by measuring the trapped mass m, the in-cylinder pressure p, and estimating the instantaneous volume of the chamber V. ( 2 ) where γ is the specific heat capacities ratio of the gases inside the cylinder, which can be approximated by dividing the gas mixture in three species, namely air, fuel, and burnt products, and modeled by polynomial expressions for the in-cylinder temperature such as suggested in [31]. The spectrogram of the in-cylinder pressure signal (P1) during a knocking cycle and knock sensors signals (A, B and C) are shown in Figure 2. In dashed white line, the resonance frequency computed from Equation (1) is represented for the first 3 radial modes. The operating conditions of such spectrograms is 1500 rpm of engine speed and 525 Nm of load. 403 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="404"?> Figure 2: Spectogram of in-cylinder pressure signal P1, and knock sensors A, B and C during a knocking cycle Analyzing Figure 2, the resonance components in the in-cylinders pressure signal are also present in the three knock sensor signals. For the cases where the knock sensor is located near the cylinder, knock sensor A for cylinder 1, the intensity of the resonance components is higher comparing with the others knock sensors (as can be seen in the color bar). It is noticeable that as the distance from cylinder 1 to the knock sensor increases, the intensity of the resonance decreases. Notice that the different peaks present in knock sensor A, B or C corresponds to the different resonance frequencies computed from Equation (1), as an example 3 resonance modes were represented. In order to assess several cycles, the correlation between the information contained in knock sensors signals and the in-cylinder pressure signals is analyzed by computing the coherence function between them. The coherence functions is the ratio between the cross power spectral density of the in-cylinder pressure (sub-index p) and knock sensor (sub-index k) signals P p,k (f ), to the product of the power spectral density of each signal P p,p (f ) and P k,k (f ). This relation is computed as [32]: ( 3 ) where f represents the different frequencies and the power spectral density of a signal x is defined as: where r xx is the auto correlation function, which is computed as a sliding inner product of the signal x with itself: ( 5 ) rxx[k] = x[n]x[n+k] 404 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="405"?> And the cross power spectral density between two signals x and y is defined as follows: where r xy is the cross-correlation function, which is computed as: ( 7 ) rxy[k] = x[n]y[n+k] The coherence function between the in-cylinder pressure and knock sensor signals (C p,k ) was evaluated at high load steady operating condition, 1300 rpm 570 Nm, over 200 cycles in Figure 3, where knocking and no-knocking cycles were part of the data-set. The black line represents the evolution of the coherence function for knock sensor A while purple and blue show the results for knock sensors B and C respectively. Three frequencies have been highlighted in red line, these frequencies correspond to the maximums of the first three resonance modes for this operation condition. Also notice that the x-axis scale it is logarithmic. Figure 3: Coherence function between in-cylinder pressure and knock sensor signals. Cylinder 1 (left) and Cylinder 3 (right). 405 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="406"?> As can be seen in Figure 3, the coherence with the in-cylinder pressure signal computed from cylinder 1 is lower for knock sensor B than for knock sensors A or C. The coherence is related with the distance between cylinder and knock sensor location. Notice that at the three frequencies highlighted in red, the coherence function reaches the maximum value. From these three comparisons, it can be seen that the mean coherence between the in-cylinder pressure and knock sensor is high, over 0.5, but in specific frequency bands, which are related with the in-cylinder resonance frequency modes computed from Equation (1). The analysis of high frequencies components of in-cylinder pressure signals and knock sensor signals was performed by computing an alternative of the Fourier transform as is described in [33]. Here, a resonance index is computed from in-cylinder pressure signal as following: ( 8 ) where α 1 and α 2 define the interval where the resonance analysis is performed, w is a window function of α 2 − α 1 length, p bp the band-pass filtered pressure, and T s (α) is the sampling period, which is constant only in time-based acquisition or if the instantaneous engine speed fluctuations are negligible, B i,j is the Bessel constant [30], D is the bore of the cylinder, V the combustion chamber volume, m the trapped mass, and p lp the low-pass in-cylinder pressure. Analogously to the Equation (8), a resonance index can be defined from the knock sensor signals, as: ( 9 ) where ks bp and ks lp are the band and low pass knock sensor signals respectively. In Figure 4, the resonance index is compared with the band-pass signal, filtered between 4.5 and 15 kHz. In black line the resonance index evolution is shown and in grey line the band pass signal. Three different signals are analyzed during the same cycle: on the top, in-cylinder pressure from cylinder 1, on the middle, the knock sensor A, and on the bottom, knock sensor C. 406 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="407"?> Figure 4: Resonance index evolution for cylinder 1 resonance evaluation: In-cylinder pressure (top), knock sensor A(middle) and C (bottom). Analyzing Figure 4 the resonance index from the three signals during this specific cycle evolves in a similar way. However, the in-cylinder pressure index maximum is located earlier than both knock indexes. When comparing knock sensor A with C index, the maximum corresponding to knock sensor A is located earlier than C, this could be due to the fact that the knock sensor A it is located closer to the cylinder 1. On the other hand, the maximum amplitude computed from knock sensor A is higher than knock sensor C. As it was shown in [13], the maximum amplitude and location of the resonance index computed from in-cylinder pressure signal it is highly sensitive to the SA. In cycles without knock, maximum amplitudes tend to be moderate and located in the surroundings of the center of combustion (MFB 50 ). On the other hand, in knocking conditions the amplitude increases and the location of the maximum happens to be located after the MFB 50 , closer to the end of combustion. Figure 5 shows the distribution of the resonance index for cylinder 1 using knock sensor A (top plots) and C (lower plots). The number of occurrences in terms of the maximum location respect to the MFB 50 and the amplitude are shown for 3 SA cases: left cases the SA is set at the reference value (SA mv ), middle and right cases show results advancing the SA 2,5 and 5,5 CAD respectively. Analyzing Figure 5, the resonance index computed from knock sensor A is more sensible to the SA change than knock sensor C, pointing out that, as expected, knock is easily identified if a sensor is placed near the cylinder. However, results with knock sensor C, still show potential for knock detection. 4 Knock recognition method The knock recognition method presented in [14] is used to develop a recognition technique based on knock sensor signal. The knock recognition method is proposed in this work is shown in Figure 6. The knock classification is performed by comparing the resonance index, I k , obtained from Equation (9), with the minimum oscillation resulting from the end gas to auto ignited, I k-min , which is estimated as: 407 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="408"?> where P eg is the pressure increase due to constant volume combustion of the end gas, and G k represents a gain from pressure increase to amplitude of the resonance indexes, which might be calibrated. Figure 5: Number of occurrences in terms of maximum location and amplitude resonance index respect to cylinder 1. Top plots Ks A and bottom plot Ks C. Figure 6: Scheme of knock recognition method based in knock sensor signal. The pressure increase due to constant volume combustion of the end gas can be written using the first law of thermodynamics as: 408 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="409"?> ( 11 ) where H p the low calorific value of the fuel, m f the fuel mass injected and MFB model the MFB from the combustion model, and κ is the adiabatic exponent and V is the cylinder volume. For this application, the minimum oscillation presented in Equation (10) instead of using the MFB computed from in-cylinder pressure signal as described in [14], the MFB is estimated from a pre-calibrated Wiebe function, which strongly simplifies the calculations and contributes to a future ECU implementation. After, following the scheme on Figure 6, the knock recognition is performed by comparing the amplitude of the maximum resonance index, with the value of the minimum oscillation evaluated at the crank angle position of the maximum resonance index, I k,min (α I ˆk). The comparison of the amplitude is performed in the crank angle position where the maximum resonance is reached because, not only the intensity of the resonance, but also the location with respect to the center of combustion characterize the cycles with knock or normal combustion (see Figure 5). 4.1 Calibration process In this section, the calibration process of the transference constant G k in Equation 10 is explained. The calibration was performed from the in-cylinder pressure measurements, where the difference between the resonance indicator I p and I p−min was computed and com‐ pared against the difference between the measurement from knock sensor signal and combustion model, i.e. I k and I k−min . These differences are represented in Figure 7, where the red dashed line represents the knock on-set detected by the method when in-cylinder pressure is used. 409 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="410"?> Figure 7: Data from calibration process: Knock sensor A (top) and knock sensor C (bottom) Notice that the linear regression used yFit has to be zero when I p −I p−min = 0, as this is the division between knocking and no-knocking cycles. In Figure 7 it can be notice that the R 2 related to knock evaluation is higher. The calibration of G k was performed by minimizing the distance of the data from the fitted line yFit, i.e min|yFit− P eg G k |. This calibration process was performed for all the knock sensors A, B and C separately. 5 Results and discussion During this section results from MFB model and knock recognition method applied to the different knock sensors signals are analyzed. First, the results from the MFB model are compared with the measurements from in-cylinder pressure. Then, the knock recognition method is applied to cylinders 1 and 3, by running the method for both knock sensors location for each case. 5.1 Minimum oscillation The objective of the MFB modeling is to follow the evolution of the minimum oscillation required to determine if a cycle is normal or knock combustion, as was shown in Figure 6. The minimum oscillation required for the end-gas to auto ignite computed from Equation (9) is represented in Figure 8 in yellow line, where the location with respect 410 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="411"?> to the CA50 and the maximum amplitude of the resonance index are also represented. Different colors are used to represent the MAPO amplitude from in-cylinder pressure signal. The left plot illustrates results from the resonance index from knock sensor A, while right plot knock sensor C. As can be seen in Figure 8, for the case of knock sensor C points are more concentrated than for knock sensor A, which translates into a difficulty in being able to differentiate the knock cycles from those of normal combustion. Figure 8: Distribution of resonance index location and maximum amplitude for knock sensor A and C. Colors represents MAPO amplitude from in-cylinder pressure 411 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="412"?> 5.2 Knock recognition The knock recognition method presented in Figure 6 is applied for cylinder 1 and knock sensors A and C. Results during this section are compared with the knock recognition when in-cylinder pressure is analyzed as described in [14]. As discussed in section 4, knocking cycles are detected when the maximum of the resonance index is above the minimum oscillation produced by the end gas auto ignition, computed from Equation (10). In Figures 9 and 10 the resonance indexes computed from knock sensors and in-cylinder pressure are analyzed for a normal combustion and a knocking cycle respectively. On the top plots, the resonance index evolution with the threshold are represented, on the left plot for knock sensors signals and on the right plot for in-cylinder pressure. On the bottom plot, the HRR and the band-pass in-cylinder pressure is represented. Figure 9: Resonance index evolution and detail cycle (HRR and p bp ) for a normal combustion cycle. Operating condition point 1300 rpm, 230 Nm (36% of full load): MAPO 0.27 bar 412 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="413"?> Figure 10: Resonance index evolution and detail cycle (HRR and p bp ) for a knocking cycle. Operating condition point 1300 rpm, 230 Nm (36% of full load): MAPO 0.26 bar As it can be seen in Figure 9 and 10, both cycles have a similar MAPO amplitude, but the resonance evolution it is not the same, i.e in normal combustion case the resonance is excited during combustion, near the maximum of the HRR, and on the other hand, for knocking case resonance is rapidly excited at the end of combustion. Notice that zoomed plots have the same y-axis length. The method was evaluated by comparing the knock probability when applying a MAPO threshold to knock sensor signal. Such threshold was estimated comparing the knock probability for different MAPO based on knock sensor signals and the one obtained by the proposed method as is shown in Figure 11. Figure 11: Knock probability computed from the different knock sensor location over five operating conditions. 413 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="414"?> The method was evaluated for in-cylinder pressure and knock sensors signals during different SA settings at steady state conditions. Results are represented in Figure 12, where the knock probability is represented as a function of the SA delay from the calibrated point. The method proposed is compared with two knock recognition methods based on in-cylinder pressure, on the one hand, the MAPO definition with a threshold of 0.4 bar, and on the other hand the high sensitivity method proposed in [14]. Figure 12: Knock recognition method over a SA sweep point 1 cylinder 1. Operating condition 1300 rpm 570 Nm. For the calibration point, SA = SA mv , the knock probability for the three methods is zero, but when advancing the SA knock probability increases for the three sensors. For knock sensor A, the knock probability is closer to the one recognized by the high sensitivity method based on in-cylinder pressure. On the other hand, for knock sensor C, the knock probability also increases when advancing the SA, but the probability is lower. For all SA settings the knock probability from knock sensor A is higher than for MAPO definition. If this threshold is applied for each knock sensor signal, the following results can be obtained. 414 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="415"?> (a) 415 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="416"?> (b) Figure 13: Confusion matrix for methods based on knock sensor signal compared with high sensitivity method based on in-cylinder pressure signal: (a) Knock sensor A. (b) Knock sensor C. As can be seen in Figure 13, the true positives obtained from knock sensor A are closer than the ones obtained from knock sensor C when the same methods are analyzed. When comparing results obtained for knock sensor A and C, it can be notice that the true positives for the proposed method are reduced on the half, while for MAPO criteria are reduced around a 20%. 416 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="417"?> 6 Conclusions In this work a high knock recognition method based on in-cylinder pressure signal was used to develop an alternative method based on knock sensor signal in a spark ignited HD engine. The method makes use of a Wiebe function to estimate the knock threshold evolution during a cycle, which is compare with a resonance index computed from different knock sensor locations signals. An analysis of the knock recognition sensitivity for two knock sensor locations has been performed over different operating conditions and spark advance settings, demonstrating that the method is able to distinguish knocking events with a high resolution for knock sensors located near the cylinder heads, case knock sensor A, while less resolution was observed by analyzing one knock sensor located between both cylinders, case knock sensor C. The main findings of this work were: • The low knocking recognition method based on in-cylinder pressure signal developed for light-duty engines is also valid for heavy-duty gas engines. • When the method is extended to knock sensor signal, the recognized cycles are less than those obtained with the low knocking recognition method but higher or equal than when applying a fixed threshold, as MAPO. • For one knock sensor per cylinder configuration the knock recognition resolution is between a high sensitivity cylinder-pressure method and the classical MAPO limit. • When the number of knock sensors is halved, the knock recognition resolution is reduced. • The proposed method exhibits a better recognition when applying a fixed thresh‐ old as MAPO to knock sensor signal, independently on the knock sensor location. As general summary of the presented work, the authors would like to point out the good performance of the proposed method for the continuous control of the knocking conditions in HD engines using low-cost knock sensors. By the way, the development of low-cost solutions for knocking control will be more and more important as the HD engine technology evolves towards complex and expensive powertrains (e.g. LNG or hydrogen fueled vehicles). The next step of this research will be the method applicability validation through dynamic tests. 7 Acknowledgments Irina Jimenez received a funding through the grant GRISOLIAP/ 2018/ 132 and BEFPI/ 2021/ 042 from the Generalitat Valenciana, Spain and the European Social Fund. 417 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="418"?> References [1] C. Bae, J. Kim, Alternative fuels for internal combustion engines, Proceedings of the Combustion Institute 36 (3) (2017) 3389-3413. [2] T. Korakianitis, A. Namasivayam, R. Crookes, Natural-gas fueled spark-ignition (si) and compression-ignition (ci) engine performance and emissions, Progress in energy and com‐ bustion science 37 (1) (2011) 89-112. [3] P. Tratzi, M. Torre, V. Paolini, L. Tomassetti, C. Montiroli, E. Manzo, F. Petracchini, Liquefied biomethane for heavy-duty transport in Italy: A well-to-wheels approach, Transportation Research Part D: Transport and Environment (2022) 107. doi.org/ 10.1016/ j.trd.2022.103288. [4] P. Matschoss" M. Steubing, J. Pertagnol, et al., A consolidated potential analy-sis of bio-methane and e-methane using two different methods for a medium-term renewable gas supply in Germany, Energ Sustain Soc 10, 41 (2020). https: / / doi.org/ 10.1186/ s13705-020-0027 6-z. [5] A. C. R. Teixeira, P. G. Machado, F. M. de Almeida Collaco, D. Mouette, Alternative fuel technologies emissions for road heavy-duty trucks: a review, Environmental Science and Pollution Research (2021) 1-16. [6] H. Wei, R. Zhang, L. Chen, J. Pan, X. Wang, Effects of high ignition energy on lean combustion characteristics of natural gas using an optical engine with a high compression ratio, Energy 223 (2021) 120053. [7] X. Li, X. Zhen, S. Xu, Y. Wang, D. Liu, Z. Tian, Numerical comparative study on knocking combustion of high compression ratio spark ignition engine fueled with methanol, ethanol and methane based on detailed chemical kinet-ics, Fuel 306 (2021) 121615. [8] M. I. Jahirul, H. H. Masjuki, R. Saidur, M. Kalam, M. Jayed, M. Wazed, Com-parative engine performance and emission analysis of cng and gasoline in a retrofitted car engine, Applied Thermal Engineering 30 (14-15) (2010) 2219-2226. [9] S. Sahoo, D. K. Srivastava, Effect of compression ratio on engine knock, performance, combustion and emission characteristics of a bi-fuel cng engine, Energy (2021) 121144. [10] Z. Chen, Y. Ai, T. Qin, F. Luo, Quantitative evaluation of nbutane concentra-tion on knock severity of a natural gas heavyduty si engine, Energy 189 (2019) 116244. [11] A. A. Quader, Lean combustion and the misfire limit in spark ignition engines, SAE transactions (1974) 3274-3296. [12] C. Park, C. Kim, Y. Choi, S. Won, Y. Moriyoshi, The influences of hydrogen on the performance and emission characteristics of a heavy duty natural gas engine, International journal of hydrogen energy 36 (5) (2011) 3739-3745. [13] J. B. Heywood, Internal combustion engine fundamentals, McGraw-Hill Edu-cation, 2018. [14] B. Pla, J. De La Morena, P. Bares, I. Jim´enez, Knock analysis in the crank angle domain for low-knocking cycles detection, Tech. rep., SAE Technical Paper (2020). [15] C. S. Draper, Pressure waves accompanying detonation in the internal com-bustion engine, Journal of the Aeronautical Sciences 5 (6) (1938) 219-226. [16] X. Zhen, Y. Wang, S. Xu, Y. Zhu, C. Tao, T. Xu, M. Song, The engine knock analysis-an overview, Applied Energy 92 (2012) 628-636. 418 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="419"?> [17] R. Novella, B. Pla, P. Bares, I. Jim´enez, Acoustic characterization of combus-tion chambers in reciprocating engines: An application for low knocking cy-cles recognition, International Journal of Engine Research (2020) 1468087420980565. [18] P. Bares, D. Selmanaj, C. Guardiola, C. Onder, A new knock event definition for knock detection and control optimization, Applied Thermal Engineering 131 (2018) 80-88. [19] F. Bi, T. Ma, X. Wang, Development of a novel knock characteristic detection method for gasoline engines based on waveletdenoising and emd decomposi-tion, Mechanical Systems and Signal Processing 117 (2019) 517-536. [20] D. Siano, M. A. Panza, A nonlinear black-box modeling method for knock detection in spark-ignition engines, in: AIP Conference Proceedings, Vol. 2191, AIP Publishing LLC, 2019, p. 020137. [21] D. Siano, D. D’agostino, Knock detection in si engines by using the discrete wavelet transform of the engine block vibrational signals, Energy Procedia 81 (2015) 673-688. [22] B. Pla, J. De La Morena, P. Bares, I. Jim´enez, Adaptive incylinder pressure model for spark ignition engine control, Fuel 299 (2021) 120870. [23] G. Brecq, J. Bellettre, M. Tazerout, A new indicator for knock detection in gas si engines, International Journal of Thermal Sciences 42 (5) (2003) 523-532. [24] X. Shen, Y. Zhang, T. Shen, Cylinder pressure resonant frequency cyclic estimation-based knock intensity metric in combustion engines, Applied Thermal Engineering 158 (2019) 113756. [25] A. Hoth, C. P. Kolodziej, Effects of knock intensity measurement technique and fuel chemical composition on the research octane number (ron) of face gasolines: Part 1-lambda and knock characterization, Fuel (2021) 120722. [26] J. Kim, In-cylinder pressure based engine knock classification model for high-compression ratio, automotive spark-ignition engines using various signal de-composition methods, Energies 14 (11) (2021) 3117. [27] K. Bizon, G. Continillo, S. Lombardi, E. Mancaruso, B. M. Vaglieco, Ann-based virtual sensor for on-line prediction of incylinder pressure in a diesel engine, in: Computer Aided Chemical Engineering, Vol. 33, Elsevier, 2014, pp. 763-768. [28] K. Akimoto, H. Komatsu, A. Kurauchi, Development of pattern recognition knock detection system using short-time fourier transform, IFAC Proceedings Volumes 46 (21) (2013) 366-371. [29] F. Bi, X. Li, C. Liu, C. Tian, T. Ma, X. Yang, Knock detection based on the optimized variational mode decomposition, Measurement 140 (2019) 1-13. [30] C. Draper, The physical effects of detonation in a closed cylindrical chamber, Twentieth Annual Report (1935) 361-376. [31] M. Lapuerta, O. Armas, J. Herna´ndez, Diagnosis of di diesel combustion from in-cylinder pressure signal by estimation of mean thermodynamic prop-erties of the gas, Applied Thermal Engineering 19 (5) (1999) 513-529. [32] O. Chiavola, G. Chiatti, L. Arnone, S. Manelli, Combustion characterization in diesel engine via block vibration analysis, Tech. rep., SAE Technical Paper (2010). [33] C. Guardiola, B. Pla, P. Bares, A. Barbier, An analysis of the in-cylinder pres-sure resonance excitation in internal combustion engines, Applied Energy 228 (2018) 1272-1279. 419 Efficient Knock recognition algorithms for heavy-duty spark ignited gas engine based on vibration signal and Wiebe function <?page no="420"?> [34] X. Zhao, Z. Li, Z. Li, L. Wang, Combustion parameters estimation based on multi-channel vibration acceleration signals, Applied Thermal Engineering 158 (2019) 113835. 420 Pierpaolo Napolitano, Irina Jimenez, Benjamín Pla, Carlo Beatrice <?page no="421"?> Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine M.Sc. Fabian Steeger 1 , Dr.-Ing. Marco Günther 1 , Dr.-Ing. Eike Stitterich 2 , Prof. Dr.-Ing. Stefan Pischinger 1 1 2 RWTH Aachen Lehrstuhl für Thermodynamik mobiler Energiewandlungssysteme (TME), Forckenbeckstr. 4 52074 Aachen Hengst SE, Nienkamp 55-85 48147 Münster Abstract: Despite extensive research in recent years, low-speed pre-ignition (LSPI) remains an issue with high need for improvement among abnormal combustion phenomena in modern spark ignition engines. Oil droplets in the combustion chamber during the compression stroke are one of the causes of LSPI. Oil ingress can be caused by sealings of intake system parts that are not 100% tight, or by recirculated engine blow-by gases. Furthermore, the lubricating oil formulation can also play an important role. Previous studies have focused on additives such as calcium as a known LSPI enhancer and molybdenum as an LSPI inhibiter. Studies on external oil droplet dosing into a single combustion chamber or directly upstream of the intake valves can be found in the literature, showing interesting dependencies of LSPI on the applied external oil dosing in gasoline spark ignition engines. To investigate the impact of external oil sources under realistic engine conditions on a test bench, a multi-cylinder series-production gasoline direct-injection engine was modified to employ an oil aerosol generator at specific locations in the air path. The objective was to simulate worn oil-lubricated components as well as possible oil reservoirs. The aerosol generator is designed to simulate droplet size distributions similar to that of blow-by aerosols from the crankcase. Furthermore, various oil additive formulations with increased calcium and molybdenum con‐ tent, as well as an aged oil used as externally dosed oil, were investigated. The test engine was operated in low-end torque conditions for several hours with applied oil dosing to capture the highly statistical abnormal combustion phenomenon. The results show an increased pre-ignition sensitivity of the test engine under low-end-torque conditions at oil dosage up to 40 g/ h. At oil dosage rates above 40 g/ h, the pre-ignition occurrence increased significantly, showing a clear <?page no="422"?> relationship between oil mass flow and pre-ignition sensitivity. However, pre-ig‐ nitions did not occur subsequently after the oil dosing has been engaged during low-end torque. This suggests that either critical mixture field conditions must be present in the cylinder, that the oil droplet aerosol is burned to form particles that build up into a hot-spot reservoir or that oil droplets agglomerate in the intake system and been transported into the combustion chamber at unique events and eventually trigger pre-ignition. The investigation on different externally dosed test oils did not reveal significantly different pre-ignition sensitivities. This can be explained by the generally lower reactivity of synthetic engine oils (base group IV) and indicates that the influence of the base group dominates the effect of the additive formulation. Introduction The reduction of climate-changing CO 2 -emissions in the transport sector has been laid down by law in a limited fleet consumption [1]. In addition to electrification, the development of low-consumption and low-emission combustion engines must contribute to achieving these goals as well. Downsizing, the representation of an engine with the same power output despite a reduction in displacement with compensation through charging, offers a possibility to reduce losses during partial load compared to a conventional gasoline engine. This trend is leading to power units with high specific outputs at high boosting levels. A disadvantage of highly charged engines is the increased sensitivity to knocking. In addition to the increasing tendency to knock, an increased pre-ignition tendency must be expected in highly charged engines as well. Pre-ignition is an irregular combustion phenomenon with auto-ignition of the gasoline-air mixture occurring before the spark discharge. This can be a singular event or a sequence of several partially intermittent pre-ignitions. Pre-ignition risk is increased in the operating range of low engine speeds and high loads. This is accompanied by high cylinder pressures and pressure fluctuations, as shown in Figure 1. The pressure curves show that the peak pressure is significantly higher than that of regular combustion and that the pressure fluctuations are significantly higher than those of knocking combustion. The probability of occurrence of irregular combustion phenomena must be limited, since even singular events can cause critical damage to the engine through high thermal and mechanical stress. Pre-ignition can be triggered by a number of mechanisms: • Ignition by hot surfaces (glow ignitions), • Ignition by particles or deposits, • Ignition by oil and fuel droplets, • Ignition by burning mixture residues and/ or • Ignition by hot residual gas. 422 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="423"?> Figure 1: Cylinder pressure traces of regular and irregular combustion The various triggering mechanisms can superimpose each other and have a different influence on the pre-ignition depending on the operating conditions. In this regard, all mechanisms must be considered in the development of modern, turbocharged gasoline engines and their influence must be minimized. Ignition by oil droplets is one of the most complex pre-ignition mechanisms and until now not fully understood. One of the reasons for this complexity is that, in addition to the combustion chamber, there are various other, often different sources for the introduction of oil droplets into the combustion process. In addition, the chemical and physical mechanism until the occurrence of a pre-ignition depends strongly on the characteristic properties of the respective engine oil. This complexity leads to unclear requirements for components that are not directly related to the combustion process, such as oil separators in the crankcase ventilation system, oil seals on the turbocharger or valve stem seals. It is clear from the above explanations that the present problem requires a close link between the combustion process development and the design of the oil lubricated engine auxiliary equipment. The objective of this paper is the detailed understanding of the influence of oil input in the intake air path on the pre-ignition tendency. Therefore, the pre-ignition behaviour will be analysed by utilizing different engine oils and external oil dosing configurations. The focus here is on droplets that find their way into the air intake system via the turbocharger shaft seal, via the valve stem seals or from the crankcase ventilation system. Thermodynamic Multi-Cylinder Engine Investigations With the target to investigate the pre-ignition behavior of different test oils in real engine operation it was inevitable to provide a state-of-the-art series production engine for testing. In order to meet this requirement, the Chair of Thermodynamics of Mobile 423 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="424"?> Energy Conversion Systems (TME) of the RWTH Aachen University provided a modern supercharged four-cylinder gasoline engine with pressure indication applied on each cylinder. The essential specifications of the multi-cylinder engine are listed in Table 1. During the 4-hour steady-state measurement, the peak pressure was monitored on all cylinders. When the peak pressure exceeded 90 bar, a measurement was triggered in which the 50 cycles before and 250 cycles after the triggering event were saved from the circular buffer storage covering 300 consecutive cycles.The identification of pre-ignition in a triggered measurement is carried out via a heat release rate analysis of the combustion according to the methodology from FVV Fuel Characteristics Numbers II [1]. In this experiment, 2% burned mixture mass fraction (MFB2%) before the onset of the ignition spark was used as the criterion for pre-ignition. Table 2 describes the operating parameters of the pre-ignition test. The operating point corresponds to a typical “low-end torque” operating point of supercharged gasoline engines. The high load at low speeds represents critical conditions for the occurrence of pre-ignition, because the chemical reaction paths that lead to the development of pre-ignition take place at higher temperatures of the unignited charge in less time. The low engine speed offers a relatively long characteristic time per working cycle before external ignition. Bore / mm 79 Stroke / mm 81.4 Number of Cylinders 4 Stroke/ Bore-Ratio / 1 1.03 Engine Displacement / cm 3 1596 Valves per cylinder 4 Compression ratio / 1 10 Table 1: Technical data of the four-cylinder series production engine Engine speed / 1/ min 1500 Load (Indicated mean effective pressure) / bar 19.5 Boost pressure abs. / mbar 1850 Air-fuel-ratio (exhaust gas) λ / 1 1.00 Inlet valve timing related to 1 mm valve lift / ° CA bTDC GE 15 Exhaust valve timing related to 1 mm valve lift / ° CA bTDC GE -19 Fuel pressure / bar 130 Start of injection / ° KW ° CA bTDC F 294 Inlet air temperature / °C 32 Table 2: Pre-ignition test boundary conditions on the four-cylinder gasoline engine 424 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="425"?> Controlled Oil Dosing into the Air Path The test oils were introduced into the air path as aerosol. For this purpose, an aerosol generator was used (Palas PLG 2100). This generator is capable of conditioning an aerosol with droplets of the size that are usually present in the crankcase or reach the intake air path of the engine via the crankcase ventilation [1] (see Table 3). Furthermore, a wide range of oil concentrations in the aerosol with oil mass flows of up to 100 g/ h was covered. Volume flow / l/ min 3 - 110 Mass flow (oil - droplets) / g/ h < 100 (white oil) Avg. droplet diameter / µm 1.5 Oil reservoir volume / l 1 Additional feature Oil reservoir heatable up to 120 °C Table 3: Technical specifications of the selected oil-aerosol generator (source: Palas GmbH ) Figure 2: Functional diagram of the selected aerosol generator (Quelle: Palas GmbH ) The operating principle of the aerosol generator is shown in Figure 2. The main element is the so-called Laskin nozzle. It is completely immersed in the oil reservoir and has up to 30 holes through which compressed air flows into the liquid oil phase. The high nozzle exit velocities lead to the formation of gas inclusions in the liquid oil reservoir containing finely atomized oil droplets. These inclusions grow in diameter and move towards the surface of the liquid oil reservoir inside the aerosol generator, where the finely dispersed oil droplets are released into the air gas phase. The generated engine oil aerosol is subsequently transported through the aerosol generator exit to the connected 425 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="426"?> oil dosing position of the engine intake air path by utilizing a pressure difference between the compressed air supply and the counter pressure at the aerosol generator exit. Further elements are the heating mat, which can heat up the oil reservoir up to 120 °C and a thermocouple, which monitors the temperature of the oil and transmits a feedback to the temperature control unit. The engine oil is filled into a laboratory glass, which is supported by the heating mat and the Laskin nozzle, thermocouple and aerosol outlet are mounted on the laboratory glass cover, which also functions as a sealing for the entire laboratory glass. The controlled oil dosing was conducted at 4 different positions of the intake air path (Figure 3): upstream compressor, through the original full load blow-by recirculation connection, upstream and downstream charge air cooler and directly into the intake manifold. The last mentioned dosing position was designed to introduce the oil aerosol directly upstream cylinder 4 and, therefore, simulate an increased oil contamination via the valve stem seal(s) of this cylinder. The oil dosing upstream the intake manifold for the remaining 3 dosing positions leads to an aerosol inflow through the original air path ducting. The objective of this methodology was the systematic evaluation of possible oil source in the air path with respect to their influence on the pre-ignition sensitivity of the test engine. A variation of the dosed oil mass flow was additionally conducted to simulate not ideally designed, worn or defective components. Figure 3: Diagram of the air path with compressor, intercooler intake manifold and the 4 selected dosing positions for the oil-aerosol generator Test Oil Selection A number of experimental studies are already known in literature which examine various base oils, ageing conditions and additives with regard to their influence on the occurrence of pre-ignition. [3-11] Five different test oils were selected: • Standard oil: reference engine oil (type 5W-20), • Calcium oil: increased Calcium content compared to Standard oil (type 5W-20), • Molybdenum oil: increased Molybdenum content compared to Standard oil (type 5W-20), • Aged oil #1: Standard oil aged in the test engine (type 5W-20), 426 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="427"?> • Aged oil #2: Alternative aged engine oil (type 5W-30). A fully synthetic oil with a viscosity class of 5W-20 is the reference engine oil from which two variants are formulated as a derivative with an increased calcium or molybdenum content. Both additives are known to influence the occurrence of engine oil-induced pre-ignition [11]. This fact is taken into account in the underlying oil selection. The ageing process of engine oils is also suspected to change the composition of the oil in such a way that the pre-ignition of oil droplets is promoted. For this reason, two different aged oils are included in this study. The ageing of the standard oil (Aged #1) is based on a lifetime of 500 hours in the four-cylinder test engine. The ageing of the alternative oil (Aged #2) was carried out by Hengst SE. Figure 4 shows an analysis from literature, where it is apparent that calcium as an additive in engine oil can increase the frequency of oil-induced pre-ignition. Figure 4: Experimental investigation of the impact of Molybdenum (left) and Calcium (right) on pre-ignition sensitivity [11] Furthermore, it is also known that an increased fuel content in the engine oil, as well as dissolved metallic abrasion due to wear (e.g. iron and copper) increases the pre-ignition rate. [12] The concentration of metallic additives in the respective test oils as well as their presumed effect on the probability of occurrence of pre-ignition can be seen in Table 4. 427 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="428"?> Table 4: List of test oils and selected results of the respective oil analysis. With the knowledge obtained from the work of Takeuchi et al. the following basic expectations can be established a priori as to how the test oils investigated here will affect the pre-ignition tendency of the test engine. The molybdenum oil has a content of 0.05% and thus a concentration which should reduce the pre-ignition rate to a minimum compared to the standard oil. On the other hand, the calcium enriched oil has a 60-70% higher calcium content compared to the other test oils. The analysis of aged engine oils showed a slightly increased fuel content, as well as lower amounts of iron and copper as metallic abrasion from engine operation. This analysis leads to the expectation that the occurrence of pre-ignition could be increased [12]. Experimental Results Prior to the pre-ignition investigations with regard to controlled oil dosing into the original intake air path, a baseline pre-ignition sensitivity of the full engine was determined due to the numerous possible causes of pre-ignition in the combustion chamber, which can superimpose the effect of the controlled oil dosing [13] (Figure 13). Figure 5 shows the average initial pre-ignition rate per hour for two different blow-by configurations depending on the different test oils from Table 1 as engine oil lubricant. The error bars in the graph represent the minimum and maximum values of several tests. 428 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="429"?> Figure 5: Averaged results of the pre-ignition baseline (bars) with the respective test oil as lubricant and two different blow-by configurations (original connected to the air path and disconnected by means of a special structure) after 4 hours of steady state operation at the low-end torque operating point and their minimum and maximum values as error bars. It can be observed that there is no clear trend between the different test oils as engine lubricant and the configuration of the blow-by recirculation does not show a clear influence on the average pre-ignition frequency. The baseline pre-ignition level of the engine is low compared to experimental investigations found in literature [10-14]. However, this engine provides a suitable basis for a controlled oil dosing via the air path into the combustion chamber as a sensitivity analysis with regard to oil drop-induced ignition. The controlled oil dosing upstream of the compressor is the furthest dosing position upstream from the combustion chamber among all examined oil injection points. Figure 6 shows the results of the experimentally determined initial pre-ignition rate per hour. The data points represent an oil dosing class defined with a tolerance of ± 5 g/ h dosed oil mass flow and the error bars are the respective absolute maximum deviations regarding the abscissa and ordinate. For comparison, the baseline pre-ignition level is shown as a horizontal line in each graph. 429 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="430"?> Figure 6: Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different test oils upstream the compressor This comparison shows a very small difference in the initial pre-ignition rate at oil input mass flows of 30 - 40 g/ h with standard oil and with molybdenum-enriched oil compared to the baseline pre-ignition level of the engine. In comparison, the calcium-enriched oil and the aged oil show a slightly increased pre-ignition rate, but no clear difference from the other test oils can be observed. However, above the concentration of 40 g/ h, an increase of the initial pre-ignition rate is clearly visible for all test oils. Between 70 g/ h and 90 g/ h there are 10-18 initial pre-ignitions per hour, which corresponds to a combustion anomaly every 3-6 minutes. This means, that only for unrealistically high dosed oil mass flows of more than 40 g/ h a significant effect on the pre-ignition frequency could be shown. The experimental setup discussed here leads to a significantly lower back pressure acting on the outlet of the aerosol generator. Since a lower aerosol generator counter pressure in this experimental study allows a higher maximum aerosol flow rate, larger oil mass flows can be achieved upstream of the compressor than is possible for the other dosing positions downstream of the compressor. The critical dosed oil concentration from which an increased tendency to pre-ignition is observed, is higher by a factor of 4 compared to a typical total oil consumption of a gasoline engine of approx. 10 g/ h [14] and shows the maximum measured initial pre-ignition rate with an externally dosed oil mass flow of 7 times of the typical oil concentration value. Based on the results in Figure 6, a proportional influence of the amount of oil introduced on the initial pre-ignition rate can be assumed. This influence is confirmed by a statistical variance analysis where the authors would like to refer to the final report for further details. There is no difference in pre-ignition tendency between the test oils visible. 430 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="431"?> The result of the controlled oil-aerosol input upstream charge air cooler is shown in Figure 7. The selection of test oils was initially reduced to the most sensitive test oils (Calcium and Standard) based on the results of the controlled oil dosing upstream compressor in Figure 6. As already mentioned before, there is a higher counter pressure for the aerosol generator for all dosing points downstream compressor, which limits the maximum possible oil dosing mass flow to approx. 40 g/ h. Compared to the baseline pre-ignition sensitivity, the number of pre-ignitions with controlled oil dosing is not increased for either of the utilized test oils. Consequently, no influence of the controlled oil dosing upstream charge air cooler on the initial pre-ignition rate compared to the baseline pre-ignition sensitivity can be detected. Analogous to the results in 6, no difference in pre-ignition sensitivity between the tested oils can be observed. Figure 7: Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different test oils upstream charge air cooler Figure 8: Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different test oils downstream charge air cooler 431 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="432"?> The pre-ignition sensitivity of the test engine is also not increased in comparison to the baseline sensitivity, when the oil dosing is conducted downstream charge air cooler (Figure 8), which is a similar result observed in Figure 7. Analogous to the experiment in Figure 7, only the most sensitive test oils were examined. The comparison of the results of the oil input upand downstream charge air cooler allows the conclusion, that the charge air cooler has no significant influence on the pre-ignition tendency of the test engine, since similar results were obtained at both oil dosing positions. The fourth dosing position was applied directly into the intake manifold of the engine. The results show a generally slightly increased initial pre-ignition rate of the controlled oil dosing compared to the baseline pre-ignition frequency, which is not dependent on the oil dosing mass flow (Figure 9). The statistical variance analysis supports, that no significant impact of the oil dosing mass flow on the occurrence of pre-ignition up to 40 g/ h can be observed. The pre-ignition sensitivity is comparable to the results of the controlled oil dosing upstream compressor for oil dosing mass flows up to 40 g/ h (see Figure 6). Also for the oil dosing position directly upstream the cylinders, no difference regarding the occurrence of pre-ignition between the tested oils can be observed. Figure 9: Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different test oils directly into the intake manifold Figure 9. Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different test oils directly into the intake manifold The previously observed pre-ignition promoting mechanism with increasing dosed oil mass flow can be investigated by an additional experimental setup. A highly efficient electrical disk separator [15] was installed between the aerosol generator and the selected dosing position in order to influence the initially dosed concentration of the oil aerosol. 432 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="433"?> Figure 10 shows the result of this investigation based on the dosing of various aged motor oils upstream compressor. The electrical disk separator is switched on (separation efficiency > 95% ) and switched off or passively applied (separation efficiency after gravimetric measurement of the separated engine oil approx. 50%). At maximum separation efficiency, the initial pre-ignition frequency is decreased to the baseline pre-ignition level and, thus, compensates the influence of the controlled oil input. At a separation efficiency of approx. 50%, a proportionally decreasing pre-ignition frequency can also be observed. Both observations confirm the influence of the oil concentration in the aerosol on the occurrence of initial pre-ignitions. Figure 10: Experimentally determined initial pre-ignition rate with controlled oil-aerosol dosing of different aged engine oils upstream compressor and the influence of an electrical disk separator applied between the aerosol generator and the dosing position. The result that the oil aerosol dosing into the intake manifold directly upstream cylinder 4 has no significant influence on the initial pre-ignition rate is unexpected, since the oil droplet transport towards the cylinders is the shortest among all 4 dosing positions. The objective by applying this dosing position is to simulate the oil input via the valve stem seal on one or more cylinders and to compare the resulting pre-ignition rate among all cylinders. For this purpose, the relative initial pre-ignition frequency for each dosing position was evaluated in Figure 11 and divided into two categories: the oil aerosol is dosed upstream the intake manifold and the aerosol is dosed directly into the intake manifold upstream cylinder 4. For both categories, a different cylinder-individual pre-ignition frequency is evident. If the oil aerosol is dosed upstream the intake manifold, cylinder 1 experiences the highest pre-ignition frequency, whereas cylinder 3 counts the lowest pre-ignition frequency. The frequency distribution across the cylinders is completely reflected for all cylinders when the oil aerosol is dosed directly into the intake manifold in front of cylinder 4. It 433 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="434"?> is particularly noticeable that the largest percentage of pre-ignition occurs in cylinder 4 as well. Figure 11: Relative cylinder-individual pre-ignition frequency The evaluation of the cylinder-individual pre-ignition frequency in Figure 11 also shows, that the intake manifold has an influence on the exposure to pre-ignition of each cylinder. The root cause of the problem cannot be clarified on the basis of this experimental approach and needs further experimental work and numerical simulations. Conclusions & Outlook Downsizing is one of the most widespread strategies to reduce CO 2 emissions from internal combustion engines, but at the same time it increases risk of abnormal combustion phenomena such as knocking and pre-ignition. The causes of pre-ignition have not yet been fully understood. Especially pre-ignition by oil droplets is one of the most complex mechanisms, which is superimposed by many other influences in the combustion chamber and also seems to be strongly dependent on the engine oil composition. The uncertainty about the effect of oil droplet induced pre-ignition leads to open questions in the field of the design of components in the oil system of gasoline engines. The aim of this work was to investigate the oil droplet influence on the pre-ignition tendency at typical low-end torque operating point. The investigations focused on the compressor, the charge air cooler and the intake manifold, where the oil droplets were introduced into the air path at various points along their path. The controlled oil input into the air path was intended to fundamentally investigate the mechanism of pre-ignition at oil droplets during engine operation, which was supplemented by the use of a test oil matrix of 5 differently 434 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="435"?> formulated and aged engine oils. Basic experimental investigations on the production engine regarding the pre-ignition sensitivity of the selected test oils at different dosing positions lead to the following findings: • No difference in pre-ignition sensitivity of the different test oils has been deter‐ mined. • With increasing dosed oil aerosol mass flow rate, the pre-ignition frequency increases, especially from approx. 40 g/ h, which, is above the typical comparable motor oil consumption of 10 g/ h. • The oil droplet transport through the charge air cooler has no influence on the initial pre-ignition rate. • The averaged cylinder-individual distribution of the relative pre-ignition fre‐ quency is reflected in comparison to the central oil droplet inflow (max. cylinder 1 = 47%) when the aerosol is dosed directly into the intake manifold upstream cylinder 4 (max. cylinder 4 = 41%) and shows that the inflow position of the oil droplets into the intake manifold has a decisive influence on the cylinder-individ‐ ual pre-ignition rate. In this paper the influence of controlled oil input into the air path, and finally into the combustion chamber, on the occurrence of pre-ignition in turbocharged gasoline engines in steady state engine operation was fundamentally investigated. However, no influence of different engine oil additive packages (enriched Calcium and Molybdenum) could be observed. Because very large dosed quantities of engine oil aerosol led to an increased pre-ignition rate in the full engine, it becomes clear that the basic pre-ignition mechanism is far more complex than the sole presence of engine oil droplets in the combustion chamber. Future planned work is to investigate the oil droplet transport through the compres‐ sor, charge air cooler and oil droplet and mass distribution into the respective cylinders with experimental optical measurements in engine operation and corresponding validated numerical investigations. These investigation are intended answer what droplet sizes are reaching the cylinders and obtain further information about the oil aerosol mass flows reaching the individual cylinders participating the combustion process. Since the avoiding pre-ignitions will remain a relevant topic in future engine developments with regard to the integration of full-load EGR, downspeeding and rightsizing in hybrid applications and for alternative fuels, there is a need for further research in this field. Of particular interest are the investigations in transient engine operation at full load (load steps), as well as the extension of the investigations to different droplet sizes. 435 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="436"?> Acknowledgements The presented work has been performed as part of the research tasks within the project “Oil Input into Combustion” defined and financed by the Research Association for Combustion Engines (FVV) e. V. (FVV Project 1283). The authors would like to thank the steering committee that accompanied the research work and all the companies involved for their support as well as the FVV for granting the financing. References [1] O. Budak, J. Dedl, and A. Heufer, Kraftstoffkennzahlen Biofuels II, Entwicklung geeigneter Kennzahlen und Korrelationen für die Charakterisierung von Ottokraftstoffen sowie alter‐ nativen Kraftstoffen mit Bioanteilen zur Beschreibung abnormaler Verbrennungsphänomene (2018). [2] Palas GmbH, Produktbeschreibung & Datenblatt PLG2100. [3] M. Kassai, T. Shiraishi, T. Noda, M. Hirabe, Y. Wakabayashi, J. Kusaka, and Y. Daisho, An Investigation on the Ignition Characteristics of Lubricant Component Containing Fuel Droplets Using Rapid Compression and Expansion Machine, SAE Int. J. Fuels Lubr. 9, 469 (2016). [4] T. Kaneko, K. Yamamori, H. Suzuki, K. Onodera, and S. Ogano, “Friction Reduction Technology for Low Viscosity Engine Oil Compatible with LSPI Prevention Performance,” in SAE Technical Paper Series (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2016). [5] K. A. Fletcher, L. Dingwell, K. Yang, W. Y. Lam, and J. P. Styer, Engine Oil Additive Impacts on Low Speed Pre-Ignition, SAE Int. J. Fuels Lubr. 9, 612 (2016). [6] A. Gupta, H. Shao, J. Remias, J. Roos, Y. Wang, Y. Long, Z. Wang, and S.-J. Shuai, “Relative Impact of Chemical and Physical Properties of the Oil-Fuel Droplet on Pre-Ignition and Super-Knock in Turbocharged Gasoline Engines,” in (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2016). [7] Y. He, Z. Liu, I. Stahl, G. Zhang, and Y. Zheng, “Comparison of Stochastic Pre-Ignition Behaviors on a Turbocharged Gasoline Engine with Various Fuels and Lubricants,” in SAE Technical Paper Series (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2016). [8] M. Kassai, K. Torii, T. Shiraishi, T. Noda, T. K. Goh, K. Wilbrand, S. Wakefield, A. Healy, D. Doyle, R. Cracknell, and M. Shibuya, “Research on the Effect of Lubricant Oil and Fuel Properties on LSPI Occurrence in Boosted S. I. Engines,” in SAE Technical Paper Series (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2016). [9] K. Onodera, T. Kato, S. Ogano, K. Fujimoto, K. Kato, and T. Kaneko, “Engine Oil Formulation Technology to Prevent Pre-ignition in Turbocharged Direct Injection Spark Ignition Engines,” in SAE Technical Paper Series (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2015). 436 M.Sc. Fabian Steeger, Marco Günther, Eike Stitterich, Stefan Pischinger <?page no="437"?> [10] K. Morikawa, Y. Moriyoshi, T. Kuboyama, T. Yamada, and M. Suzuki, “Investigation of Lubricating Oil Properties Effect on Low Speed Pre-Ignition,” in SAE Technical Paper Series (SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2015). [11] K. Takeuchi, K. Fujimoto, S. Hirano, and M. Yamashita, Investigation of Engine Oil Effect on Abnormal Combustion in Turbocharged Direct Injection - Spark Ignition Engines, SAE Int. J. Fuels Lubr. 5, 1017 (2012). [12] Toyota Motor Coporation, Investigation on Engine Oil Effect on Abnormal Combustion in Turbocharged Direct Injection - Spark Ignitions Engines (2014). [13] C. Dahnz, H. Kubach, U. Spicher, M. Magar, R. Schießl, and U. Maas, Pre-Ignition in Supercharged SI Engines (Stuttgart, Germany, 2010). [14] E. Yilmaz, Sources and Characteristics of Oil Consumption in a Spark-Ignition Engine. Dissertation (Massachusetts Institute of Technology, Cambridge, MA, 1997). [15] Hengst SE & Co. KG, Betriebsanleitung für Type: AS850E Elektrischer Tellerseparator (2015). 437 Effect of external oil sources in the air path on abnormal combustion phenomena of a turbocharged gasoline direct injection multi-cylinder engine <?page no="439"?> Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. L. Wißmann a , P. Süess b , M. Grill c , K. Herrmann b and M. Bargende a a b c Institute of Automotive Engineering (IFS), University of Stuttgart, Pfaffenwaldring 12, 70569 Stuttgart, Germany. Institute of Thermal and Fluid Engineering (ITFE), School of Engineering (HST), University of Applied Sciences and Arts Northwestern Switzerland (FHNW), Kloster‐ zelgstrasse 2, 5210 Windisch, Switzerland. Research Institute of Automotive Engineering and Vehicle Engines Stuttgart (FKFS), Pfaffenwaldring 12, 70569 Stuttgart, Germany. Abstract: Undesired self-ignition of premixed gas-air charges in dual-fuel engines caused by burning lube oil droplets continue to endanger both performance and structural integrity of those engines. Since experimental investigations in engines are demanding and associated with enormous costs a novel experimental test facility featuring dual-fuel operation was used to provide reference data for suitable 0/ 1D pre-ignition model development and validation. The test bench is equipped with an optically accessible combustion chamber, whose adaptable setup allows mounting plug-ins for specific lube oil addition. Moreover, it features adjustable operation at engine relevant conditions (com‐ pression pressure/ temperature) as well as a wide range of speed, tuneable flow (turbulence), variable gas/ air charge composition and high flexibility by pneu‐ matic driven valve train. Already established optical acquisition of ignition and flame propagation is applicable to detect inflammation and flame kernel growth of pre-ignition phenomena. Comprehensive instrumentation determines in-cylinder pressure, process gas temperature and other boundary conditions needed for 0/ 1D model validation and development. Phenomenological modelling needs a deep understanding of the underlying processes. Pre-ignition denotes a complex interaction between the cycle-to-cycle variations of flow field, turbulence, and levels of inhomogeneities, and further between reaction kinetic processes and the random existence of potential ignition kernels. The optical investigations and fundamental thermodynamic measure‐ ments performed at the experimental test facility determined pre-ignition influ‐ encing parameters. Simulation of the pre-ignition occurrence including detailed reaction kinetic mechanism is validated against test bench investigations. Initial <?page no="440"?> results of the pre-ignition model show good agreement with the optical test bench measurements. The fast 0D/ 1D model is capable to predict pre-ignition events with boundary conditions from the test bench using lube oil droplet evaporation combined with detailed reaction kinetics. 1 Introduction In view of emission legislations, carbon emissions must be reduced by improving the combustion processes and/ or by using alternative fuels with lower carbon content. Gas or dual-fuel engines using natural gas for example feature efficiencies comparable to diesel engines with reduced CO 2 as well as considerably lower particulate and NO X emissions. The underlying combustion concept of dual fuel engines is based on the ignition of a lean premixed gas/ air cylinder charge by the injection of a more reactive liquid pilot fuel or gas jet. With increasing cylinder charge pressure and temperature (to improve engine efficiency) however, the reliable engine operating range becomes increasingly limited due to “knocking” or “misfiring” effects. Moreover, unwanted pre-ignition events can occur since the premixed charge is more and more susceptible to early ignition phenomena with rising pressure and temperature levels, s, often due to self-ignition of lube oil in hot zones. Pre-ignitions can cause very steep cylinder pressure gradients and high peak pressures / temperatures that reduces the efficiency and can even damage the engine. Various pre-ignition sources and their classification with respect to automotive applications are presented in [1] and [2]. Origins thereof are often hot cylinder wall resp. valve areas as well as other exposed surfaces like the spark plug electrode. Furthermore, lube oil self-ignition in the pre-mixed air/ fuel charge is discussed. This phenomenon occurs especially in downsized and highly boosted engines at low speed and wide open throttle (WOT) conditions, where high compression temperatures and long timescales are present [3]. The examination of various lubricating oil detergents revealed that calcium can promote pre-ignition. This phenomenon was artificially reproduced in [4] and [5], where a standard fuel injector was used to insert oil into the combustion chamber of an engine. On the marine application side originators for pre-ignitions in gas/ dual-fuel engines are identified in [6] and [7]. Characteristic for this phenomena in such engines is that pre-ignitions occur sporadically within a limited crank angle range before TDC [8]. In highly efficient engines with high compression ratios the compression temperature is higher than the auto-ignition temperature of the lubricating oil (which differs not much from the auto-ignition temperature of diesel). This leads to evaporation and ignition of lubricating oil and, possibly, the further inflammation of the whole premixed air-fuel-charge. This only occurs in the limited crank angle range before TDC, because the charge reaches the oil self-ignition temperature at a certain angle before top dead center (TDC). 440 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="441"?> The experimental setup at the optically accessible engine test facility with the possibility of controlled lube oil addition allows in-depth examination of the underlying mechanisms of the phenomena. Since common fuel injectors are not suitable for this application because their hydraulic working principle relies on high injection pres‐ sures, special injectors had to be evaluated. They should allow for controlled lube oil addition even under engine relevant conditions in terms of pressures and temperatures. The influence of a variety of affecting parameters has been investigated - such as gas/ air charge composition, process gas temperatures and pressures, in-cylinder flow field and lube oil injection rate/ duration. Conclusions shall give extended insight into pre-ignition processes of premixed combustion, and the acquired reference data is used to validate and further develop corresponding simulation models. 2 Experimental Setup Fundamental investigations into pre-ignition phenomena based on data obtained from conventional engine test bench experiments is difficult since necessary information on in-cylinder flow field, evaporation and mixing processes, phase transitions and subsequent reactions are usually not known. Therefore, the optically accessible engine test facility “Flex-OeCoS” [9] is much more suitable for such experiments. Here, an optically accessible cylinder head is mounted on a motored engine block (figure 1, left). To study different combustion phenomena optical access is provided by four windows, of which the main windows have a diameter of 60 mm. Injectors can be mounted from the top (figure 1, right) or through one of the main resp. side windows by replacing them with special adapters. To facilitate premixed combustion the fuel gas (e.g. methane) is admitted into the intake ports and mixed with air by the turbulence during intake and compression stroke. Boundary conditions of the premixed charge like temperature, pressure and flow field have been studied excessively in previous investigations [10, 11], since they are essential to validate 0/ 1D models. In addition to the thermodynamic data, various optical measurement techniques can be applied as well. For the present work, simultaneous high-speed images of schlieren and OH* chemiluminescence with a resolution of 0.1 °CA were used. These images are analyzed by an adaption of the self-developed image-processing algorithm in terms of ignition location, start of ignition (SOIgn) and flame propagation. 441 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="442"?> Figure 1: The optical test rig consists of a Liebherr D944 engine block which is driven by an electric motor (left). On one cylinder an optically accessible cylinder head is mounted to investigate combustion phenomenon at engine relevant conditions (right) [9]. To study the phenomenon of lube oil triggered pre-ignition, small amounts or even single oil droplets must be injected into the combustion chamber during the compres‐ sion stroke. To achieve this, two types of injectors were either modified or developed for this purpose. For the addition of single oil droplets into the combustion chamber, the so-called piezo droplet injector (PZDI) has been developed. As the key element of the PZDI, the piezo stack from a standard GDI injector is used as actuator. The stack displaces the piston over the support plate having a centric pin as in the original GDI injector design to avoid bending stresses. Attached to the piston is a needle (⌀needle = 290 µm), which fits into the central bore (⌀bore = 300 µm) of the nozzle plate that is filled with oil. Applying a voltage to the stack leads to a length increase and therefore to a movement of the needle in the bore. The displaced oil is ejected through the nozzle bore, thereby forming single oil droplets. Disk springs provide the resetting force required to move the piston back to its initial position after the injection. Based on the experience gained from the injector tests in [12], a new cooling adapter was designed and manufactured by 3D printing (selective laser melting). With this technology, a unique design with cooling channels up to the tip could be realized. Furthermore, the nozzle plate geometry was slightly optimized to ensure proper functioning under engine-like conditions. This adapter is shown in figure 2 together with the mounted PZDI. The cooling channel geometry consists of two inlet and two outlet channels. The main purpose of the cooling water is to cool the nozzle plate with its oil reservoir to prevent the oil from evaporating into the combustion chamber since cylinder wall temperatures of up to 300°C are possible during combustion cycles. 442 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="443"?> Figure 2: PZDI mounted into the 3D printed cooling adapter for the Flex-OeCoS and detailed cooling channel geometry up to the nozzle plate of the PZDI. The second type of injector used in this investigation is a representative of the “next generation” piezo controlled common-rail injectors. This injector cannot produce single droplets, but is capable of reproducible injections of very small oil quantities over several cycles. “Normal” piezo injectors use the same hydraulic principle like solenoid injectors and therefore cannot cope with the high viscosity of lube oil. For this reason, an injector with a new innovative approach was chosen. This type uses a piezo stack that directly actuates the needle of the injector (direct acting instead of the servo concept). The original nozzle configuration of the injector consists of seven holes with 130 µm each at the circumference of the nozzle tip. The nozzles were converted into single hole nozzles by closing the original holes by laser welding and laser drilling of a new hole at the tip. Three different hole diameters were chosen (72, 101 and 205 µm, figure 3). Figure 3: The laser drilled holes in the nozzle, diameters (72, 101, 205 µm). 443 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="444"?> With this modified injector small amounts of lube oil were admitted into the optical combustion chamber during compression stroke, which leads to evaporation, subse‐ quent inflammation of the oil vapor and further propagation of the flame into the premixed methane-air charge. Optical images with schlieren and OH* chemilumines‐ cence for different air to fuel ratios are shown in figure 4. Here, a small amount of lube oil (injection duration = 75 µs, approx. 0.1 mg oil) is injected 90 °CA before TDC into the combustion chamber. During compression stroke, the oil droplets start to evaporate until self-ignition conditions are reached around 12 °CA before TDC. The resulting flame kernels lead to an inflammation of the whole charge for cases with lambda 2.0 or richer, whereas in cases with leaner mixtures the propagation is to slow and the flame of the lube oil is quenched. Figure 4: Inflammation of the lube oil in the optical combustion chamber of the Flex-OeCoS for different methane-air ratios. For lambda 2.0 an inflammation of the whole premixed charge can be observed whereas the flame for leaner mixtures extinguishes. (T comp ≈ 850 K, p comp ≈ 70 bar, SOI ≈ -90 °CA, m oil ≈ 0.1 mg) 3 Experimental Results Cylinder pressure traces (figure 5, left) and the calculated apparent heat release rate (figure 5, right) can be compared for different air to fuel ratios with identical lube oil admission parameters. The mean traces (curves) and the standard deviation (area) of the 25 cycles for each lambda are shown. A richer mixture results in an earlier and steeper increase in cylinder pressure and a higher maximum cylinder pressure. This is also reflected in the apparent heat release rate, which rises earlier and more 444 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="445"?> steeply for richer ratios due to the higher energy density of the mixture. Those results indicate the necessity of investigating this phenomenon since pre-ignitions can lead to both imperceptible resp. non-critical (lambda ≥ 1.75), or critical (lambda ≤ 1.50) events. Figure 5: Mean cylinder pressure traces (curves) for 25 cycles and the standard deviation (area) for different air to fuel ratios (left) and the respective apparent heat release rate (right). (T comp ≈ 800 K, p comp ≈ 70 bar, SOI ≈ -90 °CA, m oil ≈ 0.1 mg) Since the ignition of lube oil in a premixed charge of lambda 2.0 does not lead to a significant pressure rise before TDC (figure 5, left), the influence of the amount of lube oil was investigated in another experiment. For this, the energizing time of the lube oil injector was varied between 75 µs (i.e. 0.1 mg lube oil) and 1000 µs (1.5 mg oil) with a constant lambda of the premixed charge. Pressure traces for the cycles with a lube oil amount variation is shown in figure 6 (left). For the same engine relevant conditions, an increased lube oil amount triggers a pre-ignition with earlier and higher maximum cylinder pressure, which is also indicated in the apparent heat release rates (figure 6, right). 445 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="446"?> Figure 6: Comparison of different amount of injected lube oil into the combustion chamber with the same premixed air-methane ratio. More oil leads to an earlier and steeper pressure rise in the cylinder (left). The corresponding heat release rate (right) shows the same behavior. (T comp ≈ 800 K, p comp ≈ 70 bar, SOI ≈ -90 °CA, lambda ≈ 2.0) Furthermore, the influence of the lube oil injection time was investigated. Here, the lube oil was admitted into the combustion chamber between bottom and top dead center of the piston. This should artificially simulate different lube oil addition mechanisms like accumulated and blown-out oil of the piston crevice volume, or oil mist received from the intake port. The results are shown in figure 7 for an air-to-fuel ratio of 1.5 and a maximum compression temperature of 800 K. The start of ignition (SOIgn) is extracted from the optical OH* chemiluminescence images by a self-developed algorithm which provides the first self-ignition event in every cycle. For late injection timings (-20 °CA to 0 °CA), a reproducible ignition with a small standard deviation can be observed. This is comparable with a normal pilot fuel injection behavior. But for earlier injection timings, the deviation increases since the oil is injected before self-ignition conditions are reached. From -20 °CA to -140 °CA a nearly constant ignition position can be seen around 7.5 °CA. This represents the afore mentioned limit of pre-ignition timing for a fixed operation point (compression ratio, intake temperature, lambda). For earlier injections around bottom dead center (BDC), the pre-ignition timing is retarded, and the deviation increases due to different injection behavior and a better mixing of the lubricating oil with the charge (less fuel rich and therefore less ignitable zones). 446 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="447"?> Figure 7: The recognized start of ignition (SOIgn) for different start of injections (SOI). The mean value and the respective standard deviation were calculated from optical images of 25 combustion cycles with identical parameters. (T comp ≈ 800 K, p comp ≈ 70 bar, SOI ≈ -90 °CA, lambda ≈ 2.0) 4 Reaction Kinetic Investigation Reaction kinetics studies are performed to determine the ignition delay of the oil droplet. For the calculation the open source library Cantera [13] with the programming language Python is used. The model idea for the reaction kinetics is shown in figure 8. The black dot represents the liquid drop, the green cloud represents the oil vapor that forms around the liquid drop. 447 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="448"?> Figure 8: Reaction kinetics model idea The lubricating oil concentration in the environment varies depending on the distance to the liquid drop, therefore different oil concentrations must be expected in the investigated area. A 0D-reactor in Cantera is used to calculate ignition delays with boundary conditions based on internal combustion engine (ICE) parameters. In this work, n-hexadecane (n-C 16 H 34 ) was chosen as surrogate since sufficient agreement with the real lube oil can be expected [14]. Other work [15] shows a deviation of the ignition delay between n-C 16 H 34 and real lubricating oil of less than 10% (using unspecified commercially available lubricant base oil). For simulations in engine applications, this can be used as a basis for corrections if measurements in a real test bench are available. The POLIMI_TOT_NOx_1412 kinetic mechanism was chosen for the numerical simulation with n-C 16 H 34 as a lubricating oil surrogate. The mechanism contains a high and low temperature kinetic scheme and consists of 484 species and 19341 reactions [16-19] In comparison, a mechanism was tested which can be used up to C20 and approximately 7200 species, 31400 reactions [20]. While an ignition delay calculation with the Polimi mechanism takes a few seconds, the C20 mechanism takes several days. N-hexadecane seems to be a good compromise between accuracy and computing time. There are several approaches to determine ignition delays. In the code of this work four criteria were implemented, which can be seen in figure 9. Two of the criteria are temperature based and are reached when the temperature has a difference of 300 K and 400 K respectively to the initial temperature level. 448 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="449"?> Δ T300 Δ T400 Τ ign T400 Τ ign T300 Τ ign OH_grad Τ ign OH_Max Figure 9: Ignition criteria The two other criteria depend on the OH species. The first criterion of the two is defined by the gradient of the OH concentration the second criterion is based on the maximum OH concentration. The results from the four criteria hardly differ. The advantage of multiple criteria is that if one criterion fails, there are still other criteria available to get a result from the calculation. Furthermore, the calculation of multiple criteria influences the computing time minimally. An important result from the reaction kinetics studies can be seen in figure 10. Figure 10: Ignition delay lubricating oil variation at p=80 bar and λ=2.6 449 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="450"?> In the diagram, the ignition delay is shown logarithmically over the temperature. This type of representation was used because an Arrhenius equation which is given by: k = Ae −E a RT where k is the rate constant, T the absolute temperature, A the pre-exponential factor, E a the activation energy and R the universal gas constant in this representation means a straight line. The boundary conditions chosen are typical for a lean gas engine. Methane was selected as the fuel and the lubricating oil concentration was varied. The ignition delay reduces drastically even with small changes in the oil quantity. With expected end of compression temperatures between 700 °C and 850 °C, it is clear that preignition can only occur due to the influence of oil. At higher blending rates of oil, an NTC (negative temperature coefficient) behavior becomes observable which is common for long-chain hydrocarbons. As can be seen, between 800 K and 1000 K at 0.5% oil content, the ignition delay time does not increase any further, a plateau occurs. At 5% oil, the effect is even more pronounced - the ignition delay even increases as the temperature rises. In summary, it can be said that the oil concentration has a significant effect on the ignition delay time. Once a certain concentration has been reached, possible temperature inhomogeneities, such as those caused by residual gas, play a subordinate role in the reaction kinetics due to the NTC behavior. 5 Lube Oil Evaporation To calculate the evaporation of the oil drop, the method introduced by Pinheiro et. al. [21] was used. This method is based on the equations of Abramzon and Sirignano [22] which are known for the drop evaporation and are also used in many commercial simulation tools. The paper will not go into detail about the equations, these have already been presented in numerous publications. In the following section, the methodology used and implemented for droplet evaporation is discussed, and the added value for the pre-ignition model is shown. The evaporation model is implemented in python as part of the project Pre-ignition model [FVV Project #1394]. The physical properties for vapor and gas are also calculated using the open source package Cantera [13]. The boundary conditions for each time step are the gas composition which consists of oil vapor and ambient gas and the reference temperature and pressure. Furthermore, the diffusivity of the oil in the environment is calculated with Cantera. Evaporation enthalpy and properties of the liquid phase of the oil drop are determined from the NIST chemistry WebBook [23]. All thermodynamic data and transport properties for liquid phase, vapor phase and gas phase are assumed to be constant for each time step. The simulations with n-hexadecane were carried out in the subcritical range, i.e. the temperature of the 450 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="451"?> droplet must remain below 722 K. Also, the partial pressure of the droplet must not exceed the ambient pressure, otherwise boiling will take place instead of evaporation. 5.1 Validation of the evaporation model To validate the evaporation model, a droplet evaporation with n-heptane is calculated. To validate the model against measurement results, the measurement data from the work of Pinheiro et al [21] was used. The validation of the simulation against the measured values is performed with the boundary conditions shown in table 1. Initial droplet diameter D Start 500 μm Initial droplet temperature T Drop 300 K Ambient temperature T ∞ 623 K Ambient pressure p ∞ 101325 Pa Species N-heptane Tab. 1: Boundary conditions for the comparison between measurement and simulation Figure 11 shows the decrease in diameter D d due to evaporation relative to the initial diameter D d0 in comparison between the simulation and the measurement. The model fits the measured points very well. The evaporation follows the typical D 2 law, which represents a common behavior during the evaporation process. Since the calculation fits well with the simulation, the modeling can be used further on for the pre-ignition model. Figure 11: Evaporation model comparison between simulation and measurement [21] 451 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="452"?> 5.2 Evaporation of lube oil surrogate As already addressed in chapter 5, the species n-hexadecane is used as a lubricating oil surrogate. In the previously validated model, the data basis for the liquid phase is exchanged. Since it is the same source of the database [23] similarly good results are expected here. Good data basis is essential as it strongly affects the evaporation model. Figure 12 shows the data basis for the vapor pressure used for the model compared to data obtained by the Antoine equation with parameters of Camin, Forziati et al. [24]. Figure 12: Antoine equation vs. database [23] As can be seen, the curve obtained with the Antoine equation covers only a very small temperature range and deviates strongly from the database. The lubricating oil concentration depends directly on the vapor pressure and is described in the model with the following equation using Raoult’s law: Χ vs = p v p g where Χ vs is the surface vapor molar fraction, p v stands for the vapor partial pressure and p g the ambient gas total pressure. From the reaction kinetics studies in chapter 5, we see that even small changes in the mole fraction of the oil means a large effect on the ignition delay time in the gas phase of the oil droplet. 6 Pre-Ignition Model The pre-ignition model presented in the following section is a combination of the evaporation model and the reaction kinetic studies. The model is introduced and subsequently validated against measurements on the test bench. Since it is not known or can only be poorly estimated at which point in time the oil droplets enter the 452 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="453"?> combustion chamber, SOI (Start of injection) variations were carried out. In other words, a variation of the time at which the oil droplets are introduced into the combustion chamber which was described in chapter 4. 6.1 Model description based on an example An SOI of -50 °CA (Crank Angle) is selected as the starting point for the example calculation. A droplet diameter of 100 µm was chosen because this is a realistic value for the droplet size on the test rig. Investigations both in the simulation and on the test rig have shown that the droplet diameter has almost no effect on the timing of pre-ignition. This effect is advantageous for the model development, since the droplet diameter is difficult to determine on the test rig as well as on the full engine and represents an uncertainty. In figure 13, the droplet temperatures that occurs during evaporation is plotted over the crank angle at -50 °CA SOI. The temperature of the liquid drop T d is calculated with: m d c pl dT d dt = Q S Where m d is the droplet mass, c pl the specific heat capacity of the liquid droplet and Q S the thermal energy penetrating into the liquid phase. Due to the large gradients in the transition from liquid to gas phase, an averaging procedure is applied to calculate the temperature in droplet proximity. This temperature T m is calculated by T m = T d + α T g − T d Where T g is the temperature of the environment. For α, the value 1/ 3 recommended by Hubbard et al. [25] and Yuen et al. [26] was chosen and is an empirical value used mainly for spray combustion. The blue curve is the temperature of the liquid droplet, the orange one is at the phase transition near the liquid droplet and the green curve is in the oil vapor in the area where it ignites. 453 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="454"?> Figure 13: Temperature distribution during droplet evaporation In figure 14, the mole fraction Χ Oil is plotted against the crank angle. This is obtained from the formula presented in section 6.2. The vapor pressure p v is interpolated from the database values per time step as a function of temperature. Figure 14: Mole fraction of oil in the environment The lubricating oil concentration and temperature, in combination with the ignition delay diagram from section 1, can be used to determine when pre-ignition or ignition of the oil drop is to be expected. Figure 15 shows the combination of reaction kinetics and evaporation. The droplet is introduced into the combustion chamber at -50 °CA SOI, the mass of the droplet decreases over time. The evaporation is divided into individual time steps at which a calculation of the ignition delay time is started with the current boundary conditions, which consist of temperature, pressure and gas composition. 454 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="455"?> Figure 15: Ignition timing of the evaporating droplet The green dashed line represents the ignition delay from the reaction kinetics. The red stars represent the start of ignition (SOIgn). The minimum value of ignition represents the SOIgn for this SOI which is plotted in the comparison diagram between measurement and simulation in the following section. 6.2 Comparison of results between measurement and simulation In this section, the calculated SOIgn are compared with the measured values of the test bench. The ignition of the oil drops is evaluated optically as described in section 4. Figure 16 shows an SOI variation between bottom dead center (BDC) at -180 °CA and top dead center (TDC) at 0 °CA for one operation point. The compression curve is characterized by a maximum compression temperature of 750 K and a maximum compression pressure of 70 bar. The fuel used is methane, which is burned under stoichiometric conditions at λ = 1. For the measurement results the median is plotted which is shown as a blue curve. The median of the measurement results is used because it is less sensitive to variations than the average. The standard deviation (SD) per SOI are shown in dashed lines. 455 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="456"?> Figure 16: SOI variation and comparison between measurement (MMT) and simulation (SIM) at T = 800 K, p = 70 bar, n = 600 rpm, λ = 1.5 In the diagram, the red stars represent the minimum ignition points from the diagram in the previous section at the respective SOI. As can be seen, the simulated ignitions at the respective SOIs match the measurements very well. All points are within the range of the standard deviation of the measured values. Especially in the range below -60 °CA SOI the simulated values fit very well to the median from the measurement. This range is particularly important because above this SOI less oil input is to be expected. For an assessment of whether a pre-ignition is relevant or not, a complete SOI variation is normally not necessary. It is sufficient to calculate a few points in the middle range. Knowing when the regular ignition takes place and when a heat release starts, it can be judged whether the pre-ignition is relevant at the considered operating point. The relevance of a pre-ignition therefore does not depend on whether a pre-ignition takes place but when. It has been shown on the test rig that pre-ignition occurs as soon as oil is present. However, if the main ignition system precedes the pre-ignition, the pre-ignition is no longer critical. 7 Summary and conclusions The optically accessible engine test facility “Flex-OeCoS” [9] was introduced. For the addition of single oil droplets into the combustion chamber, the so-called piezo droplet injector (PZDI) was presented. With the combination of drop injector and test bench it was possible to produce optical images with schlieren and OH* - Chemiluminescence 456 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="457"?> for different air to fuel ratios. Cylinder pressure traces and the calculated apparent heat release rate were compared for different air to fuel ratios with identical lube oil admission parameters and for a fixed cylinder charge air to fuel ratio and a variation of lube oil injection parameters. A lube oil SOI variation was performed which was later used for the validation of the simulation. Reaction kinetic investigations were carried out. The core statement from these investigations is that pre-ignition can be ruled out for slow-running engines without the presence of lubricating oil. To calculate the temperatures and gas concentrations, a droplet evaporation model has been developed and validated against measured literature values. Exemplary at -50 °CA SOI the calculation method has been presented. The combination of the reaction kinetic studies and the droplet evaporation model constitute the pre-ignition model. With the pre-ignition model presented, the SOI variation measured on the test bench was compared with the computational results. The results of the simulation agree very well with the measured values. The developed pre-ignition model of this work can now be used to predict lubricating oil induced pre-ignitions in lean dual-fuel gas engines. 8 Acknowledgement The presented work is the scientific result of a research project “Modelling of Pre-ignition in Gas Engines” undertaken by the FVV (The Research Association for Combustion Engines eV). The research project was carried out in the framework of the industrial collective research program (IGF/ CORNET no. 257 EN). It was supported by the Federal Ministry for Economic Affairs and Energy (BMWi) through the AiF (German Federation of Industrial Research Associations eV) based on a decision taken by the German Bundestag. The project was was carried out in cooperation of the Institute of Automotive Engineering Stuttgart (IFS) at the University of Stuttgart under the direction of Prof. Dr.-Ing. Michael Bargende with the Institute of Thermaland Fluid-Engineering (ITFE) at the University of Applied Sciences and Arts Northwestern Switzerland (FHNW) under the direction of Prof. Dr.-Ing. Kai Herrmann. The authors gratefully acknowledge the support received from the FVV (Research Association for Combustion Engines eV), the working group that associates the research project initiated by Winterthur Gas & Diesel Ltd. and all others involved in the project. 9 References [1] Dahnz, C., Han, K.-M., Magar, M, “Vorentflammung bei Ottomotoren: Research Association on Combustion Engines e. V. (FVV) #931,” 2009. [2] Palaveev S., “Vorentflammung bei Ottomotoren II: Research Association on Combustion Engines e. V. (FVV) #1051,” 2013. 457 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="458"?> [3] Mayer, M., Hofmann, P., Williams, J., and Tong, D., “Influence of the Engine Oil on Pre-ig‐ nitions at Highly Supercharged Direct-injection Gasoline Engines,” MTZ Worldw 77(6): 36-41, 2016, doi: 10.1007/ s38313-016-0044-z. [4] Long, Y., Wang, Z., Qi, Y., Xiang, S. et al., “Effect of Oil and Gasoline Properties on Pre-Ignition and Super-Knock in a Thermal Research Engine (TRE) and an Optical Rapid Compression Machine (RCM),” SAE Technical Paper Series, SAE Technical Paper Series, SAE 2016 World Congress and Exhibition, APR. 12, 2016, SAE International400 Commonwealth Drive, Warrendale, PA, United States, 2016. [5] Welling, O., Moss, J., Williams, J., and Collings, N., “Measuring the Impact of Engine Oils and Fuels on Low-Speed Pre-Ignition in Downsized Engines,” SAE Int. J. Fuels Lubr. 7(1): 1-8, 2014, doi: 10.4271/ 2014-01-1219. [6] Imhof, D. and Takasaki, K., “VISUAL COMBUSTION RESEARCH USING THE RAPID COMPRESSION EXPANSION MACHINE,” MTZ ind 2(2): 28-39, 2012, doi: 10.1365/ s40353-012-0037-6. [7] Imhof, D., “Visual Combustion Studies for Environmental Friendly Marine Diesel and Gas Engines,” 2013. [8] “ASME 2012 Internal Combustion Engine Division Spring Technical Conference,” ASME 2012 Internal Combustion Engine Division Spring Technical Conference, Torino, Piemonte, Italy, 06.05.2012 - 09.05.2012, American Society of Mechanical Engineers, ISBN 978-0-7918-4466-3, 05062012. [9] Schneider, B., Schürch, C., Boulouchos, K., Herzig, S. et al., “The Flex-OeCoS—a Novel Optically Accessible Test Rig for the Investigation of Advanced Combustion Processes under Engine-Like Conditions,” Energies 13(7): 1794, 2020, doi: 10.3390/ en13071794. [10] Wüthrich, S., Humair, D., and Herrmann, K., “Enhanced instrumentation of an optical research engine with unique combustion chamber”, 14th Int. AVL Symposium on Propulsion Diagnostics,” 2020. [11] Humair, D., Cartier, P., Süess, P., Wüthrich, S. et al., “Characterization of dual-fuel combustion processes,” 2020. [12] Süess, P., Schneider, B., Wüthrich, D., Wüthrich, S. et al., “A Specifically Designed Injector for Controlled Lube Oil Addition in View of Investigation of Pre-Ignition Phenomena in Dual-Fuel/ Gas Engines,” Front. Mech. Eng. 7, 2021, doi: 10.3389/ fmech.2021.623896. [13] Goodwin, D.G., Speth, R.L., Moffat, H.K., and Weber, B.W., Cantera: An Object-oriented Software Toolkit for Chemical Kinetics, Thermodynamics, and Transport Processes, Zenodo, 2021. [14] Distaso, E., Amirante, R., Calò, G., Palma, P. de et al., “Investigation of Lubricant Oil influence on Ignition of Gasoline-like Fuels by a Detailed Reaction Mechanism,” Energy Procedia 148: 663-670, 2018, doi: 10.1016/ j.egypro.2018.08.155. [15] Ohtomo, M., Miyagawa, H., Koike, M., Yokoo, N. et al., “Pre-Ignition of Gasoline-Air Mixture Triggered by a Lubricant Oil Droplet,” SAE Int. J. Fuels Lubr. 7(3): 673-682, 2014, doi: 10.4271/ 2014-01-2627. 458 L. Wißmann, P. Süess, M. Grill, K. Herrmann and M. Bargende <?page no="459"?> [16] Cuoci, A., Frassoldati, A., Faravelli, T., and Ranzi, E., “Formation of soot and nitrogen oxides in unsteady counterflow diffusion flames,” Combustion and Flame 156(10): 2010-2022, 2009, doi: 10.1016/ j.combustflame.2009.06.023. [17] Faravelli, T., “Kinetic modeling of the interactions between NO and hydrocarbons in the oxidation of hydrocarbons at low temperatures,” Combustion and Flame 132(1-2): 188-207, 2003, doi: 10.1016/ S0010-2180(02)00437-6. [18] Frassoldati, A., Faravelli, T., and Ranzi, E., “Kinetic modeling of the interactions between NO and hydrocarbons at high temperature,” Combustion and Flame 135(1-2): 97-112, 2003, doi: 10.1016/ S0010-2180(03)00152-4. [19] Ranzi, E., Frassoldati, A., Grana, R., Cuoci, A. et al., “Hierarchical and comparative kinetic modeling of laminar flame speeds of hydrocarbon and oxygenated fuels,” Progress in Energy and Combustion Science 38(4): 468-501, 2012, doi: 10.1016/ j.pecs.2012.03.004. [20] Sarathy, S. M., Westbrook, C.K., Mehl, M., Pitz, W.J. et al., “Comprehensive chemical kinetic modeling of the oxidation of 2-methylalkanes from C7 to C20,” Combustion and Flame 158(12): 2338-2357, 2011, doi: 10.1016/ j.combustflame.2011.05.007. [21] Pinheiro, A.P. and Vedovoto, J.M., “Evaluation of Droplet Evaporation Models and the Incorporation of Natural Convection Effects,” Flow Turbulence Combust 102(3): 537-558, 2019, doi: 10.1007/ s10494-018-9973-8. [22] Abramzon, B. and Sirignano, W.A., “Droplet vaporization model for spray combus‐ tion calculations,” International Journal of Heat and Mass Transfer 32(9): 1605-1618, 1989, doi: 10.1016/ 0017-9310(89)90043-4. [23] Linstrom, P.J. and Mallard, W.G., “The NIST Chemistry WebBook: A Chemical Data Resource on the Internet,” J. Chem. Eng. Data 46(5): 1059-1063, 2001, doi: 10.1021/ je000236i. [24] Camin, D.L., Forziati, A.F., and Rossini, F.D., “Physical Properties of n-Hexadecane, n-De‐ cylcyclopentane, n-Decylcyclohexane, 1-Hexadecene and n-Decylbenzene,” J. Phys. Chem. 58(5): 440-442, 1954, doi: 10.1021/ j150515a015. [25] Hubbard, G.L., Denny, V.E., and Mills, A.F., “Droplet evaporation: Effects of transients and variable properties,” International Journal of Heat and Mass Transfer 18(9): 1003-1008, 1975, doi: 10.1016/ 0017-9310(75)90217-3. [26] YUEN, M.C. and CHEN, L.W., “On Drag of Evaporating Liquid Droplets,” Combustion Science and Technology 14(4-6): 147-154, 1976, doi: 10.1080/ 00102207608547524. 459 Development of a Predictive 0/ 1D Model for Lubricating Oil Induced Pre-Ignitions at an Optical Gas/ Dual-Fuel Engine. <?page no="461"?> Injection during compression stroke for engine knock prevention Michael Wörner, Michael Auerbach, Gregor Rottenkolber Esslingen University of Applied Sciences, Esslingen, Germany Abstract: Lower charge temperature and faster combustion fundamentally help to mitigate knock. Late fuel injection during compression stroke allows both, cooling of the charge as well as increased turbulence levels. Furthermore, very late injection reduces fuel retention time and hence radical formation. In this work, late high-pressure injection was applied to stoichiometric and lean combustion processes. An increased compression ratio compared with early intake valve closing addresses future engine concepts. In this context, the momentum of late fuel injection increases the turbulence level at the end of compression and compensates the drawback of early intake valve closing. As a result, injection within a certain time span during compression stroke enables homogenous stoichiometric combustion. This leads to improved high load performance operating with high geometric compression ratios and Miller cycle. Furthermore, a lean stratified combustion process combined with late injection close to spark angle was investigated. This strategy allows further improvement of engine efficiency and knock free combustion at moderately charged engine operating points. 1 Introduction The injection of fuel very late during compression stroke constitutes a good measure to prevent knock in gasoline engines. Already several decades ago, stratified combustion processes with knock mitigation were published, e.g. Texaco combustion process [1, 2]. Current studies proclaim fuel injection short before top dead centre (TDC) as the only possibility to avoid auto-ignition completely [3]. The major benefit of such strategies is the reduced retention time of the fuel at high temperatures, which reduces radical formation at critical areas of the combustion chamber [4, 5]. Additionally, the turbulence introduced by the spray accelerates the turbulent flame. The fast combustion supports knock prevention and reduces cycle-to-cycle variations. Sharp improvement of combustion phasing is possible resulting in reduced combustion losses [3, 5]. <?page no="462"?> However, highly increased turbulence levels and suppression of radical formation require injection timings very late in the compression stroke. This in turn requires very high fuel pressures, well-designed sprays and combustion chambers [5]. Kaminaga et al. [4] determined fuel post oxidation in the exhaust pipe at very late injection timings. The authors trace it back on poor mixture quality and fuel on the piston surface. However, reduction of these losses are shown. Kapus et al. [6] point out very low CO and particulate emissions for late injection. Additionally to the very late injections, fuel introduction during early compres‐ sion stroke enables increased mixture cooling [5, 7, 8]. Contrarily, very late injection timings do not increase mixture cooling due to compression of air with lower specific heat capacity than air-fuel mixture. Still, at injection during early compression increased fuel pressures are necessary for good homogeneity and low particulate matter [5, 8]. The work in this paper focusses on mixture preparation phenomena of high-pressure injection during compression stroke. Thermodynamic analysis on a single cylinder engine, emission measurement and video endoscopic optical investigations are applied for a deeper understanding of different injection strategies. Moreover, the interaction of such injection strategies with early intake valve closing (IVC) and an increased geometric compression ratio is investigated. The additional turbulence of the late injection addresses the reduced charge motion of the Miller cycle. Finally, a lean strategy focusses on efficiency improvement without high amount of soot at very late injection. 2 Engine setup The gasoline-direct-injection (GDI) engine has a centre-mounted injector. The spark plug location is between the exhaust valves. Table 1 contains the engine specifications. The two different geometric compression ratios were realised with the same piston (central bowl shape) by different clearance heights. Figure 1 illustrates the used valve lift profiles. The declaration numbers describe the valve opening time (VOT) in degrees crank angle (°CA) at 2 mm valve lift. The tumble flap mentioned in table 1 closes the lower half of the inlet port inside the cylinder head. It was used optionally in combination with the VOT125 to increase in-cylinder charge motion. Number of cylinders 1 Number of valves 4 Displacement 497.8 cm 3 Inlet valve diameter 30.5 mm Exhaust valve diameter 24 mm 462 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="463"?> Stroke / Bore 92 mm / 83 mm Compression ratio 12.7: 1 | 14: 1 Charge motion system Tumble flap (optional) Tab. 1: Engine specifications Fig. 1: Valve lift profiles 3 Assessment of knock The knock assessment in this work bases on the analysis of the in-cylinder pressure signal. The applied criterion is the KRAT (knock ratio), which is included in the used AVL IndiCom software and bases on the Siemens-VDO algorithm, see e.g. Scharlipp [9]. The standard parameterisation was slightly adjusted, what will be justified in the following. 3.1 Cylinder pressure oscillations Usually, if knock occurs in gasoline engines, the detonation induces oscillations to the gas volume in the cylinder. The recording of these oscillations within the measured pressure signal make knock analysis feasible. Figure 2 illustrates on the left side the resulting signals of knock free (top) and knocking combustion (bottom). It is known, that measurements with a pre-chamber ignition system can include similar pressure oscillations, but auto-ignition does not occur [10]. The mid diagrams of figure 2 point out such a phenomenon. Additionally, similar impacts were found in this work for combustion processes with late high-pressure injection. The right diagrams of figure 2 show two exemplary cycles, which are comparable to the findings of the pre-chamber ignition. Hence, an appropriate knocking criterion is necessary. 463 Injection during compression stroke for engine knock prevention <?page no="464"?> Fig. 2: Not knocking (top) and knocking (bottom) combustion events with standard spark plug (left), passive pre-chamber (mid) and late high pressure injection (right) Therefore, figure 3 includes further signal analysis. The diagrams show the standar‐ dised frequency spectra of the before discussed cycles. The spectrum of a knock free combustion (green line in the left diagram) did not include any frequencies above 4 kHz. Contrarily, regular knocking caused frequency content in segments of natural frequencies of the charge volume. For both, pre-chamber ignition and late high-pres‐ sure injection, the oscillations typically occurred only in the first natural frequency (around 7 kHz), if there was no auto-ignition detected. However, if auto-ignition occurred induced frequencies above the first natural frequency were determined. Hence, a high-pass filter with cut-off frequency at 10 kHz reduced the oscillations, which were not induced by knock. This measure also reduced the signal content in general, which had to be considered for knocking limit values. Additionally to the changed cut-off frequency from 4 to 10 kHz, the mentioned criterion based on the Siemens-VDO algorithm was chosen. The root of this matter is a split of the signal in two sections before and after maximum cylinder-pressure providing further noise cancelling. 464 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="465"?> Fig. 3: Frequency spectra of before discussed pressure signals 3.2 Test bench real time criterion Considering noise emission and damage potential, single cycles with high knocking values are relevant. Since engine knock occurrence is a stochastic phenomenon, maxi‐ mum values differ from mean values considering a certain quantity of measured cycles. This in turn requires statistical observation and hence, high number of measurement cycles as well as statistical postprocessing. The left side of figure 4 illustrates the differences in the arithmetic mean values of two measurement points with nearly equal 95% quantile. Obviously, the deviation of the mean values result from different distribution functions. Regarding the high number of measurement points within the test programme, a criterion for knock calibration directly on the test bench is required. A real time criterion predicting the correct statistical values had to be found. The diagram on the right side of figure 4 shows knock values of three different spark angles (SA) of two engine setups. Each measurement point consists 10.000 measured cycles. The arithmetic mean (dashed lines) pointed out again, that there was no consistent relationship compared with statistical 95% quantiles. However, behaviour that is more consistent was found with the following equation: K RAT _cal = x(K RAT ) + 3 * σ (K RAT ) The equation bases on the mathematical approach for quantiles of a normal distribu‐ tion. Taking the arithmetic mean value x(KRAT) plus three times the standard deviation σ(KRAT), sufficient accordance with the quantile was found (solid lines). This allowed simple knocking assessment for steady state operation directly on the test bench. Therefore, arithmetic mean and standard deviation of KRAT was calculated in real time from a moving window of 50 cycles. 465 Injection during compression stroke for engine knock prevention <?page no="466"?> Fig. 4: Statistical knock assessment 4 Stoichiometric combustion The investigations of this section were performed at a steady operating point of 1250 rpm and an IMEP n of 15 bar. The compression ratio was 12.7: 1 and valve lift profiles as described in figure 1. Fuel pressure was set constant at 1000 bar and intake air temperature at 40±2 °C. 4.1 Charge motion of the intake system The premixed flame of a gasoline engine is highly affected by the air charge motion. Therefore, the tumble flow prevailed as the most effective way to increase turbulence level at the end of compression. Hence, flame propagation gets highly turbulent and accelerated. Figure 5 shows the tumble ratio and the turbulent kinetic energy (TKE) from 3D-CFD simulation of the inlet valve lift profile VOT125 without and with adjusted tumble flap in the inlet port (marked with a T). Evident was the effect of a Miller cycle, which did not lead to any tumble flow and hence, no increased flow velocity during late compression. This in turn was followed by low TKE. On the other hand, the directed flow during intake stroke with the tumble flap increased the tumble ratio and TKE during late compression massively. A measurement of local flow around the spark plug with a voltage rise anemometry (VRA) confirmed this behaviour, see Wörner and Rottenkolber [11]. The resulting faster flame propagation due to higher TKE with the tumble flap was notable for injection during intake stroke. Figure 6 shows the results plotted over the electric start of injection (SOI), where the reference TDC is always TDC firing. Hence, SOI = -280 °CA a. TDC (below shortened with SOI -280) represents injection during intake stroke. The configuration with the tumble flap showed a shorter burn delay and especially a faster main combustion. Additionally, the smaller valve lift profile (VOT110), which means enhanced Miller cycle and thereby even lower TKE, extended the burn duration further. Contrary behaviour resulted at injection during compression stroke. Burn delay and burn duration of the three configurations showed almost no deviation for same injection timings. This in fact means there was no influence of 466 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="467"?> Fig. 5: 3D-CFD simulation of tumble ratio and TKE of inlet valve lift profile VOT125 without and with tumble flap the tumble flow on flame speed, if high-pressure injection was introduced during compression stroke. Fig. 6: Influence of injection timing and inlet configuration on burn delay and burn duration Additionally striking were some injection timings around SOI -90 and -45, where combustion speed dropped down. Further, combustion stability, which is represented by the COV IMEP in figure 7, confirmed this behaviour. Cycle-to-cycle variations increased for all inlet configurations at the mentioned injection timings. Additionally, CO emissions (figure 7) increased strongly. Most likely explanation is an adverse interaction of spray and piston at these SOIs. Thereby, lower turbulence and reduced homogenisation resulted. This led to lower combustion speed and higher CO emissions. However, the COV IMEP at injection during intake stroke proofs the improved air charge motion with the tumble flap. This also was found for CO emissions, which means improved homogeneity with higher air charge motion for injection timings during intake stroke. 467 Injection during compression stroke for engine knock prevention <?page no="468"?> Fig. 7: Influence of injection timing and inlet configuration on CO emission and combustion stability COV IMEP 4.2 Limitation of very late injection In general, greatest benefit was assumed for very late injection timings. In fact, SOI later than -40 °CA a. TDC clearly improved the phasing of the centre of combustion (MFB50) as the reduced radical formation improved knock mitigation (figure 8). Note that the general benefit of the VOT110 for all injections during compression stroke can be drawn back on reduced temperatures due to increased Miller cycle. Furthermore, combustion speed (figure 6) showed strong improvements for very late fuel injection. Especially burn delay demonstrated clearly an improvement of shifting the injection towards TDC due to higher turbulence levels. Thus, suppression of knock was supported further. However, high levels of CO emissions (figure 7) indicated an inhomogeneous charge. Additionally, strongly rising soot emissions, represented by the filter smoke number (FSN) in figure 8, point to mixing-controlled diffusion flames at very late injections. A wide range of injection timings in the compression stroke showed just small increased FSN compared to intake stroke injection. Video endoscopy pointed out soot formation located at the injector tip for both injection strategies. Contrarily, at very late injection rich mixture and not completely evaporated fuel led to strong soot formation within large areas of the combustion chamber. Hence, successful improvement of combustion phasing and flame speed was not possible without drawbacks on emissions. Figure 9 on the left side shows the indicated specific fuel consumption (ISFC) of the VOT125 configuration. Obviously, very late injection did not show any general efficiency benefit. However, injection in the range of -75 to -65 °CA a. TDC pointed out good overall performance and best efficiency. Good homogeneity, improved charge cooling and TKE levels for moderate combustion speed were combined in this area and hence enabled good performance with Miller cycle and increased geometric compression ratio. To enhance understanding of efficiency performance, a loss distribution is added to figure 9 for three injection timings. Intake injection timing (-280), best efficiency (-75) and latest injection with stable engine performance (-15). Obviously, the main benefit 468 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="469"?> Fig. 8: Influence of injection timing and inlet configuration on MFB50 and soot emission of the best efficiency point at SOI -75 are less combustion losses because of knock mitigation. Note that this cannot be rated as a fuel consumption ‘potential’ since no effort was put into an optimisation of intake injection. However, noticeable is the combustion loss of SOI -15, which was equal to SOI -75, although MFB50 and burn duration 10 - 90% was improved. The heat release curves in figure 10 further support understanding. The rate of heat release (RoHR) of the very late injection (dashed line) demonstrates the very fast combustion. However, the end of combustion was very slow and continued until exhaust valves opened, which the mass fuel burned (MFB) curve demonstrates clearly. Until 90% of energy release, the combustion was fast and combustion losses could be reduced. However, the final phase of combustion increased the losses again. The solid graphs of the SOI -75 illustrate a slower combustion. However, complete burn through of the charge was reached far earlier. The inhomogeneous charge at the very late injection led to burning of rich mixture and hence high amount of oxidation during late expansion. Fig. 9: Indicated specific fuel consumption and deviation of losses for three individual measurement points There was an additional effect determined, considering the total amount of the loss distribution of SOI -15 in figure 9. Note that the exhaust heat was calculated as the 469 Injection during compression stroke for engine knock prevention <?page no="470"?> theoretical optimum Otto cycle with isentropic exponent κ = 1.4 and therefore only depending on geometric compression ratio. Hence, the error of the total balance of SOI -15 proved post oxidation of fuel between exhaust valve and sampling point for exhaust gas measurement (CO and HC emission). In fact, the unburned fuel losses would be greater than measured, what confirmed findings of Kaminaga et al. [4]. As a result, with the used engine configuration, very late injection did not lead to any efficiency improvement due to bad mixture formation. Furthermore, described results were performed at low rpm, whereas higher engine speed made very late injection even more difficult due to reduced mixture preparation time. Fig. 10: Rate of heat release and standardised accumulated heat release of two injection timings: SOI = -75 (solid line) and SOI -15 (dashed line) However, the derived optimum injection timing for stoichiometric premixed combus‐ tion process was at SOI -75. The combination of good mixture formation, charge cooling and moderate increased turbulence levels gave advantage to injection timings in this area. Thus, similar high load performance with increased compression ratio of 14: 1 compared to a baseline configuration with a compression ratio of 11: 1 were performed. As an additional benefit, early intake valve closing and high compression ratio increased part load efficiency. 5 Stratified lean combustion Demonstrated results in the previous section consequently led to an application of a split injection strategy: A first injection during early compression in the known range for homogenous premixed charge and a second late injection close to spark angle to enhance turbulence level. Additionally, this late injection led to a lean premixed charge and therefore less radical formation. The following figure 11 shows a cylinder pressure trace at 2000 rpm and 12 bar IMEP n . Knock free combustion at MFB50 = 8 °CA a. TDC resulted. Note that the same piston as for previous investigations was used. With a 470 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="471"?> reduced clearance height, the compression ratio could be increased to almost 14: 1. On the right side of figure 11, the schematic drawing explains calibration and evaluation parameters. Fig. 11: Cylinder pressure of stratified combustion strategy and designation of electric and combustion parameters Firstly, fundamental investigations of split injection strategy were performed, e.g. variation of quantity and timing of the second injection. It was found that a high amount in the second injection was necessary for fast combustion process. Almost equal distribution of both injection durations resulted. Furthermore, spark angle had to be located close to the injection to avoid too much premixing and radical formation. However, increased particulate emissions were unavoidable due to high fuel quantity in the second injection. To improve soot oxidation processes an overall lean operation was taken into account. Figure 12 illustrates a variation of air-fuel ratio λ. Independent of air-fuel ratio, knock free combustion was possible at this operating point. The filter smoke number dropped massively with higher air excess. At λ > 1.6, almost soot free combustion could be realized. Furthermore, due to higher isentropic exponent and less incomplete combustion fuel consumption decreased as well. At λ = 1.8, misfire of single cycles occurred. Furthermore, load and speed of the engine was varied. Figure 13 illustrates the time span between electric start of the second injection (SOI2) and 2% mass fraction burned from heat release calculation (MFB2). Subsequently it is called ‘ignition delay’. The coloured lines show different engine speeds, whereas the left diagram demonstrates the known variation over air-fuel ratio and the right one a variation of the time between SOI2 and spark angle. The left diagram of figure 13 shows clearly that there was no influence of engine speed on the ignition process and the early combustion phase. Moreover, global air-fuel ratio did not affect ignition delay either. The second injection very close to spark angle led to steady mixture and turbulence levels around the spark plug and hence 471 Injection during compression stroke for engine knock prevention <?page no="472"?> Fig. 12: Indicated fuel consumption and soot emission depending on air-fuel ratio steady inflammation conditions. The 1.25 ms include the time until fuel reaches the side-mounted spark plug as well as the mixing processes and the start of combustion until 2% MFB. Variations of the time between SOI2 and SA at lean mixture of λ = 1.6 (right diagram of figure 13), show similar results. For a large time span, ignition delay increased as spark was too late after injection. This led to a longer time of premixing and knock tendency increased, which will be discussed subsequently. However, the possible range of Δα(SOI2-SA) for stable combustion decreased with increased engine speed. Due to reduced time scales, this combustion process was limited by engine speed. Fig. 13: Ignition delay depending on air-fuel ratio (left) and depending on time span between second injection and spark angle (right) Figure 14 supports understanding of the behaviour at different time spans between injection and spark angle. Note that in this case engine speed of 1250 rpm was used for a wide range of Δα(SOI2-SA). The diagrams show the RoHR of hundred single cycles and the averaged line for three combinations of second injection and spark angle. 472 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="473"?> Fig. 14: Rate of heat release for different time spans between second injection and spark angle Very early SA short before SOI2 (left diagram) led to fast combustion. The later the spark was located after injection, the slower the combustion process started, which most likely resulted from fast dropping turbulence level after end of injection. In every case, a second steep rise of RoHR occurred, whereby the tendency to knock increased for larger Δα(SOI2-SA). In this case, increased mixing time of the second injection and an extended early combustion phase resulted. However, negative Δα(SOI2-SA) led to higher amount of particulate emissions due to ignition during injection period. The following figures 15 and 16 show video endoscopy recordings of a negative Δα(SOI2-SA) and a positive Δα(SOI2-SA). Fig. 15: Video endoscopy recordings of combustion process with spark angle before second injection (negative Δα(SOI2-SA)) at 1250 rpm Obviously, the very bright lightening diffusive flame points to massive soot formation. However, in figure 16, diffusive combustion only occurred around the spark plug where the combustion process started. At the other regions of the combustion chamber, 473 Injection during compression stroke for engine knock prevention <?page no="474"?> premixed combustion occurred. In this area, soot emissions were lowered, but knock tendency increased. Fig. 16: Video endoscopy recordings of combustion process with spark angle after second injection (positive Δα(SOI2-SA)) at 1250 rpm These findings helped to improve the calibration of the two injections and the spark angle. However, the limitations of this combustion process became obvious as well. On the one side, increased soot formation and on the other side higher knock tendency led to a narrow range of the time span between second injection and spark angle. Hence, full load was not possible by this combustion process as retarded combustion phasing to reduce engine knock resulted in bad combustion stability. Very lean mixture near the walls of the combustion chamber limited this late combustion phasing. Furthermore, at increased engine speed, the range of possible timings of second injection and spark angle became smaller. Nevertheless, figure 17 shows investigations of external cooled exhaust gas recircu‐ lation (EGR) at 12 bar IMEP n . Due to lean combustion and the related problems with exhaust gas aftertreatment, the reduction of nitrogen oxide emissions was the major objective of applying EGR. Additionally, soot emissions and fuel consumption were focussed. However, in the varied range just small drawbacks in FSN and ISFC were found. Indicated efficiencies of up to 44% were performed. Limitation of the EGR rate was found at 25% roughly. This was due to increased cycle-to-cycle variations and therefore occurrence of single cycles with high knocking values. Therefore, a third injection during early combustion (red triangle markers in figure 17) was introduced. Slightly improved COV IMEP was notable. However, soot and CO emissions and therefore ISFC increased. Inversely to the variation of global air-fuel ratio (see figure 13), ignition delay Δα(SOI2-MFB2) increased with higher EGR rates due of the inertia of the residual gas. 474 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="475"?> Fig. 17: Variation of external cooled EGR rate 6 Conclusion A comprehensive study of injection during compression stroke with increased com‐ pression ratio was performed. A single injection strategy was used for stoichiometric combustion processes. Thereby, injection timings during early compression stroke showed best overall performance. Forming a homogenous mixture by high injection pressures, low CO and particulate emissions were possible. One key finding was the independence of combustion speed from intake air motion as the momentum of high-pressure fuel injection during compression stroke dominates the turbulence level during combustion. Hence, a Miller cycle with massively reduced charge motion was investigated. The Miller cycle in combination with an increased geometric compression ratio of 14: 1 and compression stroke injection showed similar high load performance as the baseline engine setup without Miller valve timings, a compression ratio of 11: 1 and intake stroke injection. Therefore, part load improvements can be realized. However, very late injection timing close to TDC did not show enhanced efficiency due to very high amount of unburned fuel and soot emissions. Hence, a stratified lean combustion process with the combination of a first injection during early compression and a second late injection close to spark angle was investigated. Knock free combustion with a compression ratio of 14: 1 was possible up to an IMEP n of 13 bar and significant efficiency improvements were performed. Nevertheless, occurrence of 475 Injection during compression stroke for engine knock prevention <?page no="476"?> knock restricted high engine load. Furthermore, limitations by engine speed due to a narrow window for second injection timing and spark angle were determined. Thus, operating range of this strategy is clearly limited. However, reduction of nitrogen oxides with external cooled EGR was possible without immense drawback on efficiency and particulate emissions. References [1] E.M. Barber, B. Reynolds, W.T. Tierney: Elimination of combustion knock-Texaco combustion process, SAE Technical Paper 510173, 1951. [2] J.M. Lewis, W.T. Tierney: Development of a Stratified-Charge Engine with Broad Fuel Tolerance, Proceedings of the First International Automotive Fuel Economy Research Conference, 285-299, 1981. [3] H. Friedl, G. Fraidl, P. Kapus: Highest efficiency and ultra low emission - internal combustion engine 4.0, Combustion Engines, 180(1), 8-16, 2020. [4] T. Kaminaga, K. Yamaguchi, S. Ratnak, J. Kusaka, T. Youso, T. Fujikawa, M. Yamakawa: A Study on Combustion Characteristics of a High Compression Ratio SI Engine with High Pressure Gasoline Injection, SAE Technical Paper 2019-24-0106, 2019. [5] P. Richardson, G. Jiayi, G. Di Liberto: Knock Prevention by Retarded Injection with Ultra-High Pressure and Fuel Injector Nozzle Development, 28th Aachen Colloquium Automobile and Engine Technology, Aachen, 2019. [6] P. Kapus, M. Certic, M. Neubauer, K. Prevedel, T. Schicker: New Gasoline Combustion Systems for Highest Efficiency and Lowest Emission, 17th Conference The Working Process of the Internal Combustion Engine: Sustainable Mobility, Transport and Power Generation, Graz, 2019. [7] T. Tabata, H. Shibata, H. Katsurahara, Y. Konishi: Evolution of Versatile Gasoline ICE Core system, Considering Future Electrification, 40. Internationales Wiener Motorensymposium, Wien, 2019. [8] G. Rösel, E. Achleitner, F. Graf, P. Rodatz, P. Senft, R. Brück, H. Stock: Lowest Real Driving Emissions: Solutions for Electrified Gasoline Engines, 28th Aachen Colloquium Automobile and Engine Technology, Aachen, 2019. [9] S. Scharlipp: Untersuchung des Klopfverhaltens methanbasierter Kraftstoffe, Dissertation, University of Stuttgart, 2017. [10] M. Sens, E. Binder: Vorkammerzündung als Schlüsseltechnologie für einen zukünftigen Antriebsstrang-Mix, MTZ - Motortechnische Zeitschrift, 80, 46-53, 2019. [11] M. Wörner, G. Rottenkolber: Voltage rise anemometry in turbulent flows applied to internal combustion engines, Experiments in Fluids 62(6), 2021. 476 Michael Wörner, Michael Auerbach, Gregor Rottenkolber <?page no="477"?> Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends Sascha Holzberger, Maurice Kettner, Karlsruhe University of Applied Sciences Roland Kirchberger, Graz University of Technology Ivica Kraljevic, Florian Sobek, Fraunhofer Institute for Chemical Technology Abstract: To further reduce CO 2 emissions from internal combustion engines, either engine efficiency can be improved or fuels can be used that are produced from renewable energies and are thus CO 2 -neutral e.g. methanol or ethanol. An increase in engine efficiency can be achieved by charge dilution. At the Karlsruhe University of Applied Sciences (HKA) a pre-chamber spark plug with an integrated controllable hot surface has been developed, called Hot Surface Assisted Spark Ignition (HSASI) which should enable higher dilution ratios and thus higher engine efficiency. A conventional glow plug that protrudes into the pre-chamber serves as the hot surface. With the help of the glow plug a higher temperature is achieved in the pre-chamber which promotes the combustion in the pre-chamber and increases the possible dilution ratio. In this work the applicability of the HSASI ignition element is investigated when using methanol and ethanol blends, specifically M25 and E85. M25 consists of 25 vol% methanol and 75 vol% RON98, while E85 contains 85 vol% ethanol and 15 vol% RON98. The experimental investigations focus on part load when dilution levels for passive pre-chamber spark plugs are impeded by flame quenching and misfire due to a cold pre-chamber. Engine test were carried out on a single-cylinder engine at the HKA. The HSASI operating mode with an active glow plug was compared with the passive pre-chamber operating mode with a non-active glow plug. The flame development angle was determined for different ignition timings and a constant air fuel equivalence ratio. The HSASI operating mode shows shorter flame development angles for both fuels with up to 3 °CA for E85 at an engine speed of 1800 rpm. Therefore, it allows for later ignition timing for E85 and M25 at a constant center of combustion. 1 Introduction Despite the increasingly stringent legal regulations regarding the sale and operation of combustion engines in passenger cars, they will continue to play a key role in mobility in the coming decades. After a slight decline in new car registrations worldwide in the <?page no="478"?> years 2017 to 2020, an increase is expected for the next few years [1]. Drivetrains with an internal combustion engine, i.e. PHEVs, HEVs, MHEVs and conventional drivetrains with only gasoline and diesel engines, will account for the majority of new registrations at around 80% in 2030 [2]. In addition to the new registrations, there are also the vehicles from the worldwide existing fleet of approximately 1.3 billion passenger cars [3]. The applications of the internal combustion engine in commercial vehicles and in maritime use are not even taken into account here. If the envisaged legal restrictions in different regions of the world are considered, it also becomes clear that so far only regions that account for about 15% of the passenger car sales market have envisaged a ban on the sale of internal combustion engines by 2040 [4]. This does not include the planned sales ban in the EU [5]. Against the background of the above and the goal of minimising CO 2 in the transport sector, it is essential to reduce CO 2 emissions of combustion engines. This goal can be achieved on the one hand by increasing the efficiency of combustion engines and on the other hand by using alternative, CO 2 -neutral fuels. The latter is also regarded by the German government as an essential component in achieving the climate goals, i.e. the targeted climate neutrality by 2045 [6, 7]. In addition to the goal of climate neutrality, the drivers for the production of alternative fuels are the reduction of dependence on limited fossil energy sources and the improvement of efficiency and emission formation during engine combustion. Two possible alternative fuels are the alcohols methanol and ethanol [8]. These can be produced in a CO 2 -neutral way and reduce CO 2 emissions from internal combustion engines either as a pure fuel or by blending with conventional fuels. There are two types of pathways for the industrial production of ethanol. Ethanol can be produced by fermenting glucose from biomass, as is already being done on a large scale in Brazil [9]. Another possibility is the production through the reaction of ethane with steam [8]. However, it should be noted here that ethane has a fossil origin as part of natural gas or as a by-product of oil refining. The disadvantage of the regenerative production of ethanol from biomass is the competition with the production of food, which is why solid agricultural waste should be used as feedstock. Methanol can be produced in a similar way to ethanol on the basis of different fossil and renewable energy sources [10]. In China, where methanol is already marketed on a large scale as an additive to conventional gasoline, it is produced using synthesis gas, whose components are CO and H 2 . However, synthesis gas in China is produced from hard coal, which is why methanol has a larger CO 2 footprint than conventional gasoline [11]. Though, methanol via synthetic gas be generated in a CO 2 -neutral way [12]. Another possibility is methanol production via CO 2 hydrogenation [13]. In both cases, H 2 and CO 2 respectively CO must be produced in a CO 2 -neutral way. CO 2 -neutral means that no fossil carbon is emitted during production and distribution of the fuel. H 2 can be produced with renewable electricity via electrolysis. Non-fossil CO 2 can be taken directly from the environment or local sources. For the conversion of CO 2 to CO, a reactor must be integrated for the subsequent synthesis [12]. An advantage of the direct use of methanol as a fuel is that no further process steps are necessary and 478 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="479"?> the fuel can be part of a holistic methanol economy [14]. In addition to the possibility of CO 2 -neutral production, alternative fuels must also be suitable for gasoline engine combustion. This includes properties such as octane number, ignition limits, burning rate, air requirement and volatility [8]. Table 1 compares the most important properties of ethanol and methanol with those of gasoline. Methanol Ethanol Gasoline LHV 1 [MJ/ l] 15.8 21.4 30-33 Stoich. air/ fuel ratio 1 6.46 8.98 14.58 Boiling point 1 [°C] 64.7 78 27-225 RON 1 109 109 88-98 Enthalpy of vaporization 1 (from 25°C) [kJ/ kg] 1168 919.6 ~ 351 Inflammability [vol%] 6.7-36.2 2 3.3-19.0 3 1.4-7.7 3 Laminar flame speed at NTP [m/ s] (λ = 1) 0.42 2 0.40 2 0.40 Tab. 1: Properties of methanol, ethanol and gasoline ( 1 [15], 2 [10], 3 [16]) Both ethanol and methanol have a higher octane number than gasoline. This mitigates knock tendency and enables an increase in the compression ratio and thus the achievable efficiency [17]. Methanol and ethanol can therefore also be used to reduce the tendency to knock [18]. In addition to increasing the octane number, the higher enthalpy of vaporization of methanol and ethanol also lowers the temperature in the cylinder and thus also the tendency to knock [19]. The laminar flame speeds of the three fuels are at a similar level. Nevertheless, studies have shown that an increased ethanol content in gasoline leads to faster combustion [20-22]. The low-boiling components in gasoline also lead to the formation of gaseous fuel at low engine temperatures, e.g. during cold starts. In contrast, due to the higher evaporation temperatures of methanol and ethanol, combustion instabilities and increased emissions of unburnt fuel fraction during cold start can occur for methanol and ethanol [8, 19, 23]. Studies on the temperatures of spark plug electrodes have also shown that lower combustion chamber temperatures are found when ethanol is used as a fuel [23]. An advantage of methanol and ethanol over gasoline is the higher hydrogen to carbon ratio of the fuel, which results in lower CO 2 emissions. However, this must be set against the lower energy density, which in turn increases fuel consumption and enables a lower mileage per liter. Furthermore, when using methanol and ethanol, the incompatibility with diverse materials and toxicity must be taken into account. In 2015, ACEA came out clearly against methanol as a fuel blend for various reasons, including the resistance of current engines and their components against methanol and toxicity [24]. 479 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="480"?> In addition to the use of CO 2 -neutral fuels, the CO 2 emissions of combustion engines can also be reduced by increasing their efficiency. Lean-burn processes can be used for this purpose [25]. To achieve stable combustion with high mixture leaning, pre-chamber spark plugs are used [26-28]. In a pre-chamber spark plug, a small portion of the air-fuel mixture is ignited in a volume, the so-called pre-chamber, which is connected to the main combustion chamber via orificies. The release of heat in the pre-chamber leads to an increase in pressure relative to the main combustion chamber, which in turn results in the outflow of gas jets from the orificies. Those gas jets penetrate the main combustion chamber and initiate the combustion. The advantage of the pre-chamber spark plug is the high energy that is supplied to the main combustion chamber and which is necessary to inflame the lean mixture in the main combustion chamber. In addition, the gas jets introduce turbulence into the main combustion chamber, which increases the burning speed and compensates for the slower flame speed of lean mixtures. A disadvantage of the pre-chamber spark plug is its large surface to volume ratio. This leads to large wall heat flows. Under full load, a hot pre-chamber spark plug can lead to pre-ignition, under partial load or cold start, a cold prechamber spark plug can lead to flame quenching and misfire [29, 30]. In order to adjust the temperature of the pre-chamber spark plug depending on the operating point, further measures must be taken. For example, active cooling can prevent excessive heating of the pre-chamber spark plug [27, 31]. Another possibility to adjust the temperature of the pre-chamber spark plug depending on the operating point is the integration of a heating element [29, 32]. The pre-chamber spark plug HSASI (Hot Surface Assisted Spark Ignition) developed at the Karlsruhe University of Applied Sciences has a controllable hot surface, more precisely a glow plug, integrated in the pre-chamber to accelerate the ignition within the pre-chamber. The design of HSASI is based on the swirl chamber spark plug introduced by Latsch [33]. HSASI enables the extension of the lean limit by Δλ = 0.1 [34]. In addition, the hot surface can be used to adjust the temperature and thus the heat range of the spark plug depending on the operating point for the reasons mentioned above. In this work, the performance of the HSASI ignition element is to be investigated when using methanol and ethanol blends in lean operation. With regard to the expected colder combustion chamber temperatures when using methanol and ethanol blends, the influence of HSASI operation, i.e. with an active hot surface, on inflammation and flame development is investigated. M25 and E85 are chosen as blends, whereas M25 consists of 25 vol% methanol and 75 vol% RON98, E85 is of 85 vol% ethanol and 15 vol% RON98. RON98 meets the DIN EN 228 standard. 2 HSASI ignition element The following section gives an overview of the HSASI ignition element. It describes the design of the pre-chamber spark plug body and the electric components of the ignition system and the glow plug controlling. 480 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="481"?> 2.1 Spark plug body The design of the HSASI spark plug is depicted in Figure 1. It differs slightly from the first prototype described in [35]. The spark plug body consist of a housing made of steel type S355 and a M14x1.25 thread. It includes a conventional but modified spark plug type ER9EH from NGK. The ground electrode of the spark plug is removed and replaced by a platinum wire with a diameter of 1 mm which is welded on the sidewall of the pre-chamber. The center electrode is extended with the same platinum wire to move the spark location further into the center of the pre-chamber. Platinum as electrode material was chosen because of its high resistance to corrosion even for higher temperatures [36]. Next to the spark plug the ceramic lead with its protective sleeve of a ceramic glow plug type CGP001 from BERU is integrated into the spark plug housing via hard soldering. The ceramic glow plug tip protrudes into the pre-chamber. The pre-chamber cap is made out of nickel and has five orificies, four radial and one axial with a diameter of 1.2 mm. The volume of the pre-chamber is 933 mm 3 which is 4.3% of the compression volume V c . Fig. 1: HSASI ignition element (left side) and pre-chamber with glow plug (red) and spark plug (blue) on the right side 2.2 Electric components An ignition coil type ZSE 032 from BERU provides the ignition energy for the spark plug. Due to limited space inside the spark plug shaft the ignition coil is connected to the spark plug via an extension cable. The temperature of the glow plug tip T CGP,tip shall be constant for a specific operation point. It is assumed that the electric resistance of the glow plug R CGP correlates with T CGP,tip [35]. Therefore, the controller algorithm of Scholl [37] is used to keep R CGP constant. The controller setup of ceramic glow plug (CGP) is depicted in Figure 2. A PI controller is implemented in the real time system ADwin Gold II. The controller sets the glow plug voltage U CGP,set via the power supply SM 18-50 from Delta Elektronika. To avoid damage to the glow plug the U CGP,set is limited to 10 V. The power supply measures U CGP which corresponds to U CGP,set as well as the glow plug 481 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="482"?> current I CGP so the actual resistance R CGP can be calculated. The controller frequency is set to 200 Hz which is sufficient to control R CGP . This is equivalent to approximately 13 controller intervention per cycle at an engine speed of 1800 1/ min. DC Power supply Controller Algorithm R = U / I DAC ADC CGP R CGP GUI R CGP,set U CGP,set U CGP I CGP T CGP,tip Fig. 2: Controller setup for glow plug controlling 2.3 Experimental setup The experimental investigations were carried out on a single cylinder test bed engine. Its specifications and the applied metrology are described below. 2.4 Engine specification The test bed engine is a naturally aspirated single-cylinder engine based on the Hatz 1B20. The cylinder head was modified and the injector bore was converted into a spark plug shaft with a M14x1.25 thread. As a result, the spark plug protrudes into the combustion chamber with an angle of 7°. A throttle and a gasoline low-pressure port fuel injector were integrated into the manifold. The injector is also suitable for methanol and ethanol injection. The fuel pressure is set to 3 bar and the end of injection to 300 °CAbTDCf (crank angle before top dead center firing). The engine has a heron cylinder head with a clearance height of 3 mm. The piston has a coin shaped piston bowl. For a better engine temperature control the cylinder head is equipped with a channel for liquid coolant. Coolant outlet temperature is kept constant at 70 °C. Fuel injection and ignition timing are controlled by the real time system ADwin Pro with a frequency of 125 kHz. Crank angles were calculated with the help of a crank angle decoder which has a resolution of 0.1 °CA and is also used for in-cylinder pressure indication system. A dynometer allows for various engine speeds. Table 2 shows further engine data. 482 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="483"?> Engine type (basis) Hatz 1B20 Geom. compression ratio 11.4 Stroke/ Bore 62/ 69 mm Displacement volume 232 cm 3 Number of valves 2 Injection Port fuel injection Tab. 2: Single-cylinder test-bed engine 2.5 Measurement setup Low frequency data, e.g. manifold pressure is sampled by the real time system ADwin Pro with a time interval of 1.5 s. Type K thermocouples measure temperatures e.g. coolant temperature. The manifold pressure is measured by a Siemens Sitrans P200 pressure sensor. The air-fuel equivalence ratio λ is determined by a wide band lambda sensor LSU4.9 from Bosch. A DEWETRON DEWE-800-CA indication system collects crank angle based data. In-cylinder pressure is measured by a Kistler sensor (6125C11). Indication system also samples the U CGP and I CGP . Table 3 shows the accuracies of the parameters which were kept constant during the measurements as well as its maximum standard deviations σ max . Measurand Unit Accuracy σ max T C,out K ±0.02 ±0.15 p man mbar ±0.30 ±1.30 λ - ±0.005 ±0.006 Tab. 3: Accuracy and standard deviation of measurand 3 Results For every operating point 200 consecutive cycles were measured. Each operating point is measured three times. Thus, every point in the following figures refers to the average value of 600 cycles. The net heat release is calculated according to Heywood [38]. Two different operating modes are compared. The first operating mode is with non-active glow plug controlling and is called passive pre-chamber spark plug. The second one is HSASI operation with active glow plug controlling. R CGP is set to the maximum value where the average value of U CGP does not exceed 3.5 V. This is done to protect the soldered glow plug from damage. Manifold pressure is set to 900 mbar by adjusting the throttle. Injection time is adjusted to get a constant air-fuel equivalence ratio 483 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="484"?> λ for each operating point of 1.35. Engine speed is kept constant at 1800 rpm. To investigate the influence of HSASI ignition on the inflammation a sweep in ignition timing was conducted. The net heat release as well as the flame development were compared between both operating modes. In addition, the combustion stability e.g. the cycle-to-cycle variations were investigated. 3.1 CGP controller Figure 3 shows the traces of U CGP and R CGP for a specific cycle. 0 °CA refers to the top dead center firing. R CGP was set to 0.813 Ω. A controller intervention can be recognised by a step in U CGP . The voltage steps have a maximum of 0.3 V. For each voltage step there is a peak in R CGP . It is assumed that those peaks do not influence the temperature of the glow plug and are due to the dynamic of the power supply. If the peaks are neglected, the actual value of R CGP differs only slightly from the set value R CGP,set . This indicates a sufficient controller set up. The ignition system disturbs the measurement of U CGP which comes apparent trough a short drop in U CGP during the charging of the primary circuit of the ignition coil. As a result, R CGP drops as well. − 360 − 270 − 180 − 90 0 90 180 270 360 Crank angle [ ° CA] 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 U CGP [V] U CGP R CGP R CGP,set 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 R CGP [Ω] Fig. 3: Traces of U CGP and R CGP as well as R CGP,set for one cycle at 1800 rpm. R CGP,set is set to 0.813 Ω. As mentioned above, to protect the glow plug from damage, R CGP was set to the maximum value which is defined by the maximum possible mean value of U CGP . This was possible due to the relatively simple monitoring of U CGP . However, this leads to different R CGP when varying the center of combustion CA50 (crank angle where 50% of net heat is released). Figure 4 shows R CGP for different CA50. With an early center of combustion R CGP increases, which indicates a higher glow plug temperature. This is due to the fact that for an earlier center of combustion more heat from the gas is transferred to the glow plug. Conversely, less heat is generated by the external power supply. With the exception of the earliest center of combustion where R CGP in the case of E85 is lower than that of M25 there seem to be no differences in R CGP between the two fuels at all other CA50. 484 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="485"?> 2 4 6 8 10 12 CA50 [ ° CAaTDCf] 0.76 0.78 0.80 0.82 0.84 R CGP [Ω] R CGP,set M25 R CGP,set E85 Fig. 4: The dependency of R CGP,set from the center of combustion at λ = 1.35 and p man = 900 mbar with a mean value of U CGP = 6 V. 3.2 Influence on flame development and heat release Figure 5 shows the flame development angle ΔΘ d (crank angle interval between ignition timing and crank angle where 5% of the total net heat was released) for different ignition timings between 14 and 20 °CAbTDCf and both operating modes. For the passive pre-chamber spark plug (PPCSP) mode with a non-active glow plug, the flame development angle for E85 shows higher values than for M25. There is no discernible trend for either E85 or M25 resulting in values of ΔΘ d at a similar level for all ignition timings. In comparision, HSASI operation seems to negate the difference for both fuels which were apparent in PPCSP mode. The values of ΔΘ d are on the same level for both fuels and lower than those of PPCSP operation. This indicates that HSASI operation compensates for possible colder pre-chamber in the case of E85 which could lead to higher values for ΔΘ d in PPCSP operation. 485 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="486"?> 12 14 16 18 20 22 IT [ ° CAbTDCf] 13 14 15 16 17 18 ∆Θ d [ ° CA] PPCSP M25 HSASI M25 PPCSP E85 HSASI E85 Fig. 5: Flame development angle for different ignition timings at 1800 rpm, p man = 900 mbar and λ = 1.35 The center of combustion for different ignition timings is depicted in Figure 6. For advanced ignition timings, CA50 moves to earlier crank angles due to advanced heat release for both operating modes. For PPCSP operation, E85 shows higher values for CA50 compared to M25 for all ignition timings which correlates with the findings for the flame development angle. As seen before in ΔΘ d , the same level in both fuels exhibits for HSASI operation. Same CA50 can be achieved under HSASI operation with approximately 1 °CA (for M25) to 3 °CA (for E85) advanced ignition timing. Thus, with an ignition timing of 14 °CAbTDCf a center of combustion of 9.2 °CAaTDCf can be attained for both fuels whereas with PPCSP operation CA50 results in 10.2 (for M25) and 13.3 °CAaTDCf (for E85). 12 14 16 18 20 22 IT [ ° CAbTDCf] 2 4 6 8 10 12 14 CA50 [ ° CAaTDCf] PPCSP M25 HSASI M25 PPCSP E85 HSASI E85 Fig. 6: Center of combustion for different ignition timings at 1800 rpm, p man = 900 mbar and λ = 1.35 Figure 7 shows the cumulative normalized net heat release for earliest (20 °CAbTDCf) and latest (14 °CAbTDCf) ignition timing for each of the operating modes. For M25 the 486 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="487"?> difference in the trace of the heat release is small between both operation modes. Heat release under HSASI operation starts slightly earlier for the same ignition timing but has no impact on the further heat release that comes apparent by a similar gradient. This indicates that the influence of the HSASI ignition is limited to the first part of the combustion i.e. to the flame development inside the pre-chamber. As already seen in Figure 6 the impact of HSASI operation on combustion is greater for E85. The heat release in HSASI operation not only starts significantly earlier, but also continues in the case of the later ignition timing with a steeper gradient. The deviation in the gradient is due to the heat release closer to the top dead center that is accompanied with higher temperatures and a faster heat release. With an earlier start of combustion, HSASI operation shows an earlier end of combustion comparing to the PPCSP mode. Fig. 7: Cumulative net heat release for late (14 °CAbTDCf) and early (20 °CAbTDCf) ignition timing at 1800 rpm, p man = 900 mbar and λ = 1.35 3.3 Influence on combustion stability An important parameter regarding engine efficiency and emission the combustion stability respectively the cycle-to-cycle variations [39]. Figure 8 shows the coefficient of variation COV IMEP for different ignition timings. COV IMEP is the ratio between the standard deviation of the indicated mean effective pressure (IMEP) and its mean value in percentage. COV IMEP is decreasing when the ignition timing is advanced for both operating modes and fuels. The reason for this is the earlier heat release near the top dead center. This results in higher gas temperatures and a more stable combustion. The effect becomes more apparent for E85 and PPCSP operation mode. Values for HSASI operation lie below those of the PPCSP operation. For latest ignition timing COV IMEP is reduced from 3.5% to 2% for HSASI operation and E85. In contrast, the effect for M25 seems to be marginal. The reduction in COV IMEP is moderate when comparing both operation modes. Nonetheless a decrease in COV IMEP from 1.5% to 1.2% is achieved for latest ignition timing. In comparision with E85, COV IMEP is lower for M25. 487 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="488"?> 12 14 16 18 20 22 IT [ ° CAbTDCf] 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 COV IMEP [%] PPCSP M25 HSASI M25 PPCSP E85 HSASI E85 Fig. 8: Cycle-to-cycle variations COV IMEP for different ignition timings at 1800 rpm, p man = 900 mbar and λ = 1.35 Figure 9 depicts the standard deviation of the start of combustion (σ SOC ) which is defined as the crank angle where 5% of the total net heat was released. E85 and PPCSP operation show higher values for σ SOC than M25 and HSASI operation. Whereas for HSASI operation values of σ SOC decrease for retarded ignition timings this trend is less apparent for PPCSP operation. With later ignition timings a higher gas temperature promotes the early flame kernel development. In addition, greater gas pressure increases the energy of the breakdown phase of the inductive ignition system which can also lead to a better inflammation [40]. However, these effects apply less to the PPCSP operation. 12 14 16 18 20 22 IT [ ° CAbTDCf] 0.8 1.0 1.2 1.4 1.6 1.8 2.0 σ SOC [ ° CA] PPCSP M25 HSASI M25 PPCSP E85 HSASI E85 Fig. 9: Standard deviation of the start of combustion σ SOC for different ignition timings at 1800 rpm, p man = 900 mbar and λ = 1.35 488 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="489"?> 4 Conclusion This work shows experimental investigations on the performance of the HSASI spark plug when using methanol and ethanol as blending fuel in a gasoline engine. Higher alcohol content in the gasoline is expected to decrease gas temperature of fresh mixture therefore deteriorate inflammation conditions. The inflammation of lean mixtures was investigated using M25 and E85 as fuel. HSASI operation was compared to the operation mode with passive pre-chamber spark plug (PPCSP). Due to higher alcohol content, E85 shows poorer inflammation with PPCSP operation. This becomes apparent with highest values for the flame development angle and the standard deviation for the start of combustion (σ SOC ). Also, E85 shows the highest cycle-to-cycle variation (COV IMEP ). The advantages of the HSASI operations result in the improvement of those parameters. Flame development angle is shortened by 3 °CA for latest ignition timing. σ SOC is reduced by up to 0.4 °CA compared to PPCSP operation. In addition, COV IMEP decreases from 3.5% to 2% for the latest ignition timing. For a constant center of combustion HSASI operation enables 2-3 °CA later ignition timings. In comparison with E85, the impact of HSASI operation on the inflammation of M25 is lower. However, the effect of the faster and more stable inflammation arises. Considering the inflammation process, HSASI operation seems to negate the difference in both fuels. This indicates that the inflammation process in PPSCP operation is mainly influenced by the gas temperature inside the pre-chamber. HSASI ignition influences the combustion mainly during the flame development process that means the gas jets penetrating the main combustion chamber seem to be independent on the operation mode. However, faster combustion can be achieved at constant ignition timing by releasing more heat closer to the top dead center. Abbreviations CA Crank angle CAaTDCf Crank angle after top dead centre firing CAbTDCf Crank angle before top dead centre firing CA50 Crank angle of 50% heat released (net release) CGP Ceramic glow plug HSASI Hot surface Assisted Spark Ignition IT Ignition timing I CGP Glow plug current IMEP Indicated mean effective pressure p man Manifold pressure R-T Resistance-temperature 489 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="490"?> R CGP Glow plug resistance R CGP,set Set value of R CGP SOC Start of combustion T CGP,tip Glow plug tip temperature T C,out Coolant outlet temperature U CGP Glow plug voltage U CGP,set Glow plug voltage set by controller V PC Pre-chamber volume Greek letters ΔΘ d Flame development angle λ Air-fuel equivalence ratio σ Standard deviation Acknowledgement The present work was part of the project “ReKra”. The authors would like to thank the Baden-Wuerttemberg Ministry of Economy, Labour and Housing for the financial support of the project. References [1] Scotiabank, “Global auto report,” 2022. [2] Xavier Mosquet, Aakash Arora, Alex Xie, and and Matt Renner, “Who Will Drive Electric Cars to the Tipping Point? ,” 2020. [3] Umweltbundesamt, “Weltweiter Autobestand,” https: / / www.umweltbundesamt.de/ bild/ wel tweiter-autobestand, July 20, 2022. [4] The International Council on Clean Transport, “Global passenger car market share of countries planning to phase out new sales of internal combustion engine vehicles: Fact Sheet Global,” 2021. [5] Springer, “EU Parliament votes to ban internal combustion vehicles from 2035,” https: / / www.springerprofessional.de/ internal-combustion-engine/ emissions/ eu-parli ament-votes-to-ban-internal-combustion-vehicles-from-203/ 23155862. [6] Bundestag, D., “Drucksache 20/ 1563: Markthochlauf synthetischer Kraftstoffe für den Verkehrssektor,” 2022. 490 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="491"?> [7] Bundesregierung, “Koalitionsvertrag 2021-2025: Mehr Fortschritt wagen,” oalitionsvertrag 2021 - 2025 zwischen der Sozialdemokratischen Partei Deutschlands (SPD" 2021. [8] Bae, C. and Kim, J., “Alternative fuels for internal combustion engines,” Proceedings of the Combustion Institute 36(3): 3389-3413, 2017, doi: 10.1016/ j.proci.2016.09.009. [9] Costa, R.C. and Sodré, J.R., “Hydrous ethanol vs. gasoline-ethanol blend: Engine performance and emissions,” Fuel 89(2): 287-293, 2010, doi: 10.1016/ j.fuel.2009.06.017. [10] Zhen, X. and Wang, Y., “An overview of methanol as an internal combustion engine fuel,” Renewable and Sustainable Energy Reviews 52(2): 477-493, 2015, doi: 10.1016/ j.rser.2015.07.083. [11] Yang, C.-J. and Jackson, R.B., “China's growing methanol economy and its implica‐ tions for energy and the environment,” Energy Policy 41(2): 878-884, 2012, doi: 10.1016/ j.en‐ pol.2011.11.037. [12] Martin Bertau, “Die Sonne im Tank - zum Potenzial Grüner Kraftstoffe aus Sicht der Chemie,” Denkströme. Journal der Sächsischen Akademie der Wissenschaft(20): 33-48, 2018. [13] Björn Kommoß, “Die Hydrierung von CO2 zu CH3OH unter überkritischne Bedingungen: Eine reaktionstechnische Untersuchung,” Disserstation, Technische Universität Darmstadt, Darmstadt, 2016. [14] Olah, G.A., “Beyond oil and gas: the methanol economy,” Angewandte Chemie (International ed. in English) 44(18): 2636-2639, 2005, doi: 10.1002/ anie.200462121. [15] Sarathy, S. M., Oßwald, P., Hansen, N., and Kohse-Höinghaus, K., “Alcohol combus‐ tion chemistry,” Progress in Energy and Combustion Science 44: 40-102, 2014, doi: 10.1016/ j.pecs.2014.04.003. [16] Heechang Oh, Choongsik Bae, and Kyoungdoug Min, “Spray and Combustion Character‐ isitics of Ethanol Blended Gasoline in A Spray Guided DISI Engine under Lean Stratified Operation,” 2010(Volume 3). [17] Speth, R.L., Chow, E.W., Malina, R., Barrett, S.R.H. et al., “Economic and environmental benefits of higher-octane gasoline,” Environmental science & technology 48(12): 6561-6568, 2014, doi: 10.1021/ es405557p. [18] Daniel, R., Wang, C., Xu, H., Tian, G. et al., “Dual-Injection as a Knock Mitigation Strategy Using Pure Ethanol and Methanol,” SAE Int. J. Fuels Lubr. 5(2): 772-784, 2012, doi: 10.4271/ 2012-01-1152. [19] Stein, R.A., Anderson, J.E., and Wallington, T.J., “An Overview of the Effects of Ethanol-Gas‐ oline Blends on SI Engine Performance, Fuel Efficiency, and Emissions,” SAE Int. J. Engines 6(1): 470-487, 2013, doi: 10.4271/ 2013-01-1635. [20] Iltesham Z Syed, Y A Mukherjee, J D Naber, and D Michalek, “Numerical Investigation Of Laminar Flame Speed Of Gasoline Ethanol Air Mixture With Varying Pressure, Temperature and Dilution,” SAE Int. J. Engines 2010(3). [21] Del Pecchia, M., Pessina, V., Berni, F., d'Adamo, A. et al., “Gasoline-ethanol blend formula‐ tion to mimic laminar flame speed and auto-ignition quality in automotive engines,” Fuel 264(1): 116741, 2020, doi: 10.1016/ j.fuel.2019.116741. [22] Bayraktar, H., “Theoretical investigation of flame propagation process in an SI engine running on gasoline-ethanol blends,” Renewable Energy 32(5): 758-771, 2007, doi: 10.1016/ j.renene.2006.03.017. 491 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="492"?> [23] Kalghatgi GT, Nakata, K., and Mogi, K. (eds.), “Octane Appetite Studies in Direct Injection Spark Ignition (DISI) Engines,” SAE Technical Paper Series, 2005 SAE World Congress, Detroit, Michigan, April 11-14, 2005, 2005. [24] ACEA, “ACEA Position Paper on Methanol as a Gasoline Blending Component,” 2015. [25] G J Germane and C G Wood, C C Hess (eds.), “Lean Combustion In Spark Ignited Internal Combustion Engines - A Review,” SAE Technical Paper Series, Fuels and Lubricants Meeting, San Francisco, October 31 - November 3, 1983. [26] Alvarez, C.E.C., Couto, G.E., Roso, V.R., Thiriet, A.B. et al., “A review of prechamber ignition systems as lean combustion technology for SI engines,” Applied Thermal Engineering 128(1-2): 107-120, 2018, doi: 10.1016/ j.applthermaleng.2017.08.118. [27] Sens, M., “Pre-Chamber Ignition as a Key Technology for Highly Efficient SI Engines - New Approaches and Operating Strategies,” in: 39. Wiener Motorensymposium. [28] Kettner, M., “Zündsysteme für magere Gemische,” Study for the Forschungsvereinigung Verbrennungskraftmaschinen (FVV) e.V., 2004. [29] Sens, M. and Binder E., “Vorkammerzündung als Schlüsseltechnologie für einen zukünftigen Antriebsstrang-Mix,” MTZ Motortechnische Zeitschrift 80(02): 46-53, 2019. [30] Kettner, M., “Experimentelle und numerische Untersuchungen zur Optimierung der En‐ tflammung von mageren Gemischen bei Ottomotoren mit Direkteinspritzung,” Ph.D. Thesis, Universität Karlsruhe, 2006. [31] Gussak L. A., Karpov V.P., Tikhonov Yu. V., “The Application of Lag-Process in Prechamber Engines,” [32] Marko F. and Koenig G., “Zuführungs- und Zündvorrichtung für einen Gasmotor und Verfahren zum Betrieb einer Zuführungs- und Zündvorrichtung für einen Gasmotor,” DE 10 2017 009 607 A1. [33] Latsch R., “The Swirl-Chamber Spark Plug: A Means of Faster, More Uniform Energy Conversion in the Spark-Ignition Engine,” [34] Holzberger, S., Kettner, M., and Kirchberger, R., “Experimental Investigations on a Pre-Chamber Spark Plug with Variable Heat Range by Integrating a Controlled Hot Surface,” The Journal of Engine Research(69), 2022. [35] Holzberger S., Kettner M., and Kirchberger R., “Hot Surface Assisted Spark Ignition (HSASI) - An Approach to Vary the Heat Range of Spark Plugs during Operation,” Reports on Energy Efficient Mobility, 2021. [36] Rager, J., “Funkenerosion an Zündkerzenelektroden,” Ph.D. Thesis, Universität des Saar‐ landes, 2006. [37] Scholl, F., “Study of Premixed Combustion Induced by Controlled Hot Surface Ignition in Stationary Gas Engines,” Ph.D. Thesis, University of Valladolid & Hochschule Karlsruhe, 2017. [38] Heywood, J., “Internal Combustion Engine Fundamentals,” McGraw-Hill, ISBN 0-07-028637-X, 1988. [39] Ozdor, N., Dulger, M., and Sher, E., “Cyclic Variability in Spark Ignition Engines - A Literature Survey,” 940987. SAE Technical Paper, 1994, doi: 10.4271/ 940987. 492 Sascha Holzberger, Maurice Kettner, Roland Kirchberger, Ivica Kraljevic, Florian Sobek <?page no="493"?> [40] Maly, R. and Vogel, M., “Initiation and propagation of flame fronts in lean CH4-air mixtures by the three modes of the ignition spark,” Symposium (International) on Combustion 17(1): 821-831, 1979, doi: 10.1016/ S0082-0784(79)80079-X. 493 Experimental Investigations on the Perfomance of the HSASI Pre-Chamber Spark Plug Using Ethanol and Methanol Blends <?page no="495"?> Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio Lukas Euchner, M.Sc, BMW Group; Laura Baumgartner, Dr.-Ing., BMW Group; Michael Wensing, Prof. Dr.-Ing., Friedrich-Alexander-Universität Erlangen-Nürnberg; Tim Russwurm, M. Sc., Friedrich-Alexander-Universität Erlangen-Nürnberg; Peter Janas, Dr.-Ing., Tenneco, Inc. Abstract: A pre-chamber fundamentally changes the combustion process by initiating the ignition in a separate chamber connected by holes with the main combustion chamber. Due to the pressure rise of the ongoing combustion inside the pre-chamber, highly reactive jets spray out via the connecting channels with a high impulse and ignite the main chamber along each of them. The widely spread ignition sources enable an enhanced burn rate, shorter burning durations and an earlier center of combustion resulting, in decreased exhaust gas temperatures. On this basis the engine running at stoichiometric air-fuel ratio can provide a significant improvement in power. A higher compression ratio can be implemented by orienting the jets on potential knocking hot spots. One of the biggest challenges is to design a pre-chamber system that reliably operates at low load while high fractions of residual gas inside the pre-chamber deteriorate the combustion process. In this range of operation, scavenging the pre-chamber with air or an air-fuel mixture shows high potentials regarding combustion stability by supplying an ignitable mixture near the spark plug and providing the advantage of lowering the mass fraction of residual gas inside the pre-chamber. However, there are many degrees of freedom to design this scavenging properly. Therefore, the conditions at ignition timing can vary widely. Detailed analyses of the load change, the ignition, and the combustion process in 3D CFD simulations is mandatory due to a lack of optical accessibility to the pre-chamber. Even in an early phase of the development process this tool can be used to enhance evaluating the ignitability of a mixture and to analyse the early flame propagation. Ignition is a highly complex phenomenon. By default, most of the CFD codes use simplified ignition models that initialize a little sphere at spark timing. In this sphere a sudden rise of temperature and pressure ignites the mixture. These <?page no="496"?> simplified spark models are not able to provide a reliable statement on the ignitability, delay of ignition and the early flame propagation. More complex models show enormous improvements in modeling the phenomena with rapidly changing conditions in the ignition gap. This publication shows the implementation of a complex ignition model. After the implementation and validation of a suitable model, an operating strategy for the active scavenged pre-chamber will be developed. In this process the problems at low load will be targeted by optimizing the timing, the duration, and the quality of the scavenged mixture. The improved conditions at ignition timing result in an increased burn rate inside the pre-chamber and a decreased delay of combustion. 1 Introduction To fulfill the increasing requirements of climate protection and environmental combability for the sector of private transport, the main development goals are to optimize the thermal efficiency and decrease the real driving emissions of future internal combustion engines. A pre-chamber combustion system can play a decisive role for this development process. The system can be seen as an enabler for rising the compression ratio, which directly increases the thermal efficiency, or for reliably igniting even highly diluted main mixtures by promoting a significantly higher ignition energy compared to the regular spark plug ignition [15]. While the central mounted regular spark plug is only capable of igniting mixtures up to an air-fuel ratio of λ = 1.6 or residual gas fractions of 25% to 30%, the pre-chamber ignition system provides the potential of enhancing these restrictions widely, depending on the pre-chamber configuration [7]. The IAV GmbH showed that even a highly diluted mixture with a mass fraction of 40% residual gas can be ignited with absolute reliability by using an active scavenged pre-chamber [16]. The diluted mixture increases the isentropic exponent, which results in improved thermal efficiency. The 1.5 liter, 3-cylinder demonstrator engine by MAHLE Powertrain can be seen as one example for this development approach. Cooper et al. [4] published results in 2020, where the pre-chamber combustion enabled the engine to run very efficiently with the application of an increased compression ratio, external exhaust gas recirculation and the Miller cycle. Nevertheless, the pre-chamber ignition system comes along with some challenges, for example the increased heat loss. When using the passive pre-chamber in a passenger car application, the operating area from low load till full load over the complete engine speed range must be reliably covered. At low load high residual gas fractions in the pre-chamber impede the ignition process. At high speed and high load, the pre-chamber system must deal with high temperatures which can lead to irregular combustion. To target the problems at low load, the geometry of the pre-chamber must be equipped with small volumes and large orifice diameters for better scavenging during the compression phase. At high load larger pre-chamber 496 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="497"?> volumes and smaller orifice diameters increase the jet impulse and shorten the combustion duration [3]. This trade-off is targeted in the present publication by optimizing the geometrical design for high load and fighting low load restrictions by scavenging the pre-chamber with a gaseous air-fuel mixture. The pre-chamber development has a long history going back to the early years of the 20 th century. Gussak published first results in the 1950s and his publications in 1975 [6] show high validity compared to present scientific papers according to pre-chamber designs and the description of the ignition process in the main chamber. During the beginning of the 21 st century the focus of the development process was on heavy duty natural gas engines operating at lean air-fuel mixtures. These engines are mainly running for purpose of power generation in steady state at a restricted operating range. These boundary conditions can simplify the technical design of the ignition system because of the lack of dynamic phenomena, low load restrictions or thermal problems operating at high load. So developers like Kawabata and Mori [8] or Wellander et al. [20] could investigate the combustion process in the main combustion chamber in more detail. By using an optical accessible engine, they could visualize the turbulent jets exiting the pre-chamber and the flame front with the entrainment of fresh mixture to the jets and the flame propagation from the outer parts of the combustion chamber towards the center parts. Nevertheless, Getzlaff et al. [5] stated that this highly complex system cannot be properly designed without the help of three-dimensional computational fluid dynamics investigations covering the load change in the pre-chamber, the distribution of the jets and the detailed analysis of the combustion phenomena. For example Baumgartner et al. [1] made numerical investigations on the mixture formation of a methane fueled pre-chamber optimizing the conditions at ignition timing at the electrodes of the spark plug. Others like Wang et al. [19] used numerical investigation to evaluate the ignition location in the main chamber depending on the fluid temperature inside the domain. One of the largest investigations on the ignition process in the main combustion chamber was published by Biswas [2]. He obtained two different ignition mechanisms depending on the Damköhler number, namely, jet ignition and flame ignition. Within the jet ignition the main mixture is ignited through hot combustion products, while the flame is quenched when passing the orifices of the pre-chamber. By increasing the orifice diameter, the ignition mechanism switches to flame ignition. The comon factor in these publications is that they use a very simplified ignition model for their numerical investigations, where a small sphere is implemented at spark timing. Within this sphere a specific amount of energy at a prescribed location must ignite the mixture. This model does not provide an accurate representation of the spark channel, nor does it give consideration of the flow field between the electrodes. A technical paper published by Laget et al. [9] used a more detailed model. In this paper the Imposed Stretch Spark Ignition Model (ISSIM) was used, which enables the spark channel modeling, but has a much simpler structure than AVL Curved Arc Diffusion Ignition Model (CADIM) used in this publication. A comparable ignition modeling 497 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="498"?> strategy was published by Shapiro et al. [17]. These results are in agreement with the present study. In the following sections the potential of the passive pre-chamber ignition system is being investigated in a validated 0D/ 1D simulation environment. The goal is to accurately predict the performance of the stoichiometric running gasoline BMW four-cylinder engine equipped with the pre-chamber. These investigations result in the definition of operating areas, where the active scavenging of the pre-chamber should be pursued. Afterwards the numerical modeling of the ignition process in AVL FIRE TM by using the AVL CADIM in combination with the three zone Extended Coherent Flame Model (ECFM-3Z) will be described and validated. With the validated three-dimensional model an operating strategy will be developed for specific areas, defined in the 1D simulations, targeting the restrictions at low load operations. In this process, the injection timing of the pre-chamber injector, the injection duration and the composition of the injected gaseous mixture will be varied. The goal is to achieve the best conditions possible at ignition timing. As Tang and Sarathy [18] stated in their 0D fundamental study on the pre-chamber combustion, the mixture in the pre-chamber should be near stoichiometric to provide turbulent jets containing a high rate of active radicals to ignite the main mixture in an efficient way. In a final evaluation, results from 1D and 3D simulations will be compared to results from engine test bench investigations. 2 Methodology 2.1 Project LEANition The project LEANition is a publicly funded project by the Bavarian Research Foundation. One of the project partners is the Friedrich-Alexander-University of Erlangen-Nuremberg. At the Institute of Engineering Thermodynamics (LTT) the pre-chamber fueling system was developed and constructed. The Tenneco Group is the other project partner responsible for constructing and manufacturing the pre-chamber prototypes, see in Figure 1. 498 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="499"?> Figure 1: Pre-chamber system LEANition project The goal of the project is to develop an active scavenged pre-chamber system suitable for passenger car sized gasoline engines running at global stoichiometric mixtures in order to guarantee the best conversion rates of the three-way catalytic converter. One of the main package restrictions on the engine side is that the pre-chamber must fit into the M12 thread of the spark plug shaft without any changes to the cylinder head. This requirement impedes the construction of the pre-chamber housing. Therefore, the insulator, the pressure sensor and the check valve must be very small to fit in the pre-chamber. At low load there is a relatively high amount of residual gas located inside the cylinder and an even higher amount in the pre-chamber. Therefore, one of the main tasks of the fueling system is to flush residual gas out of the pre-chamber. The other task is to provide a near-stoichiometric mixture at spark timing between the electrodes. The injection timing during the compression phase requires high scavenging pressures of up to 10 bar, high mass flows to efficiently flush out residual gaseous parts and gaseous mixtures to prevent high wall films inside the pre-chamber. The fueling system of the pre-chamber is operating with volatile components of gasoline fuel. The basis for this approach was set by Schumacher et al. and Russwurm et al. [12, 14]. The gaseous fuel is mixed with fresh air to receive the suitable air-fuel ratio needed. The pre-chamber scavenging is timed and dosed with a common MPI fuel injector connected to the check 499 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="500"?> valve by a small pipe. One big advantage of this system is that there is no need for a dual fuel system to scavenge the pre-chamber with gaseous fuel. The pre-chamber itself has a volume of V ≈ 1200 mm³. Variants with different numbers and orientations of orifices were investigated but will not be part of the present publication. 2.2 Test engine A BMW four-cylinder engine with an aggregated displacement volume of 2000 cm 3 and a compression ratio of 11.6 was used the basis for the 1D, 3D and test bench investigations. This turbocharged engine is equipped with direct injection and a fully variable valve train. 2.3 1D GT-Power model During the development of internal combustion engines, the use of virtual simulation tools is becoming more and more important due to the enormous time and cost savings. An important tool that links the disciplines of detailed simulation and test bench investigations is 1D modeling with GT-Power. This modeling approach displays the internal engine processes with manageable complexity and manageable effort, while at the same time providing sufficient accuracy for initial potential analyses. It includes the periphery, the four-cylinder engine with the crank train and the template for the pre-chamber. It is based on a foundational model from BMW series development and is therefore already calibrated in detail. Combustion in the pre-chamber and main combustion chamber alone requires more extensive calibration and validation effort. The non-predictive combustion model SIWiebe is used for combustion modeling in the pre-chamber, since only the center of combustion, the burning duration, and a vibe exponent need to be specified for the fixed combustion curve. The transition from the non-predictive to the semi-predictive combustion model can be achieved by storing these parameters of the Wiebe function in characteristic look-up tables as a function of rotational speed and engine load covering the entire engine map. A fully predictive combustion model is used for the main combustion chamber. The multi-zone model EngcylCombJetignition was developed specifically for pre-chamber combustion. Derived from the diesel combustion model, a free jet is modeled in the main combustion chamber. In this model, two main variables are anchored. They drive the free jet propagation in the main combustion chamber, namely the penetration depth S(t) and the velocity of the free jet tip u(t). Wenig and Roggendorf [21] summarized the formulaic connection. 500 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="501"?> S t = C s t 1/ 2 u noz d noz C d ρ noz ρ cyl 1/ 2 u t = C u t −1/ 2 u noz d noz C d ρ noz ρ cyl 1/ 2 where t is the time elapsed, u noz is the velocity at the pre-chamber nozzle, d noz is the nozzle diameter, C d is the nozzle discharge coefficient, ρ noz is the density of the fluid in the pre-chamber jet at the nozzle exit, ρ cyl is the density of the surrounding fluid in the main chamber and C s , C u are model constants. As the free jet propagates in the main combustion chamber, it is expanded and decelerated via the entrainment from fresh fluid components. At this stage two zones are implemented, namely the unburned zone and the jet zone. With the start of combustion, a third zone, the burned zone is initiated. After the start of combustion, mass continues to be entrained into the free jet. The burning rate dm jb dt for the jet combustion is then obtained as follows. dm jb dt = C df m ju k V cyl1/ 3 f y O 2 + s˙ f m ju Where m ju is the unburned mass inside the jet, k is the turbulent kinetic energy, V cyl is the main chamber volume and C df is a tuning parameter. The term k V cyl1/ 3 f y O 2 is a mixing-controlled burning rate. Burning by the turbulent flame brush is modeled by the propagation of a single spherical flame governed by an entrainment and burn-up model. As the flame brush propagates into the main chamber it entrains mass which is then burned behind the flame over a characteristic timescale. The equations governing the model are dm f u dt = ρ u A e S T − dm f b dt dm f b dt = m f u τ + s˙ j m f b τ = C l λ S L ; s˙ j = 1 m b dm jb dt ; S T = S L + C t u′ 501 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="502"?> The first equation controls the entrainment rate. The second equation governs the burning rate of the entrained mass in the jet brush. S T is the turbulent flame speed of the brush, S L is the laminar flame speed, τ is the characteristic burning timescale and s˙ j is a source term coupling the flame propagation model with the jet combustion model. After describing all these models, it is possible to visualize the relations in Figure 2. Figure 2: Visualization of the GT-Power combustion model 2.4 CFD Model Detailed analysis of the flow and combustion processes in CFD plays a pivotal role for the development of the pre-chamber ignition system. For this purpose, the following chapter deals with the construction of a valid 3D CFD modeling using AVL FIRE TM . 2.4.1 RANS Model RANS equations are solved using the finite volume method. The effects of turbulence are solved by means of Reynolds Averaged Navier-Stokes (RANS) methods, specifically the k-ζ -f-model. Combustion is modeled using the ECFM-3Z model. For RANS simu‐ lations the computational mesh size was chosen based on a grid sensitivity study. 502 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="503"?> 2.4.2 Ignition modeling Detailed ignition modeling is an important component on the way to a valid simulation model. Since the ignition conditions in the pre-chamber can vary widely depending on the prevailing flow conditions and mixture compositions, it is important to model the arc and its deflection during spark discharge. This results in a transfer of the ignition location, depending on the peripheral conditions. In addition to the mixture composition, other factors, such as the surrounding velocity field, are of decisive importance for ignition and subsequent flame propagation. The detailed ignition modeling provides the basis combustion modeling with physically correct starting conditions. Ignition modeling is normally performed in pre-chamber simulation with the highly simplified spherical ignition model. In this approach, a sphere is initialized at a predeter‐ mined ignition location at a predetermined ignition timing without any delay. During the ignition process, the values of pressure and temperature in the sphere are abruptly increased, without considering arc development and detailed chemical reactions. For this publication, the Curved Arc Diffusion Ignition Model (CADIM) was used, which was implemented as AVL CADIM in FIRE TM with an extension to the secondary electrical circuit. This model consists of four sub-models shown in Figure 3: the spark channel evolution model, the secondary electrical circuit model, the temperature diffusion model, and the ignition delay model . Figure 3: Block Diagram of the AVL CADIM Ignition Model 503 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="504"?> Lagrangian parcels are implemented between anode and cathode in the spark channel evolution model. These can be distorted by the local flow field u(x,y,z) and thus form the arc. The model for the secondary electrical circuit includes the secondary inductance, which must be specified in the model, and the spark plug. The change in available energy is calculated as: dE s dt = − R s * i s2 − V ie * i s with: V ie = V cf + V af + V gc These formulas contain the current strength i s , the resistance R s and the voltage between the electrodes V ie , which is composed of the cathode V cf , the anode V af and the breakdown voltage V gc . The length of the plasma channel l spk from the spark channel evolution model is included in the calculation of the breakdown voltage. V gc = C gc * l spk * i sC is p 0, 51 C gc and C is represent constants. Finally, the energy provided to the plasma arc can be calculated from the product of voltage and current: Q spk = V gc * i s During the breakdown phase, temperatures of several thousand Kelvin prevail in the plasma. Outside the plasma, where temperatures are slightly lower, a large number of chemical reactions take place. The main task of the temperature diffusion model is to solve the 1D heat equation in the normal direction of the arc surface. The arc curvature leads to a reduction or increase in local energy transfer, depending on whether the curvature is convex or concave. Finally, the ignition delay is determined via tabulated ignition delay times τ . This is a function of the temperature at the arc surface T , the pressure p, the air-fuel ratio ϕ and the residual gas fraction EGR. The ignition delay time τ is selected at each time step. An ignition precursor Y p is defined. The change over time is defined as follows: dY p dt = τ 2 + 4 1 − τ Y p τ The Lagrangian parcel, for which the precursor first assumes the value Y p = 1, triggers the ignition. Ignition is expected in parcels of high curvature, as this is where the temperature is the highest. The initial flame core must be defined in AVL FIRE TM . In the 504 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="505"?> present case, the electrode spacing of d = 0, 4mm was defined as the initial flame kernel size. If the flame kernel has been initialized, the combustion model is implemented, and flame propagation begins. The occurrence of misfire can also be determined using AVL CADIM. If, for example, the precursor does not assume the value Y p = 1 at any time during the breakdown phase due to an excessively high flow field or a poor mixture, no ignition will take place. This is an extremely important feature, especially for pre-chamber investigations [11]. 2.4.3 Ignition model implementation Configuring the electrical circuit is the first step in implementing the model. After determining the location of the cathode and the anode some parameters of the second electrical circuit must be specified: cathode and anode voltage, inductance and resistance of the system and the initial energy. By parameterizing the efficiency factor of the energy transmission and an ignition parameter, it is possible to influence the ignition delay and the initial flame propagation to match the pressure course of the simulation and the measurements. With one parameter setup, it is possible to compare different pre-chamber configurations regarding ignition delay and arc deflection caused by the local flow field during spark discharge, see Figure 4. Figure 4: Spark length evolution and distorted spark depending on the pre-chamber geometry 505 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="506"?> 2.4.4 Validation Figure 5: Final CFD setup matching Together with the ECFM-3Z combustion model, it is now possible to finally param‐ eterize and validate the CFD model. For the validation process, pressure matching between measurement and simulation is used inside the pre-chamber (see Figure 5), and recordings of optical measurements are used to validate the main combustion. For this purpose, results from Russwurm et al. are used, who analyzed the propagation of the flame front in the main combustion chamber on an optically accessible single-cylinder research engine through flame luminescence. The results of the validation can be seen in Figure 6. Figure 6: Validation of the CFD modeling with optical recordings Validation begins with the exit of the free jets. The comparison between optical recordings and CFD results at 720° CA show that the timing of the free jet outflow fits very well, this can be seen especially on the upper jet between the intake valves. In the next step, a very good symmetry of the four exiting jets can be detected. The 506 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="507"?> shape of the various torches also shows a great similarity between measurement and calculation. In particular, the expansion of the jets and the anomaly of the flame front fanning out when impinging the rooftop of the combustion chamber between the intake valves show a valid match. The combustion chamber is completely covered by the flame front both in optical recordings and in CFD simulations from 727.5° CA on. In the CFD results, ISO surfaces were visualized for various combustion steps. These are overlapped, this is why the intensity of the combustion cannot be fully assessed at each time step. However, there are large similarities in the distribution of combustion intensity. In the final time step of the analysis, the hotspots of the highest combustion progress in measurement and CFD are virtually exactly on top of each other. 3 Results The results of this publication are divided into three parts: potential analysis in 1D simulations, analyzing flow and combustion in 3D CFD simulations, and final evaluations which compare these simulations to results from test bench investigations. 3.1 Potential analysis of the passive pre-chamber system in 1D simulations The present model was calibrated and validated based on measurements from a previous pre-chamber examination from another project. With 1D simulations, a first fundamental comparison can be made between conventional spark ignition and passive pre-chamber ignition. In this comparison, strengths and weaknesses of passive operation are to be shown, which in conclusion can define an operating range for the active scavenged pre-chamber. Table. 1: Specification of the investigated LEANition pre-chamber At this point, the potential analysis of the passively operated pre-chamber from the LEANition project is started. For this purpose, the already calibrated spark plug 507 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="508"?> model, on which the pre-chamber model is based, is used as a basis for comparison. Positive numerical values in these comparisons therefore relate to an increase in the pre-chamber values and, in reverse, negative values describe a reduction compared to spark plug testing. Different combustion parameters and fuel consumption are compared to identify the strengths and weaknesses of a passive pre-chamber in the examined characteristic engine map. The specification of the pre-chamber used for these investigations can be found in Table 1. In a first basic comparison between spark plug and pre-chamber combustion, it makes sense to look at general combustion parameters. It is recommended to analyze the burning duration from Figure 7, which covers the range between the 5% and the 90% fuel conversion point. Figure 7: Comparison of the burning duration between regular spark plug and passive pre-chamber The classic behavior of the passive pre-chamber can be recognized quite quickly over a larger range of the engine map. At low load and with additional operation at low engine speeds, pre-chamber combustion has a burning duration of up to 40% longer than the spark plug. The main problem here is an increased residual gas content. By comparing Figure 7 and Figure 8, a very good match can be achieved between areas of extended burning duration and areas of high residual gas fractions inside the pre-chamber. 508 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="509"?> Figure 8: Residual gas fraction inside the passive pre-chamber In the critical area of low load, the simulation results show a residual gas content inside the pre-chamber of up to 32.5%. This leads to poor ignition conditions, delayed pre-chamber combustion and a low impulse of the free jets when passing over into the main combustion chamber. This results in an extension of the combustion period. With higher load, the increasingly stable and rapid combustion with decreasing residual gas content results in a significant reduction in burning duration. The optimum is towards the full load curve in the low-end torque range, where more than half the burning duration can be achieved in the simulation compared to regular spark ignition. In this area, the pre-chamber achieves the highest pressure gradients and can therefore also provide significant advantages in an earlier center of combustion compared to the spark plug. This is due to the rapid combustion in the strongly knock-limited area. Finally, the first potential analysis still provides a detailed look at the specific fuel consumption comparison between spark plug and passive pre-chamber, which is primarily intended to assess the operating range of the passive prechamber and to estimate possible areas of application and potentials of the active pre-chamber. For this reason, a specific fuel consumption comparison was made between passive pre-chamber and spark plug operation in Figure 9. The engine map can also be roughly divided into three areas, which are delimited from one another by dashed lines. 509 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="510"?> Figure 9: Comparison of the specific fuel consumption between spark plug and passive pre-chamber The three delimited areas, which are dominated by various influencing parameters, differ fundamentally in the change of specific fuel consumption. In area 1, which extends in the lower left corner of the map in Figure 9, passive pre-chamber combustion has a significantly higher fuel consumption compared to the spark plug. Looking again at the burning time and residual gas content from Figure 7 and Figure 8, a good match can be seen between the area 1 and the range of the largest residual gas content in the pre-chamber and the associated delayed combustion and extended burn duration. Together with the high migratory heat loss, there is an apparent consumption disadvantage of up to 5% in the extreme regions. Section 2 of Figure 9 is located in a region that can be considered largely as consumption-neutral in comparison of the two combustion methods. In this case, increasing combustion efficiency with more complete combustion and shorter burning duration can compensate wall heat losses, as a result of which the effective efficiency between pre-chamber and spark plug equalizes at a comparable level. This region 2 extends over the entire speed range under consideration at medium load, in which the residual gas influence in the pre-chamber decreases steadily as a result of the higher charge. At loads above the region 2, region 3 extends in which the pre-chamber combustion process can achieve a consumption advantage over conventional spark plug operation. In this area, the advantages of the pre-chamber come into full effect. The influence of the residual gas 510 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="511"?> content in the pre-chamber is virtually negligible. Due to the rapid combustion and the decrease of the distance between the ignition source and potential knock hotspots in the main combustion chamber, more efficiency-optimized center of gravity positions can be achieved in the knock-limited area. In non-knock-limited operating ranges, the advantage in consumption is achieved by the sharp increase in combustion efficiency. Maximum consumption reductions of up to 8% were apparent from the engine process simulation. As a result of this observation, it is likely that there is no need for actively scavenging the pre-chamber in region 3. Regions 1 and 2 therefore emerge from Figure 9 as potential main areas for the application of an active pre-chamber. The residual gas fraction and the mixture composition in the pre-chamber could be directly influenced by the gas scavenging, which in the first step could make pre-chamber combustion significantly more stable and could noticeably accelerate the complete burn-through of the pre-chamber charge. As a result of the improved pre-chamber combustion, an improved main combustion could therefore be achieved, which would result in an increase in efficiency. 3.2 Numerical development of an operating strategy for the active scavenged pre-chamber Figure 10: Nomenclature of the simulation cases It became clear from the previous chapters that stable combustion in passive mode can no longer be guaranteed at low load and with increasing residual gas fractions in the pre-chamber. For this reason, an operating strategy is being developed in CFD with which combustion in the pre-chamber is to be improved. This results in the goals of residual gas purging and the improvement of ignition conditions, which are pursued in the following chapter by means of CFD simulations. The indication of the simulation cases follows the nomenclature of Figure 10. The investigated operating point has an engine speed of 1000 rpm and an IMEP of 4 bar with a fixed ignition timing at 703° CA. 3.2.1 Scavenging the pre-chamber with clean air The easiest approach for flushing the residual gas mass out of the pre-chamber is to scavenge with clean air. During the compression stroke the mixture from the main 511 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="512"?> combustion chamber will be pushed into the pre-chamber. The hope is that, as a result of an early air scavenging, there is still sufficient time until spark timing to provide an ignitable mixture inside the pre-chamber. For this purpose, the variation of the end of scavenging (EOS) was embedded in the CFD. The comparison in Figure 11 shows that there is a band of 180° CA where few differences in the pressure curve are detectable. However, a significant deterioration can be seen in the event of a further retardation. This can be attributed to the air-fuel ratio in the pre-chamber, which drops again significantly with a later EOS. The reduced equivalence ratio (EQR) ensures that flame propagation is slowed down. However, the main problem of air injection can also be seen from the same figure. By having a look on the EQR in the pre-chamber in passive mode, a significantly richer mixture is noticeable. Even a very early injection of air while the exhaust valves are still open cannot prevent excessive dilution at spark timing. Figure 11: Effect of varying EOS of air injection on pressure and EQR inside the pre-chamber Unfortunately, the early EOS neutralizes the residual gas advantage in the pre-chamber, because almost the entire residual gas content can be flushed out during the injection of air, see Figure 12. However, this proportion is significantly increased again during compression by inflowing mass from the main combustion chamber, which is why an advantage of only about 5% can be achieved at the time of ignition. Even a variation of the injection duration can no longer provide any improvement. In comparison to passive operation, this results in an almost undetectable improvement in the pressure development in the pre-chamber for a duration of 5 ms and an EOS as early as possible. For these reasons, scavenging of the pre-chamber with combustible mixture is essential and will therefore be discussed further in the following chapters. 512 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="513"?> Figure 12: Effect of varying EOS on the residual gas fraction/ effect of varying the scavenging duration of the air injection on the pressure inside the pre-chamber 3.2.2 Scavenging the pre-chamber with air-fuel mixture For scavenging with an air-fuel mixture, the influence of various parameters on mixture formation and combustion is analyzed. In the investigation of air injection, it became clear that the residual gas content in the pre-chamber can be reduced at an early end of scavenging, but that it is significantly increased again in the compression as a result of the inflow of the main mixture. For this reason, the variation of the injection timing is examined in the first investigation step. Figure 13: Effect of varying EOS within the mixture scavenging on the residual gas fraction/ effect of extending this EOS on the pressure development inside the PC For this purpose, the end of scavenging is varied between the start of compression and 70° CA BTDC at a fixed injection duration of 10 ms and an HC ratio of 7% in Figure 13. The residual gas content in the pre-chamber is already slightly lowered during the intake stroke due to the scavenging gradients. Excessive inflow of the main mixture rich in residual gas is prevented by the late injection. In this study, a residual gas content of up to 20% can be achieved at an EOS of 650° CA compared to 30% in passive mode. A further delay of the EOS to 670° CA shows a significant deterioration in pressure 513 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="514"?> development and combustion progress in the pre-chamber on the right of Figure 13. The explanation for this is provided by the turbulent kinetic energy (TKE) distribution in Figure 14 for an EOS of 650° CA. Mixture flows out of the channels up to the end of scavenging. Shortly after the EOS, a flow reversal takes place via the orifices at 665° CA due to the pressure increase in the main combustion chamber during compression. As a result, the charge movement breaks down. It will then take some time for a certain level of TKE to develop again. If the EOS is delayed, there is not enough time to achieve a sufficient TKE level till ignition timing. Due to the poorer ignition conditions the combustion progress will consequently be delayed. It must therefore be noted that there must be sufficient time for mixture preparation and flow propagation between the end of injection and ignition timing in order to be able to create the most optimal ignition conditions possible. Figure 14: TKE level inside the pre-chamber after the EOS of 650° CA In order to clarify the question of whether a shorter injection period would also be possible, a variation between 5 ms and 12.5 ms is initiated in the CFD. With this investigation, the applied mass is also automatically varied in addition to the scavenging duration while maintaining the same system pressure. This results in the residual gas fraction in the pre-chamber, see Figure 15. Figure 15: Effect of varying the injection duration within the mixture scavenging on the residual gas fraction inside the pre-chamber 514 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="515"?> At first glance, there is only a slight difference of 2% between the injection, as before, with a duration of 10 ms and the shortened injection of 5 ms. Extending the scavenging duration up to 12.5 ms does not result in any additional advantages and will not be considered any further for reasons of overall engine efficiency. To better understand the effects of shortened scavenging, it is necessary to look again at the residual gas distribution in a cut along the y-axis in Figure 16. Figure 16: Residual gas content for 10 ms (top) and 5 ms (bottom) For injection durations of 5 ms and 10 ms, the residual gas distributions at 685° CA and at spark timing are compared. At both time steps, only slight differences can be seen in the pre-chamber. However, it is noticeable that the mixture in proximity to the pre-chamber has a significantly higher residual gas content when shortening the injection duration. This represents a big disadvantage. By scavenging over a longer period, a cloud of low-residual gas, fuel-rich mixture can be stored near the tip of the pre-chamber. Parts of this cloud are pushed back into the pre-chamber during compression. The entrainment at the free-jet outlet transports this mixture into the jet and thus improves the early flame phase of the main combustion. As a result, the pressure curve of the pre-chamber is shown in Figure 17. 515 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="516"?> Figure 17: Pressure curve in the pre-chamber when varying the scavenging duration This is now also the first comparison to passive operation. It should be noted at this point that the relatively long injection is accompanied by an increase in load that is no longer negligible. As a result, the load was slightly increased during passive operation, so that the same trapped total mass is obtained at the ignition timing in the simulation domain. This step provides a valid basis for comparison for all further investigations. The global air-fuel mixture continues to be kept constantly stoichiometrically. It is now clear that the ignition conditions and the pre-chamber combustion are significantly improved by the introduction of the mixture that the main combustion starts approximately 5° CA earlier and proceeds with significantly higher pressure gradients and a higher peak pressure in the further course. The very good results of the short injection can be further optimized by extending it to 10 ms. As could already be seen in the residual gas content, a further extension to 12.5 ms does not result in any noticeable added value, which is why the injection of 10 ms will be pursued in the further course of the process. As can be seen from Merker et al. , the laminar flame speed reaches its maximum at an equivalence ratio of approximatly Φ = 1.1. The previous investigations showed that this mixture could not yet be completely reached in the pre-chamber at the time of ignition. For this reason, the load change and combustion simulation with a variation of the HC ratios between 5% and 11% is now analyzed in the following consideration. End of scavenging and the injection duration are kept constant in this analysis. 516 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="517"?> Figure 18: Effect of varying the HC ratio of mixture scavenging on the EQR in the spark region Figure 18 shows the EQR curve of the four different variation steps in comparison to passive operation. The injection with a 5% mixture does not result in any differences in the EQR within the spark region at the time of ignition due to the low proportion. In compression, the mixture is diluted due to the inflowing main mixture. This results in the target value for the air-fuel mixture at the time of ignition for the 9% mixture. The other mixtures are either slightly too lean or slightly too fuel rich. As a result, the pressure profiles in the pre-chamber during combustion for the different mixtures are obtained by Figure 19. Figure 19: Pressure curve in the pre-chamber when varying the HC ratio The trends from the mixture assessment in Figure 18 continue within the pre-chamber, but the effect on combustion and thus the pressure curve is not as pronounced as might 517 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="518"?> have been expected. It is true that the shortest burning time and the shortest burning delay can be achieved at a ratio of 9% HC. However, the difference between 7% and 11% is very small. In case of the 7% mixture, the very good uniformity of the stoichiometric mixture can be cited as the reason for a rapid burn-out of the pre-chamber. The 5% ratio drops even more sharply, than expected. It should therefore be noted that there is a relatively wide range of variations in the HC content, in which the combustion process of the active scavenged pre-chamber can impress with a short burning duration and a short burning delay. Figure 20: Effect of the HC content of the scavenged mass on NOx raw emissions If the 7% threshold is exceeded, significant deteriorations can be recorded, though this is still a significant improvement compared to passive operations. Based on the findings of the CFD, the recommendation for the active scavenged 6-hole pre-chamber variant is that the greatest potential can be seen for mixture injection over 10 ms, at an injection pressure of 10 bar, with an end of injection shortly before the ignition time at 650° CA and with an HC content of the injected mass of 9%. However, initial tests on the test bench revealed that the increase by 1% in HC fraction from 7% to 8% result in 5% higher raw emissions of NO X . For that reason, the HC content should be held as low as possible. Therefore, the fraction of 7% is used moving forward. However, the analysis has already shown that there is only a very small difference in pressure development between 7% and 9%. Finally, it is important to be able to transfer this to other operating points in order to be able to develop a valid operating strategy. For this purpose, the operating point is examined at an engine speed of 1500 rpm and an IMEP of 6 bar. This point is still in region 1 defined in chapter 3.1, which is strongly characterized by high residual gas fractions in the pre-chamber. The same laws apply to the scavenging duration and the HC ratio of the mixture as for the lower-load operating point. The injection can thus be carried out over 10 ms with an HC content of 7%. The same applies to the 518 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="519"?> injection pressure of 10 bar, which is even more important with increasing load, since the final compression pressure at ignition timing increases with the load. With lower injection pressures, delayed ends of scavenging would therefore no longer be possible. The significantly improved pressure development in the pre-chamber as a result of active operation can also be seen at this examination point. The only difference for the application of the injection strategy results from a slight change in the end of scavenging. As a result of the increased engine speed, Figure 21 shows that the end of the injection must be slightly shifted by 20° CA in the direction of the earlier injector connection. Figure 21: Pressure curve in the pre-chamber when varying the EOS at engine speed of 1500 rpm The effect is due to the increased engine speed and the resulting shortened time period between EOS and spark timing. Therefore, there is not enough time to develop a correspondingly high TKE level inside the pre-chamber. This effect has already been explained based on the lower-load operating point. At the end of this chapter, the operating strategy can consequently be formulated again as a summary. For the two operating points investigated, the injection pressure of 10 bar and scavenging for 10 ms with a 7% HC ratio are generally valid. The end of scavenging alone must be shifted by 20° CA in the early direction when the engine speed is increased by 500 rpm. 3.3 Potential analysis of the active pre-chamber system in 1D simulations In order to analyze the effect of the active mixture scavenging into the pre-chamber on global operating parameters, the 1D simulation was initiated in a full model for the considered engine map area. The operating strategy from the previous numerical investigations are adopted to this analysis. 519 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="520"?> Figure 22: Comparison of the burning duration between passive and active pre-chamber The effect of active injection on flame propagation and combustion efficiency can best be seen in Figure 22 by comparing the burning durations between passive and active pre-chamber. The three areas from Chapter 3.1 are shown again. It can be observed that the reduced residual gas fraction and the mixture enrichment in the pre-chamber can significantly improve combustion in region 1. The improved ignition conditions result in rapid and intensive burn-out of the pre-chamber, which allows the free jets to emerge into the main combustion chamber with a high pulse. This leads to the potential of halving the burning duration in the area that was dominated by residual gas problems in passive mode. The transition region extends above this region 1, where active pre-chamber injection could still provide a slight advantage. In region 3, there is no need of an active scavenging due to the lack of advantages compared to passive operations. In this region the influence of the residual gas fraction inside the pre-chamber does not play a dominating role. It is particularly advantageous from an engine calibration standpoint if the active pre-chamber scavenging must only be optimized and designed for a small area. These promising results from the active pre-chamber show a very high potential in 1D and 3D simulations and need to be proved in final test bench investigations. 520 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="521"?> 3.4 Transfer of simulation results to the test bench investigations Following the simulative investigations and numerous findings, it is important to finally transfer these results to the test bench investigations. For this purpose, operation with the conventional spark plug is compared to operation with a passive, actively air-scavenged and actively mixture-scavenged pre-chamber. The previously developed scavenging strategy is used below for active mixture injection. In Figure 23 the change of variation of the IMEP is analyzed for these four operating strategies when varying the intake and exhaust phasing and by that changing the residual gas fraction inside the main combustion chamber. The configuration of inlet and outlet phasing, which was examined in the CFD, can be seen in the boxes. Figure 23: Comparison of the CoV of IMEP of the different operating strategies when varying the residual gas fraction inside the main combustion chamber by cam phasing The basis for comparison is the measurement with the conventional spark plug. Com‐ bustion stability is relatively similar across the entire test area and decreases towards the misfire limit where high residual gas fractions impede a reliable combustion initialization. The increasing residual gas content has an even larger impact on the performance of the passive pre-chamber. The reason for this lies in the enrichment of the residual gas in the passive pre-chamber, which even exceeds the content of the main combustion chamber. It becomes clear that misfire limit starts significantly earlier. In the presentable operating area, however, the variance of IMEP is at the same level as that of the regular spark plug. As has already been shown by CFD studies, scavenging the pre-chamber with air cannot provide any noticeable advantages. At 521 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="522"?> the misfire limit, the variance is rather higher than that of the passive pre-chamber and the active air scavenging is not able to extend the operating range. The promising CFD findings of mixture scavenging can also be finally confirmed in the results of the engine test. Over the complete investigated region, the combustion stability can be significantly increased. In comparison with the regular spark plug an even lower coefficient of variation of IMEP is achievable. For this reason, changes in valve timings could also be realized, which could provide a further increase in efficiency. However, the misfire limit could not be extended in this first attempt. Here, in further engine tests, a minimal adaptation of the scavenging strategy to the changed camshaft timings can be investigated or the effect of adjusting the pre-chamber geometry can be analyzed. 4 Conclusion In this publication, the importance of both 1D and 3D simulation for the development process of the pre-chamber system has been highlighted. Detailed modeling and validation of a 1D model for engine process simulations enabled the identification of potentials and limits of the passive pre-chamber over a wide range of the engine map. Three characteristic areas were identified: an area where active mixture scavenging is mandatory due to high residual gas fractions, a transition area and an area at higher loads where the pre-chamber can be operated passively. Developing a strategy for pre-chamber scavenging was done in an extensive CFD study. The results are based on a complex modeling process, which includes detailed combustion and ignition modeling. Ignition modeling was realized by implementing the AVL CADIM in FIRE TM . This model provides the ability to represent the arc between the electrodes by implementing Lagrangian parcels, which are deflected by the prevailing flow field at ignition timing. As a result, the ignition location and the ignition delay are calculated by the model depending on the mixture composition and flow conditions, which is of enormous importance for pre-chamber modeling due to widely varying ignition conditions. In combination with the ECFM-3Z combustion model the CFD modeling approach could be validated by using recordings of an optically accessible test engine. After extensive CFD investigations an operating strategy was formulated. For the two operating points investigated, the injection pressure of 10 bar and scavenging for 10 ms with a 7% HC ratio show the highest potential in increasing the pre-chamber performance. The end of scavenging alone must be shifted by 20° CA in the early direction when the engine speed is increased by 500 rpm. In conclusion, the excellent results of the actively mixture-scavenged pre-chamber were confirmed in an engine test. As a result, the active pre-chamber could be operated in the same range compared to the spark plug. However, the combustion stability was significantly increased. The limitation of the passive pre-chamber with an increase in residual gas content was illustrated and the rather disappointing results of the air scavenged pre-chamber from the CFD were also confirmed in engine tests. Scavenging the pre-chamber with an air-fuel mixture is therefore essential for operating the engine at low load. 522 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="523"?> References [1] Baumgartner, L. S., Wohlgemuth, S., Zirngibl, S., and Wachtmeister, G. 2015. Investigation of a Methane Scavenged Prechamber for Increased Efficiency of a Lean-Burn Natural Gas Engine for Automotive Applications. SAE Int. J. Engines 8, 2, 921-933. [2] Biswas, S. 2018. Physics of Turbulent Jet Ignition. Mechanisms and Dynamics of Ultra-Lean Combustion. Springer Theses Ser. Springer International Publishing AG, Cham. [3] Blankmeister, M., Alp, M., and Shimizu, E. 2018. Passive Pre-Chamber Spark Plug for Future Gasoline Combustion Systems with Direct Injection. In Ignition Systems for Gasoline Engines. Internationale Tagung Zündsysteme für Ottomotoren, M. Günther, Ed. expert verlag, Tübingen. [4] Cooper, A., Harrington, A., Bassett, M., Reader, S., and Bunce, M. Application of the Passive MAHLE Jet Ignition System and Synergies with Miller Cycle and Exhaust Gas Recirculation. SAE International 2020. DOI=10.4271/ 2020-01-0283. [5] Getzlaff, J., Pape, J., Gruenig, C., Kuhnert, D., and Latsch, R. Investigations on Pre-Chamber Spark Plug with Pilot Injection. SAE International 2007. DOI=10.4271/ 2007-01-0479. [6] Gussak, L. A. 1975. High Chemical Activity of Incomplete Combustion Products and a Method of Prechamber Torch Ignition for Avalanche Activation of Combustion in Internal Combustion Engines. Society of Automotive Engineers, Inc. [7] Kalghatgi, G., Agarwal, A. K., Leach, F., and Senecal, K., Eds. 2022. Engines and Fuels for Future Transport. Springer eBook Collection. Springer Singapore, Singapore. [8] Kawabata, Y. and Mori, D. Combustion Diagnostics & amp; Improvement of a Prechamber Lean-Burn Natural Gas Engine. SAE International 2004. DOI=10.4271/ 2004-01-0979. [9] Laget, O., Chevillard, S., Pilla, G., Gautrot, X., and Colliou, T. Investigations on Pre-chamber Ignition Device Using Experimental and Numerical Approaches. SAE Technical Paper Series 2019 2019. DOI=10.4271/ 2019-01-2163. [10] Merker, G. P. and Teichmann, R. 2019. Grundlagen Verbrennungsmotoren. Springer Fachme‐ dien Wiesbaden, Wiesbaden. [11] Pyszczek, R., Hahn, J., Priesching, P., and Teodorczyk, A. 2020. Numerical Modeling of Spark Ignition in Internal Combustion Engines. Journal of Energy Resources Technology 142, 2. [12] Russwurm, T., Schumacher, M., and Wensing, M. Active Fuelling of a Passenger Car Sized Pre-Chamber Ignition System with Gaseous Components of Gasoline. SAE Technical Paper Series 2020 2020. DOI=10.4271/ 2020-01-2045. [13] Russwurm, T., Wensing, M., Euchner, L., and Janas, P. 2021. Flame Luminesce in an Optically Accessible Engine with an Active Fuelled Pre-Chamber Ignition System. In 21. Internationales Stuttgarter Symposium. Automobil- und Motorentechnik, M. Bargende, Ed. Proceedings. Springer Fachmedien Wiesbaden, Wiesbaden. [14] Schumacher, M. and Wensing, M. A Gasoline Fuelled Pre-Chamber Ignition System for Homogeneous Lean Combustion Processes. SAE Technical Paper Series 2016 2016. DOI=10.4271/ 2016-01-2176. [15] Sens, M., Binder, E., Benz, A., Krämer, L., and Blumenröder, K. 2018. Vorkammerzündung als Schlüsseltechnologie für hocheffiziente Ottomotoren - neue Ansätze und Betriebsstrategien. 39. Internationales Wiener Motorensymposium 2018. In 39. Internationales Wiener Motoren‐ symposium 26.-27. April 2018. Band 1: erster Tag; Band 2: zweiter Tag / 39th International Vienna 523 Developing an operating strategy of an active scavenged pre-chamber system for gasoline engines running at a stoichiometric air-fuel ratio <?page no="524"?> Motor Symposium 26-27 April 2018 / in two volumes. Volume 1: first day; Volume 2: second day, B. Geringer and H.-P. Lenz, Eds. Verkehrstechnik/ Fahrzeugtechnik 807. VDI Verlag, Düsseldorf. [16] Sens, M., Günther, M., Medicke, M., and Walther, U. 2020. Der Weg zum Ottomotor mit 45% Wirkungsgrad. MTZ - Motortechnische Zeitschrift (Apr. 2020). [17] Shapiro, E., Ahmed, I., and Tiney, N. 2018. Advanced Ignition Modelling for Pre-chamber Combustion in Lean Burn Gas Engines. In Ignition Systems for Gasoline Engines. Interna‐ tionale Tagung Zündsysteme für Ottomotoren, M. Günther, Ed. expert verlag, Tübingen. DOI=10.5445/ IR/ 1000088322. [18] Tang, W. and Sarathy, M. Investigate Chemical Effects of Pre-Chamber Combustion Products on Main Chamber Ignition Performance under an Ultra-Lean Condition. SAE Technical Paper Series 2020 2020. DOI=10.4271/ 2020-01-2001. [19] Wang, N., Liu, J., Chang, W., and Lee, C.-F. The Effect of In-Cylinder Temperature on the Ignition Initiation Location of a Pre-Chamber Generated Hot Turbulent Jet. SAE International 2018. DOI=10.4271/ 2018-01-0184. [20] Wellander, R., Rosell, J., Richter, M., Alden, M., Andersson, O., Johansson, B., Duong, J., and Hyvonen, J. 2014. Study of the Early Flame Development in a Spark-Ignited Lean Burn Four-Stroke Large Bore Gas Engine by Fuel Tracer PLIF. SAE Int. J. Engines 7, 2, 928-936. [21] Wenig, M. and Roggendorf, K. Development of a Predictive Dual Fuel Combustion and Prechamber Model for Large Two Stroke Engines Within a Fast 0 D/ 1 D Simulation Environment. 29th CIMAC World Congress, Vancouver. 524 L. Euchner, L. Baumgartner, M. Wensing, T. Russwurm, M. Sc., P. Janas <?page no="525"?> Holistic knock detection and control as the key to optimum ignition timing Matthias Biehl / Marc Benzinger Robert Bosch GmbH Abstract: An engine operation at the best possible ignition timing is the objective of the holistic knock detection and control approach of the Robert Bosch Group. To achieve this target a combination of multiple functions works together to determine the knock intensity precisely, to protect the engine sufficiently, and to advance the ignition timing if possible. This combination of detection and control functions attempts to achieve the best possible ignition timing without reducing engine durability. In line with the actual and possible future emission legislations, and their verification, an engine operation at the optimum possible needs to be the objective. Furthermore, this fulfillment needs to be verified for a much wider spectrum of operation points than ever before. The holistic approach combines a reliable knock detection with the Model Based Knock Detection, a continuous knock control adaption and an ignition advance. Introduction Emissions and fuel consumption are more important than ever before, therefore, optimizing the ignition timing has an essential role in minimizing efficiency losses. To achieve this target a combination of multiple functions works together to determine the knock intensity precisely, to protect the engine sufficiently, and to advance the ignition timing if possible. After explaining the state of the art, the individual processes which, in combination, lead to an improved ignition angle will be examined. State of the art The Model Based Knock Detection (MBKD) is today the commonly used knock detection method for structure-borne noise sensors on engine control units (ECU) of the Robert Bosch Group worldwide. There are different variants of MBKD available to ensure a reliable knock detection according to the engine complexity and the <?page no="526"?> required quality regarding the determination of the knock intensity. Steps of the MBKD development are given in [1] and [2]. Based on the knock intensity, the time intervals between knock events, and the required knock control interventions to ensure a safe engine operation, a demand-ori‐ ented control reduces the ignition timing retardations to a minimum possible. In combination with a continuous knock control adaption, which allows a smooth transition between the operation points, the control ensures a reliable engine operation for regular and worse case conditions. Model Based Knock Detection The MBKD is available in three different variants (BASIC, ADVANCED and PEAK), shown in Figure 1, with a comparison regarding the classic knock detection method with a signal filtering in the time domain. Fig. 1: MBKD variants and the Classic Knock Detection approach The MBKD variants differ in their complexity and in the calculated knock feature. All the variants have the same signal acquisition, based on a time limited measuring window, and a raw signal processing via Fast Fourier Transformation (FFT). Also, the functionality to filter the raw signal in the frequency domain is equal for all the variants. The differences between the methods are primarily the result adaption. This optimization of the knock feature is not given for the variant BASIC, to reduce the complexity of the functionality for engines with a good knock sensor signal, which is free of disturbing noise, or engines with lower requirements regarding the knock detection quality. The variant ADVANCED is using the feature result adaption to reduce disturbing noises and to increase the signal noise ratio between knocking and non-knocking combustions in a way to improve the detection quality. Both variants, BASIC and ADVANCED, deliver the calculated knock intensity as a ratio between the actual combustion and the previous non-knocking ones. 526 Matthias Biehl / Marc Benzinger <?page no="527"?> The PEAK variant is additionally using the result adaption for converting the knock feature in the domain of cylinder pressure and provides the knock intensity in the unit [bar]. An example of the result of the MBKD PEAK variant is given in Figure 2. Fig. 2: Example of the MBKD PEAK result Demand-oriented knock control Based on MBKDs improved knock detection quality and the higher accuracy regarding the determination of the knock intensity, a demand-oriented knock control interven‐ tion is used in today’s software of the Robert Bosch Group. The demand-oriented knock control strategy combines the accurate information of the knock intensity with the information of the cylinder individual time intervals between several knock events (see Fig. 3). Fig. 3: Demand-oriented knock control 527 Holistic knock detection and control as the key to optimum ignition timing <?page no="528"?> This allows the knock control to differentiate between operation conditions with a low amount of knock events, with a knock tendency which is known from nominal condi‐ tions, or high knock tendencies which are given for example with a low fuel quality. Furthermore, this approach allows to consider differences in the knock tendency of each induvial cylinder and, therefore, to optimize control interventions individually. This improved functionality, compared to the long time used static control parameters, allows the demand-oriented control approach, without the need of focusing on the worst-case conditions and the cylinder with the highest knock tendency, to define the control parameters for a safe engine operation (see Fig. 4). Fig. 4: Example of ignition retardation with the demand-oriented knock control Continuous knock control adaption Compared to classic adaptation algorithms, which only trains the current operating point, the continuous knock control adaption uses a surface learning algorithm. This algorithm is capable of training adjacent operating points accordingly. Hence, when approaching the same operating point for a second time, the algorithm pre-controls the previously learned value either proportionally or completely, depending on the adaptation strategy. Smooth transitions will result, as the surface adapts in two dimensions (speed and load). Neighboring operating points in terms of load and speed are memorized proportionally prior to driving into them. Thus, operating points are pre-controlled even without having driven in them beforehand. While the range of the 528 Matthias Biehl / Marc Benzinger <?page no="529"?> surface is variable as well, it is only practical if adjacent regions do not differ too much from each other, with respect to their knock tendency. Figure 5 visualizes the comparison between the standard adaptation approach and the improved surface learning algorithm. Consequently, compared to the classic adaption approach, advantages of the surface learning algorithm are that the system is both adapted more rapidly and across a larger area in the engine characteristic map. Along with the demand-oriented knock control, the engine is thus tailored to its individual knock intensity and frequency. Combining these two functions thereby ensures engine protection while minimizing ignition angle interventions and offering a best possible pre-control. Fig. 5: Surface Learning Algorithm 529 Holistic knock detection and control as the key to optimum ignition timing <?page no="530"?> Future-oriented development The previous described safe and reliable knock detection with MBKD, as well as the demand-oriented and continious control for regular and worse case conditions, achieved with all optimizations of the last years, leave one question unanswered “Which advantages can be achieved if the conditions are more beneficial than nominal condition? ” To answer this question the Robert Bosch Group focused their development on the field of knock detection and control to optimize the iginition timing also under beneficial conditions (e.g., higher fuel quality, lower intake air temperature, higher humidity). The approach and the results of this development are introduced in the following. Iginition advance algorithm For an ignition advance to work reliably without damaging the engine or worsening fuel consumption, three boundary conditions must be considered. These boundary conditions are the centre of combustion, the acceptable mean or maximum cylinder pressure and the knock frequency. A centre of gravity of the combustion which is at a crank angle of approximately eight degrees [3] is defined as thermodynamically optimal due to having least inefficiencies [4]. Consequently, the lowest specific fuel consumption is expected at the aforementioned crank angle, since earlier ignition results in cylinder wall dissipation and retarded ignition results in energy loss due to exhaust gas temperature dissipation [5]. At low engine loads, ignition already occurs at the ideal spark timing. Consequently, no advantages arise at these operating points. However, as load increases, spark timing diverges from the optimum, which is due to ignition calibration based on its target fuel (see Fig. 6). As a consequence, ignition timing is applied to the target fuel according to the knock limit, taking into account the various degrees of freedom mapped in the Motronic software, e.g., coolant or ambient temperature. Based on the knock frequency in combination with the early ignition algorithm, component tolerances, interpolation or extrapolation impairments, or degrees of freedom not mapped in the engine control system can thus be compensated for. 530 Matthias Biehl / Marc Benzinger <?page no="531"?> Fig. 6: MFB50 difference with RON95 Higher cylinder pressure arises at an earlier spark timing. However, to avoid affecting the durability of the engine, this increased cylinder pressure must not exceed the component limits at high loads and especially at high engine speeds. The difference between the engine characteristic map without advanced ignition timing and the one with advanced ignition timing shows the area where spark timing was not at the optimum in green (see Fig. 7). Especially in the low-end torque range, the comparison illustrates that pressure differences up to 16 bar are achievable with ideal spark timing, which takes the calculated knock result as a reliable criterion. For fuels with a higher octane rating than target fuel, the knock limit shifts further towards maximum load. To ensure that the engine is not purely limited by the knocking criterion, a thermodynamically ideal ignition timing is determined with high-octane fuel during knock-free operation. In this case, the limitation is only between either the MFB50 or the maximum permissible average cylinder pressure of all cylinders PMAX0m. Figure 8 shows the consumption advantages for an identical fuel (RON95) in green. These benefits relate to steady-state operating points in stabilized condition and reach up to 8%. However, measurement tolerances which occur in the measurement must be considered. At low loads (see yellow area of Figure 8), the MFB50 is already in its optimum. However, discrepancies are present in these loads due to uncertainty in the measurement. Additional fuel-saving potential arises with superior octane rating. In dynamic operation, these advantages are also available, but depend on an adaptation strategy and its associated pre-control. 531 Holistic knock detection and control as the key to optimum ignition timing <?page no="532"?> Fig. 7: PMAX0m difference with RON95 Fig. 8: Fuel consumption benefits in steady state operation with target fuel Results With better quality fuels regarding octane rating, in this case RON101, the engine operates even more efficiently, as shown in Figures 9 and 10. Figure 9 shows the MFB50 area in green, where ignition is not yet at its optimum value. 532 Matthias Biehl / Marc Benzinger <?page no="533"?> The torque model ensures that torque homologated for the engine is not exceeded. Due to the optimized ignition angle, the torque model in turn reduces maximum air charge at full load. With RON101, maximum relative air charge is only about 160%, while for RON95 it is almost 175%. This optimized ignition angle results in fuel savings, which is visualized in Figure 11. Compared to target fuel consumption efficiencies account up to a maximum of 12% including measurement uncertainties. Ultimately, the ECU holistically handles situations ranging from best to worst-case scenarios, thereby targeting that the engine always operates at optimum ignition timing. Fig. 9: MFB50 improvement with RON101 compared to RON95 533 Holistic knock detection and control as the key to optimum ignition timing <?page no="534"?> Fig. 10: Torque improvement with RON101 compared to RON95 Fig. 11: Fuel consumption improvement with RON101 compared to RON95 Conclusion The combination of functionalities to the holistic knock detection and control approach shows clear benefits, and, furthermore, allows a change of mind regarding the objective 534 Matthias Biehl / Marc Benzinger <?page no="535"?> of the the knock detection and control. The sole function of the knock detection and control to protect the engine is outdated. The holistic knock detection and control approach evolves the classic knock detec‐ tion further to a real optimizer of the ignition timing under all operation conditions, compensating environmental influences, as well as system tolerances and ageing effects of the engine. The potential that has not been used for generations of internal combustion engines is now exposed. It would be reprehensible to leave it unused for modern and complex engines, especially under today’s political pressure and our social responsibility. Prospect The holistic knock detection and control approach has to prove its suitability and generalizability at different engine projects to ensure the methods eligibility under mass production conditions and series tolerances. This will also reveal the hidden potentials for optimization of today’s base engine calibration and can support to reduce the effort of base ignition timing calibration for different operation conditions. Furthermore, upcoming technologies like pre-chamber ignition, as well as hydrogen or synthetic fuels, will corroborate the overall benefits of this approach. References Biehl, M., Perless, E., Sloboda, R.: Artificial Intelligence for knock detection. In: 4, International Conference on Knocking in Gasoline Engines, pp. 331-346 (2013) Biehl, M., Meister, M.: Model Based Knock Detection. In: 5, International Conference on Knocking in Gasoline Engines, pp. 257-266 (2017) Bargende, M.: Schwerpunktkriterium und automatische Klingelerkennung. In: MTZ - Motor‐ technische Zeitschrift, p. 633 (1995) Thöne, H. J.: Untersuchung von Einflussgrößen auf das Klopfen von Ottomotoren unter besonderer Beachtung der internen Abgasrückführung. Techn. Hochschule, Aachen, p. 44 (1994) Spicher, U., Worret, R.: Entwicklung eines Kriteriums zur Vorausbe-rechnung der Klopfgrenze. [Vorhaben Nr. 700], Abschlussbericht. Frankfurt am Main: FVV (H. 741), p. 29 (2002) 535 Holistic knock detection and control as the key to optimum ignition timing <?page no="537"?> Knock Probability Prediction and its Potential for a Knock Control Application Nicolas Fajt, IFS - Institut für Fahrzeugtechnik, Universität Stuttgart Co-authors: M. Grill, M. Bargende Abstract: This paper presents two simulative approaches for the prediction of the knock frequency of a SI engine at the example of a single cylinder research engine. To calculate the knock frequency, a half-predictive calculation approach is utilized, that was recently published and developed based on measurement data. This calculation relies on the auto-ignition onset distribution for the single working cycles of an operating point. In order to make the calculation predictively useable within simulations, two approaches are presented. The two simulative approaches include two different methods to simulate single working cycles to which subsequently a state-of-the-art auto-ignition model is applied to determine the auto-ignition onset. Based on the auto-ignition onsets the knock frequency is calculated. The results show high accuracy of the predicted knock frequencies in comparison to the measured knock frequency for both simulation methods and significant differences in the required simulation times. A comparison to various knock control strategies shows the potential of knock frequency based knock control to increase efficiency and decrease emissions and highlights the necessity of a high calculation performance for an engine application. 1 Introduction One major focus within the development of today’s spark ignition engines is the con‐ stant reduction of CO 2 emissions. Common measures, like increasing the compression ratio or increasing the boost pressure of turbocharger systems are limited by the occurrence of knocking combustion. Therefore, reliable knock models, used within cost-efficient 0D/ 1D simulations, are crucial for the development of new engines. In this context, the presented results focus on 0D/ 1D modeling and were acquired within the scope of the FVV project “Fast Knocking Prediction for Gasoline Engines”, which has the goal to develop a methodology for a predictive, fast and robust 0D-simulation tool accounting for knocking combustion, cycle-to-cycle variations and the stochastic nature of these phenomena. <?page no="538"?> The objective of current 0D/ 1D knock models is to predict the knock-limited spark advance (KLSA), in other words the knock boundary for the specific operating conditions of an engine. Therefore, spark timing is varied and knock models (e.g. [1-4]) evaluate if an operating point is below or above the knock boundary. This knock boundary is usually defined as percentage or range of percentage of knocking engine cycles for a set of measured engine cycles, typically between 1% and 10% [1, 2, 5]. The introduced simulation approach aims for a prediction of the actual value of the knock frequency for specific operating conditions. This provides further details about the current operating conditions regarding engine knock compared to the evaluation if an operating point is below or above its knock boundary. A recently published approach for calculating the knock frequency is used, in which the ability of calculating the knock frequency based on three parameters gained from measured single working cycles was shown [6], [7]. In order to predict the knock frequency, by applying this approach to the simulation, single working cycles need to be simulated with respect to the prevailing combustion cycle-to-cycle variations. Therefore, two different methods are presented. 2 Knock Probability Calculation Method 2.1 Three-Parameter-Approach The initial method to calculate the knock probability was introduced by Hess [6], [7]. Commonly, the knock probability is determined based on measurement data by analyzing oscillations within the pressure trace signal to identify knocking cycles within a set of measured cycles. Although developed based on measurement data, the three-parameter-approach is a first step towards a method to predictively calculate the knock frequency and thereby allow it to be used within simulations. Since the calculation is based on three parameters, the method was named three-parameter-ap‐ proach. For the development of the method, measured single working cycles (SWC) of a single-cylinder engine were analyzed. The general method is present in Fig. 1. As the top right figure shows, the basis is a measurement run that contains five operating points (OPs) that differ in spark timing while all other operating conditions are similar. Each of the five operating points contains measured in-cylinder pressure of 500 single working cycles. For all single working cycles, the auto-ignition onset is determined by the latest version of a state-of-the-art two-stage auto-ignition model [8]. The required input for the auto-ignition model is obtained by performing a pressure trace analysis (PTA) for each single working cycle. As first step, the number of knocking single working cycles for the operating point with the highest measured knock frequency and earliest spark timing is calculated from the knock frequency and the number of measured single working cycles. In the presented example, this calculates to 64 knocking cycles for 500 measured cycles at a measured knock frequency of 12.8%. Based on this number an auto-ignition limit (AI limit ) is set as shown in the top left diagram of in such way that exactely 12.8% or 64 cycles (marked red) have their auto-igintion onset (AI-onset) before the set limit. AI limit is one of the three required 538 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="539"?> Fig. 1: 3-Parameter-Approach for calculation of the knock frequency, from [6] parameters and constant for the calculation of the knock frequency for all spark timing variations. The second step is applied to all operating points / spark timings of the measurement run and illustrated for OP 4 in the bottom left diagram. Instead of counting the number of single working cycles with AI-onset before AI limit , the mean AI-onset (AI mean ) and the standard deviation (σ AI ) is calculated from the distribution of AI-onsets for each operating point. As third step, the difference between AI limit and AI mean is then calculated as multiple of the respective standard deviation σ AI . In the fourth and final step, the knock frequency is calculated as the probability that values of a Gaussian distribution deviate by more than the multiple of σ AI from AI mean (as shown in the bottom right figure). The calculation was performed for 70 operating points in total and as shown in detail in [6], [7] and yielded accurate results for the knock frequency. To summarize, within the three-parameter-approach, the calculation of the knock frequency is based on the three parameters AI limit , AI mean and σ AI , which are derived from the AI-onset distribution of an operating point. The introduction of the calculation shows that single working cycles need to be simulated to be able to predictively use the three-parameter-approach in the simulation. This requirement to simulate SWCs is a major difference between the determination of the knock boundary and the determination of the knock frequency, since the knock boundary (KLSA) determination in knock models like [1-4] is based on evaluation of 539 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="540"?> the average working cycle of an operating point. For this reason, in this paper, two approaches to predict the knock frequency with the three-parameter-approach are introduced that are based on different methods to simulate single working cycles. In order to be able to evaluate the predicted knock frequencies and the simulation of single working cycles, three measurements that were used for the development of the three-parameter-approach are selected for simulation. This way, measurement data for each of the 500 single working cycles for all investigated operating points is available. As shown in Tab. 1, the operating conditions include a variation of the engine speed and the indicated mean effective pressure (IMEP). With each combination of engine speed and load containing five operating points with different spark timings while all other operating conditions are similar, 15 operating points are investigated in total. Engine Speed in RPM Indicated mean effective pressure IMEP in bar 2500 12 16 1500 - 16 Tab. 1: Operating conditions covered by the simulations. For the simulations of the single working cycles, a quasi-dimensional combustion model [9] within a real working-process calculation tool [10] is used. 2.2 Simulation of single working cycles by modelling a turbulence level distribution For SI engines large variations of the combustion are characteristic. This leads to the fact that on the real engine parameters such as burn duration, center of combustion, maximum pressure and IMEP also vary from working cycle to working cycle. As mentioned in the previous chapter, it is therefore required to simulate the cyclic fluctuations to subsequently predict a knock frequency. One simulation approach is to model a global turbulence level distribution to vary the flame propagation speed. This will result in fluctuating engine load, center of combustion, pressure gradient, maximum pressure and combustion efficiency and thereby replicate the cycle-to-cycle variations. As initial step, the simulation model is set up as usual to be able to simulate the measured average working cycles of all five operating points included in the run. Anal‐ ogous to the first step of the 3-Parameter-Approach, the turbulence level distribution is set up for the operating point with the highest measured knock frequency of the run. The utilized turbulence model calculates the progress of the turbulent kinetic energy based on a simple k-ε model and a start value for the turbulent kinetic energy that can be adjusted by the user via a scaling factor. For further details about the turbulence calculation refer to [11], [12]. Different turbulence levels in this work are realized by variation of this scaling factor, which varies the start value of turbulent kinetic energy. Therefore, the modelled distribution consists of 500 individual values for the scaling factor (the same number as measured single working cycles), as exemplarily visualized 540 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="541"?> in the left plot of Fig. 2 for 2500 rpm / 16 bar. This distribution is set up to precisely match the AI-onset distribution of the measured cycles. In Fig. 2 on the right side such AI-onset distribution is exemplarily visualized, but not for the operating point with the highest knock frequency for which the scaling factor distribution was set up, but for an operating point with lower knock frequency at 2500 rpm / 16 bar. This shows that the measured cycle-to-cycle variations can be precisely calculated although the scaling factor distribution is only calibrated at the operating point with highest knock frequency. Fig. 2: Scaling factor distribution to vary the turbulence level and auto-ignition onset distribution of the simulated single working cycles In order to determine the AI-onsets, the single working cycles are simulated by applying each scaling factor value of the distribution to a simulation, while all other calibrated simulation model parameters are kept constant. Thereby, 500 simulations are performed for one operating point. Subsequently, the simulation results are used as input for the same auto-ignition model that was used for development of the three-parameter-approach to calculate the AI-onset of every simulated single working cycle. Fig. 3: Cylinder pressure trace of simulated single working cycles using a turbulence level distribution in comparison to measured single working cycles. 541 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="542"?> In Fig. 3 the pressure traces of the simulated single working cycles in comparison to the measured cycles are shown for the operating point with highest measured knock frequency at 2500 rpm / 16 bar for which the turbulence level distribution was set up. The comparison indicates that the simulated cycles using the modelled turbulence level distribution match the measured CCV very well. In the next step, the turbulence level distribution, calibrated for the operating point with the highest knock frequency is applied to the simulations of the remaining operating points with different spark timings and lower knock frequency (here four other OPs) without additional calibration. Thus, allowing the simulation of 500 single working cycles at all operating points of an engine speed and load combination with a turbulence level distribution that is calibrated only at one operating point. As final step of the SWC simulation, similarly to the first operating point, the auto-ignition model is utilized to calculate the AI-onset of each simulated single working cycle. With the AI-onset distribution determined for every operating point, the 3-Param‐ eter-Approach can be applied. Since the input data with 500 AI-onsets for every operating point is similar to the input data used for development of the three-param‐ eter-approach, the determination of the three parameters AI limit , AI mean and σ AI as well as the final calculation oft he knock frequency is identical: 1. AI limit is determined from the AI-onset distribution. 2. AI mean is the arithmetical mean of the AI-onsets of each operating point. 3. σ AI is the standard deviation of the AI-onsets of each operating point. 4. Knock frequency = Probability that values of a Gaussian distribution exceed (AI limit - AI mean ) / σ AI . 2.3 Simulation of single working cycles by application of a CCV-model The second approach is to use an available model [13], [14] to calculate the cycle-to-cy‐ cle variations. Developed by Wenig, this model provides a phenomenological approach to predict cyclic fluctuations of the combustion using the same quasi-dimensional combustion model [9] and working process calculation tool [10] that is used for simulation of the mean cycles. With the application of this CCV-model, the number of single working cycles that have to be simulated can be reduced significantly, which also reduces calculation time significantly. In this work, the functionality of the CCV-model is described only briefly to give a basic understanding about the modeling approach and the results that are used in this work for the prediction of the knock frequency. Detailed information about the model can be found in [13], [14]. Within the model a fluctuation factor is applied on the laminar flame velocity to alter the flame propagation speed and an inflammation phase fluctuation is applied to account for the variation of the initial 0-10% burn duration. For each variation, a working process simulation is performed representing a single working cycle. One important assumption embedded in the model is that the cycle-to-cycle variation can be calculated by weighted evaluation of the results of a reduced amount of single working cycles, thus reducing the computational effort. That means a cycle-to-cycle parameter 542 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="543"?> as for example the standard deviation of the IMEP or the center of combustion can be calculated from only 15 working process calculations, whereas the previously presented SWC simulation approach calculated 500 cycles. Fig. 4: Cylinder pressure trace of simulated single working cycles using a CCV-model in comparison to measured single working cycles. Similar to the first SWC simulation approach, the CCV-model is calibrated at the operating point with the highest measured knock frequency to match the measured standard deviation of the IMEP. To all further operating points of one run the CCV-model is then applied without additional calibration. In Fig. 4 the pressure traces of the 15 calculated cycles in comparison to the 500 measured cycles are shown for the operating point with the highest knock frequency at 2500 rpm / 16 bar. The simulated pressure traces are well within the scatter of the measured single working cycles while having less maximum variation than the measured cycles. This is expected, as the total cyclic variation is determined by weighting the results of the simulated cycles. As final step, the auto-ignition model is used, as in the first method, to calculate the AI-onset for each of the 15 simulated single working cycles. Compared to the 500 cycles within the first simulative approach, now only 15 single working cycles are simulated, reducing the overall simulation duration for one operating point significantly, but due to this reduced amount of data, the strategy to determine AI limit , AI mean and σ AI has to be modified. For all operating points, instead of calculating the arithmetical mean of all AI-onset values, AI mean is calculated as weighted mean using the same weights as in the CCV-model. Equally, the standard deviation σ AI is calculated considering the weights. Due to the small amount of available values, AI limit can only be determined with a maximum resolution of 1/ 15, which corresponds to a knock frequency resolution of 6.67% using the same method as the two previous approaches. To overcome this issue, AI limit is calibrated iteratively at the operating point with the highest measured knock frequency. Therefore, a start value for AI limit is initialized at a very early timing, defined as AI mean - 8 σ AI and increased in small increments until the calculated knock frequency using the 3-Parameter-Approach matches the measured knock frequency. This ensures 543 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="544"?> an accurate calibration of AI limit and thereby the knock frequency at this operating point, without limitations to the resolution as described above. 3 Results and Discussion In Fig. 5, the calculated knock frequency in comparison to the measurement data is shown for both simulative approaches and all three engine speed and load combina‐ tions. For reasons of readability, in the following, the simulative approach modelling a turbulence level distribution will be referred to as the first simulation method and the simulative approach utilizing the CCV-model will be referred to as the second simulation method. Fig. 5: Calculated knock frequency in comparison to the measured knock frequency for simulation approach using a turbulence level distribution (a) and using a CCV-model (b) 544 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="545"?> Generally, the predicted knock frequency results are in good agreement with the measurement results for all engine speed and load combinations and both simulation approaches. Best agreement and largest deviation are both found within the second simulation method for 2500 rpm / 16 bar and 2500 rpm / 12 bar respectively. For a detailed analysis of the accuracy of the results, the deviation in center of combustion of the predicted knock frequencies from the measurement results is evaluated. This is exemplarily visualized for one OP in Fig. 5 at 2500 rpm / 12 bar with the horizontal arrow. Linear interpolation is applied between the measurement results to calculate the center of combustion for points with similar knock frequency as the simulated results. However, the predicted knock frequency for two operating points marked with ( ✱ ) is higher than the highest measured knock frequency of the respective run. Thus, no interpolation between measurement results is possible. For both these operating points an exponential fit is applied based on all five measurement results of each run that allows extrapolation of a point with higher knock frequency than the maximum measured one. Fig. 6: Evaluation of the predicted knock frequencies for the simulation approach using a turbulence distribution (a) and using a CCV-model (b) The results are presented in Fig. 6. by comparing the center of combustion of the simulation results on the y-axis to the center of combustion of the interpolated / ex‐ trapolated points for the measurement results on the x-axis. This way, high accuracy is indicated by the results being close to the angle bisector, where center of combustion of the simulation results is similar to center of combustion of the measurement for operating points with similar knock frequency. For both simulative approaches, all 15 operating points covering the three different engine speed and load combinations and the included spark timing variations are included in the evaluation shown in Fig. 6. The small mean deviation of 0.36 °CA and 0.58 °CA respectively for both simulation methods confirms the high accuracy of the predicted knock frequencies. The largest deviation found for the second simulation method and at 2500 rpm / 12 bar is 1.6 °CA. 545 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="546"?> Considering the general accuracy of other knock models of around 2 °CA in [16] and around 1.4 °CA in [1], 1.6 °CA deviation are still a good result. Additionally, further investigation revealed that this particular engine speed and load combination contains a very small cycle-to-cycle variation compared to the other operating points. Evaluated by the standard deviation of the IMEP, all five OPs at this engine speed and load have less than 40% of the cyclic fluctuations compared to the OPs of the other two engine speed / load combinations. Whereas the CCV-model calculates the cyclic variations correctly for the calibrated operating point, the predicted variations for all other four spark timings at this engine speed and load show a deviation from the measured cycle-to-cycle variation. This could have two possible reasons. One being that the calibration of the CCV-model is not optimal, considering that it is only calibrated at one operating point and a second being that the overall cycle-to-cycle variations are at such small level that the limit of accuracy of the CCV-model is reached. This indicates, that the deviations found for the predicted knock frequencies of this run can rather be attributed to inaccuracy of the CVV-model than an inaccuracy of the simulation method or the 3-Parameter-Approach itself. Another important aspect to consider is the performance of the simulation methods. All simulations are performed locally with standalone working process calculations on a quad core processor with 3.6 GHz each and 16 GB internal memory. Simulation duration for 500 single working cycles for the first method was approximately 11500 s / 3.2 h, whereas the simulation duration for 15 cycles in the second method was approximately 20 s, which is 575 times faster. The small number of only 15 compared to 500 simulated cycles is one reason for the much faster simulation. Another difference can be seen by comparing the time per cycle of both simulation approaches. The second method requires less time for the simulation of one cycle with ~1.34 s/ cycle, whereas the first method requires ~23 s/ cycle. The reason therefore is that for the first method the working process calculation tool is started individually and consecutively for each working cycle. For the second method, the calculation of 15 cycles requires only a single start of the tool, since the CCV-model is embedded into the simulation tool. Both comparisons, by total simulation duration and time per cycle, show the significant performance gain achieved by the second method. The accurate results show that the 3-Parameter-Approach can be successfully utilized for a knock frequency prediction. Key therefore is that both simulative approaches are calibrated at the operating point with the highest knock frequency for each engine speed and load combination, the same operating point for which AI limit is calibrated in the 3-Parameter-Approach. This way all calibration is carried out at a single operating point of a run and no further calibration to measurement data was required for all other spark timings at each engine speed and load combination, making the knock frequency calculation an actual simulative prediction for these OPs. For a complete predictive ability, besides required calibration of the simulation methods used to simulate single working cycles, AI limit would have to be determined phenomenologically. This could be a challenging task because at this point, as already 546 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="547"?> mentioned by Hess in [6] and [7] AI limit does not have a comprehensible physical meaning. As shown in Fig. 1, AI limit generally divides the single working cycles into knocking and non-knocking cycles. Therefore, it seems obvious to refer to this parameter as knock limit. However, the measurement results have shown that not exclusively the cycles with earliest AI-onsets are the knocking cycles. Hence, in order to avoid confusion with the knock limit the name auto-ignition limit is better suited. With the current definition, especially within the second presented simulation method, AI limit has more the character of a calibration parameter for the knock frequency calculation method. Nevertheless, all current results indicate that the 3-Parameter-Approach including the current definition of AI limit is well suited to predict the knock frequency. 4 Potential for Knock Control Application Conventional knock control is characterized by an immediate retardation of the spark timing if a knock event occurs. If no knock event occurs, the spark timing is incrementally advanced in much smaller increments compared to the retardation, resulting in a typical sawtooth profile for the spark timing over time. While this ensures safe operation of SI engines, it also includes some drawbacks. Due to the larger adjustment to later, less efficient spark timing and slow gradual advance of the spark timing over more cycles, the engine is operated for longer period at later spark timings than at the spark timing of the desired knock limit. Besides this unnecessarily retarded mean spark timing, the large retardation also leads to a wide variation of the spark timing in general. Although tighter control is possible by reducing the gains, especially for spark retardation, smaller increments also reduce the transient response capabilities. Tuning of the controller is therefore always a tradeoff between acceptable variation of the spark timing and transient response capability. Knock control based on knock frequency / probability would avoid the deterministic behavior and overcome the tradeoff difficulty. Since the knock frequency directly contains information about the probability that knock occurs in the next cycle under similar operating conditions the controller would allow knocking events without immediate spark retardation. There have been previous approaches to implement control algorithms that allow the occurrence of knock like the cumulative summation based stochastic knock controller [5] and the statistical likelihood based knock controller [15]. However, they rely on a comparison of measured to expected values. Either the number of measured knock events is compared to the expected amount for desired knock frequency after a certain time or the probability that knock occurs under the measured knock frequency is compared to the probability that knock occurs under the desired knock frequency. The cumulative summation approach therefore requires a minimum amount of cycles before the controller can adjust the spark timing. The likelihood based controller is able to react instantly, but only if knock events occur within a specific time window. If no knock event is measured although the knock frequency might be above the 547 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="548"?> desired limit, also a minimum amount of cycles would be required before the controller would react. This shows that these knock control approaches are rather reactive than predictive and rely on the occurrence of knock events. Knowledge about the current knock probability and control on this value would instead allow instantaneous and predictive adjustment of the spark timing, independent from the occurrence of knock. Further, control on the knock probability would allow a fast transient adjustment of the spark timing in both directions (spark advancement and retardation). This demonstrates, that probability based knock control has the potential to control the spark timing fast and precisely to the desired knock frequency limit, without unnecessary variation, thus increasing the efficiency and reducing the emissions. Additionally, control to desired knock frequency could be set up specific to the operating point in order to consider possible variations of the knock intensity at similar measured knock frequency to maximize the efficiency without risking engine damage. There have already been approaches to control knock based on mean knock frequency or knock intensity [17], [18], but as Peyton pointed out in [5], these concepts lacked in their transient response behavior due to the need of a sizeable buffer and a low pass filter to determine the required values to an acceptable degree of accuracy. This indicates one of the main challenges in applying a probability based knock control strategy. A high performance calculation that yields accurate results of the knock frequency is crucial to realize the theoretical benefits described above. This work shows that the 3-Parameter-Approach applied to simulations can predict the knock frequency with high accuracy. However, with σ AI , the calculation contains a parameter representative for the cycle-to-cycle variations that can only be determined from simulations of multiple cycles at this point, which takes significantly too much time. Even the second, already significantly faster, simulation approach using the embedded CCV-model has a calculation duration many magnitudes of order above the limit, which is required for an actual engine application. Further investigations are therefore performed within the FVV project “Fast Knocking Prediction of Gasoline Engines” and are crucial to determine the three parameters and calculate the knock frequency fast enough for an engine application. 5 Summary and conclusions In this study, the capability to predict the knock frequency has been investigated. A recently published approach, named three-parameter-approach, that has been solely applied to measurement data so far has been utilized to predict the knock frequency [6]. A zero-dimensional working process calculation tool was employed and two differ‐ ent methods were presented that allow for simulation of single working cycles to account for the cycle-to-cycle variations, one modelling a turbulence level distribution and one using an available phenomenological CCV-model [14]. A state of the art 548 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="549"?> auto-ignition model [7] was used to determine the auto-ignition onset for each simulated single working cycle. The simulations cover three different engine speed and load combinations. Each combination includes five operating points that differ by their spark timing while all other operating conditions are similar. The three-parameter-approach has been applied to the simulation results to calculate the knock frequency. For the simulation method using the CCV-model the procedure to determine the three parameters that are required for knock frequency calculation has been modified. This was required, since for this method only 15 single working cycles were simulated, whereas for the first method 500 single working cycles were simulated. The methods to simulate single working cycles and the three-parameter-approach have been calibrated only at a single operating point for each engine speed and load combination, more specifically the operating point with the highest measured knock frequency. The results revealed that a highly accurate calculation of the knock frequency is possible with both simulation methods. Largest deviations were still within good ac‐ curacy and were mainly attributed to the CCV-model, rather than the knock frequency calculation itself. With all models calibrated at only one operating point, the presented methods include an actual prediction of the knock frequency for the remaining four spark timings of each engine speed and load combination. The simulation duration using the CCV-model was significantly lower compared to the other method. Comparison to various knock control strategies showed that knock control based on the knock frequency has a great potential for a highly transient and precise controller that can help to increase efficiency and reduce emissions. It finally also highlighted the necessity of a fast calculation to make the strategy suitable for an engine application. Although both presented methods are at this point too slow for such application, the second simulative approach already includes a significant reduction of the required calculation time and is a good basis for further investigations. Acknowledgement The presented work was performed at the Institute of Automotive Engineering Stuttgart (IFS) at the University of Stuttgart under direction of Prof. Dr.-Ing. Michael Bargende within the scope of research project #1370 “Fast Knocking Prediction of Gasoline Engines” undertaken by the FVV (The Research Association for Combustion Engines eV). The project is conducted by an expert group led by Dr. Michael Fischer (Tenneco GmbH). The authors gratefully acknowledge the support received from the chairman, the FVV, the working group, the Institute for Combustion Engines of the RWTH Aachen University for providing the measurement data as well as all others involved in the project. 549 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="550"?> References [1] M. Hess, M. Grill, M. Bargende, and A. Kulzer, “New Criteria for 0D/ 1D Knock Models to Predict the Knock Boundary for Different Gasoline Fuels,” SAE Technical Paper 2021-01-0377, 2021, doi: 10.4271/ 2021-01-0377. [2] A. Fandakov, M. Grill, M. Bargende, and A.C. Kulzer, “A Two-Stage Knock Model for the Development of Future SI Engine Concepts,” SAE Technical Paper 2018-01-0855, 2018, doi: 10.4271/ 2018-01-0855. [3] R. Worret, S. Bernhardt, F. Schwarz, and U. Spicher, “Application of Different Cylinder Pressure Based Knock Detection Methods in Spark Ignition Engines,” SAE Technical Paper 2002-01-1668, 2002, doi: 10.4271/ 2002-01-1668. [4] D.E. Franzke, “Beitrag zur Ermittlung eines Klopfkriteriums der ottomotorischen Verbren‐ nung zur Vorausberechnung der Klopfgrenze,“ Ph.D. thesis, Technical University of Munich, Germany, 1981. [5] J.C. Peyton Jones, K.R. Muske, and J. Frey, “A Stochastic Knock Control Algorithm,” SAE Technical Paper 2009-01-1017, 2009, doi: 10.4271/ 2009-01-1017. [6] M. Blomberg, M. Hess, R. Hesse, P. Morsch, “Engine Knock Model,” Final Report on FVV Project 1313, H1253, Frankfurt am Main: Research Association for Combustion Engines e. V. (FVV), 2021. [7] M. Hess, M. Grill, M. Bargende, and A.C. Kulzer, “Two-Stage 0D/ 1D Knock Model to Predict the Knock Boundary of SI Engines,” 21th Stuttgart International Symposium, pp. 514-530, 2021, doi: 10.1007/ 978-3-658-33466-6_37. [8] M. Hess, M. Grill, M. Bargende, and A. Kulzer, “Knock Model Covering Thermodynamic and Chemical Influences on the Two-Stage Auto-Ignition of Gasoline Fuels,” SAE Technical Paper 2021-01-0381, 2021, doi: 10.4271/ 2021-01-0381. [9] M. Grill, T. Billinger, and M.Bargende, “Quasi-Dimenstional Modeling of Spark Ignition Engine Combustion with Variable Valve Train,” SAE Technical Paper 2006-01-1107, 2006, doi: 10.4271/ 2006-01-1107. [10] M. Grill, and M. Bargende, “The Development of an Highly Modular Designed Zero-Dimensional Engine Process Calculation Code,” SAE Int. J. Engines 3(1): 1-11, 2010, doi: 10.4271/ 2010-01-0149. [11] M. Bargende, “Ein Gleichungsansatz zur Berechnung der instationären Wandwärmever‐ luste im Hochdruckteil von Ottomotoren,” Ph.D. thesis, Technische Hochschule Darmstadt, Germany, 2001. [12] P. Kožuch, “Untersuchung des Zusammenhanges zwischen thermodynamischen Analyse‐ größen und optischen Lichtmesssignalen beim DE-Dieselmotor,” Final Report on FVV Project 769, FVV/ VKM, Stuttgart, 2001. [13] M. Wenig, “Simulation der ottomotorischen Zyklenschwankungen,” Ph.D. thesis, University of Stuttgart, Stuttgart, 2013. [14] M. Wenig, M. Grill, and M. Bargende, “A New Approach for Modeling Cycle-to-Cycle Variations within the Framework of a Real Working-Process Simulation,” SAE Int. J. Engines 6(2): 1099-1115, 2013, doi: 10.4271/ 2013-01-1315. 550 Nicolas Fajt, Co-authors: M. Grill, M. Bargende <?page no="551"?> [15] J.C. Peyton Jones, J. Frey, and K.R. Muske, “A Statistical Likelihood Based Knock Controller,” 6 th IFAC Proceedings Volumes, vol. 43, no. 7, pp 809-814, Jul. 2010. [16] A. Fandakov, “A Phenomenological Knock Model for the Development of Future Engine Concepts,” Ph.D. thesis, University of Stuttgart, Stuttgart, 2018. [17] Y.Y. Ham, K.M. Chun, J.H. Lee, and K.S. Chang, “Spark-Ignition Engine Knock Control and Threshold Value Determination,” SAE Technical Paper 960496, 1996, doi: 10.4271/ 960496. [18] M. Penese, C.F. Damasceno, A. Bucci, G. Montanari, “Sigma® on knock phenomenon control on Flexfuel engines,” SAE Technical Paper 2005-01-3990, 2005, doi: 10.4271/ 2005-01-3990. 551 Knock Probability Prediction and its Potential for a Knock Control Application <?page no="553"?> On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis Moritz Grüninger 1 , Peter Janas 2 , Olaf Toedter 1 , Thomas Koch 1 1 2 KIT IFKM Federal Mogul Ignition Systems Abstract: Pre-chamber spark plugs are currently under development for the usage on passenger car sized combustion engine. Its advantage over a conventional ignition system is its faster combustion due to hot and turbulent flame jets. Unfortunately, sporadic pre-ignition events have been observed during various engine testing. Since the combustion system consists of a preand main-chamber it is difficult to identify the root-cause. In the scope of this work we want to provide additional insights to the development of the pre-ignition. Our main assumptions for pre-ignition are: 1) pre-ignition can be triggered by hot pre-chamber surface components or 2) by hot and still reactive residual gasses trapped inside the pre-chamber from previous combustion cycles. In order to proof our assumptions, we conduct temperature measurements inside the pre-chamber cap, shell, centreand ground electrodes and monitor via six small sapphire windows the occurrence of the flame. Since the engine will be purposely run in pre-ignition mode and due to packaging reasons for the sapphire windows, the experiment is carried out on a medium-duty natural gas engine with a M14 pre-chamber spark plug. Furthermore, we compare the measurement results to large eddy simulations of the pre-chamber only, which will provide additional information about the conditions inside the pre-chamber. During pre-ignited cycles no critical component temperatures could be measured, and a first occurrence of the flame in the core nose region before spark timing is detected. Therefore, the origin for pre-ignition must be in the upper portion of the pre-chamber. The numerical results were also reporting higher residual gas concentration during the compression stroke in the core nose region. Finally, a new barrier-free pre-chamber design was built and tested, that did not show pre-ignition under same operating conditions, which proves the importance of internal aerodynamic effects to pre-ignition using passive pre-chamber spark plugs. <?page no="554"?> 1 Scope and Motivation Pre-chamber combustion technology is being discussed amongst the automotive industry thoroughly to improve gasoline combustion engines during the last five years. It can be divided into passive and active pre-chamber combustion technologies, where the passive pre-chamber combustion concept is seen as the preferred mid-term solution, due to its lower complexity compared to the active pre-chamber system. A literature study on passive pre-chamber combustion can be found in [1,2]. The main motivation for the passive pre-chamber system for passenger car sized engines is to run the engine with a stoichiometric mixture at full load. Due to the fast combustion, the engine is gaining on knock relief and also a reduction of the exhaust gas temperature. There are already first automotive car manufactures, which have launched passive pre-chamber combustion systems in the market, such as Maserati with the MC20 [3] where a passive pre-chamber-insert is mounted in the cylinderhead together with a dedicated spark plug and a secondary igniter on the side of the cylinder head to support in some challenging operating points. Such difficult operating points are, for example, idling, low load and catalyst heating operation, yet these are not the subject of the present study. A lot of research is currently conducted to address the aforementioned issues to run the passive pre-chamber without a second ignition source [4]. Also, the operating window of the pre-chamber can be very narrow, where the engine can rapidly move from a stable operating point to sever knocking, misfire or even pre-ignition, triggered by a one degree change of the the spark timing. The later issue is addressed in the scope of this work. In a recent study by the same authors [5] on pre-chamber jet visualizations inside a medium-duty gas engine, sudden and sporadic pre-ignition events have been observed for a stationary and stable operating point, that turned into uncontrolled pre-ignition of every consecutive engine cycle after some time. In this study the authors are discussing potential root causes of the origin of pre-ignition inside combustion engines using passive pre-chamber spark plugs as ignition source. Based on the work of Rosenthal et al. [5], the following hypotheses for pre-ignition are discussed: T1: Hot gas and radicals from previous combustion in pre-chamber T2: Hot electrodes T3: Hot cap surface (inside or outside) T4: Hot insulator Hypothesis T1 sees the culprit of the pre-ignition inside the hot trapped residual gasses, which are left from previous engine cycles and potentially not fully scavenged. Hypotheses T2 to T4 are blaming hot surface components, such as the electrodes, pre-chamber cap or insulator tip surface as the root cause for pre-ignition. In order to proof the four hypotheses, the passive pre-chamber spark plugs have been equipped with thermocouples at the centre electrode, pre-chamber cap, ground electrode and internal sealing portion to measure the component temperature. Also, to shed light on the question whether pre-ignition is happening inside or outside of the pre-chamber, 554 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="555"?> optical access by six sapphire windows to the inside of the pre-chamber has been provided around the seat portion. For the sake of simplicity, we do not perform the study on a passenger car sized engine, due to packaging limitations for the optical access. Furthermore, it is important that pre-ignition events can be consistently and purposelessly triggered without risk of engine damage. 2 Engine and Pre-Chamber Description 2.1 Engine Characteristics and Operating Point The engine used is a single-cylinder engine of a medium-duty diesel engine (type BR2000) by MTU. This engine is modified as an intake manifold-injected spark-ignited engine. Instead of the diesel injector, a spark plug is installed in the cylinder head and the compression ratio is set to 12.5: 1 by a modified piston. Natural gas from the local natural gas network is used as fuel. The use of a medium-duty engine enables the possibility to investigate the pre-ig‐ nitions closely due to its durability. Therefore, it is possible to operate the engine a certain amount of time in pre-igniting combustion. The operating point was chosen to create the same conditions as in the previous test series by Rosenthal et al. [5] in 2018 in order to allow high reproducibility. The operating point at which the pre-ignitions can be generated in a reproducible manner is shown in the following table. Speed / rpm 1500 Mean indicated pressure / bar 12.3 Q50 / °CA aTDCf ≤ 8 Energy from fuel per cycle / kJ/ cycle 5.9 Lambda / - 1.45 Tab. 1: Operating point with occurring pre-ignitions At the same operating point and Q50 later than 8 °CA aTDCf, no pre-ignition occurs. Due to this behavior, it is possible to specifically investigate the pre-ignitions with different measurement methods in a repeatable manner. A lower load point was chosen for preliminary studies of the behavior of the fiber-optic spark plug in order to protect the measurement equipment and to learn about the general effects of flame occurrence inside the pre-chamber spark plug. After the preliminary investigations, the fiber-optic spark plug was operated in the above-mentioned operating point to investigate the pre-ignitions. 555 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="556"?> 2.2 Pre-Chamber Spark Plug The passive pre-chamber spark plug used has been developed for mid-sized me‐ dium-duty gas engines, which can accommodate M14 pre-chamber spark plugs. A particularity of this passive pre-chamber spark plug is that it uses a ring type ground electrode, which consists of a flat disk with an annual hole in its center, in which a ring carrier with precious metal is installed. Together with an elongated and projected center electrode inside the center hole of the ground electrode disk, an annular sparking gap of ~ 0.3 mm is formed. This electrode arrangement has been developed with high durability requirements in mind, and a copper cored center electrode for fast heat rejection. The ground electrode disk is placed and laser welded in-between the end of the shell and the pre-chamber cap. To allow additional gas exchange between the lower and upper part of the inner pre-chamber volume, three additional kidney-shaped slots are provided. Furthermore, the pre-chamber cap exhibits four radially positioned annular holes with a diameter of 1.2 mm. The center lines of the holes are pointing towards the center electrode tip. Also, in the upper part of the pre-chamber the core-nose of the insulator is found and the shell web on which the insulator sits. The total pre-chamber volume is 813 mm 3 , with 322 mm 3 found between inner surface of the cap and ground electrode ring and 491 mm 3 behind the ground electrode ring. The total pre-chamber volume (V_pc) to total hole surface area (A_tot) ratio is 179 mm. 3 Numerical Analysis In order to better understand the contribution of aerodynamic effects inside the pre-chamber spark plug to the pre-ignition phenomena, 3D computational fluid dynamic simulations are performed. Since the flow motion inside the main combustion chamber is rather weak in this particular engine, the engine itself is excluded from the computational domain and only a small cubical volume of the main combustion chamber around the pre-chamber cap portion is considered. Time dependent pressure and temperature boundary conditions are directly applied to the bottom surface of the small cubical volume. The in-cylinder pressure of the main combustion chamber is directly taken from the experiment at pre-ignition relevant operating conditions (see Tab. 1). The in-cylinder temperature is calculated with an in-house developed 0D pre-chamber engine model, that was calibrated to match the experimental pressure traces. The 0D code is based on a previous work of Janas et al. [2] and was extended by gas exchange and combustion models, accordingly. Fig. 1 shows the full computational domain of the passive pre-chamber and the small cubical volume, which mimics the combustion chamber. 556 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="557"?> Boundary Conditions: p(t), T(t), egr(t), Ymix(t) Δx=0.13mm Δx=0.065 mm Fig. 1: Full computational domain of the pre-chamber spark plug and computational grid in a cut through the symmetry plane of the insulator along with an enlarged view of the annular spark gap. The mesh consists of 1.8 Mio. equidistant hexahedral cells with an average mesh element size inside the preand main-chamber section of 130 μm and in-between the annular sparking gap of 65 μm. The simulation is started at compression bottom dead center and simulated until 400 °CA. The complete computational domain is initialized with the measured pressure and re-calculated gas temperatures. The wall temperatures are taken directly from the measurements at the same operating point (see Tab. 2) and a gas mixture of 100% residual burnt gasses is assumed, which represents a worst-case scenario for the pre-chamber. Since it is expected that flow velocities and turbulence are rather high inside the pre-chamber, due to the penetration of the high velocity radial flow jets (Vpc/ A_tot > 100 mm) a large eddy simulation (LES) approached has been chosen of a single compression and pre-chamber combustion event. It is not the intention of the authors to calculate statistical relevant properties such as velocity fluctuations or mean flow field quantities, rather to have a more accurate and resolved flow field, which is crucial for the flame front propagation. Flame front propagation inside the pre-chamber is highly driven by the convection of the flow field and its vortex structures. By using LES, we resolve the big vortex structures and only model the smaller scales. For the turbulence closure of the smaller scales, a compressible version of the Smagorinsky subgrid scale model is used. The flame front propagation is modeled by a flame surface density approach, where for the flame wrinkling a transport equation is solved. Ignition is triggered by setting the 557 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="558"?> regress variable field b_tilde inside a small spherical volume in the vicinity of the electrode gap to 0, where b=0 represents burnt gas and b=1 fresh mixture, respectively. Inside the ring-electrode, the spark discharge could theoretically start everywhere at the edges of the ground electrode ring. Therefore, the ignition location was varied manually and the simulation results, which matched best the experimental evidences further analyzed. Three different spark positions were investigated as seen in Fig. 2. This modeling approach was already successfully applied in a previous study of Janas et al. [2]. The simulations are conducted with OpenFoam-7 [6]. 1 A A A-A 1 2 3 2 3 Fig. 2: Three ignition locations investigated for the simulation. Arrows show the flow into the pre-chamber during the compression stroke and the main recirculation zones inside the pre-chamber cap volume. 4 Thermal and optical Analysis 4.1 Temperature Measurement Setup Hypotheses T2-T4 are suspecting hot surface components, such as the electrodes, pre-chamber cap or insulator tip surface as the root cause for pre-ignition. To investi‐ gate these potential hotspots inside the pre-chamber, four pre-chamber spark plugs with one thermocouple each were manufactured. The positions of the thermocouples are inside the center electrode, ground electrode, cap and shell of the spark plug (and are marked red in Fig. 3). The signal of the thermocouples is passed to the indication system and is recorded with a temporal resolution of 0.1 °CA. Fig. 3: Positions of Thermocouples 558 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="559"?> Each of the four pre-chamber spark plugs is investigated at the previously defined operating point with and without pre-ignition, respectively the resulting glow ignition. Due to the heat capacity of the measuring equipment and measuring points (electrodes/ cap/ etc.), the signal is slightly time delayed, which means that a in-cycle resolved interpretation is not possible. 4.2 Method and optical Pre-Chamber Characteristics Optical accesses have to be installed into the pre-chamber in order to get a further understanding of the processes inside the pre-chamber spark plug. In order to obtain reliable measurement results as close as possible to the original system, the optical components have to be installed into a series M14 spark plug with only minimal changes to the geometry of the pre-chamber volume. In this tests, optical access to the pre-chamber is provided by six sapphire crystal optics and fiber optics (FO), which are glued with optically permeable adhesive (EPO-TEK®) to the sapphire crystals. The sapphire crystals themselves are glued to the spark plug shell with high temperature ceramic adhesive. The changes to the pre-chamber volume are negligible (volume changed to 826 mm 3 compared to 813 mm 3 ). Fig. 4: Internal volume of optical pre-chamber Since the aim is to gather information regarding if and where pre-ignition occurs inside the pre-chamber spark plug, the line of sight of three of the optical accesses is directed to the volume inside the cap (green, FO1/ FO3/ FO5) and the other three are directed to the volume around the ground electrode and insulator (red, FO2/ FO4/ FO6, see Fig. 5). These two different solid angles make it possible to distinguish whether the ignition occurs in the upper or lower pre-chamber volume. The fiber optics are connected to a photomultiplier, which in turn is connected to the indication system. In this way, the 559 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="560"?> signals of flame occurrence can be compared in terms of time with the ignition signal and the cylinder pressure signal. FO1 FO3 FO2 FO5 FO6 FO4 Fig. 5: Orientation of fiber optics FO1-FO6 5 Results 5.1 Thermal Analysis For the interpretation of the measurement results, we assume that the temperatures at components at a steady state operating point depend significantly on the heat conduction into the cylinder head. For this reason, the thermal capacity of the components and measurement equipment is relevant. Since the time constant of the thermocouples depends on their thermal capacity and the corresponding measuring point configuration, the measured temperatures are average maximum values at the respective measuring points. T_Shell T_Cap T_GE T_CE Stable operating point 281 °C 718 °C 710 °C 716 °C In pre-ignition 310 °C >1000 °C >1000 °C >1000 °C Tab. 2: Temperatures at measuring points Tab. 2 lists the measured temperatures at the previously defined measuring points. The values are the maximum values that occur during stable operation at the given operating point (see Tab. 1). Fraser [7] and Naber [8] have shown that for ignition delays in reasonable engine time scale of under 2 ms, gas temperatures above 1200 K 560 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="561"?> respectively 1150 K are needed for auto ignition of methane / natural gas. The literature often reports an auto ignition temperature for natural gas and air mix of around 575 °C to 640 °C [9]. Due to the determination of this temperature according to DIN 51794: 2003-05, whereby no attention is paid to the ignition delay time, the data of Faser and Naber are relevant for our application. The results show that for none of the four measuring points the critical temperatures reported by Fraeser and Naber are exeeded before the initial pre-ignitions appears. The temperatures of cap, ground electrode and center electrode are stationary and not increasing over time before pre-ignition starts. This and the findings of Faser and Naber indicate that these temperatures are unlikely to be the source of initial pre-ignitions. Therefore T2 & T3 can be excluded due to too low temperatures. However, it could be shown, that the temperature at the measuring point ‘spark plug shell’ increases before initial pre-ignitions occur. For this test the engine was set to the stable operating point and after stationary operation was reached (see Tab. 1), the ignition timing was slowly moved to early ignition to get Q50 < 8 °CA aTDCf and therefore high risk of pre-ignition. Before pre-ignitions occur, an increase in the temperature at the spark plug shell of a few degrees (~4 K) can be observed. This leads to the assumption, that the heat dissipation from the spark plug into the cylinder head is changing or is no longer sufficient. As a result, parts inside the spark plug (ceramic or (residual) gas, not the measured points) heat up and provoke pre-ignitions (T1 & T4). In addition, residual radicals in the pre-chamber could have an influence on the initial pre-ignition (T1). While in pre-ignition, temperatures above 1000 °C are measured at the electrodes and pre-chamber cap. This is expected to be the reason for the self-sustaining glow ignition following the initial pre-ignitions. It is likely that finally after some pre-ignitions the mixture ignites on the outside of the hot cap during glow ignition, as ignition timing of more than 90 °CA before TDC are registered. To get further information about the processes inside the pre-chamber the optical spark plug is used. 5.2 Optical Measurement For the interpretation of the measured optical signals, we have to assume, that all time delays result from flame motion. Delays due to light propagation can be neglected inside the pre-chamber, due to the short distances. 561 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="562"?> B1 B2 Fig. 6: Division of the volume into two sections For better understanding of the observed volumes, we marked the fiber optics directed to the cap volume (FO1, FO3, FO5) in green color and name the cap volume B2. The fiber optics observing the the volume around the ground electrode and insulator (FO2, FO4, FO6) are marked in red color and the observed volume is named B1. It is to mention, that this division is not sharp, as the fiber optics 2, 4, 6 also capture some of the cap volume and vice versa. Fig. 7 shows the signal of all six fiber optics (FO1-FO6) relative to the cylinder pressure p cyl and timing markers of ignition timing (IT), the end of the ignition spark (EOS) and five percent of the integrated heat release rate (Q5) in the main combustion chamber. Shown are the mean curves of the fiber optics signals for 100 engine cycles without pre-ignition, normalized to their respective maximum. The signal strength is not considered, because there are too many uncertainties (bonding, tolerances, orientation etc.) to be able to make reliable statements based on the signal strength 562 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="563"?> 320 340 360 380 400 420 0.0 0.2 0.4 0.6 0.8 1.0 Q5 EOS IT Signals of fiber optics normalized / - Crank angle / °CA FO1 FO2 FO3 FO4 FO5 FO6 p_cyl 020 40 60 80 Cylinder pressure / bar Fig. 7: Signal of fiber optics and cylinder pressure Shortly after the ignition spark, light emission can be detected on all six fiber optics. This indicates the inflammation of the fuel air mixture inside the pre-chamber. After respectively during combustion in the pre-chamber, hot gases/ radicals flow as turbulent jets into the main combustion chamber and ignite the fuel-air mixture in the main combustion chamber. Q5 indicates the point, at which five percent of the heat of the combustion is released. This point is often referred to as inflammation. After the fuel-air mixture has inflamed in the main combustion chamber, a second emission of light can be observed with a few °CA delay relative to the first emission in the pre-chamber. Therefore, the signal of the fiber optics can be divided into two peaks, the peak between ignition spark and Q5 and the peak after Q5. We refer to this phenomenon as O1. Each of the peaks consists of two smaller peaks, which partially superimpose each other and which we refer to as O2. At first, we will discuss the main peaks (O1). The first one is clearly due to the combustion inside the pre-chamber. The second one is more difficult to interpret. Light emission inside the pre-chamber after Q5 in the main combustion chamber leads to the theory that the radiation of combustion in the main combustion chamber may be visible inside the pre-chamber trough the holes in the cap. Due to the intensity level, this thesis is rejected. In preliminary investigations of Wippermann [10], it could be shown, that after a combustion inside the pre-chamber, fuel air mixture is flowing back into the pre-chamber. This is identical to the observations of an in-house 0D simulation for the current test series (see Fig. 8). 563 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="564"?> 300 320 340 360 380 400 420 440 0.0 2.0×10 -6 4.0×10 -6 6.0×10 -6 8.0×10 -6 1.0×10 -5 1.2×10 -5 1.4×10 -5 1.6×10 -5 1.8×10 -5 Mass in PC / kg CA / °CA mass pc-t mass pc-b mass pc-u Fig. 8: unburnt (-u), burnt (-b) and total (-t) mass inside the pre-chamber calculated with in-house 0D pre-chamber simulation tool The second theory on the second emission of light inside the pre-chamber is that the back flow of pre-mixed fuel air mass ignites and burns in the pre-chamber and generates the flame occurrence signal measured in the recent measurements. The cause of the inflammation could be hot surfaces or hot gas inside the pre-chamber due to the still ongoing but decaying combustion inside the pre-chamber. Given the case the second emission of light comes from a burning back flow, the heat input from the second combustion inside the pre-chamber must be considered in the evaluation of heat flow / thermal conduction. To check if the second peak is always present, a test at low load operation of 2,4 bar IMEP is performed. Fig. 9 shows that in this low load operation there is no high peak in the signals of the fiber optics after Q5. Fiber optics FO6 still shows a small increase in the signal after Q5, what is expected to be a minimal burning back flow. Generally, the signals are decaying after Q5 due to the burnout of the flame in the pre-chamber. This means, that in lower load operation and therefore lower temperatures inside the pre-chamber, the back flow does not inflame as in higher load operation. This supports the hypothesis of hot gas or a hot ceramic insulator inside the pre-chamber at higher load operation causing pre-ignitions (T1 & T4). 3D-Simulation in this load operation points will show the flow into and out of the pre-chamber and give further understanding of the processes. 564 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="565"?> 320 340 360 380 400 420 0.0 0.2 0.4 0.6 0.8 1.0 Q5 EOS IT Signals of fiber optics normalized / - Crank angle / °CA FO2 FO3 FO4 FO6 p_cyl 0510 15 20 25 Cylinder pressure / bar Fig. 9: Signal of fiber optics and cylinder pressure at low load operation The peaks inside the main peak (see Fig. 7, Reference O2) may originate from the different phases of combustion (radical reactions at lower wavelengths at the beginning of the combustion and formation of water at the end of the combustion at above 700 nm). Further tests are needed to investigate this hypothesis. A second hypothesis on this subject is, that the radiation of the flame front within the pre-chamber weakens when moving through the kidney holes (quenching) and then intensifies again in the other area (B1 resp. B2). In order to further verify or falsify the four theses stated at the beginning, the first pre-ignition that occurs at the critical operating point is specifically investigated. This initial pre-ignition is depicted in Fig. 10. The fiber optics detect a flame occurrence signal before ignition timing (IT), what proves that the pre-ignition originates inside the pre-chamber. 565 On the Origin of Pre-Ignition inside a Pre-Chamber Spark Plug - Optical and Thermal Analysis <?page no="566"?> 320 340 360 380 400 420 0.0 0.2 0.4 0.6 0.8 1.0 Q5 IT Signals of fiber optics normalized / - Crank angle / °CA FO2 FO3 FO4 FO6 p_cyl 020 40 60 80 100 120 Cylinder pressure / bar Fig. 10: Signal of fiber optics and cylinder pressure at operation with pre-ignition The first peaks of each of the six fibre optic signals occur one after the other, depending on the location of the flame core formation and the installation location (FO1-FO6). To determine the spatial origin of the pre-ignition, the time sequences of the occurrence of the first peak in the fiber optic signal are compared relative to each other. The assumes, that the signal of a fiber optic is strongest, when the light source is directly in its optical path. From the original assumption that all time delays are due to the movement of the flame, it follows that the fiber optic that detects the first peak first has the ignition point closest to its optical path. After calculating the sequence for every cycle, a probability distribution for the fibers is made over the measurement (~100 cycles). Fig. 11 shows the probability of the six fiber optics to be first to last for a engine operating point where no pre-ignition occurs. The higher probabilities to detect the first visible optical signal are in general for the green marked fibers (FO1, FO3, FO5), which are directed to the cap volume B2. Because fiber optic 2 observes the area between FO1 and FO3 and therefore a little volume in the cap (see Fig. 6) the probability to be upfront is also present, although the inflammation takes place in the cap. 566 Moritz Grüninger, Peter Janas, Olaf Toedter, Thomas Koch <?page no="567"?> FO1 FO2 FO3 FO4 FO5 FO6 FO1 FO2 FO3 FO4 FO5 FO6 FO1 FO2 FO3 FO4 FO5 FO6 FO1 FO2 FO3 FO4