<?page no="0"?> IAV SPARK26 IAV Symposium on Pre-ignition, Auto-ignition, Reactivity, and Knock MARC SENS (ED.) <?page no="1"?> IAV SPARK26 - IAV Symposium on Pre-Ignition, Auto-Ignition, Reactivity, and Knock <?page no="3"?> Marc Sens (Ed.) IAV SPARK26 IAV Symposium on Pre-Ignition, Auto-Ignition, Reactivity, and Knock <?page no="4"?> 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. DOI: https: / / doi.org/ 10.24053/ 9783381168620 © 2026 · expert verlag ‒ 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 Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen 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.narr.de eMail: info@narr.de Druck: Elanders Waiblingen GmbH ISBN 978-3-381-16861-3 (Print) ISBN 978-3-381-16862-0 (ePDF) <?page no="5"?> Contents Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf Experimental Investigations of Oil-induced Combustion Anomalies in Hydrogen Engines using Dosing and Engine-Based Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Vincent Boucand Lube oil formulation as a lever to mitigate combustion anomalies in Liebherr off-highway hydrogen-fueled ICE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany From Twin Spark to Pre-chamber: Ignition System Upgrade Strategy for a 6.6L Gasoline V8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Max Schüttenhelm, Karsten Michels, Thomas Garbe The Effect of Fuels with Higher Alcohol Content on the Ignition System and Combustion Anomalies in Gasoline Engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 T. Tietz, M. Günther, S. Pischinger Impact of Injection Strategy on Autoignition and Combustion Regime of Ultra-Lean SACI Combustion with RON95E10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala Reduced Emissions and Expanded Stoichiometric Operation Through Early Intake Valve Closing, Elevated Charge Motion, and Variable Geometry Turbocharging . . . 173 Lorenzo Sforza, Alessandro Nodi, Tommaso Lucchini, Gianluca D’Errico, William Northrop, Haogeng Bai, Hongchao Chu 3D ignition modelling in alternative-fuel engines: passive pre-chamber design and cycle-to-cycle variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 <?page no="6"?> Lukas Kniestedt Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle . . . . 195 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal . . . . . . . . 207 M.Sc. Celine Krnac, Dr.-Ing. Olaf Toedter, M.Sc. Moritz Grüninger and Prof. Dr. sc. techn. Thomas Koch The influence of operating parameters on spark plug wear and its prediction potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 Frank Wittmann, Felix Schönhofer, Christopher Kowis, Rafael Clemente Mallada, Sebastian Rieß, Michael Wensing Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa Knock Analysis in Hypercar Engine Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 6 Contents <?page no="7"?> Foreword Dear Colleagues, Welcome to the proceedings of IAV SPARK26, a symposium that brings together leading experts from industry, academia, and research institutions to exchange knowledge and discuss the latest developments in ignition, combustion, reactivity, and knock phenomena. The mobility sector is currently experiencing one of the greatest transformations in its history. While electrification and hybridization will play a central role in future transportation and energy systems, internal combustion engines continue to offer significant advantages in many applications. Their role is evolving through the integration of hybrid powertrains, advanced combustion concepts, and renewable fuels. More than two decades after the establishment of this conference series, the topics of ignition and combustion are more relevant than ever. Emerging technologies such as prechamber ignition, together with alternative fuels including hydrogen, methanol, ammonia and others, create new opportunities for improving efficiency and reducing greenhouse gas emissions. At the same time, they introduce new challenges. Highly reactive fuels may increase the risk of pre-ignition, while fuels with lower ignitability require advanced ignition strategies and a deeper understanding of combustion processes. Addressing these challenges is essential for the rapid deployment of sustainable energy carriers and the achievement of climate goals. Reducing greenhouse gas emissions within a short timeframe will require the use of all available technological solutions, including electrification, hybridization, and renewable fuels. As renewable energy remains a valuable and limited resource, continuously improving overall powertrain efficiency, including combustion efficiency, is of utmost importance. This IAV SPARK26 provides an outstanding platform for researchers, engineers, OEMs, suppliers, and technology partners from around the world to present innovations, share research results, and discuss practical solutions. The technical contributions collected in these proceedings demonstrate how advanced ignition and combustion technologies can support both existing vehicle fleets and future mobility concepts. We would like to thank all authors, speakers, partners and sponsors for their valuable contributions. Their commitment and expertise continue to make IAV SPARK a globally recognized forum for scientific exchange and technological progress. We hope that these proceedings inspire new ideas, foster collaboration, and contribute to the continued advancement of efficient, sustainable, and innovative internal combustion engines. Berlin, 2026 Marc Sens (IAV GmbH) & The IAV SPARK26 Scientific Advisory Board <?page no="9"?> Experimental Investigations of Oil-induced Combustion Anomalies in Hydrogen Engines using Dosing and Engine- Based Methods Peter Kappacher 1 , David Kapeller 1 , Paul Christoforetti, 1 Peter Günter 2 , Theo Maulbetsch 3 , Gottfried Lurf 4 Abstract: Hydrogen engines represent a promising propulsion technology for future applications, particularly in the heavy-duty sector, where they offer a compelling alternative to conventional diesel engines. However, several challenges remain, one of the most critical being the susceptibility of hydrogen engines to combustion anomalies such as knock, pre-ignition, and backfire. A potential contributor to these irregular combustion phenomena has been identified in the engine oil. Oil droplets can act as hot spots within the combustion chamber, potentially triggering unintended ignition of the air-fuel mixture. This study investigates the influence of different engine oil compositions on the occurrence of combustion anomalies. In an initial step, certified test oils are evaluated by introducing a defined quantity into the intake duct of one cylinder of a 2liter direct-injection (DI) hydrogen engine. For this purpose, a dedicated test methodology developed at TU Graz is applied. To validate this method, the selected oils are subsequently used as lubricants in a heavy-duty port fuel injection (PFI) hydrogen engine. Their impact on anomalous combustion behavior is then assessed under both steady-state and transient operating conditions. For steady-state operation, a dedicated anomaly assessment methodology developed within the Hylley project is employed. This method enables a rapid and reproducible evaluation of multiple parameters related to anomaly propensity. In transient operation, combustion anomalies are quantified in the standardized World Harmonized Transient Cycle (WHTC). The results reveal significant differences in anomalous combustion behavior among the investigated oil candidates in the oil-dosing experiments. These trends are consistently reproduced in both steady-state and transient tests on the heavy-duty hydrogen engine. Overall, a strong correlation between the different assessment methods is observed, demonstrating that the oil-dosing approach provides reliable and transferable insights into real engine operation across different engine configurations. 1 Graz University of Technology 2 Liebherr Machines Bulle SA 3 Fuchs Lubricants Germany GmbH 4 AVL List GmbH <?page no="10"?> 1 Introduction The global drive to reduce greenhouse gas emissions has intensified the search for sustainable, high-efficiency propulsion systems. Among the promising alternatives, hydrogen-fueled spark-ignited internal combustion engines (SI H₂-ICEs) have gained significant attention due to their potential for carbon-free operation while maintaining the advantages of established engine technologies. Hydrogen offers several favorable combustion characteristics, including wide flammability limits, high laminar flame speed, and excellent combustion efficiency. However, these same properties, combined with its very low ignition energy and high diffusivity, also render premixed and stratified SI hydrogen engines particularly susceptible to combustion anomalies. These combustion anomalies limit both the achievable full-load performance and the transient response of hydrogen internal combustion engines. Consequently, mitigating abnormal combustion under realworld operating conditions is essential to ensure robust, reliable, and high-performance engine operation, enabling H₂-ICEs to serve as a viable alternative to conventional diesel propulsion systems in heavy-duty applications. The following section provides an overview of anomalies that can occur in hydrogen engines. [1] Pre-Ignition Pre-ignition (PI) is generally defined as the ignition of the air-fuel mixture after intake valve closing (IVC) but before the spark event. Such events result in elevated cylinder pressures, which can cause engine damage. If ignition occurs before IVC, the flame can propagate into the intake system, resulting in backfire, which represents a particularly severe form of pre-ignition in port-fuel-injected (PFI) engines [2]. Additionally, the early increase in cylinder pressure caused by an occurring PI, can cause knocking within the same cycle. This results in combined pre-ignition and knock (PI-Kn), which combines the detrimental effects of both phenomena and therefore represents the combustion anomaly associated with the highest risk of engine damage [3]. The literature consistently identifies localized high-temperature ignition sources as the primary initiators of pre-ignition, including overheated exhaust valves, hot residual gases trapped in crevices, spark plugrelated hot surfaces or discharge irregularities, combustion chamber deposits, and other impurities [2], [4]. More recently, the ingress of lubricant oil into the combustion chamber has emerged as an additional contributory factor to pre-ignition. Numerical kinetic investigations have demonstrated that lubricant oil chemistry can increase the reactivity of the charge and reduce ignition delay under engine-relevant conditions, particularly at the lower and intermediate temperatures encountered during the compression stroke [5]. Experimental studies support these findings by showing that the introduction of oil droplets into the intake system significantly increases the occurrence of combustion anomalies under controlled operating conditions [6]. Based on these observations, dedicated test procedures have been developed to evaluate the combustion anomaly tendency of different engine lubricants [7]. Oil can be introduced to the combustion chamber via different ways: evaporated oil mist from the liner, the piston ring package, oil from the turbocharger, valve stem seals or crank case ventilation system are among the most important [8]. 10 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="11"?> Once introduced into the cylinder, oil can contribute to pre-ignition either through deposit formation or through the direct ignition of oil droplets. Deposits can act as ignition sources by creating localized hotspots within the combustion chamber. Furthermore, pressure oscillations during knocking combustion can detach deposit fragments, which may act as an additional trigger for pre-ignition events. [9] In the case of oil droplet-induced pre-ignition, volatile oil components evaporate at elevated temperatures, forming a locally enriched oil-air mixture. If the local conditions reach the self-ignition limits of the oil, oxidation and flame propagation can occur. The released energy may subsequently ignite the surrounding hydrogen-air mixture before the intended spark event, resulting in pre-ignition. The occurrence of this mechanism depends on factors such as droplet size, oil composition, temperature, and residence time within the combustion chamber. [7] In addition to lubricant oil effects, the exceptionally low minimum ignition energy of hydrogen makes locally rich mixture zones particularly vulnerable to spontaneous ignition when exposed to localized heat sources. Several studies have identified mixture formation as a key factor for abnormal combustion [1], [10]. Knock Hydrogen knock is primarily associated with end-gas auto-ignition and the subsequent generation of pressure waves. Experimental and optical investigations have shown that knock intensity is strongly influenced by parameters such as compression ratio, air-fuel ratio, spark timing, and engine speed. Compared to gasoline engines, hydrogen engines exhibit distinct knock characteristics due to the specific combustion properties of hydrogen, including differences in knock onset behavior and pressure oscillation frequencies [11]. Previous investigations have demonstrated that knock in hydrogen engines can, similar to gasoline engines, be mitigated by retarding the ignition timing [3]. However, the underlying knock behavior differs significantly from conventional gasoline operation. Instead of a gradual transition from stable combustion to increasing knock intensity with advancing ignition timing, hydrogen engines can exhibit a highly stochastic knock occurrence. Under knock-prone conditions, a large number of consecutive cycles may show normal combustion, while individual cycles experience severe knock events. This abrupt transition and cycle-to-cycle variability make the application of conventional knock control strategies significantly more challenging in hydrogen operation. Coupled Anomalies Increasing evidence suggests that combustion anomalies should not be considered independent events but rather interactive processes. Pre-ignition and backfire increase component temperatures and create thermal hotspots that promote subsequent knock, while severe knock can generate additional hotspots that increase the probability of preignition in following cycles [2]. This self-reinforcing interaction becomes particularly critical under rich operating conditions and highlights the importance of considering combustion anomalies as coupled phenomena rather than isolated events. Improving mixture preparation, reducing thermal hotspots, and controlling ignition sources remain essential for achieving robust and efficient hydrogen engine operation [1]. In addition, previous studies have identified various operating parameters influencing Experimental Investigations of Oil-induced Combustion Anomalies 11 <?page no="12"?> combustion anomaly behavior, including engine speed, load, air excess ratio (λ), ignition timing, intake and coolant temperatures, and intake air humidity [3], [8]. The influence of these parameters demonstrates that combustion anomaly formation is governed by a complex interaction of multiple factors rather than a single dominant cause. Although considerable progress has been made in understanding abnormal combustion mechanisms in hydrogen SI engines, several aspects of their formation and the interaction of contributing factors are not yet fully understood. This paper addresses oil induced combustion anomalies and defines a framework to assess different engine oil compositions regarding their appearance. A dedicated test procedure using a special oil dosing method into the intake manifold has been developed and published in [7]. However, this method does not address the investigation of anomaly formation in instances where oil is utilised as the lubricant within the engine. The present paper aims to verify the results obtained with the dosing method. In order to do so, dedicated measurements are obtained on a different engine using dedicated test-oils as the engine lubricant. 2 Methods 2.1 Dosing To investigate the influence of different engine oils on the occurrence of combustion anomalies, a dedicated test procedure is employed. This test procedure, together with the overall methodology, was developed at the Institute Thermodynamics and Sustainable Propulsion Systems at Graz University of Technology [7]. 2.1.1 Engine The oil dosing procedure is conducted on a Mercedes-Benz M274 four-cylinder inline spark-ignition engine with a displacement of 2.0 L. The light-duty (LD) engine features a compression ratio of 9.8 and direct fuel injection with rail pressures of up to 175 bar. The production Bosch HDEV4 injectors are centrally positioned in the combustion chamber, while prototype spark plugs are installed in an offset position relative to the injectors. Incylinder charge motion is characterized by a tumble flow. The intake air is initially compressed by a diesel turbocharger equipped with variable turbine geometry (VTG) and subsequently conditioned to the desired temperature using the original charge air cooler. The exhaust aftertreatment system is retained in its standard configuration and comprises a three-way catalytic converter and a rear silencer. Engine operation is controlled by a fully programmable ECU originally designed for motorsport applications. A schematic representation of the test bench setup is shown in Fig. 1. With the configuration described above, the engine achieves a maximum power output of 120 kW at 4000 rpm and a brake mean effective pressure (BMEP) of 22 bar at 2000 rpm [12]. 12 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="13"?> Fig. 1: Schematics of the M274 Engine 2.1.2 Dosing Setup Oil dosing is performed through a dosing lance with an inner diameter of 2 mm, which is installed in the intake port of the fourth cylinder. To ensure controlled droplet detachment, the oil is introduced normal to the flow direction in the intake port, with the dosing lance positioned centrally at the level of the right intake valve. The desired oil flow rate is supplied by a peristaltic pump. The dosed oil quantity is determined redundantly by means of an online precision balance and by calculating the oil mass using the carbon balance method during postprocessing. The applied oil mass flow rate is selected to represent the oil consumption level of a worn engine. The calculation of the dosed oil quantity based on the carbon balance is described in detail in [7]. It should also be noted that the engine’s inherent oil consumption is mathematically eliminated from the carbon balance, ensuring that only the externally dosed oil mass is considered. The online data acquisition from the precision balance enables closed-loop control of the supplied oil mass flow rate using a PID controller with feedforward control. This approach minimizes the influence of factors affecting the dosed oil quantity, primarily differences in oil viscosity, as well as manufacturing tolerances and wear of the pump tubing. A schematic representation of the oil dosing system is shown in Fig. 2. Experimental Investigations of Oil-induced Combustion Anomalies 13 <?page no="14"?> Fig. 2: Schematics of the Oil Dosing Setup 2.1.3 Dosing Test-Run The steady-state test procedure consists of three phases. Initially, a 10 min preconditioning and stabilization phase is performed. Subsequently, the oil under investigation is dosed for the same duration, followed by a 5 min post-conditioning phase. To ensure consistent operating conditions for each test, a BMEP controller is activated during the steady-state test procedure. The main engine operating parameters for the load point used during the dosing tests are summarized in Tab. 1. Throughout the entire test procedure, data from each individual combustion cycle as well as all relevant low-frequency measurement signals are recorded continuously. Combustion anomalies occurring during the oil dosing phase and thus attributable to the dosed oil are detected and evaluated during post-processing, using the methodology described in section 2.3. Fig. 3 shows the normalized oil mass flow rate calculated based on the carbon balance. As previously mentioned, the engines inherent oil consumption is eliminated from the calculation. Fig. 3: Relative specific Oil Consumption of an Oil Dosing Test 14 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="15"?> Parameter Unit Value BMEP bar 14 Speed min -1 1550 Lambda - 1.7 MFB 50 % °CA 12 Tab. 1: Engine Operating Parameters for the Oil Dosing Procedure 2.2 Engine based assessment methods 2.2.1 Engine The results obtained with the dosing method should be verified using the test oil as the engine lubricant in a completely different engine setup. For this, the samples are filled into a Liebherr H964 Heavy-Duty H 2 -engine, which is derived from a diesel platform. The engine features a port fuel injection (PFI) system with a maximum hydrogen injection pressure of 15 bar, regulated by a dual-stage pressure regulation system. Spark ignition is provided by centrally positioned spark plugs and a variable-energy capacitive ignition system. Compared to the diesel baseline, the compression ratio was reduced to 11, and the charge air system was redesigned to enable lean-burn operation. A dual-stage fixedgeometry turbocharging system with intercooling after each compressor stage and a highpressure wastegate provides the required boost pressure, while the injected fuel quantity is controlled via loadand speed-dependent lambda control. A fixed exhaust flap simulates the backpressure of a future exhaust aftertreatment system (250 mbar at rated power). Engine operation is managed by a prototype Schaeffler Protronic ECU, allowing flexible calibration for experimental studies. The engine delivers a rated power of 230 kW and a maximum brake mean effective pressure (BMEP) of 17.8 bar at 1500 rpm, with performance currently limited by combustion anomalies. Fig. 4 shows schematics of the engine. The investigated engine differs fundamentally from the M274 engine used in the oildosing experiments. Key differences include engine size (HD vs LD-engine), mixture formation strategy (port fuel injection vs. direct injection), in-cylinder charge motion (swirl vs. tumble), and the boosting system (dual-stage fixed-geometry turbocharging vs. single-stage variable turbine geometry turbocharging). These differences should demonstrate the transferability of the oil-dosing results to a broader range of engine concepts and real-world heavy-duty hydrogen engine applications. Experimental Investigations of Oil-induced Combustion Anomalies 15 <?page no="16"?> Fig. 4: Schematics of the H964 engine 2.2.2 Stationary assessment To validate the results obtained from the oil-dosing method, the combustion anomaly tendency of the investigated test oils is experimentally assessed on the H964 engine under both stationary and transient operating conditions. The engine is subjected to dedicated test procedures during which combustion anomalies are monitored and quantified. The stationary and transient test procedures are described in the following sections. Stationary assessment of the combustion anomaly tendency is conducted using a standardized test procedure developed at the Institute of Thermodynamics and Sustainable Propulsion Systems [3]. The method enables a rapid and reproducible quantification of combustion anomalies while minimizing the risk of engine damage. Although the test is performed under conditions that differ from the standard engine calibration, it has previously demonstrated good reproducibility and transferability of the obtained results [1], [3], [10]. The stationary test procedure consists of eight phases (Fig. 5). Two stabilization phases are carried out at a low-load operating point (1700 rpm, 4 bar BMEP), followed by two anomaly assessment phases at 1700 rpm and 8.4 bar BMEP. The engine is transitioned between both operating points by four quasi-stationary ramps, each lasting 20 s. Although the assessment load point is well below the engine’s full-load capability, combustion anomalies are intentionally promoted by reducing the air-fuel ratio and adapting the ignition timing. This operating strategy provides sufficient margin to the maximum cylinder pressure in case of an anomaly and therefore mitigates potential engine damage. Fig. 5 illustrates the corresponding profiles of engine load, combustion timing, and air excess ratio during the test procedure. Throughout the entire test, in-cylinder pressure is 16 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="17"?> recorded cycle-by-cycle for all four cylinders. The method described in Section 2.3 is used to detect anomalies occurring during the test, using adapted threshold values for MFB 05 % , KP Int and p max . ! " # # ## # ## # ## # ## # ## # ## $ ## ! % ! " % & &# & & & & '() " # ( ' #* " +,- # ( ' #* λ Fig. 5: Stationary combustion anomaly test procedure [3] Representative examples of the different anomaly types are shown in Fig. 6. The outcome of the procedure is the anomaly count of each type during the test. This metric serves as the experimental indicator for the combustion anomaly tendency in a specific engine configuration. It should be noted that, although the procedure reliably quantifies the occurrence of combustion anomalies, it does not allow conclusions regarding their underlying physical root causes [3]. ! " # $ $# % %# & Fig. 6: Examples for different anomaly types on the H964 engine [1] 2.2.3 Transient assessment Transient assessment is done by performing a World Harmonized Transient Cycle (WHTC). Fig. 7 presents the corresponding speed and load curves for the H964 engine. For combustion anomaly assessment, a dedicated evaluation methodology was developed within the Hylley project. During transient test cycles, slow-measuring engine data is Experimental Investigations of Oil-induced Combustion Anomalies 17 <?page no="18"?> recorded by the test bench at a sampling rate of 10 Hz, while high-resolution in-cylinder pressure data are acquired using an indicating system with crank-angle based resolution. The system continuously monitors combustion based on the same anomaly criteria applied in the stationary assessment. Whenever one or more threshold values are exceeded, the system stores the complete high-speed pressure data of the triggering cycle together with the preceding and subsequent 20 combustion cycles for all cylinders. After completion of the test cycle, all recorded combustion cycles are evaluated in a postprocessing routine using the established anomaly classification criteria (section 2.3). Each anomalous cycle is categorized as knock (Kn), pre-ignition (PI), or combined pre-ignition and knock (PI-Kn). In addition, the evaluation determines whether the peak firing pressure exceeds the safety limit of 220 bar. The resulting metrics comprise the total number of combustion anomalies, their distribution among the different anomaly types, and the occurrence of pressure limit exceedances throughout the transient test cycle. [1] Fig. 7: Denormalized WHTC for the H964 Engine [13] 2.3 Anomaly Detection The recorded in-cylinder pressure traces of each test are analyzed in a post-processing routine to identify combustion anomalies. The evaluation is based on threshold values for the maximum cylinder pressure (pmax), the crank angle corresponding to 5 % mass fraction burned (MFB 05 % ), and the knock intensity (KP Int ). Based on the exceeded threshold values, each combustion cycle is classified as knock (Kn), pre-ignition (PI), or combined pre-ignition and knock (PI-Kn), as illustrated in Fig. 8. A cycle is considered anomalous if the maximum cylinder pressure and at least one of the additional criteria is exceeded. Exceeding the MFB 05 % threshold indicates pre-ignition, while exceeding the KP Int threshold indicates knock. Simultaneous violation of all criteria is classified as a combined PI-Kn event. The threshold values are selected such that an exceedance of the maximum cylinder pressure limit (p max ) is always accompanied by a violation of at least one of the additional criteria. However, cycles exceeding the p max limit without violating the MFB 05 % threshold are only classified as knock events if the corresponding knock intensity (KP Int ) also exceeds its defined threshold. The evaluation is performed individually for every cylinder and every recorded combustion cycle. For the oil dosing method, the focus is primarily on the combustion anomalies occurring in cylinder 4, as the oil under investigation is dosed exclusively into this cylinder. 18 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="19"?> Fig. 8: Anomaly assessment 2.4 Test Oils Four different lubricant formulations are investigated in this study. The reference oil is the standard engine oil specified for the hydrogen version of the engine. In addition, two specially blended experimental oil formulations, whose compositions were derived based on findings from previous oil-dosing investigations, are evaluated. One formulation was designed to exhibit a low combustion anomaly tendency, while the other was formulated to promote an unusually high anomaly tendency. For comparison, the standard engine oil used in the diesel variant of the engine is also included in the investigations. In order to mitigate different oil consumptions during the stationary and transient tests, all of the tested oil have similar viscosity classes. Fig. 9 shows that the oil consumtions during transient testing in the WHTC varies less than 6 % across the different test oils. Tab. 2 shows key parameters of the different oil compositions. Unit Base Low reference High reference Standard diesel Viscosity class SAE 30 SAE 30 SAE 30 5W-30 Kin. viscosity 40 °C mm²/ s 72,33 120,7 77,5 73 Kin. viscosity 100 °C mm²/ s 12,14 12,23 12 12 Relative Ca/ Mg Content Ca/ Mg 50/ 50 100/ 0 100/ 0 No data Sulfate Ash % 0,9 0,7 0,7 0,95 Tab. 2: Key paramters of the test oils Experimental Investigations of Oil-induced Combustion Anomalies 19 <?page no="20"?> Fig. 9: Relative oil consumptions of different engine oils during the WHTC 3 Results This chapter presents and discusses the results obtained using the experimental methods described in the previous sections. As outlined earlier, the investigation begins with the oil-dosing experiments conducted on the M274 LD-engine. The corresponding results are presented in Fig. 10. A clear difference in combustion anomaly tendency can be observed between the investigated oil samples. As expected, the High Reference oil, which was specifically formulated to promote combustion anomalies, exhibits the highest overall anomaly count. Besides a pronounced increase in pre-ignition events, this oil also shows a significantly higher knock tendency than all other lubricant formulations. In contrast, the remaining three oils exhibit comparatively low and similar anomaly rates. Among these, the Low Reference sample produces the lowest overall anomaly count, confirming the intended effect of its formulation. The Baseand the Diesel Reference oil show a slightly increased pre-ignition tendency compared to the Low Reference oil. Overall, these three lubricant formulations demonstrate a low susceptibility to combustion anomalies under the investigated operating conditions. 20 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="21"?> Fig. 10: Dosing results for different test oils Dosing investigations are followed by stationary and transient tests on the H964 engine where the respective test sample is used as the engine oil. Prior to each measurement, the engine is thoroughly flushed before being filled with the respective test oil to minimize cross-contamination between tests. The total anomaly counts in the test runs is portrayed in Fig. 11. The ranking of the investigated lubricant formulations is consistent across all three test procedures. The High Reference oil exhibits the highest combustion anomaly tendency in every assessment method, followed by the standard Diesel oil and the Hydrogen Base oil. The Low Reference oil consistently shows the lowest tendency to induce combustion anomalies. This consistent ranking demonstrates good agreement between the different evaluation methods and supports the transferability of the oil-dosing results to both stationary and transient engine operation. Additionally, the findings indicate a high degree of transferability between fundamentally distinct engine types. Experimental Investigations of Oil-induced Combustion Anomalies 21 <?page no="22"?> Fig. 11: Anomaly counts from stationary and transient assessment 3.1 Discussion Although all three assessment methods yielded the same ranking of the investigated oil formulations in terms of their tendency to trigger irregular combustion, differences in the relative anomaly counts were observed. In the oil-dosing and in the transient experiments, the separation between the High Reference oil and the remaining formulations is considerably more pronounced than during the stationary engine tests. The observed differences can primarily be explained by the varying specific oil consumption associated with the individual test methods, as illustrated in Fig. 12. The specific oil consumption for each test method is determined using a carbon balance based on the measured exhaust gas composition. First, the absolute mass of lubricant entering the combustion chamber is calculated from the carbon-containing emissions (CO 2 , CO, HC). This value is then normalized by the corresponding engine energy output during the test to obtain the specific oil consumption. Finally, the specific oil consumptions are normalized to the value obtained from the oil-dosing method, which serves as the reference. 22 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="23"?> Fig. 12: relative specific oil consumption in the different test methods In the oil-dosing method, the dosed amount is intentionally chosen to simulate the elevated oil consumption of a worn engine. Consequently, the amount of oil entering the combustion chamber is significantly higher than the actual oil consumption of the H964 engine under stationary operating conditions. Another factor influencing the differing combustion anomaly tendency observed during the stationary assessment is the deliberately modified engine calibration. The engine is operated with a reduced excess air ratio and modified ignition timing to intentionally promote combustion anomalies. Under these conditions, lubricant-induced effects represent only one of several contributing factors to abnormal combustion, although the results clearly demonstrate that lubricant composition still has a significant influence on the combustion anomaly tendency. During transient operation, the effective oil consumption is significantly higher than under stationary conditions. A closer examination of the data suggests that frequent load changes and extended idle phases during the WHTC are responsible for most of the increase in oil consumption. During idle operation, the intake manifold pressure after the throttle valve becomes negative, promoting lubricant ingress into the combustion chamber through the piston rings or valve stem seals, which are not designed for prolonged operation under these conditions on the given engine. An examplatory load step during the WHTC is portrayed in Fig. 13. During such transient operating phases, the instantaneous normalized oil consumption can substantially exceed the reference value of the oil-dosing method. Consequently, the influence of engine oil on combustion anomaly occurrence becomes more pronounced during the WHTC. Experimental Investigations of Oil-induced Combustion Anomalies 23 <?page no="24"?> ! " Fig. 13: Oil consumption during an examplatory load step following an idle phase within the WHTC As a result, during the transient investigations with the High Reference oil, the combustion anomaly frequency and severity were sufficiently high that the WHTC could not be completed without exceeding the predefined safety limits. Consequently, the test was aborted after 821 s of the scheduled 1800 s test duration. However, by that point the recorded anomaly count had already exceeded that of all other lubricant formulations by a significant margin. Small differences in the distribution of combustion anomaly types were also observed between the applied test methods. While the relative distribution of anomaly types obtained from the oil-dosing experiments agrees well with the transient investigations, the stationary assessment exhibits a higher proportion of combined pre-ignition and knock (PI-Kn) events. This behavior is attributed to the selected operating point and reduced air excess ratio increase the likelihood that pre-ignition events develop into subsequent knock. Nevertheless, the consistent ranking of the investigated lubricant formulations and the similar anomaly characteristics across the different test procedures suggest that the same lubricant-induced mechanism governs combustion anomaly initiation in all three methods. 4 Conclusion This study demonstrates the significant influence of lubricant oil entering the combustion chamber on combustion anomaly formation in hydrogen engines, particularly with respect to pre-ignition. The investigated lubricant formulations exhibit substantial differences in their tendency to induce combustion anomalies, indicating that lubricant 24 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="25"?> composition represents an effective parameter for improving combustion robustness and extending the stable operating range of hydrogen engines. The experimental results further demonstrate the transferability of the oil-dosing methodology. Despite considerable differences in engine architecture, mixture formation, charge motion, boosting system, and operating conditions, the ranking of the investigated lubricant formulations remained consistent between the oil-dosing experiments on the light-duty engine and the stationary and transient investigations on the heavy-duty hydrogen engine. This strong agreement confirms that the oil-dosing approach reliably captures the influence of lubricant composition on combustion anomaly tendency. The results further indicate that the underlying lubricant-related mechanism leading to combustion anomalies is comparable across different engine concepts and operating conditions. In addition to lubricant formulation, the amount of lubricant entering the combustion chamber was identified as a key factor influencing combustion anomaly occurrence. The investigations revealed substantial differences in specific oil consumption between the applied test methods, with transient operation exhibiting markedly increased lubricant ingress during load changes and idle phases. These findings underline the importance of minimizing lubricant transport into the combustion chamber throughout the entire engine operating range and of using thoroughly tested engine oils suitable for hydrogen operation. Owing to its good correlation with engine-based validation tests, the oil-dosing methodology provides a fast, reproducible, and engine-independent approach for evaluating the combustion anomaly tendency of lubricant formulations. By enabling rapid screening of base oils, additive packages, and lubricant concepts without repeated oil changes or extensive engine testing, the method represents a valuable development tool for future hydrogen engine lubricants and supports the early identification of formulations with a low combustion anomaly tendency. 5 Statements and Declarations All measurements and investigations presented in this paper were conducted as part of the research project Hylley. It investigates hydrogen engines as a climate neutral and sustainable propulsion for Onand Off-Road machinery. Hylley is an international consortial research project involving partners from industry and science. The scientific partners are: • Graz Universitiy of Technology, Institute of Thermodynamics and Sustainable Propulsion Systems (ITnA), Consortium Leader • Montanuniversität Leoben, Department of Materials Scrience (MUL) • HyCentA Research GmbH (HyCentA) The industry partners are: • AVL List GmbH • Liebherr (with locations in Telfs, Deggendorf and Bulle) • Hörbiger Wien GmbH Experimental Investigations of Oil-induced Combustion Anomalies 25 <?page no="26"?> • Umicore AG • Ventrex Automotive GmbH Hylley is a COMET Project. The COMET Programme is financed by the federal government of Austria, the federal states Styria and Tyrol, the participating companies and research organisations. The responsible owners are the Federal Ministry of Innovation, Mobility and Infrastructure (BMIMI) as well as the Federal Ministry of Economy, Energy and Tourism (BMWET). COMET is managed by The Austrian Research Promotion Agency (FFG). Fig. 14: Funding of Hylley 6 References [1] Christoforetti P., Kappacher P., Plöckinger S. et. al., “Experimental and simulation-based investigation of combustion anomalies in H2 combustion engines and their effects on transient engine operation,” in 47th International Vienna Motor Symposium, Vienna, 2026. [2] Lu Y., Que J., Liu M. et.al., Study on backfire characteristics of port fuel injection single-cylinder hydrogen internal combustion engine, Applied Energy,Volume 364, https: / / doi.org/ 10.1016/ j. apenergy.2024.123110, 2024. [3] Kappacher P., Kapeller D., Christoforetti P. et. al., Experimental Quantification and Assessment of Combustion Anomalies under defined operating conditions of a Heavy-Duty Hydrogen Engine, Automot. Engine Technol. 10, 14; https: / / doi.org/ 10.1007/ s41104-025-00161-x, 2025. [4] Dhyani V., Subramanian K. A., Development of online control system for, International Journal of Hydrogen Energy, Volume 46, Issue 27, Pages 14757-14763, https: / / doi.org/ 10.1016/ j.ijhydene.2020.08.148, 2021. [5] Distaso E., Cal`o G., Amirante R. et.al., Linking lubricant oil contamination to pre-ignition events in hydrogen, Fuel, Volume 379, https: / / doi.org/ 10.1016/ j.fuel.2024.133041, 2025. [6] Rossegger, B., Grabner, P., Gschiel, K. et.al., Towards Robust Hydrogen Combustion: Precise Lube Oil Consumption Measurement as an Enabler for Tackling Pre-Ignition, SAE Int. J. Advances & Curr. Prac. in Mobility 7(4): 1607-1618, doi: 10.4271/ 2024-01-4290., 2024. [7] Gschiel K., Wilfling K. & Schneider M., “Development of a method to investigate the influence of engine oil and its additives on combustion anomalies in hydrogen engines.,” Automot. Engine Technol. 9, 3, 2024. [8] Lurf G., Murakami S., Jocham B. et. al., “Mitigating Lube Oil Ignition and Realted Flame Propagation through Combustion Parameter Optimization in Hydrogen Engines,” in 14. Dessauer Gasmotoren-Konferenz, Dessau, 2026. [9] Grabner, P., Schneider, M., and Gschiel, K., “Formation Mechanisms and Characterization of abnormal Combustion Phenomena of Hydrogen Engines,” in 2023 JSAE/ SAE Powertrains, Energy and Lubricants International Meeting, Kyoto, Japan; https: / / doi.org/ 10.4271/ 2023-32-0168, 2023. [10] Plöckinger S., Kappacher P., Schutting E., Trapp, C., “Combined numerical and experimental study of mixture formation strategies and their impact on anomalies in heavy-duty PFI hydrogen ICE.,” in Thiesel 2026 Conference on Thermo-and Fluid Dynamics of Clean Propulsion Powerplants, Valencia, 2026. 26 Peter Kappacher, David Kapeller, Paul Christoforetti, Peter Günter, Theo Maulbetsch, Gottfried Lurf <?page no="27"?> [11] Li Y., Gao W., Zhang P., Fu Z. et.al., Influence of the equivalence ratio on the knock and performance of a hydrogen direct injection internal combustion engine under different compression ratios, International Journal of hydrogen energy 46, 11982-11993, 2021. [12] Gschiel K., Analysis and Measures to reduce Combustion Anomalies on a turbocharged Hydrogen Engine with internal Mixture Formation, Dissertation, Graz University of Technology, 2025. [13] Christoforetti P., Transient behaviour optimization of hydrogen internal combustion engines, Dissertation, Graz University of Technology, 2026. Experimental Investigations of Oil-induced Combustion Anomalies 27 <?page no="29"?> Lube oil formulation as a lever to mitigate combustion anomalies in Liebherr off-highway hydrogen-fueled ICE Vincent Boucand Liebherr Machines Bulle SA, Bulle, Switzerland vincent.boucand@liebherr.com Abstract: Hydrogen is a key enabler in the roadmap to achieving net-zero emission powertrains. However, replacing a diesel engine in today’s construction machinery poses some significant challenges. Specifically, when it comes to power limitations due to combustion anomalies such as pre-ignition and knock. Abnormal combustion remains a key limiter for hydrogen-fueled Internal Combustion Engines (H 2 ICE), especially during warm weather operations and particularly at low speed and high load operation, aka Low-Speed Pre-Ignition (LSPI) conditions. While lubricant chemistry has been optimized to influence LSPI in gasoline GDI engines, the dominant drivers in H₂ ICEs are much less established. This contribution isolates lube oil formulation effects on combustion anomalies using Liebherr H 2 ICE through experimental results during steady state and transient operation. Motivated by H 2 ICE performances increase, a specific lubricant formulation is being co-developed with an external partner. The Liebherr 4-cylinder 9,0 liters H 2 ICE is equipped with a port-fuel injection system and dual-stage turbocharging system. The combustion takes places within lambda range of 1,5 to 4,0 depending on the specific operating conditions. Beyond efficiency and emissions, the off-highway sector demands high dependability: Liebherr targets combustion-anomaly-free operation across diverse ambient conditions for at least 15’000 operating hours without power derating or mechanical failure, making reliability a primary design objective alongside performance. Beyond test-bench work, some prototype hydrogen-fueled Liebherr wheel loaders equipped with the 4-cylinder 9,0 liters H 2 ICE are deployed with end customers in operating quarries in Germany and Austria, where they operate under production workloads, seasonal variations and high-dust environments. These customer trials provide in-service evidence of combustion anomaly resistance, complementing test bench results and guiding lubricant formulation for dependable off-highway operation. A test procedure was specifically developed at Liebherr Machines Bulle in Switzerland to evaluate various lube oil composition on back-to-back basis in engine test bench during steady-state and transient operation. During steady-state characterization at high-loads several thousands of combustion cycles per cylinder were collected under nominal and warm boundary conditions. Impact of the lube-oil formulation was determined by an <?page no="30"?> evaluation of abnormal combustion statistics. A similar approach during transient tests was also assessed. The main results give details about the base-oil and lube oil additive contents such as calcium, magnesium and molybdenum. Liebherr engineering team evaluated several lube oil candidates in addition to the reference lube-oil. Within the various lube oil tested the base group, lube oil viscosity and molybdenum enrichment emerges as one of the primary levers to reduce combustion anomalies especially at low speeds whereas calcium and magnesium levels exhibit minor influence. Base lube oil group changes alone were insufficient to provide significant impact on combustion anomalies. These findings align with literature indicating minimal calcium and magnesium content and a pronounced molybdenum benefit for pre-ignition and knock mitigation in H₂ ICEs. In line with the fundamental chemistry of hydrogen combustion, we emphasize that CO₂ and CO measured at the exhaust of H₂ ICEs originate from lube oil consumption and not from the fuel itself. Minimizing these lube oil derived carbon emissions is therefore essential to qualify H₂ ICEs as “nearly zero emission” solutions for construction equipment considering CO 2 produced by the engines <=1 g/ kWh. Therefore, carbon species in exhaust gas were systematically evaluated throughout the test campaign. CO₂ and CO emissions were compared across the various lube oils. Implications for off highway H₂ ICE development show that an additive driven strategy with higher molybdenum could potentially deliver significant combustion anomalies reduction at low speeds, if viscosity related friction penalties and protection against corrosion are managed. Keywords: hydrogen internal combustion engine, H₂ ICE, off-highway machinery, lubricant formulation, combustion anomalies, knock, pre-ignition, oil consumption, carbon emissions, Liebherr. 1 Introduction Decarbonization of heavy-duty off-highway machinery requires technical solutions that combine low or zero carbon operation with the power density, durability and operational flexibility expected from conventional internal combustion engines. For construction machinery, the replacement of diesel engines by alternative powertrain technologies presents several challenges, including energy storage, machine packaging, robustness, refueling time, ambient-condition capability, cooling requirements and total cost of ownership. For Liebherr off-road applications, hydrogen internal combustion engines offer several advantages. They provide a near CO₂-free fuel pathway at the point of fuel use, high power density compared with battery-electric solutions, comparable transient performances to existing diesel engines, lower hydrogen purity requirements than fuel cells, reduced sensitivity to vibrations compared to fuel cells, no liquid water emission as encountered with fuel cells, limited H 2 slip compared to fuel cells, competitive cost potential, compatibility with existing machine cooling packages and the possibility of using established after-sales and remanufacturing infrastructures. 30 Vincent Boucand <?page no="31"?> Liebherr started hydrogen ICE development in 2019 with a first six-cylinder engine. The first start of the six-cylinder hydrogen engine took place at IAV Gifhorn in December 2020, followed by commissioning of hydrogen test cells at Liebherr Machines Bulle in Switzerland and commissioning of the R9XX H₂ excavator in 2021. Further development milestones included a BAUMA innovation award in 2022, the first start of a Liebherr hydrogen wheel loader in 2024, the beginning of field testing in a quarry in Austria in 2025 and continued development of a four-cylinder engine for wheel-loader applications. Within this development program, combustion anomalies have been identified as a key technical challenge. Hydrogen’s wide flammability limits, low ignition energy and high flame speed make the combustion system highly sensitive to local ignition sources, mixture preparation, residual gas management, component temperatures and oil transport into the combustion chamber. This paper focuses on the role of lubricant formulation as a lever to mitigate such combustion anomalies in a Liebherr off-highway hydrogen-fueled ICE. 2 Liebherr hydrogen ICE The engine investigated in this work is the Liebherr H964 hydrogen engine, a fourcylinder, in-line, 9.0-litre engine operated on 100 % hydrogen. The engine has a bore of 135 mm and a stroke of 157 mm. The hydrogen engine uses a Liebherr port fuel injection system with an inlet pressure of 25 bar. The combustion system includes steel pistons with a dedicated bowl shape, a specific compression ratio and a dedicated piston-ring package. The valve train uses dedicated cam profiles with reduced overlap. The air path is based on a two-stage turbocharging system with air-to-water coolers and electronic wastegate control, combined with a low-volume intake manifold and intake throttle. Engine control is performed by a Liebherr ECU with in-house developed software. Spark ignition is provided by a capacitive ignition controller with dedicated M14 iridium spark plugs and variable spark profile. The engine also integrates a passive closed crankcase ventilation system with a filter element, two knock sensors mounted on the engine block and a specific lubricant developed with an external partner. The combination of hydrogen-specific combustion hardware, control software, boosting system, knock detection and lubricant development forms the foundation of the engine concept investigated in this paper. https: / / liebherr-my. sharepoint.com/ personal/ vincent_boucand_liebherr_com/ Documents/ Microsoft Copilot Chat Files/ IAV Spark 2026 - Liebherr H2 ICE - 20260702.pdf Lube oil formulation as a lever to mitigate combustion anomalies 31 <?page no="32"?> H964 engine cold side H964 engine hot side In terms of full-load performance, the hydrogen H964 reaches a rated power of 230 kW, which is similar to the comparable to the diesel D964 power range of 200-230 kW. The maximum torque of the hydrogen engine is 1500 Nm, compared with 1500-1965 Nm for the diesel variants. The maximum BMEP of the hydrogen engine is 21 bar between 1100 and 1400 rpm, while the diesel version reaches up to 27.5 bar at 1100 rpm for a specific application.https: / / liebherr-my.sharepoint.com/ personal/ vincent_boucand_liebherr_com/ Documents/ Microsoft Copilot Chat Files/ IAV Spark 2026 - Liebherr H2 ICE - 20260702.pdf Transient performance is a central requirement for off-highway applications. Liebherr’s four-cylinder hydrogen ICE demonstrated a transient response comparable to the equivalent diesel engine, supported by the high-performance boosting system and innovative engine control system. The transient capability fulfills the requirements of Liebherr offroad applications such as wheel loaders and excavators. 32 Vincent Boucand <?page no="33"?> H964 engine transient performances during load steps 3 Application context and field test operation The target application presented is the Liebherr L566 H wheel loader. The machine has a weight of 26 tons and a payload capacity of up to 9 tons. Its hydrogen storage system consists of three carbon-fiber tanks at 700 bar, providing 34 kg of usable hydrogen. The hydrogen pressure at the engine inlet is 25 bar, and cold-start ability has been tested on machine and cold test bench down to -15 °C. Wheel loaders are considered among the most transient applications in Liebherr’s machine portfolio. Field testing of the hydrogen wheel-loader machine is being carried out by customer Strabag in Austria over a two-year operation period. The machine is expected to save up to 100 tons of CO₂ per year, corresponding to approximately 37500 liters of diesel. A hydrogen filling station has been built on-site to refuel the wheel loader at the quarry. Liebherr L566H wheel loader At the time of the presentation, the H964 engine development program had accumulated 8361 hours on engine test benches at Liebherr Machines Bulle SA in Switzerland. Field-test machines in Austria had also started accumulating operating hours. Lube oil formulation as a lever to mitigate combustion anomalies 33 <?page no="34"?> 4 Combustion anomalies in hydrogen ICE Two types of combustion anomalies are considered in this work: pre-ignition and knock. Pre-ignition refers to abnormal ignition before the scheduled spark event, while knock refers to abnormal combustion after spark ignition. Both phenomena are relevant for hydrogen engines because of hydrogen’s high flammability and low ignition energy. https: / / liebherr-my.sharepoint.com/ personal/ vincent_boucand_liebherr_com/ Documents/ Microsoft Copilot Chat Files/ IAV Spark 2026 - Liebherr H2 ICE - 20260702.pdf 2 types of combustion anomalies investigated in this paper Potential contributors include hot spots, residual gases, electrical discharges and spontaneous ignition of oil droplets. The presentation identifies several possible oil-transport paths into the combustion chamber: cylinder liners, piston rings, recirculated blow-by gas, intake valve stem sealing and turbocharger compressor sealing. Previous measurements on an A-sample H964 engine, using specifically instrumented spark plugs, showed that knock started near the liner periphery. This finding indicated that combustion could be affected by oil transport from the cylinder liner to the combustion chamber and/ or by slower flame propagation toward the intake-valve side. 34 Vincent Boucand <?page no="35"?> Knock events sites The hydrogen engine therefore requires not only conventional combustion calibration and knock control, but also a dedicated understanding of the interaction between lubricant formulation, oil transport, oil-derived ignition sources and abnormal combustion behavior. 5 Safety and protection systems The Liebherr H₂ ICE includes several new safety and monitoring systems. The injection system checks for absence of fuel at key-on, performs flushing before start, conducts a leak check at each start, permanently monitors injectors and deactivates hydrogen supply after engine stop. The air path includes flushing at engine stop and a dedicated flushing sequence in case of a previous non-OK stop. Combustion-related safety features include knock detection and knock control, backfire detection and reaction and misfire detection with faultycylinder identification, misfire-rate monitoring and reaction strategies. Hydrogen leakage detection is performed using under hood hydrogen sensors and hydrogen sensors in the intake system on the machine. The crankcase hydrogen content was also investigated. The study indicated that after 60 seconds at low idle and two minutes with the engine stopped, hydrogen content in the crankcase is below the lower explosive limit. Machine-level safety features include an emergency switch and an air compressor for maintenance activities. Lube oil formulation as a lever to mitigate combustion anomalies 35 <?page no="36"?> These functions provide protection against hydrogen-specific risks and combustion anomalies but oil-induced abnormal combustion still requires a dedicated measurement and optimization methodology. 6 Measurement Methodology 6.1 Steady-state measurement A steady-state methodology was developed on engine test bench to quantify combustion anomalies over 10000 combustion cycles per cylinder, corresponding to 40000 cycles for the four-cylinder engine. The test points were selected at full-load operation and nominal laboratory temperature defined as 25 °C ambient. The operating points included 2300 rpm at 13 bar BMEP, 1900 rpm at 16 bar BMEP, 1500 rpm at 21 bar BMEP and 1100 rpm at 12 bar or 21 bar BMEP, depending on the test condition shown in the presentation material. Steady-state measurement with reference oil and nominal conditions To increase the occurrence of combustion anomalies and simulate operation at approximately 40 °C ambient temperature, additional tests were performed with elevated intake air, manifold air and hydrogen temperatures. These temperatures were increased by + 15 K compared with the nominal laboratory condition. For each condition, statistical analysis of the 10000 cycles per cylinder included three knock-intensity classes, IMEP standard deviation, IMEP deviation from average and excessive peak cylinder pressure events. A reference oil and candidate oil were compared using the same methodology. 36 Vincent Boucand <?page no="37"?> Steady-state measurement statistics with reference oil and candidate oil 6.2 Semi-transient measurement A semi-transient method was also developed to promote oil transport into the combustion chamber. The test sequence consisted of 20 minutes at 1 % load with vacuum in the intake manifold, followed by a rapid load increase to full load and then 20 minutes at full load. This procedure was designed to accumulate oil in the combustion chamber under low-load throttled operation and then observe combustion anomalies after rapid transition to high load. https: / / liebherr-my.sharepoint.com/ personal/ vincent_boucand_liebherr_com/ Documents/ Microsoft Copilot Chat Files/ IAV Spark 2026 - Liebherr H2 ICE - 20260702.pdf Lube oil formulation as a lever to mitigate combustion anomalies 37 <?page no="38"?> 6.3 Transient load-Step Measurement A transient measurement procedure was also defined to evaluate combustion anomalies during repeated load steps from throttled engine operation to full load. Three load steps were performed for each engine speed between 1000rpm and 2100rpm, with speed increments of 100 rpm. An example presented in the material showed that the number of knock events above or equal to MAPO = 5 bar, measured with cylinder-pressure sensors during repeated load steps from 5 % to 100 % load at constant speed, decreased from 10 events with the reference oil to 5 events with the candidate oil. https: / / liebherr-my. sharepoint.com/ personal/ vincent_boucand_liebherr_com/ Documents/ Microsoft Copilot Chat Files/ IAV Spark 2026 - Liebherr H2 ICE - 20260702.pdf Transient measurement with reference oil and candidate oil 6.4 Machine cycle measurement on engine test bench To reproduce realistic application behavior, a transient test simulating a typical wheelloader machine cycle was implemented on the engine test bench. In this cycle, knock controller correction output was evaluated using cylinder knock sensors. An example showed a reduction from 8.9 % for the reference oil to 7.3 % for the candidate oil. 38 Vincent Boucand <?page no="39"?> Machine cycle measurement with reference oil and candidate oil 6.5 Long-term knock events monitoring on field test machine Long-term knock event trends are also collected from field test machine operating in Austria with a datalogger and reference oil. Knock events detected by knock sensors followed a linear trend, indicating low sensitivity to ambient conditions, driving cycles, seasons, ageing and driver-to-driver variability during the observed period. This field-test feedback is important because it confirms that laboratory methodologies can be compared with real machine operation. It also provides a practical basis for monitoring how combustion anomalies evolve with machine usage, environmental conditions and engine ageing. Lube oil formulation as a lever to mitigate combustion anomalies 39 <?page no="40"?> Long-term knock event trends on wheel loader machine 7 Results and Discussion 7.1 Influence of oil formulation on combustion anomalies The candidate oil formulation showed a significant reduction of combustion anomalies at low engine speed during steady-state operation. The reduction was observed for knock events and excessive peak cylinder pressure events, especially at low speed and full load (1100 rpm and 21 bar BMEP). Low speed combustion anomalies with different oils At medium engine speeds, the benefit of the candidate oil on combustion-anomaly reduction was marginal most likely due to the reduced time window for knock event. This indicates that the lubricant effect is most relevant under operating conditions where oil transport, mixture conditions and combustion phasing create higher sensitivity to abnormal combustion at low engine speed and high load. The main formulation differences associated with the improved behavior were a different base group, increased viscosity, decreased magnesium content and increased molybdenum content. The investigation therefore suggests that lubricant base group, 40 Vincent Boucand <?page no="41"?> viscosity and molybdenum content are key parameters for mitigating combustion anomalies in the Liebherr H₂ ICE. Unlike lubricant strategies used to mitigate LSPI in gasoline passenger-car engines, calcium content did not reduce combustion anomalies at low speed in the Liebherr hydrogen engine. This result highlights the need for hydrogen-engine-specific lubricant development rather than direct transfer of gasoline-engine LSPI formulation strategies. 7.2 Transient operation Transient testing confirmed the relevance of lubricant formulation under dynamic operating conditions. In repeated load steps from low load to full load, the candidate oil reduced the number of knock events above or equal to 5 bar from 10 to 5 in the example shown. In the wheel-loader machine-cycle simulation, the candidate oil also reduced the knock controller correction output from 8.9 % to 7.3 %. This confirms that lubricant formulation can influence not only steady-state combustion behavior but also application-relevant transient operation. The observation is particularly relevant for wheel-loader applications because such machines operate with frequent and rapid load changes. The ability to reduce knock and combustion-anomaly occurrence during these transitions improves calibration robustness and may contribute to durability and customer operation margins. 7.3 Lube oil impact on carbon emissions Although the engine is operated with 100 % hydrogen fuel, engine-out CO₂ and CO were measured at the exhaust. These carbon-containing emissions originate from lubricant consumption rather than from the hydrogen fuel itself. Steady-state measurements showed a significant increase in specific engine-out CO₂ at high engine speeds and low loads with the candidate oil, indicating an influence of the base group and/ or viscosity on oil-derived carbon emissions. A similar trend was observed for engine-out CO concentration, which also increased significantly at high speed and low load. This reveals a trade-off between reducing combustion anomalies at low speed and controlling oil-derived carbon emissions across the engine map. Consequently, lubricant optimization for hydrogen engines must consider both abnormal-combustion mitigation and carbon-emission impact. A formulation that improves knock robustness may still need further optimization to minimize oil consumption and associated CO₂ and CO formation. 8 Conclusions Liebherr developed a dedicated methodology to quantify combustion anomalies in a heavy-duty off-highway hydrogen internal combustion engine. The methodology includes steady-state, elevated-temperature, semi-transient, load-step and machine-cycle test procedures, combined with cylinder-pressure analysis, knock-intensity classification, IMEP statistics, excessive peak-pressure monitoring and knock-sensor-based evaluation. Lube oil formulation as a lever to mitigate combustion anomalies 41 <?page no="42"?> Oil formulation optimization can reduce combustion anomalies such as knock in steady-state and transient operation, particularly at low engine speed of 1100 rpm. At medium engine speeds, the effect of the candidate oil was marginal. These results confirm that the Liebherr H₂ ICE is sensitive to oil transport from the liner to the combustion chamber. The main lubricant formulation parameters identified as beneficial for reducing lowspeed combustion anomalies were the oil base group, viscosity and molybdenum content. Calcium content did not reduce combustion anomalies at low speed, showing that gasoline LSPI lubricant strategies cannot be directly transferred to this off-highway hydrogen engine. Lubricant formulation also has an impact on carbon-containing exhaust emissions. Since the fuel is hydrogen, measured CO₂ and CO originate from lubricant consumption. The candidate oil showed increased CO₂ and CO at high speed and low load, highlighting the need to optimize abnormal-combustion robustness and oil-derived emissions simultaneously. Overall, the results demonstrate that lubricant formulation is an important development lever for hydrogen internal combustion engines. Future optimization should combine combustion-anomaly mitigation, oil-consumption reduction, emissions control and durability validation under realistic off-highway duty cycles. 9 Acknowledgements The author acknowledges the Liebherr Machines Bulle SA development teams involved in hydrogen engine development, combustion analysis, test-bench operation, field testing, engine control, safety systems. The author also acknowledges the external lubricant development partner involved in the formulation of the candidate oil evaluated in this work. 42 Vincent Boucand <?page no="43"?> From Twin Spark to Pre-chamber: Ignition System Upgrade Strategy for a 6.6L Gasoline V8 Marcus Wöbke (IAV GmbH), Antowan Zyada (General Motors), Michael Rieß (IAV GmbH), Adalbert Wolany (IAV GmbH) Abstract: This paper presents the development of a passive Pre-chamber ignition system for a production-representative 6.6 L naturally aspirated V8 gasoline engine with Twin-spark layout. One of the two spark plugs is replaced by a passive Pre-chamber while retaining the second conventional spark plug. The objective is to improve combustion efficiency and knock resistance while maintaining stoichiometric operation. A combined 1D and 3D CFD methodology was used to evaluate Pre-chamber potential and optimize chamber volume, nozzle geometry, A/ V ratio, mounting position, and jet orientation. Multiple Pre-chamber concepts were investigated at best-efficiency, full-load, and catalyst-light-off operating points. The results demonstrate that optimized passive Pre-chamber layouts substantially accelerate combustion, improve combustion phasing, and increase knock tolerance compared with the baseline twin-spark configuration. Simulation results indicate that these benefits enable a compression-ratio increase from 12.8 to approximately 14.5, providing additional efficiency potential. The study highlights the critical role of Pre-chamber geometry and jet targeting in achieving robust and efficient combustion in various operating conditions. 1 Introduction Recent trends suggest that interest in the internal combustion engine (ICE) is regaining momentum. Following an intense industry push toward battery electric vehicles (BEV), automakers are once again expanding ICE development to meet increasingly stringent current and future emissions regulations while also pursuing further improvements in fuel efficiency. After a period in which innovation in combustion engines slowed as attention shifted toward electrification, a range of new ICE technologies is now being revisited and advanced to address these challenges. One of the most prominent is Pre-chamber (PC) ignition in both configuration, a passive and an active Pre-chamber. 1.1 Prechamber Ignition Pre-chamber ignition is presented as a combustion strategy in which a small auxiliary chamber, connected to the main chamber through orifices, initiates combustion and releases high-energy jets or reactive radicals that create multiple ignition sites, thereby <?page no="44"?> increasing ignition energy, accelerating burn rate, and improving combustion stability relative to conventional spark ignition. Its main value lies in enabling lean or diluted combustion, improving knock tolerance, supporting higher compression ratio operation, and reducing fuel consumption, although these benefits are offset by added surface-area heat losses, reduced effectiveness at low load and cold start, pre-ignition and thermalmanagement concerns, and the added complexity of active Pre-chamber systems where the Pre-chamber is actively fueled by a dedicated injector inside it. The Pre-chamber systems are most effective at part and high load, where stronger scavenging promotes robust flame-jet formation, while low-load and catalyst-heating conditions remain challenging because of weak jets and higher relative heat losses. The Miller cycle, cooled exhaust gas recirculation, and homogeneous lean-burn operation are considered important complementary measures for broadening the useful operating range. Accordingly, a careful dedicated Pre-chamber design is needed. [24; 28; 30] The Pre-chamber performance depends strongly on its geometry: smaller orifices tend to produce stronger jets, diverging jets improve penetration and flame coverage, and the balance among Pre-chamber volume, nozzle diameter, total orifice area, and residual-gas volume governs combustion speed, stability, EGR tolerance, catalyst heating, and knock sensitivity, meaning that efficiency gains come from careful system optimization rather than from simply adding a Pre-chamber to an otherwise unchanged engine. [26; 31] 1.2 State of the Art Since Sir Harry Ralph Ricardo first proposed the Pre-chamber concept in 1918, numerous configurations have been investigated with the aim of transferring its combustion benefits to passenger-car engines [24,28,30,33,37,38,39]. Interest has intensified in the past decade, driven by the successful use of PC ignition in highly supercharged Formula-1 hybrid power units, where modern Pre-chambers enable ultra-fast, knock-resistant and fuelefficient combustion under strict fuel-flow limits, with Ferrari and MAHLE among the first publicly associated developers. This momentum has been reinforced in recent years by heavy research activity and a growing body of publications, and most recently by production-oriented announcements from car manufacturers, including Stellantis’s Hurricane 2.0-L inline-four turbo engine and BMW’s M Ignite concept, which both adopt passive PC architectures for road vehicles. In the Hurricane application, a PC spark plug derived from the Maserati MC20 “Nettuno” V-6 design featuring a compact press-fit copper insert with a central 0.55mm jet surrounded by eight 1-mm orifices; the PC spark claimed to enable a 238kW, 450Nm 2.0-L engine with Miller operation and high specific efficiency [44]. BMW, in parallel, has disclosed a patented PC combustion system intended to deliver high performance with reduced fuel consumption and CO₂ emissions in anticipation of Euro-7. The M Ignite implementation uses a passive PC in conjunction with variablegeometry turbocharging, where sonic-speed jets are reported to improve high-load efficiency without increasing peak power, making the engines cleaner and more efficient under demanding operating conditions [45]. Building on this context, the following subsections review representative recent publications, highlighting how PC ignition influences efficiency, emissions and operability in light-duty spark-ignition engines. 44 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="45"?> Attard et al. [33] developed an active prechamber fueled with propane designed as a direct replacement for a conventional spark plug in a modern gasoline engine. At 1500 rpm and 3.3 bar IMEP, the distributed ignition process significantly accelerated combustion, enabling stable operation at λ≈2.1 and total dilution levels up to 54 % mass fraction. Compared to stoichiometric spark ignition, the system achieved an 18 % reduction in fuel consumption and single-digit ppm NOx emissions. The study identified package optimization, fuel system integration, and operating-range expansion as important challenges for production application. Sens et al. [37] developed passive and active PC ignition systems. The concept utilizes geometric design freedoms such as PC volume, nozzle configuration, and jet orientation to accelerate combustion and improve knock resistance. Passive PC operation enabled a compression-ratio increase from 10 to 12 and delivered approximately 2-4 % fuel-consumption reduction. Active PC operation extended stable lean combustion beyond λ = 2.2 and achieved approximately 10 % fuel-economy improvement. The study identified wall heat transfer and PC thermal management as primary efficiency constraints, while lean operation requires dedicated NOx aftertreatment for production application. Overall, the results demonstrated that PC ignition is a promising enabler for high-efficiency, highcompression-ratio, dilute SI combustion systems. Stadler et al. [38] investigated a gasoline-fueled active/ passive PC ignition system at 1500rpm. The study showed that turbulent jet ignition accelerated combustion by as much as 60 %, substantially extending knock tolerance and enabling high-compression-ratio operation under stoichiometric conditions. Active PC fueling independently controlled PC mixture strength and extended the lean limit from λ≈1.4 to λ≈2.0. However, Increased heat-transfer losses from the PC remained the primary limitation. Shelby et al. [39] studied a light-duty boosted 3.7-L inline-six cylinder SI-engine with early intake valve closing (EIVC), cooled EGR, a 14: 1 compression ratio, and a centrally mounted PC ignition system. Both passive and active PC modes were shown to markedly increase dilution tolerance while maintaining rapid, stable stoichiometric combustion. Although the cylinder head also carries a conventional J-gap spark plug, the two systems were used either main-chamber or PC spark, not simultaneous dual-spark. Brake thermal efficiency up to 42 % was achieved, largely attributed to the PC system. Finally, Korkmaz et al. [24] experimentally evaluated several passive PC designs in a boosted DI gasoline single-cylinder engine under catalyst-heating, high-EGR and highload conditions. PC geometry, including chamber and reservoir volume, significantly influenced performance. Optimized designs increased the stoichiometric load limit to 18.5 bar IMEP and improved knock tolerance compared with a conventional J-gap spark plug. The best configuration achieved 45.67 % indicated thermal efficiency with 26 % EGR and no significant emissions penalty. Efficiency gain was associated with faster combustion and heat release. However, EGR tolerance and catalyst-heating performance remained highly sensitive to PC design, highlighting trade-offs between combustion robustness, residual-gas trapping, and jet ignition effectiveness. 1.3 Scope and Target The performance of a Pre-chamber in the engine is primarily governed by many variables, such as, volume, nozzle geometry, jet targeting, thermal condition, and mixture prepara- From Twin Spark to Pre-chamber 45 <?page no="46"?> tion. Larger Pre-chambers and optimized nozzle designs improve ignition strength and dilution tolerance, while the number, diameter, and angle of the holes determine jet penetration, turbulence generation, and flame distribution. Operating conditions such as load, lambda, EGR level, and boost pressure strongly influence the benefits achievable from turbulent jet ignition. The main design trade-off is that stronger jets and larger Prechambers generally improve combustion speed and knock resistance but also increase wall heat losses, which can reduce efficiency. In this work, the authors define and select a suitable passive Pre-chamber concept for a naturally aspirated, two-valve GM V8 engine. The study focuses on identifying a final Prechamber candidate that meets the specific performance, efficiency, and stability requirements of this architecture. Building on a collaboration with IAV and leveraging their extensive experience with passive Pre-chamber ignition, the primary objective is to design and assess a passive Prechamber system that preserves conventional stoichiometric operation (λ = 1) while improving combustion efficiency and robustness. A direct, single-injection strategy is employed at best-efficiency (BTE) and full-load (FL) operating points. A detailed modeling approach, combined with cylinder-pressure-based heat-release analysis and in-cylinder λ distribution within the Pre-chamber, is used to quantify changes in combustion behavior and brake thermal efficiency to identify potential limitations and required trade-offs. The work includes a systematic comparison of several side-mounted Pre-chamber designs for this engine at 2900 rpm full load and best-thermal-efficiency operating points. Variations in Pre-chamber volume, A/ V ratio, and bore layout are evaluated with respect to combustion duration, center of heat release (CA50) and knock behavior. This development and evaluation process demonstrates how side-mounted Pre-chamber ignition concepts can be optimized to avoid drawbacks at full load while providing efficiency benefits at part-load and best-efficiency points through a higher compression ratio and improved mixture utilization. The Pre-chamber concept is specifically exploited to accelerate combustion, mitigate knock, and thereby enable an increased compression ratio without sacrificing full-load fuel consumption. In addition, the Pre-chamber geometry is optimized to allow a locally richer mixture cloud to be pushed into the Prechamber, ensuring reliable ignition. 2 Potential of Pre-Chamber Spark Plug in a Twin Spark SI Engine with Regular- Grade Gasoline A GM V8 6.6L twin spark plug natural aspirated two valve engine (Table 1) with a base compression ratio of 12.8 is used for the PC ignition system study. The intake port is equipped with a tumble flap for low load operating points to improve the mixture preparation and turbulence intensity. A port fuel injection (PFI) system and a 1000 bar GDI (gasoline direct injection) injector enables a wide range of injection strategies to improve performance and to control emission formation. 46 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="47"?> Engine 4-stroke SI Number of cylinders - 8 Bore - Stroke mm 103.25 - 98 Displacement cm 3 820.5 Compression ratio - 12.8: 1 Number of valves/ cylinder - 1 intake and 1 exhaust Fuel injection system - PFI and GDI (p max =1000bar) Fuel RON - 96.5 Ignition system - Twin spark plug/ Pre-chamber Table 1: Main base engine specifications The twin-spark ICE is operated with premium fuel (RON 96.5). In a conventional sparkignited engine, the flame front propagates spherically from a central or in this case two lateral ignition sources. With Pre-chamber combustion hot jets penetrate the main combustion chamber. The jets create a big reactive surface and initiate combustion at multiple locations. As a result, knock-sensitive areas, such as the piston top ring zone, are reached by the flame significantly earlier, reducing the time for pre-reactions of unburned end gas. Compared to conventional combustion, the burn duration can be reduced by approximately 30 %. The faster combustion enables the center of heat release (CA50) to be advanced by around 4 to 5 °CA [27; 38; 41]. This potential can be used for compression ratio (CR) increase and/ or reduce required fuel quality (RON). Higher CR enables improved thermal efficiency particularly under part-load conditions. At full load, a compression ratio increase of approximately two ratio units [41] can be achieved while maintaining roughly the same fuel consumption. For evaluation of the potential to operate the ICE with a PC and a regular RON92 fuel. IAV’s sophisticated predictive 1D toolchain was applied. The 1D model in particular the predictive combustion model was refined based on engine measurement and combustion indication data. Required sub-models for predictive combustion and knock simulation were implemented and calibrated. In addition, predictive wall temperature and in-cylinder turbulence models were employed, as the ICE is equipped with a switchable intake-port flap that generates increased tumble in specific regions of the engine operating map. The applied IAV stateof-the-art knock model [16; 22] is widely published and validated, requires accurate prediction of wall temperatures and flow effects for its two-stage auto-ignition modelling and phenomenological knock occurrence criterion based on the fraction of unburnt mass in the thermal boundary layer. Furthermore, the influence of AFR, EGR, water injection and fuel quality are considered on knock propensity prediction. The knock model was calibrated and validated based on the Twin Spark CR12.8 premium fuel RON96.5 measurement at part load and at the knock limited full load. The knock model achieves with one parameter set a very good prediction of the CA50 with a maximum deviation below 2°CA under all conditions (speed, load, EGR) as shown in Figure 1. From Twin Spark to Pre-chamber 47 <?page no="48"?> Figure 1: EGR map, CA50 prediction results and deviation between measured and predicted CA50 The obtained knock threshold parameter set was applied for the PC potential prediction. The regular RON92 fuel analysis data was used in the predictive knock model and a faster Pre-chamber combustion by empirically shortened burn duration, based on various test bench results, was considered. A compression ratio (CR) study was conducted at characteristic engine operating points, namely best brake thermal efficiency (BTE), maximum torque, and rated power, as shown in Figure 2. At the BTE and rated power operating points, EGR was applied while maintaining a constant target load. At the BTE operating point, the PC system enables slightly earlier combustion phasing for CR values greater than 14.5. At rated power, the faster PC combustion achieves combustion phasing that is 1 to 2 °CA earlier compared with the twin-spark system (TS). Especially at the max. torque operating point, where the ICE is operated with wide open throttle (WOT), the knock mitigating benefit of the faster burning PC combustion is clearly evident. For compression ratios (CR) above 14, combustion phasing is advanced by 2.3-2.8 °CA, while load increases by 2.8 % at CR14 and 3.5 % at CR15 compared to the twin-spark configuration. Overall, the benefits in combustion phasing lead to improved brake efficiency. Since the CR investigation was conducted with identical EGR rates for both the PC and TS additional efficiency potential remains for the PC concept due to its enhanced combustion stability. This offers further potential to extend the dilution limit. The resulting increase in EGR rate can improve efficiency through reduced wall heat losses and reduced pumping work. Finally, a CR of 14.5 for the PC combustion system development was selected representing a good trade-off between efficiency gains and design constraints. 48 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="49"?> Figure 2: Difference between Pre-chamber and Twin spark results of a CR variation with RON92 fuel 3 CFD Simulations The simulations were carried out with the proprietary CFD Software Converge v5.1. The code employs a cut-cell Cartesian mesh generated at runtime, enabling fast and efficient automated mesh generation. [2] In section 3.13.1 the numerical methods are described, followed by the meshing strategy in section 3.2. The fuel modelling is explained in section 3.3 and the combustion model setup in chapter 3.4. 3.1 Numerical Methods A Reynolds-averaged Navier-Stokes (RANS) approach coupled with a detailed chemistry kinetics solver is used to solve the numerical problem. Turbulence is modeled using the Renormalization Group (RNG) k-ε turbulence model, which is widely used in internal combustion engine (ICE) applications, due to its improved capability to account for strongly curved flows and high shear rates. The modification in the dissipation transport equation leads to an improved prediction of flows involving large-scale anisotropic eddies. [4; 5] A second order central difference scheme is employed for spatial discretization. The temporal discretization performed with a first order implicit Euler scheme with variable time-step. For capturing knocking combustion and auto ignition the maximum time step must be limited based on CFL constraints. In this study the limits were: convection CFL=1, diffusion CFL=2 and Mach CFL=10. The Mach CFL is a compromise between computational effort and knock detection. It should be noted that for an accurate capturing of knocking combustion, including a precise representation of the auto-ignition waves, a Mach CFL smaller than 2 is necessary, but this causes very high computational effort. As a compromise a Mach CFL of 10 captures robustly the knock occurrence, with some inaccuracy for the wave front propagation, but enables a significantly reduced computational time. For pressure-velocity coupling a modified PISO-algorithm is used. The direct fuel injection is modeled with a Kelvin-Helmholtz-Rayleigh-Tayler break-up model based on the Lagrangian-Parcel-approach. [1; 2] From Twin Spark to Pre-chamber 49 <?page no="50"?> The detailed chemistry solver SAGE introduced by Senecal et. al. [3] was used for combustion modelling. The modelling of the thermal and kinetic boundary layers is described by the law-of-the-walls. The velocity boundary layer is modeled by the Standard-Wall-Function. A wall heat transfer model developed specifically for turbo charged spark ignition (SI) engines, the GruMo-UniMORE wall function, is used to calculate turbulent wall heat transfer. In [8], it was demonstrated in a high-pressure combustion cell that the GruMo-UniMORE, Han and Reitz, and O’Rourke thermal wallfunction models provide good agreement with the experimental data for Pre-chamber combustion. [1; 6; 8] Time resolved pressure and temperature traces created by 1D engine model simulations were applied as inlet and outlet boundary conditions. The chemical composition as initial and boundary conditions are calculated with a Chemical Equilibrium (CEQ) solver based on the 1D engine model results. 3.2 Mesh setup CONVERGE uses a cut-cell Cartesian grid generation methodology which automatically generates the mesh at runtime. In this study a basic grid size of 4 mm with several specific mesh refinements down to a local mesh size of 0.125 mm in the PC bores is used. Figure 3 illustrates the computational domain and details of the meshing strategy. Figure 3: Computational domain and meshing strategy 50 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="51"?> The refinements are activated based on crank angle intervals during the engine cycle to achieve a good compromise between accuracy and computational time. The adaptive mesh refinement (AMR) is used to avoid an under-resolved flow field during the simulation runtime. The minimal cell size is limited to 0.25 mm to prevent too long computational runtimes. The AMR is mainly used to resolve the flame front during combustion and the velocity field in the whole cycle. The maximum cell count during combustion reaches 4 Mio. Cells and around 600.000 cells at TDC in the compression stroke. 3.3 Fuel modelling Real gasoline fuels consist of several hundred or thousand hydrocarbon species, and their exact composition is often unknown. Therefore, suitable surrogate fuels composed of only a few representative fuel species must be developed for CFD simulations. These surrogate fuels must reproduce the relevant physical and chemical properties of the real fuel, such as evaporation behavior, density, research octane number (RON), motor octane number (MON), lower heating value and combustion characteristics like laminar burning velocity and ignition delay times. [16; 21; 22] Pati et. al [22] investigated gasoline fuels with different ethanol contents and defined optimized surrogate fuels to capture the real gasoline evaporation characteristics and vapor pressure for use in CFD simulations to evaluate the mixture preparation, specifically under cold engine operating conditions. In this work a liquid surrogate containing isopentane, iso-octane, decane, dodecane and ethanol for RON95E10 proposed by Pati et. al [22] is used, which leads to high accuracy of the evaporation behavior, but leads to significant deviations of the gaseous fuel properties (RON, MON, LHV). Furthermore, reaction mechanisms for decaneor dodecane-containing gasoline surrogate fuels that are sufficiently accurate for CFD simulations typically contain more than 300 species. To overcome this limitation, each liquid fuel species is mapped during evaporation to a predefined gaseous surrogate composition. This gaseous surrogate is selected to reproduce the relevant fuel properties, including lower heating value, knock tendency, laminar burning velocity and ignition delay time. The error introduced by this mapping approach is considered to be smaller than the errors associated with an inaccurate representation of the evaporation characteristic or of the gaseous fuel properties relevant to combustion. In this work the gaseous fuel surrogate optimization proposed by Fandakov et. al [16; 22] based on chemical fuel analysis of a regular US fuel is applied. A comparison of laminar flame speeds for different reaction mechanisms with experimental data can be found in section 3.4. The optimization algorithm minimizes the errors for the fuel´s key properties including RON, MON, H/ C-ratio, LHV and density. In Table 2 the fuel properties for the gaseous surrogate are summarized. From Twin Spark to Pre-chamber 51 <?page no="52"?> Property Liquid surrogate Gaseous surrogate RON/ - - 92.6 MON/ - - 83.9 H/ C-ratio - 1.89 Density kg/ m 3 - 746.1 LHV MJ/ kg - 41.59 Composition/ mass% Iso-octane 36.25 % Iso-pentane 29.19 % Decane 16.38 % Dodecane 6.67 % Ethanol 11.51 % Iso-octane 39.32 % n-heptane 16.58 % Toluene 33.76 % Ethanol 10.34 % Table 2: Composition and properties for liquid and gaseous fuel surrogates The GDI injector model is calibrated based on experimental measurements of the 500 bar injection system. The calibration of the KHRT spray breakup model was performed using an automated multi-objective optimization framework based on the S-Metric Selection Evolutionary Multi-Objective Algorithm (SMSEMOA). The optimization simultaneously minimizes deviations between simulated and measured spray characteristics, including penetration length and droplet size distribution. The best-performing simulation cases, identified based on the lowest overall error metrics, were subjected to a qualitative assessment by comparing the predicted spray morphology with experimentally obtained Schlieren images. [23] 3.4 Combustion model The combustion is modelled with the detailed chemistry solver SAGE included in CONVERGE CFD software package. [2; 3] The solver treats every cell as a constant volume reactor and calculates the heat release and the change of species concentration. The adaptive zoning algorithm groups cells with similar thermodynamic properties to reduce computational time. [2; 3; 7] In contrast to flamelet-based combustion models such as ECFM or G-equation formulations, detailed chemistry models describe combustion through elementary reaction kinetics and therefore allow ignition to develop locally at multiple sites in the main chamber. This formulation is better suited to conditions where the combustion regime transitions between flamelet, thin-reaction-zone, and broken-reaction-zone behavior. As a result, detailed chemistry can provide a more predictive representation of quenching-like effects, re-ignition phenomena, and radical-driven ignition processes induced by the Prechamber jets. Even premixed and/ or partially premixed combustion characteristics can be represented by the model. This enables, within certain limits, predictive modelling of the combustion process, particularly in turbulence-dominated combustion regimes where molecular diffusivity plays only a subordinate role. [17; 18; 19; 20] In this study Andrae et. al [12; 15] reaction mechanism with 201 species and 1014 reaction is used. This reactions mechanism showed a good compromise between accuracy and computational effort for three or four component gasoline surrogates (Figure 4), containing iso-octane (C8H18), n-heptane (C7H16), toluene (C7H8) and ethanol 52 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="53"?> (C2H5OH). The mechanism from Andrae et. al 2009 [12] was modified, by adding the ethanol reactions and low temperature chemistry from Andrae et. al 2016 [15]. Figure 4 compares the laminar flame speed predictions of three reaction mechanisms for RON90E0 and RON95E10 fuel surrogates with experimental data over a range of operating conditions. Abianeh et. al [14] skeletal mechanism benefits from a low computational effort but leads to large deviations for rich and stoichiometric mixtures. Cai et. al [13] mechanism with 490 species is the most accurate, but too computational expensive for typical CFD simulations. Figure 4: Comparison of the reaction mechanisms of Cai et al. (2019), Andrae et. al (2009,2016), and Samimi Abianeh (2015) at λ= 0.6-1.6 and EGR= 0 % , different temperature and pressure values for various fuel surrogates, using the experimental data of Jerzembeck et al. (2009) and Hesse et. al (2022) as references 4 Pre-chamber Layout Development The guiding principle throughout the optimization was to balance competing requirements. Designs optimized for maximum combustion acceleration under EGR diluted operation and knock resistance at high load typically favor a PC design with larger Pre-chamber volumes and lower A/ V ratios, whereas low-load and catalyst-light-off operation require ignition robustness at critically low in-cylinder pressure and temperature conditions which are often achieved through higher A/ V ratios. Therefore, the final concept selection was based on a holistic assessment of combustion phasing, burn duration, jet characteristics, mixture preparation, and operational robustness rather than any single performance metric. The study ultimately demonstrates that successful implementation of passive Prechamber ignition requires a dedicated engine-specific design approach in which jet From Twin Spark to Pre-chamber 53 <?page no="54"?> targeting, mixture access to the Pre-chamber, and combustion chamber integration are optimized simultaneously to achieve the desired efficiency and knock-tolerance benefits. 4.1 Case Study Four operating points at stoichiometric conditions (λ=1), spanning a broad range of the engine’s operational map, were selected for the Pre-chamber investigations. The main characteristics of these operating points are summarized in Table 3. A full-load (FL) operating point at 2900 rpm without external exhaust gas recirculation (EGR) was chosen to evaluate the knock mitigation potential of the Pre-chamber concepts. The influence of the different Pre-chamber layouts on thermal efficiency was assessed at the best brake thermal efficiency (BTE) operating point, which is characterized by an external EGR rate of 19.3 %. All simulations, except the catalyst-light-off (CLO) case, were performed under hot coolant conditions. Catalyst-light-off was conducted with 25 °C coolant temperature. speed Type IMEP Lambda λ EGR IGR@ST rpm bar . % % 1200 CLO 1.6 1 0 10.4 1900 BTE 8.4 1 19.3 21.6 2900 FL 11.5 1 0 3.1 3500 FL 11.1 1 9 11.2 Table 3: operating points considered for 3D-CFD simulations It is well known from publications, that Pre-chamber ignition systems struggle in CLO operation compared to conventional J-gap spark plugs in terms of combustion stability when retarding the center of heat release to late timings. The limitations of passive Prechambers under low-load operating conditions arise from multiple factors. The turbulence level inside the cylinder drops significantly and reduces the scavenging of PC, deteriorates the mixture preparation and the charge motion. As a result, the ignition kernel develops slowly, the reactive jets carry less energy and momentum, and the ability of the flame jets to reliably ignite the main-chamber mixture is diminished. A higher amount of cool residual gases enters the Pre-chamber under cold low load and low speed operating conditions. The mixture is more diluted, cooler and less reactive. The aggressive spark retard, up to 30 °CA after top dead center (TDC), further destabilizes the ignition and combustion conditions. The residual gas fraction is high, the mixture is less reactive and the temperature and pressure conditions are less favorable. The combustion stability penalties are strongly amplified for PC systems compared to conventional J-gap spark plugs. In addition, the increased wall surface area of the Pre-chamber leads to higher wall heat losses, whose relative contribution to the fuel energy becomes more significant at low engine loads. As a result, flame jet temperatures are reduced, leading to lower jet momentum and decreased penetration depth into the main chamber. [24; 25; 27; 28] The investigated Pre-chamber variants cover a systematic variation of geometric and position parameters for the CFD study. Since the base ICE is equipped with two locations for the conventional spark plugs there are two optional positions for the PC to be mounted: position A and position B as shown in Figure 5. Position B is close to the injector 54 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="55"?> which might favor an improved fuel inflow into the PC. Position A is on the opposite side between intake and exhaust valve. Both positions are off-center therefore a dedicated bore layout is applied to orientate the jets into the main chamber as shown here with PC8 on position B and PC9 at position A. PC6 has a symmetric bore layout which is placed at position B. The asymmetric bore layout has for position A and B an optimized orientation of the 4 jets into the main chamber to cover as much of the mixture in the combustion chamber as possible while avoiding direct impingement on the chamber walls and jet-tojet interaction. Contrary to the asymmetric bore layout, the symmetric bore layout has jets, which are partly oriented towards the liner and piston crown. Figure 5: PC8 at position B, PC9 at position A and symmetrical PC6 at position B The Pre-chamber volume was set to either 485 mm³ for the smaller variants or 594 mm³ for the larger variants, corresponding to a V PC / V C ratio of 0.8 % and 0.9 %, respectively. Small volume ratios under 1 % are often stated for passenger cars [24; 28], while larger ratios up to 3 % are often used for large bore engines [32]. Sha et. al [31; 32] founds a higher ignition energy and a higher jet performance for large PC volumes, caused by a higher fuel energy amount inside the PC. As discussed in [25; 27] a large volume is not always beneficial, because it must match with a suitable bore layout in terms of nozzle diameter, nozzle area and A/ V ratio. Small orifices in combination with a large Pre-chamber volume causes a worse scavenging of residual gases, a poorer filling and mixture preparation and higher thermal and flow losses. [25; 27] Pre-chamber ignition system works better at high loads and speeds, while cold conditions, low load and low speed remains problematic, due to higher residual gas dilution and lower mixture and flow energy. [25; 26; 28] The nozzle configuration was varied through the presence and diameter of a central nozzle, the lateral nozzle diameter, the number of lateral nozzles. The resulting nozzle area A/ V PC ranges from 0.079 to 0.130 cm⁻¹. In general, smaller A/ V ratios improve the knock mitigation potential at high loads by reducing the burn duration. In contrary larger A/ V ratios are beneficial for low load operating points like CLO, due to a more robust and reliable main chamber inflammation. [28; 30] A few selected variants include tangentially inclined bores to improve the scavenging and mixture preparation inside the Pre-chamber. [31; 32] Two variants (PC8 and PC9) feature an asymmetrical nozzle layout with optimized jet orientation into the main combustion chamber for the laterally offset mounting positions A and B (see Figure 5). From Twin Spark to Pre-chamber 55 <?page no="56"?> The layouts (Table 4) were evaluated at different operating points, including 2900rpm/ FL, 1900rpm/ BTE, and 1200rpm/ CLO. This design matrix therefore enables both isolated and combined assessment of the effects of Pre-chamber volume, nozzle area, jet orientation and nozzle symmetry on the combustion-relevant flow and mixture-formation characteristics predicted by the CFD simulations. Variant - 2 4 6 8 9 10 11 Volume mm 3 485 485 594 594 594 594 594 V PC / V C % 0.8 0.8 0.9 0.9 0.9 0.9 0.9 A nozzle / V PC cm -1 0.130 0.097 0.079 0.083 0.083 0.079 0.121 n central - 0 1 1 0 0 1 1 n lateral - 8 4 4 4 4 4 4 d central mm 1.25 1.25 1.25 1.55 d lateral mm 1 1.05 1.05 1.25 1.25 1.05 1.3 Tangential bore angle deg 0 15 15 0 0 15 15 Asymmetrical nozzle layout - - - yes yes - - Mounting position - B B B B A A B 2900/ FL x x x x x x 1900/ BTE x x x x x 1200/ CLO x x x Table 4: investigated Pre-chamber layouts 4.2 Results at 2900rpm FL The 2900rpm full-load operating point with a stoichiometric Air-to-Fuel ratio and without EGR dilution has a low internal residual gas mass fraction at spark timing and is therefore suitable to investigate the knock mitigation and combustion acceleration potential for passive Pre-chambers. To ensure a systematic assessment of the different Pre-chamber variants, all cases are evaluated at identical spark timing. This timing is intentionally not optimized for combustion phasing and does not correspond to the knock limit, thereby avoiding cross-influences caused by variant-specific spark timing adjustments. The evaluated PC variants cover a wide range of geometric Pre-chamber parameters. Two volume levels (PC2,4 to PC 5-8), different A/ V ratios, different central and lateral nozzle configurations, and both symmetric and asymmetric jet targeting (PC8, PC9). For PC5 a modified inner geometric shape B is investigated and directly comparable to variant PC6. PC8 and PC9 differ in the jet target. The jet target of PC8 is optimized for position B while the PC9 jet orientation is optimized for the oppositely position A. Figure 6 illustrates key parameters for the ignitibility inside the Pre-chamber. The residual gas rate is controlled by the Pre-chamber volume, the overall bore area and the mounting position. The residual gas rate increases towards higher volumes (PC4 vs. PC5), but shows a strong cross-influence of the A/ V ratio. 56 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="57"?> Figure 6: Comparison of different Pre-chamber variants on ignition and combustions parameters inside the PC at 2900rpm full load. The scavenging efficiency, expressed as the difference in residual gas rate between the Pre-chamber and the cylinder, deteriorates with increasing PC volume. A higher A/ V ratio supports the Pre-chamber scavenging. At these low residual gas rates in this operating point, this difference is not critical. However, at operating points where the residual gas is heavily diluted, particularly at low loads and low engine speeds, this difference becomes increasingly critical for the ignition capability of the Pre-chamber ignition system. Pre-chamber variants with a symmetrical nozzle layout show only minor differences of the air-fuel-distribution in the Pre-chamber and the average air-to-fuel ratio in the spark gap. The asymmetrical bore layout of PC8 enables a richer mixture, due to an improved fuel transport from the Cylinder into the side mounted PC volume. The preferred mounting position of PC8 is position B, as this allows a better and richer mixture preparation of the Pre-chamber. PC9 with identical volume and nozzle diameters is mounted on the opposite position of the cylinder, close to exhaust valve. The nozzle targeting is mirrored, while the nozzle angles are slightly modified to satisfy cylinder head geometry constraints. PC9 at position A shows a significantly leaner mixture, as this From Twin Spark to Pre-chamber 57 <?page no="58"?> location is substantially less favorable in terms of injector positioning. Despite its lean mixture, PC9, together with the asymmetric PC8, the symmetric high-load-optimized PC6 (small A/ V ratio), and the low-load-optimized PC11 (high A/ V ratio), achieves the shortest Pre-chamber combustion durations (Figure 6) among the investigated variants. This indicates that a stronger Pre-chamber combustion event can be achieved with suitable PC volume to compression volume ratio, a moderate A/ V ratio and thus a bore layout that supports a sufficient mixture preparation while maintaining a strong and effective jet generation. The integrated jet-impulse results (Figure 7) show a clear benefit for PC variants 5, 6, 8 and 9 with a larger PC volume and relative small A/ V ratio. A small number of equal sized lateral bores leads to the smallest coefficient of variation (CoV) of the bore impulse. The coefficient of variation of the jet-bore impulse quantifies the degree of non-uniformity in the distribution of integrated jet impulse among the individual Pre-chamber bores. Figure 7 shows that differently sized nozzle diameters and the usage of an additional central nozzle significantly increases the CoV of the impulse. A smaller CoV seems to be beneficial for a shorter main chamber burn duration. The asymmetrical designed bore layout of variant PC8 and PC9 shows the shortest burn durations among the smallest impulse CoV. A high jet impulse and a fast and intensive combustion inside the Prechamber are substantial prerequisites for energetic jets. However, the effectiveness of the ignition process depends not only on the total jet impulse but also on the uniformity, penetration, and reactivity of the individual jet and is influenced by various geometrical parameters such as nozzle diameter, nozzle number and Pre-chamber volume. [25] Focusing on a single metric, such as jet impulse, can be misleading, as main-chamber combustion performance results from a complex interaction between Pre-chamber geometry, engine operating conditions, and the timing of jet discharge during the combustion cycle. Figure 7 demonstrates clear benefits in terms of earlier combustion phasing and shorter burn duration for variants 6, 8, 9, and 11. Notably, these Pre-chamber configurations differ substantially with respect to nozzle arrangement, nozzle diameter, number of nozzles, and consequently A/ V ratio. Especially the optimized jet orientation of PC8 achieves a faster combustion start and leads to a shorter combustion duration. PC9, which has also optimized jet orientation but is located on position A has drawbacks due to the mixture distribution. Around the position A there is a leaner mixture in the main chamber and therefore also in the PC itself resulting in slower combustion compared with jet orientation optimized PC8 on Position B. Finally the asymmetric jet orientation of PC8 leads to best combustion results with a high fraction of fuel burned and together with a more efficient earlier combustion phasing the highest IMEP is achieved. 58 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="59"?> Figure 7: Comparison of different Pre-chamber variants on cylinder combustion and PC jet parameters at 2900rpm full load. Figure 8 illustrates OH iso-surfaces for PC6, PC9, PC9 and PC11 at different crank angles after ignition. PC8 and PC9 are equipped with four lateral nozzles, whose orientations were individually optimized for the respective mounting position. The jet targeting was selected to minimize direct and intense impingement on the piston surface, aiming to reduce wall heat losses. Furthermore, the nozzles were arranged to generate a wide jet spread and thereby improve flame coverage across the combustion chamber. From Twin Spark to Pre-chamber 59 <?page no="60"?> Figure 8: OH iso-surfaces for PC6, PC9, PC9 and PC11 at various crank angles after ignition The individually optimized bore layout of the asymmetric configurations, with PC8 installed at position B and PC9 at position A, results in a jet orientation that promotes deep jet penetration into the main-chamber mixture. The results indicate that both PC8 and PC9 achieve good cylinder-volume coverage already during the early jet-development phase at 8 °CA aIGN, while maintaining favorable conditions for flame propagation up to 20 °CA aIGN. Compared with the symmetric variants PC6 and PC11, the asymmetric layouts produce a broader and more distributed reaction zone. The individual jets remain clearly distinguishable during the early flame-development phase and merge only at later crank angles, which reduces early jet-to-jet interaction and promotes a more homogeneous coverage of the combustion chamber. The symmetric variants, by contrast, show a more centralized and clustered jet-front structure. In addition, they exhibit stronger penetration toward the piston surface and more pronounced interaction between the individual flame jets, which may limit the effective flame-propagation rate. 60 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="61"?> Figure 9 illustrates the most promising variants PC8 (low A/ V-ratio, high load optimized) and PC11 (low load optimized, high A/ V ratio) with two different ignition angles in terms of combustion phasing, burn duration and pressure traces. With the advanced spark timing, both variants are operated closer to the knock limit, although the knock limit itself is not reached. Further optimization of the spark timing is beyond the scope of this study and could be investigated in future engine test-bench experiments. Figure 9: Spark advance optimization for PC8 and PC11 at 2900rpm/ FL PC8 achieves earlier combustion phasing, represented by CA50, and a shorter burn duration due to a richer Pre-chamber mixture, optimized jet targeting that reduces wall heat losses at the piston surface, and the avoidance of early jet-to-jet interaction, which slows down the combustion. PC11 shows slightly later combustion phasing and a longer burn duration, which probably reduces its knock-mitigation potential under full-load conditions. However, the higher A/ V ratio and an additional central bore improve Pre-chamber scavenging and support a more robust ignition and combustion at low engine speed and low load operating points, especially during catalyst-light-off operation (CLO). This is consistent From Twin Spark to Pre-chamber 61 <?page no="62"?> with the observation that PC11, with its symmetric layout and higher A/ V ratio, shows a stronger increase in burn duration when the spark timing is retarded. 4.3 Results at 1900rpm BTE At the best-BTE operating point with an EGR rate of 19.3 %, the spark timing was adjusted to avoid auto-ignition while maintaining a comparable combustion phasing of CA50 = 10° CA aFTDC for all four investigated Pre-chamber concepts, as shown in Figure 10. The comparison of the required spark advance (SA) reveals clear differences between the investigated Pre-chamber concepts. The asymmetric concept PC8 achieves the target combustion phasing with the latest ignition timing, indicating an advantage in combustion initiation compared to the symmetric configurations. PC8 achieves a faster combustion also due its richest lambda level, combined with a more favorable jet orientation, promotes earlier flame development in the main chamber and accelerates the start of combustion. In contrast, PC9 does not fully benefit from its optimized jet orientation due to its location at position A. Both residual gas scavenging and fresh charge exchange within the Pre-chamber are less effective. Furthermore, the mixture around position A is globally leaner, resulting in a lean in-Pre-chamber lambda of approximately 1.15. This leads to a longer Pre-chamber combustion duration and a weaker cylinder combustion start in the lean mixture surrounding the PC and consequently causes a prolonged main-chamber combustion duration. The symmetric concepts PC6 and PC11 show comparable combustion behavior despite the larger A/ V ratio of PC11. No significant differences were observed inside the Prechamber, resulting in similar combustion characteristics and ignition timing requirements. However, the combustion initiation advantage of PC8 could not be translated into higher brake thermal efficiency. To maintain knock-free operation at a constant combustion phasing of CA50 = 10°CA aFTDC, the ignition timing of the asymmetric concepts had to be retarded, compensating part of their advantage. Consequently, no significant improvement in BTE was observed under these conditions. Instead, the primary benefit of PC8 is the reduced spark advance requirement. In comparison, the symmetric configurations PC6 and PC11 require approximately 7 °CA earlier ignition timing to achieve the same combustion phasing. This result indicates that later spark timing may offer additional potential for operation under even higher EGR rates. 62 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="63"?> Figure 10: Comparison of different Pre-chamber variants on ignition and combustions parameters inside the PC at 1900rpm/ best BTE. The advantage of a 7°CA later spark timing of the asymmetric PC8 can be attributed to differences in the in-Pre-chamber flow and combustion process and the subsequent jetinduced ignition of the main chamber. The asymmetric nozzle arrangement promotes faster combustion within the Prechamber due to a more favorable internal flow pattern as shown in the comparison between PC6 and PC8 at their individual spark timings in Figure 11. In the symmetric configurations, the centrally positioned nozzle influences flame propagation in a less favorable manner, since the inflow flows against the flame propagation direction, whereas the asymmetric arrangement generates an in-Pre-chamber flow directed from the spark gap towards the main chamber. This enhances flame propagation inside the Pre-chamber and accelerates the pressure rise required for jet formation. From Twin Spark to Pre-chamber 63 <?page no="64"?> Figure 11: In-PC flow comparison between PC6 and PC8 at spark timing at 1900rpm/ BTE. Furthermore, PC8 shows a slightly richer mixture both within the Pre-chamber and in the adjacent main-chamber region around the Pre-chamber. Due to the orientation of the asymmetric orifices, the reactive jets penetrate deeper into the main chamber and ignite a larger volume of the mixture compared with the symmetric PC6 and PC11 configurations, as illustrated by the OH iso-surfaces in Figure 12. At 8°CA after ignition the 4 jets of PC8 and PC9 are clearly separated and generate turbulence and a bigger hot and reactive surface, whereas the jets of PC6 and PC11 have high interaction with each other and with the surface of the piston which results in a merging of the jets, higher local wall heat losses where the jet hits the piston and finally a reduced hot reactive surface for further flame propagation as shown in the comparison of the different crank angles. 64 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="65"?> Figure 12: OH iso-surfaces for PC6, PC8, PC9 and PC11 at various crank angles after ignition. Operating point 1900 rpm/ BTE. 5 Challenges of Pre-chamber combustion at Catalyst-Light-off operation The energy conversion process in a spark ignited (SI) engine is a pre-mixed combustion process, which is characterized by a complex interaction between the turbulent flow field and the chemical heat release. The flow field inside a conventional SI engine is characterized by high shear flows during the intake process and the partially conserved structure and turbulence during the compression stroke. The turbulence intensity and eddy structure corrugates the flame front. This leads to an increased flame surface and results in a higher turbulent burning velocity. In a Pre-chamber combustion system the main fraction of the generated turbulence inside the main chamber occurs during the jets entering the cylinder. The resulting inhomogeneous flow field with strong shear layers at the jet boundaries increases the species and turbulence transport with a strong impact on From Twin Spark to Pre-chamber 65 <?page no="66"?> the flame front evolution and propagation. For classifying the turbulent premixed combustion a regime diagram proposed by Borghi and Peters [19] is widely used. High turbulence Pre-chamber ignition systems are characterized by covering different regimes during the combustion process. The combustion inside the Pre-chamber remains often inside the thin reaction zones, while the main chamber combustion transitions from the thin reaction zone to the corrugated flamelet or in some case to wrinkled flamelet regime. Lean mixtures leads to a reduced laminar flame speed and this shifts the combustion regime to regions with thicker flames, which show a highly fragmented flame front and can cause an instable flame behavior. [19; 20; 27; 29; 34; 35; 36] The chemical and flow field conditions under cold catalyst light off operation can be critical for a Pre-chamber ignition system in terms of an instable main chamber inflammation and are critical for a simulation approach, which must capture accurately the chemistry and turbulence characteristics. [28; 29; 34] 5.1 Turbulence-chemistry analysis methodology In this section two approaches to detect the cells prior to the flame front for a subsequent turbulent flame regime analysis proposed by Peters [19] are presented. The flow field files of the detailed chemistry simulations provide the necessary data for the post processing analysis. Inspired by the work of Piano et al. [29] a simple chemistry formaldehyde (CH 2 O) mass fraction based selection (Y CH2O > 10 -5 ) is used. The cells must be inside the cylinder or the Pre-chamber and are not part of the near-wall-exclusion mask. The multi-criteria prefront detection method is more physics based and spatially constrained. It starts from the detected flame front, searches into the unburned direction (progress variable gradient search into the unburned gas), removes wall artifacts, and grows a limited number of AMR-aware cell layers under additional chemical and thermodynamic criteria. A detailed explanation of this approach is out of the scope of this work and should be published in a later publication. The flame-front detection combines chemistry, thermodynamics and spatial information. The progress variable identifies the transition between unburned and burned gas, the gradient criterion localizes this transition spatially, and the OH, fuel, product, temperature and wall film filters remove non-flame artifacts. The detected flame front provides robust anchor cells and normals for the subsequent pre-front detection. The coordinates for the Borghi-Peters diagram were calculated for each individual cell. In a second step the cell results are binned based on the mass fraction. The laminar flame speed and flame thickness are interpolated from a tabulated flame speed table for the RON92E10 fuel calculated with the reaction mechanism discussed in section 3.4. Figure 13 illustrates the results of both pre-front cell selection methods on the laminar flame speed and the relative number of selected cells. The formaldehyde interval method leads to high temperatures inside the unburned gas and cause unrealistic high laminar flame speeds. This suggests that a substantial number of cells in the flame front are also included in the analysis. This hypothesis is supported by significant increased residual gas rate and high OH (hydroxyl radical) mass fraction values, which are typically located in the flame front [30]. In contrast, the multi-criteria pre-front detection does not show a significant increase in the OH radical signal. This indicates that the method more reliably and robustly identifies 66 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="67"?> unburned cells located ahead of the flame front. Moreover, the method identifies fewer cells during the late stage of combustion, which is consistent with the decreasing extent of unburned regions ahead of the flame front. The artifacts inside the red colored area caused by faulty cell detection inside the cylinder, before the first jet enters the main chamber. This can be improved by using a species-based gate filter. In the following analysis these artifacts are excluded. Figure 13: Comparison of laminar flame speed calculation of CH 2 O interval and multi-criteria Pre- Front detection method under CLO conditions. The identification of different turbulent flame regimes can be realized by the following dimensionless characteristic numbers, which correlates the chemical and turbulent quantities. The turbulent Dammköhler number (Da) is the ratio between the turbulent eddy turn over time and the laminar burning time and can be described as follows: Da ¼ l t = u 0 � f = s L ¼ l t s L u 0 � f Equation 1 [19] The Karlovitz number (Ka) represents the ratio between the chemical time scale and the Kolmogorov length η k . Ka ¼ u 2 � K s 2 L ¼ � 2 F � 2 K Equation 2 [19] A second Karlovitz number describing the scales in the inner reaction zone is defined as: From Twin Spark to Pre-chamber 67 <?page no="68"?> Ka � ¼ � 2 Ka ; � ¼ � R � F � 1 10 Equation 3 [19; 40] The turbulent Reynolds number follows as: Re t ¼ u 0 l t s L � F ¼ Da 2 Ka 2 Equation 4 [19; 40] For Re t <1 laminar flames are observed. For small Karlovitz numbers the flame thickness is small compared to the smallest turbulent eddies (Kolmogorov scale). The eddies could not substantially perturbate the flame structure. The turbulent fluctuations cause a corrugation of the flame front (corrugated flamelets). [19; 29; 40; 42] If Ka<1, combined with u’<s L , the turnover velocity of the largest turbulent eddies is smaller than the laminar flame propagation speed. The turbulent fluctuations are unable to penetrate the flame front. The flame structure preserves its laminar structure. For Ka>1 the smallest eddies can penetrate the flame front and modify the flame structure. [19; 2; 40; 42] For Ka δ >1 the Kolmogorov eddies could penetrate the inner reaction zone of the flame, due to a very high turbulence intensity (Broken reaction zone). The consequence is the possible breakdown of the chemical reactions locally, due to enhanced heat losses to the preheat zone and the loss of radicals. This ultimately leading to flame extinction. [19; 2; 40; 42] For Ka>1 and Da>1 (thin reaction zones), increased turbulence intensity allows the flame structure to be entrained by turbulent eddies. These eddies can incorporate islands of burned and unburned gas, resulting in a thicker and more highly corrugated flame front. [19; 29; 40; 42] For large Dammköhler (Da>1) numbers the chemistry is fast compared with the turbulence time scale. In the well stirred reactor regime (Da<1) the molecular chemical mixing dominates. This leads to slow chemistry compared to the turbulent time scale. The Combustion is limited by chemistry and molecular effects. [19; 29; 40; 42] 5.2 CFD Modeling Challenges Under Catalyst Light-Off Conditions Figure 14 presents the simulation results for three different Pre-chamber variants at the catalyst light-off (CLO) operating point and an engine speed of 1200 rpm. The simulations were conducted using a strongly advanced spark timing of 2.95 °CA aTDC, as later spark timings proved to be less favorable for the applied simulation methodology and were not successfully completed within the simulation campaign. The detailed chemistry approach with SAGE leads to a significant underestimating of the heat release and the pressure of PC6 and PC11 in the cylinder. Solely for PC8 a small pressure rise and a strong heat release gradient for the first combustion phase is observed in the main chamber. The second phase of combustion is prolonged and strongly retarded. The other PC variants show an unplausible heat release shape with a prolonged and slow 68 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="69"?> second combustion phase. The limited burned fuel fraction, ranging between 78 % and 82 %, suggests substantial shortcomings in the detailed chemistry modeling. Figure 14: Simulations results with detailed chemistry for Pre-chamber variants 6,8 and 9 at 1200rpm/ CLO The Pre-chamber combustion shows the expected trend of the heat release rate with low peak values and a prolonged combustion duration, which can be caused by the very low turbulence level. The low cylinder peak temperature levels are around 1500K support the hypothesis of a simulation model issue, not capturing the chemical and flow characteristics in accurate manner. For PC11, the combination of the lowest burned fuel fraction and a low peak heat release, despite the highest geometric A/ V ratio, is inconsistent with experimental findings [24,28,30]. These studies show improved combustion stability at late ignition timings under CLO operating conditions for increasing A/ V ratios. Figure 15 shows experimental (IAV measurements) CA50 results for different Prechamber variants at the catalyst light-off (CLO) operating point, comparing configurations with varying geometric A/ V ratios and Pre-chamber volumes. The results indicate a clear tendency toward later combustion phasing with increasing A/ V ratio. From Twin Spark to Pre-chamber 69 <?page no="70"?> Figure 15: Experimental results for various Pre-chamber variants under CLO operating conditions, showing the influence of different geometric A/ V ratios and Pre-chamber volumes on CA50. The CLO combustion takes place during the piston moves towards BDC in combination with low engine speed and low cylinder temperature values. The turbulence decays already from a very low level. The combustion is highly molecular diffusion dominated, so that the turbulence-chemistry interaction must be accurately captured. The underestimated heat release rate is a result of an under-resolved flame front. The smallest mesh size of 0.25 mm is significantly larger than the laminar flame thickness ranging form 30μm to 100μm, consequently leads to an under-estimated flame speed. As a result, the temperature gradient is under-predicted and leads on a coarse mesh to an over-prediction of the flame front thickness. As a consequence, species diffusion into the preheat and reaction zones is reduced, which can lead to an underprediction of the heat release due to increased numerical viscosity. To avoid this effect, the flame front must be sufficiently resolved numerically, requiring a mesh size on the order of the flame thickness. However, this level of refinement is computationally prohibitive and incompatible with typical RANS-based simulations. Another important aspect is the accurate representation of molecular diffusion coefficients, which is particularly challenging in multi-species simulations under cold and low-turbulence conditions. Nevertheless, a detailed discussion of these modeling challenges is beyond the scope of the present work. A mesh resolution in the order of the flame thickness is necessary to capture the molecular diffusivity coefficients under CLO conditions, because the ratio between the turbulent viscosity and molecular viscosity is approximately 1. For hot engine operating conditions and higher loads the turbulent diffusion is dominating (ν t / ν>>1) and overcompensates the lack of accurate capturing the molecular diffusion. [34; 8] Two approaches were considered to improve the understanding of PC performance under CLO conditions. The preferred technical approach is the application of a thickened flame model (TFM), which could provide a more accurate representation of combustion under these conditions. The TFM model artificially thickens the flame front to make it resolvable on the computational grid while preserving the laminar flame speed and heat release rate. The efficiency sub-model proposed by Charlotte [2] provides correction by estimating the unresolved flame wrinkling and thereby compensating for the loss of flame-turbulence interaction. However, the required experimental data for model calibration are currently unavailable, preventing its application within the scope of this work. [2; 35] 70 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="71"?> An alternative approach is the evaluation of the combustion process using the Borghi- Peters regime diagram. This method enables additional insight into the flame propagation mechanisms and combustion regimes occurring in the cylinder and will be discussed in the following sections. 5.3 Turbulent Flame regimes under Full-load and CLO conditions The comparison of the Borghi-Peters-related variables for PC8 at 2900 rpm full load and 1200 rpm CLO conditions, shown in Figure 16, clearly demonstrates the fundamentally different main-chamber combustion inflammation and combustion process. At full load the significant higher turbulence level, higher temperature and stronger jets leads to Ka>100. The small turbulent eddies can penetrate the inner reaction zone of the flame and can cause local flame extinction. With Da<1 the chemical time scale determines the reaction and combustion speed. The inflammation process happens partly inside the broken reaction zone and the thin reaction zone regimes. The turbulent flame gets thickened (1-4 mm in typical ICE cases [43]) in the latter one. The mesh resolution relative to the laminar flame thickness represents an important limitation of the present detailedchemistry RANS approach (see discussion in section 5.2). At the 2900 rpm full-load condition, the high turbulence level leads to dominant turbulent diffusion, which reduces the sensitivity of the global combustion rate to the unresolved laminar flame thickness. Nevertheless, the unresolved flame structure limits the interpretation of local quantities such as flame-front topology, local extinction, and the exact position in the Borghi-Peters diagram. Under CLO conditions, where turbulent and molecular viscosity are of comparable magnitude, the same mesh limitation becomes more critical because molecular diffusion has a stronger influence on flame propagation. Pomraning et al. [43] have shown by increasing the initial turbulence in a combustion chamber simulation reduced the mesh refinement requirements. Therefore, the Borghi- Peters diagram at full load can be used as a qualitative and comparative turbulent flame regime indicator. From Twin Spark to Pre-chamber 71 <?page no="72"?> Figure 16: Comparison of turbulent combustions regime related traces (in degree crank angle after ignition) for PC8 at 1200rpm/ CLO and 2900rpm/ FL Under CLO conditions, the same mesh-resolution limitation becomes significantly more critical. The increasing length-scale ratio indicates a growing mismatch between the physical flame-front thickness and the numerical resolution, suggesting that the flame structure is strongly under-resolved under CLO conditions. Consequently, the results should be treated with increased uncertainty. The simulations reproduce the qualitative challenge of passive Pre-chamber ignition under catalyst-light-off conditions, namely weak jet-induced turbulence, low thermal reactivity, and slow main-chamber combustion development. Under CLO conditions, main-chamber combustion starts at Damköhler numbers close to or below unity, indicating that the chemical time scale is comparable to, or longer than, the turbulent mixing time scale. Together with the low turbulence level and low thermal reactivity, this suggests that the early flame propagation is governed mainly by finite-rate chemistry and molecular transport rather than by a strongly turbulence-enhanced flame speed. The Karlovitz number initially reaches values around 10, indicating a temporary thin-reaction-zone-like behavior, but then rapidly decreases toward and below unity. This indicates a transition toward corrugated or quasi-laminar flamelet behavior with weak turbulence-flame interaction. Due to the unresolved laminar flame thickness under CLO conditions, this regime interpretation should be considered qualitative, because the quantitative prediction is additionally restricted by the unresolved laminar flame-front structure. 72 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="73"?> Figure 17 illustrates mass-weighted probability of the combustion regimes ahead of the flame front in the unburned gas for Pre-chamber variant 8, both at 2900 rpm full load and 1200 rpm under CLO conditions at various crank angles after ignition (°CAaIGN). Figure 17: mass-weighted probability of the combustion regimes ahead of the flame front (unburned gas) for PC8 at 2900rpm/ FL and 1200rpm/ CLO for various combustions timings. At 2900 rpm and full load at 6.8°CAaIGn the jets entering the cylinder. A borad distribution over across flame regimes, indicating a highly heterogenous turbulencechemistry interaction caused by the strong Pre-chamber jets and the associated shear layers in the main chamber. The spread toward high Karlovitz numbers shows that parts From Twin Spark to Pre-chamber 73 <?page no="74"?> of the flame-front region are exposed to intense small-scale turbulence and locally approach broken-reaction-zone and well stirred reactor conditions. For subsequent combustions timings, namely 10 % (CA10) and 50 % (CA50) of heat release, a substantial portion moves towards lower turbulence level and higher length-scale ratios in the thin reaction zone and the corrugated flamelet region. With progressing combustion, the flame-front evolution becomes increasingly similar to conventional spark-ignition flame propagation. Only a smaller fraction of the flame-front cells is located in the laminar flame regime, indicating that most of the combustion process is still influenced by turbulenceflame interaction. At 1200 rpm CLO the mass-weighted probability distribution differs fundamentally from the 2900 rpm full load case. The jets entering the cylinder 10.1 °CAaIGN with a centroid shifted toward the well stirred reactor regime and thin reaction zone regimes with Dammköhler numbers close to or below unity. Together with the low turbulence level and low thermal reactivity, this suggests that the early main-chamber combustion is governed mainly by chemistry-limited and molecular transport rather than by a strongly turbulence-enhanced flame speed. Initially, the Karlovitz number reaches values around 10, indicating a temporary thin-reaction-zone-like behavior. With progressing combustion at 10 % of heat release, the distribution rapidly shifts toward lower Karlovitz numbers and higher length-scale ratios, corresponding to a transition toward corrugated or quasilaminar flamelet behavior. Further combustion progress (CA50) shifts the flame toward laminar or quasi-laminar behavior with lower propagation speeds. Compared to the 2900 rpm full-load case, considerably more cells occupy the laminar regime, indicating weak turbulence-flame interaction and slow chemistry-controlled combustion under CLO conditions. Due to the unresolved laminar flame thickness, the CLO scatter distribution provides only a qualitative indication of the local combustion regime. 5.4 Turbulent Flame regimes under CLO conditions for selected PC variants The following section discusses the CLO results for Pre-chamber variants PC6, PC8, and PC11, considering the limitations introduced by the unresolved laminar flame-front. Figure 18 compares the velocity magnitude and temperature spread over the Pre-chamber orifices. The three Pre-chamber variants have the same volume, but differ in A/ V ratio, nozzle layout and jet targeting. 74 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="75"?> Figure 18: Comparison of velocity and temperature spread over PC orifices for PC6, PC8 and PC11 at 1200rpm/ CLO PC6 and PC8 with a comparable A/ V ratio show higher peak velocities and subsequently high jet momentums than PC11, while PC11 exhibits a weaker jet momentum level during the early discharge phase. The ability to generate strong local turbulence (Figure 19) by intensive jets is reduced for this variant. The additional jet-momentum evaluation shows that PC8 provides the most favorable combination of high jet momentum and low bore-tobore variation. The asymmetric PC8 layout with optimized jet targeting produces comparable hot and reactive jets at three orifices, while the fourth orifice exhibits a temporarily weaker and delayed discharge during the early jet-formation phase. The jets propagate freely with no jet-to-jet interaction in the early combustion phase into the main chamber. PC6 also reaches high peak jet velocities, but the larger coefficient of variation indicates a less homogeneous momentum distribution between the individual orifices. The OH visualization further shows strongly unequal jet development, with one or two orifices From Twin Spark to Pre-chamber 75 <?page no="76"?> producing significantly weaker jets. In addition, the PC6 jets directly impinge on the piston surface and merge with each other, reducing the effective reactive flame surface available for distributed ignition in the main chamber. PC11, despite its higher A/ V ratio and larger nozzle area, shows lower jet momentum and therefore does not achieve the same jet-induced turbulence level as PC8. Although PC11 provides the highest A/ V ratio and the largest nozzle area, the lower jet momentum limits its ability to generate strong turbulence in the main chamber. The OH visualization indicates early jet-piston interaction, jet-to-jet merging, and delayed or weak discharge from individual orifices. Figure 19 compares the mass-weighted Borghi-Peters distributions ahead of the flame front for PC8 and PC11 at 1200 rpm/ CLO. Both variants are shifted toward weakly turbulence-enhanced combustion regimes , reflecting the low temperature, low turbulence intensity, and rapidly decaying jet-induced turbulence under CLO conditions. The distributions therefore indicate that the early main-chamber combustion is not governed by a robust turbulent-flame-speed-controlled process, but rather by finite-rate chemistry, molecular transport, and limited turbulence-flame interaction. At 2 % heat release, the asymmetric PC8 variant is still mainly located in the well-stirred reactor and thin-reaction-zone regimes. This indicates stronger turbulence-flame interaction and a higher effective turbulent flame-speed contribution compared with PC11. In contrast, the symmetric high-A/ V-ratio variant PC11 already shifts further toward the corrugated flamelet, wrinkled flamelet, and laminar flame regimes, with slightly lower maximum Karlovitz numbers and reduced turbulence intensity during the jets enter the cylinder. A substantial fraction of the PC11 pre-front cells is located in a quasi-laminar regime, which explains the reduced burn velocity due to the limited turbulent enhancement of flame propagation. 76 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="77"?> Figure 19: mass-weighted probability of the combustion regimes ahead of the flame front (unburned gas) for PC8 and PC11 at 1200 rpm/ CLO for various combustions timings. From approximately 10 % heat release onward, both variants show a strong shift toward quasi-laminar flame propagation. This demonstrates the disadvantageous combustion behavior of passive Pre-chamber under CLO conditions, where the generated jets are no longer able to maintain a strongly turbulence-enhanced main-chamber combustion process. For PC11, this behavior is particularly relevant because the higher A/ V ratio primarily improves Pre-chamber scavenging, but does not necessarily generate high jet momentum or deep jet penetration into the main chamber. In the present laterally offset From Twin Spark to Pre-chamber 77 <?page no="78"?> mounting position, the symmetric nozzle layout is therefore disadvantageous. The OHbased flame visualization indicates early jet-piston interaction, jet-to-jet merging, and delayed or weak discharge from individual orifices. Consequently, the larger A/ V ratio of PC11 does not provide an improvement over the jet-targeting-optimized PC8 configuration in the present CFD geometry. The A/ V ratio should therefore not be regarded as the sole governing design parameter. Passive Prechamber performance is controlled by a complex interaction of scavenging capability, jet momentum, thermal reactivity, jet penetration, nozzle orientation, and engine-specific combustion-chamber integration, which is not yet fully understood. The experimental trend favoring PC11 under CLO-like conditions should therefore not be transferred directly to the present V8 engine geometry without considering the mounting position. In the discussed experimental configurations, the Pre-chamber spark plug is centrally located in the combustion chamber roof, which supports a more symmetric distribution of the reactive jets. In the present V8 cylinder head, however, the Pre-chamber is mounted in a laterally offset position, making jet targeting, piston interaction, and jet-to-jet merging much more influential. Under these boundary conditions, PC8 benefits from its asymmetric nozzle layout, whereas the symmetric PC11 concept loses part of its geometric scavenging advantage. 6 Conclusion This study investigated the applicability of passive Pre-chamber ignition as an upgrade path for a production-representative 6.6 L naturally aspirated twin-spark gasoline V8 engine. By replacing one of the two conventional spark plugs with a passive Pre-chamber while retaining stoichiometric operation under EGR dilution, the work aimed to improve combustion efficiency, combustion robustness, and knock resistance without introducing major engine architecture changes. The investigation combined predictive 1D simulations and detailed 3D CFD analyses to evaluate compression-ratio potential and identify an optimized Pre-chamber design. The 1D simulation study demonstrated that the faster combustion associated with passive Pre-chamber ignition substantially improves knock resistance and combustion phasing, particularly at knock-limited operating conditions. For regular RON92 fuel, the PC concept enabled a compression-ratio increase from the baseline 12.8 to approximately 14.5 while maintaining favorable combustion behavior. The faster burn rate resulted in earlier combustion phasing, improved brake efficiency, and increased full-load capability compared to the Twin spark system operated with RON96 fuel. Consequently a portion of the Pre-chamber benefit can be utilized for efficiency enhancement through higher compression ratio even though a lower fuel quality (RON92) is used, without incurring significant drawbacks in full-load performance of the baseline Twin spark engine operated with RON96 fuel. Figure 20 shows the comparison of both ignition systems at the two operating points 1900 rpm best BTE and 2900 rpm max. torque. The results were obtained from the calibrated 1D simulation model using the predicted burn rates out of the CFD investigation with PC8 and the burn rates derived from the Twin spark TPA results as inputs. Despite the use of lower-octane fuel, the PC combustion system, in combination with an 78 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="79"?> increased compression ratio, achieves an efficiency improvement of 0.4 % at the best-BTE operating point, while exhibiting a 3°CA faster combustion. Furthermore, additional dilution potential is available with the PC for further efficiency gain. At max. torque operating point the PC8 shows comparable combustion and efficiency performance despite the changes in CR and fuel quality. Figure 20: Comparison engine results with PC8 with RON92 fuel and CR14.5 vs. TS with RON96.5 fuel and CR12.8 at best BTE and max. torque operating point The CFD investigations showed that Pre-chamber performance is strongly governed by the interaction of chamber volume, A/ V-ratio, mounting position, and jet orientation. The results confirmed that combustion performance cannot be optimized by a single geometric parameter alone. In-stead, successful implementation requires a balanced design that ensures sufficient Pre-chamber scavenging, favorable mixture preparation, and effective jet penetration into the main chamber. Among the investigated concepts, the asymmetric layouts with application-specific jet targeting demonstrated the most favorable overall combustion behavior. In particular, variant PC8 provided the fastest combustion initiation, shortest burn duration, earliest combustion phasing, and highest indicated work at full load. A key finding of the study is the importance of jet targeting. Compared with symmetric configurations, the optimized asymmetric designs generated deeper jet penetration, reduced jet-to-jet interaction, improved combustion chamber coverage, and lower flame-wall interaction. These characteristics promoted faster and more homogeneous combustion development in both full-load and best-efficiency operating conditions. Furthermore, the preferred mounting position near the injector improved mixture transport into the Pre-chamber and supported a richer and more reactive local mixture, which contributed significantly to ignition quality and combustion acceleration. At the best-efficiency operating point, the optimized asymmetric concept required substantially less spark advance to achieve the same combustion phasing compared with From Twin Spark to Pre-chamber 79 <?page no="80"?> the symmetric variants. Although this advantage did not directly translate into a measurable brake-efficiency gain under the fixed CA50 comparison conditions, it indicates additional potential for extending dilution limits and increasing EGR rates in future development activities. The investigation also highlighted the limitations of passive Pre-chamber systems under catalyst-light-off conditions. Low turbulence levels, high residual-gas dilution, reduced temperatures, and increased relevance of molecular diffusion significantly weaken jet ignition effectiveness. The passive Pre-chamber performance is governed by a complex interaction of jet momentum, jet targeting, scavenging capability, thermal reactivity, and local turbulencechemistry interaction, rather than by A/ V ratio alone. An asymmetric target-optimized nozzle layout with a high A/ V ratio should be considered for CLO operation. The detailed-chemistry CFD approach used in this work was found to underpredict combustion progress under these conditions, indicating that conventional RANS-based detailed chemistry simulations are not sufficient to accurately capture cold and highly diffusion-controlled Pre-chamber combustion. Future work should therefore focus on advanced combustion-modeling approaches, such as thickened-flame methodologies, combined with dedicated experimental validation. Overall, the results demonstrate that passive Pre-chamber ignition represents a highly promising technology for future stoichiometric gasoline engines. When properly integrated and optimized for the specific combustion-chamber geometry, the system can significantly accelerate combustion, improve knock tolerance, and enable substantial compression-ratio increases. The study shows that a carefully designed side-mounted passive Pre-chamber can provide an effective pathway toward higher thermal efficiency while preserving conventional stoichiometric combustion and aftertreatment compatibility. 7 References [1] Yakhot, V.; Orszag, S. A.; Thangam, S.; Gatski, T. B.; Speziale, C. G.: “Development of turbulence models for shear flows by a double expansion technique”. In: Physics of Fluids A: Fluid Dynamics 4.7 (1992), pp. 1510-1520. issn: 0899-8213. doi: 10.1063/ 1.858424. [2] Richards, K . J ., Senecal, P. K ., and Pomraning, E., CONVERGE 5.1 Manual, Convergent Science, Madison, WI (2025) [3] Senecal, P. K.; Pomraning, E.; Richards, K. J.; Briggs, T. E.; Choi, C. Y.; Mcdavid, R. M.; Patterson, M. A.: “Multi-Dimensional Modeling of Direct-Injection Diesel Spray Liquid Length and Flame Lift-off Length using CFD and Parallel Detailed Chemistry”. In: SAE Technical Paper Series. SAE Technical Paper Series. SAE International, 2003. doi: 10.4271/ 2003-01-1043. [4] Stiesch, Gunnar: Modeling engine spray and combustion processes. Berlin and New York : Springer, 2003 [5] Clenci, Adrian C. ; Iorga-Simn, Victor ; Deligant, Michael ; Podevin, Pierre ; Descombes, Georges ; Niculescu, Rodica: A CFD (computational fluid dynamics) study on the effects of operating an engine with low intake valve lift at idle corresponding speed. In: Energy 71 (2014), S. 202-217 [6] Berni, F.; Cicalese, G.; Fontanesi, S.: “A modified thermal wall function for the estimation of gasto-wall heat fluxes in CFD in-cylinder simulations of high performance spark-ignition engines”. 80 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="81"?> In: Applied Thermal Engineering 115 (2017), pp. 1045-1062. issn: 1359-4311. doi: 10.1016/ j. applthermaleng.2017.01.055. [7] Babajimopoulos, A.; Assanis, D. N.; Flowers, D. L.; Aceves, S. M.; Hessel, R. P.: “A fully coupled computational fluid dynamics and multi-zone model with detailed chemical kinetics for the simulation of premixed charge compression ignition engines”. In: International Journal of Engine Research 6.5 (2005), pp. 497-512. doi: 10.1243/ 146808705X30503 [8] Wöbke, M., Reinicke, P.-B., Rieß, M., von Römer, L., & Sens, M. (2019). Numerical investigations of the early flame propagation of a Pre-chamber ignition system in a high pressure combustion cell. In Gedenkschrift für Prof. Roland Baar. Technische Universität Berlin. https: / / doi.org/ 10.14279/ depositonce-9822 [9] Dirrenberger, P., Glaude, P.-A., Bounaceur, R., Le Gall, H., Pires da Cruz, A., Konnov, A. A., & Battin-Leclerc, F. (2014). Laminar burning velocity of gasolines with addition of ethanol. Fuel, 115, 162-169. https: / / doi.org/ 10.1016/ j.fuel.2013.07.015 [10] Hesse, R, Schwenzer, C, Glaznev, R, Esposito, S, Fenard, Y, Pitsch, H & Beeckmann, J 2022, ‘Experimental- Based Laminar Flame Speed Approximation Formulas of Efficiency-Optimized Biofuels for SI-Engine Modeling’, SAE International Journal of Advances and Current Practices in Mobility. https: / / doi.org/ 10.4271/ 2022-24-0032 [11] Jerzembeck, S., Peters, N., Pepiot-Desjardins, P., & Pitsch, H. (2009). Laminar burning velocities at high pressure for primary reference fuels and gasoline: Experimental and numerical investigation. Combustion and Flame, 156(2), 292-301. https: / / doi.org/ 10.1016/ j.combustflame.2008.11.009 [12] Andrae, J. und Head, R. A. (2009). HCCI experiments with gasoline surrogate fuels modeled by a semidetailed chemical kinetic model, Combustion and Flame 156(4): 842-851. [13] Cai, L., Ramalingam, A., Minwegen, H., Alexander Heufer, K. und Pitsch, H. (2019). Impact of exhaust gas recirculation on ignition delay times of gasoline fuel: An experimental and modeling study, Proceedings of the Combustion Institute 37(1): 639-647. [14] Samimi Abianeh, O. (2015). Development of a New Skeletal Chemical Kinetic Mechanism for Ethanol Reference Fuel, Journal of Engineering for Gas Turbines and Power 137. [15] Andrae, J. C. G.; Kovács, T.Evaluation of Adding an Olefin to Mixtures of Primary Reference Fuels and Toluene To Model the Oxidation of a Fully Blended Gasoline.Energy & Fuels, 2016, 30 (9), 7674-7683. DOI: 10.1021/ acs.energyfuels.6b01193. [16] Fandakov, A.: A Phenomenological Knock Model for the Development of Future Engine Concepts. Wiesbaden: Springer Vieweg, 2019. ISBN 978-3-658-24874-1. DOI: 10.1007/ 978-3-658- 24875-8. [17] Colin, O., Benkenida, A. und Angelberger, C. (2003). 3d Modeling of Mixing, Ignition and Combustion Phenomena in Highly Stratified Gasoline Engines, Oil & Gas Science and Technology - Rev. IFP 58(1): 47-62. [18] Williams FA. Turbulent combustion. In: Buckmaster JD (Ed), The Mathematics of Combustion. Society for Industrial & Applied Mathematics, 1985, pp. 197-1318. [19] Peters, N.: Turbulent Combustion. Cambridge: Cambridge University Press, 2000. ISBN 978-0- 521-66082-2 [20] Peters, N.: Combustion Theory. Lecture notes, CEFRC Summer School, Princeton, 28 June-2 July 2010. RWTH Aachen University, 2010. URL: https: / / cefrc.princeton.edu/ sites/ g/ files/ toruqf1071/ files/ Files/ 2010 %20Lecture%20Notes/ Norbert%20Peters/ Peters_Summerschool_reference.pdf, Access date: 22.07.2026 [21] Pati, A., Gierth, S., Haspel, P., Hasse, C. et al., “Strategies to Define Surrogate Fuels for the Description of the Multicomponent Evaporation Behavior of Hydrocarbon Fuels,” SAE Technical Paper 2018-01-1692, 2018, doi: 10.4271/ 2018-01-1692. From Twin Spark to Pre-chamber 81 <?page no="82"?> [22] Fandakov, A., Grill, M., Bargende, M., Mally, M. P., Günther, M., Pischinger, S., Cai, L., Pitsch, H., Heufer, K. A., Ramalingam, A. K., Minwegen, H. J. & Kulzer, A. C. (2018). Development of a Model for Predicting the Knock Boundary in Consideration of Cooled Exhaust Gas Recirculation at Full Load. In: M. Günther & M. Sens (Eds.), Knocking in Gasoline Engines: 5th International Conference, Berlin, Germany (pp. 143-185). Cham: Springer. https: / / doi.org/ 10.1007/ 978-3-319- 69760-4_9. [23] Beume, N., Naujoks, B., & Emmerich, M. (2007). SMS-EMOA: Multiobjective Selection Based on Dominated Hypervolume. European Journal of Operational Research, 181(3), 1653-1669. https: / / doi.org/ 10.1016/ j.ejor.2006.08.008 [24] Korkmaz, M., Juressen, S. E., Rößmann, D., Kapus, P. E., and Pino, S., “Overview of Current Developments of Pre-chamber Spark Plugs for Passenger Car Applications,” SAE Technical Paper 2025-24-0152, 2025, doi: 10.4271/ 2025-24-0152. [25] Novella, R., Gomez-Soriano, J., Martinez-Hernandiz, P. J., Libert, C., et al., “Improving the Performance of the Passive Pre-chamber Ignition Concept for Spark-Ignition Engines Fueled with Natural Gas,” Fuel, Vol. 290, 2021, doi: 10.1016/ j.fuel.2020.119971. [26] Novella, R., Gomez-Soriano, J., Barbery, I., and Libert, C., “Numerical Analysis of the Passive Prechamber Ignition Concept for Light Duty Applications,” Applied Thermal Engineering 213, 2022, doi: 10.1016/ j.applthermaleng.2022.118610. [27] Benajes, J., Novella, R., Gomez-Soriano, J., Barbery, I., Libert, C., Rampanarivo, F., and Dabiri, M., “Computational Assessment Towards Understanding the Energy Conversion and Combustion Process of Lean Mixtures in Passive Pre-chamber Ignited Engines,” Applied Energy, Vol. 261, 2020, doi: 10.1016/ j.apenergy.2019.114433. [28] Sens, M., Binder, E., Reinicke, P., Riess, M., Stappenbeck, T., and Wöbke, M., “Pre-chamber Ignition and Promising Complementary Technologies,” 27th Aachen Colloquium Automobile and Engine Technology, Aachen, Germany, 2018. [29] Piano, A., Scalambro, A., Millo, F., Catapano, F.,Sementa, P., Di Iorio, S., and Bianco, A., “CFDbased Methodology for the Characterization of the Combustion Process of a Passive Prechamber Gasoline Engine,” Fuel 346, 2023, doi: 10.1016/ j.fuel.2023.128328. [30] Wöbke, Marcus, Reinicke, Paul-Benjamin, Rieß, Michael, Römer, Lorenz von, and Sens, Marc, “Characterization of the Ignition and Early Flame Propagation of Pre-chamber Ignition System in a High Pressure Combustion Cell,” in Ignition Systems for Gasoline Engines: 4th International Conference, Berlin, Germany, December 6 to 7, 2018, pp. 385-423, doi: 10.5445/ IR/ 1000088608. [31] Shah, A., Tunestål, P., & Johansson, B. (2015). Effect of Pre-chamber Volume and Nozzle Diameter on Pre-chamber Ignition in Heavy Duty Natural Gas Engines. SAE Technical Paper 2015-01-0867. https: / / doi.org/ 10.4271/ 2015-01-0867 [32] Shah, A. (2015). Improving the Efficiency of Gas Engines using Pre-chamber Ignition [Doctoral thesis, Lund University]. Division of Combustion Engines, Department of Energy Sciences, Faculty of Engineering, Lund University. ISBN 978-91-7623-562-1. [33] Attard, W. P., Fraser, N., Parsons, P., & Toulson, E. (2010). A turbulent jet ignition Pre-chamber combustion system for large fuel economy improvements in a modern vehicle powertrain. SAE International Journal of Engines, 3(2010-01-1457), 20-37 [34] Ravindran, A. C., & Kokjohn, S. L. (2023). The challenges of using detailed chemistry model for simulating direct injection spark ignition engine combustion during cold-start. International Journal of Engine Research, 24(1), 161-177. https: / / doi.org/ 10.1177/ 14680874211045968 [35] Posch, S., Gößnitzer, C., Lang, M., Novella, R., Steiner, H., & Wimmer, A. (2025). Turbulent combustion modeling for internal combustion engine CFD: A review. Progress in Energy and Combustion Science, 106, Article 101200. https: / / doi.org/ 10.1016/ j.pecs.2024.101200 82 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="83"?> [36] Kim, J., Scarcelli, R., Som, S., Shah, A., Biruduganti, M. S., & Longman, D. E. (2022). Numerical investigation of a fueled Pre-chamber spark-ignition natural gas engine. International Journal of Engine Research, 23(9). https: / / doi.org/ 10.1177/ 14680874211020180 [37] Sens, M., Binder, E., Benz, A., Kramer, L., Schultalbers, M., & Blumenröde, K. (2018, April). Prechamber ignition as a key tech-nology for highly efficient SI engines-new approaches and operating strategies. In Proceedings of the 39th International Vienna Motor Symposium, Vienna, Germany (pp. 26-27) [38] Stadler, A., Sauerland, H., Härtl, M., & Wachtmeister, G. (2020, April). The potential of gasoline fueled pre chamber ignition combined with elevated compression ratio. In WCX SAE World Congress Experience. SAE Technical Paper [39] Shelby, M. H., Case, M. E., & Chesney, L. A. (2024, April). Next generation high efficiency boosted engine concept. In WCX SAE World Congress Experience. SAE Technical Paper [40] Moeck, J. (2015). Verbrennungsdynamik. Lecture notes, draft version of 1 August 2015, Fachgebiet Verbrennungsdynamik, Institut für Strömungsmechanik und Technische Akustik, Technische Universität Berlin. [41] Sens, M, Binder, E.: (2019) Pre-chamber Ignition as a Key Technology for Future Powertrain Fleets. MTZ worldwide, vol. 80, no. 2, pp. 44-51, https: / / doi.org/ 10.1007/ s38313-018-0150-1 [42] Frühhaber, J., & Lauer, T. (2021). Numerical investigation of the turbulent flame propagation in dual fuel engines by means of large eddy simulation. Energies, 14(16), 5036. https: / / doi.org/ 10.3390/ en14165036 [43] Pomraning, E., Richards, K., & Senecal, P. K. (2014). Modeling turbulent combustion using a RANS model, detailed chemistry, and adaptive mesh refinement. SAE Technical Paper 2014-01- 1116. https: / / doi.org/ 10.4271/ 2014-01-1116 [44] MotorTrend (2022) Stellantis Hurricane Twin-Turbo Inline-Six Engine Deep Dive. Available at: https: / / www.motortrend.com/ news/ stellantis-hurricane-4-turbo-engine-deep-dive (Accessed: 14 August 2026) [45] BMW M (2026) BMW M Ignite: New Technology for the S58 Engine. Available at: https: / / www. bmw-m.com/ en/ topics/ magazine-article-pool/ m-ignite-technology-for-s58.html (Accessed: 14 August 2026) 8 Abbreviations and formula symbols A/ V Ratio of Pre-chamber orifice cross section area to volume AMR adaptive mesh refinement CEQ Chemical Equilibrium CFD Computational Fluid Dynamics CR Compression Ratio ECFM Extended Coherent Flame Model EGR Exhaust Gas Recirculation GDI Gasoline direct injection H/ C-ratio Hydrogen/ carbon atom ratio ICE Internal combustion engine KHRT Kelvin-Helmholtz-Rayleigh-Taylor LHV Lower heating value MC Main chamber MON Motor octane number OP Operating Point PC Pre-chamber PISO Pressure Implicit with Splitting of Operators From Twin Spark to Pre-chamber 83 <?page no="84"?> RNG renormalization group RON Research octane number SA Spark Angle SI Spark ignition TDC Top dead center TS Twin spark TPA Three pressure analysis WOT Wide open throttle 84 Marcus Wöbke, Antowan Zyada, Michael Rieß, Adalbert Wolany <?page no="85"?> Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine Giovanni Bonandrini 1 , Antonio Aliperti 1 , Davide Bizzarri 1 , Paolo Scarpato 1 Abstract: The adoption of pre-chamber systems in internal combustion engines, particularly in the passive configuration, represents a viable and effective solution for high-performance applications, both in production and motorsport contexts. For road-legal engines, the main advantages include the potential increase in power density and the extension of the operating range at stoichiometric conditions. Efficiency improvements, also linked to the capability of achieving stable lean combustion, constitute an additional benefit, which is particularly relevant for racing applications. This paper presents an investigation of a passive pre-chamber system based on the Lamborghini L411 high-performance turbocharged engine. The study aims to assess the potential benefits of such a system for road-legal applications. The first part of the work focuses on design aspects, including the overall system layout and the integration of the pre-chamber, spark plug selection, pre-chamber geometry and material. The second part describes the numerical simulation activities. A wide set of onedimensional simulations was carried out to identify optimal solutions in terms of compression ratio, camshaft profiles, and intake duct configurations. In parallel, several aspects related to permeability and tumble trade-offs, cooling, and combustion behavior were investigated through three-dimensional simulations. Finally, experimental results obtained from single-cylinder engine tests are presented. These tests focused on low-load operating conditions, where misfire issues may arise and the use of an auxiliary spark plug can be required, as well as on high-load conditions, where key aspects include knock analysis, pre-ignition risk, the capability to achieve higher in-cylinder pressures, and the extension of the λ = 1 operating range. Introduction The increasing demand for higher thermal efficiency, reduced emissions, and enhanced specific power has stimulated significant research into advanced combustion systems for spark-ignition engines. Among the most promising technologies, pre-chamber ignition systems, also known as jet ignition or turbulent jet ignition (TJI) systems, have emerged as 1 Automobili Lamborghini S. p. A. <?page no="86"?> an effective solution to overcome the limitations of conventional spark ignition, particularly under highly diluted, lean-burn, and knock-limited operating conditions. The operating principle of a pre-chamber combustion system is based on the ignition of a fuel-air mixture inside a small auxiliary chamber connected to the main combustion chamber through calibrated orifices. Following ignition, high-energy reactive jets are expelled into the main chamber, generating multiple ignition sites and significantly accelerating flame propagation. This distributed ignition process results in faster combustion, improved combustion stability, extended lean-burn capability, and enhanced knock resistance when compared to traditional spark ignition systems. Although the concept of pre-chamber ignition dates back several decades, recent developments in combustion modeling, fuel injection systems, and engine control technologies have enabled its practical implementation in both motorsport and highperformance production engines. In top-level racing applications, pre-chamber systems have contributed to unprecedented thermal efficiencies by enabling highly diluted combustion while maintaining fast heat-release rates and robust ignition characteristics. For modern production sports car engines, pre-chamber technology offers an attractive pathway toward achieving the often conflicting objectives of high specific power output, stringent emissions compliance, and improved fuel economy. The primary value of TJI for modern performance-oriented powertrains should not be viewed solely as an enabler of ultra-lean combustion, but rather as a technology capable of expanding the λ = 1 operating envelope, allowing higher loads, increased compression ratios, advanced spark timing, and improved thermal efficiency while retaining compatibility with conventional aftertreatment systems. The present work investigates the possible application of pre-chamber combustion systems to a high-performance production engine, with particular focus on their capability to extend the λ = 1 operating region and improve combustion characteristics under high-load conditions. Consequently, the paper first discusses the basic hardware adaptations required for the existing engine. Subsequently, 1D and 3D simulations of engine performance and cooling are presented to determine the optimal configuration. Finally, the results obtained from experimental testing on a single-cylinder engine are analyzed and discussed. 1 Activity Description and Scope In parallel with the development of the Lamborghini L411 combustion system [1], an investigation into the application of pre-chamber combustion technology was conducted. The study evaluated several combustion system modifications while preserving the fundamental architecture of the baseline L411 engine. The key specifications of the reference engine are reported in the table below. Lamborghini L411 experimental engine Type V8 90°, flat plane cranckshaft Intake and exhaust configuration Twin turbocharged, hot-vee exhaust Bore x Stroke [mm] 90 x 78.5 Displacement [cm3] 3,995 86 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="87"?> Lamborghini L411 experimental engine Fuel injection system GDI, max pressure 350 bar Compression ratio 9.5 Maximum engine speed [rpm] 10,000 Combustion system design symmetric pent-roof 4-valve Spark plug size M10 The pre-chamber investigations were conducted on a single-cylinder engine (SCE). The use of an SCE was motivated by the ability to achieve a more accurate assessment of combustion system behavior while significantly reducing development time and costs compared with full-engine testing. Figure 1: L411 single cylinder engine (SCE) The experimental campaign was aimed at evaluating the fundamental functionality of a passive pre-chamber combustion system derived from the Lamborghini L411 engine architecture. The main objectives of the investigation were: • Evaluation of cylinder head design solutions, including the pre-chamber insert, water jacket configuration, sealing strategy, and related components. • Assessment of cooling system design and thermal management requirements. • Optimization of the combustion system, including pre-chamber geometry, spark plug configuration, intake port design, piston geometry, and intake cam profile. • Evaluation of the potential for increased power density and extension of the λ = 1 operating range. • Assessment of engine controllability at low speed and low load conditions. • Analysis of in-cylinder pressure characteristics and combustion behavior. 2 System basic design As previously stated, the pre-chamber system was evaluated on the Lamborghini L411 single-cylinder engine (SCE). Modifications were limited primarily to the cylinder head, particularly its lower section, and to the piston. The spark plug seat was enlarged to accommodate an insert incorporating the pre-chamber internal geometry and nozzle arrangement. Several insert configurations were designed and experimentally assessed. An M10 spark plug was installed within the insert, and different spark plug variants were Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 87 <?page no="88"?> evaluated during the development process. The fuel injector remained unchanged, while multiple piston designs and intake camshaft profiles were investigated. In a subsequent evolution of the system, referred to as the wet configuration, the insert architecture was completely redesigned. This development also required modifications to both the intake ports and the cylinder-head water jacket. For both the dry and wet configurations, the design did not incorporate an additional spark plug. Such a feature would likely be required in a production-intent application to ensure compliance with emissions targets and to guarantee robust engine operation during cold start and very low-load conditions. 2.1 Pre-chamber assembly One of the primary design requirements was the accurate positioning of the pre-chamber insert within the cylinder head, as the effectiveness of the system strongly depends on the precise orientation of the individual pre-chamber jets. The insert was therefore designed to be positively located both axially and rotationally. Axial clamping was achieved by means of a threaded retaining ring, whereas circumferential alignment was ensured through a dedicated mechanical indexing feature formed by the mating geometries of the cylinder head and the insert. This arrangement enabled precise and repeatable orientation of the pre-chamber nozzles, ensuring consistent jet direction throughout the experimental campaign. Figure 2: Pre-chamber insert assembly procedure A dedicated tool is required to tighten the threaded retaining ring, and removal of the upper cylinder head assembly is not needed to replace the pre-chamber insert. The system was specifically developed to enable rapid evaluation of different pre-chamber geometries 88 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="89"?> and nozzle configurations on the single-cylinder engine (SCE) through relatively quick insert replacement. From a production perspective, however, a substantially different design approach would be required. Since frequent insert replacement would no longer be necessary, the assembly concept should be simplified and optimized for automated manufacturing processes, while also reducing component count and assembly time. The selection of the insert material represented a key aspect. On one hand, high thermal conductivity is desirable to maximize heat transfer from the pre-chamber to the surrounding cylinder head, thereby preventing the formation of local hot spots that could promote pre-ignition. On the other hand, the material must provide adequate mechanical strength and good machinability to allow the manufacture of the complex internal geometry and nozzle layout. In addition, the material must retain sufficient mechanical properties at elevated temperatures and exhibit a high melting point to ensure durability under severe operating conditions. Following a trade-off analysis of several candidate materials, a copper-based alloy was selected as the most suitable solution. The use of berylliumcontaining materials was excluded due to their high cost and the health and safety concerns associated with their toxicity. More generally, the insert cost represents an important factor in evaluating the suitability of the pre-chamber system for series production. Both material costs and the manufacturing effort required to machine the component must be carefully considered when assessing the overall feasibility of the concept. 2.2 Pre-chamber head configurations The cylinder head was developed in two different configurations. In the first configuration, called “dry” configuration, the pre-chamber insert is in direct contact only with the aluminum structure of the cylinder head, allowing the original production SI engine casting to be retained with limited machining modifications. In the second configuration, called “wet” configuration, the insert is also directly exposed to the coolant. This solution required a complete redesign of the cylinder head, including the water jacket, and therefore the manufacture of a new casting. For both configurations, and particularly for the concept in which the insert is directly cooled by the coolant, the sealing strategy represents a critical design aspect. Should the system be considered for production application, both concepts would require dedicated validation through durability and reliability testing to ensure adequate sealing performance and long-term robustness under all operating conditions. 2.2.1 First pre-chamber configuration The first configuration was based on a cylinder head casting closely derived from the production design, featuring the standard intake port (named Bst01) and specific constraints on the maximum allowable pre-chamber dimensions. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 89 <?page no="90"?> Figure 3: Dry prechamber design Axial positioning and sealing of the insert were achieved by means of a copper washer located on the combustion chamber side. This solution ensured both proper axial retention and effective sealing at the insert-to-head interface. A small radial clearance was intentionally introduced between the insert and the cylinder head to accommodate the different thermal expansion coefficients of the two materials during engine operation. As a result, direct contact between the insert and the cylinder head along the radial surfaces could not be guaranteed under all operating conditions. Consequently, the heat transfer across this interface was not precisely predictable and could vary depending on the actual thermal and mechanical operating state of the assembly. 2.2.2 Second pre-chamber configuration The second configuration features direct cooling of the pre-chamber insert and high-flow intake ports. In addition to the improved thermal management provided by direct coolant exposure, the wet pre-chamber architecture is expected to deliver enhanced performance due to the relaxation of geometric constraints, allowing greater design freedom and increased available space for pre-chamber optimization. Following the simulation activities, in the new cylinder head casting developed for the second configuration an updated intake port design (called IP4785) was also introduced. The water jacket was designed with a modular architecture, enabling compatibility with both the original and the modified intake port layouts. However, only the casting featuring IP4785 intake ports was ultimately manufactured and tested in wet configuration. 90 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="91"?> Figure 4: Wet prechamber design In the wet pre-chamber configuration, axial positioning of the insert is still achieved through a mechanical stop on the cylinder head, using a copper sealing washer located on the combustion chamber side. Owing to the increased packaging space available in this architecture, the washer dimensions were enlarged compared to the dry configuration. This modification was introduced to ensure reliable sealing performance, particularly considering the direct exposure of the insert interface to the engine coolant. 3 System definition by 1D simulations The objective of the system development was to assess the potential of combining prechamber combustion technology, a Miller-cycle strategy, and high-flow intake ports to extend the λ = 1 operating range while minimizing modifications to the base engine architecture. In particular, the proposed concept was intended to require changes only to the cylinder head, piston, and intake camshaft, while retaining the original intake and exhaust systems as well as the production turbocharger. For the 1D simulation activities, the following workflow was adopted. The first step involved the development and calibration of a dedicated combustion model capable of accurately representing Turbulent Jet Ignition (TJI) operation. Once the simulation framework had been established, a series of hardware modifications were evaluated with the aim of improving engine performance and extending the λ = 1 operating range while preserving the fundamental characteristics of the baseline engine design. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 91 <?page no="92"?> Figure 5: L411 engine 1D model The adoption of an aggressive Miller-cycle strategy required the introduction of a new high-flow intake port design, named IP4785, defined by means of a specific optimization assisted by AI [2] as an alternative to the baseline BST01 configuration. While providing significantly improved flow capacity, the new port geometry resulted in a reduction of incylinder charge motion. Compared with the BST01 intake port, the IP4785 design increased the flow coefficient by up to 25 %, while reducing the tumble coefficient by a similar percentage. This trade-off was considered acceptable in view of the enhanced combustion robustness expected from the application of pre-chamber combustion technology. 92 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="93"?> Figure 6: BST01 and IP4785 intake ducts design and flow coefficients The 1D combustion model was updated by replacing the standard predictive approach with a non-predictive model. Combustion duration and knock index were imposed as input parameters and calibrated using an initial set of 3D CFD combustion simulations. Separate calibrations were developed for conventional Spark Ignition (SI) and Turbulent Jet Ignition (TJI) operation to account for their different combustion characteristics. Preliminary 3D CFD combustion simulations were carried out to compare the prechamber concept with the standard spark plug configuration. The initial pre-chamber design was developed as a compromise between full-load and part-load requirements. In particular, the pre-chamber nozzle area was intentionally undersized with respect to the optimal full-load solution in order to achieve a more balanced performance across the operating range. The detailed results are presented in the chapter dedicated to the 3D CFD analyses. Overall, similar combustion performance was observed among the configurations investigated. However, at the LET operating point, the pre-chamber system showed a clear benefit in terms of knock mitigation compared to standard SI, thanks to the favorable effect of the turbulent jets generated by the pre-chamber. The simulations predicted an overall knock tolerance improvement of approximately 2 to 3 crank-angle degrees, combined with the reduced cycle-to-cycle variability that is characteristic of prechamber ignition systems. A full factorial analysis varying the compression ratio and the intake cam profile of the system was carried out under the following hypotheses: • Intake Port: IP4785 • Exhaust Port: Baseline Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 93 <?page no="94"?> • Intake Cam: DoE Parameter (scaled from standard cam keeping the accelerations constant as figure below) Figure 7: Intake cam profile scaling • Exhaust Cam: Baseline • CR: DoE Parameter • Lambda free • Combustion duration and knock model adapted from reference 3D simulations with TJI Figure 8: DOE table for 1D simulations The results of the DOE at the full-load operating point showed that: • Below an intake valve lift scaling factor of 0.9, the target torque cannot be achieved. • Reducing the intake valve lift scaling factor allows a reduction in CA50, partially offsetting the effect of the increased compression ratio. • Operation at λ ≈ 1 can only be achieved through the combined application of a high compression ratio and a Miller-cycle intake camshaft. • Increasing the degree of Millerization results in higher boost pressure demand and increased pre-turbine exhaust backpressure. • The adoption of a Miller-cycle strategy tends to shift the exhaust temperature limitation from the catalyst to the exhaust manifold upstream of the turbine, due to the higher expansion ratio across the turbine. 94 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="95"?> Figure 9: 1D DOE simulation results in full power point Following the DOE at the full-load operating point, the configuration featuring an intake valve lift scaling factor of 0.92 (corresponding to a high level of Millerization) and a compression ratio of 11.5 was selected for further investigation. This configuration was then evaluated through 1D simulations over the entire engine operating range and compared against the baseline engine. To fully exploit the benefits of the new combustion system, the valve timing strategy was revised, requiring a recalibration of the VVT settings, particularly with respect to Intake Valve Opening (IVO) timing. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 95 <?page no="96"?> Figure 10: 1D simulation resusts on the torque curve The results over the full torque curve can be summarized as follows: • Volumetric efficiency is reduced because of the increased Miller effect. • A slight improvement in CA50 is observed, benefiting from the enhanced combustion characteristics associated with the Miller-cycle strategy. • The engine can operate at leaner air-fuel ratios, with enrichment required only at high engine speeds. An irregular trend becomes noticeable above 8000 rpm. • Residual gas scavenging is improved, mainly due to the higher compression ratio and the increased boost demand. However, these results are strongly influenced by turbocharger and charge-air cooling limitations and therefore require validation on the full multi-cylinder engine. Overall, the 1D simulations led to the definition of a configuration capable of delivering the same torque curve as the baseline SI engine while achieving a significant extension of the λ = 1 operating range. Anyway, potential weak points of the selected best configuration are: • Assumptions adopted in the 1D TJI combustion model. • Potential impact of the higher compression ratio on LSPI tendency. • Effects of the Miller camshaft and IP4785 intake ports on mixture preparation and on combustion phasing and burn duration. • Increased heat rejection requirements, mostly for the intercooler system. 96 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="97"?> 4 CHT cooling simulations As is well known, thermal management of the pre-chamber is a key aspect of this application. On one hand, excessive pre-chamber temperatures can promote pre-ignition and lead to uncontrolled combustion events. On the other hand, an uneven heat transfer from the pre-chamber to the cylinder head may generate localized overheating of the aluminum structure, potentially resulting in crack initiation and component failure. As previously described, two pre-chamber configurations were selected and investigated through Conjugate Heat Transfer (CHT) simulations to assess the thermal behavior of the system and the resulting heat flux distribution within the cylinder head: • Dry pre-chamber, in which the insert is in contact only with the aluminum alloy of the cylinder head. • Wet pre-chamber, in which the insert is in direct contact with both the cylinder head material and the engine coolant. Figure 11: Dry and wet pre-chamber for CHT simulations The thermodynamic configurations considered in the CHT analyses also differed significantly. For the dry pre-chamber configuration, the boundary conditions derived from the 1D simulations were based on the standard BST01 intake port, production camshaft, 9.5: 1 compression ratio piston, a four-hole pre-chamber design, and rich-operation conditions. Conversely, the wet pre-chamber configuration adopted the high-flow IP4785 intake port, a Miller-cycle camshaft, a 11.5: 1 compression ratio piston, and λ = 1 operation, combined with a slightly revised four-hole pre-chamber geometry. To enable a meaningful comparison, both engine configurations were evaluated at the same performance level in terms of IMEP. Differences were also introduced in the water-jacket design to accommodate the direct cooling of the insert and the revised intake port geometry. However, the overall coolant flow layout and the coolant velocity distribution along the walls remained largely comparable between the two configurations. Herein, the results in full power conditions are presented. The wet prechamber shows a more energetic combustion rate than the dry one, due to the more favorable thermodynamic conditions (lower EGR and higher mean turbulence). In general, the heat flux generated in the “wet” system is higher, due to the higher prechamber combustion rate. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 97 <?page no="98"?> Figure 12: Pre-chamber and main chamber pressures and heat release, CHT simulations The wet pre-chamber configuration reduces the conductive heat flux from the insert to the cylinder head in the combustion chamber dome region, leading to a significant decrease in local aluminum temperatures. Figure 13: CHT simulations, head dome temperature In the wet TJI configuration, the improved cooling effectiveness of the pre-chamber significantly reduces the maximum aluminum temperature in the combustion chamber, with a peak temperature reduction of approximately 32 °C. Figure 14: CHT simulations, pre-chamber temperature The direct cooling of the pre-chamber insert reduces its maximum temperature by approximately 105 °C, despite the 38 % increase in pre-chamber wall heat transfer associated with the higher combustion rate. Spark plug temperatures should be regarded as indicative only, as the geometry was simplified and the model does not accurately 98 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="99"?> represent the detailed internal spark plug design. Nevertheless, the wet configuration exhibited a reduction in maximum spark plug temperature of approximately 100 °C, despite the higher pre-chamber combustion intensity. This result can be attributed to the improved cooling effectiveness of the pre-chamber insert and to the superior thermal characteristics of the adopted electrode design. 5 3D cfd combustion simulations for pre-chamber selection For 3D CFD combustion simulations, some basic fluid-dynamic designs have been selected: • Intake duct, standard and new IP4785 with higher permeability and lower tumble Figure 15: standard and IP4785 intake ducts • Camshaft, standard and Miller cam Figure 16 standard and Miller cam • Piston, compression ratio 9.5 and 11.5 Figure 17: CR 9.5 (standard engine) Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 99 <?page no="100"?> Figure 18: CR 11.5 (prototype, for pre-chamber tests) 5.1 Initial simulations for 1D combustion model calibration To define the baseline pre-chamber geometry and quantify the combustion benefits relative to the conventional SI system, an initial CFD simulation campaign was carried out. Two combustion system configurations were investigated: BST01 PC, featuring the addition of the pre-chamber to the baseline BST01 SI architecture, and IP4785 PC, which combined the pre-chamber with the new high-flow intake port design. Figure 19: First pre-chamber geometry, Bst01 intake ducts, CR 9.5 The initial pre-chamber design was developed based on prior experience and the geometric constraints imposed by the baseline combustion system. The pre-chamber axis was aligned with that of the original spark plug, while six nozzles were equally distributed over 360°. The nozzle inclination was selected to maximize jet penetration into the combustion chamber while ensuring sufficient clearance with the piston at top dead center. The initial sizing of the pre-chamber volume and nozzle diameters was defined as a compromise between low-load and full-load operating requirements, resulting in a relatively small pre-chamber volume and nozzle area. The different combustion system configurations were evaluated under full power conditions using identical boundary conditions (9500 rpm, λ = 0.83) to isolate the effects of the pre-chamber architecture and intake port design. 100 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="101"?> Figure 20: Trapped mass and tumble, full power point As expected, the trapped mass and tumble are the same with SI and PC for bst01 intake duct, while IP4785 shows higher flow and lower tumble. Figure 21: Lambda field at 700 degCA The lambda field of BST 01 PC is almost the same of BST01 SI, while IP 4785 PC is slightly leaner than BST01 PC and BST01 Spark in the exhaust zone. Figure 22: Turbulence and EGR close to PC electrodes during compression stroke Despite the lower tumble level, the turbulence intensity within the pre-chamber is virtually unchanged between the BST01 PC and IP4785 PC configurations. During the compression stroke, the flow entering the pre-chamber promotes dilution of the residual gases trapped within it. As a result of the similar flow field established inside the pre- Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 101 <?page no="102"?> chamber, both configurations exhibit a comparable EGR distribution despite the differences in the intake port design. Figure 23: Turbulence and EGR fields at 700 degCA With TJI operation, advancing the spark timing (SA) worsens the conditions within the pre-chamber, leading to higher EGR levels and lower turbulence intensity. This effect may become particularly critical at low-load conditions, where the combination of low turbulence and high residual-gas dilution can weaken the ignition capability of the pre-chamber. In contrast, the conventional spark-ignition configuration exhibits the opposite trend. As spark timing is retarded, the 0-10 % mass fraction burned (MFB) duration increases. This behavior is primarily associated with the lower turbulence intensity in the vicinity of the spark plug electrodes, resulting in a slower initial flame development. 102 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="103"?> Figure 24: Mass fraction burned durations as a function of MFB50 The TJI configurations (BST01-PC and IP4785-PC) exhibit a shorter 10-50 % MFB combustion duration than the conventional spark-ignition configuration (BST01 Spark), owing to the contribution of the high-energy jets discharged from the pre-chamber. Conversely, the 50-90 % MFB combustion duration is mainly governed by the mixture distribution and turbulence levels within the main combustion chamber. In this phase, the IP4785-PC configuration shows a longer burn duration, primarily because the flame front encounters locally richer mixture regions, resulting in a reduction of the laminar flame speed. Finally, a knock analysis was performed by evaluating the maximum amplitude of pressure oscillations (MAPO) as a function of MFB50. The results indicate a similar knock resistance for all three configurations. Combined with the comparable relationship between IMEP and MFB50, this suggests that the three combustion systems are capable of achieving similar performance at this operating point. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 103 <?page no="104"?> Figure 25: IMEP and MAPO as a function of MFB50 Similar simulations were performed at the LET operating point (4000 rpm, λ = 0.88). As expected, the effect of intake port permeability on trapped mass became less significant under these conditions: the higher-flow IP4785 intake port still provided an increase in trapped air mass of approximately 1.3 % compared to the BST01 configuration, while maintaining a substantially lower tumble level. Figure 26: Trapped mass and tumble, LET point Even at partial loads, the installation of the pre-chamber doesn’t affect the mixture field inside the main chamber, while the IP4785 PC system shows a rich lambda zone under the intake squish. Figure 27: Lambda field at 712 degCA 104 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="105"?> Looking at the EGR concentration at the electrodes, the LET point shows much better velues compared to the full power point: besides, due to the lower engine speed, the spark occurs later with further reduction of EGR. Figure 28: EGR at the electrodes and EGR fields at 712 degCA In the end, in this engine point showed a fast first combustion stage, thanks to a prechamber dimensioning more balanced towards the low speed/ low load application. The brilliant start of combustion is affected on IP4785 by a worse mixture field that slow down the 50-90 combustion stage. Figure 29: Mass fraction burned durations as a function of MFB50 Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 105 <?page no="106"?> Overall, the combustion performance was similar across all the configurations investigated. However, the pre-chamber systems exhibited improved knock behavior due to the beneficial effect of the high-energy jets discharged from the pre-chamber. Figure 30: IMEP and MAPO as a function of MFB50 The resulting flame propagation pattern enabled a more effective combustion of the mixture in the exhaust-side squish region, preventing the formation of the critical end-gas conditions that typically promote knock onset. The pre-chamber system provided an increase in knock tolerance equivalent to approximately 2-3 crank-angle degrees. This benefit is further complemented by the lower cycle-to-cycle variability associated with TJI combustion systems. 5.2 Bst01 configurations, CR and PCs For the dry pre-chamber configuration, the maximum allowable internal chamber diameter was limited to 9 mm due to packaging constraints. Within this dimensional restriction, several pre-chamber geometries were evaluated and optimized through CFD simulations to achieve satisfactory performance under full-load conditions. Subsequently, a dedicated study was conducted to assess the effects of compression ratio, comparing CR = 9.5 and CR = 11.5 configurations while retaining the standard camshaft profile. The prechamber design selected for this investigation, designated PCL5, featured a four-hole symmetric nozzle layout and was considered the most suitable compromise among the configurations analyzed. Figure 31: tested configuration for CR evaluation 106 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="107"?> Figure 32: Motion and turbulence results The introduction of pre-chamber does not significantly affect the tumble motion respect to the Spark configurations, while the increase of CR leads to a reduction of tumble and Omega tumble. The Spark and TJI configurations with CR=9.5 and CR=11.5 show a similar mean turbulence inside the cylinder. The configuration with CR=11.5 shows a lower mean turbulence due to a lower tumble motion during the compression stroke. Figure 33: lambda field at 700 degCA The pre-chamber system has a negligible effect on the λ distribution compared to the corresponding Spark configurations, for both CR = 9.5 and CR = 11.5. The higher compression ratio slightly alters mixture homogeneity due to the reduced tumble motion. As a result, the CR = 9.5 configuration exhibits a slightly richer mixture within the prechamber than the CR = 11.5 configuration. Figure 34: EGR at the pre-chamber electrodes Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 107 <?page no="108"?> Mixture preparation within the pre-chamber is strongly influenced by the residual gas concentration. Increasing the compression ratio significantly reduces the EGR level, owing to improved pre-chamber scavenging. In addition, the required spark timing is more retarded, further improving the mixture conditions at ignition. Figure 35: Turbulence vs EGR and EGR fields at 700 degCA Additional pre-chamber designs were investigated at CR = 11.5. While maintaining a similar pre-chamber volume, the number of nozzles was increased from 4 to 6, 8 and 10, keeping similar total nozzle flow area, symmetrical layout, and nozzle inclination. The simulations showed noticeable differences in mixture distribution, flow velocity, and turbulence structures within the pre-chamber. However, the mean values of the key parameters in the spark plug vicinity at 700 degCA were found to be broadly comparable across the configurations analyzed. Figure 36: Turbulence and EGR fields at 700 degCA 108 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="109"?> Figure 37: Turbulence vs EGR and Velocity at the electrodes at 700 degCA Combustion simulations showed very similar results across all the pre-chamber designs investigated. Only marginal improvements in knock resistance were observed for the 6hole and 8-hole configurations, while the overall combustion performance remained largely unchanged. Figure 38: IMEP and MAPO as a function of MFB50 Experimental testing of the various pre-chamber geometries in the BST01 configuration at CR = 11.5 confirmed the limited sensitivity of combustion performance to the pre-chamber design, with only minor differences observed among the configurations evaluated. 5.3 IP4785 configuration with Miller cam For the IP4785 intake port configuration combined with the Miller-cycle camshaft, the results presented in this paper focus on the LET operating point (4000 rpm, WOT). The pre-chamber design adopted for the IP4785 configuration was revised with respect to the BST01 version. The increased packaging space available in the cylinder head enabled the adoption of a larger pre-chamber internal diameter. In addition, the prechamber was positioned closer to the combustion chamber, and its volume was slightly reduced. Both configurations employed the same four-hole nozzle layout, with identical nozzle diameters. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 109 <?page no="110"?> Figure 39: In-cylinder EGR and temperature For the same intake valve lift profile (AM092), the IP4785 CR 11.5 PCL5 S5H4R LET configuration achieves lower residual gas dilution by adopting a delayed EVC. The resulting reduction in EGR leads to a decrease in the in-cylinder temperature at TDC of approximately 10 K compared to the BST01 CR 11.5 PCL5 AM092 LET configuration. Figure 40: Turbulence and EGR at the electrodes Figure 41: Turbulence and EGR fields at 700 degCA, BST01 (left) and IP4785 (right) 110 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="111"?> The conditions at the spark plug electrodes, in terms of EGR concentration, flow velocity, and turbulence intensity, are largely comparable between the two configurations. Accordingly, the combustion metrics indicate a similar combustion rate and overall burn duration. Figure 42: IMEP and MAPO as a function of MFB50 However, the lower in-cylinder temperature at TDC achieved by the new configuration results in a significant improvement in knock resistance, demonstrating the benefit of the reduced residual gas content and enhanced charge preparation. 6 Single cylinder engine testing Following the simulation campaign, two main SCE hardware configurations were selected: • Baseline cylinder head casting with limited machining modifications, dry prechamber, CR = 9.5 and 11.5, and standard camshaft. This configuration was expected to provide an initial extension of the λ = 1 operating range, while potentially suffering from pre-ignition issues and localized cylinder head overheating in the spark plug region. • New cylinder head casting featuring revised high-flow intake ports, wet prechamber, CR = 11.5, and an aggressive Miller-cycle camshaft. This configuration was expected to maximize the extension of the λ = 1 operating range while eliminating thermal criticalities in the pre-chamber area through direct insert cooling. To reproduce the exhaust backpressure characteristics of the multi-cylinder engine on the SCE test bench, dedicated exhaust pressure maps were implemented as a function of exhaust mass flow rate. Different calibrations were required for the two configurations due to the differences in intake port geometry and camshaft profiles. Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 111 <?page no="112"?> Figure 43: Exhaust pressure drop as a function of exhaust flow rate Despite the higher permeability of the IP4785 intake ports, the configuration equipped with the aggressive Miller camshaft required a higher exhaust backpressure to process the same air mass flow, owing to its altered gas-exchange characteristics. 6.1 Knock evaluation In pre-chamber ignition systems, knock assessment is complicated by the occurrence of Jet Knock, a phenomenon associated with the interaction between high-energy reacting jets issued from the pre-chamber and the unburned mixture in the main chamber. Unlike conventional knock, which originates from end-gas autoignition, jet knock is primarily related to localized rapid heat release and pressure wave generation induced by turbulent jet penetration. As a result, high-frequency pressure oscillations may be detected even in the absence of classical end-gas autoignition, leading to potential misinterpretation of conventional knock metrics. Consequently, the conventional pressure-based indicators such as the Maximum Amplitude of Pressure Oscillations (MAPO) and the Knock Peak index (KP_PK) may overestimate knock intensity when directly applied to pre-chamber combustion systems. Theoretically an accurate knock characterization therefore requires distinguishing jet-induced pressure oscillations from genuine end-gas autoignition through spectral analysis, spatial localization of the pressure source, and detailed cycle-resolved combustion diagnostics. This phenomenon was identified during the first experimental campaign performed on the pre-chamber single cylinder engine. Therefore, a dedicated knock analysis methodology was developed to accelerate test-bed activities while enabling operation closer to the actual combustion knock limit. Detailed analysis revealed that the highly accelerated combustion process generates multiple pressure oscillation modes during flame propagation, many of which are not directly related to knock onset and can therefore lead to falsepositive knock detections. To address this issue, two methodologies available in the literature were combined: • A statistical approach based on the methodology proposed in paper [3]; • A multi-frequency knock parameter approach derived from paper [4] and further developed in-house. 112 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="113"?> Based on these concepts, an innovative knock analysis methodology was implemented. A novel MAPO-based index was defined by numerically processing the three most relevant pressure oscillation frequencies associated with the combustion chamber acoustic response, corresponding to higher-order multiples of the first resonant acoustic mode. The contribution of each frequency component was subsequently combined through a statistical weighting procedure, providing a more robust discrimination between combustion-induced pressure oscillations and actual knock events. Figure 44: Traditional MAPO evaluation vs TJI effects Figure 45: Example of knock evaluation via statistical approach @4000 rpm Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 113 <?page no="114"?> An example at 4000 rpm is reported in the figures above. Prior to knock evaluation, the incylinder pressure signal was processed and angularly windowed to isolate the combustion phase potentially affected by end-gas autoignition while minimizing the contribution of pressure oscillations generated by the reactive jets emerging from the pre-chamber. This preprocessing step enables a partial separation between jet-knock-related oscillations and those associated with conventional knock. The statistical methodology then identifies the onset of knock through a deviation from the log-normal distribution observed under normal combustion conditions. In addition, the multi-frequency knock parameter divides the frequency-domain analysis into two distinct combustion phases, allowing pressure oscillations generated by the pre-chamber jet combustion process to be distinguished from those originating from knock-induced pressure waves. 6.2 Spark plug selection and low load results Spark plug selection was carried out by testing different M10 spark plug designs on the dry pre-chamber configuration equipped with the standard intake ports and the baseline piston (CR = 9.5). Four spark plug variants were evaluated: a bi-electrode design with relatively high protrusion, a semi-surface-discharge racing spark plug, and two slant-electrode racing spark plugs. All spark plugs featured an NGK heat range of 10, except for one slantelectrode variant with heat range 8, corresponding to a lower heat dissipation capability and therefore a higher operating temperature. Figure 46: Tested spark plugs To ensure that the Start of Injection (SOI) strategy did not influence the assessment of spark plug performance, SOI sweeps were performed during the evaluation. The selected spark plug exhibited the best compromise between combustion stability, knock tolerance, and pre-ignition resistance. 114 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="115"?> Figure 47: Selected points for spark plug selection and low load evaluation Figure 48: Spark plug selection and low load evaluation results with BST01 and CR 9.5 The single-electrode spark plug with NGK heat range 10 was selected for all subsequent testing. After increasing the compression ratio to 11.5, all operating points could be successfully achieved without any issues, confirming the simulation results, which indicated improved pre-chamber scavenging and a consequent reduction in residual gas concentration within the pre-chamber. Nevertheless, the dual-spark-plug concept remains a potential solution to further enhance combustion stability and robustness under low-speed, low-load operating conditions, where ignition requirements are more demanding for passive pre-chamber systems. 6.3 WOT results On the single-cylinder engine (SCE), full-load operation was defined as the corresponding load level obtained by scaling the full-load power curve of the production V8 L411 engine in its SOP configuration (800 CV). Full-load performance was assessed by comparing the standard engine configuration against the IP4785 wet pre-chamber architecture operating with an increased compression ratio (CR = 11.5). Prior to discussing the results, several considerations regarding exhaust Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 115 <?page no="116"?> thermal constraints and the interpretation of single-cylinder engine (SCE) measurements must be addressed: • In turbocharged spark-ignition engines, the maximum achievable enleanment and combustion phasing at wide-open throttle (WOT) are often constrained by turbine inlet temperature (T3), catalyst inlet temperature (T4), or catalyst brick temperature, as excessive thermal loading may compromise turbine durability, catalyst protection, and long-term component reliability. Furthermore, the relationship between T3 and T4 strongly depends on the expansion process occurring across the turbine. Consequently, the active temperature limitation at engine level may vary depending on the operating point and on the corresponding turbine pressure ratio. • Additional precautions are required when interpreting SCE results: while the singlecylinder setup allows accurate measurement of the exhaust temperature upstream of the turbine, it cannot fully reproduce the gas-dynamic interactions occurring in a multicylinder engine, including cylinder-to-cylinder pulse interference, manifold wave dynamics, and the actual turbine operating conditions. • In the present work, the exhaust backpressure was therefore imposed based on the results of a calibrated 1D engine model, providing representative boundary conditions for the investigated operating points. Owing to these limitations, the application of a common exhaust temperature limit to both SCE configurations would not provide a representative comparison of their actual fullengine potential. Consequently, a different calibration strategy was adopted. For the standard configuration, the calibration derived from the multicylinder engine and corresponding to the exhaust temperature limit was directly transferred to the SCE. For the IP4785 wet prechamber configuration with CR 11.5 and Miller cam, the engine was operated at λ = 1 while targeting the same IMEP level, without applying the multicylinder exhaust temperature constraint. Instead, only a maximum exhaust temperature threshold was imposed to guarantee safe operation and avoid exhaust valve overheating. This methodology was selected to highlight the intrinsic combustion benefits of the prechamber concept. By promoting faster combustion, improved burn rates, and enhanced knock resistance, pre-chamber ignition systems can potentially enable operation at higher compression ratios and more advanced combustion phasing, thereby increasing engine efficiency and extending the full-load operating range. Following the calibration philosophy described above, the full-load results obtained for the two configurations are presented and discussed in the following. 116 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="117"?> Figure 49: Experimental results on SCE The results show that the exhaust temperature of the IP4785 pre-chamber configuration remains substantially aligned with that of the baseline system over most of the operating range, while a noticeable increase is observed at the highest engine speeds, where the 1D engine model predicts the onset of mixture enrichment requirements. This behavior suggests that the improved combustion efficiency provided by the pre-chamber system can partially offset the thermal penalties typically associated with high-load operation. The analysis also highlights a significant increase in intake manifold pressure, indicating higher airflow demand and engine load capability. However, the resulting increase in charge density and combustion intensity leads to higher heat rejection to the cooling system, which may introduce additional constraints on the thermal management system and, consequently, on the overall vehicle integration. Furthermore, both peak in-cylinder pressure and exhaust backpressure are increased compared to the baseline configuration. These effects are expected because of the higher compression ratio and enhanced combustion process enabled by the pre-chamber architecture. As a result, dedicated mechanical design measures may be required to withstand the increased thermo-mechanical loads acting on critical engine components, including Evaluation of a Passive Pre-Chamber System for a High-Performance Turbocharged Engine 117 <?page no="118"?> the piston assembly, cylinder head, connecting rods, exhaust valves, and turbocharging system. Although the single-cylinder engine results provide a clear indication of the potential benefits of the pre-chamber concept, final validation must be performed on the complete multicylinder engine. Only full-engine testing can properly account for the interaction between combustion, turbocharger matching, exhaust thermal management, cooling system capacity, and vehicle-level operating constraints. Summary This paper presents an investigation of a passive pre-chamber system based on the Lamborghini L411 high-performance turbocharged engine. The study aims to assess the potential benefits of this technology for road-legal applications. The design activity led to the evaluation of two potential concepts, dry and wet prechamber configurations, for installation on a single-cylinder engine (SCE). An extensive 1D simulation campaign was carried out to identify a complete engine configuration capable of extending the λ = 1 operating range. This was achieved through the adoption of a Miller-cycle camshaft, an increased compression ratio, and higherpermeability intake ports. The wet and dry concepts were subsequently analyzed through Conjugate Heat Transfer (CHT) simulations. The results highlighted potential criticalities of the dry concept, as local overheating could compromise system operation and lead to component failures or pre-ignition events. In contrast, the wet design demonstrated a much more favorable temperature distribution. Both pre-chamber configurations were then investigated using 3D CFD combustion simulations, which demonstrated the potential benefits of the Miller-cycle engine equipped with a pre-chamber combustion system. Finally, experimental testing on the SCE enabled a more accurate assessment of knock behavior with the pre-chamber system, allowing for proper calibration and indicating promising improvements in terms of λ = 1 operating range extension. Nevertheless, validation on the full multi-cylinder engine is still required, and the system design will need further refinement before being considered suitable for production. References [1] Bizzarri D., Rizzi L., Bonandrini G., Kapus P., Hochfellner D., Lamborghini’s New L411, Development of a Biturbo V8 Hybrid to Generate Pure NA Feeling, 46th International Vienna Motor Symposium, 2025 [2] Forte C., Siliato A., Bizzarri D., Bonandrini G., AI-Assisted CFD Optimization of High-Tumble Intake Ports for a High-Performance V8 Hybrid Engine, International ICE Forward Symposium (ICEF), 2026 [3] Corrigan, D., Di Blasio, G., Ianniello, R., Silvestri, N. et al., “Engine Knock Detection Methods for Spark Ignition and Prechamber Combustion Systems in a High-Performance Gasoline Direct Injection Engine,” SAE Int. J. Engines 15(6): 883-897, 2022 [4] Palakunnummal, M., Priyadarshi, S., Ellis, M., and Nazha, M., “A Cylinder Pressure-Based Knock Detection Method for Pre-chamber Ignition Gasoline Engine,” SAE Int. J. Engines 14(3): 405-417, 2021 118 Giovanni Bonandrini, Antonio Aliperti, Davide Bizzarri, Paolo Scarpato <?page no="119"?> The Effect of Fuels with Higher Alcohol Content on the Ignition System and Combustion Anomalies in Gasoline Engines Max Schüttenhelm, Karsten Michels, Thomas Garbe Abstract: Disclaimer: The content of this paper does not permit any inference regarding future business activities, decisions, or corporate strategies. The decarbonization of road transport requires not only electrification but also the use of renewable fuels. In particular, ethanol and methanol offer the potential to reduce CO₂ emissions from existing and future spark-ignition vehicles while continuing to utilize the existing fuel infrastructure. This study investigates the influence of increased alcohol content in gasoline fuels on combustion, ignition systems, and combustion anomalies. Alcohol-based fuels are characterized by high octane ratings and high enthalpy of vaporization. Despite their lower heating value, they provide significant charge cooling, which substantially reduces knock tendency. This effect goes beyond the pure octane number benefit and enables engine calibration optimized for combustion efficiency. The investigations were performed on a modified 1.5-l TSI engine with an increased compression ratio. The results show that increasing ethanol or methanol content allows earlier combustion phasing, resulting in improved thermal efficiency. Compared with ETBE blends, alcohol fuels achieve superior performance at similar octane ratings because they additionally benefit from pronounced evaporative cooling. Furthermore, the behavior regarding low-speed pre-ignition (LSPI) was investigated. Higher alcohol contents significantly increase the LSPI limit, particularly in directinjection engines, where the cooling effect of the fuel is fully effective within the combustion chamber. As a result, the tendency toward pre-ignition is reduced. A comparison between a conventional spark plug and a passive pre-chamber spark plug demonstrates that high-octane alcohol fuels enable efficiency-optimal combustion phasing across the entire load range. This leads to higher thermal efficiency, lower HC and CO emissions, and improved combustion stability. Kurzfassung: Die Dekarbonisierung des Straßenverkehrs erfordert neben der Elektrifizierung auch den Einsatz erneuerbarer Kraftstoffe. Insbesondere Ethanol und Methanol bieten das Potenzial, die CO₂-Emissionen bestehender und zukünftiger Ottomotorfahrzeuge zu reduzieren und gleichzeitig die vorhandene Kraftstoffinfrastruktur weiter zu nutzen. In dieser Arbeit wird der Einfluss erhöhter Alkohol- und Etheranteile im Ottokraftstoff auf Verbrennung, Zündsystem und Verbrennungsanomalien untersucht. Alkoholkraftstoffe <?page no="120"?> zeichnen sich durch hohe Oktanzahlen sowie eine hohe Verdampfungsenthalpie aus. Trotz ihres geringeren Heizwerts bewirken sie eine starke Kühlung der Frischladung, wodurch die Klopfneigung deutlich reduziert wird. Dieser Effekt geht über den reinen Oktanzahlvorteil der Ether hinaus und ermöglicht eine verbrennungsoptimierte Motorabstimmung. Die Untersuchungen wurden an einem modifizierten 1,5-l-TSI-Motor mit erhöhtem Verdichtungsverhältnis durchgeführt. Die Ergebnisse zeigen, dass mit steigendem Ethanol- oder Methanolanteil frühere und damit wirkungsgradgünstigere Verbrennungsschwerpunktlagen erreicht werden können. Im Vergleich zu ETBE-Blends erzielen Alkoholkraftstoffe bei ähnlicher Oktanzahl bessere Ergebnisse, da sie zusätzlich von einer ausgeprägten Verdampfungskühlung profitieren. Darüber hinaus wurde das Verhalten hinsichtlich Low-Speed-Pre-Ignition (LSPI) untersucht. Höhere Alkoholanteile erhöhen das Ladedruckniveau für den LSPI-freien Betrieb deutlich, insbesondere bei Direkteinspritzung, bei der der Kühlungseffekt der Kraftstoffe vollständig im Brennraum wirksam wird. Dadurch sinkt die Neigung zu Vorentflammungen. Der Vergleich einer konventionellen Dachelektroden-Zündkerze mit einer passiven Vorkammerzündkerze zeigt, dass hochoktanige Alkoholkraftstoffe über den gesamten Lastbereich wirkungsgradoptimale Verbrennungsschwerpunktlagen ermöglichen. Dies führt zu höheren Wirkungsgraden, geringeren HC- und CO-Emissionen sowie zu einer verbesserten Verbrennungsstabilität. Insgesamt verdeutlichen die Ergebnisse, dass Kraftstoffe mit erhöhten Alkoholanteilen gegenüber herkömmlichen Kraftstoffen erhebliche Potenziale zur CO₂-Minderung, Wirkungsgradsteigerung und Reduzierung von Verbrennungsanomalien besitzen. Die Vorteile beruhen nicht nur auf der hohen Klopffestigkeit, sondern insbesondere auf der starken inneren Kühlung der Frischladung, wodurch sich neue Freiheitsgrade für die zukünftige Auslegung effizienter Ottomotoren eröffnen. 1 Introduction The recognition that CO 2 emissions from the combustion of fossil energy carriers influence the climate and make a significant contribution to anthropogenic climate change has led to extensive regulatory measures in the European Union, including the so-called Green Deal. As part of this Green Deal, the existing Renewable Energy Directive (EU) 2018/ 2001, also known as RED II, was further developed into Directive (EU) 2023/ 2413, now referred to as RED III. For the transport sector in 2030, it provides for either a 29 % share of renewable energy or a 14.5 % reduction in greenhouse gas intensity [1.1]. The German Association of the Automotive Industry (VDA) calls for a continuation of this framework through to 2050, with intermediate targets to safeguard investments in renewable energies of 90 % by 2040 and 100 % by 2050. Against this background, and in the context of the planned EU regulation to reduce CO 2 emissions in the transport sector [1.2], renewable liquid fuels, particularly biofuels and e-fuels, are being discussed as a complementary approach to the defossilization of the existing vehicle fleet and newly registered vehicles [1.3]. A relevant question in this context is the extent to which such 120 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="121"?> fuels can be taken into account in CO 2 fleet legislation for passenger cars and light commercial vehicles. Figure 1.1 shows three scenarios up to the year 2050 [1.4]. These scenarios represent the exclusive use of fossil energy in the transport sector, a strategy focused solely on electric mobility whose effect is essentially linked to fleet renewal, and a combination of electric mobility and renewable fuels. The key potential of the combined strategy lies in the fact that low-CO 2 renewable fuels can also include the existing vehicle fleet in the reduction of transport-related CO 2 emissions. Figure 1.1: Cumulative Greenhouse Gas Emissions from Road Transport in the EU [1.4] Production capacities for renewable methanol are increasing worldwide. Methanol is already used today on a large scale as a basic chemical and is also gaining increasing importance as a fuel component and as a feedstock for synthetic gasoline fuels. In China’s 15th Five-Year Plan, methanol is addressed as one element of the energy supply for the transport sector. Figure 1.2 shows the locations of methanol production plants that are already in operation, under construction, or in planning. Particularly striking are the large number and scale of the plants located in China. Figure 1.2: Biomethanol and E-Methanol Plants in operation and under construction [1.5] The Effect of Fuels with Higher Alcohol Content on the Ignition System 121 <?page no="122"?> Compared with renewable methanol, bioethanol is already available today in significantly larger quantities and is established in the fuel market. Global ethanol production amounted to approximately 116 billion liters in 2023 and is largely dominated by the United States and Brazil, which together account for around 80 % of global production [1.6]. Due to existing production, logistics, and blending structures, as well as the broad market introduction of E5 and E10 fuels, bioethanol can contribute comparatively quickly in the short term to reducing the fossil carbon intensity of gasoline fuel. At the same time, the long-term available volume potential is limited by the sustainably available biomass. Bioethanol should therefore be regarded as a short-term, readily compatible, and already introduced option, although its ultimate availability remains limited by biomass potential. Figure 1.3: Bioethanol production of the five largest producers up to 2023 [1.6] From the perspective of a vehicle manufacturer with a large existing fleet of spark-ignition engines, not only the pure availability of individual renewable fuel pathways is decisive, but also their technical compatibility with existing engine concepts, fuel specifications, and ignition systems. Against this background, the investigation of higher alcohol contents, particularly above the level of today’s E10 fuels, is relevant in order to assess the CO 2 reduction potential of renewable fuel components with an acceptable level of modification effort in the existing vehicle fleet. 2 Fuel Production The basic concept behind the production of biofuels and e-fuels is a closed carbon cycle. In the case of biofuels, plants absorb carbon from the air in the form of CO 2 . During the production process, sugars or lignocellulose processed into sugars are fermented, producing ethanol and carbon dioxide as a by-product. The ethanol can be further processed into EtG (ethanol to gasoline). On an industrial scale, methanol is produced from synthesis gas consisting of CO and hydrogen. In conventional methanol synthesis, coal and natural gas are the sources of the synthesis gases. In modern processes with a closed carbon cycle, the synthesis gases can be generated from green hydrogen produced by electrolysis and CO 2 . The latter is 122 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="123"?> provided by removing CO 2 from ambient air or from combustion processes. In these production pathways, methanol is not of biogenic origin but is an e-fuel, which also applies to the downstream product MtG. Further options include the production of synthesis gas from biomass by thermochemical gasification (pyrolysis), the recovery of CO 2 as a by-product of biogas and sewage gas production, and from the fermentation of residual biomass or the fermentation of sugars to ethanol. For methanol synthesis, the addition of renewable hydrogen is still required. This results in biomethanol and subsequently MtG (methanol to gasoline). While biofuels are limited by the available biomass, the volume potential of e-fuels during the ramp-up of the technology is limited by the availability of renewable electricity, electrolysis capacities, and sufficient availability of CO 2 . In theory, the long-term potential for e-fuels is almost unlimited. Figure 2.1: Production Cycle for Biofuels and E-Fuels EtG and MtG are produced from the corresponding alcohols by catalytic processes and are drop-in capable, i. e., they can be used in the existing vehicle fleet and comply with EN228. As a rule, the desired octane rating of RON 95/ 98 is achieved by adding ethanol within the limits of EN228. Since methanol is permitted in EN228 only up to 3 %, methanol currently plays only a subordinate role in this context. In the future, the catalytic conversion step from alcohols to gasoline could largely be omitted, thereby reducing costs and increasing the alcohol content to approximately 50 % or more. This could result in a cost-effective, high-octane fuel and create new degrees of freedom in engine development. The fuels MtG and EtG as well as the alcohols methanol and ethanol are liquid under normal conditions and can therefore be stored very effectively and transported to filling stations via the existing infrastructure. Volkswagen has approved its gasoline-engine The Effect of Fuels with Higher Alcohol Content on the Ignition System 123 <?page no="124"?> vehicles for an ethanol content of 25 vol.% since 2015 and is therefore very well prepared for the market introduction of E20. 3 Fuel Properties The properties of commercially available fuels are determined by the properties and blending behavior of an extremely large number of components. When individual components are added in large quantities, the properties of the blend are strongly influenced by these components. Table 3.1 shows important properties of typical blending components compared with the average values of gasoline without oxygen-containing components. Particularly noticeable are the extremely high octane numbers of these components. Since octane numbers significantly above 100 are very difficult to determine, many different values exist in the literature; these values should therefore be regarded primarily as a rough indication. The alcohols have significantly minor lower heating values, especially methanol. The main reason for this is the oxygen contained in the molecule, which accounts for approximately half of the molar mass in the case of methanol. Compared with gasoline, the boiling temperatures of these substances are in the lower range. The low vapor pressure is not beneficial for cold starting in winter, particularly in combination with the high enthalpy of vaporization of the alcohols. On the other hand, the high enthalpy of vaporization of the alcohols is very effective in reducing knock tendency in an engine at operating temperature. Table 3.1: Properties of blending components compared with gasoline The following figure shows the relative differences of the alcohols and ETBE compared with pure gasoline. On the one hand, the different lower heating values result in higher required injection quantities for the same energy input. On the other hand, the different enthalpy of vaporization of the blending components, in combination with the correspondingly higher required fuel quantities, lead to very different cooling effects. For example, ethanol has slightly more than twice the enthalpy of vaporization of gasoline; however, as a neat ethanol fuel, it achieves approximately 3.4 times the cooling effect due to the higher required fuel quantity. For the current market fuel E10 and the E20 fuel being prepared for market introduction, the required additional fuel quantity of approximately 3.5 % and 7 %, respectively, is significantly less relevant than the cooling effect of approximately 15 % and 32 %, respectively, compared with alcohol-free gasoline. This 124 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="125"?> effect is particularly evident in direct-injection engines, because the entire fuel quantity evaporates directly in the cylinder and therefore effectively cools the fresh charge. The octane number is determined in a standardized test engine equipped with a carburetor, which means that the predictive accuracy of this method for modern direct-injection engines is limited. For example, the cooling effect of vaporization on the gas temperature is not adequately represented. As a result, fuels with increasing alcohol content show an advantage in terms of knock tendency beyond what would be expected from the octane number alone. This advantage does not apply to fuels blended with ethers. Figure 3.1: Characteristic parameters of blends with methanol, ethanol and ETBE Engine operation with neat alcohol appears attractive because of the high octane number combined with the strong cooling effect. However, there are several reasons in favor of a blended fuel. The highly volatile components in gasoline can improve cold-start behavior, especially in winter, while the higher lower heating value of gasoline can increase vehicle range. In addition, multi-component fuels may be more robust with regard to ignition and evaporation behavior and combustion anomalies, because the individual components have different physicochemical properties and ignition delay times. Furthermore, from the perspective of fuel production, it is advantageous to be able to use different resource streams. For example, gasoline side streams from kerosene production using Fischer-Tropsch processes or MtG gasoline from MtG plants, which will be needed for the existing fleet at least in the near future, can be utilized. In addition to engine-related properties, the health-related aspects of the various components must also be analyzed and evaluated for the potential market introduction of new fuels. In general, fuels are by definition combustible substances and are therefore classified in safety data sheets as extremely flammable liquids and vapors. Gasoline is also classified as potentially causing skin irritation, genetic defects and cancer, as suspected of The Effect of Fuels with Higher Alcohol Content on the Ignition System 125 <?page no="126"?> damaging the unborn child and impairing fertility, and as toxic to aquatic organisms with long-lasting effects. While handling gasoline and its properties during refueling has been standard practice in everyday life for more than one hundred years, biofuels and e-fuels introduce new fuel components with new chemical, physical, safety-related and environmental properties at filling stations. In the event of skin contact with any fuels, whether conventional gasoline, biofuels or e-fuels, the recommended immediate measures are thorough washing, removal of contaminated clothing and, if necessary, consulting a physician. A stricter response is required for methanol, which is classified as toxic and can be absorbed not only by ingestion and inhalation but also through the skin. Particularly in the event of ingestion, a physician must always be contacted. For drivers and fillingstation customers, the framework conditions for handling methanol or methanol-containing fuels remain unclear. In contrast to conventional fuels, if larger quantities of methanol or ethanol enter the environment during transport, the alcohols are rapidly biodegraded. Due to their low viscosity and rapid dispersion, alcohols spread quickly, particularly in water, meaning that even tanker accidents involving methanol are not expected to cause major environmental damage [3.1]. Overall, crude oil and its refinery products are less favorable than methanol and ethanol with regard to environmental damage in the event of a transport accident at sea or on land. 4 Engine and Test Fuels The tests were carried out using a 1.5-l TSI engine from Volkswagen AG. For the investigations, the compression ratio was increased by installing different pistons, see table 4.1. In addition to the standard production J-gap spark plug, passive pre-chamber spark plugs (PC) were also investigated. The examined combustion phasing values were determined at a compression ratio of 14.5: 1 in combination with the respective fuel used. Due to fueland engine-dependent influences, the results shown cannot be transferred without restriction to other engine concepts or fuels and may differ significantly in such applications. For blends such as E20, it should be noted that the influence of the blending component increases with its concentration in the fuel. 126 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="127"?> Table 4.1: Engine Data The following table shows the octane numbers of the fuel blends used and their oxygen content to indicate the extent to which the respective blend currently complies with the applicable standards (E10 standard: max. 3.7 vol.% O 2 ; E20 standard: 8 vol.% O 2 ) or how far it deviates from them. Table 4.2: Fuel Data 5 Experimental Results Various tests were carried out with the fuels in order to investigate their influence at different operating points. The following figure shows the combustion phasing values using a J-gap spark plug for the operating point n = 2000 rpm and a brake mean effective pressure of 18 bar, plotted against the octane number of the fuels. A clear trend can be observed: with an increasing proportion of blending components, the octane number increases and the combustion phasing moves toward the optimum. However, at comparable octane numbers, ETBE-containing blends show markedly less favorable c. o. h. r. than blends containing alcohol fuels. The Effect of Fuels with Higher Alcohol Content on the Ignition System 127 <?page no="128"?> Figure 5.1: COHR-RON for different blends at n=2000 1/ min, BMEP=18 bar This is mainly due to the very low enthalpy of vaporization of ETBE compared with the alcohols and the associated lower internal charge cooling. In general terms, this indicates that the improvement in combustion phasing cannot be described solely by the octane number. Investigations of pre-ignition behavior (LSPI) were also carried out using a J-gap spark plug and various blends. The following table lists the calibration settings used for these tests. Table 5.1: Engine Parameter for LSPI Tests The fixed, extremely late combustion phasing provides the same time interval for prereactions leading to pre-ignition for every fuel blend, regardless of its octane number. Furthermore, due to the advanced downward movement of the piston at 40° after firing TDC, the cylinder volume is sufficiently large to avoid engine damage caused by excessive pressure peaks. This enables a large number of tests to be conducted under constant engine conditions with high reproducibility. The tests were carried out with direct injection, represented by the lighter bars, and in PFI operation, indicated by the respective darker bars. For every fuel blend, whether with ethanol, methanol or ETBE, the threshold for the first occurrence of pre-ignition is higher with direct injection than in PFI operation. While the enthalpy of vaporization can fully develop its cooling effect on the fresh charge in the cylinder with direct injection, in PFI mode part of the enthalpy of 128 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="129"?> vaporization is absorbed by the flow-guiding components between the injector and the cylinder. This effect is particularly pronounced at higher alcohol contents (E45 and M30). As described above in Section 3, this is attributable to the steadily increasing cooling effect resulting from the combination of enthalpy of vaporization and the higher fuel quantity required due to the lower heating value. In real engine operation, the strong cooling effect is accompanied by the effect of the high octane number and the resulting earlier spark timing, which limits the time available for pre-reactions leading to pre-ignition. Figure 5.2: Low Speed Pre-Ignition for different fuel blends at n=1500 1/ min The interaction between the different fuel blends and the ignition system must also be investigated for future applications. To limit the experimental effort, representative fuel blends were compared using the J-gap spark plug concept and a passive pre-chamber spark plug concept. The following figures show a comparison of the two spark plug concepts with the fuels RON 95 E10 and RON 108 M60. Instead of M60, a fuel with a higher ethanol content, such as E85, would be expected to produce very similar results. In operation with E10 RON 95, the known differences between the J-gap and prechamber spark plugs can be observed, including a reduced knock tendency, shorter combustion duration and thus higher pressure rise rates and peak cylinder pressures, as well as advantages in combustion stability. The decisive factor is the intense ignition caused by the flame jets emerging from the pre-chamber. With the M60 fuel, the high octane number made it possible not only to realize efficiency-optimal combustion phasing over the entire load range with both spark plug technologies, but also to significantly reduce the degree of Miller valve timing, i. e., early intake valve closing, and to strongly reduce valve overlap. The overall higher efficiencies can be explained by strong cooling of the gas mass in the cylinder, lower combustion temperatures and thus lower wall heat and The Effect of Fuels with Higher Alcohol Content on the Ignition System 129 <?page no="130"?> dissociation losses, as well as faster combustion that is closer to ideal heat release [5.1, 5.2, 5.3]. Figure 5.3: J-Gap and PC- Sparkplug with E10 RON95 and M60 RON 108 at n=2000 1/ min Furthermore, lower chemical losses such as HC and CO in the exhaust gas are also observed with the M60 fuel. The effective combustion of alcohol fuels with consistently optimal combustion phasing leads to higher pressure rise rates and maximum cylinder pressures, particularly at higher loads, where later combustion phasing must be applied with the RON 95 fuel due to knock. With regard to engine smoothness, the pre-chamber spark plug with the RON 95 fuel leads to comparable values for the standard deviation of indicated mean effective pressure as operation with the M60 fuel, for which only a very small difference between the J-gap and pre-chamber spark plug is observed. While the pre-chamber spark plug, through the 130 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="131"?> emerging jets, provides a very uniform and intense initiation of combustion over the individual cycles and thus low cyclic variations, the effect with the M60 fuel is based on its very high knock resistance and the resulting consistently efficiency-optimal combustion phasing. In this case, the variations occur in the range of the initial combustion phase, close to top dead center, where the volume change is small, and therefore lead to only minor cyclic variations in indicated mean effective pressure. The following figure shows the exhaust gas temperature upstream of the turbine and the raw emissions. Figure 5.4: Exhaust Gas Temperature and Raw Emissions for J-Gap and PC Spark Plugs with E10 RON 95 and M60 RON 108 at n = 2000 rpm As expected from the in-cylinder values, the exhaust gas temperature decreases with earlier combustion phasing. As mentioned above, the high knock resistance of the M60 fuel made it possible to significantly reduce the Miller degree and thus also the valve The Effect of Fuels with Higher Alcohol Content on the Ignition System 131 <?page no="132"?> overlap, which in turn substantially reduces HC and CO emissions. In the lower load range, where the engine can be operated with the same combustion phasing for both fuels, NO x emissions are slightly lower with M60 due to the fuel’s high evaporative cooling effect and the resulting lower combustion temperatures. At higher loads, this effect is overcompensated by the earlier combustion phasing and the resulting higher cylinder pressure and temperature values. Overall, this indicates that high knock resistance combined with high enthalpy of vaporization reduces the relative benefit of pre-chamber ignition. 6 Summary Against the background of increasing requirements for the defossilization of road transport, this study investigated the effects of increased ethanol and methanol contents on combustion, combustion anomalies, and ignition systems in modern spark-ignition engines. For this purpose, various alcohol and ETBE blends were analyzed using a modified 1.5-l TSI engine with an increased compression ratio. The results show that the improvement in combustion phasing cannot be described by the octane number alone. Fuels with high alcohol contents achieve significantly earlier combustion phasing than ETBE-containing blends at comparable octane numbers. The main reason is the high enthalpy of vaporization of ethanol and methanol, which leads to pronounced internal cooling of the fresh charge, especially with direct injection, thereby increasing effective knock resistance. Alcohol blends also show clear advantages with regard to low-speed pre-ignition. As the alcohol content increases, the LSPI limit rises for both ethanol and methanol blends. This effect is particularly pronounced with direct injection, as the enthalpy of vaporization becomes fully effective in the combustion chamber. The comparison between a conventional J-gap spark plug and a passive pre-chamber spark plug shows that high-octane alcohol fuels enable efficiency-optimal combustion phasing over wide load ranges. This increases efficiency and combustion stability while reducing HC and CO emissions. At the same time, the relative benefit of pre-chamber ignition decreases as the fuel’s knock resistance and evaporative cooling increase. Overall, the results demonstrate that alcohol fuels offer significant potential for improving efficiency, combustion stability, and LSPI robustness not only because of their high octane number, but particularly due to their pronounced evaporative cooling. This creates additional degrees of freedom for the future design of efficient spark-ignition engines and for the use of renewable fuel components in the existing vehicle fleet. Literature [1.1] RICHTLINIE (EU) 2023/ 2413 DES EUROPÄISCHEN PARLAMENTS UND DES RATES vom 18. Oktober 2023 zur Änderung der Richtlinie (EU) 2018/ 2001, der Verordnung (EU) 2018/ 1999 und der Richtlinie 98/ 70/ EG im Hinblick auf die Förderung von Energie aus erneuerbaren Quellen und zur Aufhebung der Richtlinie (EU) 2015/ 652 des Rates; [Online] Verfügbar unter: https: / / eur-lex.europa.eu/ legal-content/ DE/ TXT/ PDF/ ? uri=OJ: L_202302413 132 Max Schüttenhelm, Karsten Michels, Thomas Garbe <?page no="133"?> [1.2] European Commission, “Directive (EU) 2018/ 2001 of the European Parliament and of the Council of 11 December 2018 on the promotion of the use of energy from renewable sources (recast)”. 2018. [Online]. Available at: https: / / eur-lex.europa.eu/ eli/ dir/ 2018/ 2001/ oj [1.3] V. Dieterich, A. Buttler, A. Hanel, H. Spliethoff, und S. Fendt, “Power-to-liquid via synthesis of methanol, DME or Fischer-Tropsch-fuels: a review”, Energy Environ. Sci., Bd. 13, Nr. 10, S. 3207-3252, 2020, doi: 10.1039/ D0EE01187H. [1.4] Forschungsvereinigung Verbrennungskraftmaschinen e. V., “Fuelstudy 4b”, Forschungsvereinigung Verbrennungskraftmaschinen e. V., Frankfurt/ Main, Germany, 2021. [1.5] Methanol Institute & Gena (Green Energy analytics), Available at: https: / / methanol.org/ renewable/ [1.6] Global Bioenergy Statistics Report 2024, World Bioenergy Association, Available at: 241023 GBS Report Short Version.pdf [3.1] N. Wissner, S. Healy, M. Cames, J. Sutter, Methanol as a marine fuel, Öko-Institut e. V. , Berlin, 2023 [5.1] H. Menrad, A. König, Alkoholkraftstoffe, Springer Verlag, 1982 [5.2] Chr. Wouters, B. Lehrheuer, B. Heuser, St. Pischinger, Ottomischkraftstoffe mit Methanol, Ethanol und Butanol, MTZ, 2020 [5.3] A. Kuznik, F. Negüs, V. Kelich, M. Thielen, E. Wenz, Chr. Wouters, FVV Abschlussbericht Heft 1243, ICE2025+: Ultimate System Efficiency, 2020 The Effect of Fuels with Higher Alcohol Content on the Ignition System 133 <?page no="135"?> Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther Abstract: To meet todays and future environmental regulations and to reduce engineout emissions combustion engine development aims at achieving highest possible efficiencies in combination with the use of alternative fuels. For conventional spark ignition (SI) engines significant efficiency improvements are mainly reached by operation at high excess air dilution or by increasing the combustion speed. However, an increase in air dilution leads to decreasing flame speeds as well as flammability, thus restricting further improvements. Active pre-chamber (PC) ignition systems enhance combustion stability by increasing turbulence levels and extending flammability limits towards leaner operation conditions. At the same time the accelerated combustion process reduces knock tendency enabling the combination of a PC-jet-ignition combustion with the Spark-Assisted Compression Ignition (PC initiated SACI). The reactive jets emerging from the active PC combustion trigger a controlled volume reaction in the remaining charge of the mainchamber (MC), resulting in very fast combustion and indicated efficiencies higher than 45 %. The understanding of physical processes within the PC and their modelling still remains a key challenge with regard to interaction with the MC e. g. ignition mechanisms and flame propagation. The research addresses the development of a fast-running and physics-based 0D/ 1D burn rate model for PC initiated SACI combustion, based on flame propagation and Homogeneous Charge Compression Ignition (HCCI) with a dedicated focus on the active PC. The target of the modelling approach is the physical description of the correct PC ignition, turbulence and burn rate, alongside future adaption to alternative fuels like methanol, hydrogen and ammonia. The well-established two-zone entrainment model for the flame propagation inside the PC is further developed for the modelling approach and will be linked to a multi pseudozone volume reaction model, based on auto ignition (AI) integrals for the MC combustion progress. Additional sub-models are implemented to depict fuel injection, vaporization and turbulence for an active PC. The modelling is conducted via the so called FVV Cylinder Module environment developed at the Institute of Automotive Engineering Stuttgart (IFS). As part of the FVV research project “SACI II - Pre-Chamber Combustion System with Alternative Fuels” a model validation is conducted against Single-Cylinder-Engine (SCE) measurements from RWTH Aachen. First results show a strong relation between the PC-geometry and turbulence level, which most likely has an <?page no="136"?> influence on the subsequent combustion process in the MC. The developed model enables robust prediction of relevant parameters and shows good agreement in PC burn rate with first experimental data. 1 Introduction Driven by increasingly stringent CO₂ and pollutant emission regulations, the development of SI engines still continues to focus on improving thermal efficiency while reducing engine-out emissions. Lean operation in combination with renewable or carbon-neutral fuels represents a promising way towards highly efficient combustion concepts. Increasing excess-air dilution reduces heat losses, suppresses knock and enables higher compression ratios, thereby improving the engine efficiency. However, the reduced mixture reactivity and lower laminar flame speeds at high dilution levels lead to losses in combustion stability, increased cycle-to-cycle variations and misfires, limiting further efficiency improvements [1]. Active PC ignition systems have therefore emerged as a promising technology to extend lean operating limits while maintaining stable and fast combustion. In contrast to conventional spark ignition, combustion is initiated inside a small auxiliary chamber, generating multiple high-velocity reacting jets that penetrate the MC. These jets significantly enhance turbulence generation and create multiple ignition spots, resulting in a much faster and stable combustion under highly diluted operating conditions [2]. The accelerated combustion process shortens the end-gas residence time, thereby reducing knock tendency and allows the combination of the PC initiated combustion with SACI [3]. Under these conditions, the reacting PC jets trigger controlled auto-ignition within the remaining diluted MC charge, resulting in a distributed combustion process with very high heat release rates and indicated efficiencies exceeding 45 % [4-7]. The successful development and application of PC initiated SACI combustion strongly depend on predictive simulation tools which are able of representing the underlying physical processes. By reducing experimental effort and shortening development time, virtual development environments enable efficient evaluation of combustion concepts. 3D CFD simulation models provide a detailed understanding of combustion processes due to a high spatial resolution, however the computational and time requirements are significantly higher compared to predictive 0D/ 1D simulations. 0D/ 1D simulation models enable reduced model effort for complete combustion systems. This results in fast-running simulations and high efficiency for parameter studies and optimization while still providing physically meaningful predictions of the combustion process. Their modular structure furthermore simplifies the integration of advanced sub-models for a wide variation of combustion phenomena [8,9]. An introduced and established simulation environment for physics-based 0D/ 1D combustion models is the so called FVV Cylinder Module developed at the IFS [9,10]. This framework is based on modular combustion models that can be extended and adapted to new concepts while still remaining short computation times. One of the key modules regarding the scope of the research project and the upcoming sections is the wellestablished two-zone entrainment model for the calculation of flame propagation for the 136 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="137"?> PC combustion phase. Furthermore, to simulate the PC triggered fast combustion inside the MC, a volume reaction model based on HCCI and distributed AI integrals is used. By combining these approaches, a PC initiated SACI is derived as described in the following section [5,10-12]. A PC initiated SACI predictive burn rate model for gasoline fuels was already generated in a previous project at the IFS [5]. However, beside the adaption to alternative fuels and to predict the ignition mode one main reason for continuing the research project is the further development by eliminating a variety of PC simplifications to increase its predictive capability. Previously the PC combustion was represented by a simplified approach directly assuming a homogeneous mixture while neglecting relevant physical processes such as fuel injection, mixture preparation and turbulence inside the PC. Consequently, the predictive capability of the model is limited regarding different operating conditions, PC geometries and future application to alternative fuels. Therefore, this work presents the further development of a physics-based 0D/ 1D predictive burn rate model for PC initiated SACI combustion. The focus lies on the physical representation of the combustion process inside the active PC, including turbulence generation, flame propagation and fuel injection. The developed model is validated against measurements received within the research project. The following introduces the modelling approach for the combined PC initiated SACI approach as well as the detailed changes to the active PC burn rate model. Subsequently, the model changes are validated using SCE measurements from RWTH Aachen University. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 137 <?page no="138"?> 2 SCE Test Bench Setup Fig. 1: SCE with active PC configuration Fig. 1 shows the SCE, equipped with the active PC configuration used for collecting the measurement data for the model development of the PC combustion process. The SCE can be used with conventional and alternative fuels (RON95E10, methanol, ammonia, hydrogen) and is therefore suitable for current and future studies. A high geometric compression ratio of 14.7 up to 15.7 is available by changing the piston configuration, what makes it suitable for SACI operations. Further details regarding the SCE specifications for the operating points in the study are listed in Tab. 1. Cylinder number 1 Total Displacement 400 ccm 3 Stroke 90.5 mm Bore 75 mm Connecting rod length 152 mm Geometric compression ratio 14.7 - 15.7 Load (IMEP) 6 - 15 bar Start of injection 300 °CA before Top Dead Center (bTDC) Injection pressure 200 bar Tab. 1: SCE specifications 138 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="139"?> For pressure measurement, pressure quartzes are integrated in both the MC and the PC. Direct fuel injection into the MC is achieved via a Bosch injector whereas a Borg-Warner prototype injector is used for the PC injection. The six-orifices PC design is developed at the RWTH Aachen and represents a swirl forming PC configuration. Its geometrical data and specifications can be found in Tab. 2. Volume 1080 mm 3 Number of orifices 6 Orifice diameter 0.85 mm Total orifice area 3.4 mm 2 Start of injection 60 °CA bTDC Tab. 2: PC setup The SCE tests for this study are conducted using standard RON95E10 gasoline only, different excess air dilutions (λ = 2 - 3) and a load IMEP of 12 bar. For the modelling process, eleven operating points are chosen to be analysed which can be found in Tab. 3. Fuel RON95E10 Speed 2000 rpm Load 12 bar λ MC 2.0 - 3.0 Total operating points 11 Tab. 3: Experimental operating points 3 PC-SACI Combustion approach As mentioned, the focused PC initiated SACI combustion system is based on the results of a predecessor project. The underlying combustion model describes the process starting from PC ignition, the following flame propagation inside the PC, the discharge of reacting or reacted jets [13] through the PC orifices and the resulting flame propagation and volume reaction in the MC. In both chambers initial flame propagation is described by the same well-established two-zone entrainment approach developed by Tabaczynski [14]. To enable PC initiated SACI, the initial flame propagation model is merged with a volume reaction model based on HCCI combustion. Instead of a classical spark plug, PC flame propagation is used as ignition source to enhance efficiency and conversion speed. The ejected mass from the PC results in a spherical flame propagation with controls the onset of the AI effects inside the MC charge. Multiple ignition spots convert the remaining MC end gas leading to a fast auto ignited volume reaction [5]. Meanwhile, to avoid unrealistically instantaneous heat release, the auto-ignited mass is not converted immediately but according to a characteristic volume reaction time. Consequently, the SACI burn rate model consists of two interacting combustion mechanisms: a flame propagation described by the entrainment model and a distributed volume reaction model. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 139 <?page no="140"?> The share ratio between both combustion mechanisms is dependent on boundary conditions as well on the physical and chemical behaviour of the endgas mixture. By adopting an independent active PC system, MC critical boundary conditions can be shifted towards even larger dilution because of stable initial combustion processes λ PC decoupled). Although the complete PC initiated SACI combustion model is presented in the following, focus of this study lies on the development combustion process inside the PC. 3.1 Flame Propagation Combustion Model For the flame propagation, the well-established two-zone entrainment model is used for the PC combustion phase as well as the MC for the initial phase of flame propagation before triggering volume reaction. The entrainment model was combined with physical theories of Damköhler [15] and Peters [16] and finally implemented in the FVV Cylinder Module by Hann [17,11]. The model describes the division of a combustion chamber into a burnt and unburnt zone, separated by an infinitesimal thin flame front. The burn rate dm b, ent is determined via the mass m F entrained into the flame front and the characteristic burnup time τ L . The entrained mass m F depends on the fresh charge entrainment mass m E . dm b ; ent dt ¼ m F � L ¼ m E � m b ; ent � L According to Peters, the characteristic burn-up time is defined as the ratio of the integral length scale l int to the laminar flame speed s L . a τ is set to be 12. � L ¼ a � � l int s L The calculation of the integral length scale l int is based on a classical cylinder geometry and is tuned to match 3D CFD calculations carried out in Hann’s work. The baseline value of calibration parameter a l,int is 0.0222 [11]. l int ¼ a l ; int � 6 � � V cyl � � 13 To determine the entrainment mass flow dm E into the flame front the following equation is used: dm E dt ¼ � ub � A fl � u E with u E representing the entrainment velocity. The gas density in the unburnt zone (ρ ub ) is depending on the current zone mass and volume where as the flame surface A fl is calculated, using a hemispherical flame propagation inside a cylindrical combustion chamber [18]. The entrainment velocity u E gets calculated according to Peters as: u E ¼ a u � s L � 1 þ 0 : 195 � l int � L � ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ 20 : 5 s L u 0 � l int � L s � 1 ! ! 140 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="141"?> Thereby a u is the main calibration parameter for the entrainment velocity, s L represents the laminar flame speed and δ L the laminar flame thickness [16]. The turbulent fluctuation velocity u' is calculated using a quasi-dimensional k-ε turbulence model, developed in [19-21], with: u 0 ¼ ffiffiffiffiffiffiffiffiffi 2 3 � k r and the turbulence production term: dk dt ¼ dk dt � � prod �" which considers relevant flow effects inside the combustion chamber e. g. charge inflow and wall friction effects. Finally, the calculation of the burn rate can be estimated using the burned mass rate multiplied with the fuel’s Lower Heating Value (LHV): dQ b ; ent dt ¼ dm b ; ent dt � LHV 3.2 Volume Reaction Model The implemented volume reaction model is summarized in the following. It consists of two coupled sub-models: the Detailed AI Model and the Distributed AI Integral. Detailed AI Model The volume reaction model is based on a Detailed AI Model for the calculation of local ignition delay times combined with a Distributed AI Integral concept which predicts the volumetric heat release based on temperature inhomogeneities. To predict effective local ignition delay times, a chemistry-based AI model based on a two-stage three-way Arrhenius approach developed by Fandakov [22] is used. The resulting effective ignition delay times are tabulated and curve fit parameters (α, β, γ) are then used for matching a wide range of in-cylinder boundary conditions including pressure, temperature and lambda variations. � i ¼ A i � p 100 bar � � � �� � � 1 � x EGR ð Þ � � e EA ; i R � T � � A three-way Arrhenius approach is formed by adding up the three shares of ignition delay times. 1 � ¼ 1 � 1 þ � 2 þ 1 � 3 To accurately reproduce the characteristic two-stage ignition behaviour of gasoline fuels, separate Arrhenius formulations are employed for the low-temperature (LT) and hightemperature (HT) regimes. � i ; low = high ¼ A low = high � p 100 bar � � � �� � � 1 � x EGR ð Þ � � e EA ; low ; high R � T � � Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 141 <?page no="142"?> The ignition delay times are calculated at each simulation time step. The Livengood-Wu integral [23] approach is used to determine the effective AI onset, where the reciprocal values of the ignition delay times are integrated. When the LT integral reaches the threshold of 1, the HT integral starts running, whereby LT pressure and temperature increase serve as the boundary conditions for the HT integral. Finally, when the HT integral reaches 1, AI takes place. Z t 1 0 dt � low ð BC Þ ¼ 1 Z t 2 t 1 dt � high ð BC ; T incr ð BC Þ; p incr ð T incr ÞÞ ¼ 1 Distributed AI Integral Since AI onset is highly sensitive to temperature, the MC is represented by a temperature distribution instead of a single homogeneous unburnt zone. The temperature-inhomogeneity driven volume reaction is represented by the Distributed AI Integral, which was developed by Keskin [12]. The PC initiated SACI model assumes, that temperature stratification dominates the spatial ignition sequence. Therefore, the temperature field is described by a contaminated normal distribution consisting of the three following global zones i with the mass fractions ε i [5]: 1. a normal zone: describing the main fraction of unburnt gas inside the MC 2. a wall zone: representing colder mixture nearby the MC walls, and 3. a PC zone, representing the region of highly reactive mixture affected by the PC jets In contrast to the normal and wall zone, whose mass fractions ε norm and ε wall remain predefined, the PC zone’s mass fraction ε PC is calculated dynamically depending on the burned mass ejected by the PC. Each global zone is further subdivided into multiple hundreds so called pseudo-zones j, each deriving an individual unburnt mixture temperature T m,i , calculated via the temperature of the unburned zone T ub of the MC. T m ; wall ¼ T ub � 1 þ T scale b 1000 � � � � � 2 � � T T m ; norm ¼ T ub � 1 þ T scale b 1000 � � � � � " wall � T m ; wall 1 � " wall T m ; PC ¼ 1200K ; for " PC > 0 The standard deviation σ T describes the temperature inhomogeneities in the unburned mixture [12]. T scale_b represents a calibration parameter for the overall temperature level regarding uncertainties of T ub . T m,wall is assumed to be two standard deviation units below normal zone level. T m,PC is currently estimated to be 1200K and represents a simplification which will be replaced by a more sophisticated calculation approach during the study. For every pseudo-zone, the just mentioned Livengood-Wu integral is calculated using the 142 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="143"?> pseudo-zone local ignition delay time τ i,low/ high . The ignition delay time is continuously updated via the Detailed AI Model and the local thermodynamic conditions. Once the HT integral reaches the threshold of 1, direct combustion of the corresponding pseudo-zone mass fraction takes place. The mass conversion rate is calculated by summing up the burned pseudo-zones mass fractions and weighting based on the global zone shares. dm b ; vol ; i dt ¼ X n j ¼ 1 dm b ; vol ; i ; j dt dm b ; vol dt ¼ " norm � dm b ; vol ; norm dt þ " wall � dm b ; vol ; wall dt þ " PC � dm b ; vol ; PC dt Finally, the burn rate of the volume reaction respectively the PC-initiated SACI model can be stated as [5]: dQ b ; vol dt ¼ dm b ; vol dt � LHV dQ b ; PC � SACI dt ¼ dm b ; ent þ dm b ; vol dt � LHV 4 Enhanced Vaporization and Turbulence Modeling in the PC In the previous modelling approach, the PC was represented by a simplified model assuming a homogeneous mixture from calculation start and neglecting the detailed incylinder processes prior to ignition. While this approach enabled first investigations of PC initiated SACI combustion, its predictive capability is limited with respect to in-PC conditions, PC geometries and future application to alternative fuels. Thereby a set of changes, based on the use of gasoline as fuel, were added to the FVV Cylinder Module, expanding the physics based predictive combustion modelling. Firstly, an Injection Module was introduced for achieving a more flexible and physically realistic representation of the injection process and removing the simplification of homogenised and vaporized fuel. The new approach introduces a specified start and end of injection (SOI/ EOI) with a user-defined injection specification. An injection profile as a function of the crank angle and fuel mass flow rate allows to add additional degrees of freedom for possible advanced operating strategies. Multiple injections, ramping or late injections can lead to changes in mixing and stratification, ignition behaviour, jet properties and emissions. Via an external file, the injection profile is handed over to the FVV Cylinder Module, which reads the injected fuel mass flow rate with interpolation routines. The simplification, where the fuel is assumed to be gaseous, is no longer possible. Consequently, the extension using a semi-empirical vaporization module was necessary. The vaporization model is based on an existing FVV Cylinder Module approach for DI ICE [10]. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 143 <?page no="144"?> Fuel vaporization is described by a first order vaporization: dm vap dt ¼ m liq � 4 : 16 � 50 � T T ref � � a T � n n ref � � a n The vaporization rate dm vap dt is proportional to the remaining liquid fuel mass m liq . τ 50 represents the characteristic vaporization time, where 50 % of gasoline fuel is vaporized under reference conditions (T ref = 600 K and n ref = 4000 rpm). Since the reference conditions cannot be experimentally produced within the research project, no exact calibration is possible. Therefore τ 50 is estimated from published evaporation data and calculations [24-26]. The resulting value of τ 50 = 3.4 °CA agrees well with the previous assumption for gasoline fuels and is therefore adapted to the vaporization module. The vaporization rate is then converted into the corresponding vaporization heat flow, which accounts for the latent heat of vaporization and the heating of the injected fuel prior to evaporation. _ Q vap ¼ � 1 � fac wall � � � ð _ m vap � h vap þ _ Q heat Þ A key model parameter affecting the thermal impact of fuel vaporization and is the wallfactor fac wall which specifies the fraction of the evaporation energy supplied by the PC wall instead of the surrounding charge. The wallfactor is assumed to be proportional to the fraction of injected fuel that impinges on the PC wall. Consequently, the corresponding fraction of the evaporation energy is supplied by the PC wall instead of the surrounding charge. So far, a wallfactor default value of 0.4 was used, assuming that 40 % of the evaporation energy is supplied by the PC wall. In regard of depiction a more realistic vaporization and cold start behaviour for alternative fuels like methanol, this topic should not be overlooked. CFD Investigations and literature indicate that the impinging fuel is strongly affected by the PC flow field, which is governed by PC nozzle angle [27,28]. An intense swirl formation leads to retaining fuel at the centre of the swirl, thereby reducing the wall impingement. Based on the findings a more detailed swirl dependent wallfactor calculation for the vaporization module is derived. fac wall ¼ f imp � � spray � R swirl The impingement shares f imp mainly refers to the amount of injected fuel, which theoretically - and without any thermal or flow induced effects - impinging the PC wall in liquid state. This mainly takes place, if the wall distance is lower than the spray penetration depth in spray direction. Accordingly, the percentage is assumed to be high due to the present PC geometry [29,24,25]. The spray shares η spray refer to fuel, which directly vaporizes after ejection before impinging on the PC wall. η spray gets calculated via PC temperature and saturation temperature, based on semi-empirical exponential relation. The calibration parameter C accounts for unresolved effects such as spray breakup and local heat transfer. � spray ¼ 1 � exp � C � T PC � T sat ð p PC Þ T PC � � 144 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="145"?> Finally, the swirl dependent damping share R swirl is introduced. R swirl ¼ exp ð� k s � � swirl ; av Þ The damping factor gets calculated via the swirl ratio λ swirl in this case determined by 3D CFD investigations (Fig. 2). An average swirl ratio λ swirl,aν gets derived for wallfactor calculation depending on the duration of injection (DOI). According to Fig. 2 swirl ratio increases significantly towards TDC explaining high average swirl ratio for long DOI respectively late SOI/ EOI. Since no suitable value for the PC configuration was available, k s is determined by calibration via 3D CFD data, as shown in the next chapter. Fig. 2: PC swirl ratio determined by 3D-CFD investigations The CFD investigations revealed that the nozzle design generates a pronounced global swirl during compression phase. Since the swirl motion governs not only wall impingement but also turbulence generation inside the PC, the existing k-ε turbulence model whose base equation is stated as: dk dt ¼ dk dt � � prod � " was consistently extended. Following the approach of Yang, the incoming mass flow is assumed to induce a global eddy structure around the PC axis (Fig. 3). Instead of resolving the detailed flow field, the swirl is represented by its swirl angular momentum L S (SAM), which serves as the governing state variable [20]. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 145 <?page no="146"?> Fig. 3: Schematic illustration of idealized PC swirl and swirl velocity profile The resulting flow is approximated as a solid-body rotation with the circumferential velocity U S . Viscous wall friction continuously reducing the SAM. The temporal evolution of the swirl angular momentum is obtained from a balance of the inflowing, outflowing and friction-induced momentum rates. L S ¼ Z _ L S ; in þ _ L S ; out þ _ L S ; fric dt The inflowing swirl angular momentum flow rate is calculated as a function of the PC nozzle angle, inflowing mass and nozzle radius. _ L S ; in ¼ C Swirl 1000 � _ m in � R PC � U in � sin ð�Þ _ L S ; out refers to leaving mass fraction and additional reduction of SAM flow rate after PC ignition. To correctly determine the circumferential velocity U S this term is needed. _ L S ; out ¼ C Swirl � _ m out � R PC � U S _ L S ; fric represents the reduction rate due to wall friction and is estimated based on Yang [20]. C fric refers to the wall roughness and friction characteristics. _ L S ; fric ¼ 0 : 05 � h PC � d 1 : 8 PC þ C fric � 0 : 0136 � d 2 : 8 PC � � � � 0 : 8 � � 0 : 2 � U 1 : 8 S The circumferential U S velocity is calculated as: U S ¼ 2 � L S � � � MC � h PC � d PC 3 Based on the calculated swirl angular momentum, turbulence production is represented by two source terms. The turbulence production term dk dt � � prod ; inflow is determined from dimensional scaling of SAM, assuming conversion of rotational kinetic energy into turbulent kinetic energy. C S,prod governs the conversion efficiency. 146 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="147"?> dk dt � � prod ; inflow ¼ C S ; prod � L 2S � 2 � V PC � d PC = 4 ð Þ 4 The wall-induced turbulence production term dk dt � � prod ; Swirl ð wall Þ is derived from the integration of the near-wall shear production using universal wall functions and a wallfriction correlation based on the swirl velocity. dk dt � � prod ; Swirl ð wall Þ ¼ C S ; fric � 0 : 083 � d 1 : 7 PC þ 0 : 254 � h PC � d 0 : 7 PC V PC � U 1 : 8 S � � 0 : 2 � 0 : 2 � � 1 : 5 The turbulence dissipation is modelled via turbulence length scale l turb , linking the dissipation to the global turbulence level of the PC. " ¼ C � � k 32 l turb The change in turbulence kinetic energy is finally determined by the balance of turbulence production and dissipation and directly affects the turbulent fluctuation velocity as well as the entrainment velocity u E . u 0 ¼ ffiffiffiffiffiffiffiffiffi 2 3 � k r 5 PC Model Calibration and Validation Process The newly extended PC burn rate model consists of several coupled physical models. To ensure their consistent evaluation, calibration and validation procedure is performed in a stepwise manner. First, the wallfactor calculation approach is evaluated against 3D CFD results. Afterwards, the PC combustion model is calibrated using experimental data by adjusting turbulence model parameters. Finally, the calibrated model is validated under independent operating conditions. Unless otherwise stated, the combustion model parameters were retained from the previously validated SACI model [5]. Wallfactor model As described in the previous section, the wallfactor is calculated from the combination of impingement, spray vaporization and swirl related effects. The individual shares are analysed regarding their influence and compared against trends observed from 3D CFD investigations. Due to the compact geometry of the research PC (V PC = 1080mm 3 ) and a resulting short wall distance regarding the injector the impingement share f imp is set to 0.9. As seen before η spray follows the assumption of an evaporation behaviour, where the fraction of evaporating fuel before impinging the PC wall increases with increasing PC Temperature difference. Similar exponential approaches are used for heat transfer driven evaporation effects. The calibration parameter C accounting for unresolved effects of spray break-up and local heat transfer is set to 0.3. The strongest dependency is observed for the swirl related damping factor R swirl . The calibration parameter k s represents the only free parameter and therefore requires calibration. Three operating cases with different Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 147 <?page no="148"?> DOI were investigated, where fac CFD wall represents the wallfactor from CFD simulations. k s was determined by matching the calculated effective impingement share f eff imp to the corresponding CFD fac CFD wall results according to: k s ¼ � 1 � swirl ; av � ln f CFD wall f eff imp ! with f eff imp ¼ f imp � � spray ¼ 0 : 838 SOI/ CA aTDC EOI/ CA aTDC fac CFD wall / λswirl,aν/ k s / - Case 1 -60 -48 0.223 12.17 0.109 Case 2 -60 -35 0.157 14.44 0.116 Case 3 -60 -20 0.128 17.37 0.108 Tab. 4: 3D CFD operating cases and overall calculated wallfactor parameters A value of k s � 0 : 11 provides the best agreement for all investigated operating points, leading to a mean absolute deviation for the calculated vs. CFD-simulated wallfactor of 2.6 %. Overall, the dynamic wallfactor calculation enables the consideration of injection timing, various geometries and PC designs and flow-dependent effects, replacing the previously applied constant wallfactor approach. Calibration of Entrainment and Turbulence Model After calibration of the wallfactor model the PC flame propagation’s parameters are determined. Since the entrainment velocity directly depends on the turbulent fluctuation velocity, the entrainment and turbulence models are calibrated simultaneously. Unlike the wallfactor calibration, no CFD-based reference for the PC turbulence is available. Therefore, the calibration is performed using measured combustion phase as target quantity. The operating point for calibration is carried out at 2000 rpm with an IMEP of 12 bar and λ = 2.4. A Pressure Trace Analysis (PTA) is conducted resulting in a reference Rate of Heat Release (RoHR) and cumulative burn rate for calibration. The initial implementation of both new models already improves the predicted combustion behaviour by reducing burn duration and smoothing RoHR gradient. However, calibration of the newly introduced model parameters is required to obtain quantitative agreement. To better match the simulated combustion start to PTA, an entrainment delay parameter f delay is implemented. This accounts for a finite development of the PC flame before effective entrainment begins. The delay intensity can now be chosen by the user. Main parameter regarding turbulence production and resulting turbulent fluctuation velocity is the newly introduced parameter C Swirl . It controls the intensity of SAM generation from the inflowing mass and determines the overall turbulence production level. An increasing C Swirl leads to higher SAM, increased turbulence and larger u'. It was adjusted until the simulated burn rate reproduced the PTA’s burn rate. A C Swirl factor of 1.9 is identified resulting in good agreement between PTA and simulation. The tuning factors C s,fric respectively C s,prod are kept at the default value of 1. Additionally, during the calibration process it becomes apparent, that the RoHR tail towards end of combustion (EOC) inside the PC can be inadequately modelled, leading to unrealistically long combustion tails. The 148 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="149"?> original formulation of the integral length scale l int is designed for conventional combustion chambers, tending to overestimate the characteristic burn up time in the small PC. Therefore, the corresponding calibration factor a l,int is reduced to match the assumptions of Hann, whereas l int lies in a range of 0.2…2 mm. Fig. 4: Comparison between previous model (pr. SIM) and new extended PC burn rate model (SIM) with PTA Fig. 4 compares the simulated RoHR and cumulative burn rate between PTA (red) and the previous model (dashed blue) as well as the new model with model extension (dashed turquoise). While the former model overestimates the burn duration and RoHR gradient, the calibrated latter one reproduces both better. Validation Process After calibration at one operating point, the model is evaluated using a total of 15 lambda sweep operating points. The objective of the validation is not to demonstrate agreement for a single calibrated operating point but to evaluate the predictive capability of the constant parameter set for a wide lambda range. Focus lies on the PC combustion phasing only, while the PC/ MC interaction will be further investigated in future studies. Fig. 5 and 6 show the obtained pressure curve, RoHR and cumulative burn rate compared to the PTA for two representative operating points. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 149 <?page no="150"?> Fig. 5: Predictive capability of new extended PC burn rate model for pressure curve, RoHR and cumulative burn rate at λ MC =2.0 Fig. 6: Predictive capability of new extended PC burn rate model for pressure curve, RoHR and cumulative burn rate at λ MC =2.8 For both operating points the simulated pressure trace is in good agreement with the PTA. The pressure rise as well as the position of the PC pressure maximum are reproduced in an accurate way. The agreement is also represented in the corresponding RoHR. The onset as well as the overall combustion duration is matched with good accuracy. For further investigations regarding the onset inside the MC, the correct modelling of burn duration as well as SOC/ EOC represents a main target of the calibration work. The cumulative heat release confirms that the combustion phasing and the total released energy are predicted consistently. Minor deviations remain in the peak RoHR as well towards the end of combustion. The latter one can be attributed to the PTA and should be interpreted with caution. The PTA implemented in the FVV Cylinder Module is based on a modified approach for conventional engine cylinders (and geometries). When applied to the small PC volume and corresponding low injected fuel mass, the resolution of the RoHR becomes sensitive towards end of combustion. Consequently, the sudden EOC in the PTA is likely influenced by methodological limitations rather than representing the actual combustion process. The smoother predicted EOC of the simulation appears therefore physically 150 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="151"?> plausible and is considered to provide a satisfactory combustion process representation. Meanwhile an underprediction of a l,int should be avoided. Next, different regression analyses are carried out, representing the comparison between PTA and simulation for all 15 operating points. The values of RoHR, cumulative burn rate, burn duration MFB 1090 and the maximum PC pressure before MC ignition are shown in Fig. 7. The performance of the newly implemented PC burn rate model is visualized in red vs. the existing predecessor burn rate model in grey. It should be noted, that the complete parameter set was kept constant throughout the entire lambda sweep. Fig. 7: Regression analysis for predicted peak PC RoHR, cumulative burn rate, burn duration MFB1090 and peak PC pressure of the new extended PC burn rate model (red) and the previous model (grey) A generally good agreement is found, although the prediction accuracy is comparable to the previous model. The additional degrees of freedom due to the modelling approach do not affect the robustness of the model. The RoHR maximum performs slightly better and the combustion phasing remains essentially unchanged despite the new turbulence formulation, as seen in the MFB 1090 . Moreover, an improvement can be observed for the prediction of the maximum PC pressure. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 151 <?page no="152"?> The newly implemented PC burn rate model provides a more physically consistent description of the PC processes. In particular, the nozzle angle influenced turbulence production and fuel evaporation can now be depicted explicitly, while maintaining a solid prediction quality. Overall, the calibration and validation demonstrate that the model extension preserves the predictive capability while increasing its physical consistency. The newly introduced injection and evaporation module as well as the swirl formulations enable the consideration of different PC geometries and operating strategies without requiring parameter adjustments. Using one constant parameter set, good agreement with the PTA is obtained over the complete lambda sweep, indicating that the model provides a robust basis for future investigations of alternative fuels, PC/ MC interactions and ignition processes inside the MC. 6 Summary and Outlook This work represents an extension of a quasi-dimensional burn rate model for PC initiated SACI combustion, by introducing a more physical approach for the combustion processes within the PC. The previous approach assumed a homogeneous PC mixture from the start of calculation and was replaced by an injection and a vaporization module. The former enables the consideration of realistic injection strategies. Furthermore, a swirl dependent wallfactor calculation was developed to account for in-cylinder flow structures on wall impingement and fuel vaporization effects such as PC temperature dependency. The investigated swirl motion was additionally integrated into the existing k-ε turbulence model by introducing SAM as the state variable for turbulence fluctuation velocity calculation. This enables a physically motivated prediction of turbulence inside the PC with an explicit dependency on the PC nozzle geometry. Although the quasi-dimensional representation cannot depict the complete 3D flow field, it provides additional geometric degrees of freedom without increasing the computational effort significantly. The wallfactor calculation was calibrated using 3D CFD data and showed a sound match over different injection strategies and swirl intensities. Subsequently, the modified entrainment and turbulence models were calibrated using PTA results of a representative operating point. After establishing a fixed set of calibration parameters and applying it to the model environment, a lambda sweep with 15 operating points was simulated, demonstrating its predictive capability over a wide operating range. Slight improvements were observed in RoHR maximum and peak PC pressure. The implemented model extensions build a flexible base for future investigations of alternative fuels and PC geometries. Future work will now focus on interactions between PC and MC, resulting in an ignition regime definition (flame or jet ignition) [13] via a physics-based prediction of the effective PC zone temperature T m,PC . This will enable the derivation of non-dimensional criteria for differentiating flame and jet ignition as well as critical non-ignition states. 152 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="153"?> 7 Acknowledgement The presented work was carried out within the FVV project 1540 “SACI II - Pre-Chamber Combustion System with Alternative Fuels” in cooperation with the Institute of Automotive Engineering Stuttgart (IFS) at the University Stuttgart under the direction of Univ.- Prof. Dr.-Ing. André Casal Kulzer, the Chair of Thermodynamics of Mobile Energy Conversion Systems at RWTH Aachen under the direction of Univ.-Prof. Dr.-Ing. (USA) Stefan Pischinger and the Institut of Thermal and Fluid Engineering at the University of Applied Science and Arts North Western Switzerland (FHNW) under the direction of Univ.-Prof. Dr. sc. techn. Kai Herrman. The authors gratefully acknowledge the support received by the FVV Research Association for Combustion Engines e. V. and the working group under the supervision of Dr.-Ing. Jonas Villforth (Dr. Ing. h. c. F. Porsche AG). References [1] Grundlagen Verbrennungsmotoren, Vieweg+Teubner Verlag, Wiesbaden, 1 Jan. 2011. [2] Salerno, F., Bargende, M., Kulzer, A. C., and Grill, M., “A Quasi-Dimensional Burn Rate Model for Spark-Assisted Compression Ignition (SACI) Combustion,” SAE International Journal of Advances and Current Practices in Mobility, Vol. 05, No. 3, 1 Jan. 2023, pp. 1292-1311, URL: https: / / saemobilus.sae.org/ articles/ a-quasi-dimensional-burn-rate-model-spark-assisted-compression-ignition-saci-combustion-2022-24-0039#citation. [3] Robertson, D., and Prucka, R., “A Review of Spark-Assisted Compression Ignition (SACI) Research in the Context of Realizing Production Control Strategies,” SAE Technical Paper, 9 Sep. 2019. [4] Chiodi, M., Kaechele, A., Bargende, M., Wichelhaus, D., and Poetsch, C., “Development of an Innovative Combustion Process: Spark-Assisted Compression Ignition,” SAE International Journal of Engines, Vol. 10, No. 5, 1 Jan. 2017, pp. 2486-2499, URL: https: / / saemobilus.sae. org/ articles/ development-innovative-combustion-process-spark-assisted-compression-ignition-2017-24-0147#citation. [5] Salerno, F., “A quasi-dimensional burn rate model for pre-chamber-initiated SACI combustion,” 1 Jan. 2024. [6] Salerno, F., Kulzer, A., Bargende, M., Grill, M., Burkardt, P., et al., “A Quasi-Dimensional Two- System Burn Rate Model for Pre-Chamber-Initiated SACI Combustion,” SAE International Journal of Advances and Current Practices in Mobility, Vol. 06, No. 4, 1 Jan. 2024, pp. 1815-1836. doi: 10.4271/ 2023-24-0002. [7] Burkardt, P., Lehrheuer, B., and Pischinger, S., “Influence of alternative pre-chamber fuels on engine performance,” International Journal of Engine Research, Vol. 27, No. 5, 1 Jan. 2026, pp. 686-701. doi: 10.1177/ 14680874251382661. [8] Merker, G. P., and Teichmann, R. (eds.), Grundlagen Verbrennungsmotoren, Springer Fachmedien Wiesbaden, Wiesbaden, 1 Jan. 2018. [9] Grill, M., “Objektorientierte Prozessrechnung von Verbrennungsmotoren,” 1 Jan. 2006. [10] Grill, M., and Bargende, M., “The Development of an Highly Modular Designed Zero- Dimensional Engine Process Calculation Code,” SAE International Journal of Engines, Vol. 03, No. 1, 1 Jan. 2010, pp. 1-11. doi: 10.4271/ 2010-01-0149. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 153 <?page no="154"?> [11] Hann, S. (ed.), A Quasi-Dimensional SI Burn Rate Model for Carbon-Neutral Fuels, Springer Fachmedien Wiesbaden, Wiesbaden, 1 Jan. 2021. [12] Keskin, M. T., Grill, M., and Bargende, M., “Development of a Fast, Predictive Burn Rate Model for Gasoline-HCCI,” SAE Technical Paper Series, edited by M. T. Keskin, M. Grill and M. Bargende, SAE International400 Commonwealth Drive, Warrendale, PA, United States, 1 Jan. 2016. [13] Biswas, S., Physics of Turbulent Jet Ignition, Springer International Publishing, Cham, 1 Jan. 2018. [14] Tabaczynski, R. J., Ferguson, C. R., and Radhakrishnan, K., “A Turbulent Entrainment Model for Spark-Ignition Engine Combustion,” SAE Technical Paper, 1 Feb. 1977. [15] Damköhler, G., “Der Einfluss der Turbulenz auf die Flammengeschwindigkeit in Gasgemischen,” Zeitschrift für Elektrochemie und Angewandte Physikalische Chemie, Vol. 46, No. 11, 1 Jan. 1940, pp. 601-626. [16] PETERS, N., “The turbulent burning velocity for large-scale and small-scale turbulence,” Journal of Fluid Mechanics, Vol. 384, 1 Jan. 1999, pp. 107-132. doi: 10.1017/ S0022112098004212. [17] Hann, S., Grill, M., Bargende, M., and Altenschmidt, F., “A Quasi-Dimensional SI Burn Rate Model for Predicting the Effects of Changing Fuel, Air-Fuel-Ratio, EGR and Water Injection,” SAE Technical Paper Series, edited by S. Hann, M. Grill, M. Bargende and F. Altenschmidt, SAE International400 Commonwealth Drive, Warrendale, PA, United States, 1 Jan. 2020. [18] Grill, M., Billinger, T., and Bargende, M., “Quasi-Dimensional Modeling of Spark Ignition Engine Combustion with Variable Valve Train,” SAE Technical Paper, 3 Apr. 2006. [19] Fasse, S., Grill, M., and Bargende, M., “Optimierte Modellierung von Ladungswechsel und Turbulenz in der Motorprozessrechnung,” MTZ - Motortechnische Zeitschrift, Vol. 82, No. 11, 1 Jan. 2021, pp. 86-91. doi: 10.1007/ s35146-021-0720-7. [20] Yang, Q., A Quasi-dimensional Charge Motion and Turbulence Model for Combustion and Emissions Prediction in Diesel Engines with a fully Variable Valve Train, Springer Fachmedien Wiesbaden, Wiesbaden, 1 Jan. 2022. [21] Fritsch, S., Fasse, S., Yang, Q., Grill, M., and Bargende, M., “A Quasi-Dimensional Charge Motion and Turbulence Model for Spark Injection Engines with Fully Variable Valve Train and Direct Fuel Injection,” Experten-Forum Powertrain: Ladungswechsel und Emissionierung 2019, edited by J. Liebl, Springer Fachmedien Wiesbaden, Wiesbaden, 1 Jan. 2020, pp. 24-39. [22] Fandakov, A., A Phenomenological Knock Model for the Development of Future Engine Concepts, Springer Fachmedien Wiesbaden, Wiesbaden, 1 Jan. 2019. [23] Livengood, J. C., and Wu, P. C., “Correlation of autoignition phenomena in internal combustion engines and rapid compression machines,” Symposium (International) on Combustion, Vol. 5, No. 1, 1 Jan. 1955, pp. 347-356. doi: 10.1016/ S0082-0784(55)80047-1. [24] Mitroglou, N., Nouri, J. M., Gavaises, M., and Arcoumanis, C., “Spray Characteristics of a Multihole Injector for Direct-Injection Gasoline Engines,” International Journal of Engine Research, Vol. 7, No. 3, 1 Jan. 2006, pp. 255-270. doi: 10.1243/ 146808705X62922. [25] Mitroglou, N., Nouri, J. M., Yan, Y., Gavaises, M., and Arcoumanis, C., “Spray Structure Generated by Multi-Hole Injectors for Gasoline Direct-Injection Engines,” SAE World Congress, 1 Jan. 2007, URL: https: / / www.researchgate.net/ publication/ 258541019_Spray_Structure_Generated_by_Multi-Hole_Injectors_for_Gasoline_Direct-Injection_Engines. 154 F. Meyer, A. Casal Kulzer, Q. Yang, T. Tietz, M. Günther <?page no="155"?> [26] Giyad, W. Y., Shahad, H. A., Mjbel, H. M., and Al-Kayiem, H. H., “Experimental Study of Droplet Evaporation of Different Hydrocarbon Fuels,” International Journal of Heat and Technology, Vol. 43, No. 1, 1 Jan. 2025, pp. 209-218. doi: 10.18280/ ijht.430122. [27] Hu, J., Pei, Y., An, Y., Zhao, D., Zhang, Z., et al., “Study of active pre-chamber jet flames based on the synergy of airflow with different nozzle swirl angle,” Energy, Vol. 282, 1 Jan. 2023, p. 128198, URL: https: / / www.sciencedirect.com/ science/ article/ pii/ S036054422301592X. [28] Zhao, D., An, Y., Pei, Y., Shi, H., and Wang, K., “Numerical study on the asymmetrical jets formation from active pre-chamber under super-lean combustion conditions,” Energy, Vol. 262, 1 Jan. 2023, p. 125446. doi: 10.1016/ j.energy.2022.125446. [29] Lefebvre, A. H., and McDonell, V. G., Atomization and Sprays, CRC Press, Second edition. | Boca Raton : Taylor & Francis, CRC Press, 2017., 1 Jan. 2017. Physics-Based 0D/ 1D Modelling of Pre-Chamber Combustion for SACI Applications 155 <?page no="157"?> Impact of Injection Strategy on Autoignition and Combustion Regime of Ultra-Lean SACI Combustion with RON95E10 T. Tietz, M. Günther, S. Pischinger Absrtact: Despite increasing electrification, spark-ignition engines remain a key powertrain technology, requiring further improvements in efficiency and emissions. However, the trade-off between knock resistance and autoignition propensity limits conventional gasoline combustion, particularly under ultra-lean conditions. Spark Assisted Compression Ignition (SACI) enables controlled autoignition at high compression ratios by combining flame propagation with compression-driven ignition, thereby extending operating limits and improving efficiency while reducing emissions. This study investigates the influence of main chamber injection timing (SOI MC ) and prechamber injection duration (DOI PC ) on SACI combustion using RON95E10 under ultra-lean conditions. Experiments were conducted on a singlecylinder engine equipped with an active prechamber system at moderate load (IMEP = 12 bar), high compression ratio (CR = 14.7), and intake temperatures up to T intake = 45 °C. Complementary 3D-CFD simulations were applied to analyze mixture formation and local reactivity. Advancing the main-chamber injection significantly improves mixture homogenization due to enhanced tumble-driven mixing, as confirmed by CFD-based equivalence-ratio distributions. This promotes flame-induced autoignition, increases the SACI ratio, and shifts the combustion process towards increasingly volumetric combustion while maintaining nearly constant indicated efficiency. As a result, combustion duration (Δθ 10-90 ) is reduced and higher pressure rise rates are achieved, indicating an increased contribution of autoignition. The prechamber injection duration directly affects local mixture formation and thus the reactivity of the resulting jets. Increasing DOI PC leads to near-stoichiometric conditions at the sparkplug, enabling a stronger and more reactive jet. This enhances the coupling between flame propagation and autoignition in the main chamber and stabilizes SACI operation. The combined variation of SOI MC and DOI PC demonstrates that both parameters are key control variables for managing ignition delay, heat release structure, and the autoignition behavior. Optimal settings enable stable ultra-lean operation at λ = 3.0 with indicated efficiencies above η i > 44 % while maintaining controlled pressure rise rates. The results highlight that injection strategy governs the transition between flamedominated and autoignition-dominated combustion and provides an effective pathway to extend the operational limits of SACI combustion. <?page no="158"?> 1 Introduction The continuously increasing demand for higher thermal efficiencies and lower pollutant emissions requires the development of advanced combustion concepts for future sparkignition engines. While conventional spark-ignition (SI) combustion has been continuously improved over the past decades, its efficiency at lean operating conditions remains fundamentally limited by the relatively low laminar flame speed. As the air-fuel ratio increases, flame propagation becomes progressively slower, leading to prolonged combustion durations, increased cyclic variability and ultimately unstable combustion. Consequently, the achievable lean operating limit is primarily determined by combustion stability rather than by knock or thermal loading. One promising approach to overcome these limitations is the utilization of compression-assisted combustion concepts. Homogeneous Charge Compression Ignition (HCCI), for example, enables an almost volumetric combustion process through spontaneous autoignition of the entire mixture. Since a large fraction of the released chemical energy is converted under nearly constant-volume conditions, HCCI offers excellent thermal efficiencies while simultaneously reducing nitrogen oxide emissions due to the comparatively low combustion temperatures. However, the absence of a direct ignition trigger makes the combustion phasing highly sensitive to operating conditions such as intake temperature, residual gas fraction and fuel properties. Consequently, stable operation is typically restricted to a narrow load range, while transient operation remains challenging. Spark-Assisted Compression Ignition (SACI) combines the advantages of conventional SI combustion and HCCI while mitigating their respective limitations. Combustion is initiated by a conventional spark ignition inside an actively fueled pre-chamber, where a locally rich mixture is ignited reliably. The resulting pressure increase accelerates several reacting turbulent jets through the pre-chamber nozzles into the ultra-lean mixture of the main chamber. These reacting jets fulfil two essential functions. First, they act as multiple distributed ignition sources, thereby significantly accelerating flame propagation within the combustion chamber. Second, they transport heat, radicals and chemically active species into the unburned mixture while simultaneously increasing the local pressure. Once sufficiently favorable thermodynamic conditions are established, the remaining mixture undergoes flame-induced auto-ignition. Unlike conventional HCCI combustion, this auto-ignition is neither random nor uncontrolled but is triggered by the preceding flame propagation initiated inside the pre-chamber. As a result, the combustion process gradually changes from classical flame propagation to an increasingly volumetric heat release. A growing fraction of the fuel is consumed almost simultaneously, causing the heat addition to approach the ideal constant-volume combustion process of the Otto cycle. Consequently, SACI combustion enables significantly higher indicated efficiencies while simultaneously extending the lean operating limit. Furthermore, the lower combustion temperatures associated with lean operation reduce thermal NOx formation, making SACI an attractive combustion concept for future highly efficient spark-ignition engines. Despite these advantages, the combustion process becomes considerably more complex than conventional SI combustion because flame propagation and flame-induced auto- 158 T. Tietz, M. Günther, S. Pischinger <?page no="159"?> ignition occur simultaneously and continuously interact. The transition between both combustion modes is governed by the local thermodynamic conditions within the unburned mixture and therefore strongly depends on the mixture preparation in both the main chamber and the pre-chamber. The mixture formation inside the main chamber determines the local equivalence ratio distribution, evaporation behavior and temperature field prior to ignition. Consequently, the main-chamber injection timing directly influences the development of the flame front as well as the thermodynamic conditions under which flame-induced auto-ignition can occur. Similarly, the amount of fuel injected into the pre-chamber governs the energy released during the initial ignition process, the momentum and reactivity of the turbulent jets, and therefore the capability of the pre-chamber to initiate volumetric combustion within the main chamber. Both parameters therefore represent key calibration variables for controlling the transition between conventional SI combustion and SACI combustion. Although the influence of intake temperature, mixture preparation and pre-chamber fueling has been investigated in several previous studies, the resulting combustion process is generally evaluated using conventional combustion metrics such as burn duration, pressure rise rate, peak pressure or apparent heat release. While these quantities provide valuable information regarding the global combustion characteristics, they do not directly quantify the actual contribution of flame-induced auto-ignition. Consequently, it remains difficult to distinguish whether an observed combustion acceleration originates from enhanced turbulent flame propagation or from an increasing fraction of volumetric combustion. A quantitative description of the auto-ignition contribution would therefore provide a substantially improved understanding of the combustion process and enable a more systematic evaluation of the governing calibration parameters. Such an approach would allow the continuous characterization of the transition from SI to SACI combustion while simultaneously providing a physically meaningful basis for comparing different operating conditions. For this purpose, the present work introduces a novel pressure-based methodology for quantifying the contribution of flame-induced auto-ignition using only measured incylinder pressure traces. Based on the analysis of the isochoric pressure trace and its higher-order derivatives, the combustion process is separated into flame-propagation and auto-ignition contributions, allowing the definition of a continuous SACI ratio. Rather than replacing conventional combustion metrics, the proposed methodology complements them by providing direct information about the relative contribution of volumetric combustion. The developed methodology is subsequently applied to a single cylinder research engine equipped with an actively fueled pre-chamber. First, the influence of intake temperature on the onset of flame-induced auto-ignition and the lean operating limit is investigated. Building upon these findings, the effects of the main-chamber injection timing on mixture preparation and combustion development are analysed in detail. Particular emphasis is placed on the interaction between fuel stratification, flame propagation and the onset of auto-ignition. Finally, the influence of the pre-chamber fuel quantity is investigated to identify how the initial ignition energy and the reacting jet characteristics affect the transition towards SACI combustion. Unlike conventional Impact of Injection Strategy on Autoignition and Combustion Regime 159 <?page no="160"?> combustion metrics, the proposed SACI ratio enables a continuous quantification of the transition from flame-propagation-dominated combustion to volumetric SACI combustion, thereby providing a physically meaningful basis for evaluating different calibration strategies. 2 Methodology 2.1 Experimental Setup The experimental investigations were conducted on the single-cylinder research engine (SCE) shown in Figure 1, which is equipped with an actively fueled pre-chamber combustion system. The test bench was specifically developed for investigating Spark- Assisted Compression Ignition (SACI) combustion under well-defined and reproducible operating conditions. Owing to its flexible fuel supply system, the engine can be operated with conventional as well as alternative fuels, including RON95E10 gasoline, methanol, methane and hydrogen, making it suitable for investigations of future combustion concepts. Figure 1: SCE Setup The combustion system consists of a centrally mounted Bosch gasoline direct injector supplying the main chamber and an additional BorgWarner prototype injector for independent fueling of the pre-chamber. Combustion is initiated by a conventional spark plug located inside the pre-chamber. The independent fuel supply of both chambers enables separate adjustment of the main-chamber injection timing (SOI) and the pre- 160 T. Tietz, M. Günther, S. Pischinger <?page no="161"?> chamber fuel quantity (DOI), allowing a systematic investigation of their influence on the transition from conventional spark-ignition combustion towards SACI combustion. Depending on the piston configuration, the geometric compression ratio can be varied between CR = 14.7 and 15.7, enabling investigations under both conventional SI and highly diluted SACI operating conditions. The principal engine specifications are summarized in Table 1. Cylinder number 1 Total Displacement 400 ccm³ Stroke 90.5 mm Bore 75 mm Connecting rod length 152 mm Geometric compression ratio 14.7 - 15.7 Tab. 1: SCE specifications Cylinder pressure was measured in both the main chamber and the pre-chamber using piezoelectric pressure transducers, allowing a detailed analysis of the combustion process in both chambers. All combustion analyses presented in this work are based on the average of 200 consecutive engine cycles in order to reduce the influence of cyclic combustion variations. The actively fueled pre-chamber was developed at RWTH Aachen University and features a swirl-generating six-orifice design. The corresponding geometric specifications are listed in Table 2. Volume 1080 mm³ Number of orifices 6 Orifice diameter 0.85 mm Total orifice area 3.4 mm² Tab. 2: PC setup Unless stated otherwise, all experiments were conducted using commercial RON95E10 gasoline at a constant engine speed of n = 2000 min -1 and an indicated mean effective pressure of IMEP = 12 bar. The main-chamber injection was usually performed at SOI MC = 300 °CA bTDC, whereas the pre-chamber injection was fixed at SOI PC = 60 °CA bTDC. The injection pressure for both injectors is p rail = 200 bar. The global air-fuel ratio was varied between λ = 1.0 and 3.4 to investigate the transition from conventional spark-ignition combustion towards flame-assisted compression ignition. Based on the identified SACI operating regime, the main-chamber start of injection (SOI) and the injected fuel quantity in the pre-chamber (DOI) were subsequently varied systematically to investigate their influence on mixture preparation, combustion development and flame-induced autoignition. Impact of Injection Strategy on Autoignition and Combustion Regime 161 <?page no="162"?> 2.2 Numerical Methodology To support the experimental investigations, three-dimensional computational fluid dynamics (CFD) simulations were performed using the commercial software CONVERGE CFD. The computational domain comprises the complete intake and exhaust ports, intake and exhaust valves, piston, cylinder liner, combustion chamber including the actively fueled pre-chamber, the main-chamber direct injector, the pre-chamber injector and the spark plug. Figure 2: Simulation model A Cartesian base grid with a cell size of 4 mm was employed and locally refined to a minimum cell size of 0.5 mm using fixed embedding regions in combination with adaptive mesh refinement (AMR), resulting in approximately 4×10 6 computational cells. Turbulence was modelled using the k - ω - SST model, while spray breakup was described by the Kelvin-Helmholtz/ Rayleigh-Taylor (KH-RT) model. Real-gas thermodynamic properties were considered throughout the simulations, and experimentally measured intake and exhaust pressure traces were prescribed as boundary conditions. Gasoline chemistry was represented by the detailed reaction mechanism proposed by Cai et al., comprising 490 species and 2072 elementary reactions. However, since the present numerical investigations focus primarily on mixture preparation prior to ignition, combustion itself is not discussed in the following. Instead, the simulations are used to analyse the influence of different main-chamber injection timings on fuel evaporation, mixture stratification and local equivalence ratio distributions. These numerical results provide a physical explanation for the experimentally observed combustion behaviour. 2.3 Determination of the SACI Ratio One of the major challenges in analysing Spark-Assisted Compression Ignition combustion is the simultaneous occurrence of turbulent flame propagation and flame-induced auto-ignition. While conventional combustion parameters such as burn duration, peak pressure or pressure rise rate describe the overall combustion behaviour, they do not allow a quantitative distinction between both combustion mechanisms. 162 T. Tietz, M. Günther, S. Pischinger <?page no="163"?> To overcome this limitation, a pressure-based evaluation methodology was developed to determine the contribution of flame-induced auto-ignition using only measured incylinder pressure traces. The complete evaluation procedure is illustrated in Figure XX. Figure 3: Methodology for determining the SACI ratio First, an isochoric pressure trace is calculated by subtracting a scaled motored pressure curve from the measured fired pressure trace. This procedure removes the pressure variation originating from piston motion and isolates the pressure increase caused exclusively by chemical heat release. The resulting isochoric pressure curve therefore provides a significantly improved basis for identifying combustion-related events. Subsequently, the first, second and third derivatives of the isochoric pressure trace are evaluated to determine the characteristic transition from flame propagation towards autoignition. Based on the characteristic extrema and zero crossings of these derivatives, the onset and end of flame-induced auto-ignition are identified. Since turbulent flame propagation continues during the auto-ignition phase, the measured heat release cannot directly be assigned to volumetric combustion. Therefore, a reference combustion curve representing the hypothetical continuation of pure sparkignition combustion is reconstructed using a smooth parabolic interpolation between the identified transition points. The additional heat release exceeding this reconstructed reference is attributed to flame-induced auto-ignition. Finally, the flame-propagation and auto-ignition contributions are integrated separately to define the SACI ratio, which represents the fraction of the total combustion energy released by flame-induced auto-ignition: SACI ratio ¼ Q AI Q SI þ Q AI : Impact of Injection Strategy on Autoignition and Combustion Regime 163 <?page no="164"?> A value of zero therefore corresponds to conventional spark-ignition combustion without detectable auto-ignition, whereas increasing values indicate an increasing contribution of volumetric combustion. Throughout the present work, the SACI ratio is used as the primary evaluation parameter to investigate the influence of intake temperature, mainchamber injection timing and pre-chamber fuel quantity on the transition from conventional SI combustion towards SACI combustion. 3 Results and discussion The proposed SACI ratio serves as the primary evaluation parameter throughout the following investigations. Conventional combustion metrics are used to complement the analysis and to relate changes in combustion behaviour to the contribution of flameinduced autoignition. First, the transition from spark-ignition (SI) to SACI combustion is analysed by varying the global air-fuel ratio. Based on the identified SACI operating regime, the effects of the main-chamber start of injection (SOI) and the injected fuel quantity in the pre-chamber (DOI) are subsequently investigated. While the SOI primarily influences mixture preparation and fuel stratification in the main chamber, the DOI determines the combustion intensity inside the pre-chamber and therefore the characteristics of the reacting jets entering the main chamber. The combination of experimental measurements and numerical simulations enables the underlying physical mechanisms governing the transition towards SACI combustion to be identified. 3.1 Transition from Spark-Ignition to SACI Combustion The global air-fuel ratio is the primary parameter controlling the transition from conventional spark-ignition combustion towards SACI combustion. Therefore, the first part of this study investigates the relative air/ fuel ratio while maintaining a constant engine speed and load. Besides combustion efficiency and stability, particular attention is paid to the evolution of the combustion duration and the contribution of flame-induced auto-ignition. At moderate rel. air/ fuel ratios λ, combustion is almost exclusively governed by turbulent flame propagation. Increasing λ initially slows down the combustion process as the laminar flame speed decreases. Consequently, the burn delay and the overall combustion duration increase. At the same time, the indicated efficiency rises because the leaner mixture reduces heat losses while maintaining stable combustion. As the mixture becomes progressively leaner, combustion stability eventually limits further dilution at low intake temperatures. Increasing the intake temperature extends this operating range and enables stable combustion even beyond λ = 3, where indicated efficiencies of approximately η i = 44.6 % are achieved. This behaviour is directly reflected by the proposed SACI ratio, which remains close to zero at low λ, indicating predominantly flame-propagation-controlled combustion. 164 T. Tietz, M. Günther, S. Pischinger <?page no="165"?> Figure 4: General diagrams of the relative air/ fuel ratio variation for two intake temperatures Although the combustion process initially becomes slower with increasing λ, a fundamentally different behaviour is observed once the SACI operating region is reached. At an intake temperature of T in = 45 °C, the transition begins at approximately λ = 1.8, whereas at T in = 25 °C the onset is shifted to around λ = 1.9. From this point onwards, the combustion duration no longer follows the expected trend of continuously increasing with λ. Instead, the late combustion phase accelerates because flame-induced auto-ignition increasingly contributes to the total heat release. For the higher intake temperature, the burn duration between MFB 10 and MFB 90 remains nearly constant for λ > 2.4 despite the continuously decreasing laminar flame speed, clearly indicating the increasing contribution of volumetric combustion. Accordingly, the SACI ratio increases continuously with increasing λ, revealing the gradual transition from conventional spark-ignition combustion towards flame-assisted compression ignition. At identical air-fuel ratios, the higher intake temperature consistently results in larger SACI ratios, demonstrating that elevated intake temperatures promote flame-induced auto-ignition and shift the combustion process towards a more volumetric heat release. Impact of Injection Strategy on Autoignition and Combustion Regime 165 <?page no="166"?> Figure 5: Combustion information of the relative air/ fuel ratio variation for two intake temperatures The pressure traces and apparent heat-release rates further illustrate this transition. While the pressure development changes only gradually, the pressure rise rate and the heatrelease curves reveal an increasing contribution of flame-induced auto-ignition with increasing dilution. The SACI combustion process therefore cannot be identified reliably from conventional combustion metrics such as the burn duration alone. The proposed SACI ratio complements these quantities by directly quantifying the relative contribution of flame-induced auto-ignition, allowing the gradual transition from flame propagation to mixed SI/ SACI combustion to be represented by a single physically meaningful parameter. Figure 6: Pressure traces and heat release rate for different relative air/ fuel ratios at T in = 45 °C 166 T. Tietz, M. Günther, S. Pischinger <?page no="167"?> Finally, the proposed SACI ratio provides a quantitative description of this transition. With increasing λ, the calculated contribution of flame-induced auto-ignition continuously increases, allowing the combustion process to be represented by a single characteristic value rather than a qualitative interpretation of the pressure trace. This operating region subsequently serves as the basis for the investigations of the injection strategy. 3.2 Influence of Main-Chamber Injection Timing After identifying the operating range in which flame-assisted compression ignition occurs, the influence of the main-chamber start of injection was investigated. Since the injection timing primarily affects the fuel distribution before ignition, it represents an effective parameter for controlling the interaction between turbulent flame propagation and flame-induced auto-ignition. To better understand the experimentally observed combustion behaviour, the measurements were complemented by three-dimensional CFD simulations of the mixture formation. For both investigated air-fuel ratios, advancing the start of injection reduces the overall combustion duration while the indicated efficiency remains nearly constant. The combustion phasing changes only marginally. Instead, the increasing SACI ratio and the shorter burn duration indicate that earlier injection timings primarily enhance flame-induced auto-ignition rather than the initial flame propagation. In contrast, the increasing SACI ratio and the shorter burn duration indicate that earlier injection timings primarily promote flame-induced auto-ignition during the late combustion phase. Figure 7: Combustion information of the SOI variation for two different relative air/ fuel ratios The CFD simulations explain these observations by revealing significant differences in the mixture fomation prior to ignition. An earlier injection provides more time for spray Impact of Injection Strategy on Autoignition and Combustion Regime 167 <?page no="168"?> evaporation and interaction with the tumble motion, resulting in a considerably more homogeneous fuel-air distribution throughout the main chamber. In contrast, later injection timings preserve locally richer mixture regions and increase the overall stratification. Consequently, earlier SOI increases the fraction of the combustion chamber reaching favourable conditions for flame-induced autoignition, thereby promoting increasingly volumetric SACI combustion. Figure 8: Mixture distribution at 20 °CA bTDC for three different SOI This interpretation is consistent with the experimentally observed combustion characteristics, including the earlier onset of auto-ignition, the increasing SACI ratio, and the reduced burn duration for advanced injection timings. While the overall combustion phasing (MFB 10 , MFB 50 and MFB 90 ) changes only slightly, the continuously increasing SACI ratio demonstrates that an increasing fraction of the released energy originates from flame-induced auto-ignition. Improved mixture homogenisation therefore alters the combustion regime rather than merely accelerating flame propagation. An increasing fraction of the mixture participates in flame-induced autoignition, resulting in a more volumetric heat release and an increasing SACI ratio. Overall, the results demonstrate that the main chamber start of injection is an effective calibration parameter for controlling SACI combustion. By exploiting the available tumble motion during the intake process, earlier injection timings enhance mixture homogenisation and progressively shift the combustion process from flame-propagation-dominated combustion towards a more volumetric, HCCI-like combustion mode. This transition increases the SACI ratio, advances the onset of auto-ignition, and shortens the burn duration while leaving the overall combustion phasing and indicated efficiency nearly unchanged. Consequently, the injection timing changes not only the combustion phasing but the combustion regime itself by modifying the balance between turbulent flame propagation and flame-induced auto-ignition. 3.3 Influence of Pre-Chamber Fuel Quantity Following the investigation of the main-chamber injection timing, the influence of the injected fuel quantity in the pre-chamber was analysed. In contrast to the main-chamber injection, which primarily governs mixture preparation in the cylinder, the pre-chamber fuel quantity directly determines the local mixture composition at the spark plug and 168 T. Tietz, M. Günther, S. Pischinger <?page no="169"?> therefore the combustion process inside the pre-chamber. Since the resulting pressure rise generates the reacting turbulent jets that initiate combustion in the ultra-lean main chamber, the injected fuel quantity represents a key parameter for controlling flameassisted compression ignition. The objective of increasing the injected fuel quantity is not simply to increase the energy released inside the pre-chamber, but rather to establish favourable mixture conditions at the spark plug. Figure 9 illustrates the simulated equivalence-ratio distribution for different pre-chamber injection durations. With increasing DOI, the local mixture around the spark plug gradually approaches stoichiometric conditions. As the local mixture surrounding the spark plug approaches stoichiometric conditions (λ ≈ 1), the laminar flame speed reaches its maximum. This promotes a rapid pressure build-up inside the prechamber and generates highly reactive turbulent jets capable of efficiently initiating flame-assisted auto-ignition in the ultra-lean main chamber. Consequently, the conditions for flame-induced auto-ignition are substantially improved. Figure 9: Mixture distribution in the prechamber at ignition timing for for different SOI The experimental results directly reflect this behaviour. Increasing the pre-chamber injection duration progressively advances the combustion process, shortens the burn duration and increases the maximum pressure-rise rate. At the same time, the indicated efficiency improves as the stronger reacting jets accelerate the transition from conventional flame propagation towards volumetric combustion in the main chamber. The proposed SACI ratio confirms that the contribution of flame-induced auto-ignition increases with increasing pre-chamber fueling, indicating a stronger interaction between the reacting jets and the ultra-lean main-chamber mixture. Impact of Injection Strategy on Autoignition and Combustion Regime 169 <?page no="170"?> Figure 10: General information of the DOI variation A clear optimum is obtained at a pre-chamber injection duration of approximately DOI = 0.7 ms, where the highest indicated efficiency coincides with the shortest combustion duration and the highest pressure-rise rate. Beyond this point, further increases in the injected fuel quantity produce only marginal improvements. Once nearly stoichiometric conditions have been established around the spark plug, the combustion process inside the pre-chamber is no longer limited by mixture reactivity. Consequently, the reacting jets already possess sufficient momentum and chemical activity to initiate stable SACI combustion, while additional fuel mainly enriches the pre-chamber without providing a noticeable benefit. Figure 11: Combustion information of the DOI variation 170 T. Tietz, M. Günther, S. Pischinger <?page no="171"?> Overall, the results demonstrate that the pre-chamber fuel quantity controls the combustion process indirectly through its influence on the local mixture composition inside the pre-chamber. By shifting the mixture towards stoichiometric conditions, the combustion intensity inside the pre-chamber is increased, resulting in stronger reacting jets that effectively trigger flame-assisted auto-ignition in the main chamber. The corresponding increase in the SACI ratio confirms that the improved jet reactivity directly promotes flame-induced autoignition in the main chamber. The optimum pre-chamber fuel quantity is therefore not determined by the maximum possible fuel input, but by the mixture composition required to achieve the highest jet reactivity and the most efficient transition towards SACI combustion. 4 Summary and Outlook A pressure-based methodology for quantifying flame-induced auto-ignition in Spark- Assisted Compression Ignition (SACI) combustion has been presented. By separating the contributions of flame propagation and auto-ignition to the total heat release, the proposed SACI ratio provides a continuous measure of the transition between conventional spark-ignition and SACI combustion. The methodology was applied to experimental investigations on an actively fueled pre-chamber engine to analyse the influence of the global air-fuel ratio, the main-chamber start of injection and the pre-chamber fuel quantity. The experimental investigations demonstrated that the transition from SI to SACI combustion is primarily governed by the global air-fuel ratio. While increasing dilution initially slows down flame propagation, sufficiently lean operating conditions promote flame-induced auto-ignition, resulting in an accelerated late combustion phase. The proposed SACI ratio successfully captures this gradual transition and enables a quantitative comparison of different operating conditions. The proposed SACI ratio proved to be an effective tool for distinguishing changes caused by enhanced flame propagation from those originating from flame-induced autoignition. This enabled a physically meaningful evaluation of different calibration strategies. Earlier injection timings improve mixture homogenisation and promote a more volumetric combustion process, thereby influencing the balance between flame propagation and flame-induced auto-ignition. Finally, the influence of the injected fuel quantity in the pre-chamber was investigated. Increasing the fuel quantity strengthens the reacting jets emerging from the pre-chamber and improves combustion stability. However, only a limited amount of pre-chamber fuel is required to establish stable and efficient SACI combustion, while additional fueling provides only marginal improvements. Overall, the presented methodology enables a physically meaningful quantification of flame-induced auto-ignition and provides a valuable tool for analysing and calibrating SACI combustion systems. Future work will focus on extending the methodology to different fuels and operating conditions as well as applying it to further combustion concepts Impact of Injection Strategy on Autoignition and Combustion Regime 171 <?page no="172"?> 5 Acknowledgement The presented work was carried out within the FVV project 1540 “SACI II - Pre-Chamber Combustion System with Alternative Fuels” in cooperation with the Institute of Automotive Engineering Stuttgart (IFS) at the University Stuttgart under the direction of Univ.- Prof. Dr.-Ing. André Casal Kulzer, the Chair of Thermodynamics of Mobile Energy Conversion Systems at RWTH Aachen under the direction of Univ.-Prof. Dr.-Ing. (USA) Stefan Pischinger and the Institut of Thermal and Fluid Engineering at the University of Applied Science and Arts North Western Switzerland (FHNW) under the direction of Univ.-Prof. Dr. sc. techn. Kai Herrman. The authors gratefully acknowledge the support received by the FVV Research Association for Combustion Engines e. V. and the working group under the supervision of Dr.-Ing. Jonas Villforth (Dr. Ing. h. c. F. Porsche AG). 172 T. Tietz, M. Günther, S. Pischinger <?page no="173"?> Reduced Emissions and Expanded Stoichiometric Operation Through Early Intake Valve Closing, Elevated Charge Motion, and Variable Geometry Turbocharging Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, and Kevin Herrala General Motors Abstract: In anticipation of increasingly stringent emissions regulations, combustion system and boosting upgrades were investigated on a downsized, boosted DISI engine for light duty vehicle applications. The objectives were to expand the stoichiometric operating region, improve emissions and reduce fuel consumption. The engine upgrades included an early intake valve closing (EIVC or Miller Cycle) camshaft, intake port modifications to increase charge motion, and a variable geometry turbocharger (VGT). The cam timing and turbocharger specifications were selected and optimized using 1D CFD, while the combustion system changes were evaluated with 3D CFD. Cold flow simulations quantified charge motion, and combustion models assessed the effects on flame propagation. The final selected hardware package was evaluated experimentally on a steady-state dynamometer. The Miller Cycle intake cam offered two main benefits: reduced knock sensitivity at high loads, and reduced pumping losses at part load. Coupling this with a revised intake port, which generated significantly higher tumble, further improved knock sensitivity at high load. This additional knock margin enabled more advanced combustion phasing, limiting exhaust gas temperatures and reducing the need for component protection enrichment, particularly under mid-speed, high-load conditions. At these conditions, stoichiometric combustion was associated with a substantial improvement in fuel consumption. The increased charge motion also enhanced combustion stability at cold, light load conditions. While the faster burn rates required more spark retard during catalyst heating, the more robust combustion made stability less sensitive to extreme spark retard. The higher charge motion also led to a dramatic reduction in particulate matter emissions at light loads, with the benefit attributed to improved fuel/ air mixing and reduced liquid fuel impingement on combustion chamber surfaces. The full system benefits were achieved by the improved combustion system in conjunction with a VGT. The VGT provided a wide, high-efficiency boost map, which, despite the added restriction from the intake port and EIVC cam timing, enabled high specific power while retaining good low-end performance. Maintaining equal power density, the higher boost demand of this system required increased enthalpy extraction from the exhaust, which further reduced catalyst inlet temperatures, improving emissions robustness through lower catalyst temperatures in mid-speed, high-load operation. <?page no="174"?> 1 Introduction Future light-duty vehicle gasoline engines must satisfy increasingly stringent emissions requirements [1] while preserving customer-facing attributes such as rated power, lowspeed torque, towing performance, durability, and low fuel consumption. For downsized, boosted direct-injection, spark-ignition (DISI) engines, these objectives are strongly coupled. Measures that reduce engine-out emissions can increase boost demand or reduce knock margin, while strategies that maintain high specific output can increase exhaust temperature and may require enrichment to protect turbocharger and aftertreatment components. The primary development objective was to improve emissions while expanding stoichiometric operation and reducing fuel consumption on a downsized, boosted DISI engine intended for light-duty vehicle applications. Enrichment is commonly used to control exhaust gas temperature and protect components at high loads and mid-to-high engine speeds [2]. This enrichment reduces three-way catalyst effectiveness, leading to excess CO emissions, and penalizes fuel consumption, especially under high-load towing conditions. At light-load cold operation, mixture preparation and combustion stability affect both gaseous and particulate emissions, along with the ability to use aggressive spark retard for catalyst heating. Poor combustion stability results in unacceptable noise, vibration, and harshness (NVH) with retarded spark, and increases the risk of misfire, which may result in exceeding cycle emissions requirements. The desired solution therefore required improvements in both high-load and light-load combustion performance at high load, the system would benefit from additional knock margin so that combustion phasing could be advanced, reducing exhaust temperature while reducing the need for component protection enrichment. At light load and cold operating conditions, the system needed faster, more stable combustion so that the combustion system could support extreme spark retard during catalyst light off (CLO) without excessive particulate matter emissions and without exceeding stability limits. This paper describes an integrated combustion-system and boosting strategy developed to achieve the objectives outlined above. The system is based around three primary enablers: early intake valve closing (EIVC), increased intake port-generated charge motion, and a variable geometry turbocharger (VGT). Together, these changes expand the practical stoichiometric operating region, reduce gaseous and particulate emissions, lower fuel consumption in part-load and towing conditions, and reduce catalyst thermal exposure under high-load operation. 2 Integrated Technology Overview: Figure 1 shows an overview of the hardware changes studied. EIVC was used to realize Miller-cycle behavior, reducing effective compression ratio and pumping work while preserving expansion benefits. The reduced effective compression ratio limited knock sensitivity, allowing earlier ignition timing, resulting in lower exhaust gas temperatures. The reduced effective intake volume required increased boost pressure, increasing the enthalpy extracted from the exhaust via the turbocharger, reducing catalyst inlet temperatures. The revised intake port was used to increase tumble and accelerate 174 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala <?page no="175"?> combustion, further limiting knock sensitivity, while also improving combustion efficiency and stability. The VGT was needed to recover boost capability and maintain performance despite the airflow and pressure-drop penalties associated with the EIVC and high-tumble hardware. VGT EIVC cam High tumble port Figure 1: Hardware upgrade summary 2.1 Simulation-Led Technology Selection The hardware package was selected using a simulation-led process, similar to that employed in earlier studies of Miller cycle engines [3]. One-dimensional CFD was used to evaluate cam timing and turbocharger specification options. This work characterized the tradeoff between effective compression ratio, air handling, pumping work, boost demand, turbine efficiency, and exhaust gas temperature. The purpose of the 1D analysis was to identify an EIVC strategy that could reduce knock sensitivity and pumping losses while retaining sufficient power density (when coupled with a VGT) for towing-capable light-duty vehicle applications. Three-dimensional CFD was used to assess combustion-system changes, with emphasis on intake-port-generated charge motion and its effect on mixture preparation and flame propagation. Cold-flow simulations quantified the increase in tumble generated by the revised port. Combustion simulations then assessed whether the higher charge motion produced faster burn rates and even flame propagation. Combustion performance was assessed at CLO, part-load, and full-load conditions. The simulation work led to an integrated package consisting of an EIVC Miller-cycle camshaft, an intake port revised for increased charge motion, and a VGT selected to provide broad boost-system efficiency. This sequence is important: the port and cam modifications improved combustion characteristics but also increased boost and airflowsystem demands. The VGT was required to retain performance despite the increased boost demand. The increased boost also led to increased enthalpy extraction from the exhaust gas, leading to lower catalyst inlet temperatures. Reduced Emissions and Expanded Stoichiometric Operation 175 <?page no="176"?> 2.2 Early Intake Valve Closing/ Miller-Cycle Camshaft An EIVC camshaft reduces effective compression ratio by closing the intake valve earlier in the intake stroke. This Miller-cycle strategy produces two main benefits. First, it reduces knock sensitivity at high load. Second, it reduces pumping losses at low-part load [3,4,5]. In so doing, it also requires increased boost at mid-load conditions. At high load, reduced knock sensitivity enabled more advanced combustion phasing. Advancing combustion phasing improves thermodynamic efficiency and can reduce exhaust gas temperature by extracting more work before exhaust-valve opening [6]. This lower exhaust temperature reduces the need for enrichment to protect the turbocharger [7,8], particularly at mid-speed, high-load conditions. This study has shown that EIVC is not only a fuel-economy feature. In this application, its primary value is tied to reduced exhaust gas temperatures supporting an increased range of stoichiometric operation. By reducing knock and exhaust-temperature constraints at high load conditions, EIVC helps move operating points that would otherwise require enrichment back toward stoichiometric combustion. While not the main objective, there is also a fuel-consumption benefit at these higher-load points. This is primarily attributed to the reduced enrichment demand, with the effect showing up most strongly during high-load operations like towing. At lighter loads, a benefit is also seen from reduced pumping. 2.3 Elevated Charge Motion Through Intake-Port Modification The revised intake port was developed to generate significantly higher tumble. The elevated charge motion improved mixture preparation and accelerated flame propagation [9]. At high load, the faster and more repeatable burn further reduced knock sensitivity when combined with EIVC. This additional knock margin allowed more advanced combustion phasing, reinforcing the exhaust temperature and enrichment-reduction benefits of the Miller-cycle camshaft. Three different flow/ tumble tradeoff values were considered for the intake port, with the selected port design being chosen to balance the combustion benefits of increased charge motion with power requirements. At cold light load, increased tumble improved combustion robustness under catalystheating conditions [10]. While faster burn rates required additional spark retard to maintain sufficient exhaust enthalpy for CLO, the more stable combustion made the system less sensitive to extreme spark retard. The high-tumble port also reduced particulate matter emissions at light load. The benefit is attributed to improved fuel-air mixing and reduced liquid-fuel impingement on combustion-chamber surfaces [11]. 2.4 Variable Geometry Turbocharger Integration The VGT enabled the combustion-system changes to be implemented without sacrificing high specific power or low-end performance. EIVC and the more restrictive high-chargemotion intake port increased the boost demand. A fixed-geometry turbocharger would have required a stronger compromise between low-end response, high-load flow, turbine efficiency, and exhaust-temperature management. The VGT provided a wider operating range with high boosting efficiency which was essential to realizing the full system benefit [8]. 176 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala <?page no="177"?> The VGT also contributed to emissions robustness through exhaust thermal management. Maintaining equal power density with the EIVC and high-tumble system requires increased enthalpy extraction from the exhaust. This additional extraction reduced catalyst inlet temperature, particularly in mid-speed, high-load operation. Lower catalyst inlet temperature reduces thermal exposure and improves robustness for emissions durability while also supporting a broadened region of stoichiometric operation. 3 Experimental Methodology The final hardware package was evaluated experimentally using steady-state dynamometer testing. The test plan was developed to evaluate the system-level effects on: Power-density, exhaust gas temperatures under high-load knock-limited conditions, midspeed high-load stoichiometric capability, part-load fuel consumption, and catalyst heating conditions. Test setup and instrumentation were completed according to GM’s standard dynamometer-based combustion system testing protocol [12,13,14] 4 Results and Discussion 4.1 VGT-Enabled Performance In addition to the baseline combustion system, four different intake port designs, with varying flow and tumble characteristics, were investigated. The selected port offered a mid-level tumble increase among the options developed. Along with the tumble increase, however, the port was accompanied by decreased flow capacity. Figure 2 shows the coefficient of discharge as a function of bore area plotted against the tumble index. In addition to the port designs for this engine, a number of competitor intake ports are plotted, showing the latest designs to be at the high end of tumble values. Baseline Selected port CdBAcorr (intake port ow) Tumble Index CdBA values corrected by bore diameter Figure 2: Port flow vs. tumble index - investigated port designs shown as squares Reduced Emissions and Expanded Stoichiometric Operation 177 <?page no="178"?> Despite the additional restriction associated with the EIVC cam timing and high-tumble intake port, the VGT was able to offer increased boost, allowing the system to retain its high specific power along with good low-end performance. Figure 3 shows the increased compressor outlet pressure from the new system Compressor Outlet Pressure (relative) Engine Speed (low to high) Full Load Compressor Outlet Pressure New System Baseline Figure 3: Compressor outlet increase with VGT and improved combustion 4.2 High-Load Knock Sensitivity and Combustion Phasing The Miller-cycle camshaft reduced knock sensitivity at high load by lowering effective compression ratio. The increased tumble intake port further improved knock tolerance by accelerating combustion, thereby reducing the end-gas time at high temperatures/ pressures. Together, these effects created additional spark timing authority relative to the baseline combustion system. Taking advantage of these improvements allowed the fullload capability to increase as shown in Figure 4. Brake Power (relative) Engine Speed (low to high) Brake Power New System Baseline Figure 4: Increased power with revised combustion system 178 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala <?page no="179"?> The improved efficiency and reduced exhaust gas temperature showed similar benefits at high-load towing conditions, where reduced exhaust temperatures and reduced enrichment were realized. This has the practical implication of improving fuel consumption under towing conditions. Figure 5 (1D simulation anchored to dyno data) shows the reduction in turbine outlet temperature associated with the integrated system. Further, four different tumble values were evaluated, demonstrating improved temperature reduction with higher tumble. The mid-level high-tumble system was chosen for implementation as it maximized power while reducing temperatures. The figure shows that the reduction in turbine outlet temperature persists even at high ambient temperature conditions (in this case load decreases as the charge air cooler heat rejection was held constant, which accounts for the overall decreasing temperature trend) Turbine outlet temperature Ambient Temperature (low to high) Turbine Outlet Temperature vs. Ambient Temperature L4B VGT/ EIVC with c/ o port High Tumble port #1 New system Highest tumble port E ect of EIVC + VGT E ect of Increased tumble Increased tumble Figure 5: Turbine outlet temperature reductions at towing conditions 4.3 Stoichiometric Operation and Fuel Consumption The improved knock margin, and lower exhaust-temperature at boosted conditions enabled a broader stoichiometric operating region, particularly under mid-speed, highload conditions. At these points, avoiding enrichment produced a substantial fuelconsumption benefit. This high-load fuel consumption benefit would not translate directly to higher fuel economy ratings but would instead be noticed under high load manoeuvres like towing. The BSFC map shown below in Figure 6 was estimated using a 1D simulation anchored to steady-state dyno data. Relative to the baseline engine, the colour shading indicates decreased fuel consumption in blue and increased fuel consumption in red. It should be noted that fuel consumption improved over the whole map, but the improvement was particularly notable at high loads. This is primarily due to reduced enrichment. Comparing the baseline and new system stoichiometric limit lines shows the significant reduction in enrichment enabled by this system, without sacrificing any performance. Reduced Emissions and Expanded Stoichiometric Operation 179 <?page no="180"?> New system lower BSFC Baseline lowerBSFC Figure 6: BSFC contour map, with stoich limits noted At lighter loads, more relevant to fuel economy cycles, a benefit was observed primarily due to pumping reductions, but also to reduced knock in the mid-load, lightly boosted region of the map. Using an aggregate calculation scaled to ten steady-state points relevant to the FTP fuel economy cycle, a reduction in fuel consumption of approximately 1.5 % was estimated. 4.4 Cold Light-Load Combustion Stability At cold light load, such as CLO conditions, combustion stability, as measured by COV, along with hydrocarbon and Nox emissions are critical performance characteristics [15]. The elevated charge motion built into the new system increased burn rates, and improved combustion stability, but required more spark retard to achieve equivalent exhaust enthalpy. The net result was marginally improved combustion stability (as demonstrated by the COV in Figure 7), but with lower emissions (Figure 8). Notably, the robust combustion made stability (COV) less sensitive to severe spark retard, retaining acceptable COV even at extremely late spark timing. This enabled more aggressive catalyst-heating operation without the same combustion-stability penalty seen in the baseline engine. 180 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala <?page no="181"?> Worst_COVof IMEP Enthalpy Flow Park Idle Baseline New System Figure 7: Combustion stability as determined by worst-cylinder COV NOx+HC (relative) Enthalpy Flow (relative) Park Idle Baseline New System Figure 8: NO x + HC emissions 4.5 Particulate Matter Reduction Poor fuel-air mixing and liquid impingement on combustion chamber/ piston/ cylinder surfaces have been identified as major sources of particulate matter under CLO conditions [11]. Despite the lower charge motion inherent with an EIVC cam [3], implementing the high tumble port increased charge motion, resulting in improved mixing and accelerated burn rates, requiring later spark (more time for mixing/ evaporation), which may also help to reduce liquid fuel films. As shown in Figure 9, this resulted in a significant reduction in particulate matter emissions (here measured as particle number), with the improvement increasing as spark is further retarded. In the calibrated CLO space, the reduction is 75-80 %. Reduced Emissions and Expanded Stoichiometric Operation 181 <?page no="182"?> Particle Number Rate Exhaust Enthalpy Rate PN Emissions vs. Enthalpy Baseline New System 4-5x reduction Figure 9: CLO particle number emissions 5 Conclusions A combustion-system and boosting package consisting of EIVC Miller-cycle valve timing, elevated intake-generated charge motion, and a variable geometry turbocharger was developed for a downsized, boosted DISI light-duty truck engine. The integrated system expanded stoichiometric operating capability while improving emissions and fuel consumption. The EIVC camshaft reduced high-load knock sensitivity and part-load pumping losses. The revised high-tumble intake port further reduced knock sensitivity, accelerated flame propagation, improved cold light-load combustion stability, and reduced particulate emissions through improved mixture preparation. The VGT provided boost-system efficiency and operating-range flexibility required to maintain high specific power and low-end performance despite the added restrictions associated with an EIVC intake cam and a high-charge motion intake port. The combined hardware package reduced the need for component protection enrichment under mid-speed, high-load operation. Additional knock margin allowed more advanced combustion phasing, which lowered exhaust gas temperature and supported stoichiometric combustion. At the same time, increased exhaust enthalpy extraction through the VGT reduced catalyst inlet temperature, improving emissions aftertreatment system robustness. The work demonstrates that future emissions and fuel-consumption objectives can be addressed through integrated combustion-system and boosting development rather than through aftertreatment escalation alone. The benefits arise from the interaction of Millercycle valve timing, charge-motion-enhanced combustion, and variable-geometry boosting, and should therefore be evaluated as a complete system. 182 Justin Ketterer, Andrew Aerts, Hugh Miller, Sharon Li, Muniappan Anbarasu, Kevin Herrala <?page no="183"?> 6 References [1] EPA, U. S., Multi-Pollutant Emissions Standards for Model Years 2027 and Later Light-Duty and Medium-Duty Vehicles. 89 Fed. Reg. 27,842 (2024). [2] Nose, H. . I. T. . K. S. . S. A. . e. a., Fuel Enrichment Control System by Catalyst Temperature Estimation to Enable Frequent Stoichiometric Operation at High Engine Speed/ Load Condition, presented at SAE 2013 World Congress & Exhibition, Detroit, MI, 2013 (unpublished). [3] Ketterer, J., Gautier, E. & Keating, E., The Development and Evaluation of Robust Combustion Systems for Miller Cycle Engines, presented at SAE Technical Paper 2018-01-1416, Detroit, MI, 2018 (unpublished). [4] Zheng, B., Yin, T. & Li, T., Analysis of Thermal Efficiency Improvement of a Highly Boosted, High Compression Ratio, Direct-Injection Gasoline Engine with LIVC and EIVC at Partial and Full Loads, presented at JSAE/ SAE International Powertrains, Fuels and Lubricants Meeting, 2015 (unpublished). [5] Ouyang, X. et al., A Comparative Study on Influence of EIVC and LIVC on Fuel Economy of A TGDI Engine Part I: Friction Torques of Intake Cams with Different Profiles and Lifts, presented at SAE International Powertrains, Fuels & Lubricants Meeting, 2017 (unpublished). [6] Ayala, F. A., Gerty, M. D. & Heywood, J. B., Effects of Combustion Phasing, Relative Air-fuel Ratio, Compression Ratio, and Load on SI Engine Efficiency, presented at SAE World Congress, Detroit, MI, 2006 (unpublished). [7] R. Welzmuller, M. W. T. S. M. K. M. K., The new 6-cylinder engine from BMW, presented at 31st Aachen Colloquium Sustainable Mobility 2022, Aachen, 2022 (unpublished). [8] P. Scheller, D. S. D. W. a. T. T., The New EA211 1.5L TSI evo2 84 (1), 16-23 (2023). [9] Ruhland, H. . L. T. . D. J. . B. A. . e. a. ., A Study on Charge Motion Requirements for a Class-Leading GTDI Engine, presented at 13th International Conference on Engines & Vehicles, Capri, Italy, 2017 (unpublished). [10] Lee, D. . a. H. J., Effects of Charge Motion Control During Cold Start of SI Engines, presented at Powertrain & Fluid Systems Conference and Exhibition, Toronto, ON, 2016 (unpublished). [11] Ketterer, J. a. C. W., On the Nature of Particulate Emissions from DISI Engines at Cold-Fast-Idle. SAE Int. J. Engines 7 (2), 986-994 (2014). [12] Halsall, S., Luchansky, K., Zeng, Y., Davis, R. & Herrin, R., Development of the Combustion System for the General Motors Fifth Generation “Small Block” Engine Family, presented at SAE World Congress, Detroit, MI, 2013 (unpublished). [13] Davis, R. S. & Patterson, G. J., Cylinder Pressure Data Quality Checks and Procedures to, presented at SAE World Congress, Detroit, MI, 2006 (unpublished). [14] Patterson, G. J. & Davis, R. S., Geometric and Topological Considerations to Maximize Remotely Mounted Cylinder Pressure Transducer Data Quality. SAE Int. J. Engines 2 (1), 414-420 (2009). [15] Cedrone, K. & Cheng, W. K., SI Engine Control in the Cold-Fast-Idle Period for Low HC Emissions and Fast Catalyst Light Off. SAE Int. J. Engines 7 (2), 968-976 (2014). Reduced Emissions and Expanded Stoichiometric Operation 183 <?page no="185"?> 3D ignition modelling in alternative-fuel engines: passive pre-chamber design and cycle-to-cycle variability Lorenzo Sforza, Alessandro Nodi, Tommaso Lucchini, Gianluca D’Errico 1 , William Northrop 2 , Haogeng Bai, Hongchao Chu 3 Abstract: The use of alternative fuels in spark-ignition engines requires dedicated optimization strategies, strongly dependent on fuel properties. In this context, reliable and computationally efficient 3D numerical models effectively support experimental activities, enabling a deeper understanding of thermo-fluid dynamic processes in production engines. This work addresses two key aspects of the ignition process in spark-ignition engines: the use of passive pre-chambers in neat ammonia engines, and the adoption of ultra-lean mixtures in engines fuelled with synthetic fuels or hydrogen. In the first case, turbulent jet ignition (TJI) based on a passive pre-chamber represents a simple and cost-effective solution to compensate for the high ignition energy and low laminar flame speed of ammonia. The analysis focuses on key design parameters - prechamber volume, number of nozzles, and nozzle diameter - and their impact on combustion and indicated efficiency. For each configuration, the numerical approach is validated against experimental in-cylinder pressure and heat release rate. Results show that the selected design promotes an organized tumble motion, limiting residual gas accumulation near the ignition region and enhancing flame propagation towards the nozzles. This leads to improved efficiency and reduced cycle-to-cycle variability (CCV) compared to conventional spark ignition. A reduction in the nozzle area-to-pre-chamber volume ratio (A noz / V pc ) further increases efficiency and combustion stability, at the cost of higher thermal losses, particularly from pre-chamber and piston surfaces. In the second case, ultra-lean combustion is investigated as a solution to maximize efficiency in synthetic-fuel engines and for abating NO x emissions in hydrogen engines. Considering that mixture ignitability represents a design limit, a RANS-based model is developed to predict the impact of the ignition system on CCV. The spark discharge is modelled using Lagrangian particles, whose velocity is initially perturbed through a divergence-free, spatially correlated stochastic field and then evolved using the Simplified Langevin Model (SLM) to account for stochastic interactions with the flow. The approach is first assessed against DNS simulations in a simplified geometry, where 1 ICE Group, Department of Energy, Politecnico di Milano, Italy 2 T. E. Murphy Engine Research Laboratory, Department of Mechanical Engineering, University of Minnesota, USA 3 Institute for Combustion Technology, RWTH Aachen University, Germany <?page no="186"?> an ultra-lean hydrogen-air mixture is ignited between two wire electrodes under high cross-flow. Finally, the model is validated in an engine configuration by comparing two different ignition systems. Keywords: alternative-fuel engines; spark ignition; turbulent jet ignition; passive prechamber; neat ammonia; ultra-lean combustion; synthetic gasoline; hydrogen; cycle-tocycle variability; 3D-CFD; RANS. 1 Introduction and motivation The decarbonization of the energy and transportation sectors is driving a growing interest in carbon-free and carbon-neutral fuels suitable for large-scale deployment in internal combustion engines. Because the combustion behaviour of these fuels differs markedly from that of conventional hydrocarbons, their use in spark-ignition (SI) engines requires dedicated optimization strategies, strongly dependent on fuel properties. In this context, reliable and computationally efficient 3D numerical models effectively support experimental activities, enabling a deeper understanding of the thermo-fluid dynamic processes taking place in production engines. This work addresses two key aspects of the ignition process in SI engines: the use of passive pre-chambers in neat ammonia engines, and the adoption of ultra-lean mixtures in engines fuelled with synthetic fuels or hydrogen. Ammonia (NH 3 ) is an attractive carbon-free energy carrier, thanks to its established production and distribution infrastructure and its favourable storage characteristics, and it avoids the cryogenic storage and extreme pressures required for hydrogen. However, its low laminar flame speed, high ignition energy and narrow flammability limits make stable and efficient combustion difficult under typical engine conditions, so most studies rely on dual-fuel strategies. Since the burning velocity of NH 3 responds favourably to increased turbulence, turbulent jet ignition (TJI) with a passive pre-chamber - which provides multiple ignition sites and strong local turbulence without any auxiliary fuel injection - is a simple and cost-effective route to stable neat-NH 3 operation, provided the pre-chamber geometry is properly tailored. For engines fuelled with synthetic (e-)fuels, ultra-lean operation is one of the most effective levers to maximize thermal efficiency, while in hydrogen engines lean mixtures are essential to abate NO x emissions. Drop-in synthetic gasolines, produced from renewable energy and captured CO 2 , are of particular interest as carbon-neutral fuels compatible with the existing SI engine architecture; iso-octane is commonly adopted as a representative surrogate for this fuel class. Under ultra-lean conditions, however, mixture ignitability becomes a design limit and the ignition system strongly affects the cycle-to-cycle variability (CCV) of combustion. Predicting this effect calls for an ignition model able to reproduce the stochastic nature of the early flame kernel within an affordable RANS framework. 186 L. Sforza, A. Nodi, T. Lucchini, G. D’Errico, W. Northrop, H. Bai, H. Chu <?page no="187"?> 2 Numerical methodology 2.1 Turbulent combustion model All simulations are carried out with Lib-ICE, an OpenFOAM-based suite for internal combustion engine simulations, within a URANS framework using the standard k-ε turbulence model. The flame front propagation is described through the transport equation of the unburnt gas mass fraction (regress variable b); the reaction rate is modelled with the Flame Area Evolution (FAE) approach (Weller et al., 1998), in which the flame wrinkling due to turbulence is accounted for through the wrinkle factor Ξ = S t / S u . The laminar-toturbulent transition after ignition follows the Herweg-Maly formulation (Herweg & Maly, 1992), later refined by Sforza et al. (2023), evolving Ξ from unity to its equilibrium value, modelled according to Gülder (1991). The laminar flame speed S u is retrieved from look-up tables generated with Cantera over the full range of engine-relevant conditions, adopting a detailed kinetic mechanism for NH₃ oxidation (Stagni et al., 2020). A multi-mesh movinggrid strategy captures the full engine cycle, generating and deforming successive meshes with conservative field mapping, while crank-angle-resolved pressure and temperature boundary conditions are provided by a 1D engine model. 2.2 Ignition modelling Two ignition sub-models are used, according to the objective of each case study. For the passive pre-chamber (Case 1), a simplified energy-deposition model initializes the regressvariable distribution to trigger flame propagation, since the focus is on the geometrydriven turbulent combustion rather than on the discharge itself. For the ultra-lean case (Case 2), an advanced spark-ignition model is adopted to predict the impact of the ignition system on CCV within a RANS simulation. It follows a multiphysics, coupled Lagrangian-Eulerian approach: the spark channel is discretized into Lagrangian particles introduced along the spark-gap centerline, while the turbulent flame is treated on the Eulerian grid. At each discharge (ignition and any restrike), a near-DNS particle velocity is reconstructed as the sum of the mean RANS flow velocity and a perturbation that is stochastic but spatially correlated among the channel particles and divergence-free; the perturbation is generated through a Wiener process using an Ornstein-Uhlenbeck equation discretized as a first-order autoregressive model. Each particle velocity is then evolved with the Simplified Langevin Model (SLM), which retains the memory of previous perturbations and represents the stochastic diffusion caused by the interaction with the surrounding mixture. The local channel diameter is evolved from mass and energy conservation, accounting for the energy released after breakdown, the support of the electrical circuit and the early laminar-flame stage; the secondary side of an inductive ignition system is modelled to provide the delivered power, and a restrike model reproduces channel shortenings. Finally, the energy delivered to the gas - net of electrode heat losses - is coupled to the combustion model through the ignition source term of the b equation. 3D ignition modelling in alternative-fuel engines 187 <?page no="188"?> 3 Case 1 - Passive pre-chamber in a neat ammonia engine 3.1 Engine and pre-chamber configurations Experiments were carried out on a Cummins 6.7 L inline six-cylinder medium-duty engine, modified to operate as a single-cylinder research unit fuelled with neat, vaporized NH 3 . The investigated operating point (1800 rpm, 75 % load) is representative of stationary power generation for a generator set on a 60 Hz grid. The stock configuration was converted to conventional SI by replacing the injector with a spark plug and then to TJI by installing a passive pre-chamber with a modular nozzle-cap design, allowing an independent variation of pre-chamber volume, nozzle number and nozzle diameter. The main engine data are summarized in Table 1. Parameter Value Active cylinders [-] 1 Displacement [cm³] 1115 Bore [mm] 107 Stroke [mm] 124 Compression ratio [-] 18 V pc / V cyl ,TDC [%] 5, 10 Pre-chamber nozzles [-] 6, 8 Nozzle diameter [mm] 1.6, 1.9 Fuel [-] Vaporized NH 3 Equivalence ratio ϕ [-] 1.025 Engine speed [rpm] 1800 Engine load [%] 75 Intake pressure [bar] 1.3 Intake temperature [°C] 80 Tab. 1: Engine geometrical data and operating condition (Case 1). 3.2 Baseline pre-chamber vs conventional spark ignition For the baseline pre-chamber (PC5V-N8-1.9mm), the filling process during compression establishes a pronounced tumble motion inside the pre-chamber (Fig. 1 (a)), driven by the non-uniform mass flow through intakeand exhaust-facing nozzles. This flow organizes the exhaust-gas-residual (EGR) stratification (Fig. 1(b)) and, although ignition occurs in a relatively diluted, low-reactivity region, the high turbulence generated during filling raises the turbulent flame speed to 20-30 times the laminar value (Fig. 1(c) and (d)), effectively compensating for the intrinsically slow NH 3 kinetics. The combustion in the main chamber then develops in three phases: a pre-jet expulsion of unburnt mixture, a jetdriven phase with multiple reacting jets and distributed ignition sites, and a postimpingement phase where the merged turbulent front interacts with the piston bowl. 188 L. Sforza, A. Nodi, T. Lucchini, G. D’Errico, W. Northrop, H. Bai, H. Chu <?page no="189"?> Fig. 1: Pre-chamber distributions at spark timing for the baseline PC5V-N8-1.9 mm: (a) velocity magnitude, (b) EGR, (c) laminar flame speed S u , (d) turbulent flame speed S t . Compared with conventional SI, the baseline TJI configuration shortens the combustion duration by about 45 %, reduces the cycle-to-cycle variability (CoV of IMEP from 3.8 % to 1.1 %) and increases the indicated efficiency from 36.7 % to 39.9 % (Fig. 2). A combustionregime analysis on the Borghi-Peters diagram shows that NH 3 attains thickened-wrinkled regimes comparable to those of high-reactivity hydrocarbon fuels, but at markedly lower turbulence intensity. Fig. 2: Numerical-experimental average-cycle pressure traces (cylinder and pre-chamber) and incylinder AHRR for conventional SI (SA = 46 °bTDC) and baseline TJI (SA = 14 °bTDC). Grey lines: 300 single-cycle measurements. 3.3 Effect of the nozzle-area-to-volume ratio and heat losses Starting from the baseline, three isolated geometric variations were considered - larger pre-chamber volume, fewer nozzles and smaller nozzle diameter - each reducing the nozzle area-to-pre-chamber volume ratio (A noz / V pc ). The measured engine performance is summarized in Table 2. A consistent trend emerges: reducing A noz / V pc increases the pressure difference driving the jets, hence jet momentum and turbulence in the main chamber, accelerating combustion. Increasing the pre-chamber volume gives the largest benefit, reducing combustion duration by more than 65 % with respect to SI while 3D ignition modelling in alternative-fuel engines 189 <?page no="190"?> achieving CoV of IMEP below 1 % and indicated efficiency above 40 %; reducing the nozzle number yields similar but weaker gains, whereas reducing the nozzle diameter also improves performance but pushes the flame towards the model validity limits, where nozzle flame quenching through the nozzles and re-ignition in main chamber - not represented in the model - become likely. Configuration A noz V pc [mm²/ mL] SA [°bTDC] IMEP [bar] η i [%] CoV IMEP [%] p max [bar] CoV P max [%] SI - 46 8.7 36.7 3.8 101.4 10.0 PC5V-N8-1.9 mm 6.7 14 8.8 39.9 1.1 83.4 7.9 PC10V-N8-1.9 mm 3.5 9 8.6 40.5 0.6 95.5 4.4 PC5V-N6-1.9 mm 4.7 12.5 8.8 40.4 0.5 88.9 5.5 PC5V-N8-1.6 mm 4.7 11.5 8.9 40.3 0.5 89.5 6.9 Tab. 2: Measured engine performance over 300 cycles for conventional SI and the four passive prechamber configurations. The efficiency gains within the TJI design space remain moderate (+ 0.4-0.6 percentage points over the baseline pre-chamber) because they are partly offset by wall heat losses, which grow monotonically as A noz / V pc decreases (from + 3 % for the baseline up to + 23 % for the largest pre-chamber, relative to SI), driven mainly by the pre-chamber walls and by jet impingement on the piston (Fig. 3). Pre-chamber optimization for neat NH 3 therefore requires a combined consideration of geometry, flow organization and flame stability, rather than a single governing parameter. Fig. 3: Total heat losses over the engine cycle (normalized to combustion energy) for conventional SI and the four pre-chamber configurations, decomposed by chamber surface. 190 L. Sforza, A. Nodi, T. Lucchini, G. D’Errico, W. Northrop, H. Bai, H. Chu <?page no="191"?> 4 Case 2 - Ultra-lean combustion in synthetic-fuel and hydrogen engines 4.1 Case setup The advanced ignition model was assessed on a pent-roof, high-tumble single-cylinder SI engine operated with an ultra-lean premixed charge (equivalence ratio ϕ = 0.58), using isooctane as a surrogate for synthetic gasoline. The full power cycle is computed within the same URANS/ k-ε framework and with the same multi-mesh moving-grid strategy adopted for Case 1 - with mesh deformation over the engine cycle - so that a realistic flow field is obtained at spark timing. At fixed engine and mixture conditions, two ignition systems were compared: an enhanced coil (EC), able to sustain a higher secondary current for a user-defined duration, and a standard coil (SC) with the usual triangular current profile (Table 3). For each system, an ensemble of independent single-cycle discharges was simulated to reconstruct the combustion CCV. Parameter Value Engine type pent-roof, high-tumble single-cylinder SI Compression ratio [-] 12 Stroke-to-bore ratio [-] 1.3 Fuel iso-octane-air (synthetic-gasoline surrogate) Equivalence ratio ϕ [-] 0.58 Spark advance [°CA bTDC] 29 Enhanced coil (EC) i S,0 = 120 mA; Δt = 1 ms Standard coil (SC) i S,0 = 80 mA; Δt = 0 ms Tab. 3: Engine, mixture and ignition-system parameters (Case 2). 4.2 Impact of the ignition system on cycle-to-cycle variability The preliminary results confirm that the ignition system markedly influences combustion stability under ultra-lean conditions. The enhanced coil sustains a more repeatable early flame kernel and, consequently, a narrower cycle-to-cycle distribution of the in-cylinder pressure, whereas the standard coil produces a wider spread of pressure and heat-release traces - i. e. a higher CCV - and a lower average IMEP (Fig. 4). 3D ignition modelling in alternative-fuel engines 191 <?page no="192"?> Fig. 4: Cylinder pressure (top) and secondary current with in-cylinder AHRR (bottom) for the enhanced (EC) and standard (SC) coil: experimental mean, max and min (black) and computed single cycles (red for EC, blue for SC). This behaviour originates from the different discharge waveforms of the two systems: the enhanced coil delivers a higher secondary current over a longer time (a plateau sustained for about 1 ms), maintaining higher inter-electrode voltages and thus a longer, more energetic spark channel; the standard coil, by contrast, discharges through a shorter, triangular pulse that extinguishes the channel about twice as early. Figure 5 shows a quantitative comparison between the measured IMEP and its coefficient of variation and the corresponding computed values, as a function of the number of single-cycle discharges considered. The proposed approach captures the experimentally observed trend of a lower CCV when the enhanced coil is used in place of the standard one, even with the few numerical cycles considered; the absolute IMEP is, however, overpredicted for both systems - an aspect targeted by the ongoing model update. 192 L. Sforza, A. Nodi, T. Lucchini, G. D’Errico, W. Northrop, H. Bai, H. Chu <?page no="193"?> Fig. 5: IMEP (left) and its coefficient of variation (right) vs number of single-cycle discharges, CFD vs experiment, for the enhanced (EC, red) and standard (SC, blue) coil. 5 Conclusions Two complementary aspects of the ignition process in alternative-fuel SI engines were investigated with reliable and computationally efficient 3D-CFD models. For the neatammonia engine, a passive pre-chamber enables stable and efficient TJI combustion without any dual-fuel support: relative to conventional SI, the combustion duration is reduced by up to 65 %, the CoV of IMEP falls below 1 % and the indicated efficiency exceeds 40 %. The nozzle area-to-pre-chamber volume ratio is the key governing parameter - lower values increase jet momentum and turbulence and thus efficiency - but the benefit is bounded by wall heat losses and by the specific geometric pathway used to reduce it. For ultra-lean synthetic-fuel and hydrogen engines, the RANS-based Lagrangian-Eulerian spark model reproduces the impact of the ignition system on cycle-to-cycle variability, correctly capturing the lower CCV achieved with an enhanced coil and supporting the design of ignition strategies at the ignitability limit. References Sforza, L., Lucchini, T., D’Errico, G., Gianetti, G., Beatrice, C., & Tunestål, P. (2023). A Numerical Methodology for the Design of Active Prechambers in Spark-Ignition Engines. Combustion Science and Technology. https: / / doi.org/ 10.1080/ 00102202.2023.2219450 Weller, H. G., Tabor, G., Gosman, A. D., & Fureby, C. (1998). Application of a flame-wrinkling LES combustion model to a turbulent mixing layer. Symposium (International) on Combustion, 27(1), 899-907. Herweg, R., & Maly, R. (1992). A fundamental model for flame kernel formation in SI engines. SAE Transactions, 1947-1976. Stagni, A., Cavallotti, C., Arunthanayothin, S., Song, Y., Herbinet, O., Battin-Leclerc, F., & Faravelli, T. (2020). An experimental, theoretical and kinetic modeling study of the gas-phase oxidation of ammonia. Reaction Chemistry & Engineering, 5, 696-711. 3D ignition modelling in alternative-fuel engines 193 <?page no="194"?> Abbreviations AHRR apparent heat release rate CAD crank angle degree CCV cycle-to-cycle variability CFD computational fluid-dynamics CR compression ratio EC/ SC enhanced coil/ standard coil EGR exhaust gas residuals FAE flame area evolution IMEP indicated mean effective pressure NO x nitrogen oxides PC pre-chamber RANS Reynolds-averaged Navier-Stokes SA spark advance SI spark ignition SLM Simplified Langevin Model S u / S t laminar/ turbulent flame speed TDC top dead centre TJI turbulent jet ignition 194 L. Sforza, A. Nodi, T. Lucchini, G. D’Errico, W. Northrop, H. Bai, H. Chu <?page no="195"?> Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle Lukas Kniestedt Abstract: The Argon Power Cycle (APC) represents a promising technology for future energy systems. It enables the emission-free reconversion of hydrogen into electrical energy within a closed argon-oxygen cycle. The high isentropic exponent of the working gas increases the thermodynamic efficiency potential but also promotes the occurrence of knock and auto-ignition. In engine test bench investigations, the isentropic exponent, compression ratio, and system pressure were evaluated as key parameters for influencing the combustion behavior. The results demonstrate that stable spark-ignited operation of the APC can be achieved through a targeted adjustment of the thermodynamic boundary conditions. 1 Introduction Achieving the climate targets of the Paris Agreement and limiting global warming to a maximum of 1.5 °C require enormous technological and societal efforts and therefore represent one of the greatest challenges of our time. In particular, a substantial reduction in carbon dioxide emissions is necessary to meet these goals. A significant contribution can be made by the energy sector, whose emissions can be considerably reduced through the increasing deployment of renewable energy sources. However, renewable energy systems provide only limited baseload capability. In particular, wind and solar power generation are subject to daily and seasonal fluctuations. As the share of renewable energy continues to increase, situations with either surplus electricity generation or insufficient power supply will occur more frequently. To maintain grid stability, suitable solutions for storing excess electrical energy must therefore be developed. For the long-term storage of large amounts of energy, battery systems are currently of limited suitability. A promising alternative is the conversion of electrical energy into hydrogen by electrolysis and its subsequent storage. When required, the stored hydrogen can be reconverted into electricity, for example by using fuel cells or internal combustion engines. The application of hydrogen combustion engines is currently limited by their lower efficiency compared to fuel cells and by the formation of nitrogen oxide emissions. By replacing air as the working medium with an inert gas-oxygen mixture, however, nitrogen <?page no="196"?> oxide emissions can be completely avoided. Furthermore, the careful selection of the inert gas offers the potential to increase the thermal efficiency of the combustion process. A promising implementation of this concept is the Argon Power Cycle (APC). In this system, argon is used as the inert working gas and also gives the concept its name. Replacing air with argon significantly changes the thermodynamic conditions inside the combustion chamber. In particular, the temperatures and pressures occurring during compression and combustion increase considerably. While this offers the potential for improved efficiency, it also increases the tendency for combustion anomalies such as preignition and knock. The effects of these altered boundary conditions on combustion stability and achievable engine performance therefore require further investigation. The objective of the present study is the experimental investigation of the Argon Power Cycle fueled with hydrogen. Both direct injection into the combustion chamber and port fuel injection are considered. By varying different operating parameters on an engine test bench, their influence on efficiency, combustion stability, and load capability is analyzed. Particular attention is given to the occurrence of knock and its impact on the operation of the Argon Power Cycle. 2 The Argon Power Cycle The Argon Power Cycle (APC) is based on a conventional internal combustion engine. Instead of air, an argon-oxygen mixture is used as the working medium. If the exhaust gas were discharged to the environment after combustion, argon consumption would become so high that economical operation of the system would no longer be feasible. Therefore, the exhaust gas is recirculated to the intake side within a closed loop. During the combustion of hydrogen with oxygen, water is the only reaction product formed. Argon is inert with respect to both substances and does not participate in the chemical reaction. To prevent the accumulation of water within the cycle, it must be continuously removed from the system. For this purpose, the exhaust gas is cooled after combustion, allowing the water vapor to condense and be separated. After water removal, the working gas mainly consists of argon and excess oxygen, which are subsequently returned to the engine. Since argon does not react with either hydrogen or oxygen, no undesired reaction products are formed. In contrast to conventional air-breathing engines, the working gas contains no nitrogen, eliminating the possibility of nitrogen oxide formation. As a result, the APC operates with zero local emissions. In addition to this advantage, argon also offers the potential to increase the thermal efficiency of the combustion process. The theoretical efficiency of idealized combustion processes is primarily influenced by the compression ratio and the isentropic exponent of the working gas. This relationship can be derived from the following equation for the constant-volume cycle. � theo ¼ 1 � 1 " �� 1 196 Lukas Kniestedt <?page no="197"?> Figure 1: theoretical efficiency constant volume cycle As a reference, air is considered, which mainly consists of the diatomic gases nitrogen and oxygen and has an isentropic exponent of approximately κ = 1.4. A high isentropic exponent is beneficial for increasing thermal efficiency. The monatomic noble gases helium, neon, and argon exhibit the maximum value of κ = 1.67. For application in a closed-cycle engine, argon offers additional practical advantages. It is relatively abundant in the atmosphere with a concentration of nearly one percent and is already widely used as a shielding gas in welding applications. As a result, argon can be provided at reasonable cost. Figure 1 illustrates the influence of the compression ratio and the isentropic exponent on the theoretical efficiency. A schematic representation of the Argon Power Cycle is shown in Figure 2. The closed loop connecting the exhaust and intake sides of the engine can be clearly identified. The exhaust gas cooling system for water separation is also included. By replacing air as the working medium, the oxygen required for combustion is no longer inherently available and must be supplied separately. This is achieved with a gas mixer. Hydrogen and oxygen are provided by an electrolysis system, which generates both gases in the proportions required for the combustion process. The APC therefore enables the direct utilization of electrolysis products for the reconversion of renewable energy into electricity. Since hydrogen, oxygen, and argon must be supplied and stored for operation, the application of the system in mobile sectors is associated with increased complexity. In stationary energy applications, however, the additional space requirement is of minor importance. The APC combines the emission-free reconversion of hydrogen with the advantages of internal combustion engines, including high power density, robustness, and the potential for high efficiencies. It therefore represents an attractive alternative to other hydrogen reconversion technologies. Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle 197 <?page no="198"?> Figure 2: scheme of the Argon Power Cycle The feasibility of the Argon Power Cycle has already been demonstrated at WTZ Roßlau. [1] Initial investigations were carried out using high-pressure direct hydrogen injection in combination with compression ignition. Building on these results, subsequent research focused on the application of low-pressure hydrogen injection and spark-ignited combustion concepts. [2] 3 Experimental Setup The investigations were conducted on a research engine test bench at WTZ Roßlau. A Ford 1.0 L EcoBoost (B7DA) engine was used as the test platform, with its main engine specifications summarized in Table 1. Parameter Value Stroke 82 mm Bore 71,9 mm Number of Cylinders 3 Displacement 1,0 l Compression Ratio 10,0 Valvetrain DOHC, variable Valves per Cylinder 4 Fuel Hydrogen Port Fuel Injection Pressure 4,5 bar Direct Injection Pressure 11,5 bar Table 1: Main specifications of the Ford 1.0 L EcoBoost (B7DA) 198 Lukas Kniestedt <?page no="199"?> For the experiments, the production engine was converted into a hydrogen-fueled closedcycle engine. For this purpose, the stock turbocharger was removed and a closed workinggas circuit connecting the exhaust and intake sides of the engine was installed. This arrangement allows the working gas to be conditioned after combustion and recirculated to the engine. For the investigations with hydrogen port fuel injection (PFI), the intake manifold was equipped with three injectors to enable cylinder-individual fuel delivery. Since the test engine was originally designed with direct injection, only the fuel injectors had to be replaced with hydrogen-compatible components for the investigations using direct hydrogen injection (DI). The engine was coupled to a dynamometer via an electric machine. In motored operation, the electric machine was used to drive the engine to the desired speed without combustion. During fired operation, it served as a load unit to maintain a constant engine speed independent of the generated power output. Since the APC is primarily intended for stationary applications for the reconversion of renewable energy, all experiments were conducted at a constant engine speed of 1500 rpm. This speed represents a typical operating point for generator-coupled applications in the European power grid. To analyze the combustion process, pressures in the intake system, combustion chamber, and exhaust system were recorded with crank-angle resolution. Temperatures and gas compositions were measured on a time-resolved basis and continuously logged. Only steady-state operating points were considered for the evaluation in order to ensure reproducible boundary conditions and to systematically investigate the influence of individual operating parameters. 4 Results 4.1 Initial Operation of the APC The increased isentropic exponent of the working gas when using argon as the inert gas enhances the theoretical efficiency potential of the system. This advantage results from the altered thermodynamic conditions during compression and combustion. Compared to operation with air, significantly higher pressures and temperatures occur inside the combustion chamber when argon is used as the working gas. However, these operating conditions also have a strong influence on the tendency toward combustion anomalies. Figure 3 shows the calculated pressures and temperatures during a polytropic compression process for different working gases. The test engine used in this study has a production compression ratio of ε = 10. When the working gas is changed from air to argon, the pressure at the end of compression increases by approximately 20 bar. At the same time, the gas temperature rises by about 650 K. Although the oxygen and hydrogen required for combustion slightly reduce the effective isentropic exponent of the working gas mixture, the overall effect remains essentially unchanged. Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle 199 <?page no="200"?> Figure 3: Influence of the working gas on compression pressure and temperature When operating with pure argon at a compression ratio of ε = 10, compression temperatures exceeding 1100 °C are reached. As a result, the auto-ignition temperature of hydrogen is significantly exceeded, making controlled spark ignition impossible. During the initial experiments, the homogeneous hydrogen-oxygen mixture ignited before the intended ignition timing. Consequently, the combustion phasing could no longer be controlled. In addition, the mechanical loading of the engine increased considerably due to the resulting pressure oscillations. Figure 4 shows an exemplary in-cylinder pressure trace, the calculated heat release rate, and the results of the knock analysis performed by the combustion analysis system. Knock is characterized by high-frequency pressure oscillations superimposed on the regular cylinder pressure trace. To quantify knock intensity, these oscillatory components are separated from the original pressure signal using a high-pass filter. The maximum pressure oscillation amplitudes are then plotted for 200 consecutive engine cycles. Figure 4: Combustion analysis of the APC with pure argon and a compression ratio of ε = 10 200 Lukas Kniestedt <?page no="201"?> Combustion could not be controlled by the spark plug. As a result of auto-ignition, the center of combustion occurred very close to top dead center (TDC). Throughout the entire measurement period, additional pressure oscillations with amplitudes between 1 and 4 bar were detected. For engine operation, a knock limit of 2 bar pressure oscillation amplitude was defined. This limit was clearly exceeded at the operating point shown. Under these conditions, only an indicated mean effective pressure (IMEP) of 3.2 bar could be achieved. The results of the initial commissioning demonstrate that stable and controlled operation of the APC with pure argon and the production compression ratio is not feasible. Consequently, suitable measures had to be developed to restore spark-controlled combustion while simultaneously reducing the tendency toward knock to an acceptable level. The following investigations therefore focus on approaches to lower the end-ofcompression temperature and improve combustion stability. 4.2 Investigation of Knock Behavior Following the initial investigations, it became evident that controlling combustion anomalies is one of the key development challenges of the Argon Power Cycle. The theoretical analysis of end-of-compression temperatures identified two major influencing factors: the isentropic exponent of the working gas and the compression ratio. Consequently, two additional inert gas mixtures with reduced isentropic exponents were introduced for the subsequent investigations. Despite the reduction in isentropic exponent, both mixtures remained chemically inert with respect to oxygen. Figure 5: Variation of inert gas composition and engine load As shown in Figure 5, experiments were conducted at different load points using the two additional working gas mixtures. The investigated mixtures exhibited isentropic exponents of κ = 1.60 and κ = 1.54, respectively. By modifying the working gas composition, the thermodynamic conditions within the combustion chamber could be adjusted in a targeted Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle 201 <?page no="202"?> manner. In particular, the temperatures during compression were reduced, enabling stable spark-ignited operation while maintaining approximately constant combustion phasing. The results show that significantly higher engine loads can be achieved with decreasing isentropic exponent. At the same time, the measured indicated efficiencies demonstrate the fundamental advantage of the APC compared to a conventional air-breathing engine. Despite the reduction in isentropic exponent, a noticeable efficiency benefit is retained. These investigations demonstrate that the knock behavior of the engine can be effectively influenced through the selection of the working gas. Reducing the isentropic exponent lowers the temperatures reached during compression and consequently decreases the likelihood of auto-ignition and knock. However, the theoretical efficiency potential of the process is reduced as well. As a result, a trade-off arises between combustion stability, maximum achievable engine load, and maximum attainable efficiency. Reducing the compression ratio leads to a similar effect as lowering the isentropic exponent of the working gas. Both measures reduce the pressures and temperatures occurring during compression and can therefore decrease the tendency toward combustion anomalies. Compared to modifying the working gas composition, however, changing the compression ratio requires a mechanical modification of the engine. In addition, a more significant reduction in compression ratio is necessary to achieve a noticeable improvement in combustion stability. Figure 6 shows the knock analysis for operation with pure argon, a compression ratio of ε = 8.5, and direct hydrogen injection. Even at an IMEP of 7 bar, the ignition timing has to be shifted significantly toward later combustion phasing in order to remain below the defined knock limit of 2 bar. This demonstrates that reducing the compression ratio improves combustion stability but is not sufficient to completely eliminate combustion anomalies. Figure 6: knocking behavior with Argon and reduced compression ratio A comparison of both approaches indicates that the adaptation of the working gas composition has a more pronounced effect on combustion stability than the reduction of the compression ratio. Nevertheless, lowering the compression ratio can further extend 202 Lukas Kniestedt <?page no="203"?> the stable operating range when combined with an optimized working gas. The optimum combination of working gas composition and compression ratio is application-specific and requires balancing maximum load capability against achievable efficiency. In addition to the measures discussed so far, which require either mechanical modifications to the engine or the selection of a suitable working gas, a final parameter was investigated that can be adjusted during engine operation. For this purpose, the experiments were conducted using a reduced compression ratio and an optimized inert gas mixture. To reduce knock intensity, the system pressure of the closed-cycle circuit was gradually increased. Increasing the system pressure raises the mass trapped within the combustion chamber. As a result, peak combustion temperatures are reduced, thereby decreasing the tendency toward knock. A comparable effect is well known from boosted air-breathing engines and can, in principle, also be applied to the Argon Power Cycle. Figure 7: Knock behavior at different system pressures Figure 7 shows an excerpt of the cylinder pressure analysis and the corresponding knock evaluation at identical engine load and varying system pressures. The increase in system pressure is clearly visible in the cylinder pressure traces throughout the entire engine cycle. At the same time, the heat release characteristics are affected. Due to the lower combustion temperatures and the increased dilution of the mixture, the heat release curve becomes flatter and broader. This indicates a slower and more uniform combustion process. The evaluation of the pressure oscillations shown on the right-hand side of Figure 7 further demonstrates the beneficial effect of increased system pressure on combustion stability. As the system pressure is increased, knock intensity is significantly reduced, resulting in more stable combustion. These results demonstrate that, in addition to the choice of working gas and compression ratio, the system pressure also has a considerable influence on knock behavior. In Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle 203 <?page no="204"?> contrast to the other two measures, however, system pressure can be adjusted during engine operation and is therefore particularly well suited for the fine-tuning of combustion stability under varying load conditions. 5 Conclusion The Argon Power Cycle (APC) represents a concept for the emission-free reconversion of hydrogen into electrical energy. By employing a closed working-gas cycle with argon and oxygen, nitrogen oxide emissions can be completely eliminated. At the same time, the high isentropic exponent of argon increases the theoretical efficiency potential compared to conventional air-breathing internal combustion engines. However, the resulting increase in in-cylinder pressures and temperatures leads to a higher tendency toward combustion anomalies, particularly auto-ignition and knock. Within the scope of this study, a Ford 1.0 L EcoBoost engine was converted into a hydrogen-fueled closed-cycle engine and investigated on an engine test bench. The experiments included the initial commissioning of the APC as well as a systematic evaluation of different approaches for influencing knock behavior. Initial investigations with pure argon and the production compression ratio of ε = 10 revealed that controlled spark-ignited operation was not possible due to the extremely high end-of-compression temperatures. Auto-ignition of the hydrogen-oxygen mixture occurred during the compression process, resulting in a significant exceedance of the defined knock limits. To improve combustion stability, the isentropic exponent of the working gas, the compression ratio, and the system pressure were subsequently varied. The results show that reducing the isentropic exponent has the strongest impact on knock tendency while still allowing high engine loads at elevated indicated efficiencies. Reducing the compression ratio produces similar effects but requires more extensive hardware modifications and proves less effective. Furthermore, increasing the system pressure was shown to significantly reduce knock intensity and can therefore be used as an operational parameter for adjusting combustion stability. Overall, the investigations demonstrate that stable spark-ignited operation of the Argon Power Cycle is feasible when the thermodynamic conditions within the combustion chamber are carefully adjusted. The selection of the working gas, compression ratio, and system pressure provides effective means of balancing the trade-off between combustion stability, maximum load capability, and efficiency. Consequently, the Argon Power Cycle represents a promising approach for the emission-free reconversion of green hydrogen in stationary energy applications. 6 Acknowledgements The authors gratefully acknowledge the financial support provided for the research project “Zero-Emission Closed-Cycle Engine”. The project was funded by EuroNorm GmbH under the auspices of the German Federal Ministry for Economic Affairs and Energy (BMWi). 204 Lukas Kniestedt <?page no="205"?> 7 References [1] Cech, Manuel; Knape, Matthias; Wilfert, Tom; Reiser, Christian (2021): Der abgasfreie Wasserstoff-Kreislaufmotor. In: MTZ Motortechnische Zeitschrift Z 82 (4), S. 44-49. DOI: 10.1007/ s35146-021-0645-1. [2] Kniestedt, Lukas; Cech, Manuel; Mahler, Kay; Reiser, Christian (2022): Experimental investigations on increasing performance and efficiency of the spark-ignited Argon Power Cycle. In: 12th Dessau Gas Engine Conference. Investigation of the Knock Behavior of the Spark-Ignited Argon Power Cycle 205 <?page no="207"?> Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal Federico Ricci 1* , Stefano Papi 1 , Massimo Dal Re 1 , Massimiliano Avana 2 Zusammenfassung: Um den steigenden Anforderungen an höhere Motorwirkungsgrade und geringere Umweltbelastung gerecht zu werden, gewinnen innovative Zündsysteme in der Automobilindustrie zunehmend an Bedeutung. Unter diesen stellt die Vorkammerzündung eine Schlüsseltechnologie für zukünftige Ottomotoren dar, da sie im Vergleich zu konventionellen J-gap-Zündkerzen die Klopfneigung reduziert, die Turbulenz im Brennraum erhöht und eine schnellere sowie stabilere Verbrennung ermöglicht. Die verbesserte Zündfähigkeit unterstützt zudem den Betrieb mit mageren Gemischen, erhöhtem Ladedruck und Abgasrückführung, wobei der Hauptvorteil auf der volumetrischen Zündung durch energiereiche reaktive Jets aus der Vorkammer beruht. Gleichzeitig erfordert der Übergang zu Smart-Mobility-Konzepten fortschrittliche Strategien zur zustandsbasierten Wartung, um klassische zeit- oder laufleistungsbasierte Inspektionen zu ergänzen. Zündkerzen sind während des Betriebs hohen thermischen, chemischen und elektrischen Belastungen ausgesetzt, die zu fortschreitendem Verschleiß, Änderungen des Elektrodenabstands und einer Drift des Zündverhaltens führen können. In dieser Arbeit wird ein von Tenneco entwickeltes Smart- Ignition-Coil-System zur Überwachung des Zündkerzenverschleißes untersucht. Die Spule erzeugt ein Diagnosesignal, dessen Dauer dem Zeitintervall zwischen dem Beginn der Spulenladung und dem Ende der Funkenentladung entspricht. Dieses Signal kann zur Bewertung der Zündqualität und zur Erkennung von Fehlzündungen genutzt werden; in der vorliegenden Arbeit wird insbesondere seine zeitliche Entwicklung als indirekter Indikator für verschleißbedingte Änderungen der Zündkerze betrachtet. Zur Validierung des Ansatzes wurde ein umfassender experimenteller Rahmen auf Basis einer 2x2-Testmatrix entwickelt, der sowohl konventionelle J-gap-Zündkerzen als auch Vorkammer-Zündsysteme umfasst. Die Methodik beginnt mit kontrollierten Labortests, um den Zusammenhang zwischen Elektrodenabstand, physischem Verschleiß und Variationen des Diagnosesignals zu untersuchen. Anschließend wird der Ansatz auf motorische Anwendungen übertragen, um die Methode unter realen thermischen, chemischen und verbrennungsbedingten Belastungen zu bewerten. Zusätzlich wird ein KI-gestützter Post-Processing-Ansatz eingesetzt, um die langfristige Drift statistischer Merkmale des Diagnosesignals zu verfolgen und die Interpretation des Verschleißfort- 1 CHAMPION IGNITION TENNECO, Via Della Scienza 6/ 8, 41012, Carpi, Italy * Corresponding Author: federico.ricci@tenneco.com 2 Department of Engineering, University of Perugia, Via G. Duranti 93, 06125, Perugia, Italy <?page no="208"?> schritts zu unterstützen. Alle experimentellen Daten werden durch mikroskopische und röntgenbasierte Inspektionen kreuzvalidiert, wodurch eine direkte Korrelation zwischen Signaländerung und physischem Zustand der Hardware hergestellt wird. Abstract: To meet the growing demands for higher engine efficiency and reduced environmental impact, innovative ignition solutions are becoming essential in the automotive industry. Among these, pre-chamber ignition technology has emerged as a key enabler for next-generation internal combustion engines. The pre-chamber ignition system offers superior combustion characteristics over traditional J-gap spark plugs, specifically by reducing the tendency for engine knock, enhancing in-cylinder turbulence, and ensuring a more rapid and consistent burn. Moreover, the increased ignitability facilitates lean-burn operation and high-pressure intake conditions without compromising exhaust gas recirculation stability. This performance is primarily connected to the mechanism of volumetric ignition triggered by high-energy reactive jets. However, the transition toward Smart Mobility requires advanced predictive maintenance strategies to replace traditional scheduled inspections. Spark plugs are subject to both extreme thermal and chemical stress, leading to gradual performance drift. Therefore, the integration of alternative systems is required for monitoring spark plug status to improve engine control, thus facilitating the detection of early problems while ensuring optimal performance throughout the vehicle’s operational life. In this context, a specific Smart ignition coil, developed by Tenneco, has been utilized in this activity to meet this target. The device produces a diagnostic signal whose duration corresponds to the total interval between the start of coil charging and the end of spark discharge. This signal is used for combustion diagnostics, including misfire detection and spark quality evaluation. Previous studies have successfully demonstrated the synergy between this hardware and artificial intelligence for predicting in-cylinder characteristics and identifying transient engine anomalies. Building upon these foundations, the present work shifts the focus toward the temporal evolution of the ignition signal to monitor spark plug wear. To ensure the reliability of the proposed approach, a comprehensive validation framework was implemented across a 2x2 experimental matrix, covering both conventional J-gap and pre-chamber ignition systems. The methodology follows a defined approach, starting from laboratory tests to establish a fundamental correlation between electrode gap wear and diagnostic signal variations and an AI-based approach for data post-processing aimed at simplifying the proposed method. The study then moves to real-world engine applications, where the methodology is assessed under actual thermal and chemical stresses using standardized cycles for signal acquisition. This process was first validated on traditional J-gap spark plugs and subsequently extended to pre-chamber systems to verify its adaptability to complex ignition architectures. To provide a rigorous ground truth for the AI model, all experimental data were cross validated through physical and optical inspections, ensuring a precise correlation between signal degradation and the physical hardware state. 208 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="209"?> 1 Introduction The continuous pressure to reduce greenhouse gas emissions while maintaining high power density and drivability is pushing spark-ignition (SI) internal combustion engines (ICEs) toward increasingly demanding operating conditions [1]. Lean combustion, high dilution levels, exhaust gas recirculation, and increased intake pressure are among the most effective strategies to improve thermal efficiency and reduce fuel consumption [2]. However, these approaches also move the combustion process closer to its stability limits, increasing the requirements for energy delivery, discharge robustness, and low cycle-tocycle variability [3]. Under such conditions, the ignition event becomes a critical enabler for stable combustion, since small variations in kernel formation may result in combustion instability, misfire, and reduced engine efficiency. Within this context, pre-chamber ignition has emerged as one of the most promising technologies for next-generation SI engines [4]. Unlike the local combustion generated by conventional J-gap spark plugs, prechamber ignition promotes ignition of the main chamber through multiple high-energy reactive jets generated inside a small auxiliary volume [5]. This mechanism increases the effective ignition surface, enhances turbulence in the main chamber, and accelerates flame propagation, with consequent benefits in terms of ignition robustness and combustion stability. These characteristics make pre-chamber ignition attractive for advanced combustion concepts, where conventional ignition systems may struggle [6]. Moreover, J-gap spark plugs operate in a highly aggressive environment, where repeated electrical discharges, high temperature, pressure fluctuations, chemical exposure, and combustion deposits progressively modify the electrode geometry and surface condition. Electrode erosion typically leads to gap growth and changes in breakdown and discharge characteristics [7]. Several studies have shown that spark plug wear can increase the required breakdown voltage, alter the spark discharge process, and eventually compromise ignition reliability [8]. Moreover, high-energy ignition systems, often required to support lean or diluted combustion, may further accelerate electrode erosion and reduce spark plug service life [9]. From an industrial perspective, this degradation mechanism is particularly relevant because spark plug replacement is still mainly managed through scheduled maintenance intervals or post-failure inspections [10]. Such approaches do not account for the actual operating history of the component, which may vary significantly depending on engine calibration, fuel type, driving profile, combustion mode, and environmental conditions [11,12]. The development of a non-invasive diagnostic method capable of monitoring spark plug wear during operation would therefore represent an important step toward condition-based maintenance and smart powertrain health monitoring without additional intrusive sensors [13]. Ignition coils are of particular interest because their voltage and current signals are directly affected by electrical breakdown, spark discharge, and the interaction between the plasma channel and the in-cylinder environment [14,15]. Previous research has demonstrated that coil signals can be used to provide information related to combustion quality, in-cylinder pressure features, and abnormal combustion events, confirming the potential of the ignition system as both an actuator and a virtual sensing platform [16]. Building on these findings, the present work investigates the use of a Smart ignition coil [15,16] diagnostic signal for spark plug wear monitoring. Since spark plug aging modifies the electrical and physical conditions at the electrodes, the temporal Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 209 <?page no="210"?> evolution of this signal is expected to contain information about progressive degradation of the ignition device. The main objective is therefore to assess whether this diagnostic feature can be used as an indirect indicator of spark plug wear and as an early marker for service requirement detection. The experimental plan includes both conventional J-gap and pre-chamber ignition systems, allowing the methodology to be evaluated on ignition architectures with different discharge configurations. The work starts with controlled laboratory tests, where the relationship between electrode gap variation and diagnostic signal evolution can be isolated and analyzed under repeatable conditions. The methodology is then extended to engine applications, where the diagnostic strategy is evaluated under real thermal, chemical, and combustion-related stresses. A further objective of the work is to investigate the use of artificial intelligence as a monitoring tool for detecting long-term drift in the diagnostic signal. Rather than replacing the deterministic diagnostic logic, the AI-based approach is intended to track changes in the signal characteristics over time and support the identification of deviations from the expected reference behavior. Finally, physical and optical inspections are used as ground truth to cross-validate the diagnostic results and to correlate the observed signal degradation with the actual hardware state. In this way, the study aims to demonstrate the feasibility of using smart ignition diagnostics as a non-invasive tool for predictive maintenance, enabling early detection of spark plug wear and supporting future condition-based service strategies for both conventional and advanced ignition systems. 2 Materials and methods All experimental activities were performed using an inductive ignition coil developed by Champion Ignition, also referred to as Hy2Fire® because it was originally designed for hydrogen internal combustion engine applications [15]. The Hy2Fire® coil uses an inductive architecture with a high-voltage transformer and a primary-side IGBT to control the ignition event. The transformer raises the supply voltage to spark-discharge level, while the magnetic core transfers high ignition energy to the secondary side (Figure 1). A key feature of the Smart ignition coil is the generation of a dedicated diagnostic signal representative of the ignition process. This signal can be processed by the ECU to verify proper coil charging, detect electrical faults, and evaluate spark discharge behavior without external primary and secondary current measurements. In this work, variations in the Discharge Feedback duration and dispersion are statistically analyzed as potential indirect indicators of progressive spark plug wear. 210 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="211"?> Figure 1: (a) Hy2Fire® Design (b) signals composition The experimental activity was designed to investigate the relationship between spark plug wear and the evolution of the Smart ignition coil diagnostic signal. The images in Figure 2 show the spark plug architecture used in this activity: the conventional J-gap spark plug at the top and the pre-chamber device at the bottom. Figure 2: J-gap spark plug and pre-chamber ignition device used in the experimental activity. The methodology included three complementary activities: • a controlled laboratory aging campaign on selected J-gap spark plugs (Test A); • an engine validation on additional J-gap spark plugs under firing conditions (Test B); • and a pre-chamber validation based on repeated optical inspections and ambientpressure diagnostic characterization before and after engine operation and aging (Test C). For Test A, each sample was initially characterized through microscopic analysis and Xray inspection (step 1), and ambient-pressure diagnostic acquisition (step 2). Optical investigations were performed to provide an independent physical reference for the assessment of spark plug wear and to cross-validate the diagnostic information extracted from the coil signal. Two complementary inspection techniques were adopted for step 1 (Table 1): Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 211 <?page no="212"?> • Optical analysis: X-ray imaging was carried out using a Nordson Quadra 3 Pro system and was used to evaluate the geometrical condition of the plugs. • Physical analysis: Microscopic inspection was performed using a Nikon SMZ25 microscope equipped with a Photonic F3000 illumination system, in order to document the visible wear state of the electrodes and the firing-end region. For brevity, the results section reports only the gap values measured at different electrode positions (left, middle, right), which are the relevant data for wear assessment (see Table 1). The minimum value is used in the results analysis, as it identifies the position where discharge is more likely to occur due to locally higher electrical conductance. Front view Side view Top view Physical analysis Optical analysis Table 1: Example of the optical investigation views used for microscopic and X-ray inspection. The ambient-pressure characterization (step 2) was performed to assess the behavior of the Smart coil diagnostic signal under repeatable non-firing conditions. In this test configuration [16], the spark plug was directly connected to the Hy2Fire® coil plug boot and operated in ambient air. The dwell time was fixed at 2.5 ms for this campaign. The main current supplied to the coil was provided by an EA-PS 2042-20B DC power supply. The main electrical quantities of the coil were monitored by means of a Tektronix MSO46 oscilloscope. The primary current was measured using a Tektronix TCP0030A current probe, while the secondary voltage was measured using a Tektronix P6015A high-voltage probe connected to the coil plug boot through a customized high-voltage adapter. The diagnostic signal generated by the coil was recorded using a HIOKI 8431-20 Memory HiLogger. The resulting dataset was then used to calculate the mean value and standard deviation of the diagnostic signal duration, which represents the reference statistical indicators for the analytical wear-monitoring methodology described in the previous section. The spark plugs were then subjected to an aging campaign (step 3) with continuous diagnostic signal recording and selected samples removed at predefined intervals for 212 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="213"?> optical inspection (step 1) and diagnostic characterization (step 2). This allowed the temporal evolution of the diagnostic signal to be correlated with progressive electrode wear. The aging tests were performed on a dedicated testbed cell designed to reproduce controlled electrical, thermal, and pressure conditions during continuous spark operation. The setup was installed inside a Weisstechnik ClimeEvent climate chamber and equipped with the main instrumentation required to supply and monitor the coil. The coil was powered by an EA-PS 2042-20B DC power supply, while the primary current and secondary voltage were monitored using a Tektronix MSO46 oscilloscope with a TCP0030A current probe and a Tektronix P6015A high-voltage probe. The diagnostic signal was continuously recorded using a HIOKI 8431-20 Memory HiLogger. The aging testbed cell and its experimental configuration are illustrated in Figure 3. Figure 3: Aging test cell AI-assisted progressive forecasting was used to predict the future coil diagnostic response of an unseen spark plug from its early aging history. The model did not estimate operating hours or remaining useful life but forecasted statistical descriptors of the Discharge Feedback duration as indicators of wear-related diagnostic drift. During Test A, the signal was extracted cycle by cycle and grouped into blocks of 1000 valid events, from which means, standard deviation, coefficient of variation, percentiles, interquartile range, skewness and kurtosis were computed. A Random Forest regression model was trained on the aging histories of four spark plugs, while the fifth was used as the unseen validation specimen. Generalization was assessed through leave-one-spark-plug-out and walk-forward validation, using only the diagnostic history available at each cutoff; future measurements were retained only for retrospective comparison. The main characteristics of the framework are summarized in Table 2. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 213 <?page no="214"?> Diagnostic descriptors Derived predictor variables RF configuration Prediction Mean_us Baseline value Regression RF Future Mean_us Std_us Latest value 1000 trees Future Std_us Median_us Long-term slope MinLeafSize = 5 Future Median_us P05_us Short-term slope OOB enabled Future P05_us P95_us Individual polynomial forecast LOSO validation Future P95_us IQR_us Population polynomial forecast Walk-forward Future IQR_us FutureHour, CutoffHour, ForecastHorizon_h One RF per descriptor Future diagnostic prediction Table 2: Random Forest forecasting framework adopted for progressive diagnostic drift prediction. Test B was used to transfer the methodology from ambient conditions to engine operation. New, artificially increased-gap and engine-aged J-gap spark plugs were compared under stabilized firing conditions, and the cycle-by-cycle diagnostic signal was statistically analyzed to assess whether the approach remained sensitive to spark-plug-related variations in a real combustion environment. The engine-based validation was performed at the University of Perugia on a conventional 999 cm³ turbocharged three-cylinder Smart W451T with compression ratio of 10: 1 and fueled with gasoline E5 in PFI mode [15,17]. The experimental configuration was based on gasoline operation using E5 fuel and a J-gap spark plug architecture. During the dedicated validation activity, the engine was operated at stabilized conditions in order to minimize combustion variability and improve the sensitivity of the diagnostic signal to spark-plug-related effects. Test C extended the same concept to a passive pre-chamber ignition device. The diagnostic response (step 3) was analyzed before and after engine operation and aging (step 2), and compared with X-ray and microscopic evidence (step 1), to verify whether the ambient-pressure diagnostic method could also be used to monitor the progressive physical condition of pre-chamber spark plugs. 3 Test campaign 3.1 Test A Table 3 summarizes the Test A workflow for five J-gap spark plugs. Two samples were subjected to 300 h of aging, with continuous coil diagnostic acquisition throughout the test. Green cells indicate performed activities, while gray cells indicate activities not carried out at the corresponding aging time. At fixed aging steps, the test was temporarily interrupted, and the spark plugs were progressively removed to perform step 1 (microscopic and X-ray inspections), followed by step 2 (ambient-pressure diagnostic characterization). These inspections provided the ground truth, while the diagnostic measurements verified the sensitivity of the signal to aging-induced wear and gap variation. The histories acquired from J-gap 01 to J-gap 04 were used for AI training, while J-gap 05 was retained as the unseen sample for progressive diagnostic-drift prediction. 214 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="215"?> Test [h] 0 75 150 225 300 Step [n°.] 1 2 3 3 1 2 3 1 2 3 1 2 3 1 2 ID Spark J-gap 01 J-gap 02 J-gap 03 J-gap 04 J-gap 05 Table 3: Workflow adopted for the spark plug wear-monitoring campaign. Table 4 summarizes the main characteristics of the aging test. Feature Specification Battery Voltage [V] 14 Frequency [Hz] 50 Output Voltage [kV] 20 +/ -1.5 Test duration [h] 300 Temperature [°C] 25 Pressure [bar] 10 Medium [type] N 2 Table 4: Specification of the aging test. Within this workflow, two complementary data uses were defined: Analytical plug-wear evaluation: spark plugs removed at selected aging intervals were characterized under ambient-pressure and ambient-temperature conditions. These measurements provided discrete diagnostic states to be compared with the corresponding X-ray and microscopic evidence. Continuous aging-data exploitation: continuous diagnostic histories acquired during the aging test were used for AI-based monitoring. The histories of the first four spark plugs served as training data, while the fifth spark plug was retained as the unseen specimen for progressive diagnostic-drift prediction (Table 5). Training data AI model Unseen spark plug Validation J-gap 01, J-gap 02, J-gap 03 and J-gap 04 diagnostic history Random Forest regression J-gap 05 diagnostic history up to the current time Predicted future diagnostic evolution vs J-gap 05 measured future diagnostic evolution Table 5: Data split adopted for AI-based diagnostic-drift prediction, including training histories, unseen validation specimen and comparison between predicted and measured future diagnostic evolution. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 215 <?page no="216"?> 3.2 Test B The main operating conditions of Test B adopted for the wear-monitoring validation on the metal engine were: • Engine speed = 1500 rpm • Load = idle condition • ignition timing = -40 CAD aTDC. • Coil Dwell Time = 2500 μs. This calibration was selected to obtain stable combustion and a low coefficient of variation of IMEP, thereby reducing the influence of cycle-to-cycle combustion dispersion on the diagnostic feature. The test matrix included spark plugs with different wear-related conditions, including: • a new reference plug with nominal gap of 0.9 mm • plug with increased gap up to 1.3 mm • plug with increased gap up to 1.6 mm • and an engine-aged plug with over 100 working hours For each spark plug configuration, 500 consecutive acquisitions were recorded to assess diagnostic and combustion repeatability over a representative set of engine cycles. As shown in Figure 4, both quantities exhibit limited dispersion, confirming stable operating conditions and a reduced influence on combustion variability. This supports a more direct interpretation of diagnostic-signal variations as spark-plug-related effects. Figure 4: Cycle-by-cycle dispersion of the coil diagnostic signal and P cyl 3.3 Test C Test C investigated one passive pre-chamber ignition device to assess whether the ambient-pressure diagnostic method could be extended to pre-chamber wear monitoring. The device underwent X-ray and microscopic inspection, and ambient diagnostic characterization before and after engine operation and the aging test. This sequence allowed the diagnostic-signal evolution to be compared with optical and radiographic evidence, supporting the evaluation of pre-chamber wear through variations in the ambient diagnostic response. For confidentiality reasons, detailed X-ray images are not displayed 216 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="217"?> in the present work; only the information related to the spark gap and relevant for the wear assessment is reported. 4 Results 4.1 Test A Analytical plug-wear evaluation For the sake of brevity, only the microscopic images of two representative samples are reported in Table 6: the first spark plug removed from the ageing campaign (J-GAP 01) and the spark plug that reached the end of the test (J-GAP 04). • sharpening of central and ground electrode Visual examination following completion of the durability test showed moderate sharpening of the firing edges on both the center and ground electrodes. The wear pattern was uniform and symmetrical, indicating stable spark discharge conditions throughout the test. No signs of electrode melting, deposit-induced damage, cracking, or abnormal erosion were observed. The reduction in edge radius is consistent with normal electrode wear mechanisms associated with prolonged engine operation and did not result in any significant change in spark gap or electrode alignment. • surface oxidation on ceramic and electrode Only limited surface oxidation was observed on both the ceramic insulator, and the spark plug J-gap electrode. The ceramic surface retained its original appearance for the most part, showing only slight discoloration. No notable deposits, cracking, or signs of thermal damage were evident. Similarly, the electrode exhibited only minor oxidation, visible as a thin and localized oxide film on the exposed metallic surfaces, particularly in the vicinity of the spark gap. The electrode edges remained well defined, with no significant erosion, melting, or material degradation observed. • surface discharge on ceramic insulator evidence surface discharge was observed on the core nose of the ceramic insulator. The discharge manifested as faint tracking marks and localized discoloration extending along the insulator surface, indicating that flashover events had occurred intermittently during operation. These surface traces are consistent with electrical leakage paths developing across the ceramic under high-voltage conditions. The affected area exhibited slight darkening and superficial deposits, but no significant cracking, chipping, or structural damage to the ceramic material was observed. The discharge marks remained confined to the insulator surface and did not appear to penetrate the ceramic body. The limited extent of the tracking suggests that the flashover events were occasional rather than continuous throughout the test. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 217 <?page no="218"?> Physical Analysis - J-Gap 01 Front view Side view Top view 0 h After 75 h of aging Physical Analysis - J-Gap 04 Front view Side view Top view 0 h After 300 h of aging Table 6: Microscopic comparison of representative J-gap spark plugs before and after aging. The diagnostic trend over time, calculated according to the methodological criteria described in Section 3.1, is reported in Figure 5 and compared with the X-ray inspection results (minimum value of the spark gap; see Table 1 for reference). The results in Figure 5 confirm a consistent relationship between spark gap evolution and the Smart ignition coil diagnostic response. Since the spark plugs start from different reference gap values and undergo different aging durations, the comparison is expressed as percentage change relative to the initial condition. The orange bars indicate the gap increase, while the blue bars show the corresponding decrease in diagnostic feedback duration. Overall, the increase in electrode gap is associated with a reduction in diagnostic duration, confirming the sensitivity of the diagnostic feature to wear-related changes in the discharge region. 218 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="219"?> Figure 5: Percentage variation of spark gap and Discharge Feedback duration with respect to the new reference condition. Continuous aging-data exploitation Figure 6 shows the progressive AI forecast of the mean Discharge Feedback for the unseen spark plug. At each cutoff, the prediction is recalculated and re-anchored to the latest available measurement, while future measurements remain hidden and are used exclusively for validation. By combining the aging patterns learned from the reference spark plugs with the progressively acquired history of the unseen specimen, the AI framework adapts the forecast to its specific diagnostic evolution. The agreement with the experimental data improves as new measurements become available and the prediction horizon decreases, demonstrating the capability of the proposed approach to continuously update the expected diagnostic drift rather than providing a fixed population-based trajectory. This adaptive capability represents the main added value of the AI framework, enabling early forecasts to be progressively refined during operation. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 219 <?page no="220"?> Figure 6: Progressive AI forecast of the mean Discharge Feedback duration for the unseen spark plug. 4.2 Test B Figure 7 compares the filtered mean diagnostic duration for the investigated spark plug configurations. The three graphical representations provide complementary information on diagnostic length behavior. The bar chart summarizes the cleaned datasets through their mean value and standard deviation, showing a progressive reduction in diagnostic length from the new 0.9 mm condition to the new 1.6 mm condition. The KDE analysis confirms this trend using the actual sample distributions and reveals a substantial overlap among the new 0.9 mm, aged 0.9 mm, and new 1.3 mm conditions. In contrast, the new 1.6 mm population is clearly shifted toward lower diagnostic length values and exhibits a narrower distribution. The Gaussian reconstructions reproduce the same ranking observed in both the bar chart and KDE results, providing a compact representation of the distribution location and spread. Overall, all three visualizations consistently indicate that increasing electrode gap is associated with shorter diagnostic length. The strong overlap between the new and aged 0.9 mm conditions is consistent with the expected wear mechanism. During the relatively short aging campaign, thermo-electrical and chemical degradation mainly induce local electrode surface modifications, while the average electrode gap remains substantially unchanged. Consequently, the diagnostic length exhibits only a limited variation between the new and aged 0.9 mm spark plugs. 220 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="221"?> Conversely, the systematic shift observed for increasing electrode gap indicates that the diagnostic length is primarily governed by electrode spacing. These results suggest that the effects associated with gap enlargement can be partially distinguished from the thermo-electrical and chemical aging processes occurring during the early stages of spark plug wear. Figure 7: Comparison of the cleaned mean diagnostic duration, measured distribution and Gaussian reconstruction for the investigated spark plug configurations, used as reference for the Test C prechamber validation assumptions. 4.3 Test C The results obtained for the pre-chamber ignition device confirm the findings observed in the two previous activities. As shown by the optical inspections (Table 7), the progressive evolution of electrode gap is reflected in the corresponding variation of the coil diagnostic response (Figure 8). As shown in Table 7, the pre-chamber device exhibits a progressive modification of the firing-end region after engine operation and aging. Compared with the new condition, the inspected surfaces show evidence of localized wear and oxidation. These changes indicate that the pre-chamber electrodes are affected by the combined thermal, chemical, and electrical stresses experienced during operation, although the enclosed geometry makes direct wear observation and quantitative assessment more difficult than for conventional J-gap spark plugs. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 221 <?page no="222"?> Physical Analysis - Pre-chamber new After engine test After ageing Front view Top view Table 7: Microscopic comparison of the pre-chamber ignition device in new condition, after engine operation, and after aging. Accordingly, Figure 8 reports the percentage variation of the pre-chamber gap measured by X-ray inspection and of the Discharge Feedback duration with respect to the new reference condition. After engine operation, a reduction in diagnostic duration is already observed together with an increase in the measured gap. After the aging activity, both variations become more pronounced, confirming the progressive evolution of the discharge region. As already observed for the J-gap spark plugs, the increase in gap is associated with a decrease in diagnostic duration, supporting the applicability of the proposed methodology also to pre-chamber ignition devices. Figure 8: Percentage variation of pre-chamber spark gap and Discharge Feedback duration with respect to the new reference condition. After engine operation, a reduction in diagnostic duration is observed, indicating an initial drift associated with exposure to real firing conditions. After the aging activity, the diagnostic values decrease further, consistently with the more advanced wear state evidenced by the optical and radiographic inspection. This progressive reduction in diagnostic duration and increase in pre-chamber gap confirms that the proposed methodology can be extended from conventional J-gap spark plugs to more complex ignition architectures. This aspect is particularly relevant for the pre-chamber configuration, 222 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="223"?> where direct assessment of gap wear through conventional inspection methods is less straightforward than for a conventional J-gap spark plug. Microscopic analysis mainly provides information on the accessible surface condition and can reveal local superficial wear, while X-ray inspection allows the internal geometry and the spark gap to be observed, although it makes a quantitative assessment of progressive gap wear more challenging. The condition and quality of the electrodes are particularly important for combustion performance, as they directly influence spark formation and ignition reliability. In pre-chamber spark plugs, however, electrode conditions can be affected not only by wear but also by end-gas fouling and deposits developing within the pre-chamber environment, further complicating the interpretation of inspection results. In addition, unlike conventional J-gap spark plugs, the pre-chamber architecture does not allow direct spark-gap measurement with a feeler gauge, limiting the possibility of evaluating wear through standard mechanical inspection methods. Conversely, for a J-gap spark plug, the combined interpretation of microscopic images, X-ray views, and feeler-gauge measurements enables a more direct and reliable assessment of the wear state For this reason, the diagnostic response becomes fundamental for pre-chamber devices, since it provides an additional and functional indicator of the progressive evolution of the discharge region, complementing the partial information obtained from optical and radiographic inspections. 5 Conclusions The wear detection capability of the Smart ignition coil was evaluated using both conventional J-gap spark plugs and pre-chamber spark plugs under controlled laboratory conditions as well as during engine testing. The assessment demonstrated the applicability of the diagnostic approach across different spark plug technologies and operating environments. To enhance prediction accuracy and robustness, an AI-based algorithm was developed and implemented. The algorithm improves wear forecasting by accounting for the increasing variability of diagnostic signals as spark plug degradation progresses, enabling earlier and more reliable prediction of component wear over its service life. Wear detection was successfully validated through laboratory testing, which primarily focused on electrical erosion mechanisms affecting the spark plug electrodes. Additional validation was performed through engine testing, where the spark plugs were exposed to a broader range of degradation factors, including thermal stresses, vibration, and chemical interactions with combustion products. Despite the different wear mechanisms involved, both laboratory and engine test results showed a comparable impact on the diagnostic parameters, allowing reliable wear detection with a consistently high level of confidence in the data analyzed. While the overall wear detection methodology has been proven effective, the classification and differentiation of individual wear mechanisms require further investigation. Additional work is needed to improve the ability to distinguish between electrical, thermal, mechanical, and chemical degradation processes and to quantify their respective contributions to the observed diagnostic signatures. Monitoring Spark Plug Wear Using Smart Ignition Coil Diagnostic Signal 223 <?page no="224"?> 6 References [1] EUROPEAN PARLIAMENT ; COUNCIL OF THE EUROPEAN UNION . REGULATION ( EU ) 2023/ 851 OF 19 APRIL 2023 AMENDING REGULATION ( EU ) 2019/ 631 AS REGARDS STRENGTHENING THE CO 2 EMISSION PERFORMANCE STANDARDS FOR NEW PASSENGER CARS AND NEW LIGHT COMMERCIAL VEHICLES IN LINE WITH THE UNION ’ S INCREASED CLIMATE AMBITION . OFFICIAL JOURNAL OF THE EUROPEAN UNION , L 110, 2023, PP . 5-20. [2] DERNOTTE , J .; NAJT , P . M .; DURRETT , R . P . DOWNSIZED - BOOSTED GASOLINE ENGINE WITH EXHAUST COMPOUND AND DILUTE ADVANCED COMBUSTION . SAE INT . J . ADV . CURR . PRACT . MOBIL . 2020, 2, 2665-2680. [3] ATIS , C .; SCHOCK , H . 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 %). SAE INT . J . ADV . CURR . PRACT . MOBIL . 2021, 3, 1569-1584. [4] JEONG , S .- J . CFD SIMULATION OF PRE - CHAMBER SPARK - IGNITION ENGINES - A PERSPECTIVE REVIEW . ENERGIES 2024, 17, 4696. [5] MILLO , F .; ROLANDO , L .; PIANO , A .; SEMENTA , P .; CATAPANO , F .; DI IORIO , S .; BIANCO , A . EXPERIMENTAL AND NUMERICAL INVESTIGATION OF A PASSIVE PRE - CHAMBER JET IGNITION SINGLE - CYLINDER ENGINE . SAE TECHNICAL PAPER 2021-24-0010, 2021. [6] KARAGEORGIOU , D .; PRUNIER , T .; MYSLIVECEK , M .; VESEL , C . ASSESSMENT OF PASSIVE TJI TECHNOLOGY ON A MILD HYBRID POWERTRAIN AND ITS PERFORMANCE ON COLD OPERATING CONDITIONS . IN : SENS , M . ( ED .), 6 TH INTERNATIONAL CONFERENCE ON IGNITION SYSTEMS FOR SI ENGINES - 7 TH INTERNATIONAL CONFERENCE ON KNOCKING IN SI ENGINES , EXPERT VERLAG , 2024, PP . 157-196. ISBN 978-3-381-12991-1. [7] ZHANG , K .; FEUK , H .; ÄNGEBY , J .; EHN , A .; RICHTER , M . INVESTIGATION OF ELECTRODE WEAR DURING SPARK DISCHARGES USING PLANAR LASER - INDUCED FLUORESCENCE . JOURNAL OF PHYSICS D : APPLIED PHYSICS 2025, 58, 505501. [8] SHARMA , A .; BREDEN , D .; CRESS , J .; RAJA , L . PREDICTIVE BREAKDOWN MODELING FOR SPARK PLUG DESIGN . SAE TECHNICAL PAPER 2020-01-0781, 2020. [9] TAMBASCO , C .; HALL , M .; MATTHEWS , R . SPARK DISCHARGE CHARACTERISTICS FOR VARYING SPARK PLUG GEOMETRIES AND GAS COMPOSITIONS . SAE TECHNICAL PAPER 2022-01-0437, 2022. [10] TAMBASCO , C .; HALL , M .; MATTHEWS , R . EFFECTS OF SPARK PLUG OPERATING CONDITIONS ON ELECTRODE EROSION AND SURFACE DEFORMATION . SAE TECHNICAL PAPER 2024-01-2100, 2024. [11] KUČERA , M .; GUTTEN , M .; KORENČIAK , D .; KÚDELČÍK , J . USING THE SPARK PLUG AS A SENSOR FOR ANALYZING THE STATE OF THE COMBUSTION SYSTEM . SENSORS 2025, 25, 4198. [12] NAD , M .; DANKO , M .; KONIAR , D .; FRIVALDSKY , M . OPTIMIZATION OF THE IGNITION SYSTEM DIAGNOSTICS METHODOLOGY . VEHICLES 2026, 8, 71. [13] ZHU , D .; DENG , J .; WANG , J .; WANG , S .; ZHANG , H .; ANDERT , J .; LI , L . DEVELOPMENT AND APPLICATION OF ION CURRENT / CYLINDER PRESSURE COOPERATIVE COMBUSTION DIAGNOSIS AND CONTROL SYSTEM . ENERGIES 2020, 13, 5656. [14] YONEYA , N .; AMAYA , K .; KUMANO , K .; SUKEGAWA , Y .; UCHISE , Y .; JITSU , H .; FUJIYAMA , Y . MISFIRE DETECTION TECHNOLOGY WITH DEEP NEURAL NETWORK BASED ON IGNITION COIL SIGNALS . SAE INT . J . ENGINES 2023. [15] RICCI , F .; PAPI , S .; AVANA , M .; DAL RE , M .; GRIMALDI , C . SMART IGNITION COIL FOR H 2 ICE APPLICATION WITH COMBUSTION DIAGNOSTIC SYSTEM FOR MISFIRE DETECTION . SAE TECHNICAL PAPER 2025-24-0058, 2025. [16] PAPI , S .; RICCI , F .; DAL RE , M .; CHANDELIER , M . SMART IGNITION COIL DIAGNOSTIC SYSTEM FOR H 2 ICE COMBUSTION DETECTION . IN : SENS , M . ( ED .), 6 TH INTERNATIONAL CONFERENCE ON IGNITION SYSTEMS FOR SI ENGINES - 7 TH INTERNATIONAL CONFERENCE ON KNOCKING IN SI ENGINES , EXPERT VERLAG , 2024, PP . 95-124. ISBN 978-3-381- 12991-1. [17] R ICCI , F., P API , S., B ATTISTONI , M., Z EMBI , J., D AL R E , M., & G RIMALDI , C. (2023). E NGINE E FFICIENCY M EASUREMENTS U SING A 100 K H Z R ADIO F REQUENCY C ORONA I GNITER (N O . 2023-24-0041). SAE T ECHNICAL P APER . 224 Federico Ricci, Stefano Papi, Massimo Dal Re, Massimiliano Avana <?page no="225"?> The influence of operating parameters on spark plug wear and its prediction potential M.Sc. Celine Krnac, Dr.-Ing. Olaf Toedter, M.Sc. Moritz Grüninger and Prof. Dr. sc. techn. Thomas Koch Abstract: During engine operation, the ignition system has to provide ignitability of the air-fuel mixture without any misfires to guarantee the engines full use-case i. e. in decentralized energy supply systems. A major issue regarding its durability is the wear of the spark plug electrodes. The continuous erosion of the material increases the electrode gap and will cause misfires at a certain point. By reducing the volume of erosion, maintenance intervals can be extended, direct and indirect maintenance costs and materials can be reduced. To understand the fundamental mechanisms of wear, five wear models have been developed up to date: the particle-ejection model, the spark footpoint mechanism, erosion by vaporization, sputtering, and the formation and removal of an oxidation layer. All of these models are based on idealized, parallel electrodes and do not account for the complex interactions in real spark plugs. The experiments conducted here therefore differ from the investigations reported in the literature. Real spark plugs were used in these experiments, which are intended to provide insights into the actual ignition system. In this presentation, the influencing factors are divided into the categories of material wise, electrical parameters, and engine-related factors. Material-related parameters, as well as the coil activation duration (a subcategory of electrical factors) and the electrode gap (a subcategory of engine-related factors), are examined in comparison. The influence of activation duration and electrode gap is investigated on spark plugs with different electrode materials. An endurance test bench capable of accommodating up to four spark plugs simultaneously enables the investigation of wear over the entire life span under engine-like conditions. The spark plugs are operated at a slightly increased ignition frequency for approx. 26 million ignitions each. A single spark is detected using a high-speed camera in order to make statements about the radiation intensity of the ignition spark. After 193 hours of operation, the spark plug electrodes from the endurance test are analyzed using optical measurement methodes. The combination of the endurance test and the constantvolume chamber allows for a simultaneous evaluation of the influencing factors and provides a practical optimization approach for ignition systems. The analysis of the results shows that, for the longer coil activation duration, as expected the eroded volume of platinum and nickel is significantly greater than that for the <?page no="226"?> shorter activation duration. In tests of platinum and nickel erosion at varying electrode distances, it was shown, that nickel and platinum electrodes exhibit the greatest wear as the gap distance increases. The observations illustrate that, in addition to the geometrical and material property the control strategies also have a decisive influence. The influence of the plasma behavior inside the spark is affected direct or indirect by the combination of the rection chamber and the control strategy. The results make it possible to better understand wear and its underlying mechanisms and to investigate them under conditions closer to reality than before. 1 Motivation To ensure a stable engine operation, a misfire free ignition system is required. It can be problematic because the continuous wear of the electrode leads to a significantly higher ignition voltage. If the ignition system can not provide the voltage, misfires appear. Due to higher pressures and temperatures, downsized engines cause increased wear on the anodes and cathodes of spark plugs. That’s why a fundamental understanding of the spark plug wear is essential. Previous works tried to explain the physical and chemical mechanisms behind the erosion of the spark plug electrode with five different models: the particle-ejection model, the spark footpoint mechanism, erosion by vaporization, sputtering, and the formation and removal of an oxidation layer. However, these investigations only considered idealized models of parallel electrodes while this work includes the complex interaction in a series production spark plug in detail. Therefore, in this thesis a real spark plug is used, to examine the factors affecting wear. The different factors can be divided in three different main groups: material-related, electrical and engine-related. Several factors that influence wear, including material properties, coil activation duration, and spark plug gap, will be investigated. Those are directly influencing the plasma and gas effects during the spark, which has a significantly effect of the wear mechanisms. Therefore, the theoretical background has to be investigated and modeled behind these mechanisms and the corresponding measurement setups. An endurance test bench is used to show the erosion of the electrode materials. Afterward all results are presented and discussed. 2 Theoretical Background and Measurement Setups This chapter reviews the theoretical foundations of spark plug wear models, as explored in prior research, and outlines their key influencing factors and compares it to the experimental setups of the applied measurement methods. 2.1 Spark plug wear mechanisms The wear models described in this subchapter differ in the underlying physical and chemical processes. By integrating all five mechanisms, a comprehensive description of wear can be achieved. 226 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="227"?> Particle-Ejection-Model During the spark breakdown, a conductive plasma channel is formed, followed by the socalled arc phase. If the ion beam strikes the electrode surface during this arc phase, the energy input raises the local temperature around the spark footpoint. Upon exceeding the melting temperature, the metal melts locally, forming a melt pool (see Figure 1). The force exerted by the ion beam pushes the melt pool radially outward. This outward force is counteracted by the material’s recoil pressure. The transition to the glow phase, characterized by the cessation of the ion beam, induces a motion of the molten material away from the electrode surface, driven by the recoil force, opposing surface tension. If the surface tension is insufficient to counteract this motion, molten metal detaches from the melt pool in the form of droplets. These ejected droplets may redeposit on the electrode surface, creating crater-like and spherical structures. This erosion mechanism is referred to as the Particle-Ejection Model. [1], [2] Material-related parameters including the melting temperature, thermal conductivity, and surface tension influence melt pool formation and droplet emission. High enthalpy of fusion and melting temperature require greater energy input for melting, resulting in a smaller melt pool and fewer droplets. Enhanced thermal conductivity distributes the supplied energy more rapidly across the electrode, lowering peak temperatures and limiting the melt pool size. Elevated surface tension supresses droplet detachment, reducing droplet emission. Fig. 1: Particle-Ejection-Model recreated based on [2] Electrical parameters significantly affect the process as well. Extended coil activation duration increases the total energy available in the system, leading to a larger melt pool and enhanced droplet emission. High current density intensifies local heating, facilitating melting, while a higher current (i. e., more ions per unit of time) increases the ion beam force on the melt pool, leading to a greater emission of metal droplets. Engine-side parameters, including gas pressure, density, and combustion chamber flow, also play a critical role. An increase in gas density or pressure reduces the mean free path of ions, causing them to dissipate their kinetic energy more rapidly through collisions and decreasing droplet formation. The in-cylinder flow displaces the spark footpoint, causing a reduction of local energy density, which further suppresses droplet emissions. Erosion by evaporation A futher wear mechanism is the erosion of electrode material through evaporation. Energy input heats the material, and if the melting and subsequently the boiling The influence of operating parameters on spark plug wear and its prediction potential 227 <?page no="228"?> temperature are exceeded, the electrode material can be fully converted into the gas phase. [1], [3], [4] The process is influenced by material-related parameters, such as melting and boiling temperature and thermal conductivity [4]. Lower melting and boiling temperatures reduce the energy threshold required for evaporation, therby increasing erosion under identical energy input. High thermal conductivity of the spark plug electrode quickly distributes the supplied energy across the electrode, lowering local peak temperatures. If temperatures do not locally exceed the boiling point, material evaporation cannot occur. Fig. 2: Erosion by evaporation Electrical parameters also play a critical role. Higher current density J ¼ electric current I cross � section area A [5] increases local heating at higher currents and smaller ion beam cross-sections, enabling more material to reach and exceed the melting and boiling temperatures and intensifying evaporation. Coil activation duration influences the energy input into the system, extended durations result in larger erosion volumes Erosion caused by sputtering Within the conductive plasma channel, the electric field accelerates ions toward the center electrode, where they transfer their kinetic energy to surface atoms through collisions. If the binding energy of the material is lower than the transferred energy, surface-near atoms can be ejected from the metal lattice. This process is known as sputter erosion. Fig. 3: Erosion by sputtering based on [6] 228 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="229"?> Material-related factors like sublimation energy, threshold energy, atomic mass, and lattice structure define the erosion potential. Lighter atoms, with lower atomic masses, require less energy to remove them from the crystal lattice, while the specific lattice structure determines the local binding strength. [1] On the electrical side, a higher current means more ions capable of sputtering the lattice. Engine-side parameters, such as high gas density and pressure, the mean free path of the ions shorten, leading to more frequent collisions and energy loss before impact. Consequently, less energy is available to dislodge atoms, reducing erosion. The interplay of these influencing factors allows targeted control of sputtering. Formation and removal of an oxidation layer and footpoint machanism The high temperatures inside the combustion chamber, combined with a high oxygen environment, especially in atomic form, promote the formation of an oxidation layer on the metal surface of the electrode, see Figure 4. This process begins with the physisorption of oxygen molecules (1), followed by their dissociation and chemisorption onto the metal (2-3), the resulting oxides then vaporize (4). A process that constitutes erosion through the removal of the oxidation layer. If these oxides redeposit (5), the reactions proceed in reverse. [3] Fig. 4: Formation and removal of an oxidation layer based on [1] The process is strongly influenced by material-related parameters, particularly oxidation resistance, which determines the electrodes’s tendency to form an oxide layer. Engine-side factors such as high atomic-oxygen concentration promote the dissociation of O 2 and its chemisorption onto the metal surface, causing the oxide layer to form more rapidly. High temperatures are essential for both dissociation and the subsequent vaporization of oxides, without sufficient thermal energy, the oxidation layer remains stable, slowing down erosion. Higher pressure or gas density shortens the mean free path of oxygen atoms, increases their local concentration at the metal surface, and promotes the formation of a thicker oxide layer, which in turn accelerates material removal. The influence of operating parameters on spark plug wear and its prediction potential 229 <?page no="230"?> Fig. 5: Spark footprint mechanism In contrast, the spark-footpoint mechanism describes localized overheating at the spark impact point. A thin oxidation layer forms there and evaporates at sufficiently high temperatures, allowing the spark to advance through the material by relocating the footpoint. This progressive erosion affects deeper electrode layers. In-cylinder flows, can cause the spark to break off, causing the spark footpoint to migrate across the surface. This distribution of energy reduces the temperature concentration at a single point and minimizing erosion in both mechanisms. 2.2 Measurement Setups and Experimental procedure The measurement setup comprises an endurance test bench and a pressure chamber equipped with optical access, providing a comprehensive assessment of electrode wear behavior, which is described in detail below. Major focus of the measurements is a usage in real spark plugs as an more usage correlaleted testing environment in order to applicate the already developed general models to a concrete use case. Endurance Test Bench The endurance test bench enables testing spark plugs to a predetermined number of ignitions in order to draw conclusions about electrode wear. The setup features a pressure chamber containing up to four spark plugs, providing conditions comparable to real engine usage. Power is supplied by a 12 V automotive battery that maintains a stable voltage throughout the test, while a function generator controls the ignition coils at a predefined frequency. A microcontroller counts the generators’s signal edges and sequentially distributes the trigger signal to the coils. 230 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="231"?> Valve Valve Thermocouple Temperature Monitoring Pressure Chamber Ignition Coils & Spark Plugs Function Generator Pressure Regulator Compressed Air Leak Fan Pressure Sensor Compressed Air Monitoring Microcontroller Optocoupler Control Box Battery + Charger Power Grid 30°C 8 bar Fig. 6: Schematic diagram of the endurance test bench setup A control box manages power supply to the battery, generator, microcontroller, and temperatureand pressure sensing elements, while a fan prevents ignition coil overheating, monitored via a thermocouple. The camber pressure is regulated by a pressurereducing valve, and a controlled leakage permits controlled air exchange, flushing radicals and metallic particles from the test chamber. Pressure chamber with optical access The optically accessible pressure chamber enables high-speed camera recordings of the spark discharge during the spark phases. These images allow the analysis of the radiation intensity and spark behavior within the electrode gap, providing insights into spark plug wear mechanisms. The spark plug is positioned within a constant-volume chamber, to ensure reproducibility discharge condition, featuring a square cross-section with constant pressure and no airflow. To maintain these conditions, the piston and turbulence plate is secured to prevent axial motion. After each spark discharge the chamber is flushed and refilled with fresh gas via magnetically actuated valves connected to a gas mixer, with the pressure set beforehand. High-Speed Camera FFM Ignition Coil Constant-Volume Combustion Chamber Spark Plug Gas Inlet Gas Blender Gas Outlet Fig. 7: Schematic diagram of the pressure chamber with optical access The influence of operating parameters on spark plug wear and its prediction potential 231 <?page no="232"?> A non-return valve prevents backflow to the gas mixer. Afterwards, the air is released from the chamber via a magnetic valve. The spark plug is connected to the ignition coil via a spark plug wire, while optical access to the plasma is provided through glass panels in the pressure chamber walls. Experimental procedure Electrode Electrode Gap Activation Duration Nickel Ø 2,4 mm Small High Intermediate Low High Large High Platinum Ø 0,6 mm Small High Intermediate Low High Large High Tab. 1: Parameter variation The target is a better understanding of the mechanisms causing wear and the factors affecting them. To this purpose, this study specifically varied the electrode gap and the duration of the ignition coil activation on both test benches. These parameter variations are presented in Table 1. The endurance tests run for a total of approx. 26 million ignition sparks. The spark plugs are photographed from different angles before and after 193 hours using a light microscope. These camera images serve as the basis for the analysis to determine the amount of eroded material. 3 Results In this chapter the results of the spark plug wear are presented. First the erosion by different electrode gaps is presented. After that, the variation of the coil activation duration was examined. 3.1 Variation electrode gap The results of the endurance test and the spectroscopic analysis of electrode wear are summarized and presented below. Electrode wear is not only affected by the electrode gap, as seen down below, but also influenced by the electrode material. The experiments verify, that platinum is more resistant to erosion than nickel. Targeted variation of the electrode gap demonstrates that it has a significant influence on the amount of electrode material removed. Platinum As expected, platinum exhibits significantly different wear behavior depending on the electrode spacing. Figure 8 compares the electrode in its initial state with its condition after 193 hours of operation at various gap distances. For enhanced comparative analysis, 232 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="233"?> the figure includes overlays (A-D 2), where the new electrode is visualized together with a binary mask representing the corresponding center electrode at each gap distance. These masks were generated using a custom MATLAB script, which processes microscopic images through binary thresholding to segment the electrode, followed by color-coded visualization for improved contrast and analysis. The platinum electrode shows the most severe wear at the smallest gap, where the platinum tip is almost completely removed by erosion (B1-2). At larger spacings, the wear of the platinum tip decreases notably, but a slight bevel forms on the electrode surface. Fig. 8: A-D 1: Light-microscopy images of the platinum center electrode at different electrode gaps, comparing the initial state and after 193 hours of operation; A-D 2: Comparative analysis of the new electrode with generated erosion masks after 193 hours Figure 9 shows the top views of a new platinum electrode and electrodes after 8 houres of ignition sparks at different electrode distances. Platinum tends to form spherical droplets and crater-like structures due to the particle-ejection process, where molten droplets are ejected and solidify upon redeposition. This effect is influenced by material properties such as surface tension. High surface tension suppresses droplet detachment, while low surface tension facilitates it. As the electrode spacing increases, the tendency toward spheroid formation decreases. At the largest gap, the electrode surface closly resembles the new electrode, indicating minimal wear. However, top views do not reveal radial material erosion occurring near theceramic. Figure 10 visualizes the spark intensity within the electrode gap, showing that intensity decreases as the gap increases. At the smallest gap, the stable and focused spark results in concentrated high energy input at the spark footpoint. This increased energy input and the associated rise in local material temperature leads to a more frequent occurrence of the particle-ejection mechanism described above. This is particularly evident in the pronounced crater-like surface structures. Additionally, a grayish layer forms across the entire surface, which is presumably platinum oxide. This layer forms at high temperatures in the presence of sufficient oxygen, and its vaporization contributes to electrode erosion. The influence of operating parameters on spark plug wear and its prediction potential 233 <?page no="234"?> Electrode Gap Increase 8h 8h 8h 0,1 mm 0,1 mm 0,1 mm 0,1 mm 0,1 mm 0,1 mm 0h Fig. 9: 8-hour endurance run, top view of the platinum center electrode under varying electrode gap According to prior research, the formation and evaporation of an oxide layer is the primary wear mechanism. In areas near the ceramic, a radial notch forms in the material, indicating a different wear mechanism in nickel, though it does not exhibit droplet-shaped or spherical structures, potentially involving oxide layer formation and removal or direct vaporization. Sputtering is unlikely due to the presence of the oxide layer. The location of the eroded material depends strongly on the spark location, which varies with the gap size. 1 0,8 0,6 0,4 0,2 0 Electrode Gap Increase Fig. 10: Radiation intensity of the spark discharge post-breakdown under varying electrode gap, platin electrode Nickel At the smallest gap, the spark predominantly jumps between the ground electrode and the platinum cylinder, while at lager gaps, sparks occur laterally on the center electrode and the ground electrode. The results indicate that the electrode gap has a significant influence on the wear behavior, enabling associated effects such as reaching the material’s melting temperature. Nickel The same wear behavior observed for platinum was also noted for nickel. Erosion of the center electrode increases as the distance to the ground electrode decreases. Figure 11 compares the surface and geometry of a new electrode (A 1) with electrodes after a 193hour endurance test at various gaps (B-D 1). Overlays (A-D 2) combine the new electrode with binary masks of the worn electrodes at each gap distance for enhanced comparison. 234 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="235"?> Fig. 11: A-D 1: Light-microscopy images of the nickel center electrode at different electrode gaps, comparing the initial state and after 193 hours of operation; A-D 2: Comparative analysis of the new electrode with generated erosion masks after 193 hours At a small gap, the electrode surface is heavily eroded, taking on a distinct rounded shape. This rounding decreases with increasing electrode gap. At the largest gap, the cylindrical center electrode shows clear signs of wear on the mantle surface (see Fig. 11), suggesting that the spark discharge in our experiments does not occur exclusively at the end surface but predominantly on the lateral surface, likely due to the increased spark gap. If the gap becomes too large, the ignition system’s energy may be insufficient to ionize the gas volume along the entire electrode gap and form a stable conductive plasma channel. Consequently, spark breakdown tends to shift to areas with a locally smaller gap from a discharge point of view, leading to lateral material removal on the mantle surface. This results in a distinctly rounded edge on the side facing the hook (C-D 1), caused by the accelerated formation of an ionized channel in the areas with reduced distance. Figure 12 presents a top view of the mass electrode in its initial state and after an 8-hour endurance test. The surface shows small craters alongside manufacturing-related grooves. These craters, visible in side views of Figure 11, likely resulting from spark footpoint interactions. The formation and evaporation of an oxide layer at the spark footpoint creates small holes or grooves, while the dark layer within these depressions indicates the presence of an oxide layer. This indicates the spark footpoint mechanism at work. Additionally, a notably darker gray layer covers the entire surface during the endurance test. Given the high-temperature, oxygen-rich conditions, this is likely an oxidation layer, futher supporting the wear mechanism involving oxid formation and vaporization. The influence of operating parameters on spark plug wear and its prediction potential 235 <?page no="236"?> Fig. 12: 8-hour endurance run, top view of the nickel center electrode under varying electrode gap Figure 13 shows the spark’s radiation intensity at various distances. Increasing the electrode gap reduces the spark intensity between the electrodes, indicating a reduced population of excited radiating species within the spark plasma. Electrode Gap Increase 1 0,8 0,6 0,4 0,2 0 Fig. 13: Radiation intensity of the spark discharge post-breakdown under varying electrode gap, nickel electrode According to Paschen’s law, the breakdown voltage increases with increasing distance, potentially causing instability or failure of the spark. Consequently, less energy is available for the discharge process. The temperatures are no longer sufficient to vaporize the oxide layer and the material at the root. Therefore, material-removing wear decreases as the gap width increases. Sputtering cannot be ruled out either, especially at smaller gaps. It leads to higher electric field strengths in the gap, allowing more of the ions’ kinetic energy to be transferred to the electrode surface through stronger acceleration. This process increases the number of collisions with sufficient energy to detach atoms from the crystal lattice. Evaporation and melting of the entire electrode surface as a wear mechanism is highly unlikely, since the wear occurs very locally, in areas of spark discharge (see Figure 12). The particle-ejection model is also improbable, given the absence of typical droplet formation. Instead, localized overheating at the spark footpoint likely drives oxide layer formation and vaporization, the dominant wear mechanism. Smaller gaps concentrate this energy, amplifying erosion in the form of grooves and craters. 3.2 Variation coil activation duration In addition to the electrode gap, the ignition coil’s activation duration is another key factor influencing electrode wear. 236 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="237"?> Fig. 14: A-D 1: Light-microscopy images of the nickel center electrode at different coil activation duration, comparing the initial state and after 193 hours of operation; A-D 2: Comparative analysis of the new electrode with generated erosion masks after 193 hours The activation duration determines the amount of energy stored in the ignition coil and subsequently made available to the electrodes during the discharge process, thereby affecting thermal and electrical stress on the electrode surfaces. Experimental results show that longer activation durations lead to increased material removal for both nickel and platinum (see Fig. 14 and 16). In contrast, shorter activation durations significantly reduce wear on the nickel electrode, this effect is particularly noticeable in the upper region toward the ground electrode (see Fig. 14). Coil Activation Duration Increase 1 0,8 0,6 0,4 0,2 0 Fig. 15: Radiation intensity of the spark discharge post-breakdown under varying coil activation durations, nickel electrode Figure 15 and 17 illustrates the temporal evolution of spark radiation intensity following the spark breakthrough. A reduction in activation duration lowerd spark intensity and shifts the spark footpoint, decreasing energy input and thermal loads. Consequently, the temperatures necessary for material formation and evaporation, as well as for oxidation and evaporation, are not achieved. Furthermore, energy availability for ion acceleration is diminished, causing ions to possess insufficient kinetic energy to displace atoms from the crystal lattice. This reduction also weakens the platinum particle-ejection mechanism. As a result, overall wear is markedly reduced. For platinum, a longer activation duration produces a distinct notch near the ceramic, accompanied by more pronounced rounding (see Figure 16), The influence of operating parameters on spark plug wear and its prediction potential 237 <?page no="238"?> where as a shorter duration eliminates it. The rounding of the upper edges of the electrode is less pronounced in comparison, thereby indicating reduced erosion at lower coil activation durations. The platinum electrode exhibits minimal visible wear. Fig. 16: A-D 1: Light-microscopy images of the platinum center electrode at different coil activation duration, comparing the initial state and after 193 hours of operation; A-D 2: Comparative analysis of the new electrode with generated erosion masks after 193 hours A grayish layer, likely an oxidation film, has formed on the electrode surface. With a shorter activation duration, the layer is thinner, appearing as a faint film, and the electrode tip shows no detectable wear, indicating insufficient spark intensity (see Figure 18). The spark channel is low-intensity with a shorter duration, and the moving spark footpoint distributes energy more evenly, reducing localized thermal stress. 1 0,8 0,6 0,4 0,2 0 Coil Activation Duration Increase Fig. 17: Radiation intensity of the spark discharge post-breakdown under varying coil activation durations, platin electrode Consequently, peak temperatures required for oxidation are not attained, nor are the melting and boiling points sufficiently reached, thereby limiting wear mechanisms, such as oxidation layer formation, evaporation, overall surface evaporation, and the spark footpoint erosion. The effectiveness of the particle-ejection model and sputtering-induced erosion is significantly restricted due to lower available energy in the system. These observations apply equally to the nickel electrode. 238 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="239"?> Coil Activation Duration Increase 0,1 mm 0,1 mm 0,1 mm 8h 0,1 mm 0,1 mm 0,1 mm 8 h 0,1 mm 0 h Fig. 18: 8-hour endurance run, top view of the platinum center electrode under varying coil activation durations After an 8-hour endurance test, electrodes exposed to shorter activation durations exhibit minimal structural changes beyond production-related features, such as minor spherical formations (Figs. 18-19), further confirming reduced wear. The wear mechanisms, discussed in section 3.1, persist in a weakened form under these conditions. Fig. 19: 8-hour endurance run, top view of the nickel center electrode under varying coil activation durations 4 Summary and Outlook Platinum and nickel spark plug electrodes were tested under engine like physical condition but without flame and fuel interaction at varying electrode gaps to evaluate wear behavior. Platinum spark plug electrodes show the highest wear at small gaps, where focused spark energy leads to severe erosion, crater formation, and near-complete material removal. As the gap increases, wear decreases significantly, indicating a strong dependence on spark intensity and energy distribution. The particle ejection mechanism intensifies due to high local energy input and temperatures at the spark footpoint, leading to pronounced crater-like surface structures. Additionally, a grayish platinum-oxide layer forms on the electrode surface under high-temperature conditions with sufficient oxygen, and its vaporization contributes significantly to material erosion. The location and intensity of wear are strongly influenced by the electrode gap and spark position. For platinum electrodes, melting and evaporation of the entire surface is unlikely, as material removal is highly localized to spark discharge zones. For nickel spark plug electrodes, one primary wear mechanisms dominates, localized erosion at the spark footpoint, where excessive temperatures cause oxide layer vaporiza- The influence of operating parameters on spark plug wear and its prediction potential 239 <?page no="240"?> tion and material removal. Maximum wear occurs at small electrode gaps, where concentrated spark energy leads to pronounced rounding and crater at the ceramic interface. At larger gaps, erosion shifts to the side surfaces due to insufficient ionization, reducing overall wear intensity. Additionally, sputtering contributes to material degradation by accelerating ions that detach atoms from the electrode lattice, particularly in tight gaps where electric field strength peaks. Evaporation or melting of the electrode surface is unlikely, as wear is confined to highly localized spark discharge regions. Similarly, the particle-ejection model is largely inapplicable, given the absence of typical droplet formation. The ignition coil’s activation duration significantly influences electrode wear, as it determines the energy available during the discharge process and, consequently, the thermal and electrical stress on the electrode surfaces. Experimental results demonstrate that a longer activation duration leads to increased material removal at the electrodes for both nickel and platinum. Even at shorter activation durations, nickel electrodes exhibit significantly higher wear compared to platinum, as evidenced by more pronounced structural changes and erosion. The observed wear mechanisms, such as material evaporation, oxidation, and particle ejection, are intensified by higher discharge energies associated with longer activation durations. Conversely, shorter activation durations limit these processes, as evidenced by the minimal structural changes observed on the electrode surfaces. These findings align with the reduced spark intensity, lower thermal loads, and weakened wear mechanisms reported in this study. The results confirm that optimizing coil activation duration can substantially mitigate electrode wear and extend component lifespan. Building upon prior studies that investigated influencing factors such as oxygen content, further research could analyze these variables using real spark plugs to derive more precise insights into wear mechanisms. Additionally, long-term endurance tests with pure nickel and platinum electrodes may help eliminate potential cross-influences of intermetallic trace minerals and validate observed trends. The electrode models reported in the literature are only limitedly transferable to the behaviour of real spark plugs, since the increased complexity of the overall system is not fully captured in these models. The goal is the development of a predictive wear model that quantitatively describes input parameters and their impact, enabling accurate prediction of wear behavior under defined operating conditions. References [1] J. Rager, Funkenerosion an Zündkerzenelektroden (Zugl.: Saarbrücken, Univ., Diss., 2006) (Saarbrücker Reihe Materialwissenschaft und Werkstofftechnik Bd. 5). Aachen: Shaker, 2006. [2] E. W. Gray and J. R. Pharney, “Electrode erosion by particle ejection in low-current arcs,” Journal of Applied Physics, vol. 45, no. 2, pp. 667-671, 1974, doi: 10.1063/ 1.1663300. [3] S. Baus, Funkenerosionsmodell von Nickelbasislegierungen (Zugl.: Saarbrücken, Univ., Diss., 2011) (Saarbrücker Reihe Materialwissenschaft und Werkstofftechnik Bd. 24). Aachen: Shaker, 2011. 240 M.Sc. C. Krnac, Dr.-Ing. O. Toedter, M.Sc. M. Grüninger and Prof. Dr. sc. techn. T. Koch <?page no="241"?> [4] F. L. Jones, “Electrode Erosion by Spark Discharges,” Br. J. Appl. Phys., vol. 1, no. 3, pp. 60-65, 1950, doi: 10.1088/ 0508-3443/ 1/ 3/ 302. [5] W. Demtröder, Experimentalphysik 2. Berlin, Heidelberg: Springer Berlin Heidelberg, 2017. [6] G. Bräuer, “Sputtering,” in CIRP Encyclopedia of Production Engineering, L. Laperrière and G. Reinhart, Eds., Berlin, Heidelberg: Springer Berlin Heidelberg, 2014, pp. 1138-1143. The influence of operating parameters on spark plug wear and its prediction potential 241 <?page no="243"?> Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing Albert Guggenheimer, Maurice Kettner, Karlsruhe University of Applied Sciences Steffen Kuhnert, Nicolas Wippermann, multitorch services GmbH Abstract: In-cylinder pressure is essential for combustion analysis and closed-loop combustion control. Conventional piezoelectric transducers require dedicated combustion-chamber access and involve considerable cost and integration effort. This paper investigates a strain-gauge spark plug (SGSP) for indirect in-cylinder pressure measurement based on the elastic deformation of the spark-plug housing. The deformation is detected by a sputtered thin-film full bridge on a 0.3 mm Inconel plate welded into the housing. The SGSP is evaluated in quasi-static pressure tests, a constant-volume combustion chamber (CVCC) and a single-cylinder engine, where its signal is compared with a piezoelectric reference transducer under motored and fired operation. The quasi-static tests detect housing strain during tightening and loosening as well as a distinct response to hydraulic pressurization. In the CVCC, the SGSP follows the dynamic pressure rise but exhibits operating-point-dependent sensitivity and a slowly varying negative signal component consistent with thermally induced strain. In the engine, good agreement with the reference is achieved at 0.7 MPa indicated mean effective pressure (IMEP), with a mean coefficient of determination of 0.988 and a mean root-mean-square error of 0.096 MPa. CA50 (crank angle with 50 % apparent heat release) differs from the reference by less than 0.5 °CA at 2200 and 2500 min -1 . At 0.3 MPa IMEP and under motored operation, electronic noise and ignition interference prevent reliable cycle-resolved evaluation. The results demonstrate the feasibility of deriving pressure and combustion-phasing information from spark-plug housing strain when the pressure-induced signal sufficiently exceeds the interference level. 1 Introduction In-cylinder pressure is one of the most important parameters in the development and calibration of internal combustion engines. It gives direct access to the combustion process and allows the calculation of the heat release rate, the indicated mean effective pressure and characteristic combustion parameters such as the apparent heat release. Cylinder pressure is also the key input for closed-loop combustion control, which allows the engine to be operated closer to its efficiency and knock limits [1]. These parameters can be computed in real time by integrated electronics [2]. <?page no="244"?> The most common in-cylinder pressure sensors are piezoelectric transducers. They are costly and require a dedicated bore into the combustion chamber and are therefore mainly used in engine development and on engine test benches. For series applications, the sensor should instead be integrated into a component that already has access to the combustion chamber. In diesel engines, this approach is established in the form of the pressure sensor glow plug, where a piezoresistive element measures the force transmitted through the movable heater rod [3]. In spark-ignition engines, the spark plug offers a comparable access to the combustion chamber. Since the spark plug already provides access to the combustion chamber, it has long been considered as a carrier for a pressure sensor. Different measuring principles have been proposed, which can be divided into three groups by the location and the effect that is measured. In the first group, a dedicated sensor is placed next to the center electrode and is in contact with the combustion gases. The pressure acts on a small diaphragm and is converted into an electrical charge by a piezoelectric element [4] or into a change of reflected light intensity by a fiber-optic element [5]. In the second group, the pressure is derived from the change of the thread preload. The combustion pressure reduces the force in the thread connection between the spark plug and the cylinder head, which unloads a pre-stressed element at the spark plug seat. This has been shown with a piezoelectric sensor that replaces the spark plug gasket [6, 7] and with a pre-stressed element integrated into the spark plug seat [8]. In the third group, the pressure is derived from the elastic deformation of the spark plug housing. The combustion pressure acts on the insulator and presses it against a shoulder in the upper housing, which creates an elastic strain in the housing. This strain can be measured with a pre-stressed piezoelectric element that is unloaded by the deformation [9] or with a fiber-optic strain element in a groove of the housing [10]. The same effect could be measured with strain gauges on the housing surface. This strain gauge principle has been patented for spark plugs [11, 12]. Strain-based pressure measurement has also been shown on other engine components. The in-cylinder pressure has been derived from the strain on the outer cylinder wall [13], from the load on an engine stud [14] and from the compression of a force rod in the cylinder head [15]. In these concepts, the sensor measures the deformation of the engine structure. In a spark plug with integrated strain measurement, the load path lies inside one component. In all these strain-based concepts, the mounting preload and the thermal expansion superimpose the pressure-induced strain and have to be considered in the signal evaluation. Other indirect methods use the engine block vibration, the change in crankshaft speed or the ion current at the spark plug to estimate combustion parameters such as knock intensity or combustion phasing [7, 16]. The aim of this work is to investigate whether the pressure that acts on the front of the insulator can be determined from the elastic deformation of the spark plug housing. A conventional spark plug is used, in which a thin-film strain gauge full bridge is sputtered on an Inconel sensor carrier. The sensor is tested on three test benches. Quasi-static pressure tests give the basic sensitivity. Combustion tests in a constant-volume combustion chamber show the response to a dynamic pressure rise without a static thermal load. On a single-cylinder engine, the signal is compared with a piezoelectric reference transducer under motored and fired operation. 244 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="245"?> 2 Experimental Setup 2.1 Spark plug with integrated strain measurement The spark plug housing consists of an upper and a lower section. The insulator is clamped between both sections by the hot-lock process, creating a mechanical preload along the housing. The cylinder pressure acts on the insulator nose and creates an axial force that presses it against the upper housing shoulder, causing elastic tensile strain in the upper housing section. The strain in the upper housing section is therefore a measure of the cylinder pressure. The strain measured at this position comprises a static strain resulting from spark-plug tightening, a slowly varying thermally induced strain and a cyclic pressure-induced strain. The static and thermal strains are substantially larger than the pressure-induced strain. The static strain represents a fixed installation offset, whereas the thermal strain produces a temperature-dependent offset. The comparatively small cyclic strain contains the pressure-related measurement signal. Fig. 1: Spark plug with sensor carrier To measure this strain, a pocket was machined into the upper housing section to accommodate a sensor carrier. This strain sensor consists of a sputtered thin-film strain gauge full bridge on a 9 x 5 mm Inconel 718 carrier with a thickness of 0.3 mm. The carrier is laser-welded into the pocket of the housing, as illustrated in Figure 1. The figure shows a top view and a projected side view of strain gauge spark plug (SGSP), with detail Z showing and enlarged view of the integrated sensor. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 245 <?page no="246"?> Since the pressure-induced strain is very small, the bridge signal is amplified by an amplifier circuit with a nominal gain of approximately 600. A shielded cable supplies the bridge and transmits the measurement signal. To reduce the electromagnetic interference by the ignition discharge, the initial tests were performed with a reduced electrode gap. 2.2 Static hydraulic pressure chamber The quasi-static tests represent the first validation step and isolate the mechanical pressure response from combustion-induced thermal and electromagnetic effects. Recording the complete sequence of tightening, pressure loading and loosening allows the installation-related offset to be distinguished from the pressure-induced voltage change. The tests therefore assess the basic pressure sensitivity, repeatability and reversibility of the sensing principle under controlled conditions. Static pressure tests were performed on a hydraulic test bench to evaluate the sensor response without dynamic pressure fluctuations or thermal loads. The test bench consists of a closed pressure vessel with a threaded port matching the spark plug thread (M10x1.0). The SGSP was installed and tightened to the specified torque. Hydraulic oil was then supplied to the vessel, and the static pressure was increased up to 30 MPa with a hand pump. The pressure level was monitored with a mechanical pressure gauge. The sensor signal was recorded during tightening, during the step-wise pressure increase and during the subsequent loosening of the spark plug. In each pressure ramp, the pressure was held constant at 10, 20 and 30 MPa for 10 seconds. 2.3 constant volume combustion chamber Dynamic pressure measurements without engine-specific interference were performed in a constant volume combustion chamber (CVCC) at Karlsruhe University of Applied Sciences (HKA). The vessel has a volume of approximately 580 cm³ and a maximum operating pressure of 25 MPa. It provides ports for gas supply, gas exhaust, a piezoelectric reference pressure transducer (Kistler 6125C) and the spark plug. All tests were initially performed at ambient temperature. The vessel was first evacuated with a vacuum pump. Hydrogen was then added, followed by compressed air until an air-fuel equivalence of 1.26 was reached. Three mixtures were tested by setting the hydrogen partial pressure to 0.05, 0.1 and 0.2 MPa. The initial pressure in the vessel therefore increases with the hydrogen partial pressure, which varies the peak combustion pressure. After closing all valves, the mixture was ignited by the SGSP. The amplified sensor signal and the reference pressure were recorded synchronously with a DEWETRON DEWE-800-CA indication system at a sampling rate of 200 kHz. After each measurement, the exhaust valve was opened to vent the vessel. 2.4 Engine test bed Engine tests were performed on a naturally aspirated single-cylinder research engine based on a Hatz 1B20 diesel engine, which was modified for spark-ignition operation with gasoline port fuel injection. The original injector bore was re-machined into a spark plug shaft. The compression ratio is 11.4. The engine is coupled to a four-quadrant electric dynamometer, which allows both motored and fired operation. A real-time ADwin-Pro system running custom control code 246 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="247"?> manages ignition and injection timing at an execution frequency of 125 kHz. The coolant outlet temperature was controlled to a constant 70 °C at all operating points. The air-fuel equivalence was kept at 1.0 and the combustion phasing was set to 13 °CA aTDCF at all fired operating points. High-speed data were recorded with the DEWETRON indication system at a crank angle resolution of 0.1 °CA. These include the cylinder pressure from the piezoelectric reference transducer (Kistler 6125C), the amplified SGSP signal, the ignition signal and the injection pulses. Measurements were performed at engine speeds of 2200, 2500 and 2800 min -1 under motored conditions and at fired operating points with an indicated mean effective pressure (IMEP) of 0.3 MPa and 0.7 MPa. Each measurement comprises 300 consecutive engine cycles recorded at a crank angle resolution of 0.1 °CA. Unless stated otherwise, all curves are cycle averages over these 300 cycles. 3 Results 3.1 Static pressure test The static hydraulic measurement recorded the complete mechanical strain history during installation, pressure loading and removal of the SGSP. In this measurement chain, a negative output voltage corresponds to compression within the spark plug housing, whereas a positive voltage shift denotes tensile strain caused by pressure. As shown in Figure 2, tightening the SGSP (t = 8 to 10 s) creates a large compressive strain. The detailed view of the loading sequence in Figure 3a shows three consecutive pressure cycles with plateaus at 10, 20 and 30 MPa, followed by a sudden pressure release to 0 MPa. 0 50 100 Time in s 4 2 0 U SGSP in V Tightening Pressure Loosening Fig. 2: SGSP signal over tightening, pressure increase and loosening To evaluate the transfer characteristic over the pressure range, the incremental sensitivity was calculated for each pressure step. As shown in Figure 3b, the sensitivity across all loading steps lies between 12.5 and 14.2 mV/ MPa. The unloading step forms a narrow cluster around 12.5 mV/ MPa, which shows that the signal during pressure release scales in the same way as during loading. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 247 <?page no="248"?> 25 50 75 100 Time in s 3.8 3.6 3.4 U SGSP in V (a) 0 10 10 20 20 30 30 0 Pressure step in MPa 12 14 16 Sensitivity in mV/ MPa (b) Single cycles Mean ± SD Fig. 3: (a): detailed pressure ramps; (b): boxplot of Sensitivity over pressure ramp The unloading phases in Figure 3a show a zero-point hysteresis. After the first two pressure cycles, the signal does not return completely to the initial baseline of −3.92 V and remains offset by approximately 70 mV. This residual shift is probably attributed to mechanical settling under high static forces rather than to plastic deformation of the housing. This is supported by the disassembly step in Figure 2 at t = 120 s, where the signal returns to its initial baseline of around 0 V after loosening the spark plug. No permanent offset of the unloaded measurement chain was detected after removal of the spark plug. 3.2 Constant Volume combustion chamber The SGSP was tested in the CVCC at three operating points, referred to as OP1 to OP3, with hydrogen partial pressures of 0.05 MPa, 0.1 MPa and 0.2 MPa. Each operating point was repeated three times. All measurements of an operating point were aligned to the start of ignition, the raw signals were averaged and the filter was applied to the mean signal. The mean curve is the basis of the evaluation, while the individual measurements give the scatter. The signal was zeroed in an interval before the ignition happens. The interval in which the amplifier is saturated by the ignition was then blanked out and bridged by cubic Hermite interpolation. Fixed notch filters were applied against mains and switching harmonics, followed by a zero-phase Butterworth low-pass filter with a cut-off frequency of 500 Hz. The conversion of the SGSP voltage into pressure is based on a linear regression of the SGSP voltage against the reference pressure. The regression window starts after the end of the blanked interval and ends at the peak pressure. Figure 4 shows the mean curves of OP3 in the first 15 ms after the start of ignition. The reference pressure is shown in green, the filtered SGSP signal in blue and the raw SGSP signal in gray. The gray bar marks the blanked interval and the red dashed line the ignition trigger. The blanked interval covers the ignition event and ends before the onset of the pressure rise. The SGSP reproduces the pressure rise, but the signal leads the reference pressure and reaches its maximum earlier. After the maximum, the SGSP signal decreases more rapidly than the reference pressure. This deviation is probably caused by heating of 248 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="249"?> the spark plug, which induces thermal strain and may partially reduce the effective mechanical preload. 10 5 0 5 10 15 Time after start of ignition in ms 0 25 50 U SGSP in mV Pressure Heating p Piezo U SGSP raw U SGSP filtered Start of ignition 0 2 4 p Piezo in MPa Fig. 4: averaged reference pressure and SGSP signals at OP3 Figure 5 shows the same operating point over one second. After combustion, the SGSP signal drops to approximately −420 mV and recovers only partly within this period, whereas the reference pressure returns towards its initial level. The pressure-induced amplitude at OP3 is approximately 60 mV. The slow negative component therefore exceeds the pressure-induced component by about a factor of seven and acts in the opposite direction. Both components can be observed separately in the CVCC because they act on different time scales. Although it cannot be isolated during the pressure rise, the slow negative component observed after combustion may already contribute to the SGSP signal before peak pressure is reached. It may therefore reduce the SGSP voltage near peak pressure and affect the pressure estimate. 0 500 1000 Time after start of ignition in ms 400 200 0 U SGSP in mV Heating Cooling p Piezo U SGSP raw U SGSP filtered Start of ignition 20 0 p Piezo in MPa Fig. 5: averaged reference pressure and SGSP at OP3 over one second after start of ignition Figure 6 shows the SGSP voltage as a function of the reference pressure for the three operating points, together with the linear regressions of the mean curves. At all three operating points, the measurement points bend downwards at the upper end of the pressure range. This supports the hypothesis of the influence of the thermal component, Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 249 <?page no="250"?> which grows with the released heat and therefore acts most strongly at the highest pressures of each measurement. At OP1 and OP2, the effect is limited to the last part of the range and the regression describes the remaining points well. At OP3, the deviation from the regression line is larger over a wider range, which is reflected in an RMSE about three times higher than at OP2 and the fitted slope is considerably lower. Fig. 6: SGSP voltage over reference pressure; lines denote linear regressions of the mean curves Figure 7 shows the sensitivity of the individual measurements. OP1 and OP2 agree within the scatter of the repetitions, whereas OP3 is considerably lower. Two effects change simultaneously between the operating points. The peak pressure increases and the higher amount of hydrogen releases more heat and raises the temperature of the housing. This behavior is consistent with an additional thermally induced strain component. However, pressure-dependent nonlinearity and thermal effects cannot be separated with the present measurements. OP1 OP2 OP3 Operating points 0 10 20 Sensitivity in mV/ MPa Single measurements Mean ± SD Fig. 7: SGSP sensitivity at the three operating points Figure 8 compares the peak pressure of both signals. The SGSP underestimates the peak pressure by 9.0 % at OP1 and by 6.4 % at OP2 and overestimates it by 19.0 % at OP3. The overestimation at OP3 follows from the reduced slope of the regression. Since the peak pressure is calculated from the maximum SGSP voltage and this slope, a low slope results in a higher pressure. The scatter of the repetitions is small at OP1 and OP2 and increases at OP3. 250 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="251"?> OP1 OP2 OP3 Operating points 0 2 4 p max in MPa p max, Piezo p max, SGSP Fig. 8: Peak pressure obtained from the reference transducer and the SGSP at the three operating points The CVCC results show that the SGSP follows a dynamic pressure rise and that the transfer characteristic is close to linear over most of the pressure range at the two lower operating points. At the highest operating point, the sensitivity decreases and the deviation from the linear fit increases. The tests also show the magnitude of the thermal component that is added by the combustion. This component is not a constant offset. It grows with the released heat, develops within the same time as the pressure rise and decays slowly afterwards, as the recovery over one second shows. 3.3 Engine tests The subsequent engine tests therefore examine whether the pressure-induced strain component remains recoverable under periodic operation, engine vibration and ignition-related electromagnetic interference. Signal processing was applied cycle by cycle. First, the SGSP signal was zeroed using a low-pressure interval of the respective engine cycle. As in the CVCC evaluation, ignitionaffected samples were blanked before zero-phase filtering. However, the periodic engine data and the two-part interference pattern required a cycle-resolved processing procedure with two separate blanking intervals. The first interval covered the short switching transient of the ignition coil, while the second covered the spark discharge and subsequent restrikes. The quieter interval between both transients was retained. Depending on the operating point, a total of 28.5 to 41.6 °CA was blanked. The blanked intervals were bridged by piecewise cubic Hermite interpolation, followed by a zero-phase Butterworth low-pass filter with a cut-off frequency of 1 kHz. Finally, a moving average was blended into the signal outside the high-pressure region. This procedure preserves the pressure gradients and the location of peak pressure around firing top dead center. Figures 9 to 12 compare the cycle-averaged reference pressure and SGSP signals at 2500 min -1 under motored operation and at 0.3 and 0.7 MPa IMEP. The reference pressure is shown in green, the processed SGSP signal in blue and the raw SGSP signal in gray. Gray bars indicate blanked crank angle intervals and the red dashed line marks the start of ignition. SGSP voltage and reference pressure are plotted on separate axes to compare the signal shape. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 251 <?page no="252"?> Fig. 9: Cycle-averaged reference pressure and SGSP signals at 2500 min -1 under motored operation. Under motored operation (Figure 9), the SGSP reproduces the compression and expansion sides of the pressure trace. Deviations are largest during gas exchange, where noise from the unshielded in-house developed circuit board is strongly amplified by the nominal gain of approximately 600. The pressure-induced signal amplitude in this phase is too small to reconstruct the gas-exchange loop. At 0.3 MPa IMEP (Figure 10), the SGSP signal is strongly disturbed. Both the coil charging and the subsequent discharge couple into the measurement chain. With a pressure span of approximately 1.9 MPa, this operating point has the smallest useful signal relative to the ignition-induced interference. Blanking, interpolation and filtering produce a continuous curve, but do not recover the reference pressure trace with sufficient accuracy. Fig. 10: Cycle-averaged reference pressure and SGSP signals at 2500 min -1 and 0.3 MPa IMEP At 0.7 MPa IMEP (Figure 11), the pressure-induced component dominates the remaining interference. The SGSP reproduces the pressure rise and the subsequent pressure decay without an apparent phase shift, and no substantial difference in gradient is visible after the blanked interval. Despite the interpolation across part of the rising pressure trace, the processed signal retains the dynamically relevant course of the cylinder pressure at this operating point. 252 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="253"?> Fig. 11: Cycle-averaged reference pressure and SGSP signals at 2500 min -1 and 0.7 MPa IMEP Figure 12 shows the cycle-averaged SGSP voltage as a function of the reference pressure together with its linear regression. Compression and expansion sides are shown separately, and reconstructed samples within the blanked windows are excluded. A separate linear regression is fitted to each individual cycle to determine its sensitivity and subsequently convert the corresponding SGSP signal back into pressure. The regression of the cycleaveraged transfer characteristic is used to convert the cycle-averaged SGSP signal. Under motored operation, the measurement points lie above the regression line at low pressure and below it at high pressure, which indicates a curved transfer characteristic. The same direction of curvature is visible at 0.7 MPa IMEP. At 0.3 MPa IMEP, the relationship is strongly distorted at all three engine speeds, whereas the 0.7 MPa operating points agree well with the linear fit. Fig. 12: SGSP voltage over reference pressure for the compression and expansion sides; lines denote linear regressions of the cycle-averaged transfer characteristics Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 253 <?page no="254"?> Figure 13 shows that R 2 increases with the covered pressure range and is therefore not suitable for comparing the absolute fit error between operating points. The RMSE describes the deviation of the cycle-averaged pressure-voltage relationship from the fitted linear model and therefore represents the accuracy of the linear calibration, not the cycle-to-cycle repeatability. The 0.7 MPa operating points provide the best fit not only in relative but also in absolute terms. Fig. 13: R2 and RMSE over the load points Figure 14 shows the cycle-wise sensitivity obtained by linear regression over the highpressure part of each cycle. At 2200, 2500 and 2800 min -1 , the median sensitivities were 36, 38 and 49 mV/ MPa under motored operation, 41, 63 and 60 mV/ MPa at 0.3 MPa IMEP and 89, 90 and 98 mV/ MPa at 0.7 MPa IMEP. The standard deviation ranged from 1.8 to 2.2 mV/ MPa under motored operation and from 1.7 to 2.3 mV/ MPa at 0.7 MPa IMEP, but increased to 7.6 - 26.7 mV/ MPa at 0.3 MPa IMEP. At this load, the ignition interference is large relative to the pressure-induced signal, and the observed scatter therefore cannot be interpreted as intrinsic sensor repeatability. The speed dependence was consequently evaluated only for motored operation and 0.7 MPa IMEP, for which the median sensitivity increased by 36 % and 10 %, respectively, between 2200 and 2800 min -1 . Fig. 14: Cycle-wise SGSP sensitivity at the investigated operating points 254 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="255"?> The static sensitivity of 12.5 - 14.2 mV/ MPa between 10 and 30 MPa was considerably lower than the engine-derived values. This is consistent with the observed curvature of the transfer characteristic, whose slope decreases towards higher pressures, while the amplifier gain was identical in all tests. The sensitivity ratio between 0.7 MPa IMEP and motored operation decreased from 2.6 to 2 with increasing engine speed. This apparent load dependence partly results from the different regression ranges, which extend to approximately 2.7 MPa under motored operation and 3.7 MPa at 0.7 MPa IMEP. For a curved transfer characteristic, the fitted slope depends on the evaluated pressure interval. Differences in housing temperature and pressure gradient may also contribute. Fig. 15: COV of peak pressure for the reference pressure transducer and the SGSP in % Cycle-to-cycle repeatability was assessed using the coefficient of variation of peak pressure, COV p,max . Figure 15 compares the values obtained from the piezoelectric sensor and the SGSP-derived pressure Figure 16 compares peak cylinder pressure, the location of peak pressure (ϴ p,max ) and CA50 for the individual cycles. Under motored operation, the reference transducer shows nearly constant peak pressure and ϴ p,max , whereas the SGSP signal form vertical point clouds. The ϴ p,max,SGSP exceeds the reference value by a factor of approximately 20 because the pressure-induced strain signal is small relative to the electronic noise. Consequently, a cycle-averaged pressure trace may appear plausible, even if the reconstruction of individual cycles is not repeatable. At 0.3 MPa IMEP, the peak pressure results at 2500 and 2800 min -1 lie well away from the diagonal, while the 2200 min -1 results scatter around it. The ϴ p,max results at 2200 and 2500 min -1 are distributed closer to the diagonal, whereas the 2800 min -1 results show a systematic positive offset. At 2200 and 2500 min -1 , the COV p,max, SGSP is about twice the reference value, which reflects the strong influence of the ignition interference on cycle-resolved metrics. At 0.7 MPa IMEP, peak pressure, ϴ p,max and CA50 correlate well with the reference values at all three engine speeds, and the COV p,max,SGSP is close to the reference value. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 255 <?page no="256"?> Together with the cycle-wise correlation in Figure 9, this indicates that the SGSP captures the magnitude of the real cycle-to-cycle variation. A comparable COV alone does not establish absolute pressure accuracy. Fig. 16: Cycle-resolved comparison of peak cylinder pressure, θ p,max and CA50 obtained from the piezoelectric reference transducer and the SGSP; the diagonal represents ideal agreement Figure 17 compares the combustion phasing derived from both signals. At 0.7 MPa IMEP, CA50 agrees within 0.5 °CA at 2200 and 2500 min -1 , and the cycle-wise scatter is close to that of the reference transducer. At 0.3 MPa IMEP, the scatter of the SGSP values increases strongly and reaches crank angles before firing top dead center at 2500 min -1 . The combustion phasing can therefore not be determined reliably at this load. 256 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="257"?> Fig. 17: CA50 from the cycle-averaged pressure traces of the reference transducer and the SGSP; error bars denote the standard deviation over 300 consecutive cycles At 2800 min -1 , the SGSP overestimates CA50 at both loads, by 6.3 °CA at 0.3 MPa and by 3.5 °CA at 0.7 MPa, while the deviation at the two lower speeds stays below 2 °CA. This follows the width of the blanked interval, which increases with engine speed. The spark duration is nearly constant and therefore covers a larger crank angle interval at higher speed. Part of the late compression and of the early pressure rise is consequently reconstructed by interpolation, so that CA50 depends on the interpolation and on the filter settings rather than on measured values alone. The combustion parameters derived from the SGSP signal should be interpreted with this in mind, in particular at the highest engine speed. Figure 18 compares the apparent rate of heat release (RoHR n ) and the normalized cumulative RoHR n derived from both pressure signals at 2500 min -1 and 0.3 and 0.7 MPa IMEP. The gray bars again mark the blanked crank angle interval. At 0.7 MPa IMEP, the RoHR n from the SGSP follows the reference in location and overall shape. At 0.3 MPa IMEP, it breaks down and becomes negative over parts of the evaluation window. As a consequence, the normalized cumulative RoHR n,SGSP also becomes negative before the onset of combustion and decreases again after reaching its maximum. This behaviour prevents a robust assessment of the combustion phasing at 0.3 MPa IMEP. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 257 <?page no="258"?> Fig. 18: RoHR n and norm. cum. RoHR n derived from the piezoelectric reference pressure and the SGSP signal at 2500 min -1 and 0.3 and 0.7 MPa IMEP The heat release evaluation starts before ignition and therefore integrates across the bridged interval of the SGSP signal. As Figure 11 shows, CA05 and CA10 fall within this interval at 0.7 MPa IMEP and rely on reconstructed values, whereas CA50 and CA90 lie outside the blanked window. Values within the blanked window must therefore not be interpreted as directly measured early-combustion information. 4 Summary and outlook The present SGSP and signal processing chain provided usable pressure traces and cycleresolved combustion metrics at 0.7 MPa IMEP. At this load, the linear transfer characteristic reached a mean coefficient of determination of 0.988 and a mean RMSE of 0.1 MPa over the three engine speeds, which corresponds to about 2.6 % of the peak cylinder pressure. The mean COV p,max was 5.4 % for the SGSP compared with 4.9 % for the reference transducer, and CA50 agreed within 0.5 °CA at 2200 and 2500 min -1 . Under motored operation, the pressure trace was still reproduced with a mean coefficient of determination of 0.930, although the individual cycles were not repeatable. At 0.3 MPa IMEP, the mean coefficient of determination dropped to 0.468 and the mean RMSE increased to 0.35 MPa, which is in the order of the pressure rise caused by the combustion itself. Neither the pressure trace nor the combustion phasing could be determined reliably at this load. The evaluation of the gas exchange loop was not possible at any operating point because of the noise level. At the lower load and under motored operation, the pressure-induced signal was too small relative to the interference. Two sources of interference dominate. The measurement electronics contribute broadband electronic noise, which is amplified together with the measurement signal by the nominal gain of approximately 600. In addition, the sensor carrier is located directly on the spark plug housing, so that the strain gauge bridge and its leads act as an antenna for the ignition discharge. Ignition transients of several volts acted on a pressure-induced signal of a few tens of millivolts at 0.3 MPa IMEP and caused amplifier saturation with recovery effects extending beyond the nominal spark duration. 258 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="259"?> Several approaches could improve these results. A further development of the measurement electronics towards a lower susceptibility to electromagnetic interference appears promising, as does a reduction of the coupling of the ignition discharge into the bridge, which would limit the extent of the blanked interval and therefore the share of reconstructed data. A higher strain at the measuring position would improve the signalto-noise ratio and would allow a lower gain, which in turn reduces the amplification of the remaining noise. The results at 0.7 MPa IMEP show that the cylinder pressure can be reproduced from the housing strain. Whether the principle is also applicable at low load cannot be assessed from the present measurements, since the signal quality there was governed by the measurement chain rather than by the sensing principle. Acknowledgement The present work was part of the project “CyPress”. The authors would like to thank the German Federal Ministry for Economic Affairs and Climate Action (BMWK) for the financial support of the project within the Central Innovation Programme for SMEs (ZIM). References [1] M. C. Sellnau, F. A. Matekunas, P. A. Battiston, C.-F. Chang, and D. R. Lancaster, “Cylinderpressure-based engine control using pressure-ratio-management and low-cost non-intrusive cylinder pressure sensors,” SAE Technical Paper 2000-01-0932, 2000. doi: 10.4271/ 2000-01-0932 [2] D. Vollberg, P. Gibson, G. Schultes, H.-W. Groh, and T. Heinze, “Smart in-cylinder pressure sensor for closed-loop combustion control,” J. Sens. Sens. Syst., vol. 11, pp. 1-13, 2022. doi: 10.5194/ jsss-11-1-2022 [3] M. Borgers et al., “Drucksensor-Glühkerze für Dieselmotoren,” MTZ - Motortechnische Zeitschrift, vol. 65, no. 11, pp. 888-894, 2004. [4] A. Bertola, R. Dolt, and J. Höwing, “The Use of Spark Plug Pressure Transducers for Engine Indication Measurements - Possibilities and Limits,” SAE Technical Paper 2015-26-0041, 2015. doi: 10.4271/ 2015-26-0041 [5] M. T. Wlodarczyk, “Fiber optic-based in-cylinder pressure sensor for advanced engine control and monitoring,” Combustion Engines, vol. 51, no. 3, pp. 3-8, 2012. [6] M. Kondo, A. Niimi, and T. Nakamura, “Indiscope - A New Combustion Pressure Indicator with Washer Transducers,” SAE Technical Paper 750883, 1975. doi: 10.4271/ 750883 [7] E. Corti, M. Abbondanza, F. Ponti, and L. Raggini, “The Use of Piezoelectric Washers for Feedback Combustion Control,” SAE Int. J. Adv. & Curr. Prac. Mobility, vol. 1, pp. 2217-2228, 2020. doi: 10.4271/ 2020-01-1146 [8] Y. Shimasaki et al., “Study on Engine Management System Using In-cylinder Pressure Sensor Integrated with Spark Plug,” SAE Technical Paper 2004-01-0519, 2004. doi: 10.4271/ 2004-01-0519 [9] A. L. Burrows and G. Tinwell, “Spark plug with combustion sensor,” U. S. Patent 9,304,059 B2, Apr. 5, 2016. [10] T. Bae, R. A. Atkins, H. F. Taylor, and W. N. Gibler, “Interferometric fiber-optic sensor embedded in a spark plug for in-cylinder pressure measurement in engines,” Applied Optics, vol. 42, no. 6, pp. 1003-1007, 2003. doi: 10.1364/ AO.42.001003 [11] Delphi, “Spark plug mounted thick film strain gauge,” U. S. Patent 6,834,538 B2, 2004. [12] N. Wippermann and S. Kuhnert, “Vorkammerzündkerze,” European Patent EP 4 542 016 A1, Apr. 23, 2025. Measurement of In-Cylinder Pressure through Elastic Deformation of the Spark Plug Housing 259 <?page no="260"?> [13] R. Amirante, C. Casavola, E. Distaso, and P. Tamburrano, “Towards the Development of the In- Cylinder Pressure Measurement Based on the Strain Gauge Technique for Internal Combustion Engines,” SAE Technical Paper 2015-24-2419, 2015. doi: 10.4271/ 2015-24-2419 [14] L. Romani, G. Lenzi, L. Ferrari, and G. Ferrara, “Indirect Estimation of In-Cylinder Pressure through the Stress Analysis of an Engine Stud,” SAE Technical Paper 2016-01-0814, 2016. doi: 10.4271/ 2016-01-0814 [15] D. Vollberg, D. Wachter, T. Kuberczyk, and G. Schultes, “Cylinder pressure sensors for smart combustion control,” J. Sens. Sens. Syst., vol. 8, pp. 75-85, 2019. doi: 10.5194/ jsss-8-75-2019 [16] N. Cavina, N. Rojo, A. Businaro, and R. Cevolani, “Comparison between Pressureand Ion- Current-Based Closed-Loop Combustion Control Performance,” SAE Int. J. Engines, vol. 12, no. 2, pp. 219-230, 2019. doi: 10.4271/ 03-12-02-0016 260 Albert Guggenheimer, Maurice Kettner, Steffen Kuhnert, Nicolas Wippermann <?page no="261"?> Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits Dimitrios Karageorgiou 1,* , Miguel Monserrat Diaz 1,2 , Matej Myslivecek 1 , Dr. Maxime Tarot 1,3 , Dr. Salim Sebai 1,3 Abstract: With the automotive industry challenged to meet CO2 reduction targets and contribute to the global effort of mitigating climate change, there is a need to use Internal Combustion Engine (ICE) technology that can simultaneously enhance performance and reduce GHG emissions. Active Turbulent Jet Ignition (TJI) is one of the main concepts to achieve these goals. It relies on a pre-chamber (PC) containing a sparkplug and an Air/ Fuel mixture injector which initiates the combustion. A flame kernel develops in the PC and hot burnt products are expulsed from PC to the Main Chamber (MC), thus accelerating MC combustion compared to conventional sparkignition engines through, (i) dispersed ignition spots (ii) MC turbulence production emerging from high PC jets velocity. These phenomena reduce the knock propensity and, when combined with EGR or lean dilution, they increase indicated efficiency. Nonetheless, flame transport from PC to MC through small orifices involves intense wall interaction and turbulent strain. This may lead to flame quenching and MC misfire, especially under cold start conditions: low load and late ignition. This operating point challenges the implementation of pre-chambers in passenger car ICEs, and this publication is part of an investigation aimed to better understand the physical fundamental reasons why turbulent jets fail to ignite the MC under these conditions. An active TJI system named Valvijet is vertically mounted on a Rapid Compression Machine (RCM). On the one hand, Valvijet consists of a check valve which separates PC and MC during the gaseous mixtures’ formations, thus allowing an improved control on their compositions and masses. On the other hand, the RCM allows (i) to investigate the phenomena at well-defined boundary conditions without unwanted perturbations, such as charge motion or turbulence; and (ii) to visualize the jets propagation, ignition and combustion using OH* chemiluminescence and Schlieren techniques. As a result, an analysis of a PC ignition timing sweep reveals that jets’ characteristics remain quite similar as PC ignition timing is varied. However, jets failure to ignite the MC become more frequent as PC ignition timing is retarded due to lower MC temperature and pressure which increases too much the heat transfer rate from the jets to the MC 1 Aramco Europe, 232 Av. Napoleon Bonaparte, 92852 Rueil-Malmaison Cedex, France * Corresponding Autor: dimitrios.karageorgiou@aramcoeurope.com 2 Davidson Consulting, 40 rue Fanfan la Tulipe, 92100 Boulogne-Billancourt, France 3 EMITECH - EMC France, 4 Allée de la Rhubarbe, Achères 78260, France <?page no="262"?> mixture. Furthermore, an analysis of a PC load sweep with constant late ignition timing proves there is an optimum PC energy which maximizes the chances of successful jet ignitions, proving the injected PC mass is one of the valuable levers to operate cold start with active pre-chambers. Finally, the TJI quenching hypothesis is formulated: for any MC state there is a range of combinations of three hot jets properties which maximize their chances to ignite the MC: turbulence, enthalpy (thermal energy), and concentration of active species. 1 Introduction In recent years, research and development in the transportation sector has focused primarily on reducing green-house gas (GHG) emissions across various technologies and associated supply-chains. The internal combustion engine (ICE) remains an integral part of the global transportation technology portfolio and recent developments show that new synergies can be found across technologies to improve the trade-off between GHGemission reduction potential, efficiency and cost. One of the enabling technologies for high-efficiency spark ignited (SI) engines is Turbulent Jet Ignition (TJI). An air and fuel charge burns in a passive or active prechamber (PC) and discharges hot, chemically active turbulent jets into the main chamber (MC) through one or more orifices. Compared to single ignition kernel of a spark plug (SP), these jets provide a large effective ignition surface and multiple distributed ignition sites, typically shortening combustion duration and thus improving tolerance to knock, enabling increase in engine geometrical compression ratio and in specific cases to dilution by air or recirculated exhaust gases (EGR) (1), (2). In passive pre-chambers (pPC), no auxiliary fuel is supplied. This simplicity makes passive designs attractive in terms of cost and robustness, and they have been shown to extend dilution limits and improve combustion in gasoline engines with small amount of hardware changes (3), (4). Optical studies have clarified the torch-ejection sequence and the strong influence of orifice layout on ignition development (3), and earlier dividedchamber work demonstrated the importance of spark location within the PC for pressure rise and jet characteristics (4). Despite the higher energy available for ignition for TJI compared to SP, certain operating points at the limits of MC mixture ignitability remain a challenge. One of them is the phase of catalyst heating. For modern vehicles with SI gasoline engines, regulated emissions are largely determined by the first tens of seconds after start before the three-way catalyst reaches its light-off temperature. Once it is attained, it converts HC, CO, and NOx with efficiencies typically above 95 % (5). The reduction of the duration of the phase of catalyst heating window is therefore decisive to comply with emission norms. SP is known to perform well, allowing to achieve a spark retard of 80 CAD of MFB50 or more, leading to sufficiently high exhaust enthalpy and fast rise in catalyst temperature, despite the challenging conditions of low MC charge temperatures, poor fuel vaporization and mixture preparation or residual/ wall-wetting effects. Operating the pPC at these conditions adds further limiting factors. Passive PC mixture can be colder, more stratified and globally more diluted by residuals than the mixture 262 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="263"?> around SP electrodes at the same time and same MC conditions, making ignition and subsequent PC combustion and jet development difficult. On top of that, cold PC surface and characteristics of the jet interaction with MC charge may further increase the risk of flame quenching during and after the jet ejection through orifices, necessitating measures in PC design that tend to shift the trade-off to less favorable engine performance at full load due to reduced knock mitigation potential. On the other hand, active PC adds an auxiliary injection of fuel or air and fuel directly into the PC, enabling mixture control near the spark electrodes to certain extent independently. The additional injection can improve ignitability and extends the ignitability range of the mixture inside pre-chamber, particularly at lean and low-load conditions. It has been demonstrated that active fueling strategies improve the MC combustion retard potential compared to passive pre-chamber, reaching similar performance as SP at low load (~ 2 bar IMEP net ) conditions (6), (7). Active PCs have been integrated into more advanced concepts such as pre-chamberinitiated SACI strategies (8) or high efficiency engine demonstrator using premixed airfuel PC injection system (7). These benefits come at the cost of added complexity (such as injector integration, control of small injected quantities). More fundamentally, even in active designs, the PC and MC remain coupled during engine compression phase due to continuous mass and enthalpy exchange through the PC orifices. The conditions inside PC at spark timing therefore reflect a combination of injected fuel, MC to PC in-flow, residual gas distribution and mixing and turbulence inside the PC (9), (10). Mixture control is improved relative to passive designs, nevertheless it remains imperfect. Fundamental studies show that TJI MC ignition by the jets is governed by the competition between chemical and mixing timescales. Depending on PC conditions and orifice geometry, a jet exhibits a continuum of ignition behavior between two limit cases: a reacting, radical-rich flame (“flame ignition”) at one extreme; at the other, it emerges as quenched jet of hot products that must re-initiate reactions downstream (“jet ignition”) (1). Intermediate regimes may show characteristics of both mechanisms. The transition is often described by the Damköhler number (1). Jet quenching within the orifice and possible re-ignition in the MC are now recognized as key to characterize the practical operating limits and sensitivity to dilution (10), (11). Across this literature, PC equivalence ratio repeatedly appears as a first-order lever affecting jet reactivity, ignition probability, and kernel location (1), (11). Quantifying these effects in firing engines is complicated by the continuous PC to MC exchange, motivating thermodynamic methods that separate PC and MC heat release using dual-pressure measurements (9). Prior work consistently identifies the pre-chamber mixture as a dominant control parameter for jet reactivity, the flame-versus-jet regime, and the ignition/ quenching boundary. However, existing passive and active configurations have not allowed its effect to be isolated. Consequently, how independent pre-chamber mixture conditions govern jet ignition versus quenching has remained experimentally inaccessible. This question is of general importance but is particularly relevant to catalyst-heating combustion, where the unfavorable conditions in the main chamber make successful jet ignition especially challenging. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 263 <?page no="264"?> To address these fundamental limitations, we characterize the TJI combustion on a rapid compression machine (RCM) setup using the Valvijet technology (12). Compared to permanently open orifices of a conventional active pre-chamber, this TJI system introduces a pressure driven valve, separating the pre-chamber from the main combustion chamber, thus enabling an independent mixture preparation. Using the RCM isolates the fundamental jet-ignition and quenching physics from engine-specific flow complexities and piston-motion effects, while enabling repeatable thermodynamic initial conditions (1), (8), (12). Using the RCM isolates the fundamental jetignition and quenching physics from engine-specific flow complexities and piston-motion effects, while enabling repeatable thermodynamic initial conditions (1), (8), (12). The work aims to characterize limits of combustion of an active TJI technology in the conditions of challenging ignitability of late combustion. Rapid compression machine coupled with OH* and Schlieren optical methods is used to understand how the combination of pre-chamber and main chamber conditions affect the jet characteristics and the propensity of jets to a) ignite the main-chamber charge, b) quench and reignite or c) quench. Pre-chamber with pressure driven valve is used to isolate both combustion chambers. It enables more precise mixture preparation and zero inflow from the main chamber during late compression stroke, thus further simplifying the consequent pre-chamber velocity and turbulence fields 2 Testing set-up and methodology The testing bench consists of an RCM coupled with an optical set-up able to record the main chamber with the Schlieren or OH* chemiluminescence techniques through a flat window on the piston head. There are two combustion chambers: the pre-chamber (PC) where combustion is initiated with a spark, and the main chamber (MC) where the combustion is aimed to be triggered by the jets emerging from the PC through small nozzles. The pre-chamber is vertically mounted on the cylinder, aligned with its axis, thus enabling the jets and MC ignition visualization through the piston window. A check valve separates the chambers, and it opens when the PC pressure overcomes with a certain margin the MC pressure. Figure 1 is a schematic of the PC tested: it has 6 nozzles of 0.8 mm diameter each and a volume of 700 mm 3 , so an A/ V ratio of 4.31 m -1 . The check valve stroke is 0.8 mm, and it is initially held closed by a magnet located above the spark plug (not shown). Spark plug Gas injector Piezoelectric pressure sensor Check valve (closed position) Figure 1: Schematic of the active pre-chamber (Valvijet) 264 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="265"?> The point of using a check valve is forming the mixtures separately in PC and MC, thus allowing (i) an improved control on their composition and mass with any PC ignition timing and (ii) a decoupling of the combustion process and the uncertainties linked to the mass transfer between the chambers prior to PC ignition. The MC is a cylinder of 925 mm 3 with flat piston and cylinder heads. This way, there are two injection systems, one per chamber. Both mixtures are injected gaseous, premixed and preheated. Isooctane and pure nitrogen are chosen as fuel and inert dilution gas (synthetic EGR) respectively. The inlet mixtures temperatures and final pressures at the end of the filling processes are controlled, and a forced vacuum precedes the mixture formation to eliminate the exhaust gases produced in the previous test. The MC mixture is formed with a slow filling process (17 seconds) before releasing the piston. Therefore, this initial charge is assumed to be quiescent and uniform in temperature and composition. The mass flow of air, fuel and nitrogen (if used) are measured with three sensors separately before mixing them upstream the MC. The PC mixture is constantly formed upstream the PC in a mixer device which guarantees the charge homogeneity and vaporization. A constant mass flow is established along the line through a secondary outlet to avoid any condensation and to guarantee the conditions upstream the PC injector are always the same. In the RCM used, the piston is released from a chosen position and pushed by the pressure built underneath in an oil chamber, leading to a repetitive compression and expansion piston displacement. This movement is ruled by the piston inertia and the difference of pressure in between the chambers below and above it. Figure 2(A) shows the machine connected to the PC and MC intake lines. Figure 2: RCM outside view connected to the intake PC and MC lines (A) Mirrored cylinder head of the RCM (B) Concerning the optical set-up, a phantom camera v710 is targeted to a 90 degrees mirror which allows to point at the piston window. This, added to mirrored surface of the cylinder head (Figure 2(B)), allowed the installation of a single mirror axes-coincident Schlieren visualization system, schematized in Figure 3 (12). Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 265 <?page no="266"?> The OH* chemiluminescence optical set-up also uses this same mirror to visually access the MC through the same piston window. Figure 3: Schematic of the testing setup. A light filter of wavelength 300 - 320 nm is placed in front of the intensifier, which opens with the same acquisition frequency as the camera, 51 kHz, leading to the highest exposure rate. The camera and amplifier are driven by a synchronizer, which also sends a different activation signal to start the PC coil charge and, later, to open the PC injector. The input signal required by the synchronizer to drive all these components is produced when the piston overcomes a certain chosen position. This input is calibrated to have the spark-plug electric arc at the desired MC pressure. This way, the PC ignition timing is given as the MC pressure at which the coil is discharged. Due to the nature and disposition of the techniques used, Schlieren and OH* chemiluminescence cannot be used at the same time. Figure 4 shows an example of two frames from two different tests captured with each optical system. In these examples, the PC jets are penetrating an MC filled with air. They correspond to #3 in Table 1. Figure 4: Schlieren picture (left) and OH* chemiluminescence picture (right) in the RCM during 20 μs of jet penetration. 266 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="267"?> Concerning the sensors, the pressure in MC and PC are measured with two fast acquisition piezoelectric sensors. The PC’s is installed radially, facing the spark plug (Figure 1), and the MC’s is installed in the cylinder head so that it faces the piston head perpendicularly. The injector current and coil voltage are also measured. Figure 5 shows the evolution of all these variables during a normal test. Concerning the methodology, every RCM test is performed as follows (example of reference in Figure 5): I. A vacuum pump empties the MC and PC until the pressure in the vacuum line inlet drops to 40 mbar. According to the MC and PC pressure sensors readings, the check valve stays open since shortly after the beginning of this process and until its end. Some air is injected in the MC only during the first seconds to guarantee the scavenging of the intake line and MC. II. The MC is slowly filled with the target mixture. It can be air, air and fuel, or air, fuel and extra nitrogen introduced to imitate the EGR of an actual engine. The target MC pressure is controlled with a fast valve and a slow valve to avoid an overshoot III. The oil below the piston is pressurized, shooting it in a quick compression. During it, the piston overcomes a certain input position, which sends the trigger signal to the synchronizer. As a result, this device sends immediately a continuous signal to charge the coil (see red curve starting to rise 5.5 ms before its discharge in Figure 5) and two identic PWM signals: one to the camera and another one to the intensifier. After a fixed time delay, it sends another continuous signal to actuate the PC injector until the input pulse width is over (see current rising in the green line of Figure 5). IV. The coil is discharged and the electric arc is produced, which ignites the PC mixture after a short delay. This moment can be identified when the coil tension drops back towards 0 V, at 0 ms in Figure 5. V. The PC combustion increases the PC pressure, and the check valve opens at some point after the instant in which the PC pressure overcomes the MC pressure (cross pressure point, at 0.5 ms in Figure 5), connecting the PC with the MC. This opening takes around 80 μs. a. According to previous studies (12), a jet of unburned PC mixture is expected to precede a jet composed by burned gases, active species, and/ or flame. They are named cold and hot jets respectively. Their sequential MC penetration is logical since the spark plug is on the top of the pre-chamber, so its fresh mixture is compressed towards the bottom check valve as combustion develops inside. This way, unburned gases exit the PC earlier than the flame and/ or combustion products. VI. The jet exit is visible with optical techniques as well as the successful MC ignition(s). Sometimes, the valve stays open until the end of the test, and some other times it closes at some point during the MC pressure rise due to MC combustion. Figure 5 is an example of the second case: the check valve closes again at 2.4 ms since the PC pressure is not following the MC pressure anymore. Right after, the PC pressure decreases due to heat transfer loses through the walls. Then, at some point during Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 267 <?page no="268"?> the MC expansion (at 4.5 ms in the example of Figure 5) the valve reopens because the MC pressure drops faster until it is lower than that of the PC (12). Figure 5: Sensor signals out of a normal test in the RCM. The operating point is #2 in Table 1. 2.1 Operating points The inputs are chosen to study the TJI quenching phenomena in catalyst heating conditions: • The MC initial pressure is set as low as possible (1 bar) to allow the piston to bounce back in an expansion with a speed similar enough to that of an engine. • The PC mixture composition was chosen to be phi=1.2 to (i) guarantee PC combustion, (ii) maximize its repeatability between tests and (iii) match the conditions of the available tests with a Single Cylinder Engine (SCE) run with Valvijet (data not presented in this publication). This phi was found to be the most reliable choice in the engine. • The MC mixture composition was just air in some tests and air and fuel at phi=1 in some others. They were chosen to study the PC jets characteristics and the interaction of these jets with a reactive mixture, respectively. Every operating point is repeated twice so that it can be recorded with both Schlieren and OH* chemiluminescence techniques. • Two sweeps were performed: - Timing variation of the electric arc to ignite the PC charge. The PC coil discharge is triggered during the expansion, this is, after top piston position, aTPP (equivalent to the TDC of an actual engine, aTDC) and it is progressively 268 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="269"?> retarded from a timing which guarantees MC ignition by all jets to another one in which MC ignition is impossible (MC misfire), so that this transition can be well understood. These are #3 - #12 in Table 1. - PC load variation. The pressure at the end of PC injection is varied to understand the effect of the injected PC energy on the jets’ characteristics and their consequent MC ignition chances and flame development. These are the only tests recorded only with OH* chemiluminescence. They were run with a PC ignition timing which delivers high chances of MC ignition failure with the PC load of the previous ignition timing variation sweep. These are #13 - #18 in Table 1. Two additional operating points were added to study the MC ignition at middle-high load. It uses 1.36 bar of initial MC pressure, PC ignition timing before the top piston position (bTPP) and 25 % of extra nitrogen (in mass) inside of the MC to imitate the inert gas effect of the EGR in a real engine. These are #1 - #2 in Table 1 # PC MC Test Nbr φinj Load [J] ST [bars MC] Reference operating point N2 [% in mass] φMC Pint [bar] Schlieren OH 1 1.2 23.2 20 (bTPP) Middle-high load air 0 1.36 Ok Ok 2 1.2 23.2 28 (bTPP) 25 1 3 1.2 20.1 28 (aTPP) Late combustion air 1 1 4 1.2 20.1 24 (aTPP) 5 1.2 20.1 20 (aTPP) 6 1.2 20.1 26 (aTPP) 7 1.2 20.1 22 (aTPP) 8 1.2 20.1 28 (aTPP) 0 9 1.2 20.1 24 (aTPP) 10 1.2 20.1 20 (aTPP) 11 1.2 20.1 22 (aTPP) 12 1.2 20.1 26 (aTPP) 13 1.2 22.0 22 (aTPP) 1 n.a 14 1.2 16.7 22 (aTPP) 15 1.2 14.7 22 (aTPP) 16 1.2 13.3 22 (aTPP) 17 1.2 9.3 22 (aTPP) 18 1.2 8.0 22 (aTPP) Table 1: RCM operating points list. Table 1 summarizes the definition of every operating point whose data is presented in this publication, including the optical technique used. There are 20 operating points and from 7 to 20 tests of each. In this publication, the operating points will be constantly referenced using the first column of this table (e. g. #3 for the third test listed in Table 1). Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 269 <?page no="270"?> The terminology STXX means the spark in the PC was triggered when the MC pressure reached XX bars, before Top Piston Position (bTPP) if the operating point is Middle-high load (#1 and #2) and after the Top Piston Position (aTPP) if it is Late Combustion (#3 - #18). 3 Analysis methodology This section aims to explain the post-processing techniques used in the investigation. 3.1 Optical post-processing: detection of jets and flame boundaries. The optical data has also been post-processed to obtain the jet exit timing, the evolution of the jet’s penetration and the flames areas, and the ignition locations and delays with respect to jet exit. The coordinates of each jet boundary were extracted with different thresholds of light intensity to be able to notice relevant differences on jets luminosity and to be able to overlap the boundaries, as shown in Figure 6. 3% 10% 30% 50% Schlieren + = Figure 6: Schlieren (top) and OH (bottom) contours from two similar tests. Pressure traces are shown on the right; the solid red line marks the image time, and dashed lines the first visible jet signal. It should be noted the jets observed with OH* are closer to the center of the PC than those seen with Schlieren. This is because there is a ring of non-mirrored cylinder head surface around the PC cap, so right above the nozzle’s outlet. Therefore, Schlieren signal cannot be obtained out of the first millimeter after the nozzle outlet. 3.2 Optical post-processing: detection of ignitions and application of colormap. A colormap was applied for the OH* videos to notice differences in light intensity, as shown in Figure 7. The highest value of the color map is 65535 because these are 16 bits videos and the lowest is 0. Ignition locations and timings are found out of the OH* videos only, and they are marked with red crosses which appear at their detection and stay fixed until the end of the video, as shown in Figure 7 too. The criteria to detect them consists of (i) finding blobs which overcome a combined threshold of pixels and intensity which grows for 100 μs after its detection and (ii) computing its centroid. The thresholds are chosen based on 270 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="271"?> comparisons in between the tests with the MC filled only with air (#8 - #12) and their homologues with reactive mixture (#3 - #7). Figure 7: Top: raw OH chemiluminescence frames. Bottom: post-processed frames showing ignition locations marked by red crosses. An ignition density probability map was built to find the locations and jets where ignition chances are the highest/ lowest. An example is available in Figure 8, where the central colormap has the size and shape of the piston window and shows the ignition locations probability density, and the outer annular segments indicate the probability of every jet to ignite the MC. The density color map is computed with a Gaussian KDE (Kernel Density Estimation) (2.3) whose bandwidth is calculated by adjusting the Scott’s rule with a factor (2.4). This method is no more than a simple sum of Gaussian distributions centered on every ignition point. f ð x Þ ¼ 1 N X N i ¼ 1 1 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð 2 �Þ d j H j q exp � 1 2 x � x i ð Þ T H � 1 x � x i ð Þ � � (2.3) H ¼ factor � H Scott ¼ factor � N � 1 3 � � (2.4) Where x and x i are any grid point and ignition coordinates respectively, N is the number of ignition locations, H is the bandwidth, H Scott is the baseline bandwidth calculated with the Scott’s rule for the 2D space case and Σ is the covariance matrix of the ignition locations coordinates. The factor was adjusted so that the ignition probability density in between the jets is null, because no ignition was ever observed in these locations so far. A factor of 0.25 was chosen for this study, and the difference with a factor of 1 is presented in Figure 8. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 271 <?page no="272"?> Figure 8: Gaussian KDE density map of ignition location probability This way, the KDE density is in mm -2 and its interpretation is relative. For example, in the second colormap of Figure 8, jet 4 wedge is found to have an area with values of 0.002 whereas jet 2 wedge has an area with values of 0.001 distributed along an axially larger surface. This means the chances of finding an ignition in this area of jet 4 are two times higher than those in the mentioned area of jet 2. The operating point, number of tests available and ignitions found to build each colormap are always indicated in the title. 3.3 Optical post-processing: detection of jets intensities. The intensity of the jets at each time step is evaluated out of the recorded videos with both techniques, Schlieren and OH* chemiluminescence, by computing the average radial intensities. The schematic of Figure 9 illustrates the calculation process: first the frames are split in wedges which are divided radially, then the average light intensity is calculated for each of these wedge portions. A B C Jet 2 of frame 4 A B C A B C Figure 9: Schematic to illustrate the calculation of the jets’ intensity colormaps. Therefore, every colormap corresponds to a jet of a test, every line corresponds to a time step (frame captured), and every column corresponds to a different portion of the given jet’s wedge. As shown in Figure 9 right, a per-jet average colormap is computed for every operating point. This way, jet penetration can be appreciated in the first 4~5 frames/ time steps with visible light in this example. 272 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="273"?> 3.3.1 PC and MC combustion repeatability The PC and MC combustion repeatability is evaluated using the dispersion of the pressure traces of every operating point, separately. This way, for any operating point, there are N pressure traces sampled at the same K times in between the PC ignition and the end of the positive difference of pressure (ΔP = P PC - P MC ) for the PC; and between the jets exit seen by the camera and the end of the MC combustion (MFB 100 %) for the MC (2.4). P i t j � �; i ¼ 1 ; . . . ; N (2.4) t ign PC < t j < t end � P > 0 ! PC t jet exit by camera < t j < MFB100 MC ! MC � Firstly, the ensemble-average pressure is calculated with (2.5). P t j � � ¼ 1 N X N i ¼ 1 P i t j � � (2.5) Next, the pointwise standard deviation is computed with (2.6). s P t j � � ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 N � 1 X N j ¼ 1 P i t j � � � P t j � � � � 2 vuut (2.6) This result is used to better visualize the pressure variation dispersion by plotting the traces with a shaded region defined as (2.7), with a percentile band of 5-95 % to better resist outliers. P ð t Þ � s p ð t Þ (2.7) Figure 10: Example of PC and MC combustion repeatability evaluation using the pressure traces of two tests. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 273 <?page no="274"?> An example is shown in Figure 10, in which the operating point on the left has a better MC and PC combustion repeatability than the one on the right. Then, this time-dependent standard deviation is collapsed into a scalar value: the RMS (Root Mean Squared) trace dispersion (2.8). � trace ; RMS ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 K X K j ¼ 1 s 2 P t j � � vuut (2.8) However, comparisons of this variable in between operating points run with different pressure levels would be biased. For example, when comparing MC combustions of points with different PC ignition timing, or when comparing PC combustions of points with different PC load (injected mass before ignition). To compensate for this, the RMS is normalized (2.9). D P ¼ 100 � trace ; RMS � P ref (2.9) � P ref ¼ max t P ð t Þ � � � min t P ð t Þ � � In the example of Figure 10, the normalized RMS of the operating point of the plot on the left is 9.1 % for PC and 4.5 % for MC, whereas the values for the plot on the right are 14.9 % for the PC and 18.5 % for the MC. The lower values, the better combustion repeatability. 4 Experimental results and discussion When igniting after aTPP, the PC jets characteristics and the MC conditions change as ignition timing is retarded in such a way the MC ignition chances reduce. Firstly, the jets properties are studied out of tests in which the MC is filled with air in section 4.1. (#8 - #12 in Table 1). Secondly, the MC ignition mechanism as well as its spatial and time variability is investigated using tests in which the MC is filled with reactive mixture in section 4.2. (#3 - #7). Thirdly, the input PC energy is better understood as a lever capable of allowing the jets to ignite the MC in section 4.3. (#13 - #18). Finally, a hypothesis is formulated to explain the reason why the PC jets cannot ignite the MC when the ignition timing is too retarded and when the reactivity conditions at the vicinity of burnt gas jets are insufficient for MC ignition. This is the quenching hypothesis, and the CFD simulation results presented in section 5 are essential to assess it. 4.1 Jets characterization To start with, the exit timing of cold and hot jets is studied. As defined in section 2, cold jets are made of unburned/ fresh PC mixture, whereas hot jets are made of one or more of these three: burned gases, active species and radicals and flame (1). On the one hand, Schlieren visualization should serve as an indicator of cold jet exit because it detects density gradients, which should be produced by the turbulent mass transfer of PC fresh mixture penetrating the MC after the check valve opens. On the other hand, OH* 274 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="275"?> chemiluminescence should be a valid indicator of hot jet exit because OH* is both a radical and an active species, which can be expelled by the PC as part of a flame which survived the nozzle stretching or as part of the combustion products of a flame which quenched in the hole. This way, #1 tests (Middle-high load operating point) are grouped to form couples of Schlieren and OH* recordings whose PC combustion was very similar, so that the jets exit timing can be better understood under normal operating conditions (ignition bTPP). Figure 11 shows a representative couple of this operating point. As observed on the right, the Schlieren signal appears right after the start of the valve opening is expected, after the cross-pressure point. The frame section on the bottom left shows the development of the Schlieren signal 80 μs after the first light detection. In the next frame, recorded 20 μs later, an intense first OH* light is detected with a similar surface to that of the Schlieren signal, thus indicating the OH* signal travels 4 times faster than the Schlieren’s. 20 s 20 us 80 us 3% 10% 30% 50% (first OH* signal of this jet) Schlieren OH Figure 11: Right: PC and MC pressure traces from two similar tests #3 (Schlieren, blue; OH*, red); vertical lines mark key events. Left: Overlaid jet contours near one nozzle over 20 μs, with the lower image recorded 20 μs earlier. This earlier detection of jet exit timing with the Schlieren than with the OH* optical setup is coherent with the statement that the first signal observed with Schlieren is a valid indicator of cold jet exit timing, whereas the first signal seen with OH* is a correct indicator of hot jet exit timing. The faster initial propagation of the jet when it is observed with OH* chemiluminescence may be because (i) the PC flame reaches the nozzles at a time when the pressure difference in between the chambers is closer to its maximum value, and/ or because (ii) part of this hot jet is a kernel flame which survived the nozzle stretching and propagated a short distance, meaning the reaction speed was added to the one of the bulk flow. Similarly, when this analysis is repeated for tests with ignition aTPP (Late Combustion operating point #8 - #12), the first Schlieren signal is detected earlier than the OH* one, but closer to it and sometimes even at the same time. An example of #10, like that of #3 presented in Figure 11, is presented in Figure 12. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 275 <?page no="276"?> 20 s 20 us 60 us 3% 10% 30% 50% (first OH* signal of this jet) Schlieren OH Figure 12: Right: PC and MC pressure traces from two similar tests #10 (Schlieren, blue; OH*, red); vertical lines mark key events. Left: Overlaid jet contours near one nozzle over 20 μs, with the lower image recorded 20 μs earlier. Figure 13 shows the PC and MC pressure traces during PC combustion of all the tests with any PC ignition timings tested and recorded with OH* chemiluminescence and ignitable MC mixture (#3 - #7). They rank from 28 to 20 bars of MC pressure aTPP. Note these curves were shifted to have their cross-pressure point at the same time so that a proper simultaneous visualization is possible. The vertical discontinuous colored lines indicate the first hot jets exit of each test, which is always very close to the peak PC pressure. This suggests the hot jets are always produced at a similar PC combustion timing regardless of the PC ignition timing. Figure 13: PC and MC pressure traces of #3 - #7 tests. Vertical colored lines indicate the hot jet exit of each test. All data is time-shifted to have the cross-pressure points aligned (black discontinuous line). 276 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="277"?> Figure 14 (left) shows one colored bar per test indicating the value of the maximum pressure difference between the chambers. This value barely changes as PC ignition timing is retarded due to the counter effect of two simultaneous consequent phenomena: shorter PC combustion with the check valve closed, which will lead to lower PC peak pressure, and lower MC pressure due to a lower piston position. This suggests hot jets have similar penetration speeds and. This is coherent with the values of jet tip speed measured out of the Schlieren recordings because they barely change as PC ignition is retarded. The maximum value for every test with only air in the MC and the same PC ignition timings and mixtures (#8 - #12) are shown in Figure 14 right. Therefore, jets turbulence remains very similar as ignition timing is retarded. Figure 14: Left: maxi pressure difference between PC and MC. Right: maximum jets tip penetration speed. The right axes indicate the standard deviation of the corresponding variables. There is one bar per test and one color per PC ignition timing. Figure 15 and Figure 16 show the shots-averaged radial intensity obtained out of the Schlieren and OH* chemiluminescence recordings respectively (see section 2.2 for definition and calculation details), with the earliest (left) and the latest (right) PC ignition timings (28 and 20 bars of MC pressure aTPP: #8-10) on both figures. These are tests with the MC filled with air only: #8 - #12. Figure 15: Shots-averaged radial intensity obtained out of the Schlieren recordings of #8 (ST28) on the left and #10 (ST20) on the right. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 277 <?page no="278"?> Figure 16: Shots-averaged radial intensity obtained out of the OH* recordings of #8 (ST28) on the left and #10 (ST20) on the right. The subtraction of these Schlieren intensities (Figure 17 left) demonstrates the Schlieren values change 12.5 % of the maximum intensity recorded. This low value further supports the statement that the hot jets turbulence does not significantly change as PC ignition is delayed. The same subtraction for these OH* intensities (Figure 17 right) evidence there is an increased jet-to-jet heterogeneity as PC ignition is delayed. The hot jet signal of jets 4 and 5 decreases whereas that of the other jets increases, meaning there is a preferential flame propagation path towards nozzles 1, 2, 3 and 6 which becomes more noticeable as the PC ignition is triggered later. However, these light intensity differences do not overcome the 20 % of the maximum shots-averaged signal intensity recorded. Figure 17: On the left, jet-to-jet subtraction of the Schlieren colormaps of Figure 15 as left minus right. On the right, jet-to-jet subtraction of the OH* colormaps of Figure 16 as left minus right. To summarize, jets’ characteristics remain quite similar as PC ignition timing is varied. This is achieved thanks to the active homogeneous injection system coupled with a check valve which allows chambers separation and, therefore, independent mixture formation. 4.2 Turbulent jet ignition mechanism Before attempting to explain the reason why the jets fail to ignite the MC more frequently as the PC ignition timing is retarded, the TJI mechanism must be identified. As explained 278 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="279"?> in section 1, MC ignition can be triggered by (i) active species, radicals and burned products of a flame which quenched in the nozzle, (ii) a flame which survived the nozzle stretching or (iii) a combination of the previous. In the present work, the Yamaguchi classification (13) is considered (see Table 2). Flame-Regime Observation/ Characteristics Well-dispersed burning (1) • Flame quenching • Radicals ignite the MC mixture (chemical chain reactions) • Long induction period Composite ignition (2) • Flame kernels initiate MC combustion • Thermal and chemical effects responsible • Faster burn rate Flame kernel torch ignition (3) • Flame torch ignite MC mixture • No flame quenching • Turbulent flames entrain MC Flame front torch ignition (4) • Laminar flames pass through orifices • Almost SI like combustion Table 2: Classification for different combustion regimes with active pre-chamber and ignition of main chamber (13). OH* chemiluminescence visualization helps to determine whether the PC flame survives the nozzles stretching or not (1). This way, the MC ignition and combustion of tests of middle-high load (#2) and late combustion (#3 - #7) are analyzed, putting special attention on the OH* light variation between hot jet exit and MC ignition. 4.2.1 Middle-High load Figure 18 shows the OH* signal evolution during the first frames of a representative #2 (PC ignition at 28 bar of MC pressure bTPP) test, where frame indices and time steps are given relative to the first frame in which light is detected, which is the hot jets exit timing: • Frame 1 (0 μs). All hot jets penetrate the MC at the same time. Light is quite intense in all the jets from the nozzle to the tip, with values above 50 % of the maximum in most of them. • Frame 2 (20 μs). Light drops in four jets (1, 2, 5 and 6) indicating they have not any kernel, whereas it increases in the other two (3 and 4) indicating MC ignitions. • Frame 3 (40 μs). Light drops further in two jets (1 and 5), whereas it increases in jets 2 and 6 indicating MC ignition. • Frame 4 (60 μs). Light increases in the jets 1 and 5 indicating MC ignition. At this time, the flames initiated by 2, 3 and 4 have already propagated enough to doble the size of these two recently created kernels. • Frame 8 (140 μs). The flames of all jets are developed but still not merged. Observe the flame initiated by jet 3 (one of the first which ignited the MC) has developed as much surface as the flame initiated by jet 1 (one of the last which ignited the MC). Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 279 <?page no="280"?> Figure 18: OH* signal of the first frames of a representative #2 shot (frame numbers and time are given relative to the first frame in which light is detected). This way, jets 3 and 4 produce composite MC ignitions: a kernel flame survives the nozzle stretching and, along with the radicals and combustion products, MC ignition is triggered with a short induction period (time in between hot jet exit and MC ignition ≤ 20 μs). However, the other jets produce an ignition closer to well-dispersed burning: they have decreased light in between hot jet exit and MC ignition, indicating their initial kernel flames quenched (at least partially), and the MC ignition is triggered by radicals and burned products. As expected, this mechanism leads to longer induction periods (≥ 20 μs and ≤ 80 μs), but to faster MC combustion which is visible in the flame development of jet 1 compared to that of jet 3. It should be noted that most of the MC ignitions observed in #2 tests are rather welldispersed burning type. However, this trend reverses as ignition timing is advanced, but this study is not presented in this publication since its main purpose is to address the TJI quenching in low load and late combustion conditions. 4.2.2 Late Combustion When this analysis is repeated for the Late Combustion operating point with PC coil discharge at the same MC pressure as #2, 28 bars but now aTPP, the identified dominant TJI mechanism changes. Figure 19 shows the OH* signal of the first frames of a representative #3 shot, where frame numbers and time are given relative to the first frame in which light is detected: • Frame 1 (0 μs), 2 (20 μs), 3 (60 μs) and 4 (80 μs). Hot jets penetrate the MC at different times: 1 and 2 show up at least 20 μs later than the others. The OH* signals are weak since they do not exceed 10 % of the maximum possible value, and no MC ignition is detected yet. Observe the light intensity of jets 2 and 3 is progressively increasing whereas that of the other jets drops down. There is no zone exceeding 10 % threshold in Frame 4. • Frame 5 (100 μs) and 6 (120 μs). MC ignition is triggered in jets 2, 3 and 4; and for 2 and 4 in two visible locations. • Frame 14 (300 μs). MC ignition is triggered in jet 5, closer to the nozzles. • Frame 33 (380 μs). MC ignition is triggered in jet 6. Shortly after, jet 1 ignites too (not shown). Both appear very close to the nozzles. 280 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="281"?> 20 s 20 s 3% 10% 30% 50% 20 s 20 s 20 s 180 s 380 s 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 Figure 19: OH* signal of the first frames of a representative #3 shot (frame numbers and time are given relative to the first frame in which light is detected). This way, the low OH* signal at hot jet exit evidence the kernel flame is not surviving through the nozzles. This difference of #3 compared to #2 might be due to (i) the lower MC pressure at jet exit since the MC volume is increasing, thus leading to decreased laminar flame thickness and therefore longer quenching distance in the nozzle (1) (14); and due to (ii) the lower injected PC load, which leads to hot jets with lower enthalpy and mass of active species because less energy is released into the PC to form them. Therefore, the dominant mechanism in #3 is well-dispersed burning, and this stays true when the PC ignition timing is retarded with constant PC load (#4, #5, #6 and #7). Furthermore, it should be noted that hot jets intensities are slightly reduced when PC ignition is retarded, which is coherent with the analysis of the previous section. ST28 ST26 ST24 ST22 ST20 Figure 20: HRR of all tests of the PC ignition timing sweep with reactive mixture in the MC (#3 - #7). The different colors represent the test number at the same conditions. Figure 20 presents how MC slow combustions become more frequent when PC ignition is retarded, thus penalizing the MC combustion repeatability, evaluated with the normalized RMS (see section 2.2 for definition, see Appendix Afor calculation details) in Figure 21 left. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 281 <?page no="282"?> When PC ignition was delayed to 22 bars of MC pressure aTPP (ST22), misfires appeared (Figure 21 right), thus heavily deteriorating the repeatability. Finally, with 20 bars of MC pressure aTPP at PC ignition (ST20), MC misfire chances are almost 100 %. Misleading repeatability due to high number of misfires Figure 21: Left: normalized RMS of the MC pressure. Right: number and proportion of MC misfires. Data obtained out of all tests of the PC ignition timing sweep with reactive mixture in the MC (#3 - #7). Figure 22 demonstrates this evolution of MC combustion speed is related to the number of jets which succeed in igniting. Figure 22: Scatter plot with one point per test and one color per PC ignition timing shows a relationship in between the number of jets which ignite and the time between hot et exit and MC MFB 50. The variation of the MC ignition delay with the PC ignition timing is studied through the boxplot of Figure 23(A) (left axis), in which every input value corresponds to the first ignition of every jet which succeeds in igniting the MC. This MC ignition delay is the time in between the hot jet exit detection and the corresponding MC ignition of every jet and test. The right axis of Figure 23(A), indicates the ratio of jets which have failed in igniting the MC along all the tests of every PC ignition timing. 282 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="283"?> Figure 23: Left axis: box plot of MC ignition delays for every PC ignition timing. There is one input value per jet which succeeds in igniting the MC. Right axis: rate of jets which failed to ignite the MC. There is one input value per jet. All shot of #3 - #7 recorded with OH*chemiluminescence are included. As observed, more jets fail to ignite the MC as PC ignition is retarded, and the MC ignition delay to hot jet exit increases. Consequently, these MC ignitions are detected further from nozzles, as proved in Figure 23 (B) (left axis). Figure 24 displays a point per MC ignition detected on a 2D map of the combustion chamber where the PC cap and the piston window limits are marked with black lines. The colored arcs indicate the average ignition radius of every jet and group of PC ignition timing. It should be noted that all MC ignitions are produced around one of the jets axes. Figure 24: Left: MC ignition distance and non-igniting jet fraction. Right: Ignition locations and average ignition radius for each jet and PC ignition timing. Data from OH* measurements of shots #3-#7. Finally, the chances of every jet to successfully ignite the MC was found to be more heterogeneous as PC ignition timing is retarded. As shown in the ignition location probability density maps of Figure 25, with PC ignition timing of 28 bar aTPP, all jets have the same chances to ignite the MC and at roughly the same distance to their correspond- Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 283 <?page no="284"?> ing nozzle. However, delaying this timing leads to progressively reduced chances of igniting with jets 2 and 3. Furthermore, the jet with the highest chances to ignite switches from 6 to 5 when retarding the PC ignition from 24 to 22 bar. This anticlockwise switch could be explained by a different location of the PC flame front at the instant the check valve opens, to be investigated further with CFD. It should be noted that OH* jet-to-jet intensities were already found to be more heterogeneous as PC ignition timing was retarded in the tests with the MC full of air, analyzed in the previous section. This is potentially related to the jet-to-jet MC ignition chances heterogeneities observed in this analysis of the tests with the MC full of reactive mixture. Figure 25: Ignition location probability density maps of #3 - #6, with an annular segmented ring indicating the ratio of successful MC ignitions achieved with every jet. 4.3 Quenching hypothesis 4.3.1 PC ignition timing sweep The OH* signal intensity recorded out of Late Combustion tests with a reactive mixture in the MC (#3 - #7) were compared to those with the same PC ignition timing and load but with just air in the MC (#8 - #12). This was achieved by making unique couples of the tests of both groups with the same PC ignition timing, looking for pairs with the most similar PC combustion. As a result, the hot jets penetrating an MC filled just with air were found to have a higher OH* signal than those of an MC filled with reactive mixture. Figure 26 compares the first frames in which OH* is detected in two tests with very similar PC combustion and a PC ignition timing of 28 bar of MC pressure aTPP: on the bottom the 284 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="285"?> recording with just air in the MC, and on the top the homologue test with reactive mixture in the MC. Since both tests have also the same PC pressure trace during the combustion, the exact same hot jets characteristics are expected. However, their OH* signal is more intense in the videos with the MC full of air. This is because, in those with MC full of air and fuel, part of this same OH* mass is quickly consumed by the reactive mixture, potentially as part of the reactions necessary to ignite it. If this comparison is repeated for a couple of tests with the latest PC ignition timing (20 bar aTPP), the same OH* light intensity difference is observed in the hot jets in between the reactive and the non-reactive MC mixture tests. However, with this PC ignition timing the air and fuel MC mixture never ignites. Figure 27 is an example, in which the tests were chosen again to have a very similar PC combustion and therefore hot jets. On the bottom, the recording with just air in the MC, and on the top, the homologue test with reactive mixture in the MC. PC ignition 28 bar of MC pressure aTPP: MC filled with reactive mixture vs with air Figure 26: First four frames with OH* signal of two tests with very similar PC combustion and the same PC ignition timing: 28 bars of MC pressure aTPP. On the top: MC filled with reactive mixture (#3), on the bottom: MC filled just with air (#8). Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 285 <?page no="286"?> PC ignition 20 bar of MC pressure aTPP: MC filled with reactive mixture vs with air Figure 27: First four frames with OH* signal of two tests with very similar PC combustion and the same PC ignition timing: 20 bars of MC pressure aTPP. On the top: MC filled with reactive mixture (#5), on the bottom: MC filled just with air (#10). Since the OH* light intensity difference is still present, there are MC reactions consuming the OH* of the hot jets in the tests with the MC full of reactive mixture, which might be required to ignite the MC. However, they don’t lead to visible OH* production due to ignitions of this mixture. This suggest the TJI quenching is partial: MC ignition reactions start, since the OH* is consumed, but they do not conclude in an MC combustion flame onset. On the one hand, cold and hot jets characteristics have minor differences when PC ignition timing changes, as demonstrated in section 4.2. On the other hand, MC conditions at jet exit change significantly. As shown in Figure 28, the mixture around the hot jets of tests with PC ignition timing at 20 bars aTPP should have a temperature and pressure of 75 C and 8 bars respectively lower than this mixture of tests run with 28 bars aTPP of PC ignition timing. The temperature was calculated using the ideal gas law and the pressure was measured. Figure 28: MC calculated temperature (with ideal gas law) and measured pressure at hot jet exit of Late Combustion tests with the same PC load and phi but different ignition timings, marked with colors and bars groups (#3 - #7). 286 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="287"?> The lower MC temperature, the quicker heat transfer from the hot jets to the surrounding fresh mixture; and the lower the pressure, the lower the density of unburned gas mixture around the hot jets and therefore the stronger expansion of the hot jets, which increases further their heat transfer rate. These effects are responsible of decreasing the chances of igniting the MC with jets as PC ignition timing is retarded until they are null. This TJI partial quenching explanation will be checked through CFD simulation results in section 5.2 4.3.2 PC load sweep If this TJI quenching hypothesis was correct, a change in the jets’ properties would increase the chances of igniting the MC. These new jets characteristics should be aimed at slowing down their heat transfer to the MC mixture and/ or at increasing their thermal energy (enthalpy and temperature) and/ or active species concentrations. This way, a sweep of reactive mixture injected into the PC was performed towards higher (#13) and lower (#14 - #18) values, with a constant PC ignition timing of 22 bar of MC pressure aTPP. The original PC load is 7.43 mg of reactive mixture at phi 1.2, which is 20.1 J of initial input PC energy used to produce the hot jets. As shown in Figure 29, the number of MC misfiring tests is reduced with lower PC loads until there are none with 8.0, 9.3 and 13.3 J of PC load, whereas MC misfires are guaranteed with higher loads: 22 J. It should be noted that a PC load below 8.0 J was not possible with this PC ignition timing because the PC pressure could not overcome the MC pressure with it, meaning the check valve remained closed. Figure 29: Number and proportion of MC misfires obtained out of the tests with different PC loads (energy injected into the PC before triggering its ignition) A quick analysis of the HRR (Figure 30) highlights every PC load leads to two types of MC combustion: fast and slow ones, being the former more frequent with 9.3 and 13.3 J. To continue, PC and MC combustion repeatability is studied with the method explained previously (see details of the calculations for these tests in Appendix B). As a result, Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 287 <?page no="288"?> Figure 31 shows the normalized RMS of the MC pressure traces of every PC load sweep test. As observed, a PC energy before ignition in the range of [8.0 - 13.3] J maximizes MC combustion repeatability without misfires, achieving the minimum dispersion with 9.3 J. Fast Slow Fast Slow Figure 30: Four figures, one per PC load, showing the HRR of every test. Fast and slow MC combustions are identified as groups marked with arrows. The different colors represent the test number at the same conditions. Figure 31: Normalized RMS of the MC pressure traces during combustion of the PC load sweep tests. This way, the MC combustion repeatability is closely related to the variations on the MC combustion speed. This parameter was found to be affected by three factors: (i) the number of jets which succeed in igniting the MC, (ii) the TJI mechanism, and (iii) the jets penetration. They are all analyzed next for the tests of this PC load sweep. To start with, Figure 32 compares the time in between hot jets exit and MFB 50 of the MC, to the number of jets which ignite in every test of the PC load sweep. It demonstrates that there is a trend and that at least three jets must ignite the MC to produce a combustion assigned to the fast groups identified in Figure 30. 288 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="289"?> Figure 32: Scatter plot with one point per test and one color per PC load. It shows a linear relationship in between the number of jets which ignite the MC and the time between hot et exit and MFB 50 of the MC. It is clear the number of jets which succeed in igniting the MC is the main factor affecting the MC combustion speed. In addition, the TJI mechanism and the jet penetration have a secondary effect on the MC combustion speed, which is well visible when tests of different PC loads are compared. Concerning the TJI mechanism, well-dispersed burning ignition is always dominant, but indications of possible flame kernel torch ignitions appear with the lowest PC loads: 8.0 and 9.3 J. In these cases, both ignition types coexist, and this makes a noticeable difference in the MC combustion speed. This can be appreciated when two tests with 8.0 and 13.3 J of PC load in which the same number of jets ignite are compared. Comparison to understand TJI mechanism effect: PC load 8.0 J with 4 jets igniting Figure 33: Top left: scatter of Figure 32 indicating the test whose data is displayed in the other figure and frames, it corresponds to #18. Top right: PC and MC pressure traces of this test with vertical lines indicating the times at which the frames below correspond, ordered from the earliest (bottom left) to the latest (bottom right). Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 289 <?page no="290"?> Comparison to understand TJI mechanism effect: PC load 13.3 J with 4 jets igniting Figure 34: Top left: scatter of Figure 32 indicating the test whose data is displayed in the other figure and frames, it corresponds to #16. Top right: PC and MC pressure traces of this test with vertical lines indicating the times at which the frames below correspond, ordered from the earliest (bottom left) to the latest (bottom right). As observed in the example of Figure 33, the first light visible with the lowest PC load is detected when the pressure difference between PC and MC drops back to 0. This suggests all light observed is due to MC combustions, which starts from nozzles 1, 2 and 4 vicinity and propagate axially and radially at relatively low speed, thus indicating flame kernel torch ignitions. Jet 3 seems to have the same ignition type and an additional well-dispersed one. On the contrary, jets 5 and 6 fail to ignite the MC. Figure 34 shows the first light visible, with the higher PC load, is detected when the pressure difference between PC and MC is close to its maximum value and all hot jets are visible before they trigger the MC ignition. The evolution of the OH* intensities is like the one presented in the PC ignition timing sweep of the previous section: TJI mechanism is well-dispersed burning, and MC ignitions occur in jets 1, 2, 3 and 4 at a certain distance from their corresponding nozzles, triggering a relatively faster flame propagation. This difference of 2 ms of half of MC combustion duration (visible in the scatter plot) in between these two tests is also because the flames initiated by each jet cover a bigger MC volume and mass before merging in the case of the higher PC load, leading to more flame surface during combustion. This is visible when comparing the last frames of the Figure 33 and Figure 34 (taken at the instant flames merge), and it is a logical consequence of the ignitions distance to their nozzles: the shorter distance between the ignitions triggered by different jets, the shorter distance is traveled by their corresponding flames before merging so the lower flame surface during combustion. It should be noted that the different TJI quenching mechanisms is a direct consequence of the ΔP built between the chambers, which determines the jets speed and turbulence, and it is obviously proportional to the PC load. The lower ΔP when the PC flame reaches the nozzles, the lower bulk flow speed inside of the orifices, so the lower heat transferred 290 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="291"?> through their walls and, therefore, the lower chances of quenching the flame when it goes through them. Comparison to understand jet penetration effect: PC load 9.3 J with 6 jets igniting Figure 35: Top left: scatter of Figure 32 indicating the test whose data is displayed in the other figure and frames, it corresponds to #17 (9.3 J of PC load). Top right: PC and MC pressure traces of this test with vertical lines indicating the times at which the frames below correspond, ordered from the earliest (bottom left) to the latest (bottom right). Concerning the jet penetration, its effect on the MC combustion speed is easily understood when comparing tests with loads 9.3 and 13.3 J when the same number of jets ignite the MC and the TJI mechanism of the 9.3 J test is also fully well-dispersed burning. Figure 35 and Figure 36 show the pressure traces and a few frames during the first stages of MC combustion of two tests with these characteristics. Consistent with the previous observations, jets with a PC load of 13.3 J penetrate more due to a higher ΔP between the chambers, leading to ignition locations further from the nozzles. After the ignitions, the burning mass is also translated further from the nozzles thanks to the higher jets flow inertia. Consequently, more mass entrains the separate flames before they merge since they will cover a bigger MC volume, leading to a half of MC combustion 1.8 ms shorter (visible in the scatter plot). Furthermore, the higher pressure difference between the chambers obtained with 13.3 J increases the hot jets turbulence, thus increasing the flame wrinkling during the first stages of MC combustion. This leads to faster initial flame propagation and OH* production rate, which is visible in the recorded intensities right after every ignition. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 291 <?page no="292"?> Comparison to understand jet penetration effect: PC load 13.3 J with 6 jets igniting Figure 36: Top left: scatter of Figure 32 indicating the test whose data is displayed in the other figure and frames, it corresponds to #16. Top right: PC and MC pressure traces of this test with vertical lines indicating the times at which the frames below correspond, ordered from the earliest (bottom left) to the latest (bottom right). Finally, Figure 37 (right axis of both plots) proves hot jets have the highest chances to ignite the MC with a PC load in between 9.3 and 13.3 J. Furthermore, Figure 37 (left axis) shows the ignitions achieved with these loads happen the closest to the nozzles (right) and with the shortest delays (left) from hot jets exit. Figure 37: Box plots showing the evolution of the MC ignition delay to hot jets exit (left, left axis) and the distance of the jets’ ignitions to their corresponding nozzle (right, left axis) with the PC load (one box plot per load). The right axis indicates the ratio of jets which failed to ignite the MC with every PC load. In addition, the ignition location probability density map of Figure 38 demonstrates a PC load of [9.3 - 13.3] J homogenizes more the jet-to-jet chances of igniting the MC. It is interesting to see the jets with the highest chances to ignite change as PC load is decreased in a clockwise sense, from jets 4 and 5 with 16.7 J to jets 1, 2 and 6 with 8.0 J. Once again, this could be explained by a different development of the PC flame front at the instant the check valve opens. 292 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="293"?> Figure 38: Ignition location probability density map (inner circle) with percentage of successful jet ignitions (outer segments) of tests with different PC input energy: #13 - #18. This way, it has been demonstrated that there is a range of PC loads, [9.3 - 13.3] J, which maximizes the chances of igniting, with turbulent hot jets, an expanding MC mixture with relatively low reactive mass and pressure. Inside of this range, there is a “sweet point” of 9.3 J, which makes MC combustion slower while maximizing its repeatability. This would be preferred in an engine because the objective of igniting late with low MC load is to convert a higher proportion of the fuel energy into exhaust gas enthalpy to warm up the catalyst quicker in a cold start procedure (5). To conclude, the TJI quenching hypothesis is presented: there is a range of combinations of three hot jets properties which maximizes their chances of igniting the MC: turbulence, enthalpy (thermal energy), and concentration of active species. This is true even when the hot jets attempt to ignite late an expanding low-density MC mixture. This way, on the one hand, MC ignition was impossible with the highest PC load (22.0 J) due to an excess of turbulence which leads to a strong mixing and therefore a too fast heat transfer from the jets to the MC mixture; even though the thermal energy and concentration of active species of the hot jets is higher than with lower loads. On the other hand, MC ignition chances drop down with the lowest PC load (8.0 J) due to an insufficient amount of thermal energy and/ or active species concentrations of the hot jets. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 293 <?page no="294"?> This hypothesis of the TJI quenching limits will be further studied in an incoming publication, in which advanced optical post-processing and CFD modeling will be combined to determine accurate MC ignition limits based on the hot jet properties in any given MC conditions. To support this effort, the next section describes the CFD model developed in the present work and assesses its predictive capabilities through a first validation against experimental observations, followed by preliminary simulation results. 5 Numerical study of jet ignition behavior Computational fluid dynamics (CFD) is a relevant tool to study quantities which may not be available to direct measurement in an experimental context. However, accurate modelling of the phenomena occurring in the experimental facility is necessary for correct representativity. This work focuses on the modelling of the ignition of the main chamber (MC) by burnt gas expulsion from the prechamber (PC), and the possibility of misfire, in a context of late combustion (LC). Comparison between experimental data - considered as a baseline for CFD - and numerical data aims at validating qualitatively the ignition modelling approach and provide support to RCM optical observations. While experimental results suggest high ignitability (i. e. low misfire propensity), fast, and repeatable combustion for ST28 (28 bar of MC pressure aTPP), lower ignitability (i. e. high misfire propensity), slow and more unstable combustion is displayed by ST20. Intermediate behavior is attributed to ST24. Involving gradual transition between two limit behaviors, three late combustion operating points are selected for CFD simulations: ST28, ST24, ST20 (#8, #9, #10). These points are of particular interest to evaluate the capability of the combustion modelling approach to capture ignition/ misfire events in the MC. 5.1 Numerical setup Converge version 3.0 is used here. Large-eddy simulation (LES) approach based on Sigma (15) model is selected to simulate the RCM facility. A cartesian mesh of cell size of 200 μ is applied in the PC, in the ducts, and in the region of the expulsed jets Figure 39. The flameturbulence interaction (FTI) is modelled through Pitsch correlation (16) incorporated in Gequation (17), coupled with a kinetic solver (SAGE) for reactions computation. Note that ignition in the MC relies on an ad-hoc ignition criterion, designed to represent the possibility of flame quenching in the ducts connecting PC and MC. Details on the criterion will be provided in a future work. Laminar flame speed (S L ) table used in G-equation is based on computation of collection of canonical 1D flames with Cantera solver, using Sarathy mechanism (18) containing 62 species. Gas state is defined by Redlich-Kwong nonideal gas law. Heat exchanges rely on O’Rourke model (19), and blow-by is considered via Namazian model (20). The fuel used in the simulation is iso-octane, coherently with the one used in the RCM tests. 294 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="295"?> x =3.2mm x =0.2mm Figure 39: Schemic of the computational domain at the associated mesh. 5.2 Validation of the numerical setup MC blow-by model is validated on a non-firing RCM shot, based on pressure trace comparison. Numerical results reveal minor deviation from experimental baseline Figure 40. Same procedure is used to validate the modelling of the PC high-velocity gas injection and the PC combustion. The former displays marginal deviation from measurements Figure 40. Pressure differential between PC and MC noted ΔP is also provided Figure 41. Despite underestimation of ΔP peak, the phasing of the latter is in satisfactory agreement with experiment. PC combustion and ensuing burnt gas expulsion to the MC are considered acceptably representative of the experiment. Figure 40: Left: MC pressure trace comparison between experiment and CFD on a compression shot with no combustion. Right: PC pressure trace during gas injection and PC combustion for ST28. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 295 <?page no="296"?> Figure 41: Pressure differential ΔP comparison between experiment and simulation. Comparison between experiment and simulation is focused on the capability of the combustion model to represent MC ignition across the three Late Combustion operating points. However, as visible on Figure 41, PC gas expulsion can be segregated into two stages: (i) PC ignition generates successively sharp ΔP increase and decrease as burnt gas expulsion occurs due to the opening of the PC valve. This stage is referred as “main stage of burnt gas expulsion”. (ii) ΔP oscillations around zero develop as the open valve allows PC and MC pressure to equilibrate. This second stage involves an oscillatory flow between the two regions, and multiple - low velocity - expulsions of gas from the PC (not show here). Note that this study is limited to the “main stage of burnt gas expulsion”. 5.3 Simulation Results 5.3.1 MC ignition predictivity In this section, a description of the MC ignition is provided at the end of the “main stage of burnt gas expulsion” introduced in section 5.2. This corresponds to t=1.460ms, t=1.320ms, and t=1.300ms after PC spark-ignition timing (SIT) for ST28, ST24, and ST20, respectively, see Figure 42. The ignition in the MC is evaluated through a progress variable noted C, describing the local state of the gases. Values of zero and one are attributed to unburnt and burnt gas state, respectively, and intermediate values to partially burnt mixture. In this study, a value superior/ inferior to C=0.5 is considered as an ignited/ non-ignited region. 296 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="297"?> Figure 42: ST28 (left), ST24 (middle), ST20 (right) at the end of “main stage of burnt gas expulsion”. Top row represents temperature fields in a slice in the middle of the computational domain. Bottom row represents a top view of the domain with three iso-surfaces of progress variable, such as C=0.1 (lighter yellow), C=0.5 (darker yellow), C=0.9 (orange). Presence of ignition kernels is observed for ST28 and ST24 in respectively three and two of the six burnt gas jets expulsed from the PC. ST20 displays no ignition kernel, suggesting weaker reactivity. Importantly, this trend is compatible with observations drawn from the RCM optical data, section 4.2.2, suggesting qualitative representativity of the ignition modelling approach. Nonetheless - quantitatively - the simulation strongly underpredicts the ignitability of ST28 and ST24 as number of jets leading to MC ignition is inferior to experimental observations Figure 25. Also, comparison of PC and MC pressure traces provided in Figure 43 suggests overprediction of the pressure elevation ensuing jets exit, for both ST24 and ST20. Modelling of flame propagation in the MC should thus be improved. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 297 <?page no="298"?> Figure 43: Experimental (left) and simulation (right) PC and MC pressure traces. PC behavior is comparable. MC pressure increase after jet exit is overestimated. 5.3.2 OH fields analysis As described in section 2, OH* radical luminescence is associated to combustion areas. Evaluation of the local OH mass concentration in simulation data is provided to support experimental optical analysis. High OH concentration is attributed to combusting regions, low concentration to non-combusting ones. OH fields from simulation are provided in Figure 44 at two instants: during jet exit (left), and at the end of “main stage of burnt gas expulsion”. Comparably to experiment, low OH concentration is encountered at the vicinity of the PC ducts during the jet exit, and higher values are observed in the more peripheral ignition pockets for ST28 and ST24. For ST20, in absence of MC ignition pockets, OH concentration appears solely at the exit of the PC ducts, in low concentration. Note that this is coherent with experimental observation, Figures 26 ,27. Also, OH fields at passing through the PC/ MC ducts display strong concentration decrease, suggesting intense consumption of OH radical, and strong weakening of reactivity promoting flame quenching. 298 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="299"?> Figure 44: Fields of OH mass fraction along a plane crossing the middle of the computational domain. ST28 (top row), ST24 (middle row), ST20 (bottom row) taken during jet exit from PC (left), and at the end of “main stage of burnt gas expulsion (right)”. 5.4 Conclusions on the simulation results: A set of simulations aimed to represent three experimental points referred as ST28, ST24, ST20 involving strong MC ignitability differential - ST28/ ST20 identified the most/ least ignitable conditions - were carried out. Results suggest that the developed combustion modelling approach is capable to capture qualitatively the ignitability differential observed in experiment: simulation displays higher number ignition spots emerging from expulsed PC jets for ST28, then for ST24, then for ST20, compatibly with experimental measurements. It is highlighted that these results are a preliminary step for future modelling efforts. Also, OH concentration fields display a strong decrease in the ducts interfacing PC and MC. This suggests weakening of reactivity and local flame quenching. Also, OH concentration is low in the portion of the jets located at the vicinity of the ducts, and increases in the more distant ignition kernels, similarly with optical observations. 6 Conclusion In this investigation, an innovative experimental setup of active Turbulent Jet Ignition coupled with a rapid compression machine (RCM) and advanced optical diagnostics was developed to investigate the ignition and quenching limits of active pre-chamber combustion under cold-start-like conditions. The combination of independent pre-cham- Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 299 <?page no="300"?> ber and main-chamber mixture preparation, enabled by the check-valve concept (Valvijet), together with OH* chemiluminescence and Schlieren visualization, provided unprecedented access to the fundamental mechanisms governing jet ignition and quenching. (i) Retarding PC ignition timing progressively reduces MC ignitability, increases ignition delay, shifts ignition locations further from the nozzles, and ultimately leads to misfire. (ii) Global jet characteristics remain nearly unchanged with ignition retardation, as both the pressure difference between chambers (ΔP) and the hot-jet ejection timing remain similar. However, jet-to-jet heterogeneity increases, with OH* measurements indicating preferential activity in specific jets and increasing nonuniform ignition probabilities. (iii) Well-dispersed burning was identified as the dominant TJI mechanism under all late-combustion conditions. MC ignition is initiated by hot combustion products and active species transported by the jets rather than by flame kernels directly propagating through the nozzles. (iv) The progressive reduction in MC ignitability with ignition retard is primarily caused by the decreasing MC temperature, pressure, and density at jet arrival, which increase heat losses from the hot jets to the surrounding mixture. (v) PC load (injected mass into the PC before its ignition) strongly affects jet properties. Lower PC loads reduce jet turbulence, thermal enthalpy, and active-species content, whereas higher PC loads generate more energetic and turbulent jets. Jet-to-jet heterogeneity is reduced within an intermediate PC-load range, although complete homogeneity is never achieved. (vi) The number of jets successfully igniting the MC was identified as the primary parameter controlling MC combustion speed, thus impacting the most on the MC combustion repeatability. Ignition mechanism, jet turbulence, and ignition location and translation also influence combustion development, but their contribution remains secondary. (vii) An optimum PC-load range was identified, maximizing ignition probability and combustion repeatability. Within this range, a narrower “sweet spot” provides the best compromise between ignition robustness, ignition delay, and combustion development. (viii) The results support the existence of upper and lower TJI quenching limits governed by the balance between jet turbulence, thermal enthalpy, and activespecies concentration. Excessively high PC loads increase turbulence and heat transfer to the surrounding mixture, promoting quenching despite the higher chemical and thermal potential of the jets. Conversely, excessively low PC loads lead to insufficient thermal energy and active-species concentrations to sustain MC ignition. Finally, a CFD framework was developed and validated against the experimental measurements. The simulations reproduced the main features of the PC combustion and jet ejection processes, correctly capturing the qualitative ignitability trend observed experimentally, with ignition predicted for ST28 and ST24 and no ignition for ST20. The 300 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="301"?> simulated OH fields were also consistent with the optical observations and provided further support for the proposed quenching mechanism. Overall, the combined experimental and numerical analyses indicate that successful TJI operation relies on an optimum balance between jet turbulence, thermal energy, and chemical reactivity. Future work will build upon advanced optical diagnostics and CFD simulations to establish quantitative ignition criteria and predictive quenching limits for active pre-chamber combustion systems operating under highly diluted and cold-start conditions. 7 Bibliography 1. Biswas, S. Physics of Turbulent Jet Ignition: Mechanisms and Dynamics of Ultra-lean Combustion. s. l. : Springer Theses, Springer International Publishing, 2018. 2. A Review of Pre-Chamber Initiated Jet Ignition Combustion Systems. Toulson, E., Schock, H. J. and Attard, W. P. 2010, SAE Technical Paper 2010-01-2263. 3. Simultaneous Direct-photography of Flame Propagation inside Pre-chamber and Main-chamber in Gasoline Engine with Passive Pre-chamber. Hokimoto, S., Moriyoshi, Y., Kuboyama, T., Egashira, S. and Nagai, Y. s. l. : SAE Technical Paper 2024-32-0095, 2024. 4. Investigation of Spark Position Effects in a Small Pre-chamber on Ignition and Early Flame Propagation. Wallesten, J. and Chomiak, J. 2000-01-2839, s. l. : SAE Technical Paper, 2000. 5. Heywood, J. B. Internal Combustion Engine Fundamentals. 2nd ed., . s. l. : McGraw-Hill Education, 2018. 6. The use of active jet ignition to overcome traditional challenges of pre-chamber combustors under low load conditions. Bunce, M., Cairns, A. and Blaxill, H. 2020, International Journal of Engine Research. 7. Development of a High Efficiency SI-engine Demonstrator Based on Advanced active TJI Technology. Karageorgiou, D., Myslivecek, M., Rieß, M., Reinicke, PB. [ed.] A. C., Reuss, HC., Wagner, A. (eds) Kulzer. s. l. : Springer Vieweg, 2025. 2025 Stuttgart International Symposium on Automotive and Engine Technology. 8. Burkardt, P., Mirsch, N., Fleischmann, M., Preu ker, M. and Salerno, F. (Pischinger, S., Heufer, K. A. and Casal Kulzer, A., dir.). SACI Combustion System with Active Pre-Chamber Investigation of the CO2 reduction potential of a combustion system based on spark-assisted compression ignition (SACI) in combination with an active pre-chamber. . s. l. : FVV Project no. 1431, 2024. 9. Method for pressure trace based thermodynamic analysis of pre-chamber combustion. Balmelli, M., Rogers, D., Hilfiker, T., Wright, Y. and Soltic, P. 312, 2024, Energy Conversion and Management. 10. Kyrtatos, P., Bardis, K., Bolla, M., Denisov, A., Wright, Y., Herrmann, K. and Boulouchos, K. Transferability of Insights from Fundamental Investigations into Practical Applications of Pre-chamber Combustion Systems. [book auth.] ETH Zürich. Aerothermochemistry and Combustion Systems Laboratory. 8.3. 11. Computational analysis of flame initiation, quenching, and re-ignition in a prechamber natural gas engine under varying EGR-dilution levels. Varma, A. R., Singh, S., Rajasegar, R. and Srna, A. 2024, Fuel, Vol. 375. 12. Innovative prechamber system with valve for future high. Karageorgiou, D., Cao L., Sankesh, D., Gastaldi, P., Myslivecek, M., Rabhi, V. [ed.] M. Sens (ED.). s. l. : Narr Francke Attempo Verlag GmbH, 2022. IAV International Conference on Ignition Systems for Gasoline Engines. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 301 <?page no="302"?> 13. Ignition and burning process in a divided chamber bomb. Shigeki Yamaguchi, Norio Ohiwa, Tatsuya Hasegawa. 2, 1985, Combustion and Flame, Vol. 59, pp. 177-187. 14. Using singular values to build a subgrid-scale model for large eddy simulations. Nicoud, Franck, et al. 8, 8 2011, Physics of Fluids, Vol. 23, p. 085106. 1070-6631. 15. A consistent level set formulation for large-eddy simulation of premixed turbulent combustion. Pitsch, H. 4, 2005, Combustion and Flame, Vol. 143, pp. 587-598. Special Issue to Honor Professor Robert W. Bilger on the Occasion of His Seventieth Birthday. 16. Large-eddy simulation of premixed turbulent combustion using a level-set approach. Pitsch, H. and Duchamp de Lageneste, L. 2, 2002, Proceedings of the Combustion Institute, Vol. 29, pp. 2001-2008. 17. Reduced Gasoline Surrogate (Toluene/ n-Heptane/ iso-Octane) Chemical Kinetic Model for Compression Ignition Simulations. Sarathy, M. Atef, N., Alfazazi, A., Badra, J., Zhang, Y., Tzanetakis, T. and Pei, Y. s. l. : SAE International, 2018. WCX World Congress Experience, Detroit, Michigan, United States. 18. Amsden, A. A. KIVA-3V: A block-structured KIVA program for engines with vertical or canted valves. s. l. : Los Alamos National Laboratory, New Mexico, 1997. 19. Flow in the Piston-Cylinder-Ring Crevices of a Spark-Ignition Engine: Effect on Hydrocarbon Emissions, Efficiency and Power. Namazian, M. and Heywood, J. B. 1982, SAE Technical Paper 820088. 20. Settles, G. S. Schlieren and Shadowgraphy Techniques. Penn State : Springer, 2001. ISBN 978-3- 642-63034-7. 21. Using singular values to build a subgrid-scale model for large eddy simulations. Nicoud, F., Toda, H. B., Cabrit, O., Bose, S. and Lee, J. . 8, 2011, Physics of Fluids, Vol. 23. 302 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="303"?> Appendix A. Combustion repeatability of the PC ignition timing sweep tests These are the pressure traces in both chambers of the tests of the PC ignition timing sweep: #3 - #7. As explained in section 2.2, the shadow region marks the point-wise standard deviation and the thicker line is the corresponding average. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 303 <?page no="304"?> These are the normalized RMS values for every operating point of this sweep. Note the values for ST20 are misleading due to the high rate of MC misfires. It’s interesting to see the dispersion of the MC combustion follows that of the PC. 304 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="305"?> Appendix B: Combustion repeatability of the PC load sweep tests These are the pressure traces in both chambers of the tests of the PC load sweep: #13 - #18. As explained in section 2.2, the shadow region marks the point-wise standard deviation, and the thicker line is the corresponding average. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 305 <?page no="306"?> 306 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="307"?> These are the normalized RMS values for every operating point of this sweep. Note the values for 20.1 and 22.0 J are misleading due to the high rate of MC misfires. It’s interesting to see the dispersion of the MC combustion also follows that of the PC. Experimental and Numerical Investigation of active Turbulent Jet Ignition quenching limits 307 <?page no="308"?> 308 Dimitrios Karageorgiou, Miguel Monserrat Diaz, Matej Myslivecek, Dr. Maxime Tarot, Dr. Salim Sebai <?page no="309"?> Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber Frank Wittmann 1,2 , Felix Schönhofer 1 , Christopher Kowis 1,2 , Rafael Clemente Mallada 1 , Sebastian Rieß 1,2 , Michael Wensing 1,2 Abstract: An actively fueled single-hole pre-chamber ignition system is studied to investigate how the gas composition inside the pre-chamber influences the emerging jets and the transition between flame ignition and jet ignition. The pre-chamber is fueled with a homogeneously premixed methane-air mixture with equivalence ratios ϕ =0.9, 1.0, 1.2 and 1.5. Methane and air are mixed using two mass flow controllers and supplied to the pre-chamber using a DI-injector. The mixture inside the pre-chamber is ignited by a spark plug. Additionally, the oxygen content of the air is varied to investigate different residual gas fractions inside the pre-chamber prior to ignition. Therefore, experiments with an oxygen content of 21 %, 18 % and 15 % have been conducted in order to mimic the real operating conditions of a pre-chamber engine with 0 %, 14.29 % and 28.57 % residual gas remaining inside the pre-chamber while running. The pre chamber is mounted to a continuously scavenged constant flow chamber (CFC) with high optical accessibility. The CFC enables highly controlled the modynamic conditions, with the purge gas temperature maintained at a constant 573 K and the pressure regulated to 1 MPa. The CFC is purged either with nitrogen or air. The pre-chamber is equipped with a miniature piezoelectric pressure sensor, providing information about the pressure rise caused by the ignition and combustion inside the pre-chamber. High-speed Schlieren imaging was used to visualize the penetration of the hot gas jet exiting the pre-chamber. Additional OH* chemiluminescence imaging was used to identify if the emerging jet contains active radicals and, if so, to provide insight into the position of the reaction zone. Together, the jet penetration speed in the main chamber and the pressure trace inside the pre-chamber serve as indicators of the pre-chamber heat-release rate and the resulting blowdowndriven jet momentum. The jet penetration and pressure trace are evaluated and compared for different oxygen contents showing that ignition stability is influenced by the available oxygen. The OH* signal is only detected for rich mixtures and when using air instead of nitrogen as purge gas in the CFC, leading to the conclusion that the equivalence ratio and available oxygen are important factors when distinguishing between flame ignition and jet ignition. The experiments highlight the central influence of gas composition on jet formation and ignition mode for pre-chamber ignition systems. 1 Professorship for Fluid System Technology, Friedrich-Alexander-University Erlangen-Nürnberg Cauerstraße 4, 91058 Erlangen, Germany 2 Erlangen Graduate School in Advanced Optical Technologies (SAOT) <?page no="310"?> Introduction The internal combustion engine (ICE) remains a fundamental pillar of global transportation and power generation, yet it faces increasingly rigorous regulatory frameworks aimed at the radical reduction of greenhouse gas emissions, particularly carbon dioxide (CO 2 ), and criteria pollutants such as nitrogen oxides (NO x ) [1, 2]. To meet these mandates while transitioning toward carbon-neutral mobility, the industry has focused heavily on lean-burn and highly diluted combustion strategies [3, 4]. By operating with a high global excess air ratio (ϕ) or substantial levels of exhaust gas recirculation (EGR), engines can achieve significantly higher thermal efficiency than conventional stoichiometric systems [5, 6]. The fundamental thermodynamic advantage of operating under lean or diluted conditions is derived from the ideal Otto cycle, where the theoretical thermal efficiency (η th ) is defined as: � th ¼ 1 � 1 CR �� 1 where CR represents the compression ratio and γ is the specific heat ratio, defined by the ratio of specific heat at constant pressure to the the specific heat at constant volume, C p / C v [4, 7, 8]. Diluted mixtures enhance effciency through two main mechanisms. First, they increase the value of γ because air and triatomic residual gases like CO 2 and H 2 O have different thermodynamic properties than fuel-rich fuel-air mixtures. Second, lower peak temperatures reduce wall heat transfer losses and lead to a higher specific heat ratio too as except for noble gases the specific heat ratio decreases when the temperature rises. Therefore it is beneficial to have a cool process gas in order to improve the efficiency by using the higher specific heat ratio [4, 6, 8, 9, 10]. Furthermore, lean operation leads to a dethrottling effect reducing pumping losses especially during part-load operation [6, 7]. Conventional spark-ignition (SI) systems encounter severe operational barriers when dilution levels are increased or the mixtures are leaned out as the minimum ignition energy (MIE) required to ignite the mixture increases [11]. In ultra-lean mixtures, the flame propagation speed decreases heavily, causing longer ignition delays and increasing the cycle-to-cycle variability that is usually quantified by the coefficient of variation in indicated mean effective pressure, COV IMEP . With that the risk of partial burns or total misfires increases [3, 12, 13]. Pre-chamber assisted combustion, or turbulent jet ignition (TJI), helps to overcome the limitations of conventional SI systems. By initiating combustion within a small separated volume, highly reactive turbulent hot jets are expelled through one or more small transition bores into the main combustion chamber. These emerging jets serve as a locally distributed ignition source, providing multiple ignition kernels and a greatly enlarged surface area compared to a standard spark plug [14, 15]. This way pre-chamber ignition can provide an ignition energy in the order of magnitude of 80 J, whereas conventional SI engines deliver only up to 120 mJ [4, 11]. Minimum ignition energies roughly 3 orders of magnitude higher than typical SI engines enable ignition while ensuring fast and stable flame propagation in very lean or highly diluted mixtures [14, 16]. Pre-chamber systems can be categorized into passive and active systems [3, 17, 18]. Passive systems rely entirely on the pressure difference created during the compression stroke to force the main chamber mixture via the transition bores into the pre-chamber. While mechanically simpler, passive systems often suffer from poor 310 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="311"?> scavenging, leading to high concentrations of residual exhaust gases that compromise ignitability and limit the engine’s tolerance for EGR. On the other hand, active prechamber systems use an additional fuel supply to precisely control the mixture composition in the PC independently of the main chamber equivalence ratio. This active control is used to set the pre-chamber’s equivalence ratio and the present residual gas fraction at the moment of the spark discharge and thereby determining the reactivity of the emerging jet [9, 16, 19]. The underlying physics of TJI identifies two primary ignition mechanisms: flame ignition and jet ignition. Flame ignition - which can be described as ignition by a reacting jet - occurs when the pre-chamber flame survives the high stretch rates and thermal heat losses associated with passage through the transition bore. The resulting jet contains wrinkled turbulent flames and a high concentration of active radicals, such as H*, O*, and OH*, which initiate near-instantaneous combustion across the entire surface of the emerging jet. Conversely, jet ignition - also known as ignition by a reacted jet - involves a quenched flame where only hot combustion products and reactive species are ejected. In this mode, re-ignition in the main chamber is typically delayed and initiates at the lateral boundaries of the jet downstream from the nozzle as the hot products mix with the fresh charge [3, 13, 16]. To characterize the geometric interaction between the prechamber and the main combustion chamber, the area-to-volume (A/ V) ratio is used. This parameter dictates the fluid-dynamic and thermochemical behavior in the two volumes. The ratio is defined as the quotient of the total cross-sectional area of all transition bores of orifices and the internal volume of the pre-chamber [7, 8, 20]: A = V ¼ P A orifice V PC Gussak et al. established that the most favorable results for ignition stability and engine efficiency are typically achieved when this ratio lies in the range of 0.03 to 0.04 cm 2 cm −3 [8, 21]. This optimized range leads to a balance between the pressure build-up necessary for a high-momentum jet discharge and the efficient gas exchange required to maintain an ignitable mixture [5, 19, 21]. Vera-Tudela et al. also studied that too small transition bores lead to high heat losses and consequently, to a longer ignition delay in the main chamber. Especially configurations with low A/ V ratios and therefore small orifices are prone to thermal quenching shifting the ignition mode form flame ignition to jet ignition [18]. Extensive research has been conducted to characterize the complex physics of prechamber assisted combustion, focusing on the interplay between thermochemical reactivity and fluid dynamics. Gussak et al. pioneered this field with the development of the LAG (Avalanche Activated Combustion) process, identifying that chemically active products of incomplete combustion from fuel-rich pre-chamber mixtures (ϕ ≈ 1.4 − 2.5) are essential for promoting avalanche chemical reactions in ultra-lean main charges [21]. Yamaguchi et al. categorized the ignition process in the main chamber into four distinct patterns - chemical chain, composite, flame kernel torch, and flame front torch ignition - concluding that the composite ignition mode, which utilizes both active radicals and thermal effects, is most favorable for extending lean limits [22]. Biswas et al. employed simultaneous high-speed Schlieren and OH* chemiluminescence imaging to define two fundamental ignition mechanisms: flame ignition (reacting jet) and jet ignition (reacted jet), while establishing that a critical global Damköhler number (e. g., Da = 140 for Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 311 <?page no="312"?> methane and Da = 40 for hydrogen) serves as the limiting parameter for successful reignition [13]. Additionally, Biswas and Qiao demonstrated that utilizing supersonic hot jets from converging-diverging nozzles can shorten ignition delays and extend the lean flammability limit of H 2 / air mixtures down to ϕ = 0.22 [12]. In engine-relevant applications, Attard et al. implemented active pre-chambers in light-duty gasoline engines, achieving stable combustion at ϕ ≈ 0.5 with a 20 % improvement in peak efficiency compared to conventional SI modes [23]. More recently, Korb et al. investigated internal charge motion and found that radial nozzle orientations are superior to tilted designs for generating the tumble flow necessary for efficient pre-chamber scavenging, while Cai et al. established an ignition regime boundary for methane/ air mixtures at a global Da of 0.1, highlighting that successful ignition under engine-like conditions requires heat release rates sufficient to compensate for high scalar dissipation [1, 24]. A major challenge in the practical application of TJI is the presence of remaining exhaust gas from previous cycles. Residual gas act as both a thermal sink and a chemical diluent, slowing the early stages of flame kernel development in the pre-chamber and reducing the peak pressure rise necessary for high-momentum jet discharge. Effective scavenging is needed to flush these residuals and ensure a high-reactivity mixture near the spark gap inside the prechamber [6, 9, 19, 20, 25]. While geometric parameters such as orifice diameter and volume have been extensively investigated, there is a critical need for fundamental studies that isolate the influence of gas composition - specifically the interplay between fuel enrichment and residual gas fractions - on the jet dynamics and the transition between ignition modes. This paper presents an experimental investigation on the influence of the pre-chamber gas composition on the ignition process using an actively fueled single-hole pre-chamber system. This study systematically varies the pre-chamber equivalence ratio (ϕ = 0.9, 1.0, 1.2 and 1.5) to explore the effects of fuel enrichment. To mimic the presence of residual gas fractions found in real engine applications, the oxygen content of the air was varied to 21 %, 18 % and 15 % prior to ignition. Those oxygen contents represent exhaust gas rates of 0 %, 14.29 % and 28.57 % respectively. Methodoloy The experiments are performed in a constant flow chamber that is usually used to investigate diesel sprays or hydrogen jets and can provide constant ambient conditions of up to 100 bar and 1000 K with continuous scavenging of either air or pure nitrogen. The constant gas flow has a velocity of around 30 mm s −1 . Temperature and pressure can be adjusted independently. The CFC provides thermodynamic boundary conditions similar to engine application and with five windows with a diameter of 125 mm a high optical access so that the pre-chamber discharge can be investigated by the means of High-speed Schlieren imaging and recording the OH* chemiluminescence signal. The pre-chamber is mounted to the bottom of the CFC and can be cooled or heated by an external temperature control. In this work the test bench is operated with pure air scavenging as well as pure nitrogen at 1 MPa and 573 K at all operating points. The pre-chamber temperature is regulated to 363 K. Outside the CFC air and fuel are premixed at a defined ratio using a gas mixer with two mass flow controllers. The pre-chamber is fueled by the premixed methane-air-mixture using a DI-injector that is supplied by the gas mixer. The schematic 312 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="313"?> of gas supply to the pre-chamber is shown in Figure 1. The sequence of mixture preparation begins with two gas reservoirs: one for the fuel gas and one for air. In the subsequent step, the pressure of both gases is reduced to the desired level using pressure control valves. Following this, the mass flow rates of both gases are regulated by two electronically controlled mass flow controllers to achieve the target equivalence ratio within the pre-chamber. Two check valves and a solenoid valve are incorporated to prevent uncontrolled outflow or backflow of gases. Immediately before the DI-injector, an additional check valve and a filter that simultaneously acts as a flame arrestor are installed. The DI-injector is a Bosch single-hole GDI injector. Fig. 1: Schematic of the pre-chamber fueling and gas mixer To make sure that there are no exhaust gases remaining between the different ignitions, the pre-chamber is flushed with fresh air-fuel-mixture after each ignition. The mixture inside the pre-chamber is ignited by a NKG spark plug with M8 thread size. A miniature Piezocryst T1-A1 piezoelectric pressure sensor is mounted to the pre-chamber. The pressure signal of the miniature pressure transducer is amplified using a Kistler 5018A charge amplifier. The pressure sensor is positioned next to the spark plug on the back side of the pre-chamber. Figure 2 shows the geometry of the single-hole pre-chamber and the positioning of the pressure sensor and the spark plug. Fig. 2: Internal pre-chamber geometry The volume of the pre-chamber is 1.086 cm 3 , it has a 4.2 mm long throat with a 4.5 mm diameter. The tranistion bore has a diameter of 1.0 mm and a length of 1.3 mm. This results in a cross-section area-to-volume ratio A/ V of 0.0072 cm 2 cm −3 for the used prechamber. As the single-hole test pre-chamber is used to investigate the behavior of one isolated transition bore the area-to-volume ratio is reduced compared to the recommended values between 0.03 to 0.04 cm 2 cm −3 . Using a pre-chamber with four to six transition bores like the tested one results in a recommended area-to-volume ratio. The exhaust gas rates Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 313 <?page no="314"?> of 0 %, 14,29 % and 28,57 % are achieved by adjusting the oxygen content of the air used with the gas mixer to contain 21 %, 18 % and 15 % oxygen. The equation to describe the relation between EGR rate and oxygen content is: EGRrate ¼ 1 � O 2 � content 21 With this approach no hot exhaust gases are used to mimic the EGR but additional nitrogen as an inert gas to dilute the mixture. This isolates the influence of charge dilution on the combustion without handling real exhaust gases. Nitrogen differs from exhaust gases containing CO 2 , CO, H 2 O, NO x and unburned hydrocarbons with different specific heat and chemical properties [11]. As the focus of the experiments are the ignition characteristics and not NO x or soot the simplification using nitrogen to dilute the mixture is deemed reasonable. Optical Setup The optical setup of the experiments is depicted in figure 3. A single pass Schlieren setup is used with a green LED (532 nm) light source. The Schlieren images are recorded with a Photron SA-Z high-speed camera, equipped with a Tamron SP 70-200 mm f/ 4.5-5.6 lens, operating at a framerate of 40 kHz with a shutter time of 1.00 μs. In addition to the Schlieren images, the chemiluminescence of the hydroxyl molecule is recorded. To achieve simultaneous recording, a dichroic mirror is installed into the light path of the Schlieren setup. The dichroic mirror transmits the light form the green LED used as light source for the Schlieren imaging, but reflects shorter wavelength, such as the OH*-signal at 307 nm. The OH*-signal is filtered out using an OH filter (Schott UG-11). The remaining signal then passes a high-speed intensifier (LaVision HS-IRO) and is recorded by a second Photron SA-Z equipped with a UV-objective form Sill Optics that has a focal length of 105 mm. The intensified OH*-signal is still weak, therefore a shutter time of 1000 μs and a relatively low framerate of 1000 Hz had to be chosen. With the low framerate, the OH*signal for these experiments is used as a qualitative measure to show whether the emerging jet contains active radicals. 314 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="315"?> Fig. 3: Measurement setup The experiments are performed using pure nitrogen and air as scavenging gas in the CFC. The results shown in this paper are acquired while constantly flushing the CFC with air. Ambient conditions in the CFC are regulated to be 1 MPa and the scavenging gas temperature is controlled to be 573 K. The wall temperature of the CFC is maintained at 473 K. An external temperature control is used to keep the pre-chamber at a constant 363 K. The supply pressure of the methane-air mixture is kept at 2 MPa. For each oxygen content used with the gas mixer the equivalence ratio ϕis varied to be ϕ = 0.9, 1.0, 1.2 and 1.5. Each combination of equivalence ratio and oxygen content is ignited in 16 independent measurements. All of those igntions are then analyzed separately, but for the final result an average value is calculated. The acquired images were processed using a self-developed MATLAB program, that uses the intensity gradients in the Schlieren images to extract the jet structure. A detailed description of the image processing methodology is provided in Rieß [26] and Peter [27]. The data processing program calculates the penetration depth based on the image resolution of 9.55 px mm −1 . Together with the frame rate the jet propagation speed can then be calculated. Figure 4 shows an example of the Schlieren images. The turbulent hot jet is clearly visible exiting the single-hole pre-chamber. Fig. 4: Example image of the raw Schlieren recording of the pre-chamber Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 315 <?page no="316"?> Results In general the results show that the penetration speed of the pre-chamber jet is highest at 21 % oxygen, noticeably slower for 18 % oxygen and considerably slower for 15 %. The results of the experiments are plotted in the following figures comparing the three different oxygen concentrations for each tested equivalence ratio. The figuers 5 till 8 show the penetration depth of the tip of the pre-chamber jet. The standard deviation is included by the shaded area around the averaged value. The plotted penetration depth data was smoothed using a Savitzky-Golay filter with a third-order polynomial and an odd window length selected to balance noise reduction and feature preservation. Figure 5 shows the three oxygen contents for the equivalence ratio ϕ = 1.0. As expected the penetration of the jet is slower for 18 % oxygen. Calculating the propagation speed shows that the prechamber jet propagates at only roughly half the speed compared to the mixture using 21 % oxygen. Looking at the penetration depth for 15 % oxygen it is apparent that the momentum of the jet is reduced by one order of magnitude. The corresponding pressure signals are plotted in figure 9. The unfiltered pressure difference between the pre-chamber and the main chamber (CFC) is shown. Standard deviation is indicated by the shaded areas. The left graph in figure 9 shows the pressure build-up for the experiments with 15 % oxygen, the center one for 18 % and the right one for 21 %. As observed in figures 5 through 8, the steepest slope of the penetration curves - representing the maximum jet velocity - aligns temporally with the timing of peak pressure difference (ΔP) shown in figure 9. This correlation confirms that the jet momentum is primarily blowdown-driven, where the thermodynamic work of the pre-chamber combustion is converted directly into kinetic energy at the nozzle [7]. It solidifies the difference between the three oxygen contents. The pressure rise for ϕ = 1.0 at 21 % is considerably higher than for 18 % oxygen. Whereas the pressure does not rise noticeably for the experiments with ϕ = 1.0 at 15 % oxygen. Due to the lack of sufficient oxygen (15 % O2), the laminar burning velocity of the methane-air mixture is reduced to a critical level where the rate of chemical heat release is roughly equaled or exceeded by the thermal heat losses to the pre-chamber walls. Consequently, the flame propagation is too slow to produce a significant pressure differential (ΔP), leading to a phlegmatic discharge with jet momentum reduced by an order of magnitude compared to the undiluted case. From a fluid-dynamic perspective, this failure indicates that the global Damköhler number has dropped below the critical threshold required for sustained ignition [3]. In this 15 % oxygen regime, the chemical reaction timescales are too slow to compete with the high scalar dissipation rates and turbulent mixing occurring at the orifice exit, preventing the formation of self-sustaining ignition kernels in the main chamber [3, 13]. 316 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="317"?> Fig. 5: Penetration depth of the pre-chamber jet for different oxygen contents at equivalence ratio ϕ = 1.0 Figure 6 shows the penetration depth for slightly fuel-rich conditions at a equivalence ratio ϕ = 1.2. It shows a similar behavior like the previous figure. Again the experiments done with 21 % oxygen penetrate the CFC the fastest, while the ones with 18 % oxygen propagate roughly with half the speed. Comparing the propagation behavior to the ones at ϕ = 1.0 (Fig. 5) slower jet penetration and thus lower momentum can be identified for the pre-chamber charges with 21 % and 18 % oxygen. Comparing the jet penetration for 15 % oxygen it can be seen that the slightly fuel-rich mixture propagates faster by roughly 50 % compared to the stoichiometric mixture at ϕ = 1.0. Based on the laminar burning velocity of methane a slight reduction of the jet penetration is expected as the pressure build-up in the pre-chamber is reduced by the slower burning rate of the mixture. The pressure traces depicted in figure 9 underlay that observation. The pressure difference between the prechamber and the main-chamber is significantly reduced when using air with less oxygen. Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 317 <?page no="318"?> Fig. 6: Penetration depth of the pre-chamber jet for different oxygen contents at equivalence ratio ϕ = 1.2 Figure 7 shows the penetration depth of the pre-chamber jets at fuel-rich conditions with an equivalence ratio of ϕ = 1.5. It is clearly visible that the propagation speed of the prechamber jet is reduced even further by the richer mixture for both 21 % and 18 % oxygen. For 15 % oxygen the charge inside the pre-chamber could not be ignited by the used spark plug. Therefore no pre-chamber jet could be generated. The reduction in penetration depth is supported by the pressure signal in figure 9. The pressure difference between prechamber and main chamber is marginal and delayed for the experiments with 21 % oxygen. There is no distinguished peak like for ϕ 0.9, 1.0 or 1.2 but only a shallow rise. For 18 % oxygen the pressure signal for combusting with ϕ = 1.5 is barely detectable by the used pressure transducer and for 15 % oxygen no pressure difference between the prechamber and the main chamber could be recorded as no ignition started. 318 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="319"?> Fig. 7: Penetration depth of the pre-chamber jet for different oxygen contents at equivalence ratio ϕ = 1.5 To see if the effects of the dilution with nitrogen in the pre-chamber mixture are also notable at leaner conditions, the experiments are done at an eqivalence ratio of ϕ = 0.9. The results are depicted in figure 8. As for the stoichiometric and the slightly rich air-fuel mixtures the penetration depth shows the same behavior for ϕ = 0.9. The propagation speed of the pre-chamber jet is fastest for 21 % oxygen, reduces to roughly half the speed for 18 % and further slows down to about one fifth for 15 % oxygen. The behavior is similar to the results that were acquired at ϕ = 1.2. Comparing the penetration depth of the prechamber jet of ϕ = 0.9 to the experiments done at a stoichiometric ϕ of 1.0 it can be noted, that for 21 % and 18 % oxygen the penetration is slower. Based on the laminar burning velocity of methane this behavior is expected. The penetration for 15 % oxygen seems to be less effected by the dilution at ϕ = 0.9 than 1.0, indicated by a faster propagation speed. The pressure data for ϕ = 0.9 in figure 9 for 15 % is so low that no real conclusions can be drawn from that. For 18 % oxygen the pressure rise caused by the combustion inside the pre-chamber for ϕ = 0.9 is higher than the other equivalence ratios. Like all pressure traces no sharp peak can be identified in the pressure signal for 18 %. Only for 21 % oxygen the pressure traces show peaks right after the ignition. For ϕ = 0.9 the pressure rise is nearly identical with the one at stoichiometric conditions except the height of the peak is slightly lower. Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 319 <?page no="320"?> Fig. 8: Penetration depth of the pre-chamber jet for different oxygen contents at equivalence ratio ϕ = 0.9 Fig. 9: Comparison of the mean pressure difference between the pre-chamber and the main chamber for different oxygen contents and equivalence ratios The penetration depth and the pressure difference give information about the momentum of the jet that is exiting the pre-chamber via the transition bore towards the main chamber. To distinguish between the ignition modes described by Biswas knowledge about active radicals is necessary. While Biswas defines flame ignition as a ‘reacting jet’ containing wrinkled turbulent flames and active radicals like OH, jet ignition involves a quenched flame where only hot combustion products and potentially unreacted species are ejected [3]. Therefore the OH* chemiluminescence was recorded. Because the CFC was scavenged with air, no new radicals were formed via main-chamber combustion, resulting in a weak but traceable OH*-signal. In figure 10 the OH*-signal can be seen. It is plotted over the 320 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="321"?> forty averaged Schlieren images that correspond to the OH* time frame. It shows the OH*signal for one ignition at ϕ = 1.2 while using 21 % oxygen recorded 10 ms after the electronic ignition signal in the pre-chamber. The OH*-signal is only visible with some distance to the nozzle tip. That might be due to the comparably low frame rate to get signal. Further experiments need to be done to get more information about the spatial distribution of the active radicals. Fig. 10: Single OH*-signal placed over corresponding averaged Schlieren images at ϕ = 1.2 while using 21 % oxygen recorded 10 ms after the electronic ignition signal Figure 11 shows the number of ignition events where OH* was detected depending on the oxygen content and the equivalence ratio. Each operating condition was ignited 16 times. Generally the OH* chemiluminescence signals reveal a strong dependency of the ignition behavior on both the equivalence ratio and the oxygen concentration. The maximum number of ignition events with detectable OH* occurred at ϕ= 1.2, regardless of dilution. This aligns with the foundational research of Gussak et al., who identified that slightly rich pre-chamber mixtures are optimal for the production of chemically active products of incomplete combustion [21]. At the stoichiometric point (ϕ = 1.0) with 21 % oxygen, fewer OH* events were detected despite this condition producing the highest jet exit speeds. This finding supports Gussak’s observation that stoichiometric pre-chamber mixtures can be ‘phlegmatic’. The rapid consumption of radicals during the intense pre-chamber burn leaves few active species for the emerging jet, making jet ignition by hot products the dominant mechanism. Conversely, for fuel-rich points without dilution, the sustained OH*-signal indicates a transition to flame ignition, where the flame survives the nozzle passage. The reduced number of OH* detection at 18 % oxygen demonstrates that dilution increases the likelihood of jet ignition. This behavior is consistent with the quenching mechanisms describedby Vera-Tudela et al. [18]. The higher simulated EGR rate reduces the combustion temperature and pressure rise, facilitating thermal quenching as the flame propagates through the transition bore [18]. The observed lower pressure rise and slower jet penetration under 18 % and 15 % oxygen further underline the reduced reaction kinetics Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 321 <?page no="322"?> caused by the presence of inert diluents, which act as a thermal sink and slow the chemical timescales relative to the mixing scales. At leaner conditions (ϕ = 0.9), the number of ignition events with active OH radicals exiting the pre-chamber decreases, especially at 18 % and 15 % oxygen. Fig. 11: Number of ignition events - out of 16 - with OH* detected depending on oxygen content and equivalence ratio Conclusion This study presented an experimental investigation into the influence of pre-chamber gas composition on the ignition process of an actively fueled single-hole pre-chamber system. By systematically varying the equivalence ratio (ϕ = 0.9, 1.0, 1.2 and 1.5) and the available oxygen content (21 %, 18 % and 15 %), the research clarified how fuel enrichment and residual gas fractions dictate the transition between flame ignition and jet ignition modes. The experimental results lead to the following primary conclusions: 1. Thermochemical Sensitivity: The presence of active radicals (OH*) in the emerging jet is highly sensitive to both the equivalence ratio and dilution. A slightly rich mixture of ϕ = 1.2 was found to be the most robust operating point, providing the maximum number of reacting jets regardless of the oxygen level. This supports the foundational theory of chemically active jet ignition, which posits that rich-burn intermediate species are essential for providing the chemical energy necessary for stable ignition in diluted environments. 2. Ignition Mode Transition: While stoichiometric mixtures (ϕ = 1.0) achieved the highest jet exit speeds, they exhibited ‘phlegmatic’ behavior with minimal detectable radicals. Under these conditions, jet ignition by hot products is the dominant mechanism. Conversely, increasing dilution (simulating higher EGR rates) facilitates 322 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="323"?> thermal quenching at the nozzle, forcing a transition from flame ignition to jet ignition as combustion temperatures and pressure rise rates decrease. 3. Dilution Limits: High residual gas fractions (mimicked by 15 % O2) significantly reduce reaction kinetics, leading to slower jet penetration and, in fuel-rich cases, total ignition failure within the pre-chamber. These findings highlight that the effectiveness of Turbulent Jet Ignition (TJI) is fundamentally governed by the chemical state of the jet at the nozzle exit. Ultimately, these results confirm that the transition between reacting and reacted jets is not merely a function of fluid dynamics, but a complex interplay between chemical reaction rates and thermal quenching at the nozzle. These thermochemical sensitivities underline the necessity of active scavenging strategies that can precisely maintain optimal pre-chamber reactivity even in the presence of high residual gas fractions. Furthermore, the performance of passive pre-chamber systems requires further detailed investigation, as their inherent scavenging limitations make them particularly sensitive to increased EGR rates, which often result in higher residual fractions within the pre-chamber than the main chamber average and can severely restrict the stable dilution range. To further understand the initiation of these processes, future research will utilize optical measurements inside the pre-chamber volume to correlate internal flame development with the resulting jet reactivity. Acknowledgements The authors gratefully acknowledge funding of the Erlangen Graduate School in Advanced Optical Technologies (SAOT) and by the Federal Ministry for Economic Affairs and Energy in the IGF/ CORNET research project ‘PIONEER’. Special thanks are due to David Lindemann for his assistance in evaluating the measurements. Nomenclature ϕ Equivalence ratio η th Theoretical thermal efficiency γ Specific heat ratio ΔP Pressure difference A Area C p Specific heat at constant pressure C v Specific heat at constant volume CFC Constant flow chamber CO 2 Carbon Dioxide CO Carbon Monoxide COV IMPE Coefficent of variation in indicated mean effective pressure CR Compression ratio Da Damköhler number DI Direct injection EGR Exhaust gas recirculation GDI Gasoline direct injection H 2 Hydrogen Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 323 <?page no="324"?> H 2 O Water ICE Internal combustion engine LED Light emitting diode MIE Minimum ignition energy NO x Nitrogen oxides OH* Hydroxyl radical PC Pre-chamber SI Spark ignition TJI Turbulent jet ignition V Volume V PC Pre-chamber volume References [1] J. Cai, K. Sun, Y. Feng, M. Jia, Z. Lu, L. Shi und T. Wang, “Turbulent jet ignition of ultra-lean methane/ air mixture under engine-like condition,” Physics of Fluids, Bd. 33, November 2021. [2] P. Sharma, Q. Tang, R. Sampath, P. hlaing, M. E. Marquez, E. Cenker und G. Magnotti, “Optical Study of Pre-Chamber Assisted Internal Combustion Engine at Lean Limits,” SSRN Electronic Journal, 2022. [3] S. Biswas, Physics of Turbulent Jet Ignition, Cham: Springer International Publishing AG, 2018. [4] M. Silva, X. Liu, P. Hlaing, S. Sanal, E. Cenker, J. Chang, B. Johansson und H. G. Im, “Computational assessment of effects of throat diameter on combustion and turbulence characteristics in a pre-chamber engine,” Applied Thermal Engineering, Bd. 212, p. 118595, July 2022. [5] Schumacher, Moritz and Russwurm, Tim and Wensing, Michael, “Pre-chamber Ignition System for Homogeneous Lean Combustion Processes with Active Fuelling by Volatile Fuel Components,” in Ignition Systems for Gasoline Engines, 2018. [6] T. Russwurm und M. Wensing, “Active Pre-chamber Ignition Systems: An Enabler for New ICE Combustion Concepts,” in 9th European Combustion Meeting (ECM2019), 2019. [7] Q. Tang, R. Sampath, M. E. Marquez, P. Sharma, P. Hlaing, M. B. Houidi, E. Cenker, J. Chang, G. Magnotti und B. Johansson, “Optical diagnostics on the pre-chamber jet and main chamber ignition in the active pre-chamber combustion (PCC),” Combustion and Flame, Bd. 228, p. 218-235, June 2021. [8] T. Russwurm, Aktive Vorkammerzündung für Personenkraftwagen, Düren: Shaker Verlag, 2023. [9] T. Russwurm, A. Peter, L. Weiß, S. Rieß und M. Wensing, “Optical Investigations on the Jet Propagation of an Actively Fuelled Pre-Chamber Ignition System with Highspeed Schlieren Imaging,” in THIESEL 2020.Thermo-and Fluid Dynamic Processes in Direct Injection Engines.8th- 11th September, 2020. [10] M. Schumacher und M. Wensing, “A Gasoline Fuelled Pre-Chamber Ignition System for Homogeneous Lean Combustion Processes,” in SAE Technical Paper Series, 2016. [11] J. B. Heywood, Internal combustion engine fundamentals, New York [u. a.]: McGraw-Hill, 1988. [12] S. Biswas und L. Qiao, “Prechamber Hot Jet Ignition of Ultra-Lean H2/ Air Mixtures: Effect of Supersonic Jets and Combustion Instability,” SAE International Journal of Engines, Bd. 09, p. 1584-1592, April 2016. [13] S. Biswas, S. Tanvir, H. Wang und L. Qiao, “On ignition mechanisms of premixed CH 4 / air and H 2 / air using a hot turbulent jet generated by pre-chamber combustion,” Applied Thermal Engineering, Bd. 106, p. 925-937, August 2016. 324 F. Wittmann, F. Schönhofer, C. Kowis, R. C. Mallada, S. Rieß, M. Wensing <?page no="325"?> [14] D. Sen und S. Biswas, “Optical prechamber-equipped high-pressure large-bore optical combustor for fundamental combustion studies,” Review of Scientific Instruments, Bd. 96, March 2025. [15] A. Ramachandran, R. Soni, M. Roßmann, M. Klawitter, C. Gößnitzer, J. Woisetschläger, A. Tilz, G. Pirker und A. Wimmer, “High Speed Particle Image Velocimetry in a Large Engine Prechamber,” Flow, Turbulence and Combustion, Bd. 113, p. 1003-1023, September 2024. [16] S. Zhu, S. Akehurst, A. Lewis und H. Yuan, “A review of the pre-chamber ignition system applied on future low-carbon spark ignition engines,” Renewable and Sustainable Energy Reviews, Bd. 154, p. 111872, February 2022. [17] T. Russwurm, A. Peter, L. Strauß, S. Rieß und M. Wensing, “Investigations on an active prechamber ignition system in a combustion chamber,” International Journal of Engine Research, Bd. 24, p. 2422-2436, September 2022. [18] W. Vera-Tudela, C. Barro und K. Boulouchos, “Investigations on spark pre-chamber ignition and subsequent turbulent jet main chamber ignition in a novel optically accessible test rig,” International Journal of Engine Research, Bd. 23, p. 1543-1555, May 2021. [19] T. Russwurm, T. Achenbach und M. Wensing, “Investigations on lean and EGR-diluted combustion with active Pre-Chamber Ignition using a 1D-simulation model,” in International Conference on Ignition Systems for Gasoline Engines - International Conference on Knocking in Gasoline Engines, Berlin, 2022. [20] J. Antolini, P. Sementa, C. Tornatore, F. Catapano, B. M. Vaglieco, J. M. Desantes und J. J. López, “Effect of passive pre-chamber orifice diameter on the methane combustion process in an optically accessible SI engine,” Fuel, Bd. 341, p. 126990, June 2023. [21] L. A. Gussak, V. P. Karpov und Y. V. Tikhonov, “The Application of Lag-Process in Prechamber Engines,” SAE Transactions, Bd. 88, p. 2355-2380, 1979. [22] S. Yamaguchi, N. Ohiwa und T. Hasegawa, “Ignition and burning process in a divided chamber bomb,” Combustion and Flame, Bd. 59, p. 177-187, February 1985. [23] W. P. Attard, N. Fraser, P. Parsons und E. Toulson, “A Turbulent Jet Ignition Pre-Chamber Combustion System for Large Fuel Economy Improvements in a Modern Vehicle Powertrain,” SAE International Journal of Engines, Bd. 03, p. 20-37, May 2010. [24] B. Korb, K. Kuppa, H. D. Nguyen, F. Dinkelacker und G. Wachtmeister, “Experimental and numerical investigations of charge motion and combustion in lean-burn natural gas engines,” Combustion and Flame, Bd. 212, pp. 309-322, 2020. [25] C. Müller, S. Pischinger, S. Tews, A. Müller und K. Habermann, “Analysis of experimental results with an active pre-chamber ultra-lean burn SI engine,” International Journal of Engine Research, Bd. 22, p. 3103-3127, December 2020. [26] S. Rieß, Einfluss von Kraftstoff und Abgasrückführung auf Gemischbildung und Verbrennung von dieselmotorischen Sprays, Bde. %1 von %22017,2, Aachen: Shaker Verlag, 2017. [27] A. Peter, Charakterisierung der Gemischbildung und Zündung in Diesel- und Dual-Fuel- Brennverfahren, Düren: Shaker Verlag, 2022. Influence of the Pre-chamber Gas Composition on the Ignition Process of an Active Pre-chamber 325 <?page no="327"?> Knock Analysis in Hypercar Engine Development D. J. Corrigan, G. Lepri, C. Mecheri & D. Villa (Bugatti-Rimac) Abstract: Hypercars are the highest performance echelon of road transport. Such cars have very high power output to enable performance way in excess of mainstream automotive machinery. Internal Combustion Engines (ICEs) remain dominant in this vehicle type and their performance is limited by a number of factors. One of these, famously, is knock. Both Turbocharged (TC) and Normally Aspirated (NA) ICEs can be used for hypercars. It is easier to achieve high power density with a TC ICE. It is often considered that TC ICEs are more knock prone. Achieving the target vehicle performance with an NA ICE comes with a number of distinct challenges however, meaning that this engine type will also be knock limited. A number of technologies can be considered for knock mitigation, but it is shown that many of these are not applicable to an NA ICE. Performance will be gained for a given hardware configuration if knock is measured, understood and managed. This is true both in a development context and in series production. This paper therefore describes hypercars and their powertrain technology before focusing on the particular challenges of large, high performance NA ICEs with a high cylinder count. It is shown that a total vehicle output of 1800 PS can be reached with a hybridised NA ICE powertrain running at the knock limit. 1 Introduction: What is a hypercar? 1.1 Hypercar history The term supercar has been used as early as the 1910s (1) and refers to a vehicle with a very high performance compared to mainstream machinery. Hypercar is a more recent term with the McLaren F1 perhaps being the first example. These are cars in an even higher category than supercars, typically with extreme top speeds (the McLaren F1 reached 387 kph in 1998) and very high power outputs (the Bugatti Veyron 16.4 was the first series production car with over 1000 PS). Bugatti is arguably the leading hypercar brand and since the introduction of the Veyron in 2005, has brought vehicles to the market with ever increasing power outputs, see Figure 1. <?page no="328"?> Figure 1: Power output of Bugatti hypercars since 2005 1.2 Modern hypercar targets High output is not the only requirement of a hypercar powertrain. The powertrain should also be an emotional device. This is somewhat harder to define but is considered one of the reasons why fully electric hypercars, such as the Rimac Nevera, are less successful in the market, despite power output above 2000 PS. For many customers, it is the ICE that gives this emotional resonance. The engine designer should not therefore consider only what configuration may give the most power, but also what would give the best experience to the customer. 2 Hypercar ICE technology 2.1 To boost or not to boost? If we consider the McLaren F1 as the first hypercar, this had an NA ICE of 6.1 L capacity in a V12 configuration. It had a power output of 627 PS. Whilst this was a very large number when the car was launched in 1992, this is far below the level expected in the hypercar class today, as can be seen in Figure 1. Bugatti vehicles since the EB 110 of 1991 have all been turbocharged. The base engine architecture of the 2005 Veyron and the 2023 Mistral is the same: an 8.0 L 16 cylinder in a W configuration. Higher pressure boosting has been the main factor enabling the output to increase from 1001 PS to 1600 PS in this time. Specific output of the latest applications is 200 PS/ L. This specific output figure is high, but not unusual for modern performance TC ICEs and in fact even 300 PS/ L was brought to market in 2025 in the Ferrari F80. Boosting comes with two main problems from the customer point of view however. Firstly, a turbocharged engine features a turbine in the exhaust line. This tends to dull the exhaust sound - one of the main emotional connections to the product. Transient 328 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="329"?> response also suffers greatly in comparison to an NA ICE, as there is a lag between when the driver makes a torque request to when the turbocharger reaches its speed target. This can be, at least partially, mitigated by use of parallel-sequintial boosting systems (2) or electrified turbochargers. It was decided to target 1800 PS total powertrain output with Tourbillon but for 800 PS of this to come from electrification. This allows a more modest ICE target of around 1000 PS - a figure achievable with a high performance NA engine of reasonable size for the hypercar class. 2.2 Hypercar NA ICE architecture Power outputs are often discussed based on “Specific Output” meaning power normalised by swept volume i. e. PS/ L. As power is just the product of torque and speed, both of these can be worked on to improve output. The two ways to increase torque are by increasing engine swept volume or by increasing the mean effective cylinder pressure (MEP). It was decided that the Tourbillon ICE should be of broadly similar capacity to the Chiron, hence around 8 L. The obvious way to increase MEP is by boosting, but this was not desired for reasons discussed above. Adequate MEP would therefore be achieved by a combination of positive tuning, a high compression ratio and appropriate combustion phasing. There is no fundamental limit on the rotational velocity of an ICE but there is a standard guideline for the Mean Piston Speed (MPS), as defined below. MPS ¼ 2 Stroke : RPM 60 � � A maximum value considered for a high performance ICE is 25 m/ s (3). This means that for high engine speed, a short stroke is required. The stroke is of course directly proportional to the swept volume as a product with the bore area. For a fixed engine capacity, we can therefore say that the maximum engine speed is driven by the available bore area. This parameter should be as high as possible; in fact high performance NA engines, particularly in racing classes such as F1, normally have very high bore areas in comparison to their swept volume - a high Bore/ Stroke ratio. A key limit for ICE installation in a vehicle is total engine length. This will obviously depend on the bore size and the number of cylinders per bank, along with additional smaller factors such as interbore spacing, bank stagger, timing drives, casting thicknesses etc. Considering reasonable values for the above, a direct relationship can be obtained between ICE length and maximum engine speed, as shown in Figure 2. A V16 configuration was assumed. The bore size was swept from 80 to 100 mm. Knock Analysis in Hypercar Engine Development 329 <?page no="330"?> Figure 2: Relationship between ICE length and maximum speed for a V16 configuration The final configuration chosen for Tourbillon is shown in Table 1. The bore size selected, 92 mm, was to limit the engine length and drives directly the maximum engine speed achievable. Parameter Value Unit No. Of Cylinders 16 - Layout V16 - Bore diameter 92.00 mm Stroke 78.55 mm Bore/ Stroke Ratio 1.17 - Maximum Speed 9300 rpm Maximum MPS 24.4 m/ s Swept Volume 8.355 L Compression ratio 14.4 - Table 1: Architectural choices for Bugatti Tourbillon ICE 2.3 Airflow and combustion limits The specific power target of the ICE to achieve 1001 PS with an 8.355 L swept volume is 120 PS/ L. For the given engine architecture, this can only be met by harnessing Unsteady Gas Dynamics (UGD) to have a Volumetric Efficiency (VE) well in excess of 100 % (4). This effectively means having higher than atmospheric pressure, hence temperature, at the intake valve during its opening duration, and especially shortly before Intake Valve Closing (IVC). VE values of around 120 % can be achieved exploiting this wave action. ICE performance and efficiency depend also on the Compression Ratio (CR). Whilst there can be geometric limits on this parameter, the main issue in maximising CR is knock (5). Knock can be considered a race between the flame front and autoignition of mixture in the regions of the combustion chamber furthest from where combustion begins. This region is known as the end-gas. The autoignition reactions are much more likely to occur 330 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="331"?> at elevated end-gas temperatures. Both a high intake temperature and a high compression ratio will lead to this condition. High intake temperatures can be driven not only by positive tuning but also the installation of the ICE in-vehicle. The ICE installation and cooling systems for a hybrid hypercar are very complex, as shown in Figure 3. This may lead to less than ideal temperature conditions for the ICE and the intake air. Figure 3: Tourbillon cooling systems layout 3 Knock in Hypercar ICEs 3.1 Knock reduction technologies A number of knock reducing technologies have been developed by engineers over the years but not all of these can be applied to NA ICEs. An incomplete list is given in Table 1. Further discussion on these technologies in the context of extension of stoichimetric operation of the ICE is given in (6). Technology Applicability Miller cycle Turbocharger required Atkinson cycle Negative impact on VE De-tuning/ anti-tuning Negative impact on VE High octane fuel Market availability dependent High tumble combustion system Negative impact on VE Prechamber ignition Likely additional complexity/ weight if secondary igniter required Water injection Increased weight, component availability Gasoline Direct Injection (GDI) Potential mixing issues at high speed Table 2: Knock reduction technologies and applicability to NA ICEs Atkinson cycle, de-tuning and antituning would have an excessively deterimental effect on VE. It was decided to base the development on relatively high octane fuel - RON 98 Unleaded Gasoline. This has good market availability in most Tourbillon target markets. Knock Analysis in Hypercar Engine Development 331 <?page no="332"?> The ICE still needs to be compatible with lower octane fuels however (this will be further commented upon later). A high tumble combustion system would increase combustion speed but at the detriment of VE. Gasoline Direct Injection (GDI) should give a useful knock improvement over Port Fuel Injection (PFI) (7). It can be difficult to achieve adequate homogeneity of the air/ fuel mixture at high engine speeds with a high VE intake port design however, as shown in Figure 4. An increase in tumble and a modification of piston bowl design would improve the situation but with deterimental impact on port flow coefficients. Figure 4: Poor air/ fuel homogeneity for high speed high VE ICE with GDI fuelling Water injection is a technology interesting both for knock mitigation and stoichiometric operation extension and has been extensively research by Bugatti-Rimac as documented in (6). It may lead to unstable combustion and misfires however, and implies a vehicle mass increase due to the additional water tank. Prechamber Ignition (PCI) is a technology that is now ubiquitous in Formula 1. The technology has a lengthy history (8) and is now appearing in high performance road cars. Issues with low load running and especially catalyst heating mean that, for a modern passenger car engine, a secondary igniter is generally also required. This adds complexity and mass. That said, Bugatti-Rimac has performed research on PCI systems and these remain interesting for future applications. 3.2 Knock analysis in development As the hardware technologies for knock mitigation are more limited for NA ICEs, it is even more important to have effective calibration and control, including in series production. The reference measurement during ICE combustion development is pressure-based indication with piezoelectric transducers. Due to the large valve sizes of the Tourbillon engine, only M5 sized transducers, such as the Kistler 6054, can be mounted. 200 kHz bandwidth charge amplifiers are used together with high frequency data acquisition systems (Kistler KiBox and AVL Xion). Analysis of some of these data will now be discussed, highlighting some issues particular to this engine type. Figure 5 shows cylinder pressure and apparent heat release from an early prototype Tourbillon engine at 9000 rpm. 100 cycles were recorded. High frequency content can be seen around firing Top Dead Center (TDCF), largely driven by combustion chamber 332 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="333"?> acoustic phenomena (9). High frequency content can also be seen however, in other crank angle regions, far from combustion. Continuous Wavelet Transforms (CWTs) were used to perform Time-Frequency Analysis (TFA) to gain greater insight. A CWT was executed for each cycle, then the CWT values were averaged to give the mean TFA over the acquisition. The results are shown in Figure 6. A number of broadband impulses can be seen across the 720˚ average cycle. The strongest of these is at around -110˚ after TDCF. This corresponds to the IVC angle of the analysed cylinder. The other impulses are driven by the IVC of neighbouring cylinders, with the stength depending on the distance from the cylinder analysed in the figure. This IVC noise is common on high speed engines with high valve lift thus high seating velocities. It was a major issue for cylinder pressure analysis on NA F1 engines. The noise is most problematic in the compression curve, where it can lead to issues with estimation of the compression polytropic coefficient. This may also lead to pegging inaccuracies (depending on the method employed) and errors in heat release analysis, but it is not a major issue for knock. Figure 5: Cylinder pressure and heat release analysis at 9000 rpm Knock Analysis in Hypercar Engine Development 333 <?page no="334"?> Figure 6: Continuous Wavelet Transform analysis at 9000 rpm Another peculiarity of this engine type occurs at the opposite end of the engine speed range. Figure 7 shows data at 1000 rpm. Note that the high frequency content is much reduced in comparison to 9000 rpm. The combustion is also notably more retarded. This is because at low speed there is more time for autoignition to occur as the flame front propogation is much slower. The combustion must therefore be phased appropriately to manage this. The normalised Heat Release Rate (HRR) has a clear inflection point. Mathematical fitting of HRR data, or the cumulative Mass Fraction Burned (MFB) related function, can be performed using Wiebe functions (10). If there is only a single combustion process, a single Wiebe fit should be adequate. A triple Wiebe fit was attempted for the HRR data in Figure 7. Whilst a single Wiebe fit can be done by sweeping the fit parameters manually, multiple fitting is best performed using a global optimiser. A Genetic Algorithm (GA) was used in Matlab to this end. The results are shown in Figure 8. It can be seen that two of the Wiebe functions are making the vast majority of the contribution of fitting the target curve. These are of quite similar size. This could suggest two distinct combustion modes in the same cycle - the flamefront reaction and autoignition. This is despite little evidence of the high frequency oscillations in these cycles. In fact, this seems to be a sort of Spark Assisted Compression Ignition (SACI) type combustion mode, common to very high compression ratio NA engines at very low speed. 334 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="335"?> Figure 7: Cylinder pressure and heat release analysis at 1000 rpm Figure 8: Triple Wiebe fit of 1000 rpm Heat Release Rate data Once data has been acquired, the first step in knock analysis is normally producing a cyclic index of the knock severity. The Maximum Amplitude of Pressure Oscillation (MAPO) is commonly used. This is alternatively referred to as Knock Peak. Thresholding must then be performed to separate cycles into knocking and non-knocking. As seen above, pressure oscillations increase with engine speed, rendering recognition of those Knock Analysis in Hypercar Engine Development 335 <?page no="336"?> driven by autoignition more difficult. A common thresholding approach is therefore to set a limit as a function of engine speed, e. g. a cycle is classified as knocking if its MAPO value in bar is greater than RPM/ 1000. This is simple, somewhat intuitive and easy to apply, but also arguably quite abritrary. An alternative is to separate the cycles based on a mathematical technique. It is often said that knocking cycle data follow a log-normal distribution. This too is an oversimplification. A more accurate statement is that both knocking and non-knocking cycles follow distinct log-normal distributions and this fact can be used to separate between them. This was first shown in (11). A method of implementing this threshold, is performing a lognormal fit of all the cycles in an acquisition, and detecting those at high amplitude which do not align with the fit (12). This analysis is shown at 2000 rpm in Figure 9. It can be seen that the majority of cycles follow the fit line (in fact for nonknocking datasets, all cycles follow the line). Those that do not belong to another population - knocking cycles. A threshold is therefore set based on distance from the fit line. It can be seen for this dataset that this corresponds to a little over 0.3 bar. Visual interpretation of the bandpass filtered pressure, CWT analysis of the strongest cycle and the cyclic relationship between Knock Peak and MFB50 all confirm that knocking cycles are indeed present in the dataset. This is despite the fact that there are no cycles above the commonly applied threshold at 2000 rpm: 2 bar. Figure 9: Knock analysis at 2000 rpm 336 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="337"?> In a development program there is rarely time to examine individual acquisitions in such detail. A data pipeline is therefore required from the cylinder pressure sensor ideally all the way to a datalake accessible to the whole Research and Development (R&D) department. This pipeline is illustrated in Figure 10. Data from the test cell, the Engine Management System (EMS) and various domain controllers is automatically combined and uploaded to a datalake after each acquisition. The data is then available through a Webapp to all stakeholders without needing any special software. An example of an acceleration test is shown in Figure 11. The target values of 0-200 kph in 5 s and 0-300 kph in 10 s are confirmed, and a certain non-zero level of knocking is permitted to help achieve these goals. More advanced users can interface with the datalake directly through Matlab and Python for data-mining and analysis. Such an approach is invaluable for durability cycle analysis. Figure 10: Indicating data pipeline Knock Analysis in Hypercar Engine Development 337 <?page no="338"?> Figure 11: Acceleration test cycle including knock signals 3.3 Production ICE knock control High precision piezoelectric sensors are the reference for development but for production another technology is required. A number of approaches can be applied, as shown in Table 3. In addition to knock feedback, it must also be possible to diagnose a misfire for Onboard Diagnostics (OBD) reasons. Ion sensing can perform the required tasks without the necessity of additional sensors on the engine. This was therefore selected for Tourbillon. Technology Knock Feedback Combustion Feedback Comments Open loop No No Conservative calibration required, non optimum performance Accelerometer Yes No Additional technology required for misfire detection Cylinder pressure Yes Yes Cylinder head modifications required. Additional complexity and mass. Ion sensing Yes Yes No additional mass penalty Table 3: Production knock management approaches Ion sensing has been used in combustion research since the 1920s (13). Whilst dedicated sensors can be used, it is also possible to measure the ion current across the electrodes of a spark plug once the discharge is completed. A bias voltage of the order of 100 V is required and the resulting current is some mA. During combustion, a first peak on the ion signal is driven by the flame front. This is why the technology was originally used for flamefront 338 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="339"?> detection. Thereafter, if knock occurs, oscillations can be seen on the signal due to pressure wave activity in the combustion chamber driving temperature oscillations in the spark plug gap (14). An example from the Tourbillon ICE is shown in Figure 12. It is possible to correlate these oscillations to knocking events. Typically a laboratory grade piezoelectric sensor is used as the reference. Figure 12: Comparison of cylinder pressure and ion current signals during a knocking cycle The low current of the ion measurement makes it somewhat susceptible to electrical noise sources on the engine. The strongest of these is of course the coil discharge. On most engines, combustion events are rather spaced out. The Tourbillon V16 engine fires every 45˚CA however. Indeed, on some cylinders coil discharge noise interferes with the ion current detection window as shown in Figure 13. Work on the wiring loom improved but did not fully resolve the issue. Implementation of cylinder individual detection thresholds allowed robust knock detection from the ion current signal taking acount of this additional, deterministic noise, as shown in Figure 14. Figure 13: Issue with ion current detection window on 16 cylinder engine with even firing Knock Analysis in Hypercar Engine Development 339 <?page no="340"?> Reliable onboard knock detection in production vehicles allows calibration engineers to push harder with the base spark advance of the ICE. This directly gives higher power output to the customer in nominal conditions. The knock detection system can also be used to cope safely with lower octane fuel if RON 98 is not available. Figure 14: Individual cylinder thresholding for robust knock detection considering neighbouring cylinder spark noise 4 Conclusions • ICEs remain the dominant prime mover in most hypercar powertrain architectures • NA ICEs may be preferred to give a more exciting product to the customer. It is more difficult to obtain very high output with an NA engine than a TC one. Electrification can contribute to overall power output but the ICE itself must also be optimised. • The ideal hypercar NA ICE architecture is a tradeoff between a number of factors, one of which is maximum engine speed versus instalability (overall engine length). • A high compression ratio may be beneficial for performance but this, together with pressure wave tuning and difficult heat management, means high performance NA engines can be knock limited - knock is not only a TC ICE concern. • Fewer knock mitigation technologies are appropriate for NA hypercar ICEs hence high attention to knock measurement, calibration and control is fundamental. Tolerating some knock will result in higher performance. • Production engine knock control systems allow higher performance base calibrations to be applied whilst still tolerating knock aggravators which may occur in the field e. g. lower octane fuel. • The Bugatti Tourbillon has achieved its vehicle acceleration targets with a hybridised knock limited NA ICE powertrain - showing that turbocharging is not the only way to match modern hypercar performance expectations. 340 D. J. Corrigan, G. Lepri, C. Mecheri, D. Villa <?page no="341"?> 5 Bibliography 1. Oxford English Dictionary. Supercar (n.). Oxford English Dictionary. [Online] July 2023. https: / / doi.org/ 10.1093/ OED/ 5950338984. 2. The 16-cylinder High-performance Engine in the Bugatti Chiron. Dürheimer, Wolfgang, et al. s. l. : Springer Nature, 2017, MTZ Worldwide, Bd. 78, S. 18-25. 3. Pignone, G. A. und Vercelli, U. R. Motori ad alta potenza specifica. Le basi concettuali della tecnica da competizione. Vimodrone, Italy : Giorgio Nada Editore, 2023. 4. Blair, Gordon P. Design and Simulation of Four-Stroke Engines. Warrendale : SAE International, 1999. 5. Knock: A Century of Research. Corrigan, Dáire James und Stefano, Fontanesi. 2022, SAE International Journal of Engines. 6. Water Injection as an Enabling Technology for Stoichiometric High Performance Normally Aspirated Engines. Corrigan, Dáire James, et al. s. l. : SAE International, 2025. 17th International Conference on Engines and Vehicles. 7. Fuel Injection Strategies to Increase Full-Load Torque Output of a Direct-Injection SI Engine. Yang, J und Anderson, R. W. 1998, SAE Technical Papers. 8. A review of pre-chamber initiated jet ignition combustion systems. Toulson, E., Schock, H. J. und Attard, W. P. 2010. SAE 2010 Powertrains Fuels & Lubricants Meeting. 9. Pressure Waves Accompanying Detonation in the Internal Combustion Engine. Draper, C. s. l. : Institute of Aeronautical Sciences, 1938. Power Plants and Propellers Session, Sixth Annual Meeting, I. Ae. S. 10. Review of the development and applications of the Wiebe function: A tribute to the contribution of Ivan Wiebe to engine research. Ghojel, J. 2010, International Journal of Engine Research. 11. A New Method to Automatically Optimize the Knock Detection Level in the Knock Control System. Iwata, T, Sakakibara, K und Haragucti, H. s. l. : SAE International, 1989. Passenger Car Meeting and Exposition. 12. Engine Knock Detection Methods for Spark Ignition and Prechamber Combustion Systems in a High-Performance Gasoline Direct Injection Engine. Corrigan, Dáire James, et al. s. l. : SAE International, 2022, SAE International Journal of Engines. 13. The Velocity of Flame Propagation in Engine Cylinders. MacKenzie, D. und Honaman, R. K. 1920, Pre-1964 SAE Technical Papers. 14. Spark Control for Ion Current Sensing. Weißgerber, T. und Kortendiek, G. s. l. : Springer, Cham., 2016. Ignition Systems for Gasoline Engines. Knock Analysis in Hypercar Engine Development 341 <?page no="342"?> ISBN 978-3-381-16861-3 IAV SPARK, the Symposium on Pre-Ignition, Auto-Ignition, Reactivity, and Knock, has been a cornerstone of the international engine development community for more than two decades. Bringing together leading researchers, engineers, and industry experts from around the world, the event serves as a unique platform for sharing the latest insights and innovations in ignition systems, knock phenomena, and the efficient combustion of conventional and renewable fuels in internal combustion engines. The success story began with two independent conferences: the International Conference on Knocking in SI Engines and the International Conference on Ignition Systems in SI Engines, launched in 2003 and 2012 respectively. Both events quickly established themselves as highly regarded forums for international knowledge exchange. Given their strong thematic connection, the next logical step was to unite them into a single, more powerful platform. In 2022, the two conferences were merged for the first time under the motto “Two Events - One Goal.” The response from the global engine development community was overwhelmingly positive. Building on this success, the format was continued in 2024, once again demonstrating the strong demand for a shared forum that fosters interdisciplinary collaboration and innovation. Together with our distinguished and increasingly international Advisory Board, we continued to drive the evolution of the event forward. In 2025, the conference crossed the Atlantic for the first time. Hosted in Detroit, the event attracted a global audience and further strengthened its position as a leading international forum for engine combustion and ignition research. For the first time under the new name: IAV SPARK. Today, IAV, the Advisory Board, and I are especially proud to welcome you to IAV SPARK26 in Berlin. Together, we are creating a place where ideas are born, innovations are discussed, and the foundations for the next generation of efficient and sustainable combustion technologies are laid. We invite you to be part of this inspiring community and to help shape the future of mobility.