1. Fachtagung TestRig - Juni 2022 121 Data analysis in Hardware-in-the-loop applied in a complete common rail system for testing of fuel-component compatibility Daniel Correa-Sanchez, M. Sc. Tec4Fuels GmbH, Aachen, Germany Chandra Kanth Kosuru, M. Sc. Tec4Fuels GmbH, Aachen, Germany Dr.-Ing. Hajo Hoffmann Tec4Fuels GmbH, Aachen, Germany Dr.-Ing. Klaus Lucka Tec4Fuels GmbH, Aachen, Germany Abstract: Hardware-in-the-loop testing has been the prime focus of Tec4Fuels over several years. Tec4Fuels has developed a Co- CoS (Complete common-rail system) test bench to conduct compatibility testing of various fuels with the fuel leading components. The test bench is at complete flexibility to incorporate any kind of fuel injection components with minimum effort. The test circuit consists of the following main components in-tank pump, filter, high pressure pump, rail, and injector. The fuel is condensed (after injection) in a reactor and collected back in the tank, to avoid combustion. This provides the benefit of using as low as 30 liters of fuel for testing. Enhancements to the data acquisition system such as information handling in the context of Industry 4.0 and the Internet of Things (IoT) are currently being implemented. Sensors are used to collect as much information as possible, then the data is analyzed to find possible correlations between all the variables and to better understand the dynamics of the system. This big data is also gathered to feed a machine-learning oriented program. Each trial has its own characteristics even if the same initial conditions are used. With every single test performed, more useful information is collected, the predictive analytics model is improved, and a further step towards prescriptive analytics is done. 1. Hardware-in-the-loop testing in Tec4Fuels. Real conditions in engines or heaters can be simulated in the laboratory to research specific interactions or behaviors consuming less resources. This increases the understanding of such systems. It opens the path to further research specially during long operation periods [1]. CoCoS (complete common rail system) test benches check the durability and effects of different fuels and their components. These tests benches are from recent development and applied in the automotive industry. The know-how of the previous Hardware-in-the-loop test benches, such as ENIAK and ATES [2], is enforced and improved to provide the required quality in endurance testing. It is important to mention that the fuel is reintroduced into the system after the injection. This allows to perform long endurance tests with a reduced quantity of fuel. Both dieseland gasoline-common-rail-systems are emulated in different test benches. 1.1 CoCos Test bench. The test bench consists of a fuel tank, an in-tank fuel pump, a filter, an electric motor, a high-pressure pump, a common rail, and an injector. An adjustable control system makes the components work together to perform continuous fuel circulation. The overall setup of the system is simplified and presented in the following figure. 122 1. Fachtagung TestRig - Juni 2022 Data analysis in Hardware-in-the-loop applied in a complete common rail system for testing of fuel-component compatibility Figure 1. General diagram [3]. The in-tank fuel pump is switched on to provide fuel at low pressure to the high-pressure pump. The fuel passes through a filter to avoid small particles. The high-pressure pump is coupled to an electric motor. The rotation speed is translated into pressure applied to the common rail. Then the fuel is distributed to the corresponding injector. The other outputs of the rail are blocked. The injector is activated by a current profile supplied by the injector control unit, which is explained in the next section. The heating rod is attached to an aluminum block to be heated. This emulates an engine’s cylinder mounting of the injector. The injector sprays the fuel into a Reactor where the fuel is condensed and sent back to the fuel tank. The fuels return line of the high-pressure pump is cooled to regulate the fuel tank temperature. The recollection of the fuel leads to a considerably fewer fuel usage. It also allows short testing times in the global development of the project, due to the high-stress operation on the fuel and on the components. The complete testing cycle is divided in three main phases: continuous operation, pause, long pause, and cycled operation. In the continuous operation phase the fuel circulates repeatedly through the components. At the beginning of the test, this phase is usually longer. The stage duration is fixed and can range from 90 to 300 minutes. Afterwards, a settled pause varying between 45 and 120 minutes occurs. Regularly, the duration of each phase is settled to fit several alternating continuous and pause phases, and one long break within 24 hours accomplishing a partial testing cycle. During the cycled operation, partial cycles are performed to achieve endurance tests ranging from 100 to 500 hours, aiming to keep test times as short as possible. Figure 2. Example of a partial test cycle within 24 hours. 1.1.1 High-Pressure Pump The inlet pressure is between 3 and 6 bar, and the highpressure pump can elevate this value up to 2500 bar. A rotating shaft in the center moves the radial-positioned pistons inside, which intakes the low-pressure fuel and press it into the next pipe with high-pressure. The pressure is regulated by a metering unit control valve, which is a solenoid. Electrically it opens or closes depending on the desired pressure, thus regulating the flow of fuel from low pressure part to the high-pressure part of the pump. The feedback is in this case given by the pressure sensor on the rail. 1.1.2 Common Rail The rail can withstand high-pressures even exceeding 2500 bar depending on the model. Multiple injectors can be coupled to the rail. In this test, only one injector is assembled, and the remaining outlets are blocked with especially designed pellets. It has an electrical pressure sensor incorporated by the manufacturer. 1.1.3 Injector heating It is designed to resemble the conditions around the injector when it is inserted in the cavity of an engine during normal operation. The injector heating consists of the heating block and the heating rod. The heating block is made of aluminum due to its light weight and thermal capacity. On the right side of Figure 3 the parts are described. Figure 3. Injector and its heating element [2]. 1. Fachtagung TestRig - Juni 2022 123 Data analysis in Hardware-in-the-loop applied in a complete common rail system for testing of fuel-component compatibility 1.1.4 Injector The injector is the main component under investigation. Different fuels have different effects after long operation times. Before and after an endurance test, a flow rate measurement in the injector is performed to compare the deviations generated due to the constant operation. Internal and external deposits can be formed, and their causes are further investigated [2]. The nozzle temperature is recorded and used as an indirect measure of flow rate during operation. Figure 4. Actual components in the test bench [3]. The activation time of the injectors is led by the injector control unit. The injector is activated at a frequency range between 10 and 40 Hz. It is adjusted to emulate two-thirds of the engine revolutions per minute (rpms) for a two-stroke motor when it reaches 100 km/ h. 2. Data acquisition and user interface During operation, the values of the sensors are being logged each second. For 400 hours of operation at least 1,440,000 data points per sensors are obtained. Considering that a standard Excel sheet has “only” 1,048,576 rows, it exemplifies the size of the data to be handled. The data is saved in a CSV file locally at the test bench, which connected to the network. This allows the data to be saved in a database in the server located remotely. The information is gathered by a Linux-based data acquisition software. This software allows a connection among several devices in the same network. It also deploys a webserver which could be accessed from any browser in any device in the network. Therefore, the status of the test bench can be monitored and evaluated by the engineer at any time. The complication of data loss is reduced to its minimum because any value from the sensors is recorded simultaneously in the database located remotely as well as in the computer at the test bench. This concept of data acquisition process is shown in Figure 5. Figure 5. Simplified overview of the data acquisition procedure [3]. The control panel allows the user to observe the numeric values of the sensors in real time. It is shown in the next picture. Figure 6. Representation of a simple user interface. To acquire an immediate outlook of the current process it is useful that the values are being charted. The time frame can be adjusted to observe the possible trends or patterns. The next figure shows the temperature developing from several thermocouples. The user can select a specific time frame or filter the variables. Figure 7. Example of temperatures changing in a partial cycle. 3. Evaluation of the data 3.1 Interpretation Data without a meaning is merely empty content. Each measured variable expresses a condition in a component during a trial directly or indirectly. 124 1. Fachtagung TestRig - Juni 2022 Data analysis in Hardware-in-the-loop applied in a complete common rail system for testing of fuel-component compatibility Humans can identify changes in recurring patterns easily through comparison. A trained eye can identify problems such as in the following example immediately. The data was taken from another test bench in order to exemplify periodical time series. The cycle of this test bench consisted in heating up a sample until 750°C and then cooling it down to 400°C with air flux. Temperature in the sample is represented in red, pressure in blue. It is observed, that after several successful completed cycles, the time required to reach the breaking point is slowly increasing, thus the system starts to behave in an undesired manner. No abortion criterion was achieved because the temperature and the pressure were within their operative ranges. Figure 8. Temperature and pressure measured and recorded during endurance test. But an algorithm could detect if the next cycle is slightly starting to deviate from the previous series. A notification could be triggered even if the human operator cannot detect the anomaly in the cycle. The following example demonstrates an increase in the injector temperature and a correlated decrease in its nozzle temperature. Deposits inside the spray holes is a possible reason [4]. Unmounting the injector and reviewing its status is unpractical due to the nature of the test, thus an indirect measurement is practical. Figure 9. Temperature in the injector nozzle vs temperature in the injector. It has to be noted that, it can be hard to notice to a human operator depending on the visual representation of the chart. The temperature in the injector in this case was overpassed during a short time outside the normal values (after the initial overshot). Although the small peaks can be effortlessly visually detected, this undesired condition is not distinguished by exclusively setting a range. Figure 10. Nozzle temperature. In the case of diesel systems, the metering unit regulates the high-pressure in the rail. The pressure sensor progress graph will show small pressure drops, but they will be automatically adjusted by the metering unit. The percentage of aperture is controlled and can be logged. By recording and graphing this value, it is feasible to correlate its variation to possible deposits formation. The corresponding fluctuations are marked in blue circles. The temperature in the injector nozzle increases and then decreases for a very short time, which is hard to spot at first sight. Figure 11. Metering unit (black), injector temperature (red), and tank temperature (yellow). An algorithm can supervise two or more variables simultaneously, towards warning the user of a possible failure. 3.2 Open points and future implementation The predictive analytics part can be improved once enough data from many components during many tests can be obtained. Classifying the results and ranking them by the user is thought to be implemented. Therefore, the training data can be obtained and used later to evaluate the performance of future tests. Machine-learning can be more accurate when enough data is available. It can be used to predict the possible outcome, or to identify previously defined behaviors which are not desired or trend to malfunction. Because of the well-defined input system, it is possible to add sensors with wireless capabilities without adding complexity to the data administration. More sensors mean more possible correlations between the variables and more behaviors to analyze, which consequently will improve the understanding of the system. 1. Fachtagung TestRig - Juni 2022 125 Data analysis in Hardware-in-the-loop applied in a complete common rail system for testing of fuel-component compatibility There are tools such as TensorFlow® or PyTorch® which are open source and available to generate predictions using machine learning via neural networks. It is planned to implement different time series analysis methods such as FFT (Fast Fourier Transform) or first derivative and second derivative analysis [5]. 4. Conclusions Hardware-in-the-loop systems were designed to test possible outcomes under harsher conditions. The development time of Complete common-rail systems (CoCoS) are abbreviated by testing and analyzing the interaction with different fuels using a reduced quantity. Data analysis is essential to monitor and to assure that the testing parameters remain stable. Post-analysis helps to understand the behavior and response of the tests and feeds the main database with identified problems. The experience acquired in each trial run is fundamental to categorize certain patterns in the time series as possible problems. Bibliography [1] Tec4Fuels, „Index,“ [Online]. Available: http: / / www.tec4fuels.com. [Accessed 04 01 2021]. [2] H. Hoffmann, „A Contribution to the Investigation of Internal Diesel Injector Deposits“, Aachen: Shaker Verlag, 2018. [3] D. Correa-Sanchez, Enhancements on a test bench built as a cost-efficient tool in the introduction of alternative fuels., Aachen: Fachhochschule Aachen, 2020. [4] H. Hoffmann and K. Lucka, “Hardware-in-the-loop testing: Complete Commonrail System (CoCoS) and component testing as rapid and cost-efficient tool in the introduction of alternative fuels in the automotive sector,” in 11. Tagung Einspritzung und Kraftstoffe 2018 , Springer Fachmedien Wiesbaden, 2019. [5] T. Górecki and M. Łuczak, “Using derivatives in time series classification,” Data Mining Knowledge Discovery, no. 26, pp. 310-331, 2013.