A Novel Method of Data Synchronization during Transient Engine Testing for ECU Development

Speakers Information- Controls, Measurement & Calibration Congress A Novel Method of Data Synchronization during Transient Engine Testing for ECU Development Jensen Samuel J, Paul Pramod S, Ramesh A IIT Madras, Chennai Anand Mammen Thomas Research& Innovation Centre, DRDO, Chennai Jishnu Gopal S Thiagarajar College of Engineering, Madurai ABSTRACT In any modern Diesel engine development, data acquisition during transient operation is a challenging task. The acquired data can range from engine position synchronous parameters such as in-cylinder, intake manifold, exhaust manifold and fuel rail pressures, fuel injection timings and duration, to time based data such as temperatures, fuel flow rate, turbo speed and VGT position feedback. In addition, data acquired through on-board diagnostics will also be of interest. These data are generally acquired at different rates and resolutions on different hardware platforms. While postprocessing the acquired data to analyze engine performance during transients such as cold starting, warm up and acceleration it becomes vital to synchronize the data against a common time base. This paper discusses a novel method and a Matlab GUI that was used to accomplish this task of synchronizing data from three sources namely Kistler KiBox, ETAS ES650 and ETAS ES592 (OBD). In order to synchronize all the signals a square pulse generated during cranking was recorded on all the platforms. A program with GUI was also written in Matlab to post process the data sets from each platform. This software reduces the data to the needed resolution and plots all the signals against a common time base even though the signals were obtained at different rates and in different forms (analog / digital). Experiments were conducted on a 2.2L state-of-the-art common rail, turbocharged, intercooled engine with EGR and VGT under cold start conditions with its stock ECU. The acquired data has been successfully used to arrive at control strategies for starting and idle speed governing of the same engine using an ETAS Flex-ECU open engine controller. INTRODUCTION Modern direct injection diesel engines make use of sophisticated electronic controllers for managing parameters like fuel injection timing and duration, fuel injection pressure, intake air boost pressure and charge-air oxygen concentration. Any experimental study of a modern diesel engine under transient operating conditions like cold starting, warm up, acceleration, etc., involves acquiring and synchronizing data from different sensors and instruments which work at different sampling rates and resolutions [1]. During the development of new control strategies for transient operation, crank angle synchronous data like in-cylinder pressure, manifold pressures, fuel injection timings and duration are to be captured and analyzed [2]. Along with this, data acquired at fixed time rates like temperatures, flow rates and position feedback from different valves are required to be correlated to understand the effect of control inputs on the engine performance variables. Control inputs like fuel injection timings and duration, frequency and duty cycle of pulse width modulated (PWM) inputs to actuators for Inlet Fuel Metering Valve (IMV), Variable Geometry Turbocharger (VGT) and Exhaust Gas Recirculation (EGR) dictate how the engine behaves during these transients and hence their variation with time has to be captured. The influences of all the time varying input parameters generated by the controller is understood by studying the measured values of performance variables like specific fuel consumption, engine speed, rail pressure, boost pressure, EGR mass flow rates, emission levels and exhaust gas temperature. Acquiring slow variables at high sampling frequencies only increases the amount of data to be processed. Similarly not acquiring fast variables like cylinder pressure, turbo speed, position of valves during transients at high frequencies will lead to loss of valuable information. Further measurements done by sampling important variables at definite crank angle intervals need to be

transformed to an appropriate time base as crank position is not directly linked to time and depends on the instantaneous speed as well. Available data acquisition systems do not allow large number of such varied types of signals to be acquired at high and low speeds on the time basis and also simultaneously capture certain other variables on the crank angle basis and analyse them on the same time base. Hence, researchers are often confronted with the problem of using different data acquisition systems simultaneously and then synchronizing the signals for representation on a common time base. The simplest way of synchronizing data from these platforms is to match the corresponding time stamps available with the signals. However, the available time stamp from each of the platforms might have offsets due to differences in the device s clock time settings and differences between the times at which recording of data from different systems is initiated. [3] Since huge volumes of data are handled and the process including data acquisition has to automated, special techniques to synchronize the data acquired on different platforms are to be combined with suitably developed post processing software. The present work is in this direction wherein data acquired from two platforms: a) On the crank angle basis on Kistler Ki box and b) On the time basis from ETAS data acquisition modules have been synchronized and post processed to present results on a common time base during starting of a common rail diesel engine. The processed data has been successfully used to arrive at control strategies for starting and idle speed governing of the reference engine using an ETAS Flex-ECU open engine controller. This technique and software can also be employed to process such data at other transient and steady operating conditions. EXPERIMENTAL SETUP Experiments were conducted on a Mahindra mhawk 2.2L state-of-the-art common rail, turbocharged, inter-cooled engine with EGR and VGT under cold start conditions with an aim to develop and demonstrate new control strategies. The test engine was coupled to an AVL alpha 160 eddy current dynamometer, as shown in Fig. 1. Fuel flow rate was measured using an AVL fuel flow meter working on the Coriolis principle and throttle position was set using an in-house developed throttle actuator. INSTRUMENTATION: Fig.1 Experimental Test Setup An uncooled piezoelectric pressure sensor was flush mounted on to combustion chamber of cylinder 1. A piezoresistive pressure sensor was mounted on to the intake manifold of cylinder 1 for referencing the in-cylinder pressure, while a water-cooled piezo-resistive pressure sensor was mounted on the exhaust manifold. An optical crank angle

encoder, as shown in Fig. 2, was fitted on to the crank shaft pulley to track the engine s position. The turbocharger was instrumented with an eddy principle based speed sensor and the VGT position was tracked using an in-house developed sensor working on potentiometric principle. In-house developed current sensors (Fig. 3) working on Hall-effect principle were instrumented on the engine s wiring harness at different locations to sense the magnitude and timing of the current signals given to various actuators such as Fuel injectors, Variable Geometry Turbine, Exhaust Gas Recirculation and Fuel Inlet Metering Valve solenoids. DATA ACQUISITION Fig. 2 Crank Angle Encoder Fig. 3 In-house developed Current Sensor Combustion related crank angle synchronous data like in-cylinder pressure, manifold pressures, high resolution engine speed, common rail fuel pressure and fuel injection timings and duration were acquired using Kistler s KiBox combustion analyzer which has a maximum sampling rate of 1.5MHz per channel. [4] Fuel Injector Cylinder Pressure Sensor Crank Angle Encoder Current Clamp Rail Pressure Sensor Intake Manifold Pressure Sensor Exhaust Pressure Sensor Inlet Metering Valve Turbo Speed Sensor Amplifier VGT Actuator Amplifier Reference Engine Current Sensor Current Sensor Charge Amplifier Combustion Analyzer Current Sensor Fig.4 High Speed Data Acquisition EGR Actuator

These signals have been processed by the respective amplifiers and are given as analogue voltage inputs to the Kibox. In addition to the combustion related signals, analogue signals like VGT position, rail pressure and PWM inputs given to the EGR and VGT actuators were also acquired on Kistler KiBox. The sampling rate in time with Kistler Kibox will be a variable as the speed of the engine is never constant. Further, digital inputs like turbo speed and a specially generated pulse for synchronization were also taken into KiBox through its digital input channels. The layout of these signals connected to Kistler Kibox is seen in Fig.4. Slow analog signals like fuel flow rate and temperatures at different locations in combustion air path were acquired using ETAS ES650 module which has a maximum sampling rate of 2000Hz. In addition, data from the On-Board Diagnostics (OBD) module were also acquired using ETAS ES592 module through the CAN input, as shown in the schematic in Fig. 5. OBD Data Ambient air Temp. Air Mass Flow Rate Boost Pressure Rail Pressure Barometric Pressure Engine Speed Load Coolant Temp ETAS ES592 ETAS ES650 Fuel Flow Meter ECU Thermocouples Compressor Outlet Intercooler Outlet Intake Manifold Exhaust Manifold Turbo Outlet Coolant Inlet Coolant Outlet Oil Sump ETAS INCA Fig.5 Low speed data acquired on ETAS ES592 and ES650 modules TECHNIQUE FOR SYNCHRONIZATION OF DATA FROM DIFFERENT MODULES Since the data is acquired on two platforms namely Kistler Kibox and ETAS data acquisition modules the time has to be matched. Since the time at which each platform starts measuring is unknown, synchronization has to be done during post processing. However, a common signal that is acquired on both the platforms can be used for referencing the time. This technique was adopted here. The instant the supply is given to the starter motor is taken as τ 0. This was done by taking the input to the starter motor solenoid and generating a pulse and feeding the same to both the platforms. For this an opto-isolator based pulse generator was developed as shown in Fig.6. The typical output of the pulse generator during engine cranking is shown in Fig. 7. When the starter motor is off, the photo-transistor in the opto-isolator does not conduct and thus the output voltage goes to nearly 5V.The moment the starting key is turned ON to crank the engine, the photo-transistor within the opto-isolator conducts and the output falls a very low value. This could be observed from the sudden fall in voltage shown in Fig. 7. The moment the starting key is released to turn OFF the starter motor, the output voltage increases as shown in Fig.7. However, releasing the starter key in the direction of its spring forces creates a bouncing effect and hence the voltage fluctuates about this point. The inductance of the starter motor solenoid produces the observed gradual change in voltage with time seen in Fig.7. The

start of cranking of the starter motor shown by the initial sharp dip in the voltage is taken as time τ 0. The developed pulse generator is also shown in Fig. 8. R 1 + - M Starter Motor R 2 R 3 Fig. 6 Pulse Generator Circuit Synchronization of the time based on the engine cranking signal will enable determination of some important parameters like time taken for ECU synchronization with engine position, engine cranking duration, rail pressure fluctuations at engine start, engine speed and acceleration during starting, heat release rate of the first firing cycle, VGT position during starting, injection timings and durations, number of fuel injection pulses at start, turbocharger speed and acceleration, fuel flow rate and engine idling speed. The data helps in evaluating how different actuator control signals and engine parameters vary during transients. Start of Cranking End of Cranking Key Bounce Solenoid Inductance Fig. 7 Pulse Generated during engine Cranking Fig. 8 Pulse Generator Setup

DATA POST-PROCESSING The pulse generated during starting, as described in the earlier section, was recorded on all the platforms as discussed and used to offset the time stamps and to bring signals to a common time base. To this end, a software module with Graphical User Interface (GUI) was developed in Matlab platform as shown in Fig. 9 as a part of this work. This software uses the Matlab Integration Package (MIP) given by Kistler to access the data recorded on KiBox, within the Matlab environment. Similarly the data from the ETAS platform is also imported in ASCII format. The data from Kistler Kibox is on the crank angle basis (0.1 degree intervals for cylinder pressure and 1 degree intervals for intake and exhaust pressure, PWM signals of the VGT, EGR and IMV actuators). This means that the data is at non-uniform time and crank angle intervals, however with time stamps. The GUI seen in Fig.9 indicates the different data tracks (Select Data Tracks block) and allows the user to choose the ones to be processed and plotted. The start and end times for processing and plotting can also be specified by the user after which the software computes based on the signals from Kibox the number of cycles of engine data that will be correspondingly processed. It may be noted that some channels, due to restrictions in the data acquisition platforms may be sampled at higher than required sampling rates. The data from such channels has to be suitably trimmed to get it down to the desired sampling interval so that high volumes of data are not handled unnecessarily. The Excel File Creation block reads the size of the data (number of data points) of any selected channel and allows the user to select a scaling factor for data reduction of that channel. It also suggests the minimum scaling required for exporting the data to excel. The software resorts to interpolation when needed. The software identifies the start of the common pulse (from the opto-isolator) recorded on each platform and the corresponding time stamp for all other channels acquired on that platform is then offset to time τ 0. The processed data is presented in a graphical form by software which also could be exported to other tools for further analysis and presentation. There is a separate Plot block as seen in Fig.9 to preview the data that was processed. Figure 10 shows a typical raw data acquired at different sampling rates and the processed data output by the software. On the left in Fig.10 we see the data acquired through Kistler Kibox on the crank angle basis. It is also seen that certain data like cumulative heat release and engine speed are at a different crank angle resolution as compared to cylinder pressure. Similarly data from ETAS which were obtained on the time basis also had different resolutions. The output of the software, after all these data were post processed by the developed software, is shown in Fig.10 on the right. We see that each data set (parameter) has its own time stamp with which it can be plotted as shown later in Figs. 11 and 12.. Fig. 9 In-house developed post-processing software tool

Fig. 10 Data at different sampling rates and common time-stamp (right); Data after post-processing (left) RESULTS AND DISCUSSION Figures 11 & 12 show the results obtained after different channels were synchronized using the developed technique and software. The data represented was obtained during first 2 seconds of cold starting of the reference common rail diesel engine. Figure 11 shows the key signal (inverted) obtained from ETAS along with the cylinder pressure and engine speed obtained from Kistler Kibox. It may be noted that t=0 is at the key signal s rising edge. The variation in cylinder pressure during starting along with the corresponding speed is seen. The first cycle that fired had peak pressure of about 120 bar which led to acceleration of the crank shaft. Once the speed crosses a threshold value the idle speed controller reduced the main injection duration, and its effect is seen in the lowered cylinder peak pressure. The engine s initial acceleration increased the engine speed to about 1000 rpm and the speed was immediately reduced to 800 rpm for idling as seen in Fig. 11. Such events have been clearly captured on a common time base with this software. The first firing occurred after about 1s as seen in Fig. 11 which is influenced by the time taken by the engine controller to synchronize with the crank shaft position. It is evident that the engine was cranked for about 1.5s, about 0.5s after the first firing took place as seen in Fig. 11. Figure 12 also indicates the signals obtained from the OBD, ETAS and Kistler Kibox plotted on the same time base. Though the set rail pressure for idling was 250 bar the data indicates that the actual rail pressure shot up to 600 bar before it settled to reach about 250 bar during idling as seen in Fig. 12 and from engine idling trends. The fuel consumption rate measured by the Coriolis force meter indicates its correspondence with the variation in rail pressure due to the influence of the rail pressure controller. We see that when the rail pressure shoots up the fuel consumption starts falling because the inlet metering valve is closed by the controller. The steps in the fuel consumption rate plot are due to the low frequency (20Hz) at which the coriolis meter provides its output. The air flow rate follows the speed of the engine. The rail pressure drops at the end of each injection and increases for each pumping action which is also evident from the number of pumping actions per engine cycle. Similarly other data like injection current profile, other PWM outputs, boost and manifold pressures, temperatures and turbo speed were observed and plotted as a function of time. Thus the developed software could post process data and provide information in a form that can help analyze the influence of different parameters.

Fig. 11Key Signal, Speed and Cylinder Pressure for the first 2 Seconds Fig. 12 Fuel rail pressure, fuel flow rate and air flow rate for the first 2 Seconds

CONCLUSION A useful technique and a software tool were developed to synchronize the data acquired from different hardware platforms sampling at different rates and resolutions. This software employing the developed technique has been successfully demonstrated for studying the starting transients of the reference engine. This technique and software can also be employed to process such data at other transient and steady operating conditions. This software is thus a very valuable tool for analysis of several complex engine variables obtained from different platforms for use in engine research and development environments. ACKNOWLEDGMENTS We thank Defence Research and Development Organisation (DRDO) India, for funding this research work. We also thank Mahindra & Mahindra for their technical support in carrying out this work. REFERENCES [1]-Rakopoulos, Diesel engine transient operation, textbook. [2]- LinoGuzella and Christopher Onder, Introduction to modeling and control of internal combustion engines, textbook [3]-David Ashlock, Synchronizing Controller Area Network and Analog Input Measurements for In-Vehicle Data Logging, National Instruments Application Notes. [4]-Kistler Kibox Manual. CONTACT Jensen Samuel J Department of Mechanical Engineering, IIT Madras, Chennai 600036. Email: jensenmech@yahoo.co.in ABBREVIATIONS ECU Electronic Control Unit EGR Exhaust Gas Recirculation GUI Graphical User Interface IMV Inlet Metering Valve OBD On-Board Diagnostics VGT Variable Geometry Turbine