Vehicle calibration optimization using a dynamic test bed with real time vehicle simulation

Transcription

1 Vehicle calibration optimization using a dynamic test bed with real time vehicle simulation G. Burette 1, F. Perez 2, K. Bansal 2 1: D2T Powertrain Engineering Direction Ingénierie du GMP Technopôle du Madrillet 52 Av Galilée F-768 St Etienne du Rouvray Tel: : D2T Powertrain Engineering Direction Ingénierie et Equipements de Moyens d essais Z-A de Trappes-Elancourt 11 rue Denis Papin F-7819 Trappes Tel: Abstract: Nowadays, most engine calibration tests are carried out on a chassis dynamometer. For the initial calibration phase, these tests require extensive resources, such as complete prototype vehicles and associated test beds, while the phase itself focuses only on the powertrain. In order to lower the overall cost of engine development, D2T offers test bed solutions designed to reduce the number of chassis dynamometer tests thanks to real time simulation models included in the MORPHEE 2 automation system. Using a conventional dynamic test bed application (powertrain with no clutch and a simplified gearbox), D2T models simulate driver behaviour and actuation of the clutch, gearbox and final drive without the drawbacks and running costs of a chassis dynamometer. Indeed, it is easier to modify mechanical specifications chosen to run for the models and then to test other solutions by simulation. Moreover, the initial development phase (optimum settings on cycle, optimization of pollution settings in transient states, validation of DPF or NOxTrap supervisor, etc.) can be carried out in first on a dynamic engine test bed in order to reduce the number of tests on the chassis dynamometer. This presentation aims to describe this new approach using a comparative study between rolling tests on a chassis dynamometer and the same tests simulated on an engine test bed. The most restrictive tests being exhaust emissions standards approvals, the quality of the results will be evaluated using a comparison of pollutant emissions. Keywords : Simulation, Dynamic test bed 1. Introduction D2T currently offers a simulation tool in the MORPHEE 2 1 automation system that allows dynamic cycles to be carried out. We have assessed these capabilities in the initial phase of a calibration adjustment on a Euro 5 diesel engine. This phase makes high demands in terms of chassis dynamometer tests. All these tests provide a macroscopic assessment of the integration of adjustments resulting primarily from design of experiments. If it is possible to scan the same adjustment zones (injection and air loop) as during a test started in NEDC conditions, it is possible to evaluate and optimise the calibrations upstream of the phase relative to the availability of complete vehicles. The first part of this paper presents the SIMUDYN 2 simulation tool offered by D2T. The second part gives a comparative study between the tests carried out using SIMUDYN 2 and tests carried out on a chassis dynamometer, in order to quantify the discrepancies between the two test methods. Lastly, the article covers development of the same simulation tool for hybrid vehicle applications (HyHIL project). 2. SIMUDYN 2 The work described in this paper is based on the SIMUDYN 2 use on an engine test bed in order to represent the vehicle dynamics and the driver behaviour instead of executing NEDC test on chassis dynamometer test bed. SIMUDYN 2 is D2T application software based on MORPHEE 2 real time simulation execution capabilities. For our use case, SIMUDYN 2 provides the following features: Vehicle dynamics simulation for a vehicle equipped with a manual gearbox Driver simulation, management of the throttle, brake pedal, and gear shifting for the manual gear box Parameterization of the simulation The following figure (fig#1) explains the principle of SIMUDYN 2. 1 MORPHEE 2 is a test bed automation system developed by D2T SIA International Conference May 26 & 27, 21 1/8 Page

2 Fig #2 : The vehicle model fig#1 SIMUDYN 2 is based on three real time models: The first concerns the vehicle dynamics simulation The second represents the driver action on the throttle and the brake pedal during standard driving phases The last manages the gear shifting, the vehicle take-off of the driving cycle. Execution of standard NEDC test procedure Exhaust gas analysis based on raw gas measurements to identify mass and specific emissions During the NEDC test execution, SIMUDYN 2 takes the control of the engine test bed. This control is based on speed/throttle position standard control mode. It means that SIMUDYN 2 takes into account the engine dynamics to control the test bed during the gear shifting and free idle phases. This control is assumed by a real time controller developed in Matlab-Simulink which is integrated into SIMUDYN 2. This model is designed to be directly connected to the Dynamometer. In a vehicle there can be phases where the engine speed is determined directly by the vehicle velocity. This means that vehicle provides a set-point of Angular velocity which can be given to the Dynamometer. However, there are also the phases where the engine speed is determined by the equilibrium of the torque provided by the engine and the torque transmitted to the vehicle. In order to simulate these phases correctly, the equilibrium must be simulated as well. This gives rise to the need of a model which converts the Torque set-point coming from the vehicle to an angular velocity setpoint which can directly be given to the Dynamometer. The transition between the phases where the vehicle provides the angular velocity setpoint and the phases where it provides torque setpoint is instantaneous. Thus, the BenchControl model handles both the phases and switch between them according to the demand of the vehicle. The vehicle model simulates all parts of the vehicle including the Gearbox, Final Drive and wheels. This model gives a relation (direct or indirect) between the engine angular velocity and the vehicle velocity. This relation depends upon the type of gearbox used and the phase of operation. Care is taken to simulate the case of Gear =. Also, the change in direction of torque transmission (Engine to Wheels or Wheels to Engine) is taken into account. This model considers the shafts of the drive-train as rigid bodies. Thus, in this model, their vibrations are neglected. The information of torque goes from the clutch towards the wheels and the information of the speed flows from the wheels to the clutch. This means that the clutch decides the Torque transferred to the gearbox which converts this to the torque transferred to the Final Drive, which in turn decides the torque transmitted to the wheels. This Torque is used to calculate the velocity of the vehicle which gives the angular velocity of the wheels. This is translated back to the angular velocity downstream the clutch by the Final drive and Gearbox. In this case of vehicle model, the most important part to be simulated with a great precision is the clutch. The clutch is modelled as an elastic element with a damper in parallel. The model takes as an input the angular velocity upstream and downstream the clutch. Using these values, it computes the torque transmitted through the clutch. The maximum torque transmitted by this system is limited by the maximum friction force in between the clutch plates. The damper is modelled with a constant damping coefficient while the elastic element s stiffness is a function of the relative angular displacement between 2 plates as shown in Fig#3 k = f (Angle) D MaxTorque = f (Clutch) Fig#3 SIA International Conference May 26 & 27, 21 Page 2/8

3 The Driver model This model emulates the driver of the vehicle. It is based on two separate models which assume the following actions: The first model which is the Driver model manages the accelerator and brake position in such a way that the vehicle follows the velocity set-point within the required velocity margin. The second model which is the Gearboxmanagement model is dedicated to a vehicle equipped with a manual gearbox, manages the Gear-shifting, vehicle take-off and vehicle stopping processes. The driver model manages the accelerator pedal in a way similar to a real driver. That is to say: a minimal change in the accelerator position without sudden changes. Also, care is taken that brake and accelerator are not managed simultaneously. Thus, whenever the accelerator position is non-zero, the brake position is equal to zero and vice-versa. The driver takes into account the change in various driving conditions such as the slope of the road. This model does not manage the phases of starting and stopping of the vehicle and the gear-shifting. Thus, this model is independent of the type of gearbox present in the vehicle. This model has a bypass through which it outputs the brake and accelerator set-point given by the user. This model is based on standard PID principles and integrates the Road Load simulation to control the vehicle velocity. Concerning the Gearboxmanagement model it manages the Gear-shifting, vehicle starting and vehicle stopping processes. A vehicle start is triggered when all the conditions are conducive for it. During the vehicle starting phase, the clutch position must be managed by GearBoxManagement and 1st gear is selected. Starting of the vehicle is defined in such a way that no other decision concerning gearbox are made during this phase. This implies that all the Gear-shift decisions is postponed until the vehicle starting is complete. At low vehicle speeds, the clutch position is managed by disconnecting (partially or fully) the engine from the wheels. This clutch management based on a controller allows to drive a vehicle at very low speed without risking to stop the engine on the test bed. Care is taken for the strategy of gear-shifting. The strategy is robust with all the possible cases taken into account. Also during the gear-shifting phase, the clutch and accelerator position are managed. Once the gear shifting process is started, any change in parameters does not lead to incoherent state. This can be done by reading the parameters only at the beginning of the gear shift process. The gear-shifting phase is based on simulation parameters which can be adapted by the user to be compatible with every driver behaviour and vehicle characteristics. It is possible to utilize the model in two modes: Manual and Automated. In the Manual mode, the gear-shifting decision is manual, i.e. comes from the exterior of the model. This is corresponding to a manual gearbox where driver makes the decision of gear-shifting. In Automated mode, the decision of gear-shifting is made by the model. In Automated mode, Gear= cannot be reached. The Gear = is only possible in the manual mode. In addition to aforementioned functions, the model is also be able to manage external demand for clutch disengaging. An external demand for clutch disengaging can come due to various reasons, such as in order to stop vehicle. In some Road cycles such as NEDC, there are certain periods where clutch has to be disengaged in a vehicle with manual gearbox. During these phases, the model manages both the accelerator and clutch positions. Once the period of forced clutch disengagement is finished, clutch is engaged depending on the new conditions. 3. Setting up the road cycle test Following an NEDC test conducted on a chassis dynamometer, we configured SIMUDYN 2 on a dynamic test bed in order to recreate comparable and usable results. During this phase, we concentrated on assessing pollutant emissions without the post treatment system, so as to quantify each pollutant throughout the whole duration of the cycle, and to eliminate the difference in characteristics of DOC (Diesel Oxidation Catalyst). We therefore used a Euro 5 diesel vehicle configuration (called C1) for which the emissions are known, and are proved to be robust during a NEDC test. In the first instance, it is essential to provide the parameters that define the vehicle (coast-down, gear ratio, final drive ratio, wheel size, etc.). This phase ensures that the right zones are swept in terms of speed / indicated torque. Then, we have tuned the setting to define the driver. With regulator parameters (PID) and action time on the pedals, it is possible to modify the kind of the driving in order to create a driver who can be penalising or benefiting on pollutant emissions. The next figure shows two kinds of driving. Each case respects the tolerance of driving cycle but the second is worse for pollutant emissions than the first. SIA International Conference May 26 & 27, 21 Page 3/8

4 Vehicle speed (km/h) Vehicle speed (km/h) Vehicle speed SIMUDYN time (s) Vehicule speed maxi Vehicule speed mini VEHICULE SPEED Indicated torque Vehicle speed SIMUDYN torque (Nm) Torque (Nm) It is essential for the configuration of the cooling circuit to remain as close as possible to that of the vehicle, in other words with an operational thermostat box, a comparable volume of liquid and an operational heater loop. In addition to having the water circulating in the same circuits as on the vehicle, the heater loop also provides a means of accelerating the engine cooling. Since the thermostat is operational, there is no longer any flow in the main loop below the thermostat regulating temperature. Figure #6 shows the temperature of the coolant measured by the ECU during a test on a chassis dynamometer and a test on an engine test bed. The maximum temperature difference between the two tests is 6 C when the engine thermostat is opened. This difference has only a minor impact, since combustion strategies are generally established before this opening process. The average difference during the increase is in around 1 C time (s) Coolant temperature SIM2 vs Chassis dyno Vehicule speed maxi Vehicule speed mini VEHICULE SPEED Indicated torque Fig # For the following test campaign, we have fixed the second driver configuration. Figure #5 shows the engine speed and indicated torque detected by the ECU during the end of NEDC. The average difference between chassis dyno test and SIMUDYN test is less than 2Nm over the whole cycle T C 6 Indicated torque SIM2 vs Chassis dyno Chassis dyno ZREGIME Time (s) SIM2 R_EC.SPEED Chassis dyno ZTEASMOT SIM2 B_ZTEASMOT Fig #6-2 ES (rpm) time (s) Chassis dyno ZREGIME SIM2 R_EC.SPEED SIM2 B_ZCI Chassis dyno ZCI_DEM SIM2 B_ZDRI Fig #5 Since combustion settings are dependent upon the temperature of the engine fluids, it is important to use configuration methods that enable the creation of an identical rise in temperature between the two tests. It requires modifications for assembling the engine on a static test bed. If the engine sensors detect the same temperature increase profiles as in a real time test, it is ensured that the same operating points are swept TQI (Nm) The engine air intake temperature also affects the combustion strategies and has a strong influence on pollutant emissions. During tests under steady running conditions, this temperature is regulated using the results of a table based on engine speed and engine torque. This type of mapping does not take into account the geometry of the front face of the vehicle. In order to overcome this problem, we have added mapping for air temperature settings, based upon the vehicle speed and engine power. These settings result from measurement of the temperature at the exchanger outlet measured on a vehicle during an NEDC test. If the vehicle is not available, it is possible to use a simulation model to estimate the evolution of this temperature over the course of a cycle according to the engine architecture (Model developed thanks to a simulation tool such as AMESim). Figure #7 shows the mapping profile drawn up for the C1 configuration. SIA International Conference May 26 & 27, 21 Page 4/8

5 Air intake T C Engine power (kw) 4 Fig. # Vehicle speed (km/h) The air is cooled by injecting of cold water onto the air / air intercooler. The blowing of warm air onto the exchanger has been added in order to accelerate increases in temperature (during the EUDC). Figure #8 shows the air temperature at the outlet of the air / air intercooler measured by the ECU during a test on a chassis dynamometer and during a test on an engine test bed. The average difference is around 1,3 C, with the maximum difference being 5 C at the end of the EUDC. studied in subsequent tests (variable rate ventilation on the turbo face, for example). Figure #9 shows the exhaust gas temperature measured at the engine outlet Exhaust gas temperature SIM2 vs Chassis dyno Chassis dyno ZREGIME Time (s) SIM2 R_EC.SPEED Chassis dyno ZTECSMOT_B SIM2 B_ZTECSMOT_B Exhaust gas temperature SIM2 vs Chassis dyno 1st ECE Air inlet temperate SIM2 vs Chassis dyno Time (s) 5 5 Chassis dyno ZREGIME Chassis dyno ZTECSMOT_B SIM2 R_EC.SPEED SIM2 B_ZTECSMOT_B Time (s) Chassis dyno ZREGIME SIM2 R_EC.SPEED Chassis dyno ZTAIEMOT SIM2 B_ZTAIEMOT Fig #8 The oil is cooled by an oil / water cooler, if the increase in water temperature is adequate the oil behaves in the same way as it does in the vehicle. Exhaust gas temperature Without forced air ventilation in the test cell, the temperature of the exhaust gases follows a profile comparable to that in a vehicle. Since the configuration of our engine incorporates a turbo that is relatively confined, the fact that the engine is not in its vehicle environment has only a minor impact on the exhaust gases emitted from the engine. This factor can be more problematic if the exhaust line is more exposed on the vehicle, and may be T C Emissions results fig #9 One of the advantages of carrying out these tests on an engine test bed is having the possibility of repeating a higher number of tests than on a chassis dynamometer. By motoring the engine with the dynamometer and simultaneously forcing the cold water to circulate around it, it is possible to reduce the temperature of the fluids to produce the same initial conditions as a new NEDC. Following this method of forced cooling, the temperature increase curves are comparable with each other, and enable a high level of test repeatability. Graph #5 shows a mass estimate of each pollutant over three successive cycles. The first test is carried out after a night of soaking in the test cell, and the two subsequent tests are conducted after forced cooling. The fourth value is the same mass estimate for pollutants upstream of the post treatment system during a test on a chassis dynamometer. SIA International Conference May 26 & 27, 21 Page 5/8

6 12 NOx Specific [mg/km] CO Specific [mg/km] CO2 Specific [g/km] THC Specific [mg/km] 6 Coolant temperature SIM2 vs Chassis dyno SIM2 Cold start SIM 2 Restart #1 SIM 2 Restart #2 Chassi dyno Fig. # T C 2 It can be seen that the test results do not show large discrepancies. It is possible to realise 4 tests per 12 hours, as opposed to 2 tests on a chassis dynamometer per 24 hours. The variation of emissions between two tests methodologies can be explained by a little difference in engine stage between that used on the vehicle and that used on the engine test bed. However, emissions of CO, CO2 and HCs are comparable to those in tests on vehicles (difference of -.61% for CO, -.65% for CO2 and -11.8% for HCs). On the other hand, The NOx difference can be explained by the kind of driving. The torque requirements are greater during transients with SIMUDYN, Higher emission peaks of NOx can be noted during these transient phases. Given the excellent repeatability of tests with SIMUDYN 2, our goal of being able to compare different methods is reached. We did not linger to configure a more precise driver but evaluate the deviation with a static driver configuration Comparisons with a different vehicle configuration This second phase aims to quantify the differences in pollutant emissions from the same engine mounted on a different vehicle configuration. We have therefore used the emissions results for a cycle in a vehicle fitted with the same engine, but for which the gearbox and the final drive are different. The configuration (C2) of SIMUDYN 2 has been carried out with a 5-speed gearbox, superior final ratio, and more restrictive inertia and road load. In the same way as for the C1 configuration, the temperature table for the intercooler outlet has been adapted to this new type of vehicle. The engine test bed temperature increase profiles show the same consistency as for the C1 configuration (average coolant temperature difference of 1.1 C, average air temperature difference of 1.6 C, average difference of 2.4Nm in indicated torque)). Figure #11 shows the temperature increase curves for this new configuration. 1 ES (rpm) Chassis dyno ZREGIME SIM2 R_EC.SPEED Time Chassis dyno ZTEASMOT SIM2 B_ZTEASMOT (s) Indicated torque SIM2 vs Chassis dyno Chassis 7 dyno ZREGIME 8 9 SIM2 R_EC.SPEED 1 11 SIM2 B_ZCI 12 Chassis dyno ZCI_DEM time SIM2 (s) B_ZDRI Fig #11 The following bar graph (Fig #12) shows the pollutant emissions over three successive tests NOx Specific [mg/km] CO Specific [mg/km] CO2 Specific [g/km] THC Specific [mg/km] SIM2 Cold start SIM _2 SIM _3 Chassi dyno Fig #12 The same type of difference can be observed as during tests on the C1 configuration with respect to the results obtained with the chassis dynamometer (difference of -6.97% for CO, -.65% for CO2, for HC). The following table (fig #13) gives details of the relative differences for each pollutant between the two configurations tested. NOx Specific CO Specific CO2 Specific THC Specific DEVIATION [g/km] [g/km] [g/km] [g/km] C2 - C1, chassis dyno TQI (Nm) HC + Nox Specific [g/km] C2 - C1, SIMUDYN Fig #13 Intermediate conclusion SIA International Conference May 26 & 27, 21 Page 6/8

7 This study shows us that, with SIMUDYN 2, it is possible to draw up new methodologies to characterise calibrations on an engine test bed. This does not replace the chassis dynamometer test, but may be reduce the number of test on chassis dynamometer. Furthermore, the simulation included in this tool is open. Existing models can be replaced and new components can be integrated easily. It is therefore possible to simulate the components of a hybrid power train in order to observe their impact on rolling cycles. 4. HyHIL Project In parallel of this study some works are realized to simulate hybrid vehicle on engine test bed. These studies are executed through the HyHiL project in partnership with IFP, Renault, LMS-IMAGINE and G2ELAB. The project objectives are to develop software tools to simulate Hybrid vehicles on an engine test bed to reduce the investments in terms of time and costs for hybrid vehicle project development. This kind of test bed allows: To federate around the engine test bed the several competences required during a hybrid vehicle development To reduce the risks of the project evaluating and confirming the technical choices before their development To anticipate the control strategies developments testing then directly with the real thermal engine To reduce the tests performed with a real vehicle and the development of physical prototypes. For this project we developed real time models to simulate the hybrid vehicle dynamic, and the control strategies into a simulated hybrid Energy Management System. High dynamic models of the hybrid vehicle were developed taking into account the oscillation along the powertrain and integrating and adding 2D vehicle to take pitch and changing load on tires. The complete simulation was integrated into MORPHEE 2 and required several model execution frequencies up to 2 khz for the vehicle model. To be able to execute all models on the engine test bed we developed a prototype of the next MORPHEE 2 automation system version, which is able to distribute the real time models execution on multiprocessor platforms. The following figure (fig #14) explains the principle of the HyHiL test bed. Fig #14 The use of all of these developments makes possible the characterization of fuel consumption, pollutant emissions and also to test the engine control drivability strategies directly on the engine test bed with a great precision. One of the most interesting features of this application is to make it possible to compare several levels of hybrid powertrain applied on the tested engine. The following figure (fig #15) shows results obtained on the test bed for three different architectures for the same vehicle with the same thermal engine: full thermal engine configuration, micro-hybrid configuration, and full-hybrid configuration. Thermal Engine Micro-hybrid Full-Hybrid Vehicle Fuel cons. Vehicle Fuel cons. Vehicle Fuel cons. mass [kg] [L/1km] mass [kg] [L/1km] Gain mass [kg] [L/1km] Gain NEDC % % FTP % % 5. Conclusion Fig #15 This paper shows that results obtained with SIMUDYN 2 are appropriate to lead new methodologies in ECU calibration process reducing the number of tests on chassis dynamometers. Moreover, this approach increases the independence of the calibration process from the availability of vehicle prototypes. These tools are only software add-on for MORPHEE 2, that s why it is not required to install any additional hardware on the test beds. Until now, only a few test beds in a test center had the capability of executing real time simulation. With MORPHEE 2, every test bed is able to integrate real time simulation, not only for vehicle dynamics simulation, but also to simulate for example the complete communication between the ECU and the vehicle, to simulate post-treatment equipments, etc. SIA International Conference May 26 & 27, 21 Page 7/8

8 One illustration of these new capabilities of engine test bed based on real time simulation is the HyHiL project. It demonstrates this for hybrid vehicle development project. This shows that the importance of the engine test beds in the vehicle development projects with keep on increasing even with the on-going paradigm shift in the vehicle technology. 6. Acknowledgement The authors would like to thank PSA for the technical support provided in order to be able to input in the models the most accurate data concerning the vehicles and the powertrain design. We want also to thank the Conseil Général des Yvelines, the Direction Générale des Equipements and Mov eo who make possible the HyHiL project. 7. References A. Del Mastro, M. Castagné, A. Chasse, P. Moulin : "Drive Cycle Reproduction on High Dynamic Test Bed in order to perform Transient Calibration" A. Del Mastro, A. Chasse, P. Pognant-Gros, G. Corde, F. Perez : Advanced Hybrid Vehicle Simulation : from Virtual to HyHIL test bench SIA International Conference May 26 & 27, 21 Page 8/8