The increasing number of passenger car variants and derivatives in all global markets with regionally specific legislative requirements (e.g. Real Driving Emissions in Europe), different market specific driver behaviors and customer expectations together with the vehicle manufacturers’ efforts to achieve global market and brand specific drivability characteristics as unique selling propositions, require an objectified evaluation and classification approach for the drivability capabilities of modern cars. The utilization of systems, which allow the customer to adjust the product-specific drivability capabilities additionally in a multistep way by different drive modes, is a major driver in this context. To streamline the vehicle development process, and to reduce manpower and prototype hardware resources an objectified drivability evaluation and classification approach, which is based on physical criteria, is developed at the RWTH Aachen University in cooperation with FEV.
For the demonstration of the reliability and maturity of this approach, results of sensitivity tests, which are based on ECU calibration changes with regard to the longitudinal vehicle drivability behavior and carried out with a real vehicle on a proving ground, are examined. Additionally, by reproducing this automated drivability measurement program with the same internal combustion engine (ICE) on a modern and highly dynamic Engine-in-the-Loop (EiL) test bench, the transferability of the holistic method is demonstrated. Hence, as introduction, the derived physical based criteria for the objectification of positive longitudinal drivability load change maneuvers are presented.
For the evaluation of positive load change maneuvers, seven different measures have been developed to describe the recognized quality of a tip-in maneuver. These attributes are shown in Figure 1 and 2.
Each of the seven drivability attributes, which are describing the different physical influences on the human body during a tip-in maneuver, have been defined and specified during various workshops with well experienced drivability calibration engineers for gasoline, diesel and hybrid powered vehicles. Furthermore, results and statements from the relevant international literature on the subject of longitudinal vehicle drivability have been taken into account during the precision process of the attributes.
Drivability results on the road and at the engine in the loop test bench
For the analysis of the transferability of engine related longitudinal drivability development and evaluation tasks from the real prototype vehicle to the EiL test bench, both, the vehicle and the EiL setup, were equipped with the same ICE and ECU version (incl. the same market related data set version). At the EiL test bench, the whole vehicle, with the exception of the real ICE, was simulated on the real-time co-simulation platform.
Initially, three different ECU longitudinal drivability calibration data sets, one sportive, one comfortable and one in between (medium), were calibrated in the real prototype vehicle (Veh) on the test track. Subsequently, those three data sets were applied to the ECU at the EiL test bench (TB). For both test scenarios and each of the three ECU data sets, the same engine load change maneuvers have been performed automatically by the use of FEV’s “TOPEXPERT Vehicle Test Automator” (VTA).
For the analysis of the comparability between both setups, first of all, the intake manifold pressure (IMP) build-up curves for all six different test scenarios are illustrated in Figure 3. All intake manifold pressure build-up curves of the EiL tests and the tests with the prototype vehicle show a very similar progression for each of the three different drivability calibration data sets. Furthermore, the differences between the various calibration data sets are quite obvious for both test scenarios. However, for all ECU calibration data sets the boost pressure build-up seems to be slightly slower at the EiL test bench compared to the real vehicle.
The boost pressure build up at the EiL test bench is slower, because the corresponding ECU is missing some information. Firstly, the alternator is running unloaded at the test bench, and does not deliver any electrical power. Therefore, no additional torque is requested from the ICE (by the ECU). In the real vehicle, even though all electric loads are switched off, the alternator requires about 3 Nm of torque from the ICE to supply electricity to the board net of the real vehicle. Further, torque losses in the transmission and the other drivetrain components are indeed simulated in the real-time capable models at the EiL test bench, but the data is not communicated to the ECU, since they are generally compensated by the driver model during the execution of emission cycles. The ECU in the real vehicle receives the required information, and it tries to compensate these losses by requesting more boost pressure to deliver the same drive torque as would occur without these loses. Additionally, due to safety reasons, the ambient pressure inside the test cell has to be slightly below the environment pressure level (approx. minus 40 mbar). Nevertheless, the curves for the three calibrations data sets show the same tendencies for both setups and they are consistent to each other. To further analyze the reaction of the ICE at the EiL test bench and in the test vehicle, the course of the throttle valve position (TP) of the sportive ECU drivability calibration is compared to the course of the comfortable one in Figure 4.
For the sportive, as well as the comfortable drivability calibration, the reaction of the engine on the test bench is in good agreement with the reaction of the engine in the test vehicle. Even the small hook in the course of the throttle valve position in the comfortable calibration can be observed in both test setups. In principal, the courses of the throttle valve position for both drivability calibration data sets are significantly different. With the sportive drivability calibration, the throttle valve opens almost immediately (0.2 s after tip-in start), while in the comfortable calibration the throttle valve is used to shape and slow down the drive torque build-up and thus to reduce the build-up of the vehicle longitudinal acceleration. Further, the differences between the two ECU calibrations are clearly visible in the longitudinal acceleration signals for both test scenarios. The build-up of the longitudinal acceleration for the sportive ECU calibration is distinctly faster than for the comfortable ECU calibration, while the comfortable ECU calibration produces smaller load change oscillations (less backlash and surging).
In Figure 5, the longitudinal acceleration measured in the real vehicle is compared to the simulated longitudinal acceleration of the virtual vehicle at the EiL test bench for the sportive and comfortable ECU calibrations. For both test scenarios, the differences between the two ECU calibrations is clearly visible, but there are a few differences between the simulated and the measured acceleration values. As an initial load change reaction, the acceleration of the virtual vehicle at the EiL test bench is slightly faster because any of the mechanical play in the drive train components is not simulated with the models. As the drivability maneuver progresses, the simulated longitudinal acceleration of the virtual vehicle rises slower than the corresponding acceleration of the real vehicle. The reason is the already mentioned slower boost pressure build up at the EiL test bench. Due to the simulation of the elasticities of the drivetrain components, similar oscillations can be observed in the longitudinal acceleration signals for both test scenarios. Considering the clear differences in the longitudinal acceleration signals between both ECU calibrations at the EiL test bench, it is expected that the objective and physically criteria are good measures to analyze the longitudinal drivability in virtual and real test scenarios. For the analysis with regard to reproducibility, 50 of the same positive load change maneuvers (2nd gear, 2,500 RPM, 0 percent to 50 percent pedal) have been performed for each ECU calibration in the two test scenarios. For all these measurements, the drivability attributes were calculated. In Figure 6 and Figure 7, the attribute “jolt” is compared to the “time to reaction”, and the attribute “backlash” is compared to the “duration of acceleration build-up phase”. The influence of the different longitudinal drivability calibration data sets on the results of the EiL test bench are shown in Figure 5, and subsequently the data of the corresponding vehicle tests are presented in Figure 7.
As expected, it becomes clear for both test scenarios that the sportive ECU drivability calibration data set leads to an increase of the attribute “jolt”. Also, the “duration of acceleration build-up phase” is reduced by the sportive ECU calibration. The “time to reaction” is not affected by the prepared different ECU drivability calibration data sets, because the comfortable ECU calibration does not delay the initial reaction of the ICE to changes in the accelerator pedal position (it only shapes the torque build-up). The “backlash” of a comfortable ECU drivability calibration is supposed to be lower compared to a sporty one. However, the ECU drivability functionalities of the chosen vehicle were not able to depict this correlation for the attribute “backlash”. Nevertheless, this behavior is determined in both test scenarios. The sensitivity of the ICE towards changes in the ECU drivability calibration at the EiL test bench is nearly identical to the sensitivity of the ICE in the real vehicle. The attribute “time to reaction” is shorter at the EiL test bench due to the missing play in the virtual powertrain components in comparison to the real vehicle. Overall, the influences of the different ECU drivability calibration data sets on the longitudinal drivability attributes are clearly visible in the same way and on the same level for the EiL test bench and for the real vehicle.
As a major advantage for the EiL test bench application, it can be stated that the reproducibility of the individual load change drive maneuvers is distinctly better than in real vehicle tests where there is typically a larger dispersion of the drivability attributes (Figure 5 and Figure 6). The reason is that various environmental influences the real vehicle is exposed to, such as wind, road conditions and temperature changes, which are not present, if not explicitly required and simulated, at the EiL test bench.
The basic comparability between the operation of a real vehicle and the automated drivability testing and evaluation approach via an EiL test bench could be demonstrated with an analysis of relevant engine operating data (throttle valve position, intake manifold pressure, engine speed etc.) for the same load change maneuvers. Also, different ECU drivability calibrations regarding e.g. a very sporty or a more comfortable throttle response can be reproduced very well with the EiL method in comparison to the real vehicle test. Exactly the same statement can be made based on the results of the physical based and objectified drivability evaluation approach for the sensitivity of the reaction of the internal combustion engine to changes in drivability calibration at the EiL test in comparison to the real vehicle. As a major advantage for the EiL test bench application, it can be stated that the reproducibility of the individual load change drive maneuvers is distinctly better than in real vehicle tests, where there is typically a larger dispersion of the drivability attribute results. Thus, the reliability, maturity, validity and transferability of the holistic method to different fully, partially or neither virtual testing scenarios is demonstrated.