DEVELOPMENT Real Driving Emissions Testing with Virtual Prototype Vehicles on the Test Bench IPG Automotive AUTHORS Dipl.-Ing. Christian Lensch-Franzen is Head of Engineering at APL Automobil-Prüftechnik Landau GmbH in Landau (Germany). Dipl.-Ing. Michael Friedmann is Project Engineer in the Basic Development/Research Team in the Engineering Department at APL Automobil-Prüftechnik Landau GmbH in Landau (Germany). Dr.-Ing. Christian Donn is Manager Business Development at IPG Automotive in Karlsruhe (Germany). Dipl.-Ing. Christian Rohrpasser is Engineer for Test Systems and Engineering at IPG Automotive in Karlsruhe (Germany). 36
With virtual vehicle prototypes and the integration of real driving simulation on es, new paths can be explored in propulsion system development. In cooperation with IPG Automotive, APL widens the scope of approaches to the development of RDE-optimised propulsion systems. REPRODUCIBLE POWERTRAIN DEVELOPMENT THROUGH VIRTUALISATION Pollutant emission and fuel consumption of vehicles under real on-road driving conditions are increasingly becoming the focus of public and legislative attention. The challenge lies in the development of robust vehicle propulsion systems which ensure adherence to emission limit values and the manufacturers specifications throughout the vehicles lifetime, in addition to the requested driving dynamics, drivability and durability. A variety of influencing factors and growing system requirements necessitate advanced methods for propulsion system development in order to facilitate the robust design and calibration of the entire system across the complete range of use in terms of emissions, FIGURE 1 [1, 2]. The quantification of the influence of individual measures or changes to the propulsion system plays a key role in the development process for example internal engine modifications, electrification or exhaust gas aftertreatment systems on the hardware level, as well as calibration variants with regard to a parameterisation and operating strategy that is optimal for RDE on the software level. The evaluation and specific development of technically optimal measures for different vehicle concepts require representative real driving conditions in addition to the reproducibility and robustness of tests. Due to a range of influences that are external and dependent on the process, there is no such reproducibility in on-road testing. The emission comparison of one single module of the same, five times driven RDE cycle shows despite subjectively comparable driving style and similar traffic conditions a variance of 11 % in the cumulated normalised particle emission. Tests on the test bench, in contrast, have proved to achieve a high reproducibility as well as a high level of automation, which is shown in the smaller variance of 4 %, FIGURE 2. Therefore, the objective is to transfer representative real driving flexibly and precisely to es, thereby enabling system development in a reproducible environment. DEVELOPMENT PROCESS METHODOLOGY On the es of APL, real driving is transferred into the test environment in a way that is test-specific and optimally adapted to the requirements. Analogous to the test matrix in FIGURE 3, there is a distinction between the type of and the level of complexity of the subsystems to be tested. For this purpose, a multidimensional testing matrix is generated which comprises either purely simulative studies in the office, component es, FIGURE 1 Use of simulation along the development process chain under RDE boundary conditions ( APL) ATZ worldwide 10 2017 37
DE VELO PMENT Real Driv ing Emissions APL Trackkit Complete RDE cycle Single module Vehicle speed road [km/h] 150 100 50 0 Particle number road [%] 100 80 60 40 20 0 0 10 20 30 40 50 Distance [%] 60 70 Particle number cumulated [%] 100 % 80 90 11 % 100 4% 80 % 60 % 40 % 96.2 % 100.0 % Trip 15 Fahrt Road 1 Fahrt 2 Trip 25 Road 91.3 % 89.4 % 91.3 % 93.6 % 91,1 % 20 % 0% Trip 35 Fahrt Road 3 Trip 45 Fahrt Road 4 Trip 55 Fahrt Road 5 Mittelwert Straße AV Road Mittelwert AV EnginePrüfstand FIGURE 2 Exemplary comparison of the reproducibility of measurements taken on the road and engine based on the particle number ( APL) endurance testing of systems, complex engine es, holistic powertrain testing or the examination of the whole vehicle on the chassis dynamometer or on the real road. In the field of endurance testing which aims at ensuring the operational stability using adequate means, the target course and load spectrum can be specified as rotational speedaccelerator pedal value sequences in the simplest case. In the case of a specified vehicle speed curve, at least a simple longitudinal dynamics model needs to be used and parameterised correspondingly. For the reproducible modelling of driving cycles that are less dynamic and required by law, such as the NEDC, this may be an effective and cost-efficient approach. The variants described above, however, have the disadvantage that essential aspects which influence emissions 38 such as traffic flow or driver behaviour/ style are inevitably integrated in the specified speed profile and cannot be separated according to their share in the component load or emission spectra after being recorded in the vehicle. For the robust, flexible and precise transfer of real on-road driving to es and for the variation of parameters and influencing factors, the implementation of free real driving simulation as offered by IPG Automotive s CarMaker is indispens able. Here, a tuneable driver model and different levels of detailing of traffic simulation with reproducible stochastic traffic events can be used as described in the following section. Due to the seamless use of the simulation environment in the office and on engine, powertrain and chassis dynamometers, the same models can be used for example for driver characterisation and the parame- terisation of driving robots across platforms as well. APL adopts this approach in the field of complex function testing and calibration validation with a focus on system robustness, the comprehension of mechanisms leading to emission genesis, the study of stochastic pheno mena and a target-oriented parameter variation. Good modelling even allows developers adopting the approach of model-based testing to shift parameter variation to fields which are interesting for calibration purposes but for which there is no exact information from real measurement data. The approach of real driving simulation is taken in the development stages illustrated in FIGURE 4, from the conceptual phase to SOP. This ensures that the influences of vehicle properties, driver, traffic, environment boundary conditions and the route are considered
throughout the entire development process for the design of the hardware as well as the development of the operating strategy and calibration. Once the process chain has been completed for one product, the insights gained serve for the concept evaluation and estimation of the performance of derivatives. Derivatives of already certified powertrains can thus be tested for system robustness and critical ranges can be identified early on in the development process. REAL DRIVING SIMULATION AND ITS USE ON TEST BENCHES The real driving simulation environment contains real-time capable models which allows for the realistic and precise modelling of a diverse range of vehicle types including their handling characteristics, the driver behaviour, the traffic situation and the road and its surroundings in the virtual world. With the integration on es, virtual test driving enables the generation of a flexible, semi-virtual RDE development environment, FIGURE 5, and thus offers a considerable potential for an increase in efficiency in propulsion system development. The foundation of the method for semi-virtual RDE tests presented here is the generation of virtual test tracks which are built from real routes based on measurement or map data and which also contain traffic lights and traffic signs that are relevant for vehicle speed in addition to curves and elevation profiles. With a vehicle model (virtual prototype) adapted to the real driving resistances based on the component data or vehicle coast-down curves and, if required, validated using real driving measurement data, virtual test driving is subsequently performed. The intelligent driver model which enables the reproducible representation of different driver types while autonomously observing traffic signs, traffic lights and the traffic plays an equally central role as the modelling of traffic conditions that are stochastic and reproducible at the same time. In addition to a deterministic traffic model and the optional coupling with microscopic traffic simulation (PTV Vissim), a phenomenological approach is available for this purpose which allows for the road segment-specific modelling of different statistic traffic densities. Furthermore, the driver model features a function for the comparison of road and tests thanks to which a measurement-based target speed profile can be retraced on the virtual route. Due to the synchronised real-time coupling between real driving simulation and es, a high-performance closed-loop integration of the real systems to be tested into the virtual environment is achieved, thereby creating a semi-virtual development environment, FIGURE 6. Individual subsystems such as the internal combustion engine on the or the whole vehicle on chassis dynamometers can thus be integrated into the simulation environment and tested in quasi-real operation. With this method, the complex modelling of an IC engine which would be necessary for the evaluation of fuel consumption and emissions in transient real operation is not required in early development stages, which is an essential advantage [3]. This method allows for the precise or stochastic variation of influencing factors relevant in real driving such as driver behaviour, vehicle properties and environment boundary conditions and the exact repetition of tests as required. On component es (for example engine, transmission or battery es) in contrast to test benches for the complete propulsion system or the vehicle, the remaining elements of the powertrain except the device under test are also part of the simulation environment [4] and can therefore be varied flexibly. This enables a virtual electrification of the propulsion system or tests in different virtual Simulation environment required/constructive for powertrain development at optimal time and costs Simulation environment optional according to objective Simulation environment not effective/high degree of complexity Hardware existent Test environment Office (MiL) Component Endurance Complex engine Powertrain Chassis dynamometer Real driving Driver Environment Vehicle Chassis Exhaust aftertreatment Auxiliary units Cooling system Powertrain Gearbox Electric motor Battery Internal combustion engine FIGURE 3 Test matrix for the transfer of real driving into the test environment ( APL) ATZ worldwide 10 2017 39
DEVELOPMENT Real Driving Emissions FIGURE 4 Optimised whole vehicle development process under RDE boundary conditions ( APL) FIGURE 5 Virtual components of the RDE development environment ( IPG Automotive) proto type vehicles, which allows for the evaluation of the behaviour of the device under test in different hybrid and vehicle variants [5]. An effective use of component test benches thus becomes possible already 40 in early stages of the development process since it allows for fundamental decisions regarding concepts and components, without the need for real prototype vehicles. Throughout the course of the development process, an increasing number of real components/assemblies can be integrated into the development and validation process on the propulsion system test bench up to whole vehicles on chassis dynamometers. The boundary conditions and scenarios of the test, however, remain
FIGURE 6 Virtual electrification using the example of an enginein-the-loop with different vehicle and powertrain variants ( IPG Automotive) the same, which leads to significant savings in time and cost in the overall process in addition to the good comparability of the test results. For development and application activities, real driving is thus consistently transferred from the road to the, where driving can be performed with automation and independently of weather conditions or influences depending on the time of day. SUMMARY Reproducibility of emission measurements cannot be attained in test driving on real roads due to a multitude of influences that are external and depend on the process, while being a fundamental requirement for effective propulsion system development. Two particular strengths of tests on es, in contrast, are high reproducibility as well as a high level of automation. Thus, with an increasing proportion of simulation, the presented method allows to transfer representative real driving flexibly and exactly to es throughout the entire development process. Based on systematic variations, the influence of vehicle-specific aspects (for example overall weight, ECU software data status, operating strategy, hardware variants), driver behaviour as well as traffic and environment boundary conditions on consumption, emissions and driving performance can be quantified and critical real operating conditions identified early on. As a result, with the virtualisation described, the utilisation of es and with it the efficiency of the propulsion system development process can be increased significantly. REFERENCES [1] Lensch-Franzen, C.; Gohl, M.; Mink, T.: Impact analysis of fuels, operating fluids and combustion parameters; focus raw emission behaviour. 4 th International Engine Congress, Baden-Baden (Germany), 2017 [2] Lensch-Franzen, C.; Gohl, M.; Becker, Mink, T.: The interaction between tribology, thermodynamics and emissions under real driving conditions (RDE). 11 th International MTZ Conference The Powertrain of Tomorrow, Frankfurt/Main (Germany), 2017 [3] Disch, C.; Koch, T.; Spicher, U.; Donn, C.: Engine-in-the-Loop as a Development Tool for Emissions Optimisation in the Hybrid Context. In: MTZworldwide 75 (2014), No. 10, pp. 40-46 [4] Donn, C.; Bensch, V.: Real-Time Capable Model Environment for Developing and Testing Hybrid and Battery Electric Vehicles. 11 th International MTZ Conference The Powertrain of Tomorrow, Frankfurt/ Main (Germany), 2017 [5] Donn, C.; Pfeffer, R.; Bensch, V.: Model-Based Testing on the Engine Test Bench Semi-Virtual Examination of Hybrid Powertrain Systems in Real Driving Conditions. 2017 JSAE Annual Congress, Yokohama (Japan), 2017 ATZ worldwide 10 2017 41