1.1. Motivation
Due to the constant social, technological, and legal changes, pressure is increasing on the automotive industry ( Figure 1. ). Climate change and environmental awareness are the two most important factors. 1, 2 ]. In order to limit emissions of harmful gases to human health and the climate, legislators and governments are developing regulations. 1 ]. The European Union has defined a plan to reduce CO 2 Emissions in the transport sector, specifically for light commercial vehicles and passenger cars. Regulation (EU) 2019/631 has been introduced. 3 The CO 2 The fleet target has been set at 95 g/km. This will be further reduced to 60 g/km by 2030. 3 ]. Manufacturers who do not meet these targets will face significant fines 3 ]. The EU Commission also published a proposal on the adaptation of the European Market’s Pollutant Emission Regulation. 4 [Euro 7] The EU Commission aims to limit further the pollution caused by harmful emissions components. In some cases, the limits of Euro 6d ISC FCM will be further reduced. 4 ]. Moreover, emissions of other features, such as NH 3 N 2 O and CH 4, will be limited 4 ]. The range of driving situations in which the law applies will be significantly expanded. 4, 5 ]. ] According to the latest information, Euro 7 legislation could be implemented as soon as July 2025 4 ]. This will, combined with the rapid technological advancement, result in a faster time to market 6 ].Figure 1. Motivation for Virtual Calibration. Automotive manufacturers must respond to these short-term changes, specifically by adopting vehicle concepts and adding improved emission-reducing systems [ 7, 8 ]. Internal combustion engines (ICEs) can significantly reduce fuel consumption and pollution by electrifying their powertrains. 7, 9 ]. These powertrains are more complex than conventional powertrains, which makes the development of them even more challenging. 10 ]. The increased number of possible states in which the engine can be operated results in a greater number of control functions that are interrelated. 11, 12 ]. To find a calibration that is compatible with all disciplines of development, it’s necessary to test and validate on the road as well as in existing facilities such as chassis dynamometers. Vehicle complexity is driven by increasing customer expectations and technological advancements. 6 To counter this trend, automotive companies must implement robust and flexible methods and processes for developing and calibrating tasks. 13, 14, 15 ]. In this context, virtualized test environments are an important area of research. By moving development tasks to a virtualized environment, time and costs associated with calibration can be reduced up to 20% 16, 17 ].
1.2. Virtual Calibration for Automotive Control Units
As shown in the diagram, the vehicle development process can be divided into four major stages. Figure 2. After the definition and conception of hardware and software begins the calibration phase, which is usually divided into three stages. Often, the results of calibration are generated and validated using different test benches. To achieve this, prototype vehicles and component test benches on chassis dynamometers can be used. Later phases involve the operation of the prototype vehicle on the road in order to evaluate the close-to-customer performance. In an ideal world, all collected data is stored centrally within a measurement database. Figure 2. Virtualized methods using X-in-the-loop test rigs (XiLs) can be integrated with existing processes. Figure 2. ). These virtualized testbenches consist of a real hardware target component (device under testing, DUT) and a simulation environment. Different setups are possible for XiL Testbenches, which vary in their virtualization level and accuracy relative to the real system.
Previous reports have provided a detailed description of each test environment. 18, 19 ]. The simulation environment is continuously optimized and refined by the available measurement data during the development process. During the development phase, the simulation environment is continually optimized and refined based on the measurement data available. In order to improve the reliability of generated results, parts of the target hardware are added during development. Thus, so-called hardware- (HiL), engine- (EiL), or powertrain-in-the-loop (PiL) test benches can be derived from existing test benches ( Figure 2. ). Different calibration disciplines have been able to integrate test benches of different virtualization levels into the calibration process. 20, 21, 22, 23, 24 ]. A prototype vehicle is the only way to achieve a final release of target hardware and software. Virtualized testbenches are complementary to a holistic approach and can reduce costs and time.
1.3. Use Case: Virtual calibration of the operating strategies of a hybrid-electric vehicle
This is a high-potential use case of virtual calibration. Wu et al. have partially shown this in various studies. [ 25 ], Merl et al. [ 26 ], Kuznik et al. [ 27 ], and Schmidt et al. [ 28 The operating strategy controls the interface between the electric and conventional drive systems (Chapter 2.2) for P2-plug-in hybrid vehicles (PHEVs). The operating strategy determines the interface between electric and conventional drive systems in the P2-plugin hybrid vehicles (PHEVs), which are the subject of this paper (Chapter 22). Figure 3. ). The hybrid operating strategy, among other things, is responsible for start/stop decisions, load point shifting in hybrid operation, and predictive operating strategy. The P2-PHEV has many states, compared to hybrid topologies with fewer possible states. This makes the optimal calibration of these vehicles more difficult. The hybrid operating strategy is well-suited to the use of virtual methods in the development process. Figure 3. The hybrid operating strategy has a large influence on other disciplines of calibration, such as emissions or driving ability. 29 ] ( Figure 3. ).
A holistic approach is therefore required to calibrate. From the beginning of the optimization process, it is important to consider relevant quality criteria from other disciplines. Wang et al. have investigated similar approaches in open-loop and control function simulation. [ 30 ], Duan et al. [ 31 [ ] and Gorke et. al. [ 32 ]. These approaches, however, did not use real hardware targets but instead used simulations. This holistic approach reduces development costs and time compared to the conventional iterative method. In this scenario, the vehicle as a whole is replicated within a simulation environment. All functions that are relevant to representative operation, including the control functions that need to be calibrated and other parts to simulate ( Figure 2. ). The model must run faster than real-time criteria to enable time-efficient optimization. The parameterization is primarily based on simulation results and measurement data. Figure 2. ). This setup differs from the MiL in that it uses real hardware and software of the target system to control the model. The calibration becomes more mature and can be transferred to the target system. Due to the target hardware being used, a real-time simulation environment is required. Accelerating the simulation is not possible. This setup can still save time, as it allows different test environments to be evaluated without the need to condition real components. The calibration can also be optimized using the pre-calibration of the MiL environment. Final validation of the virtual results is done with a vehicle prototype and then released for homologation.