Multiphysics Simulation and Model-based System Testing of Automotive E-Powertrains

Abstract

Model-Based System Testing emerges as a new paradigm for the development cycle that is currently gaining momentum, especially in the automotive industry. This novel approach is focused on combining computer simulation and real experimentation to shift the bulk of problem detection and redesign tasks towards the early stages of the developments. Along these lines, Model-Based System Testing is aimed at decreasing the amount of resources invested in these tasks and enabling the early identification of design flaws and operation problems before a full-vehicle prototype is available. The use of Model-Based System Testing, however, requires to implement some critical technologies, three of which will be discussed in this thesis. The first task addressed in this thesis is the design of a multiplatform framework to assess the description and resolution of the equations of motion of virtual models used in simulation. This framework enables the efficiency evaluation of different modelling and solution methods and implementations. In Model-Based System Testing contexts virtual models interact with physical components, therefore it is mandatory to guarantee their real-time capabilities, regardless of the software or hardware implementations. Second, estimation techniques based on Kalman Filters are of interest in Model- Based System Testing applications to evaluate parameters, inputs or states of a virtual model of a given system. These procedures can be combined with the use of Digital Twins, virtual counterparts of real systems, with which they exchange information in a two-way communication. The available measurements from the sensors located at a physical system can be fused with the results obtained from the simulation of the virtual model. Thus, this avenue improves the knowledge of the magnitudes that cannot be measured directly by these sensors. In turn, the outcomes obtained from the simulation of the virtual model could serve to make decisions and apply corrective actions onto the physical system. Third, co-simulation techniques are necessary when a system is split into several subsystems that are coordinated through the exchange of a reduced set of variables at discrete points in time. This is the case with a majority of Model-Based System Testing applications, in which physical and virtual components are coupled through a discrete-time communication gateway. The resulting cyber-physical applications are essentially an example of real-time co-simulation, in which all the subsystems need to achieve real-time performance. Due to the presence of physical components, which cannot iterate over their integration steps, explicit schemes are often mandatory. These, however, introduce errors associated with the inherent delays of a discrete communication interface. These errors can render co-simulation results inaccurate and even unstable unless they are eliminated. This thesis will address this correction by means of an energy-based procedure that considers the power exchange between subsystems. This research work concludes with an example of a cyber-physical application, in which real components are interfaced to a virtual environment, which requires the application of all the MBST technologies addressed in this thesis.