Reduced Order Modeling with AI for Model-in-the-Loop Simulation

Model-in-the-Loop (MIL) simulation is a testing technique widely used in the automotive industry for developing and testing control systems. It involves integrating a mathematical model of a physical system being developed into a virtual environment and performing rigorous testing. MIL simulation offers benefits in system design and development, but it also has several computational challenges that can impact the efficiency of the simulation. This paper focuses on enhancing the computational efficiency of a MIL simulation environment for Electric Vehicle (EV) Powertrain system using a Reduced Order Model (ROM) with Artificial Intelligence (AI). The EV MIL simulation environment has been developed using a first-principles model in MATLAB/Simulink to analyze the system performance. However, system level MIL simulation performance is poor due to high-fidelity plant models. This paper's objective is to accelerate system level MIL simulation performance by leveraging ROM using AI techniques. To achieve this, the high computational high-fidelity plant model has been replaced with trained Neural State Space (NSS) ROM model using MATLAB/Simulink tool. Simulation results show that the AI driven ROM model achieved an accuracy of 85% compared to the estimation data. This implementation has improved the overall MIL simulation computational performance by three times. This approach enables a simulation engineer to analyze or simulate complex systems by significantly reducing computation time while preserving the expected fidelity within a controlled error margin. Keywords: Model-in-the-Loop, simulation, testing control system, MIL, system design, computational efficiency, EV, powertrain system, Electric Vehicle, artificial intelligence (AI), MATLAB, Simulink