Technical Paper
Development of reduced order modeling technology with AI and its application to Model-Based-Development
2024-04-09
2024-01-2850
This paper presents the reduced order modeling techniques with AI for MBD. In vehicle development, some detailed physical models are replaced with the reduced order models to speed up simulations. In recent years, with the development of reduced order modeling technology with AI, it is expected that reduced order modeling work will become more efficient and that simulation speed will increase. However, there are limits to the simulations (MILS/HILS/bench systems) that can be embedded. Therefore, by utilizing a model format (ONNX: Open Neural Network Exchange) that can be commonly used among machine learning frameworks and a standard interface standard for simulation tools (FMI: Functional Mock-up Interface), the limitations of embedding can be eased. This research targeted a vehicle model used in engine surge simulation. By learning the MILS simulation results with LSTM, a reduced order model was created in ONNX format.