Reduced Order Modeling Technology with AI for Model-Based-Development

2024-01-2850

04/09/2024

Features
Event
WCX SAE World Congress Experience
Authors Abstract
Content
This paper introduces reduced-order modeling techniques with Artificial Intelligence (AI) for Model-Based Development (MBD). In vehicle development, detailed physical models are replaced by reduced-order models (ROM) to expedite simulations. With recent advancements in AI-based reduced-order modeling, it is expected that modeling work will become more efficient, leading to reduced simulation times. However, the range of simulations (Model-in-the-Loop Simulation - MILS, Hardware-in-the-Loop Simulation - HILS, bench-system) compatible with ROM is limited. To overcome this limitation, this study leverages the ONNX format (Open Neural Network Exchange), a universally supported format among machine learning frameworks, and the Functional Mock-up Interface (FMI), a standard interface format for simulation tools, to enable general-purpose embedded technology with ROM.
This study employs a vehicle model in engine surge simulations to validate AI-based reduced-order modeling for MBD. In MILS simulations, the ONNX-format model, trained using Long Short-Term Memory (LSTM), is integrated into an FMI-format model compatible with the simulation environment. This FMI-format model is then incorporated into MILS/HILS/bench systems, confirming its capability for accurate simulations. Thus, we have successfully established AI-based reduced-order modeling technology for comprehensive MBD.
Meta TagsDetails
DOI
https://doi.org/10.4271/2024-01-2850
Pages
5
Citation
Inagaki, T., Nasu, T., Takeshige, M., Iwata, M. et al., "Reduced Order Modeling Technology with AI for Model-Based-Development," SAE Technical Paper 2024-01-2850, 2024, https://doi.org/10.4271/2024-01-2850.
Additional Details
Publisher
Published
Apr 09
Product Code
2024-01-2850
Content Type
Technical Paper
Language
English