Gamma Technologies has published a blog explaining how GT-SUITE’s Machine Learning Assistant can convert a high-fidelity Finite Element (FE) thermal model of an electric motor into a NeuralODE-based metamodel for system-level EV simulation. The approach uses Design of Experiments (DOE) data generated from the FE model, covering component losses, ambient temperature, coolant conditions, and oil spray cooling, to predict motor hotspot temperatures and coolant outlet temperature.
The trained metamodel is integrated into the system model using the MetamodelHarness and validated against the FE model across different driving conditions. It can also be exported as FMU or C code for third-party integration or ECU deployment. This method aims to preserve FE-level thermal accuracy while supporting broader, system-level scenario exploration.
Image generated by: ChatGPT