Thermal Neural Network for Temperature Modeling in E-Motors

Schaeffler is pursuing and developing new approaches to temperature modeling in electric machines. In this session, see how they compare different AI methods to achieve this task with a focus on thermal neural networks (proposed by Wilhelm Kirchgässner, Oliver Wallscheid, and Joachim Böcker in “Lumped-Parameter Thermal Modeling With State-Space Machine Learning,” Eng. Appl. Artif. Intell., vol. 117 [2023]).

This thermal neural network combines data-driven and simulation-driven approaches and provides several advantages in terms of flexibility, initialization strategy, and plausibility. Explore the use of MATLAB^® for concrete scenarios along with the required modifications and plausibility checks conducted with domain experts. Finally, see how this neural network can be implemented in Simulink^® and how C code can be generated to run the network on a microcontroller.