This work presents a systematic approach to identify the thermal behavior of arbitrary automotive component systems. The proposed methodology leverages experimental temperature data and prior knowledge from Computational Fluid Dynamics (CFD) simulations to achieve a consistent system identification. The key aspects of the approach include thermal behavior identification through minimizing the least-squared error between the predicted thermal lumped parameter model and the experimentally measured temperature data, ensuring a robust and accurate representation of the system's thermal characteristics. The identified system model is then utilized to generate transient system responses for defined use-cases, enabling a comprehensive understanding of the thermal behavior under various operating conditions. The identification algorithm is based on the least-square programming algorithm from SciPy, providing a robust and efficient computational framework.
Ensuring the reliability and durability of automotive components is crucial, as they must withstand the wide range of temperatures encountered during operation. To this end, the temperature-critical components are experimentally tested and simulated using CFD. The proposed methodology offers the capability to understand thermal interactions in experimental data and to generate transient responses based on stationary CFD simulations. Additionally, this work lays the groundwork for predicting temperatures in future vehicles with physics-informed neural networks.
The method is tested with experimental temperature data and a numerical model of the component space of one control unit in the A-pillar of BMW's current 7 Series.