Automated, holistic design of power modules
Brief description
Power electronics is a key component for a CO2-neutral society. Power electronic converter stages are required for energy conversion. On its way from the source to the consumer, the energy passes through several such converter stages - which is why the highest possible efficiency is essential. The project aims to pave the way for an AI-based module layout of power electronic converter stages in order to further improve the efficiency of power electronic modules.
The design of a power module is a multi-physical design problem where a trade-off between cooling, low parasitic elements and mechanical design has to be found. There are many changing variables to adjust the different physical and electrical properties of the module. Until now, the geometric design and optimization has mostly been carried out in the traditional way using manual 3D simulation coupled with linear parameter variation - a very time-consuming and resource-intensive process. At the same time, 3D simulations require high computing power and a lot of time. By combining a data structure with analytical optimization and AI-supported analysis, hundreds of thousands of layouts can be evaluated in seconds, complex design problems can be solved efficiently and new insights can be gained - all while further increasing efficiency. The aim of the project is a bidirectional transfer of knowledge between the university and the company. Semikron-Danfoss shares its knowledge of module design, while the university contributes the use of AI to the module design process.