Topology optimization of electric motors based on finite element computation

Abstract

Topology Optimization (TO) is a powerful tool for engineers to help them explore suitable structures. More importantly, it can find the topologies of electric motors that never existed before. This thesis work is oriented toward the TO of electric motors by developing various aspects of the subject. First, an optimization framework for TO is developed and tested. Since a complete optimization process for an electric motor requires a motor performance evaluator and an optimization algorithm working together, a coupling is done using both. Furthermore, a TO methodology is developed and tested based on the binary encoded genetic algorithm (GA) and filtering process. A well-known Synchronous reluctance motor (SynRM) test case is used to accomplish the tests and validate the tools and methodology. Afterward, the methodology is applied to an asymmetric rotor interior permanent magnet (AIPM) motor, representing a more realistic test case. A high-resolution interpolation and edge-smoothing method are employed to increase modeling accuracy. An asymmetric rotor pole is presented for different problem formulations. Finally, deep learning (DL) and physics-informed generative adversarial network (PIGAN) are investigated for faster magnetic field approximation for the simulation of coaxial magnetic gear (CMG) and permanent magnet linear synchronous motor (PMLSM). These examples allow us to experience the feasibility and efficiency of employing DL algorithms for the performance evaluation of electric motors, which can significantly benefit the optimization work.