Kriging-Assisted Multiobjective Optimization of Embedded Magnetic Encoder in PM Synchronous Machines

Abstract

The embedded magnetic encoder (EME), located in stator slot of permanent magnet (PM) synchronous machines (PMSMs), monitors the PM flux-density in the cavity and feedback the rotor angle for vector control. To reduce the total harmonic distortion (THD) of flux-density and improve the angle estimation accuracy, the topology optimization of PMSMs can be carried out. However, as THD is set as the only objective during optimization, the other characteristics, e.g., back electromotive-force (back-EMF) due to PMs, and winding inductance of PMSMs will uncontrollably change. Hence, a systemic design optimization of PMSM combined with EME is necessary. Different from the traditional multi-objective optimization of electric machines based on the time-consuming finite element analysis (FEA), in this paper a Kriging-assisted surrogate model is proposed to replace FEA during optimization, achieving time saving and high prediction accuracy simultaneously. The decision variables are carefully selected. e.g., a Cosine function reshaping method is proposed and proper linear Hall installation position is analyzed. In the preliminary optimization, the Pareto front with minimum THD and maximum rotor saliency is obtained. Further, by modifying the THD objective with retaining the third harmonic component, the THD becomes lower and rotor saliency is increased in the optimal case. Finally, three different rotors are manufactured to verify the effectiveness of proposed reshaping method and improved optimization strategy.