A Novel Surrogate-Assisted Multi-Objective Optimization Algorithm for an Electromagnetic Machine Design

Abstract

To design electric machines, the motor performance, cost, and manufacturing have to be considered. Hence, researchers have called this the multi-objective optimization (MOO) problem in which the goal is to minimize or maximize several objective functions at the same time. In order to solve the MOO problem, various algorithms, such as nondominated sorting genetic algorithm II and multi-objective particle swarm optimization, have been widely used. When these algorithms are applied to the electric machine design, much time consumption is inevitable due to many times of function evaluations using a finite-element method. To solve this problem, a novel surrogate-assisted MOO algorithm is proposed. Its validity is confirmed by comparing the optimization results of test functions with conventional optimization methods. To verify the feasibility of its application to a practical electric machine, an interior permanent magnet synchronous motor is designed.