Kriging-Assisted Multi-Objective Design of Permanent Magnet Motor for Position Sensorless Control

This paper presents a novel multi-objective evolutionary algorithm (MOEA) based on surrogate models. Kriging surrogate models are created and updated with the solutions from the MOEA for reducing the number of evaluations of the computationally expensive finite-element analysis models. The proposed method is applied to solve a multi-objective design problem of a surface-mounted permanent magnet motor in order to improve its sensorless control capability without comprising other performances, such as output torque and torque ripple. The Kriging-assisted MOEA is proved to have the potential of providing good solutions with a limited computation time budget.