Surrogate-Assisted Optimization of a Five-Phase SPM Machine With Quasi-Trapezoidal PMs
Yiming Ma, Jin Wang, Libing Zhou, Kang Shuai
Abstract
In this article, an improved surrogate-assisted design optimization method is proposed and applied to the torque performance optimization of a five-phase surface-mounted permanent magnet machine with quasi-trapezoidal permanent magnets. It integrates the Taguchi and surrogate-assisted methods; the training data for surrogate models are generated by the Taguchi method to reduce reliance on the finite-element analysis (FEA), in turn, the local optimum problem caused by the Taguchi method is tackled by the surrogate model. The generalization ability of surrogate models is enhanced by the parameter transfer technique. Using the fuzzy best-worst method, the decision process of multiple solutions is placed before optimization so that the multiobjective optimization can be transformed into a single objective to facilitate the search process for the optimal design. The optimization results are validated by the test results from a prototype. Compared with the FEA-based parametric scanning method, the optimization cycle is shortened by 94% and a globally optimal design can be obtained.