Review of surrogate model assisted multi-objective design optimization of electrical machines: New opportunities and challenges

Abstract

This paper overviews surrogate model-assisted multi-objective design optimization techniques of electrical machines for efficient, accurate, and robust design optimization to ease design issues due to unprecedentedly increasing machine performance requirements. Firstly, the mechanism of surrogate-assisted modeling is introduced by comparing it with conventional physical modeling approaches. The relevant techniques are then categorized and subsequently reviewed in terms of the design of experiments, surrogate model construction, and multi-objective optimization algorithms. The potential application prospects for machine design optimization are highlighted. Finally, three surrogate-assisted modeling methods, i.e., transfer learning-based models, gradient sampling-based multi-fidelity models, and search space decay-based surrogate models, are quantitively compared by applying them to the design optimization of a five-phase permanent magnet synchronous machine.