Two-Stage Surrogate Model for Finite-Element-Based Optimization of Permanent-Magnet Synchronous Motor

The use of finite-element-method (FEM) simulation in electrical machine optimal design is affected by two main problems: the computation time in FEM simulations and the large number of parameters of the electrical machine. Here, we propose a surrogate model to use with electrical machines, based on statistical multiple correlation coefficients (R2) analysis and moving least squares (MLS) approximation. In the context of an optimization process, which needs a large number of evaluations strongly depending on the number of parameters, the computation effort is small compared to the time that can be saved. We validate this method by applying it to the optimal design of a synchronous machine. The results show that the torque per weight ratio improved by about 13% in comparison with that obtained by classical optimization methods.