Two-Level Surrogate-Assisted Differential Evolution Multi-ob-jective Optimization of Electric Machines Using 3D Finite Element Analysis (FEA).

Abstract

Many parameters are considered in electric machine design and an optimization algorithm can be used. These usually need thousands of design evaluations before meeting the termination criterion. Time consuming 3D finite element analyses (FEAs) are not tenable although machines with 3D flux paths, such as axial flux and transverse flux, cannot be accurately evaluated with 2D models. One solution is to use surrogate models rather than 3D FEA; however, the accuracy of surrogate models reduces for a large and nonlinear search space. Another solution can utilize algorithms that find the global optima with a minimum number of design evaluations. A combination of these two solutions is proposed here.