Principal Component Optimization With Mesh Adaptive Direct Search for Optimal Design of IPMSM

In this paper, a novel global search algorithm, implementing principal component optimization (PCO) in combination with mesh adaptive direct search (MADS) is proposed. The PCO is based on principal component analysis, a widely used multivariate technique in statistics, while MADS is a highly effective local search algorithm, implemented to find local optima. Their integration reduces computation time and improves convergence reliability to a global optimum. The proposed algorithm is verified via application of a test function and comparison to conventional hybrid algorithms such as particle swarm optimization with MADS. Finally, the algorithm is applied to the optimal design of an interior permanent magnet synchronous machine for minimization of total harmonic distortion and torque ripple via finite element analysis.