Surrogate Model-Based Design Optimization of Consequent-Pole Asymmetric Rotor Hybrid Interior Permanent Magnet Synchronous Machine

This paper presents design optimization of a consequent-pole asymmetric rotor hybrid interior permanent magnet synchronous machine (CPAR-HIPMSM) based on the surrogate model to improve torque performance. First, the influence of key design parameters is analyzed through finite element model (FEM), and the sample database for surrogate model is obtained. Second, a Bayesian optimization improved support vector regression (SVR), termed as BO-SVR, is employed to establish the surrogate model for rapid and accurate torque performance calculation. Third, sensitivity analysis based on Spearman correlation coefficient (SCC) is implemented to classify the importance of design parameters for optimization. Finally, the sequential subspace northern goshawk optimization (SS-NGO) is combined with BO-SVR for torque performance optimization of CPAR-HIPMSM. Compared with the conventional direct optimization method based on FEM and non-dominated sorting genetic algorithm-II (NSGA-II), the proposed BO-SVR+SS-NGO method significantly improves optimization efficiency and obtains the optimized machine with increased average torque and reduced torque ripple. Its effectiveness is validated by both prototype experiments and FEM simulations.