Analysis and Optimization of Permanent Magnet-Assisted Synchronous Reluctance Machine Based on Neural Network Fitting

This paper proposes an optimization procedure for improving the performance of a permanent magnet-assisted synchronous reluctance machine (PMASynRM) including average torque, torque ripple, efficiency and the total harmonic distortion of back electromotive force. Constraints of maximum stress of the rotor core and maximum field intensity in the magnets are also considered to ensure both mechanical and magnetic safety of the machine. Surrogate models of the PMASynRM, generated by neural network fitting, are applied to the procedure, and their accuracy is confirmed by the linear regression analysis. Pattern search algorithm is used for the optimization. Validity of the optimal solution and improvement of the machine performance are verified with finite element analysis. A PMASynRM that undergoes the same design optimization method is verified by the experiment.