Efficient Multi-Objective Optimization of an Asymmetric Rotor IPMSM with Hybrid Poles

<p dir="ltr">This paper presents a novel hybrid poles asymmetric interior permanent magnet (HPAIPM) motor topology for electric vehicle applications featuring the combined utilization of rare-earth and ferrite permanent magnets (PMs). It fully utilizes the magnetic-field-shifting effect in improving torque density by adding a ferrite at the upper part of the V-shaped asymmetric rare earth PM pole and optimizes the harmonic distribution of the air-gap magnetic field to achieve a significant reduction in torque ripple. In response to the optimization challenges of complex rotor structure motors, this work applies surrogate-assisted methods to achieve rapid optimization. Firstly, a constraint space Latin hypercube design algorithm is proposed to solve the problem of selecting sampling points in a confined space composed of strong constraints on design variables. This not only avoids the generation of invalid points but also optimizes the spatial distribution characteristics of sampling points, reducing the prediction error of the surrogate model by 29.1%. Secondly, a surrogate modeling method based on transfer learning neural network is given, which freezes some hidden layer coefficients of the neural network to avoid repeated training of surrogate models corresponding to different ferrite materials, reducing the number of training samples by 60%. According to the optimal design scheme, the proposed HPAIPM topology can reduce torque ripple by over 80% and maintain high torque density compared to conventional interior PM and asymmetric interior PM topology. Finally, a small prototype of the proposed HPAIPM topology is designed, manufactured and tested for validation.</p>