Machine Learning Technique Based Multi-Level Optimization Design of a Dual-Stator Flux Modulated Machine With Dual-PM Excitation
Yao Meng, Shuhua Fang, Zhenbao Pan, Wei Liu, Ling Qin
Abstract
This article proposes a new machine learning technique based multi-level optimization (MLT-MLO) method to optimize a dual-stator flux-modulated machine with dual-PM excitation (DS-FMDPMM). The proposed MLT-MLO method using multi-level optimization can effectively alleviate the calculation burden caused by the multiple design variables in DS-FMDPMM. In addition, the proposed MLT-MLO method combines the support vector machine regression (SVR) and the non-dominated sorting genetic algorithm-II (NSGA-II) to conduct the motor optimization, which can effectively reduce the calculation time and improve the optimization efficiency. Moreover, before the optimization, a simplified analytical model is built to determine the design variables and a sensitivity analysis is carried out to reduce the workload. The topology of DS-FMDPMM and the flowchart of the proposed MLT-MLO method are introduced first. Then, based on the proposed MLT-MLO method, the DS-FMDPMM is comprehensively optimized for high torque production and low torque ripple. Finally, the finite element (FE) and experimental validations are carried out, which verify the effectiveness of the proposed MLT-MLO method.