LAUSR

In this article, we present a two-step strategy for the torque performance improvement of a dual-stator permanent magnet arc motor (DS-PMAM), which is used in some industrial applications, such as robot joint. The proposed strategy can achieve high average torque, low torque ripple, and, eventually, improve the quality of manufactured products. In the first step, the no-load torque model caused by the end effect is analytically established. In order to suppress the inductance unbalance and the torque ripple, two DS-PMAM models with different winding connection methods are investigated and compared in terms of inductances, backelectromotive force, and torque characteristics. In the second step, a new sensitivity analysis method based on the shapley additive explanations value is first proposed to evaluate the sensitivity of each structural parameter to different optimization objectives. Then, an efficient optimization design method is proposed by combining the machine learning algorithm named eXtreme gradient boosting (XGBoost) and the intelligent optimization algorithm called nondominated sorting genetic algorithm-II (NSGA-II). The XGBoost is innovatively introduced to efficiently approximate the function relationship between the optimization objectives and the structural parameters. The NSGA-II is adopted to determine the optimal combination of the structural parameters and motor performances. Finally, based on the optimal results obtained by the two-step strategy, a prototype of DS-PMAM is manufactured and tested to validate the proposed strategy.