Combined Gradient Boosting Decision Tree and Differential Evolution Algorithm for Optimization Design of Permanent Magnet Arc Motor

Permanent magnet arc motor (PMAM) is widely used in some scanning systems, such as the astronomical telescopes and large antennas. In order to address the optimization design of the PMAM, a new multi-objective optimization design method by combing gradient boosting decision tree (GBDT) and differential evolution algorithm (DEA) is proposed in this paper. Specifically, the motor structure of the PMAM is firstly presented. And then, a finite-element (FE) model of the PMAM is built to obtain sample data. Based on the sample data, the powerful machine learning algorithm called GBDT is innovatively introduced to establish surrogate model, which can identify the mapping relationships between the optimization objectives and the structural parameters. Subsequently, a popular intelligent search algorithm named DEA is employed to conduct the optimization design. Finally, both FE simulation and prototype experiment are performed to verify the effectiveness and advantages of the proposed method.