Thrust Performance Improvement of PMSLM Based on Lasso Regression With Embedded Analytical Model
Zhilei Zheng, Jiwen Zhao, Lijun Wang
Abstract
To solve the problem that both analytical model (AM) and finite element model are difficult to give consideration to the calculation accuracy and efficiency at the same time in motor optimization work, a novel Lasso regression with the embedded AM, called EAM-LR, is proposed to quickly and accurately calculate the thrust performance of the permanent magnet synchronous linear motor (PMSLM) in this article. First, the thrust performances of PMSLM are analyzed by the AM to determine the variation range of structural design parameters. Based on the variation range, a finite-element sample database is established. Then, combined with the finite-element sample database, the analytical mapping functions derived from AM, are integrated into Lasso regression to establish EAM-LR. The accuracy of EAM-LR was verified by comparison with AM, traditional lasso regression, and well-known extreme learning machine. Finally, combined with the EAM-LR, the chaotic golden section algorithm is introduced to search the optimal structure parameters of PMSLM, and the control simulation and prototype experiment prove the effectiveness of the proposed method.