Surrogate Model based Optimization of DSPMLSM with Multi-Sampling Points Adding Rule based on Support Vector Machine

In this paper, the force and force-mass ratio of a long primary double-sided permanent magnet linear synchronous machine (DS-PMLSM) is optimized under the constraint of efficiency-power factor production based on Kriging Surrogate model. Efficiency-power factor production is always treated as “Black-box” constraint which is evaluated by Kriging model. Hence, the inaccuracy of the evaluation of constraint based on Kriging model can cause the unfeasible solution in optimization results. In order to increase the accuracy of the optimization, a multi sampling points adding rule based on Support Vector Machine (SVM) is proposed. The effectiveness of the proposed method is verified by finite element model (FEM) and the optimization result is also validated experimentally base on a prototype.