基于多目标有效全局优化算法的直线感应电动机优化设计

Optimal Design of a Linear Induction Motor Using Multi-Objective Efficient Global Optimization

直接针对有限元模型的电机优化设计周期长、效率低。多目标有效全局优化算法基于Kriging代理模型引导采样,从而可以减少优化过程中有限元模型的迭代次数,缩短优化设计周期,并且保证了优化结果的准确度。同时将并行计算的方法与多目标有效全局优化算法相结合,进一步提高了算法的效率。最后成功地将该算法应用于直线感应电动机的三维有限元模型的优化设计中,得到了经过有限元模型分析值组成的二维Pareto最优前沿解集,为基于有限元模型的直线感应电动机优化设计提供了新方法。

Integration of finite element models（FEM） in the optimal design process of electrical machine is complex and time-costly. The multi-objective efficient global optimization（MEGO） algorithm uses Kriging surrogate model as a guide on the optimization problem. The computationally expensive FEM is replaced by Kriging surrogate model, which can reduce the iterations of the FEM in the optimal design process. A Parallel strategy is integrated with MEGO in order to further save the time of optimization. A multi-objective optimization is achieved, by applying the MEGO algorithm to a 3D FEM of the linear induction motor. A 2D Pareto set composed of 3D FEM solutions is obtained with an affordable time-cost. This paper provides a new method for the optimal design of linear induction motor with FEM.