An adaptive optimization method using Kriging model and Latin hypercube design and its application to optimum design of PMLSM

This paper presents a computationally efficient optimal design algorithm for electromagnetic devices by combining Kriging interpolation approximation of the objective function and genetic algorithm. In this method, the design space is successively reduced with the iteration, and Pareto-optimal sampling points are generated by using Latin hypercube design (LHD). The proposed algorithm is applied to the optimum design of permanent magnet linear synchronous motor (PMLSM) and the computational efficiency is investigated.