Model selection strategy for the sequential optimization design of electromagnetic devices

Model selection strategy is investigated for sequential optimization method (SOM) to deal with the optimization design problems of electromagnetic devices. Four kinds of approximate models, response surface model, radial basis function model, Kriging model and artificial neural network model are considered respectively. From the analysis of two mathematic test functions and an IEEE TEAM benchmark problem, we can see that the total cost of finite element analysis can be obviously saved by SOM with all kinds of approximate models. Moreover, the efficiency of SOM depends a little on the types of approximate models.