Surrogate Modeling and Two-Level Infill Criteria Applied to Electromagnetic Device Optimization

The surrogate-based optimization is largely used in engineering design to find optimal performance characteristics of computational expensive simulations such as the finite element method. An original multiple surrogate-based optimization criterion is proposed. The approach was tested on the Zakharov benchmark function and on the Loney solenoid benchmark problem. Simulation results and comparisons with some traditional Kriging-based global optimization methods show that the proposed method can be used very effectively in electromagnetic design optimization.