Fast Design of Multilayered Shields Using Surrogate Model and Space Mapping

Designing an efficient shield using simulation requires an expansive computational cost due to large ratio of width to thickness. This paper presents a two-level surrogate based optimization process to efficiently design multilayered shields. The method first exploits space-filling sampling to obtain a global optimal start point. The coarse model is then adopted to be the first-level surrogate. Two-dimensional (2-D) simulation is selected as the coarse model to avoid the time consuming high-fidelity 3-D calculation. Sequential design is followed where response surface approximation (RSA) is applied to construct the surrogate of the coarse model. The evaluation time of the RSA is negligible and it is the second-level surrogate. Finally, the optimum design found by the surrogate is refined using space mapping technique to give an approximation of the location of the optimum design of the original problem. By taking advantages of both surrogate modeling and space mapping, the proposed approach can significantly improve the efficiency of the design process. The method is applied to optimize a double shield and a U-shaped double shield to show its efficiency.