Enriched Serial-Loop Optimization Method for Efficient Reliability-Based Electromagnetic Designs

摘要

An enriched serial-loop method for reliability-based design optimization is presented to substantially improve computational efficiency as well as numerical accuracy when applied to electromagnetic design problems in the presence of uncertainties. Two numerical improvements are made over the original serial-loop optimization method, which employs a series of cycles of an equivalent deterministic optimization and reliability analysis. One is a feasibility check technique for the probabilistic constraints conducted only at the first design cycle. It identifies inactive constraints, which need not be considered after the next design cycles. The other is a reliability improvement scheme for slightly violated probabilistic constraints, which is performed at the end of iterative design cycles. When an obtained optimum is unsatisfactory to all probabilistic constraint conditions prescribed, the design point is slightly shifted toward a feasible design region by utilizing probabilistic information obtained. Finally, the TEAM Workshop Problem 22 is tested to compare the proposed method with the existing ones from a numerical efficiency and accuracy point of view.