An Improved Robust Optimization Algorithm: Second-Order Sensitivity Assisted Worst Case Optimization

The demand of performance robustness against the uncertainty in design variables is required during the optimization process. From the viewpoint of fast robustness evaluation for the electromagnetic problem, based on the worst case scenario approximation and design sensitivity analysis, a new robust optimization algorithm, the sensitivity-assisted worst case optimization (SA-WCO), is developed. In order to keep higher accuracy of the robustness prediction, the second-order sensitivity implemented by the hybrid direct differentiation-adjoint variable method is introduced into the robustness assessment. The performance and robustness of the proposed SA-WCO algorithms are investigated through a numerical experiment with the TEAM Problem 22.