A direct coupled solution methodology for efficient robust optimizations of inverse problems under uncertainty

In the existing resolution methodology for robust design optimizations, the procedures for solving robust optimization and uncertainty quantization as well as the use of high fidelity models are completely decoupled and independent from each other. As a result, the overall cost is typically the product of the costs of the three approaches. Such methodology is simple but more expensive than necessary. To develop an efficient robust optimizer, a direct coupled solution methodology based on an evolutionary algorithm is proposed. Stochastic approximation method is employed to minimize the computational burdens when computing the gradient information in designing the exploiting phase. Numerical results are reported to showcase the merits of the proposed methodology.