Uncertainty-Based Multidisciplinary Design Optimization for Feedback-Coupled Systems Under Both Parametric and Metamodeling Uncertainties

Abstract

Abstract Uncertainty-based multidisciplinary design optimization (UMDO) is an effective methodology to deal with uncertainties in the engineering system design. In order to shorten the design cycle and improve the design efficiency, the time-consuming computer simulation models are often replaced by metamodels, which consequently introduces metamodeling uncertainty into the UMDO procedure. The optimal solutions may deviate from the true results or even become infeasible if the metamodeling uncertainty is neglected. However, it is difficult to quantify and propagate the metamodeling uncertainty, especially in the UMDO process with feedback-coupled systems since the interdisciplinary consistency needs to be satisfied. In this paper, a new approach is proposed to solve the UMDO problem for the feedback-coupled systems under both parametric and metamodeling uncertainties. This approach adopts the decoupled formulation and it applies the Kriging technique to quantify the metamodeling uncertainty. The polynomial chaos expansion (PCE) technique is applied to propagate the two types of uncertainties and represent the interdisciplinary consistency constraints. In the optimization approach, the proposed method uses the iterative construction of PCE models for response means and variances to satisfy the multidisciplinary consistency at the optimal solution. The proposed approach is verified by a mathematical example and applied to the fire satellite design. The results demonstrate the proposed approach can solve the UMDO problem for coupled systems accurately and efficiently.