Reliability-Based Design Optimization Using Confidence-Based Model Validation for Insufficient Experimental Data

Abstract

The conventional reliability-based design optimization (RBDO) methods assume that a simulation model is able to represent the real physics accurately. However, this assumption may not always hold as the simulation model could be biased. Accordingly, designed product based on the conventional RBDO optimum may either not satisfy the target reliability or be overly conservative design. Therefore, simulation model validation using output experimental data, which corrects model bias, should be integrated in the RBDO process. With particular focus on RBDO, the model validation needs to account for the uncertainty induced by insufficient experimental data as well as the inherent variability of the products. In this paper, a confidence-based model validation method that captures the variability and the uncertainty, and that corrects model bias at a user-specified target confidence level, has been developed. The developed model validation helps RBDO to obtain a conservative RBDO optimum design at the target confidence level. The RBDO with model validation may have a convergence issue because the feasible domain changes as the design moves (i.e., a moving-target problem). To resolve the issue, a practical optimization procedure is proposed. Furthermore, the efficiency is achieved by carrying out deterministic design optimization (DDO) and RBDO without model validation, followed by RBDO with confidence-based model validation. Finally, we demonstrate that the proposed RBDO approach can achieve a conservative and practical optimum design given a limited number of experimental data.