Sampling-Based RBDO Using the Dynamic Kriging (D-Kriging) Method and Stochastic Sensitivity Analysis

This study presents how to carry out RBDO when surrogate models are used to represent true performance functions. The Dynamic Kriging (D-Kriging) method is used to generate surrogate models, and stochastic sensitivity analysis is introduced to compute sensitivities of probabilistic constraints with respect to the design variables, which are the mean values of the input independent or correlated random variables. To apply D-Kriging and stochastic sensitivity analysis for the sampling-based RBDO, which requires Monte Carlo simulation (MCS) to evaluate probabilistic constraints and sensitivities, this paper proposes new efficient strategies such as a local window for surrogate model generation, sample reuse, filtering of constraints, and an adaptive initial point for pattern search. Since the D-Kriging can accurately approximate true responses and there is no approximation in the estimation of probabilities, the sampling-based RBDO can yield very accurate optimum design. In addition, newly proposed strategies help find the optimum design very efficiently. Numerical examples verify that the proposed sampling-based RBDO can find the optimum design more accurately and efficiently than existing methods.