Reliability-based global optimization using response surface

Abstract

Reliability-based design optimization (RBDO) is one of design optimization techniques considering uncertainty and has gained attention in the engineering design. To reduce the number of function evaluations in the RBDO, many studies using response surface have been conducted. It has been recognized that sampling strategy plays an important role for highly accurate reliability analysis as well as the reduction of function evaluations. In this paper, reliability-based global optimization with a novel sampling strategy is proposed. The proposed sampling strategy consists of two parts. In the first sampling strategy, several sampling points are added around unexplored region in order to achieve the uniform distribution of the sampling points. Second sampling strategy adds a new sampling point at one of Most Probable Target Points (MPTPs) for improving the local accuracy, in which the new sampling point is selected by the distribution of sampling points. In addition, a deterministic formulation of the RBDO is conducted to search global optimal solution efficiently. Through numerical examples, the validity of the proposed method is examined.