New collocation method for stochastic response surface reliability analyses

Abstract

The stochastic response surface method (SRSM) is widely used in engineering reliability analyses due to its efficiency and accuracy. The selection of collocation points in the SRSM has great significance, as it may strongly affect the computed results. This paper investigates the performance of different selection strategies in SRSM, and proposes a new collocation method. First, two commonly used collocation methods—the regression-based collocation method and the linearly independent collocation method—are briefly reviewed; and their limitations in application to reliability analysis are discussed. Then, an improved collocation method that achieves a better tradeoff between efficiency and accuracy is proposed. Four examples are employed to test the performance of the proposed collocation method; and a comparative study is conducted to demonstrate its advantages with respect to some other existing collocation methods.