Improved Adaptive Response Surface Method for Reliability Analysis of Structures

Abstract

An adaptive response surface method for improved reliability estimate of structures is explored in the present study. The approach primarily follows the iterative algorithm of obtaining improved converged centre point for the design of experiment (DOE) in which the bounds of random variables during iterations are set with a multiplier of the standard deviation of the corresponding random variables to obtain new DOE. However, there is no specific basis for such selection. An attempt has been made here to propose a strategy to select the bound for new DOE during each iteration. The boundary nearer to the new centre point is kept the same as the physical boundary for each random variable. The other boundary is set in such a way that both the upper limit and the lower limit of the corresponding random variable are equidistance from the new centre point of DOE. Further, a zone-based metamodelling strategy is explored to minimize such inaccuracy for approximating limit state function. The improved performance of the proposed approach with regard to the accuracy and efficiency of the proposed algorithm is demonstrated by considering two example problems.