Comparative Investigations of Modified Bivariate Dimension Methods for Statistical Moments Assessment of response

Abstract

Evaluating the statistical moments of performance functions which aims at keeping the tradeoff of accuracy and efficiency is still a challenge in structure reliability analysis. This paper proposes several modified bivariate dimension reduction methods (BDRM), based on two-dimensional and one-dimensional Gauss-Hermite quadrature. Compared to the original BDRM, evaluating central moments by these modified BDRMs requires less computational effort. One numerical example is investigated, which demonstrate that two of the modified BDRMs can achieve good balance between accuracy and efficiency for statistical moment assessment.