Investigation On Response Surface Methods With High Efficiency For Structural Reliability Analysis

This dissertation focused on the improvement of efficiency of response surface methods, including the cooperative response surface method (CRSM) and the stochastic response surface method (SRSM) for structural reliability analysis. The SRSM based on stepwise regression analysis and the CRSM based on optimal probabilistic collocation points are presented in this dissertation.The main works and contribution are listed as follow:1. In order to circumvent the disadvantage of the SRSM with great numbers of undetermined coefficients for multi-variate complex problems, the improved stochastic response surface method is developed based on stepwise regression analysis. The number of terms and undetermined coefficients in the expansion of the nodal displacement decrease in the improved SRSM.2. Based on techniques of spatial average discretization and Nataf transformation, the SRSM was further investigaed for reliability analysis of structures with non-Gaussian random fields in this paper.3. Based on the techniques of Nataf transformation and determination of optimal probabilistic collocation points, the CRSM was further investigaed for reliability analysis of structures with non-Gaussian random variables. Examples show that the presented methods are of high accuracy, global applicability and fast convergence for structural reliability analysis involving correlated non-Gaussian random variables.