Approximated Evaluation of Failure Probability for Structural System Using Conditional Safety Indices.

Abstract

This paper deals with the approximated evaluation method for failure probability of structural systems using conditional safety indices. In general, it is difficult to compute an exact solution of the structural reliability problem with correlations of design variables and limit state functions. It is expected that the conditional safety index proposed will be a useful measure for the safety of complicated structural systems. The merit of this new measure is that the characteristics of statistics and geometry are identical on standardized space. Therefore the conditional safety index yields a practical effective interval of system reliability. When the high-order product sets of a failure event is obtained, the new assessment of the bounds for the estimation of failure probability is also significant. However, the evaluation process of conditional safety indices requires a sequential conditional correlation matrix so that it is the complicated process to calculate high-order simultaneous probability. This paper also proposes the method for transforming the vector of original safety indices into the vector of conditional safety indices using cholesky decomposition of the correlation matrix. In addition, it is suggested that the multi-integral can be approximated by the product of simple integrals using conditional safety indices.