General strategy for structural systems reliability analysis

Abstract

A general strategy for estimating the failure probability of the class of structural systems for which each deterministic structural analysis is expensive is proposed. The strategy is interactive, sequential, and iterative. It relies upon the user to interpret and generate information about the system through the use of appropriate techniques such as directed experimental design, sampling techniques, response surface, and first- and second-order reliability methods. In the strategy, one tries to obtain increasingly accurate representation of the more important failure regions and use these representations to arrive at an estimate of the system failure probability. The goal is to obtain an estimate with the minimum number of structural analyses. There are no concrete steps in the strategy. The steps taken for each problem are decided by the user. The strategy was applied to two examples. The purpose of the first example, which is a simple truss structure with brittle components subjected to a static load, is to demonstrate the strategy in greater detail. In the second example, a first estimate of the probability of failure of an offshore structure was obtained using the strategy.