An application of genetic algorithms to reliability analysis in rare events

Abstract

An application of Genetic Algorithms (GAs) to reliability analysis in rare events is presented. The application is directed towards the enhancement of the computational capability in the analysis. As an exemplified application, GAs are utilized for searching the point of maximum likelihood of failure probability, referred to as the design point. The design point is then used as the center of the sampling density in Monte Carlo Simulation (MCS) by which the computation of low probability becomes efficient. The most advantageous aspect of GAs is that the explicitness of limit-state-function in terms of basic random variables is not required in their search operation. The determination of the design point is thus made possible in the problems involved with complex systems where implicit limit-state-functions are naturally encountered. Consequently, the reliability analysis of broader classes of systems with the interest in rare events can be realized and efficiently accomplished by such an application of GAs.