Reliability analysis of discrete-state performance functions via adaptive sequential sampling with detection of failure surfaces

Abstract

Reliability analysis of discrete-state performance functions via adaptive sequential sampling with detection of failure surfaces Miroslav Vořechovský• accurate estimation of failure probability with error prediction, resistance to multiple design points, and strong nonlinearities • the method needs at least a categorical limit state function (including no response output) for continuously distributed inputs • adaptive sequential extension of experimental design using the ψ criterion, minimizing the number of limit state evaluations • the novel criterion automatically balances between exploration and exploitation in the input space • efficient importance sampling in important rings, excluding the safe ball and unimportant exterior • global sensitivity indices quantifying the importance of individual variables to rare event probability • adaptive expansion and refinement near the failure surface avoiding clusters of points (no redundant information)