Propagation of uncertainties in problems of structural reliability

Abstract The problem of controlling a variable Y such that the probability of its exceeding a specified design limit L is very small, is treated. This variable is related to a set of random variables Xi by means of a known function Y = ƒ(X i ) . The following approximate methods are considered for estimating the propagation of error in the Xi's through the function ƒ(·) : linearization; method of moments; Monte Carlo methods; numerical integration. Response surface and associated design of experiments problems as well as statistical inference problems are discussed.