Structural reliability analysis from sparse data

Abstract

Over the past several decades, major advances have been made in probabilistic methods for assessing structural reliability with a critical feature of these methods being that probability models of random variables are known precisely. However, when data are scant it is rear to identify a unique probability distribution that fits the data, a fact that introduces uncertainty into the estimation of the probability of failure since the location of the limit surface in the probability space is also uncertain. The objective of the proposed work is to realistically assess the uncertainty in probability of failure estimates of the First Order Reliability Method (FORM) resulting from the limited amount of data.