Modelling Infrastructure Deterioration Risks Using Bayesian Mixtures

Abstract

A major aspect of assessing the long-term reliability of deteriorating structures is the need to integrate the results of different inspections in time, within the models used to analyse the progress of deterioration. This article presents a framework for such an analysis. The article focuses on reinforced concrete structures subject to chloride corrosion, as this represents the single largest form of deterioration of infrastructure in Europe and North America. The analysis focuses on special extensions of the Empirical Bayes method where the nonobservable parameter is a discrete random variable with a relatively small number of outcomes. Various likelihood functions are derived. They are based on mixtures of “deterioration scenarios.” The authors show how the method can be used to calibrate the response of a stochastic deterioration model, and to update a time-dependent structural reliability analysis. An example relating to long-term chloride corrosion in a reinforced concrete bridge slab is presented.