Bayesian approaches to lifetime prediction

In this paper a framework to predict the remaining lifetime for existing waterways infrastructures based on stochastic modeling of deterioration processes and Bayesian analysis is presented. The application of the Bayes’ theorem is motivated by the availability of expert knowledge as well as the collection of both qualitative and quantitative data from the structure. An original method is proposed to derive the prior statistical parameters of the gamma distribution describing the stochastic deterioration process, based on the assumption that the lifetime distribution can be approximated by the Birnbaum-Saunders statistical model. An appropriate Bayesian Network is finally implemented to improve the classification of the structure with respect to its proneness to damage. The outcome of the research work is to assist the owners of large infrastructural network in planning and prioritizing maintenance interventions.