Contaminant source identification in water distribution networks: A Bayesian framework

Abstract

This work presents a Bayesian model updating approach for handling contaminant source characterization problems in the context of water distribution networks. The problem is formulated in a Bayesian model class selection framework where each model class represents a possible contaminant event. The parameters of each model class characterize the contaminant mass inflow over time in terms of its intensity and starting time. The class with the highest posterior probability is interpreted as the most plausible location for the contaminant injection. The evidences of the model classes are estimated using the transitional Markov chain Monte Carlo (TMCMC) method. The approach provides additional insight into the current network state in terms of posterior samples of the parameters that describe the contaminant event. The effectiveness of the proposed identification framework is illustrated by applying the contaminant source detection approach to a couple of water distribution systems.