Integrated approach for pipe failure prediction and condition scoring in water infrastructure systems

With an increasingly aging water infrastructure, decision makers are directing their attention to better ways to model and predict asset failure. Failure modeling is a discipline that addresses complex deterioration processes to better inform asset management practices. These deterioration processes and contributing factors have been addressed using a variety of models. One statistical model that has been explored in this thesis is the logistic regression model. The proposed approach consisted of developing a logistic regression model to estimate pipe-level failure probabilities in a flexible time interval. The approach further used the probabilities to estimate a Mean Time to Failure and assign pipe condition scores according to a methodology suggested by Opila and Attoh-Okine [1]. This thesis contributes with a practical and systematic methodology to capitalize on failure records and generate actionable failure probabilities and condition scores to integrate in asset prioritization strategies.