Search

Two case studies are presented, where Proportional Hazards and Poisson-type models are applied to predict failure probabilities in deteriorating water pipes. It is shown that breaks follow a nonhomogeneous Markov Process during the early stages of deterioration, that is, the failure probability is a function of time and also depends, among other things, on the number of previous breaks. During the later stages of deterioration a constant, but highly varying among pipes, break rate characterizes the failure pattern. In this latter case breaks could be reasonably approximated as Poisson arrivals. The results help in identifying factors that contribute to increased break rates and provide a high level of detail necessary for understanding failure pattern variability at the individual pipe level. The proposed methodologies are expected to provide a useful quantitative tool for improving maintenance strategies.