Conceptual Water Main Failure Risk: Self-Excitation, Pipe Age, and Statistical Modeling Performance

Abstract

Statistical water main failure models that improve our understanding of main breaks may help water utilities allocate resources more efficiently. A variety of statistical models have been developed, but few actively seek to replicate empirical main break behavior. Furthermore, the prevailing conceptual model of how failure risk changes over the lifetime of a water main, which includes self-excitation, is based on limited empirical evidence. We investigate self-excitation and pipe aging behavior using data describing a large cohort of water mains, present a statistical model that includes self-excitation, and compare the performance of several published models both with and without self-excitation. The failure data suggest that temporal clustering is occurring, which may be caused by self-excitation; however, the modeling results suggest that including self-excitation in failure models may not be worth the additional required resources. Researchers and practitioners should investigate their data and assess their specific goals and available resources to determine which modeling approach is most appropriate.