Empirical Assessment of Unobserved Heterogeneity and Polyvinyl Chloride Pipe Failures in Water Distribution Systems

Abstract

An understanding of the failure patterns of pipes in water distribution systems is critical to cost-effective system-maintenance planning. Failure patterns, which typically measure the frequency of water main breaks in a water distribution system, can vary widely depending on the type of pipe material being considered, and statistical analysis of pipe frequency-of-failure data is complicated by limited data on soil conditions, freeze-thaw cycles, construction quality, and so on, which manifests itself as unobserved heterogeneity. The current paper considers failure frequencies in polyvinyl chloride (PVC) pipes using pipe-break data from a 21-year period in Indianapolis, Indiana. Failure frequencies were studied using a random parameters negative binomial (and a latent class negative binomial) to account for possible unobserved heterogeneity in the data and to assess the systemwide monthly frequency of PVC pipe breaks as a function of a number of observable explanatory variables. The results indicate the importance of pipe diameter, length, and age, and environmental conditions, in estimating monthly break frequency. This paper provides an important demonstration of method (the application of heterogeneity models), and the model estimation results provide insights into pipe-failure frequencies that can be used as guidance for both maintenance-crew allocations and targeted inspection of PVC pipes. Future databases, which may include detailed information on soil conditions, freeze-thaw cycles, construction quality, and other factors, can provide additional heterogeneity-related insights using the methods proposed in this paper.