Applying Survival Analysis to Pipeline Data: Gaps and Challenges

摘要

A major challenge in water pipe infrastructure asset management is to determine the deterioration process for water pipes. The deterioration processes are usually modeled with a decay curve, which describes how pipe performance changes with time. Decay curves can also be used to predict the future performance of a water pipe. The methodology used to develop decay curves should be interpretable, reproducible, and validated. Many studies have been conducted to apply survival analysis to pipeline data, but most of the existing studies ignore the left-truncation problems in pipe break history records. Instead, in the modelling process they assume the first recorded failure to be the real first failure. This assumption leads to incorrect inferences and systematic biases by ignoring missing truncated failure records (i.e., breaks and leaks). In this paper, we summarize the identified major gaps and challenges when applying survival analysis to pipeline data by examining the published literature, practices, and the water pipeline break history data we collected from utilities across the U.S.