A Comparison Study of Water Pipe Failure Prediction Models Using Weibull Distribution and Binary Logistic Regression

Abstract

The aging water infrastructure is one ofthe main causes of the increasing water pipe breaks nationwide, resulting in significant social and economic costs. To sustain water infrastructure assets in good condition, it requires a proactive decision making process for capital improvement project planning, maintenance, rehabilitation and replacement. Because physical inspections involve extensive labor and financial resources, it is not always a feasible option to cover entire water pipe inspection and condition assessment. This paper presents two statistical water pipe failure prediction models developed forwater system managers to assisttheir long-term decision making process using Weibull distribution and binary logistic regression. The models presented are developed based on water pipe failure data from approximately 100 years old cast iron pipesfroma municipality in the Midwest, United States.