Bayesian analysis for causes of failure at a water utility

Abstract

Purpose

The purpose of this paper is to inform facility managers of the type of failure affecting certain pipe types more than others. This is useful in asset management as preventive maintenance can be undertaken for those pipe types that experience high probabilities of failure.

Design/methodology/approach

The probability of a specific pipe type failing given the cause of break, age at failure, pipe diameter, and type of soil at the location of the break was found using inventory and main break data from the Honolulu Board of Water Supply (HBWS). Bayes’ theorem was then applied to find the posterior probabilities of failure starting from the prior probabilities of failure.

Findings

It was observed that the greatest probabilities of failure involved corrosion, pipes aged between 20‐30 years, 8″ pipes, and pipes in fill material. The pipe types were ranked and scored based on their probability of failing due to break cause, age, diameter, and soil type. Cast iron pipes were shown to have the highest probability of failing. As such, attention should be given to replace segments of cast iron pipes as they reach the end of their service lives.

Practical implications

This study serves to address a major query in asset management at a public utility, that of which pipes should be selected for replacement when they reach the end of their service life. In addition, this study helps to understand the causes of failure for the various types of pipe.

Social Implications

The importance of having reliable water supply at low cost has immense social implications in modern communities. To deliver such service, water pipe assets have to be managed efficiently.

Originality/value

This paper addresses the probability of failure in a straightforward manner that the water utility can easily apply to its own data, both in its design and asset management.

Citation

Singh, A. (2011), "Bayesian analysis for causes of failure at a water utility", Built Environment Project and Asset Management, Vol. 1 No. 2, pp. 195-210. https://doi.org/10.1108/20441241111180433