Bayesian network to predict failure of Sydney Water's critical water mains

As water mains age, their potential for failure increases which can result in the loss or reduction of water to customers. In addition, the flow of water from a main break can cause damage to other infrastructure and impact on community and social processes. The management of water supply assets requires that mangers make decision under significant uncertainty. Decisions are required for planning of investment, replacement and rehabilitation strategies. Burst behaviour is difficult to model due to the many uncertain factors associated with failure including soil condition, pipe material, age and water pressure. Sydney Water is currently moving to a risk based approach whereby a quantitative assessment of likelihood and consequence are made on individual critical water mains to assess the overall risk and to inform replacement and renewal strategies. This paper uses a Bayesian network to assess the probability of failure of Sydney Water's critical water mains. Validation of the model used actual failure history from 1998-2006