Probabilistic identification of failures in water distribution networks using Bayesian belief networks

Pipes in water distribution networks (WDNs) are subject to various risks of failure. These failures can be based on structural, consequential, hydraulic, water quality and human error factors. One way to estimate the overall failure risk is to use Bayesian Belief Networks (BBN) that combine the individual probabilities of different risk factors. In this research, we build a BBN for identifying failure risk based on literature review and supported with engineering judgement. The proposed model is applicable to any kind of WDN regardless of structural elements (e.g. pipe materials) or environmental characteristics (e.g. local climate). We calculate the failure risk for each pipe segment in a real-world WDN by running our BBN model individually. Then we create a failure risk map that indicates high-risk and sensible areas, offering effective decision-support to different stakeholders for planning Maintenance, Rehabilitation and Replacement (MRR).

 Click here to watch recorded presentation on YouTube.