Integrating failure prediction models for water mains: Bayesian belief network based data fusion

Due to incomplete and partial information, data/information from multiple sources with different credibility or confidence, and the involvement of human (expert) judgment for the interpretation and integration of data/information, uncertainties become a major concern for the development of water main failure prediction model. To reduce these uncertainties, a new Bayesian belief network based data fusion model is developed for the failure prediction of water mains. To accredit the proposed framework, it is implemented to predict the failure of CI and DI pipes of the water distribution network of the City of Calgary. Analysis results indicate that ∼6.16% and 8.20% of the total 18,762 CI and DI pipes are at high and very high failure rates, respectively. The proposed model can be integrated with the geographic information system of the utilities and capable of identifying the most 'vulnerable' and 'sensitive' pipes within the distribution network as well as estimate the total number of pipes that need maintenance/rehabilitation/replacement (M/R/R) actions.