Predicting water pipe breaks using neural network

The relationships between pipe breaks of service pipes and mains and several factors were examined. Historical pipe breaks, and water and soil temperatures were also modeled by an artificial neural network to predict pipe breaks for efficient management and maintenance of the pipe networks. It was observed that the breaks of pipes increased after the temperatures of water and soil crossed in spring and fall. The pipe breaks were closely related to water and soil temperature, especially mains were affected more than service pipes. The fittings and valves were susceptible to the temperatures and needed to take measures for preventing breaks. The prediction of the pipe breaks by the ANN model built had a good performance except that the sensitivity was not good when the pipe breaks rapidly increased or decreased. The ANN model gave a good performance and was to be useful to predict the patterns of pipe breaks on a seasonal basis.