DATA-DRIVEN MODEL FOR PREDICTING FAILURE RATE OF UNDERGROUND POTABLE WATER DISTRIBUTION SYSTEM

The failures of underground water pipe networks create severe consequences for water network operators. Thus, forecasting pipeline failure rates is crucial to ensure a reliable, sustainable water supply system and cost-effective network maintenance and replacement plans. Thus, this paper proposes an innovative approach based on artificial neural network (ANN) for predicting underground potable water pipe failure rates. First, the factors affecting the pipe failure are identified from the literature and validated by subject-matter experts. Second, ANN is developed to predict failure rates. The model is further optimized to determine the most effective sequence and choose the optimal subset of input variables. The study findings reveal that the ANN model is highly efficient in predicting the failure rate of underground water pipes with high accuracy. The developed model can assist decision makers to prepare long-term corporate pipe maintenance and replacement plans.