Urban Water Distribution Network Failure Prediction using Artificial Intelligence

Abstract

In this study Artificial Neural Network (ANN) and Adaptive Neuro Fuzzy Inference Systems (ANFIS) have been used to predict the failure trend of pipe network and to access the present condition of water distribution system. To predict the number of failures in pipelines different methods can be used. Data driven modeling is the most recent method adopted in different fields for prediction where historical data are available. Soft computing techniques like ANN and ANFIS have the potential of exploiting the tolerance for imprecision, uncertainty and partial truth to achieve tractability, robustness, low solution cost and better support with reality. This approach is tested and verified in a real life water distribution system in Trivandrum city. The case study demonstrates the entire process from data aggregation to model development. The work included the use of two ANN networks namely cascade and feed forward back propagation network for the prediction of water pipelines failure. Seven indicators were identified as input parameters in the water main failure. They include age of pipe, number of previous failures, pipe length, diameter, thickness, material and demand. The performance of the models is evaluated by using coefficient of correlation and mean absolute error. The study reveals that the predicted results will help the authorities to take decision regarding the repair and replacement of pipes in the distribution system.