Assessment of Remaining Useful Life of Pipelines Using Different Artificial Neural Networks Models

Abstract

Water distribution networks have a significant effect on public health and safety. Recent reports state that the 21st century is estimated to be the end of effective life for most water distribution networks in the United States. It is essential to implement accurate and cost-effective models that can predict deterioration rates along with estimates of remaining useful life (RUL) of the pipelines, to perform necessary intervention plans that can prevent disastrous failures. This study presents a computational model that predicts the RUL of water pipelines using an artificial neural network (ANN) model that has been developed using the Levenberg-Marquardt backpropagation algorithm. The model is implemented, tested, and trained using data collected from the city of Montreal. Results show that pipeline age, condition, length, diameter, material, and breakage rate are the most important factors in the prediction of RUL. Because the model shows robustness and accuracy in estimating the RUL of water pipelines in the case study, it can be used to support the municipality of Montréal, Quebec, Canada, in future planning.