Investigating the Causes of Deterioration in Polyethylene Pipes in the Mashhad Urban Water Distribution Network and Predicting Pipe Failure Rates Using Machine Learning

Jafarzadeh, Mohammad Reza; Zohorian, Soheil

doi:10.22048/wss.2026.587891.1031

Investigating the Causes of Deterioration in Polyethylene Pipes in the Mashhad Urban Water Distribution Network and Predicting Pipe Failure Rates Using Machine Learning

Document Type : Research Article

Authors

Mohammad Reza Jafarzadeh

Soheil Zohorian

Dept. of Civil Engineering, Ferdowsi University of Mashhad, Iran

10.22048/wss.2026.587891.1031

Abstract

Urban water distribution networks are among the most critical infrastructures of society, and pipe failures pose a major challenge to the sustainability and reliability of water supply services. Polyethylene pipes are widely used because of their corrosion resistance, lightweight structure, and flexibility; however, environmental and operational factors can accelerate their deterioration and increase failures. This study investigates the factors affecting polyethylene pipe deterioration in the Mashhad water distribution network and predicts pipe failures using machine learning algorithms. Real-world data, including pipe age, diameter, length, installation depth, hydraulic pressure, and historical failure records, were collected from the Mashhad water distribution network. Environmental variables, including surface traffic and temperature fluctuations, were incorporated into the dataset. Four machine learning algorithms—Random Forest (RF), Light Gradient Boosting (LightGBM), Artificial Neural Network (ANN), and Support Vector Classifier (SVC)—were evaluated. The results showed that all models achieved high predictive performance. Among them, LightGBM produced the best results, with an accuracy of 96.1%, a recall of 95.7%, and an area under the ROC curve (AUC) of 0.980. The RF model also performed well, achieving an accuracy of 94.3%, a recall of 92.9%, and an AUC of 0.971, while the ANN and SVC models achieved accuracies of 88.4% and 89.7%, respectively. Feature importance analysis identified pipe age, temperature fluctuations, hydraulic pressure, and pipe diameter as the most influential factors affecting pipe failures. These findings provide a reliable framework for identifying high-risk pipes and supporting proactive maintenance strategies to improve urban water network reliability.

Keywords

Polyethylene pipes

Mashhad urban water distribution network

Random Forest

Light Gradient Boosting Machine

Subjects