About our Sessions

Aging water infrastructure poses critical challenges for municipalities, leading to escalating maintenance costs, service disruptions, and public health risks. To address this, the City of Sugar Land has developed a pioneering machine learning-based approach to enhance the predictive maintenance of water mains by leveraging historical break data, pipe characteristics, and environmental factors. This project integrates advanced models such as Decision Trees, Random Forest, and XGBoost, chosen for their ability to handle complex failure patterns and assign risk scores to assets with high interpretability. Traditional linear regression models often assume constant failure rates, limiting predictive accuracy. To overcome this, we employ a dynamic degradation function using the Weibull Reliability model, which estimates conditional failure probabilities based on evolving infrastructure conditions. Additionally, we introduce the Hydraulic Impact Factor (HIF), incorporating pressure, velocity, headloss, and water quality to refine risk assessments and optimize yearly benefit-cost prioritization for data-driven asset management. The model is trained using historical data exported from CityWorks, with planned integration for real-time risk assessment and proactive maintenance scheduling. This methodology enhances prediction accuracy, reduces operational costs, and minimizes service disruptions. By embedding machine learning within the Integrated Asset Management System (IAMS), we establish a new benchmark in water infrastructure management. Our findings will showcase how cities can revolutionize water main reliability assessments using AI, setting a precedent for smarter, more resilient municipal infrastructure planning.