An Innovative Geostatistical Framework for Prioritizing Inspections in Water Distribution Networks

Aging water infrastructure networks face significant challenges, with the United States experiencing frequent pipe failures and substantial water loss. This study leverages geospatial and geo-statistical analysis to prioritize pipeline inspection in Watertown, Connecticut. By integrating Geographic Information Systems (GIS) with machine learning, various factors influencing pipeline health—such as pipe age, material, length, diameter, soil pH, and groundwater depth— are analyzed. Historical pipe break data between 2015 and 2021 are mapped to identify spatial correlations with environmental and operational attributes. A K-Means clustering algorithm classifies pipelines into high, medium, and low inspection priority categories based on critical attributes. Findings reveal that corrosive soil pipes with low diameters and high break frequencies pose significant risks. The study demonstrates the effectiveness of GIS-driven data integration and analysis in optimizing maintenance strategies, reducing system-wide impacts, and improving resilience. These insights provide a replicable framework for water utilities worldwide to address aging infrastructure challenges. Keywords: water distribution network, pipe inspection, risk prioritization, geospatial statistics, unsupervised learning