Implementing Network Science to enhance water distribution network pipe roughness calibration

<p dir="ltr">This study presents a methodology for improving pipe roughness calibration in water distribution networks (WDNs) by integrating hydraulic and network science metrics. Initial calibration was based on conventional grouping by pipe diameter and DMA location. To enhance this approach, network clustering using the k-means algorithm was applied, utilising a combination of numerical and categorical attributes, including graph-theoretic measures and flow statistics. The clustering performance and attribute contribution were assessed using clustering efficiency scores and the Attribute Contribution Score (ACS). Results indicate that the inclusion of graph-based metrics improved model calibration, particularly in reproducing pressure trends during peak demand periods. The proposed heuristic demonstrates that combining hydraulic and topological attributes in a simple clustering framework can support more accurate and insightful roughness calibration in WDNs.</p><p dir="ltr">This paper was presented at the 21st Computing and Control in the Water Industry Conference (CCWI 2025) at the University of Sheffield (1st - 3rd September 2025).</p>