Predictive Maintenance Model for Sewer Networks Based on Cross-Correlation Analysis of Rainfall and Flow Data: A Case Study of Palermo's Sewer System

<p dir="ltr">This paper presents a predictive maintenance model for urban sewer networks, demonstrated through a case study of the Acqua dei Corsari treatment plant in Palermo. The methodology integrates hydrometric level data from one monitoring station with rainfall data from eight surrounding weather stations. The experimental results reveal characteristically low Spearman correlation coefficients, with relatively higher values for urbanized area stations compared to geographically closer locations. This pattern validates that sewer infrastructure connectivity dominates over natural watershed delineation in determining rainfall-runoff relationships. The observed lag times exhibit significant variability, indicating unreliable temporal correlation patterns. Analysis of baseflow conditions demonstrates temporal stability without systematic deterioration trends, establishing robust reference conditions for anomaly detection. Given correlation-based limitations, the methodology emphasizes hydrometric-only indicators, particularly recession time analysis, recession ratio, and peak level variations. The low correlations and variable lag times confirm the inadequacy of traditional watershed-based approaches, highlighting the necessity of focusing on infrastructure-dependent hydraulic performance indicators. Results indicate that for systems where behaviour is determined by hydraulic infrastructure and operational management, hydrometric-only indicators provide more reliable diagnostic capabilities than cross-correlation metrics. The framework provides a practical approach for predictive maintenance in complex urbanized catchments where detailed network mapping is unavailable.</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>