Analysis of drainage system defects using co-occurrence patterns

Abstract

Urban drainage systems face unprecedented challenges from ageing infrastructure, environmental stressors, and increasing service demands. Traditional condition assessment approaches aggregate diverse defect patterns into single condition scores, obscuring the underlying deterioration mechanisms that drive failure processes. This study develops a network science framework to identify systematic defect co-occurrence patterns using CCTV inspection data from Hong Kong’s drainage network. Constructing weighted defect co-occurrence networks and applying Louvain clustering algorithms, we identified four distinct deterioration mechanisms: Lining-Deformation Co-occurrence Pattern (18.4% prevalence), Structural-Hydraulic Co-occurrence Pattern (54.5%), Root-Joint Co-occurrence Pattern (22.3%), and Connection-Sediment Co-occurrence Pattern (4.9%). Multi-method validation using five alternative weighting schemes demonstrated robust clustering consistency (Adjusted Rand Index = 0.438–1.000), while statistical analysis revealed significant associations between mechanisms and infrastructure characteristics (diameter, age, material) and contextual factors (district, land use, traffic intensity). These findings support the development of mechanism-informed management strategies, enabling utilities to move from purely reactive condition-based approaches toward targeted interventions informed by systematic defect association patterns. While cross-sectional analysis cannot establish causal relationships, the observed co-occurrence patterns provide actionable intelligence for risk-based asset management.