Analytics-driven approaches supporting asset management of sanitary sewer networks

Abstract: Sewer blockages can cause overflows and flooding, with consequences such as damage to property and environmental pollution, risks to public health and economic loss. Despite the causes being understood, blockages in sewer networks may occur unpredictably. The responsible staff for sewer networks at water utilities need to efficiently determine the most effective action (what), the specific network location needing attention (where), the optimal timing for intervention (when), and the appropriate remedial task (how), especially given the unpredictability of blockages. Today a reactive approach to asset management and maintenance is often adopted. Additionally, data availability, quality and interoperability between systems are not always at levels that can support decided objectives, proactive maintenance planning and asset management of pipe networks. Thus, the aim of this thesis is to propose and evaluate approaches that can support analytics-driven maintenance planning and asset management for sewer networks. These approaches aim to contribute to mitigating the impact of siloed data structures and enhance the understanding of blockage root causes from a spatial perspective.In this thesis, the challenges of data management in the asset management of pipe networks were investigated through focus group workshops and questionnaire surveys. A conceptual framework was developed based on findings from focus group workshops and surveys. The framework combines data quality assessments, interoperability evaluations between asset management tools, data collection, and informational benefits analysis. This framework aimed to identify the presence of data silos and plausible pathways towards more data-driven data management strategies. A performance assessment combining performance indicators associated with blockages and partial least squares regression (PLS) was conducted to draw inferences that could be useful at a strategic level. Furthermore, a spatial heterogeneity assessment of blockages and factors affecting blockages was carried out. This approach combined network kernel density estimation (NKDE), network k-function, and geographically weighted Poisson regression (GWPR). Lastly, a vulnerability assessment was carried out that combined topological analysis using edge-based centrality measures and network cross-k-function. These approaches were applied to three sewer networks.The focus group workshops and questionnaire surveys identified several challenges affecting data management in the context of pipe network asset management. Many of the challenges could be ascribed to issues related to data quality and interoperability. Results from the preliminary application of the conceptual framework showed how it could be applied for identifying data silos and pathways to data-driven decision-making towards proactive management blockages in sewers. The observed spatial trends and patterns from network k-function analysis and network kernel density estimation showed spatial variability in the occurrence of blockages (single occurring and recurring). Geographically-weighted Poisson regression analysis showed spatial heterogeneity in factors influencing blockage propensity. The network cross-k-function analysis demonstrated that pipes with historical blockage incidents tend to be clustered around critical pipes with higher centrality values. These results could support vulnerability assessments in sewer networks and the development of targeted maintenance strategies. These approaches together could aid data-informed maintenance planning and asset management at the strategic, tactical and operational levels.