Anticipatory Detection and Characterization of Urban Stormwater System Flood Risk: A Model-Based Fault Diagnosis Approach

Degraded components in an urban stormwater system increase the risk of flood but also measurably influence the downstream flow of water. Installation of flow sensors within the system has been shown to be useful for detection of fault formation, but due to the complex dynamics of stormwater flow, a methodology for using flow measurements to further diagnose the fault type, location within the system, and its severity is undeveloped. This research explores the interpretability of the fault signals in flow measurements for detection and diagnosis under increasing fault severity and sensor noise for the purpose of assessing the potential flood risk and develops a framework for practical implementation. A numerical model of a stormwater system built in US Environmental Protection Agency’s Storm Water Management Model (EPA SWMM) serves as a synthetic testbed for detection and diagnosis experiments. Detection and diagnosis were found to be accurate at low noise levels and for moderate to severe faults, pointing to the potential for stormwater sensing to augment current maintenance approaches and reduce flood risk.