Modeling and Physics-Enhanced Fault Detection in Wastewater Pump Stations

Monitoring wastewater pump stations is essential because they are critical infrastructure. However, monitoring is still often performed manually due to the lack of suitable algorithmic methods and data. This paper introduces a high-fidelity, physics-enhanced simulator of a three-pump wastewater station that captures transient hydro-mechanical dynamics at a one-second resolution. The simulator is fully parameter-driven, adaptable to other wastewater stations, and capable of generating datasets for data-driven analytics. It can also generate balanced faulty datasets when real failures are scarce or confidential. A comparison with high-frequency SCADA data from a municipal station shows strong agreement across key operational metrics. Furthermore, the paper proposes robust statistical and mathematical frameworks for fault detection and isolation, including a nested-model F-test to detect pump degradation or system faults, and a tangent residual approach to distinguish pump faults from system faults using operating-point kinematics. This framework enables what-if studies, facilitates early fault diagnosis based on flow rate and head, and provides actionable insights for condition-based maintenance in wastewater pumping infrastructure.