Bentley Fellow Bentley Systems Inc., United States
Abstract Submission: Anomaly Leak Finder (ALF) has been developed as a high-fidelity Digital Twin (DT) solution in collaboration with PUB, Singapore's National Water Agency, to significantly enhance the leak detection and localization across 3 of Singapore’s supply zones. These networks comprise over 1,000 km of underground pipelines, are continuously monitored by a sophisticated array of 95 smart sensors. ALF leverages near real-time hydraulic time-series data with combination of data-driven prediction (DDP) and physics-based simulation (PBS) models, offers an innovative means of reducing non-revenue water (NRW) losses by detecting and localizing hidden leaks before they escalate into major disruptions. ALF’s cloud-agnostic architecture enables seamless deployment across various cloud providers, ensuring flexibility and scalability (e.g.: Azure, AWS, GCP). The ALF architecture is composed of a robust backend API and an intuitive frontend UI. The backend, designed as a microservices architecture, comprises a suite of near real-time analytical capabilities, handling core functions such as (1) data retrieval and processing, (2) predictive modeling (hybrid extended Kalman filter and deep neural network), (3) anomaly detection and (4) anomaly localization. These containerized functions ensure scalability across various cloud platforms. The frontend web application provides real-time visualization of sensor data, detected anomalies and localized events with six core modules including dashboard, event, sensor, report, health monitoring and administration. This paper highlights ALF’s scalable architecture, demonstrating how it empowers users to make proactive decisions that effectively reduce NRW losses and ensuring the long-term sustainability of smart water distribution networks.
.jpg "Zheng Yi Wu photo")
