By: Kanishke Noel, senior director, intelligent networks with Mueller
Digital transformation is increasingly central to how water utilities manage assets, control costs and maintain service reliability. In Singapore, PUB, the National Water Agency, oversees a potable water network spanning 5,500km and serving more than 1.5 million customers. Proactive leak detection and asset maintenance have thus become critical to strengthening network resilience while minimising operational disruption and unnecessary capital expenditure.
To support these objectives, PUB has deployed a new-generation Transmission Pipe Leak Monitoring (TPLM) system across 200km of large-diameter transmission mains. These critical assets transport high volumes of water under significant pressure, meaning that even minor leaks can lead to substantial water loss and costly damage if not identified early.
The system, EchoShore-TX by Echologics, leverages acoustic hydrophone sensors to detect leaks through both correlation and standalone logging methods. Each installed node — comprising an electronic module, hydrophone, and combined cellular and GPS antenna — collects acoustic data and transmits it to a cloud-based platform at regular intervals. This continuous data flow enables remote monitoring at scale, reducing the need for labour-intensive field inspections while improving visibility across the network.
A defining feature of the system is its use of physics-driven algorithms combined with advanced data processing techniques. Rather than relying solely on conventional AI, this approach integrates physical modelling to enhance the quality and reliability of insights. According to Marcin Kloc, data analyst at Echologics, “Anyone can deploy an AI model and feed it a bunch of data, but it takes real scientific skill to figure out how to use physics and math to transform the data to mine gold out of chaff.”
This integration of physics-based analytics has translated into tangible operational gains. By refining how acoustic data is interpreted, the system reduces background noise and false positives, enabling engineers to focus on high-probability leak events. At the same time, it supports more consistent scaling across deployment sites, allowing PUB to expand monitoring coverage without a proportional increase in analytical workload.
The improvements are evident in the system ‘speed to notification’, a factor in limiting water loss and avoiding escalation. Earlier iterations required four consecutive noise recordings before a leak alert could be triggered, typically resulting in a detection window of 2-3 days. With the introduction of higher-frequency acoustic sampling and spectral data analysis, the updated system has shortened this timeframe to 1-2 days, enabling faster response and reducing the likelihood of secondary damage.
Further enhancements have expanded detection capabilities beyond traditional correlation methods. The introduction of single-channel detection allows the system to identify leaks that do not propagate between sensors. In one instance, this capability enabled the detection of an air-valve leak of less than 5L/min at the sensor location — an issue that would have otherwise gone unnoticed due to the absence of correlatable data.
PUB has emphasised the importance of standardising performance evaluation. A structured set of metrics is used to assess both detection capability and classification accuracy, covering the full spectrum of outcomes from correctly identified leaks to missed events. Two key indicators — leak classification performance and leak sensitivity — are used to measure how effectively the system distinguishes real leaks from normal operational noise and how comprehensively it detects leaks within a given area.
This data-driven framework supports more informed decision-making, enabling PUB to optimise operational resources while maintaining service standards. It also provides a basis for benchmarking performance across different systems and deployments, ensuring accountability and continuous improvement.
The value of early leak detection is significant for large-diameter mains, where replacement costs are high and failures can be disruptive. By identifying and addressing issues at an early stage, utilities can defer capital expenditure, reduce non-revenue water (NRW), and avoid the costs associated with emergency repairs.
PUB’s deployment illustrates how advanced monitoring technologies, underpinned by robust analytics, can deliver measurable returns. Beyond improving operational efficiency, such systems enable utilities to adopt a more predictive approach to asset management, aligning long-term infrastructure planning with immediate performance gains.