HWM Global nominated for Water Industry Award

Along with our partners at Hydraulic Analysis Group and Thames Water, we are delighted to be nominated for the Leakage Initiative of the Year award at the upcoming Water Industry Awards!

The nomination recognises a collaboration where advanced machine learning is used to identify leaks or classify other sources of noise from audio files collected by PermaNet SU devices.

Following Thames Water’s open source data release, Hydraulic Analysis Group and HWM Global worked together to develop a Machine Learning (ML) solution to enhance leak detection performance.  The objectives were to reduce false alarms, increase confidence in genuine leak alerts, and improve the identification of detached loggers.

The initiative was delivered through the development and deployment of an audio multi classification ML model.  Using an ensemble learning approach, the solution combined multiple audio feature extraction methods with spectrogram-based classification to improve detection accuracy while minimising false positives.  The model was trained on verified leak data alongside twelve non leak profiles, followed by validation during live testing.

Over the past 12 months, the ML solution has been embedded into Thames Water’s leakage operations and has delivered strong, consistently tracked benefits.  Ongoing refinements on reducing false positives near sources of electrical noise and pressure reducing valves will increase the above benefits.  In addition, a secondary use case is progressing to further enhance detached logger identification and site visit efficiency.

This technology has since been embedded into a further number of water companies who are now also benefitting from enhanced leak detection performance through Machine Learning.