Institute of Communications and Connected Systems


Leakage detection in water networks

 Clear white background. Artists impression of water bursting from a steel pipe. Encircled on the right hand side of the image is a computer screen with red and blue binary numbers floating in lines in front of a scientific calculation.

15 March 2021

Integrating model-based and machine learning techniques to detect and isolate leakages in water distribution networks

Funder N/A
Amount N/A


Research topics Fault detection| Utilities | Water Networks | Machine Learning


Water is a fundamental and increasingly scarce resource for life. Water transmission pipelines lose an average of 20% to 30% of the water transmitted through them, up to above 50% in old systems with poor maintenance. Therefore, there is a need to prevent the threat of leaks and minimize their damages through extensive research in leak detection technology.
A major challenge for leak detection is the lack of measurements, as sensors are placed sparsely in the network and meter readings are available only in some cases. The project will investigate if machine learning can be used to complement sparse sensor data to detect the presence and localise leaks in a water distribution network (WDN).
The project will include: exploration of the available data and topology; design of a novel detection technique; testing the detection method on known leakage events; analysis of the results and possible comparisons.


View Principal Investigator's Publications