Author(s): Ignasius Axel Hutomo; Ioana Popescu; Leonardo Alfonso

Linked Author(s): Ioana Popescu

Keywords: Coustic Sensor; Artificial Neural Network; Hydraulic Model; Leakage; Water Distribution Networks

Abstract: Despite ongoing research and practical efforts, water losses in water distribution networks remain alarmingly high, impacting water quantity and quality. Traditionally, water utilities have employed separate approaches using acoustic sensors and hydraulic models to address leaks. However, the integration of these methods and their potential for mutual improvement have not been thoroughly studied. This research proposes a novel approach to improve leak location accuracy by integrating acoustic sensor data and hydraulic modelling within a machine learning framework, using data from an actual use case. Results show that by combining selected acoustic statistical data in time and frequency domains, and various hydraulic features from physical modelling as inputs, a 94% accuracy in the leak location of leaks above 1 L/s can be achieved. This represents a substantial improvement relative to the accuracy achieved by the acoustic methods alone (84%) or by the hydraulic modelling data alone (64%).

DOI: https://doi.org/10.3850/iahr-hic2483430201-522

Year: 2024