Signal Processing and Pattern Recognition for Leak Detection in a Water Distribution Network

Leaks are a constant problem in water distribution systems, resulting in wasted resources, environmental impacts, and financial losses. Thus, it is crucial to develop effective and agile methods to detect network leaks. In this context, this study proposes a leak detection methodology using three different processes. The first consists of treating monitoring data through independent component analysis, and the other two detection processes use the interquartile range (IQR) and matrix profile (MP) techniques, respectively. The methodology is evaluated based on a set of benchmark data. The results indicate that the proposed approach is effective in detecting leaks, with some cases being detected in a few minutes after the beginning of the leak. It is worth mentioning that the IQR method presented better performance in detecting leaks with abrupt onset, whereas the MP method was more efficient in leaks with gradual increase in flow. In summary, the proposed methodology offers a robust and promising approach for fast and accurate leak detection in water distribution networks.