Anomaly Recognition in Water Pipe Systems Using a Characterization Framework and Persistent Hydraulic Transient Waves

AbstractAnomaly detection in pipe networks is critical for targeted maintenance within water distribution systems. A paired impulse response function (termed paired-IRF) technique to detect anomalies in pipe networks has been previously proposed by the authors, and its high sensitivity and accuracy have been proved through laboratory experiments. In this paper, numerical experiments were conducted to individually investigate the features of the response induced by different anomalies (including leaks, blockages, and wall-deteriorated sections) in the paired-IRF trace. With these features characterized from these case studies, an anomaly characterization framework was developed to automatically determine the type, location, and size of anomalies by recognizing the features in the paired-IRF trace. The robustness of the framework was tested on a numerical pipe network where leaks, junctions, a short section of blockage, and a wall deteriorated section existed. These different types of anomalies were successfully detected and classified using the characterization and recognition process.