Author(s): Andrea Menapace; Maurizio Tavelli; Daniele Dalla Torre; Maurizio Righetti

Linked Author(s): Andrea Menapace, Maurizio Righetti

Keywords: Rtificial Neural Networks; Leaks Detection; Numerical Hydraulic Modelling; Sustainable Water Management; Transient Signal Processing; Water Distribution Networks

Abstract: The digitalisation, along with the implementation of smart grids, is driving a new era of efficient and sustainable water distribution systems. Implementing smart sensors, transmission systems, and integrated support systems allows for the development of cutting-edge technologies in real-time monitoring and control, ensuring prompt detection and accurate localisation of anomalies and water losses par excellence. Transient test-based techniques consist of the generation and monitoring of small pressure waves in the pipelines with the final aim of identifying any anomalies through the analysis of the transient signals. To enable this technology to be suitable for real-world applications, it is necessary to develop techniques for processing transient pressure signals for real-time anomaly detection. Thus, a methodology embedding an unsteady hydraulic model for generating big training datasets with machine learning algorithms exploiting this synthetic transient data for automatically identifying leaks in pipes is proposed. The feasibility of this methodology is demonstrated by two different complexity tests, which ensure promising results in the case of both single conducts and simple networks.

DOI: https://doi.org/10.3850/iahr-hic2483430201-272

Year: 2024