Leak detection and localization in water distribution systems via multilayer networks

Area of Science:

• Civil Engineering
• Network Science
• Water Resource Management

Background:

• Water distribution networks (WDNs) face increasing management challenges due to their size and complexity, leading to a high risk of failures.
• Effective leak detection and localization are critical for maintaining WDN integrity and minimizing water loss.

Purpose of the Study:

• To propose a methodology for leak detection and localization in WDNs using graph correlation and multilayer network analysis.
• To enhance the management of complex WDNs by identifying failures and optimizing sensor placement.

Main Methods:

• Correlating monitored data to create a temporal graph and classifying vertices for anomaly detection using z-score and interquartile range algorithms.
• Employing a multi-graph approach combining sensor data and network topology for leak localization.
• Utilizing Dynamic Time Warping to compare monitored and simulated leak data for identifying potential leak sites.

Main Results:

• The methodology successfully detects anomalies within 15 minutes of a leak's onset.
• Leaks are localized with an accuracy of 50 meters from the actual leak location.
• Multilayer network analysis provides insights into leak location, sensor coverage, and reduces the need for extensive hydraulic simulations.

Conclusions:

• The proposed graph correlation and multilayer network analysis method is effective for leak detection and localization in WDNs.
• This approach offers significant advantages for WDN management, including improved failure identification and reduced simulation costs.
• The study demonstrates the value of multilayer network analysis in optimizing WDN operations and maintenance.