Novel Physics-Informed Indicators for Leak Detection in Water Supply Pipelines

Area of Science:

• Engineering
• Environmental Science
• Acoustics

Background:

• Accurate monitoring of urban water supply pipelines is essential for public health and resource management.
• Existing leak detection methods often depend heavily on data-driven features, limiting their robustness.
• Understanding the physical basis of leakage noise sources is key to developing more reliable detection techniques.

Purpose of the Study:

• To propose a novel physical indicator for identifying leaks in urban water pipelines.
• To overcome the limitations of current data-driven leak detection approaches.
• To validate the effectiveness and robustness of the proposed physical indicator through experimental testing.

Main Methods:

• Establishment of an integral form of the leakage source noise power spectral density.
• Rigorous theoretical analysis to develop a physical indicator for leak detection.
• Experimental validation using Support Vector Machine (SVM) and eXtreme Gradient Boosting (XGBoost) models.

Main Results:

• Leak detection models trained with physical features achieved high accuracies: 99.89% (SVM) and 99.97% (XGBoost) in experiments.
• Field tests on a 701 m pipeline demonstrated strong performance: 97.92% accuracy for SVM and 99.31% for XGBoost.
• The proposed physical indicator proved effective and robust in identifying pipeline leaks.

Conclusions:

• The developed physical indicator offers a robust alternative to purely data-driven methods for urban water pipeline leak detection.
• The approach enhances the reliability and accuracy of leak monitoring systems.
• This research contributes to safer and more efficient urban water management.