Leakage identification and localization in water supply networks based on pressure wave phase characteristics
Considering the various limitations of existing pipeline leakage detection methods in practical applications,this paper proposes a leakage detection method that uses the phase of pres-sure waves from the transient frequency response of the pipeline network as the sample parameters for training a neural network.By actively disturbing the valves to generate leakage characteristic signals and simulating potential leakage conditions using a transient flow hydraulic model,the leak-age samples required for training the ANN detection model are obtained.To improve the detection accuracy,a genetic algorithm is employed to optimize the structure of the ANN model.The theory and methods described are applied to leakage detection in typical water supply networks of residen-tial areas.The results show that when the single-point leakage flow rate reaches more than 5%,the detection accuracy is generally greater than 80%,and the detection accuracy is higher in cases where the pipeline network scale is smaller and there are fewer water use nodes.
Water supply networkLeakage detectionFrequency responseValve disturbancePhase spectrum