Author(s): Kaiyi Tan; Yiyi Ma
Linked Author(s):
Keywords: Leak detection acoustic signal processing neural network water distribution system
Abstract: Leakage detection is an important task in the maintenance of urban infrastructure, and traditional detection methods have limitations such as low detection accuracy and high cost, which makes it difficult to be applied on a large scale. In this study, a neural network-based pipeline acoustic signal recognition method is proposed for the problem of leakage detection of urban water distribution pipes. The wavelet transform- Mel-Frequency Cepstral Coefficients method (WT-MFCC) is adopted to preprocess the pipeline acoustic signal, aiming at extracting effective leakage features from the signal. The study establishes several neural network models with different structures and proposes a practical evaluation protocol for the evaluation of model performance. The results show that the best-performing model achieves over 95% testing accuracy and has good applicability and generalization ability, which can be applied to the actual water distribution network inspection situation.
Year: 2025