Water Pipeline Leak Detection Based on a Pseudo-Siamese Convolutional Neural Network: Integrating Handcrafted Features and Deep Representations

The detection of leaks in water distribution systems (WDS) has always been a major concern for urban supply companies. However, the performance traditional leak classifiers highly depends on effectiveness handcrafted features. An alternative method is to use convolutional neural network (CNN) process raw signals directly obtain deep representations that may ignore prior information about leakage. study proposes novel approach WDS using ground acoustic signals, and demonstrates combining features pseudo-siamese (PCNN) model. Mel frequency cepstral coefficient (MFCCs) are selected as additional time- frequency-domain (TFD) Based results model evaluation, optimized PCNN performs better than other methods, with an accuracy 99.70%. A quantitative analysis representations. Model visualization interpretation show feature fusion occurs feedforward PCNN, hence improving model’s performance. present work can effectively support development intelligent equipment WDS.