Signal processing techniques for detecting leakage in urban water supply pipelines: Denoising and feature enhancement

As urban water supply pipelines continue to expand, an increasing number of pipelines encounter issues such as aging and corrosion, resulting in frequent leakages. Accurately identifying leakage signals is challenging due to significant background noise complicating signal isolation. To address this issue, this paper proposes a signal denoising and feature enhancement method to capture leakage signals and amplifiy leakage feature based on rime optimization algorithm, variational mode decomposition and teager energy operator-symbolic entropy (RIME-VMD-TEOSE). First, the RIME is employed to optimize VMD, enabling the adaptive selection of key parameters. The bubble entropy of the denoised signal is computed to construct a fault feature vector, which is fed into the RIME-optimized extreme learning machine (ELM) for leakage condition identification. Second, the Teager energy operator is applied to enhance the intrinsic mode functions (IMFs) obtained from the RIME-VMD, and the symbolic entropy of the enhanced signals is calculated to construct a leakage feature vector, which is then input into the RIME-optimized ELM for leakage pressure identification. Finally, the proposed signal denoising and enhancement methods were validated using pipeline experimental data, achieving a recognition accuracy of 95.71 % for distinguishing between large leakage, small leakage, normal, and knock signals, and 97.69 % for identifying pipeline leakage pressures of 0.4 MPa and 0.6 MPa. These results demonstrate the effectiveness of the proposed methods in improving the accuracy and reliability of pipeline leakage detection and pressure identification, contributing to better monitoring and maintenance of urban water supply systems.