Locating Water Supply Pipeline Leaks Using Multivariate Variational Mode Decomposition and Multichannel Singular Spectrum Analysis

Environmental noise interference makes it difficult to extract leakage features from vibration signals during pipeline leaks, resulting in inaccurate leak localization. In order to reduce noise in the signal, improve the cross-correlation between signals, and enhance the accuracy of leak localization, this article proposes a method for locating water supply pipeline leaks using multivariate variational mode decomposition (MVMD) and multichannel singular spectrum analysis (MSSA). First, the improved sparrow search algorithm (ISSA) is applied to optimize the initial parameters of the MVMD algorithm. The MVMD algorithm is used to decompose multichannel vibration signals and obtain multiple intrinsic mode function (IMF) components. Then, the MSSA algorithm is used to denoise and reconstruct the IMF components within the frequency band of the leakage vibration signal. Finally, the cross-correlation (CC) algorithm is employed to analyze the reconstructed vibration signal and determine the location of the pipeline leakage point. The average positioning error of this method is 2.03%, which verifies the feasibility and accuracy of the method and has important guiding significance for the actual pipeline leakage positioning.