Extracting acoustic leakage signals in buried pipes

Abstract

This paper develops a method for separation and extraction of pipeline leakage signals. The independent component analysis is applied to separate acoustic signals, which is then coupled with the image similarity distance matching fusion algorithm for the extraction of the signals due to leakage. Acoustic signals generated by real pipeline leakage and common interference noise are collected respectively in order to form a multi-channel matrix with random mixing of leakage signal and uncorrected noise signals. It is demonstrated that the suspected leakage signals extracted after re-sequencing match well with the original leakage signal in terms of coherence and peak cross-correlation. They are further verified based on the cumulative energy distribution and the energy ratio at low and high frequencies. By effective separation of leak signals from the uncorrected background noise signals, the method presented in this paper provides the basis for the improvement of pipeline leak localization.