Study on extracting pipeline leak eigenvector based on wavelet packet

Abstract

Pipeline leak detection is an important part of pipeline safety, which is usually carried out by extracting feature vectors of leakage signal. However, the complexity of the leakage acoustic emission signal makes the extraction of feature vectors very difficult. To solve this problem, the authors propose an improved wavelet packet algorithm to extract the feature vectors which are constituted by five time-frequency domain parameters: time-domain energy, frequency-domain energy, frequency-domain peak, kurtosis coefficient and variance. Many experiments have been performed to extract feature vectors based on the proposed algorithm, with the results showing the proposed algorithm to be efficient enough to overcome the mixing effects caused by traditional wavelet packet when reconstructing the single sub-band signal. Thus, the proposed algorithm can accurately extract the feature vectors. The study of this article provides a good foundation for the subsequent work such as pipeline leak detection and positioning analysis.