Leakage Detection in Pipelines Based on Synchrosqueezed Wavelet Transform

Abstract

Because of uncertainty, the diagnosis of gas pipeline leakage is a difficult problem. In order to solve the difficult problem to distinguish leakage in gas pipelines, an approach of leakage detection using Synchrosqueezed Wavelet Transform (SST) and Stacked Contractive Auto-encoder (SCAE) was proposed. First, the sound signal in gas pipelines is subjected to simultaneous compression wavelet transform to obtain time-frequency images n grey-scale and normalized the image. Then, the time-frequency representations were compressed to the appropriate size. The compressed time-frequency matrix should be expanded into a column as the SCAE’s input. The classification model of SCAE and SoftMax was established to realize the leakage detection in pipelines. The experimental results indicated that this method could effectively identify small leakage in pipelines, and the method could effectively improve the fault recognition rate and reduce the training cost.