Feature Extraction And Recognition Of Pipeline Defect Signal In Compressor Station

The pipe in compressor station are mainly used to transport flammable and explosive hazardous materials.Usually he pipes are in a harsh environment,so they often produce defects that are likely to have extremely serious consequences.In order to prevent the occurrence of major accidents in pipes,the regular defect detection is needed.for pipes.Collecting detection signals and analyzing the characteristics of the signals In this paper,feature extraction of pipe detection signals will be performed and analyzed.The main research contents and conclusions are as follows:(1)The time-frequency distribution of the hole defect and crack defect signals in the pipe is analyzed.Studying the highest energy layer characteristics of the Wigner-Ville distribution of pipe defect signals.The energy concentration frequency of the hole defect is larger than the energy concentration frequency of the crack defect.The area of the highest energy layer of the hole defect is smaller than the area of the highest energy layer of the crack defect.The defect types can be determined based on the energy concentration frequency and the product of the limit points in the highest energy layer.(2)By collecting a large number of pipeline defect signals.The time-frequency matrix of Wigner-Ville distribution is partitioned in the direction of frequency axis,and the singular value of the divided local matrix is obtained to construct the eigenvector matrix of the defect signal.The Support Vector Machine(SVM)is used to train with the eigenvector matrix to get training model and realize classification of defects.Fifty groups of sample data were tested,and the accuracy rate was 92%.(3)Taking two defects of pipeline in station library as an example,the eigenvector matrix of defect signal is constructed by using the WVD-SVM method,and defect recognition of pipeline in station library is realized by testing.(4)In order to study the feature extraction of defect signals,the defect signal feature extraction software based on Py Qt and Eric is developed.The software realize time domain feature extraction and time-frequency distribution design of defect signals,and provide good analysis methods for subsequent defect signal analysis.