Research On Pipeline Defect Signal Analysis Method Based On Optimized EL

A considerable number of oil and gas pipelines in China have been put into operation for many years and have entered a period of high accident incidence.Once an accident occurs,it will cause serious consequences.If the oil and gas pipelines are inspected and repaired before the accident occurs,the possible losses will be It is greatly reduced,so it is of great significance to do a good job in the detection of oil and gas pipelines and the analysis of pipeline signals.In this paper,the magnetic eddy current defect signal of oil and gas pipeline is taken as the research object,and the noise reduction,feature extraction,feature dimension reduction and defect diagnosis based on ELM are studied on the defect signal.The main research contents of this paper are as follows:(1)Aiming at the preprocessing of the magnetic eddy current signal of oil and gas pipelines,the wavelet analysis noise reduction method is used to denoise the magnetic eddy current signals collected in the laboratory and the gas transmission station site respectively,and the wavelet analysis method is used to verify the complex content of the gas transmission station.The noise reduction effect of magnetic eddy current signal is evaluated,and its noise reduction effect is evaluated.Variational mode decomposition(VMD)method is used to extract the characteristic parameters of laboratory signals in time-frequency domain.Three methods of MIC,KPCA and LTSA are used to reduce the dimension of the extracted feature parameters,and it is proposed to use the relationship between the feature dimension and the accuracy rate after dimension reduction to select the dimension after dimension reduction.Use the LSAT dimensionality reduction method to reduce the dimension of the best accuracy to 10 dimensions,as the input sample data of the following ELM model.(2)Aiming at the problem of determining the number of neurons in the hidden layer in ELM,a traversal method is proposed to determine the optimal number of neurons in the hidden layer of the laboratory magnetic eddy current data sample is 30.The input weights and thresholds of ELM are optimized by three algorithms of PSO,CS and WOA.The three algorithms of PSO-ELM,CS-ELM and WOA-ELM are used to diagnose oil and gas pipeline defects respectively.The diagnosis accuracy of PSO-ELM is 92.38 %,the diagnostic accuracy of CS-ELM is 96.13%,and the diagnostic accuracy of WOAELM ranks first,and the diagnostic accuracy reaches 96.88%.Compared with ELM,the accuracy of WOA-ELM in the diagnosis of magnetic eddy current pipeline defect data is increased by about 19.4%,and the accuracy and stability of defect diagnosis are greatly improved.(3)In order to facilitate the practical application of oil and gas pipeline signal processing,make the operation of pipeline defect diagnosis more simple and easy to use,realize the visualization and tooling of signal analysis and processing,and develop a set of visualization programs for magnetic eddy current pipeline defect signal analysis.The main functions that can be realized include wavelet noise reduction in different threshold determination methods of signals,extraction of time-frequency domain and timefrequency feature parameters,feature dimension reduction,and defect prediction and classification.And do segmentation processing,you can use a single function alone.