Research On Quantization Method Of Pipeline Magnetic Leakage Internal Detection Signal Based On Neural Network

Long-distance pipeline transportation has become an important form of transportation for oil and gas transportation in China.Although pipeline transportation is highly efficient and inexpensive,pipeline leakage accidents occur from time to time due to long use of pipelines,rusting and external damage,which bring serious safety hazards to social production and also cause great economic losses to the country.Pipeline leakage magnetic detection technology has been widely used in the detection of pipeline defects because of its lower cost,high sensitivity and other advantages.The problem of quantification of pipeline defects has been the focus of research on this technology,and is also a difficult problem to be solved.The quantification of pipeline defects can provide a more intuitive understanding of the degree of damage to the pipeline and ensure the safety of pipeline operation.In this thesis,we take the characteristic amount of leakage signal at the pipeline defects as the research object,and realize the quantification of pipeline defects by constructing a neural network model.To achieve the quantification of pipeline defects based on the pipeline defect leakage signal,this thesis conducts an in-depth study of the extracted radial and axial components of the pipeline defect leakage signal,and determines the defect leakage signal feature quantities that measure the length,width and depth of the pipeline defects.In order to achieve the accurate quantification of pipeline defects,the relationship between pipeline defect leakage signal feature quantity and pipeline defect size is established by using neural network model in this thesis.After that,the neural network and intelligent optimization algorithm are studied and compared,and the PSO-RBF neural network model is proposed to improve and optimize the RBF neural network by particle swarm algorithm.The PSO-RBF neural network model is then used to quantify pipeline defects of different pipe diameters to verify the quantification effect of the PSO-RBF neural network model on pipeline defects,and the error analysis is made on the quantification results of pipeline defects.Finally,the quantification results of the PSO-RBF neural network model for pipe defects are compared with those of the RBF neural network model for pipe defects.Meanwhile,in order to verify the better applicability of PSO-RBF neural network model in quantifying pipeline defects,the quantization error of PSO-RBF neural network model and the current commonly used convolutional neural network model for quantifying pipeline defects are compared in this thesis.The experimental results prove that the PSO-RBF neural network model proposed in this thesis can quantify pipe defects of different pipe diameters,especially the quantification ability of pipe defect length and depth is better,the quantification error of pipe defect length is about 0.2mm at minimum,and the quantification error of pipe defect depth is about0.4mm at minimum,compared with the traditional RBF neural network model for pipe defects.The quantization accuracy of pipeline defects is greatly improved compared with the traditional RBF neural network model;meanwhile,the comparison of the quantification error of the PSO-RBF neural network model and the convolutional neural network model for pipeline defects also reflects that the PSO-RBF neural network model is more applicable in the quantification of pipeline defects.Therefore,the PSO-RBF neural network model meets the requirements of pipeline defect quantification for practical engineering applications and has good application prospects in the field of defect leakage signal processing for pipeline leakage internal detection.