Research On Defect Region Identification And Profile Reconstruction In Petroleum And Gas Pipelines Based On Magnetic Flux Leakage Testing

Petroleum and natural gas are very important energy and chemical raw ma-terials,they play a pivotal role in industrial and agricultural production,national economic development and national defense construction.Since the oil and natural gas are inflammable and explosive dangerous goods,it is necessary to pay special attention to the safety of transportation.Because of its good safety,high reliability and high cost performance,pipeline transportation has gradually become the most important and the most commonly used transportation method for oil and natu-ral gas.With the increase of service time,the possibility of corrosion and leakage increases.In order to maintain the normal operation of the pipeline,it is neces-sary to carry out regular inspection and maintenance to prevent leakage and other accidents.At present,the commonly used pipeline nondestructive testing technology main-ly includes:magnetic flux leakage testing(MFL),eddy current testing(ECT)and ultrasonic testing(UT).MFL testing has become the most widely used nondestruc-tive testing(NDT)method because of its low environmental requirements,high reliability and high precision.In addition to ensuring that the detector hardware can acquire high resolution MFL signal,it is necessary to do a good job of MFL signal processing.In the process of MFL signal processing,defect diagnosis and defect profile reconstruction are two important and key links.The defect diagnosis completes the function of identifying the defects from all the abnormal regions.The defect profile reconstruction is used to estimate the real profiles of defects,which provides a reference for the estimation of the severity and the urgency of mainte-nance.In this paper,these two key problems in MFL testing for oil and gas pipelines are studied.In order to improve the accuracy of the defect diagnosis,this paper proposes a convolutional neural network(CNN)based method.In some previous papers,feature based methods are usually used to diagnose the defects.In these feature based methods,the features of the abnormal regions are extracted and used as the input of the model to determine whether suspected region is a real defect.This kind of method are noise sensitive and cannot balance identification accuracy and robustness,thus they can't achieve a good accuracy.The CNN based method use the whole MFL image as the input of the model,it can automatically extract the distinguishable features to diagnose abnormal region type.This method skips the handcrafted feature extraction process and saves some computation time and storage space.Benefited from the special structure of the proposed model,this method is robust for noise,shift,scale and distortion variances of input MFL images,which ensures a higher accuracy than the traditional methods.In this paper,three kinds of different defect profile reconstruction methods are proposed,which are suitable for different industrial background.Defect pro-file reconstruction method based on genetic tabu search hybrid algorithm(GTSA),RBFNN error adjustme.nt and modified harmony search algorithm(MHS),respec-tively.The defect profile reconstruction method based on genetic tabu search hybrid algorithm uses radial basis function neural network(RBFNN)as the forward model,and uses the genetic tabu search hybrid algorithm to update the predicted defect profile.The forward model based on RBFNN and genetic tabu search hybrid algo-rithm can guarantee that the method has a fast reconstruction speed.Due to the large number of training samples required in the training RBFNN forward model,the precondition of using this method is that it has a large number of prior magnetic leakage data.Simulation and experimental results show that the proposed method can obtain a good defect profile reconstruction accuracy in a short time.The defect profile reconstruction method based on the RBFNN error adjustment uses the finite element method(FEM)as the forward model,and the RBFNN based error adjustment strategy is used to update the predicted defect profile.This method also requires a large amount of prior magnetic leakage data as support.The finite element method is used as the forward model,so the arbitrary irregular profile can be reconstructed.The RBFNN based error adjustment strategy can ensure that the estimated profile quickly approach the real defect profile,thus this method can achieve the high efficiency.The defect profile reconstruction method based on the modified harmony search algorithm breaks through the limitation of requiring a large amount of prior magnetic flux leakage data.In this method,the finite element method is used as the forward model,and the improved harmony search algorithm is used to update the estimated defect profile.The improved harmony search algorithm can improve the intelligence level and optimization performance of the original harmony search algorithm from three different aspects.At the same time,an improved sum-square error function is proposed,which is used as the fitness function,which can effectively improve the fitting accuracy of the predicted and real profiles in the defect response segment.The simulation and experimental results show that the method based on improved harmony search algorithm is effective and can be applied to the practical engineering.Finally,the work of this paper is summarized.According to the current sit-uation of oil and gas pipeline MFL testing,the future development direction and development trend are predicted and forecasted.