Research On Oil Pipeline Defect Identification Method Based On Magnetic Flux Leakage Internal Detection

In order to ensure the safety and efficiency of oil pipeline transportation and reduce the potential leakage risk caused by various reasons such as wear,corrosion and accidental damage,it is necessary to carry out regular inspection and maintenance of the pipeline to avoid the energy waste and environmental pollution caused by pipeline leakage;it is necessary to detect the anomalies in the pipeline and identify the defects before the pipeline leakage occurs,so as to reduce the potential leakage risk of pipeline.The pipeline shall be repaired to ensure the safe use of the pipeline.At present,the magnetic flux leakage(MFL)technology is usually used to detect metal loss defects in steel pipelines.As one of the most commonly used non-destructive detection technologies,MFL provides reliable basis for evaluating the safety of pipelines,predicting the service life of pipelines,and repairing and maintaining pipelines.In this paper,the magnetic flux leakage(MFL)internal inspection data of long-distance pipeline is studied,and the magnetic flux leakage(MFL)image is transformed into magnetic flux leakage(MFL)image,the magnetic flux leakage image is detected and recognized intelligently,and the three-dimensional contour of the detected defect area is reconstructed.In view of the above problems,a lot of research and innovation work has been carried out.This paper studies the method of pipeline anomaly edge extraction.In the intelligent recognition of magnetic flux leakage(MFL)image defects,abnormal edge extraction is a very important link,and the accuracy of abnormal edge directly affects the subsequent inversion evaluation.Due to the existence of data noise,the accuracy of edge extraction,especially the edge detection of complex anomalies,is greatly reduced.Moreover,in the face of huge magnetic flux leakage data,the general machine learning algorithm takes more time.Wavelet multi-scale edge detection method is widely used in industrial anomaly extraction.Therefore,aiming at the problem of abnormal edge extraction in magnetic flux leakage(MFL)internal detection,a wavelet transform edge extraction algorithm based on data fusion is proposed,which combines the traditional wavelet multi-scale maximum edge extraction with data fusion,and adds feature layer fusion and decision-making laye fusion to the algorithm Finally,the abnormal edge in MFL internal detection is extracted accurately.This paper studies the extraction method of pipeline micro anomaly region.Aiming at the small abnormal area in pipeline,a method based on U-net depth network is proposed.U-net network is an improved full convolution neural network,which can extract the detailed features of the image with a small amount of data.It can effectively extract the small abnormal area in the pipeline magnetic flux leakage detection.In order to improve the accuracy of extraction,this paper improves the U-net network model and proposes a training method based on confrontation network.The proposed method can accurately and completely extract the small abnormal area,and retain the detailed features of the abnormal area in the magnetic flux leakage image,and has strong robustness,high accuracy and efficiency.This paper studies the identification method of pipeline components and defects.Aiming at the identification of components and defects in pipeline inspection,a deep network defect recognition method based on convolution neural network is proposed.Through the improved convolution neural network algorithm,the recognition accuracy of pipeline components and defect images can be improved,and the accuracy index can reach more than 90%.This method not only has high recognition sensitivity for samples with insignificant signal-to-noise ratio,but also has good displacement robustness and distortion robustness for magnetic flux leakage images.This paper studies the method of pipeline defect contour reconstruction.In magnetic flux leakage(MFL)testing,it is very important to accurately reconstruct the defect contour from the measured MFL signals.Defect contour reconstruction can be used for quantitative research on defects.It has certain practical significance for defect size evaluation and visualization display of defect reconstruction in actual projects.In this paper,a 3D contour reconstruction method based on deviation estimation for random forest defects is proposed.In this method,the random forest algorithm is used to estimate the reconstruction contour deviation according to the deviation between the estimated signal and the actual signal,and the defect contour is updated by optimizing the parameters,and finally the three-dimensional defect contour can be reconstructed.The proposed method has a good effect on the accuracy of defect contour reconstruction.In this paper,the edge of magnetic flux leakage(MFL)anomaly is extracted by wavelet transform based on data fusion,and the subtle abnormal area of MFL image is further extracted by U-net network;the components and defects of MFL image are intelligently identified by improved convolution neural network;the 3D contour reconstruction method of random forest defects based on deviation estimation is used to reconstruct the detected defect area.The purpose of intelligent detection and identification of long-distance pipeline defects is presented to ensure the safety of pipeline transportation.