Intelligent Diagnosis Method For Defects Of Magnetic Flux Leakage Pipeline Inspection With Complicated Conditions

Energy is an important resource for national economic development.The transporta-tion of energy is the prerequisite of its exploitation and applications.Compared with oth-ers,pipeline transportation has many advantages,such as higher transportation capacity,less consumption of resources,short period of construction,higher safety and reliabili-ty,and environmental protection.Pipeline transportation has a history of more than 150 years ago.However,during the long term use of pipelines,it is unavoidable to encounter problems of corrosion and wear,which will destroy the original structure of pipelines to form different degrees of damage.If this kind of damage is not found and repaired in time,the pipeline is likely to evolve into perforation or broken under high pressure and high temperature conditions and the damage will result in leakage of energy.Therefore,it is necessary to diagnose the pipes regularly.The electromagnetic non-destructive testing technology,including ray detection,eddy current detection,ultrasonic testing and magnetic flux leakage(MFL)detection,is often used in the fault diagnosis of pipelines.Compared with other detection methods,the magnetic flux leakage detection has simple process,lower requirements of detection conditions,and can detect many types of defects.This technology is the most widely used pipeline fault diagnosis technology at present.The difficulty of this technology is not only to measure the leakage magnetic field of faults,but also to quantify the collected magnetic leakage signals accurately.Aiming at several signals processing,this paper makes a thorough study on the quantification of defects in pipeline fault diagnosis combined with complicated detection conditions.Be-cause the research is aimed at the practical problems encountered in pipeline detection,these methods are more practical and can be easily converted to application.As the primary goal of pipeline fault diagnosis,the fault area is necessary to lo-cate accurately.In this paper,a intelligent defect area detection method based on space series and random forest prediction is proposed for seamless pipelines.By using this method,the signal in the normal state can be predicted,and the position of the defect is determined by comparing the difference between the predicted signal and the original measured signal.Compared with other methods,this method can greatly improve the detection accuracy of defect areas,reduce the false detection rate and miss detection rate.After locating the defect area,we need to estimate the size of the defect prelimi-narily.The traditional methods often use feature extraction technology.When solving complex defects,they have low precision,because feature extraction procedure relies on prior knowledge and the ability of designer.Inspired by the idea of convolutional neural network,a novel visual transformation convolutional neural network is proposed in this paper.It mainly solves the key problem of transformation from industrial sampling data into images that can be understood by the depth learning algorithm.By adding the vi-sual transformation layer,the defect features in the industrial data can be identified more accurately.In addition,a new mesher magnetic dipole model is proposed to simulate the industrial process,and a large amount of training data can be generated.Compared with other traditional methods,it is proved that this method has higher accuracy of size estimation in solving complex shape defects.In order to further obtain more accurate defect profiles,the paper proposes two so-lutions combined with the working conditions encountered in the experiment.The first is that the detector will vibrate continuously inside the pipe during the detection process.This vibration will lead to the change of the sensors' liftoff and cause the distortion of the MFL signal.The distorted signal will greatly reduce the inversion performance of the defect profile.Therefore,this paper proposes a modification method of liftoff based on finite element model(FEM).The main idea of this method is to introduce a mapping algorithm to correct the original measurement signal and get the "pure" signal.This map-ping method is realized by constructing an iterative defect model which is close to the real defect.Experimental results show that the proposed method can effectively improve the signal-to-noise ratio of the signal,and can more accurately inverse the defect profile.Besides,the relative motion between detector and pipeline will cause eddy current effect.The effect will become more obvious with the increase of detector velocity.The velocity effect may lead to distortion of the defect signal and reduce the inversion accu-racy of the defect profile.In this paper,an effective method for inversing arbitrary defect profiles in different velocity conditions.In the proposed method,the FEM considering ve-locity effect is employed as the forward model.A weighting conjugate gradient algorithm is applied to update the defect profile iteratively in two gradient directions.The experi-mental results show that the FEM considering velocity effect can achieve better inversion accuracy.Moreover,the performance of this method is more stable and robust.Finally,the main work of this paper is summarized.Combined with the rapid de-velopment of science and technology at present,the future development direction and development trend of MFL testing technology are forecasted and prospected.