Advances in magnetic flux leakage testing technology

ZHANG Peng,WEI Minghui,GUO Zhiyong,LIU Zhongxiang

(1. School of Mechatronic Engineering，Southwest Petroleum University，Chengdu 610500,China；2. Department of Biomedical Engineering,Southern University of Science and Technology，Shenzhen 518055，China)

Abstract：Magnetic flux leakage (MFL) testing technology has the advantages of simple principle，easy engineering implementation and low requirements on the surface of the detected workpiece.Therefore，it has been one of the research hotspots in the field of non-destructive testing (NDT) and widely used for testing long distance pipelines.This paper presents the development of MFL tesing technology from the aspects of basic theory,influencing factors,magnetization technology,signal processing,etc.The problems to be solved and the future development are summarized，which can provide reference for the research and system development of MFL testing technology．

Key words：non-destructive testing (NDT);magnetic flux leakage (MFL) testing;magnetization and detection;signal processing

0 Introduction

Modern non-destructive testing (NDT) technology is developing towards rapid and high-precision testing.Research and development of fast,high-precision,intelligent,effective and perfect evaluation NDT systems are of great significance.In modern industry,pipelines are widely used to transport oil,natural gas,water and other materials due to their advantages of economy,high efficiency,safety and large capacity[1].The safe operation of pipelines plays an important role in people’s life,industrial and agricultural production,scientific and technological development and national defense construction.

The basic requirements for piping are safety and efficiency.However,the working conditions of pipelines are generally poor.Because the pipeline is chronically subject to internal transportation fluid pressure,corrosiveness,air humidity and other adverse conditions,accidents such as leakage,explosion,deformation,corrosion of the pipeline,frequenly happen,which will cause great threats and losses to people’s life and the safety of industrial production[2].

At present,the main technologies for in-pipe detection include ultrasonic testing,eddy current testing and magnetic flux leakage (MFL) testing.The ultrasonic testing process is complex,and it needs coupling and is easy to be interfered by the environment[3].Eddy current testing is suitable for testing the parameters of materials,however,it is only sensitive to vertical cracks and suitable for surface testing[4].Compared with these two testing methods,MFL testing has the following advantages.

MFL testing method can accurately identify many kinds of metal defects and all kinds of characteristics of the whole pipeline.Moreover,it can also detect the abnormal cracks of the pipeline.MFL testing does not need coupling agent,has strong anti-interference ability and fast detection speed,and is suitable for fast detection of large-area and long-distance pipelines[5].

In 1933,Zuschlag proposed the idea of using magnetic sensor to detect the magnetic leakage field.However,the technology was not widely accepted until 1947 when Hastings designed the first MFL detection system.In the early 1950s,this technology was widely used in industrial detection.Since then,it has developed from the qualitative identification to the quantitative research of material defects.The theoretical study of magnetic leakage field began in 1966 when Shcherbinin et al.proposed and verified the magnetic dipole method,which approximated the defect as point,line and strip magnetic dipole.Shcherbinin et al also proposed the theory of interaction between the magnetic field and the defect and then experimentally verified it[6-7].However,the magnetic dipole model is not suitable for the calculation of nonlinear defects,therefore it is difficult to solve the problems caused by the complex MFL,including angle,depth and width of defects.Hwang et al.introduced the finite element method to the calculation of the leakage magnetic field in 1975.For the first time,he connected the amplitude of the leakage magnetic field with the permeability of the material,and analyzed the influence of the width,depth and angle of the rectangular groove on the leakage magnetic field.By the end of the 20th century,a lot of researches had been done on the distribution of defect magnetic field,influencing factors and defect inversion.Russian scholar Doubov put forward the idea of magnetic memory detection technology.Because this technology does not need magnetization or demagnetization and can detect early failure and damage,it is still one of the research hotspots[8].At the beginning of the 21st century,Sophian,a British scholar,proposed the pulsed magnetic flux leakage (PMFL) technology combining the advantages of MFL detection and pulsed eddy current detection,which not only magnetizes the ferromagnetic material,but also realizes the deeper defect detection and improved the sensitivity[9].In the following year,Wilson et al.proposed the pulsed electromagnetic method for defect detection and characterization of the pulsed magnetic resistance (PMR),PMFL reaction surface defect location information and eigenvalue,and PMR reaction sub-surface defect characteristic information,so as to realize the distinction between the surface defect and the sub-surface defects[10].

1 Principles and models

1.1 Principle of MLF testing

The basic principle of MFL testing is that after the ferromagnetic material is magnetized,the permeability of the defects on or near the surface of the pipe wall becomes smaller,and part of the magnetic flux will leak to the surface of the material to form a MFL field.The magnetic sensor measures the MFL field,analyzes the measured signal,and then obtains the information of the defect[11-12].The principle diagram is shown in Fig.1.The MFL detector is composed of permanent magnet,steel brush,yoke iron and magnetic sensor,and they form a closed loop with the pipe wall.If there are no defects in the pipe wall,the magnetic flux will rarely leak into the air and will not form a leakage magnetic field.If not,there will be an MFL into the air to form a magnetic leakage field.The magnetic sensor determines the defect characteristics by detecting the magnetic leakage signal[13].

Fig.1 Schematic diagram of MFL testing

1.2 Models of MFL testing

At present,the calculation methods of MFL testing mainly include the analytical model of magnetic dipole based on magnetic charge theory and the numerical method of solving Maxwell’s equation.Scholars have formed two schools in this field：one is the classical theoretical school focusing on the magnetic dipole method;the other is the engineering approximation school focusing on the numerical calculation of solving the Maxwell equation.

1.2.1 Analytical method

According to the theory of magnetic dipole,the magnetic leakage field of defect is produced by the opposite magnetic dipole,and the defect can be approximated as point,line and strip magnetic dipole.This theory is one of the basic theories of analytical method of magnetic leakage testing.There are many kinds of defects in industrial environment.One of the main applications of magnetic dipole theory is to explore the distribution of leakage magnetic field in different defect shapes to expand the detection range.

As shown in Fig.2,taking two-dimensional space as an example,two magnetic dipoles are located on both sides of the defect,and the magnetic charge density is -σand +σ,respectively,thus the leakage magnetic field generated at pointPis

Fig.2 Magnetic dipole model

(1)

whereB,σandμ0denote the strength of leakage magnetic field,magnetic charge density,magnetic constant,respectively;n1andn2respresent the unit vector from -σandσto pointP,respectively.

This model solves the problem of the spatial distribution of MFL,and simulates the holes and pits.The surface magnetic dipole model can be used to calculate the rectangular open crack[14-16].However,it is still difficult to calculate the complex defects.In order to expand the application range of magnetic dipoles,Mandache et al.modified the analytical model of circular defects and applied it to the holes,correlating the normal component MFL profile of circular defects with the defect size along the magnetic field direction[17],as shown in Fig.3.

As shown in Fig.3(a),according to the magnetic dipole theory,the magnetic field dHgenerated by the charge dpat a distanceris expressed as

(a) Dipole diagram of defects in cylindrical holes with radius R

(a) Top view of defect

(2)

The magnetic field component at pointp(x,y,h) is expressed as

(3)

Thex-axis component of the leakage magnetic field is offset due to symmetry,and only the normalz-axis component on the linex=0 mm is considered.Then the magnetic field on the positive and negative pole sides of the defect can be expressed as

(4)

(5)

Dutta et al.[18-19]proposed a three-dimensional analytical MFL model of the surface fracture defect field of ferromagnetic specimen based on the magnetic dipole model,and the magnetic charge on the surface element can be expressed as

dp=σdS=MndS=|M|sinθdS,

(6)

whereMis the magnetization vector,which is a constant across the defect surface;andnis local unit normal vector of areaS.

Then we can obtain

r=(x-Rcosθ)i+(y+Rsinθ)j+(h-z)k.

(7)

Finally,substituting the above formulas into the MFL vector formula,we can obtain

(8)

H(r)=Hxi+Hyj+Hzk.

(9)

Although this model can simulate and analyze the three-dimensional leakage magnetic field distribution of defects,it can only be used for simple defects such as cylinders and rectangles.

As shown in Fig.4,Trevino et al.[20]optimized the above formula to improve the description of the normal vector from one parameter to two parameters,which is expressed as

dp=σdS=Mn(s)dS(s)=|M|sinθsinβdS(s).

(10)

Then the final mathematical model is

(11)

Based on the above basic model,the following researchers further expand the shape of defects,including spherical,concave-convex defects,etc[21].In fact,according to the difference of magnetic material distribution and magnetic properties,the magnetic charge in the material and the defect wall after magnetization of the defective material are non-uniformly distributed,and the external magnetic field strength and magnetic induction strength are nonlinear,which is difficult to describe in the dipole sub-model.Some scholars have proposed to use the unit magnetic dipole moment as an integral element,and transform the integral problem of the whole defective material into the superposition of several elements after integration,so as to obtain the calculation equation of the magnetic field at any field point[22-23].

1.2.2 Numerical method

Due to the magnetic dipole model is not suitable for calculating nonlinear defects,it is difficult to solve the problems of MFL caused by complex angle,depth and width of defects.By constructing the equivalent energy function of the electromagnetic vector partial differential equation,the finite element method transforms the problem of discrete region and energy minimization into an algebraic system of solving the vector magnetic potential of each grid node as an unknown quantity[24-25].In 1982-1986,Förster,used the experimental method to verify the calculation of the finite element magnetic leakage field proposed by Hwang et al.,and modified the influence of the defect magnetic leakage crack width[26].Katragadda et al.made a comparative analysis on the axisymmetric geometry and the three-dimensional asymmetric magnetized body under the condition of neglecting the influence of the velocity.Li Y (Lreland) and Dutta S M(UK) also published their research results on the simulation of MFL.Kang from Huazhong University of Science and Technology,Li from Tsinghua University,Yang from Shenyang University of Technology and their team have also studied the distribution of defect leakage magnetic field[27].

Due to the limitations of both analytical method and numerical method,modern scholars have also explored other models.Wang et al determined the magnetic moment of the solenoid by using the b-h characteristics of the material,and simulated the magnetic leakage field with the semi-solenoid as the source.Then,they presented a solenoid analytical model (see Fig.5) for MFL field based on the molecular current model of a magnetic medium.Their team also proposed an electromagnetic model based on the magnetization mechanism of the magnetic medium to simulate the magnetic leakage field.By introducing solenoid model and Jiles-Atherton model,the leakage magnetic field of complex defects can be accurately calculated[28-29].

Fig.5 Solenoid model

2 Factors affecting magnetic leakage

The signal of magnetic leakage field of defect is affected by many factors,such as the dimension of defect,the direction of extension,the magnetization of pipeline,the distance between defect and sensor,the stress change of material,the magnetization of pipeline and the running speed of detector.

Fig.6 shows the vector distribution of defect leakage magnetic fieldsBxandBy.The transverse axis is the defect width,and the longitudinal axis is the magnetic induction intensity[30-31].

Fig.6 MFL signal component

In addition,during the operation of oil and gas pipeline,the leakage magnetic field will be affected by the changes of defect stress,line pressure,bending or residual stress,especially in the stress concentration area.Based on the metal magnetic memory effect and reverse magnetostriction effect,the weak magnetic field on the surface of ferromagnetic structure under the action of stress is detected to determine the material damage and stress state.Because it can detect tiny defects and the equipment is simple,it has been one of the hot researches.In recent years,the research subjects include the influence mechanism of different stress on the leakage magnetic field,how to improve the reliability of detection under the condition of different location and susceptibility[32-34].Under fluid pressure,weld cracking is one of the main forms of pipeline failure.Cui et al proposed a fluid solid magnetic coupling method based on virtual crack closure technology(VCCT) to study the dynamic application of fluid pressure,the coupling effect of fluid pipe weld structure and magnetic field,and the characteristics of magnetic leakage field in different stages of weld crack growth[35-36].

In the process of high-speed detection,there is velocity induced eddy current,which leads to the distortion of defect signal and is more obvious with the increase of velocity and plate thickness.In order to solve the distortion of sensing signal caused by the velocity induced eddy current,in addition to the study of the influencing factors such as the speed and the thickness of the plate,the change characteristics of the leakage signal caused by the motion induced eddy current are also studied and analyzed.Finally,the influence of the sensing signal characteristics is analyzed and compensated[37-41].In addition,skin effect is also a common phenomenon that affects the effect of alternating excitation.Skin depth does not decrease infinitely with the increase of excitation frequencyf,and when the excitation frequency reaches 1 kHz,the skin depth is stable at about 1 mm[42].

3 Testing technology and equipment

3.1 Magnetization technology

The function of the excitation system in the detector is to magnetize the tested material and then generate a leakage magnetic field around the defect.The design requirements of the magnetization system are to produce enough magnetic field strength in the tested material to keep even between the inner and outer walls and to keep the same magnetic field in the detection,etc[43].The magnetization methods can be divided into AC magnetization,DC magnetization and permanent magnetization.

1) AC magnetization.AC magnetization is widely used because of its simple structure,easy control and low cost.However,in the process of detection,it is easy to produce the convergence effect and eddy current field.The magnetization frequency and detection speed are important factors that affect the reliability.The increase of detection frequency can increase the detection sensitivity,but the detection depth will decrease[44-46].

2) DC magnetization.DC magnetization can be divided into DC pulsating current magnetization and DC constant current magnetization.DC constant current magnetization is simpler in structure than DC constant current magnetization,but its excitation current is larger,generally a few amperes or even a hundred amperes.Although DC magnetization has no specific requirements for detection speed,its magnetization capacity is limited and demagnetization is required every time it is used.

3) Permanent magnet magnetization.Permanent magnet is used as the excitation source for permanent magnet magnetization.Its magnetization characteristics are the same as that of DC magnetization,but the intensity regulation is not as good as that of DC magnetization.Permanent magnet is usually made of permanent ferrite,aluminum,nickel cobalt permanent magnet and rare earth permanent magnet.Because of its high energy,small volume,no electricity and other characteristics,it has been well used in MFL detection.

In addition to the above traditional magnetization methods,pulse MFL detection is also one of the hot research directions in recent years.In essence,this technology is also a kind of alternating MFL detection technology,with a wide spectrum and deeper penetration depth,which can effectively improve the detection sensitivity of subsurface defects.It is also better than traditional MFL in estimating defect parameters[46-48].

In the high-speed MFL detection,the effective time of magnetization is very short,and the phenomenon of magnetic after effect will lead to the inability to realize the magnetic saturation quickly.In addition,the study of the factors that determine the relationship between the measured component of the signal and the speed and the reduction of the influence of eddy current on the metal loss nearby will be the subject of future research.The multi-stage magnetizing coil can reduce the variation of magnetic field B strength and suppress the eddy current effect.Moreover,it is also one of the methods to increase the magnetization time and to reduce the influence of magnetic aftereffect[49-50].

Furthermore,some new ideas and magnetization structures for complex environment have been studied.As shown in Fig.7,the AC and DC magnetization are combined,and the DC magnetic field magnetizes the tested object to detect the leakage information of the magnetic field in the transverse (circumferential) direction of the crack;then the AC magnetic field is used to form the eddy current field of the magnetic circuit to detect the disturbance signal of the secondary induced magnetic field in the longitudinal (axial) direction of the crack,so as to realize the blind spot free detection of the crack[51-52].In order to deal with small and medium-sized pipeline detection in urban environment,reducing the weight of detector and magnetic adhesion is the main idea to reduce blocking,for example,using automatic driving robot system,using differential coil to replace heavy and magnetic coil,and optimizing the configuration of permanent magnet[53-55].

Fig.7 Composite detection structure diagram

3.2 Sensors and detection methods

With the development of the technology of magnetic field and electronic sensor,the technology of magnetic field measurement has been greatly improved.According to different measurement principles,it can be divided into electrical induction method,magnetic resistance effect method,Hall effect method and magnetic resonance imaging.

Electromagnetic induction method.Based on Faraday’s law of electromagnetic induction,this technology uses induction coil to measure DC magnetic field,AC magnetic field and pulse magnetic field.

Magnetic resistance effect method.This technology comes from the variation characteristics of material resistance under the action of magnetic field,mainly using semiconductor relief elements and ferrometallic thin film relief elements[56].

Hall effect method.This technology is based on Hall effect to measure magnetic field.It mainly uses Hall element with mature manufacturing process,better stability and temperature characteristics to measure leakage magnetic field.

Magnetic resonance imaging.By absorbing or radiating electromagnetic waves of a certain frequency in a magnetic field,some microscopic particles will cause resonance.Therefore,the strength of the magnetic field is obtained by measuring the degree of resonance.Due to the different types of particle resonance,devices can be of various types.

Magneto optical method.This approach utilizes the magneto-optical and magneto-stricture effects.It mainly uses optical fiber sensor which can work in harsh environment.

Magnetic sensor is a device that converts magnetic signal into electrical signal.There are many kinds of magnetic sensors.According to different measurement requirements and methods,the types of sensors are also different.The main types of sensors include fluxgate,Hall element,induction coil,magnetic diode and transistor,magnetoresistance,etc.The measurement ranges of different magnetic sensors are listed in Table 1.

Table 1 Measurement ranges of magnetic sensors

In order to achieve the best effect of MFL testing,ferromagnetic materials must be magnetized to be saturated or near saturated.However,realization of magnetization saturation or near saturation in the actual engineering environment is affected by many factors.Research mainly focuses on improvement of magnetization technology and study of unsaturated magnetization state characteristics.

The research on the improvement of magnetization detection technology mainly focuses on sensor technology.Through the use of high-sensitivity and micro-magnetic sensor,the array MFL sensor has been developed,which can provide multiple scanning signals and realize the accurate inversion and quantification of defect detection results after the algorithm processing.Different from the laboratory conditions,the vibration of the detector in the in-service pipeline will affect the fluctuation of the lift off value of the sensor.The study of a new sensor array method can reduce the impact of this fluctuation,and even detect the weak magnetic signal when the location of the shallow defect is on the far side surface or far away from the magnetic sensor[57-59].

The research of MFL detection technology for unsaturated magnetization mainly focuses on the characteristics and formation rules of MFL signals.For example,Wu et al.studied the influence of internal defects of steel plate on the distribution of dynamic permeability and the formation mechanism of local hysteresis loop under the condition of variable excitation,thus establishing the relationship between dynamic permeability and local hysteresis loop.This method only needs to provide a shallow bias magnetization field without saturation magnetization,which provides an idea for breaking through the limitation of alternating MFL method,and also plays a positive role in reducing the detector volume[60-61].

4 Detection signal processing

4.1 Preprocessing

The actual signal of MFL detection will be polluted and even submerged by various noise sources.Therefore,before identifying the MFL signal,it is necessary to preprocess the signal,including the foreign matter signal processing and waveform interpolation.Because of the diversity of interference sources,the processing methods are also different.For noise,many filtering methods are used,such as wavelet or other filtering methods[62-63].To eliminate errors,correction algorithm,weighted gradient algorithm and decoupling algorithm are often used[64-66].In addition,it is very important to select high compression ratio,high speed and high precision on-line compression and defect hazard identification methods in special detection environment such as pipeline[67-68].

4.2 Identification and quantitative analysis

The MFL detection signal is mainly used to identify and invert the defects,including the distinction of different types of defects,qualitative analysis of MFL characteristic signals obtained and quantitative description of defect shape parameters,which is a typical anti-electromagnetic field problem.According to the characteristics of the inverse problem of MFL testing,the solutions can be divided into direct method and indirect method.The indirect method is still the main method of defect quantification.At present,there are three methods to solve the quantitative problem of MFL detection:mapping method,iterative method and pattern classification method.

1) The mapping method needs to establish the standard defect feature libraryx,and then use the model or physical experiment to get the main features of these defects,including signal peak and valley values,defect significance,contrast,center point,etc.,forming the standard signal libraryyand then establishing the mapping relationship betweenyandxaccording to the statistical or neural network methods.However,this method relies too much on the consistency and accuracy of statistical samples and training samples,but it lacks scalability,and quantitative accuracy of the defect with complex structure in detecting the actual shape is low[69-70].

2) The iterative method is based on the model and uses the forward model (magnetic dipole model,finite element model,etc.) to solve the forward problem with good behavior in the feedback cycle.The method first estimates the initial parameters of defects,and then solves the forward problem to predict the MFL signal.If the error is greater than the predetermined threshold,adjusting the defect parameters and repeating the iterative calculation makes the error small at a predetermined threshold.Different forward models have different characteristics.The calculation speed of the magnetic dipole model is fast,but the accuracy is not high;the finite element model has high accuracy,but the calculation amount is huge,especially in the application of three-dimensional model,which limits its practical application.The successful application of neural network theory in pattern recognition provides a new way to overcome this difficulty.The radial basis function neural network (RBFNN) is used to solve the forward problem.On the one hand,the calculation efficiency is improved;on the other hand,the accuracy and adaptability are improved by using the closed-loop structure.In addition,in recent years,some scholars have also proposed many improved models and hybrid models.In addition to the traditional gradient descent algorithm and simulated annealing algorithm,the cuckoo algorithm in other fields is also used in MFL inversion[71-74].

3) Pattern classification includes clustering algorithm and neural network method.The main features of defect signals are extracted,the concept of basic signal and the method of basic combination signal are proposed,and the inverse problem is decomposed into a limited number of defects.At present,neural network has been widely used in the field of nondestructive testing signal classification[75-76].

Fig.8 Radial basis function neural network

However,these methods have some defects.The direct method is limited by poor conditions of electromagnetic inverse problem,therefore,it cannot get the only stable solution,and the indirect method is still the main method of defect quantification.In addition,some scholars applied other methods to MFL inversion,for example,using function convolution to get the spatial spectrum characteristics of defects[77],defect width estimation based on image processing and morphological methods[78],and using improved SSD network to extract small target[79].

5 Conclusions

In this paper,we summarize the research progress of MFL testing technology,including MFL testing theory,excitation and testing technology and defect signal processing,which provides a certain technical basis and reference for the research and engineering application of MFL testing technology.At present,MFL detection has made great achievements in practical application,but further research is still needed in theory and engineering application.

At present,the theoretical research of MFL is not closely connected with the applied technology.There is a lack of research on the detection phenomenon.In addition,doubtful points needs to be solved such as the formation mechanism of the leakage magnetic field,the mathematical relationship between the defect characteristics and the leakage magnetic field signal,the theoretical model of different types of defect leakage magnetic field and so on.

For physical mechanism of MFL NDT,the in-depth application research of magnetic refraction,magnetic diffusion,magnetic compression and other magnetic effects is of great help to obtain more comprehensive information of magnetic leakage field and further improve the detection performance of magnetic leakage equipment.

Generally,more accurate magnetic leakage identification and inversion technology requires a large amount of computation and takes a long time.High-speed MFL testing under the premise of accuracy will be one of the research directions.