This report was prepared by Battelle as an account of work sponsored by the Gas Research Institute (GRI). Neither GRI, members of GRI, Battelle, officers, trustees, or staff of Battelle, nor any person acting on behalf of either:

a. Makes any warranty or representation, expressed or implied, with respect to the accuracy, completeness, or usefulness of the information contained in this report, or that the use of any information, apparatus, software, method, or process disclosed in this report may not infringe privately owned rights; or

b. Assumes any liability with respect to the use of, or for damages resulting from the use of, any information, apparatus, software, method, or process disclosed in this report.

Reference to trade names or specific commercial products, commodities, or services in this report does not represent nor constitute an endorsement, recommendation, or favoring by GRI or Battelle of the specific commercial product, commodity, or service.

Research Summary Form

Table of Contents without Links (Table of Contents with Links)

References

Introduction

Pipeline operators use a wide variety of methods to evaluate, inspect, and monitor the hundreds of thousands of miles of transmission pipelines now in operation worldwide^[AGA] . Such activities include right-of-way surveys, cathodic protection surveys, leak detection programs, excavations to look for pipe corrosion or protective coating failures, hydrostatic tests, and the use of in-line inspection tools that travel through the pipe. Combinations of these procedures constitute an overall integrity assurance program of the pipeline operator.

Magnetic flux leakage (MFL) is the oldest and most commonly used in-line inspection method for finding metal-loss regions in gas-transmission pipelines. MFL can reliably detect metal loss due to corrosion and, sometimes, gouging. In addition, while not designed for this purpose, MFL can sometimes find other metallurgical and geometric conditions^{[Bubenik98, Grimes92, Nestleroth99, Papenfuss91]} .

Brief History of MFL
Summary of MFL Capabilities

This report presents the underlying principles and current status on the use of MFL for pipelines as they are understood by the authors. A significant development effort is underway at in-line inspection service companies and by GRI and other research organizations^[GRI97]. These efforts will undoubtedly lead to an enhanced understanding of the topics discussed herein and to continuing advances in the capabilities of commercial MFL in-line inspection tools. (See Pigging Products and Services Association for information on current in-line inspection companies.)

Report Organization

This report is an update to the widely distributed MFL topical report first prepared in 1992 ^[Bubenik92] . It includes additional information and details on MFL inspection technology. This updated report was written in a Web format to let readers quickly access information of interest to them.

The Table of Contents lists the main sections of the report. The body of the report is done in the style of an Executive Summary. That is, it contains brief descriptions and major conclusions. Within each section, there are links to background and more detailed information. In addition, there is an on-line glossary. Words shown in italics are contained in the glossary. In the written version of this report, the links and glossary are available as separate Appendices.

Links are identified with a document icon (), a figure icon (), an underline, or a button. Typically, document links open in place of the current document (which can be accessed again by pressing the back key); figure links open in a separate window; and underlined links (without an icon) redirect the user to another location on the same page or to an external Internet link. The text on a button will identify its use; buttons can redirect the user, open windows, or launch an external program.

Nondestructive Testing Termonology

Overview of Pipeline Inspection Using MFL Tools

An understanding of magnetism, flux, and flux leakage is needed to understand the capabilities of MFL inspection systems. This section presents an overview of magnets and flux leakage as they apply to MFL inspections. ^{[Bozorth51, Dobmann87]}

MFL starts with a magnet. A magnet has two ends, called north and south poles. The poles exert forces on steel pieces and on other magnet poles. This force of attraction is caused by the magnetic field. Flux lines are used to show the strength and direction of the force of a magnetic field. They are tensor quantities (that is, they have both magnitude and direction) and they are drawn parallel to the direction of the magnetic force. The spacing of flux lines is called the flux density. A large number of flux lines represents a strong magnetic field. Flux Field Around a Magnet The figure at right illustrates the flux lines around a magnet and its poles as calculated using finite-element analyses [Brauer88, Trowbridge91]. The magnet is indicated by the light colored bar near the top of the figure. The curved gray regions attached to the poles are steel pieces, which can be used to channel magnetic flux in a particular direction. The flux lines are the curved lines from the poles, through the steel and surrounding media. For the case shown, some of the flux lines go directly between the poles, but most pass through and between the steel pieces.

Flux Field Around a Magnet in Contact with a Pipe When a magnet is placed next to a pipe wall, most of the flux lines pass through the pipe wall. That is, the pipe wall is a preferred path for the flux. While most of the flux lines concentrate in the pipe wall, a few pass through the surrounding media. The lines that do not pass through the pipe wall are referred to as the air-coupled field or, for gas transmission pipelines, the gas-coupled field.

Flux Field Around a Magnet in Contact with a Pipe with a Defect Flux leakage at a metal-loss region is caused by a local decrease in the thickness of the pipe wall. At a metal-loss region, the flux carried by the thin section is less than that carried in the full wall. Flux leaks from both surfaces of the pipe. In addition, the shape of the gas-coupled field is changed. A sensor positioned on the inside (magnet side) of the pipe is typically used to measure the magnetic field adjacent to the pipe wall. At a metal-loss region, the sensor records a higher flux density or magnetic field, which indicates the presence of an anomaly. In this manner, an MFL tool detects an anomaly that causes flux to leak. The measured leakage field depends on the radial depth, axial length, circumferential width, and shape of the anomaly, as well as the magnetic properties of the nearby material. To characterize the anomaly, the measured leakage field must be analyzed.

MFL Process Flow

MFL tools apply the principles of flux leakage inside a pressurized and flowing gas-transmission pipeline. A magnetizing system applies a magnetic field along a length of pipe as the tool moves through the line. Defects distort this applied field, producing flux leakage. Sensors measure flux leakage, and a recording system stores the measurements. Last, the measurements are analyzed to estimate the defect geometry and severity.

Inspection Objectives

MFL inspections are typically used to detect, locate, and characterize metal-loss and other anomalies in natural gas-transmission pipelines. There are many types of defects, and not all of these anomalies can be detected or characterized by MFL.

Typical Pipeline Defects

MFL is most often used for detecting and sizing metal loss. The severity of a metal-loss region is a function of its geometry, the pipe geometry, and its mechanical properties. Standard criteria, such as ASME B31G ^{[Kiefner72, ASME B31G]} and RSTRENG ^{[Kiefner89, Vieth93]} , have been developed for estimating the failure pressure of metal-loss regions. Other criteria have been, or are being, developed for other types of defects ^[Stephens99] . Understanding failure criteria is important in order to understand the detection and characterization accuracy requirements for MFL tools.

Detection and characterization requirements should be based on the condition of the pipeline and on the operator's maintenance and repair strategy ^{[Grimes96, Hodgeman96, Nestleroth99, Transportation Research Board88, Turner96, U.S. Government Accounting Office92, Ulrich96]}. Some operators are interested in identifying locations where defects are forming, and they place a strong emphasis on detecting small imperfections that can grow into defects. Others are more interested in identifying large defects that may affect the current integrity of a line, and they place a stronger emphasis on sizing or characterization accuracy. High detection reliability is almost always needed, particularly for defects that threaten the integrity of a pipeline. Good characterization accuracy is needed when inspection results are used to prioritize sites for field investigation or remedial action.

Detection Threshold Requirements
Characterization Accuracy Requirements

Accurately determining the location of a defect is needed for field assessments and repairs. Identifying pipeline features such as girth welds, wall thickness changes, valves and off-takes can help in the location of defects and verifying the accuracy of as-built and maintenance documentation. Typically, requirements on location accuracy depend on the difficulty with which excavations are made and the ease with which marker systems can be placed during an inspection.

Location Accuracy Requirements

False calls are indications that are incorrectly classified as anomalies. False calls can be minimized by proper pipeline feature identification and analysis. Missed calls are the opposite of false calls. Missed calls are far more serious and can result from blind areas due to high velocities, mechanical failures, and failures of sensors or data acquisition systems. Typically, there is a trade off between false-call and missed-call accuracies.

False-Call and Missed-Call Accuracy Requirements

MFL Inspection Tool Components

MFL concepts are simple, but their application in gas-transmission pipelines requires sophisticated inspection tool technology. MFL tools for in-line inspection of pipelines are self-contained units incorporating a number of related systems. The tools can be either segmented, with two or more pieces joined by flexible connectors, or single piece, where all components are contained in a single, rigid package.

Typical MFL Tools (courtesy of Pipetronix)

Shown above are three typical MFL tools. The tool in the foreground is a segmented tool, where six individual segments are joined with flexible connectors. The flexible connectors between the segments allow data and power transfers. The tool in the background is a single-piece tool, where all of the components are contained in a single, rigid package. A two-segmented tool is shown between the single-piece and segmented tools. Single-piece tools are usually longer than segmented tools. Typical single-piece tools are 7 to 10 feet long, while segmented tools are 7 to 16 feet long. Some specialty tools are up to 30 feet long.

Bends in pipelines limit the maximum length of a tool or its segments because long tool segments cannot pass through tight bends. Segmented tools are commonly used in small-diameter lines, where space and bend requirements preclude the use of longer rigid tools. Segmented tools are also used for larger-diameter lines with tight bends. Some pipeline operating companies believe that segmented tools raise the risk that a tool can become stuck at a pipeline connection, where two lines intersect in a tee configuration. So, these companies must balance the increased flexibility of a segmented tool with the perceived risks of a stuck tool.

Single-piece and segmented MFL tools incorporate the following systems:

• Drive System. Gas pressure pushing on a drive cup at the front of the tool propels the cup, which in turn pulls the rest of the tool through the line. The amount of pressure needed to move a tool through a line depends on the age and condition of the cups, the weight and magnet strength of the tool, the presence of pipeline features such as bends, valves, and dents, and the internal condition and dryness of the line.

 

• Magnetizing System. Either permanent magnets or battery-powered electromagnets are used to magnetize the pipe being inspected. The ends of the magnet are connected to metal brushes or plates that rub against the wall and transmit the magnetic field to the pipe.

 

• Sensor System. A sensor system records the leakage field during the inspection. A change in the leakage field indicates a possible defect.

 

• Data Conditioning and Recording System. Data condition and recording systems process and store the sensor measurements for later playback. Data systems are either analog or digital.

 

• Power System. Most MFL tools use rechargeable battery systems to provide power for the sensor, data conditioning, and recording systems.

 

• Other Systems.

Running an Inspection Tool

Getting the MFL inspection tool into and out of a pressurized pipeline requires special components. Most commonly, the devices are called pig launchers and receivers and are installed at compressor stations or other easily accessible locations.

During an inspection ^[Fisher98], control of the gas and tool velocity is important for providing good results. Tool position can be monitored during the run with in-line or external sensors. Monitoring the tool's position is important in the event that a tool becomes stuck.

After the tool is captured in the receiver, the tool is inspected to verify that all components are in working condition at the end of the run. In addition, some of the data are examined to determine whether the tool operated successfully throughout the run. The data are then downloaded, checked for quality and completeness, and analyzed

Launching, Running, and Retrieving MFL Tools

Implementing MFL Technology in Pipelines
Factors That Affect Capabilities

A number of factors affect MFL detection and characterization accuracy. These factors can be grouped in five areas:

• Magnetization: the relationships between the magnetization system, the pipe material, and the applied flux field.

 

• Leakage: the relationships between the applied flux field, the anomaly, and the leakage field.

 

• Measurement: the relationships between the leakage field, the sensor, and the measured signal.

 

• Recording and display: the impact of recording resolution and playback methodologies.

 

• Analysis: the process used to classify anomaly types and characterize metal-loss geometries from the measured signal.

The output or results of each area affects the input and results of the next area. In addition, all five of these areas have theoretical capabilities and limitations. In designing commercial MFL inspection systems, inspection tool designers try to reach these limits within economic constraints.

Magnetization

The magnetization system in an MFL tool applies a magnetic field to the pipe material that interacts with anomalies to produce flux leakage. The design goal for a magnetization system is to produce a magnetic field that is

• strong enough to cause a measurable amount of magnetic flux to leak from the pipe at anomalies,

 

• uniform from inside to the outside surface of the wall thickness so that the measured signal is more linearly related to anomaly characteristics, and

 

• consistent in magnitude along the length of a pipe so that flux leakage measurements can be compared at different locations during an inspection run.

In general, detection is most strongly affected by the field strength, while good characterization requires a field that is strong, uniform, and consistent. The applied field is defined by relationships between the magnetizing system and the pipe material, and variations are introduced by operating parameters such as velocity and stress. The following sections summarize the key relationships between the variables that impact magnetization.

Background

The relationship between the applied magnetic field and the flux density in the pipe is nonlinear. At low applied field levels, a small change in applied field produces a large change in flux. At medium levels, the relationship is highly nonlinear. At high levels, large changes in applied field produce small changes in flux.

Typical Magnetization Curve

MFL requires that magnetic flux be diverted out of the pipe at an anomaly. The presence of an anomaly does not guarantee that flux will leak. For example, corrosion causes a reduction in the amount of flux carrying material, but the reduction in material alone may not cause flux leakage because the remaining material may still be able to carry all of the magnetic flux.

An essential factor for flux leakage is a change in permeability. Permeability is a measure of the ability of magnetic flux to diffuse through (or permeate) a magnetic material. It is related to the slope of the magnetization curve. A reduction in wall thickness coupled with a reduction of permeability causes the flux to flow in alternative paths. One such path is out of the material, hence flux leakage.

In flux leakage testing, the term saturation is often used to imply permeability is decreasing and flux leakage is occurring. Saturation is defined in this report as the magnetization level beyond which an increase provides no significant change in flux density. It occurs after the peak in permeability and beyond the knee of the magnetization curve.

Background Information on Permeability and Saturation

Using this definition, the magnetization curve can be divided into three sections:

• A low magnetization level, below knee of the magnetization curve and below saturation
• A medium level, at the knee of the magnetization curve and near saturation
• A high level, above knee of the magnetization curve and above saturation.

Applied Magnetic Field Strength

Flux Leakage at Three Magnetization Levels

As expected, magnet strength has the strongest impact on the applied field. The magnetization systems in corrosion tools are usually designed to produce magnetic saturation in the pipe wall so that a reduction in material will cause flux to leak. In mechanical damage tools, the magnetization system may be designed to produce lower levels.

For a given magnet strength, an increase in wall thickness will decrease the flux density in the pipe. So, the strength of the magnetization system must be tailored to the wall thicknesses of the pipe to be inspected. Thick-walled pipe can be difficult to inspect because it requires a high magnet strength to attain saturation. Also, inspection results from heavy wall pipe used at road crossings can be difficult to interpret because the flux density is different than that in the rest of the pipeline.

Variations in wall thickness will change the applied field strength, especially when the tool is designed to operate at medium magnetization levels. Typical wall thickness variations in welded pipe are small, but variations in seamless pipe can range from 5 to 20 percent. These variations increase or decrease the applied flux density.

Further Information on the Effects of Magnetization Level on Flux Leakage

Other Parameters Affecting Applied Field Strength

A number of other parameters affect the applied field. These parameters include

• Material Property Variations: Permeability variations arise from small changes in carbon content, alloying elements, and impurities. The variations between batches or heats can be significant ^{[Kiefner96, Nestleroth98a]}. These variations make data interpretation difficult for systems that are designed to operate at medium magnetization levels.
  
  

• Coupling: Brushes are used to couple the magnetic energy into the pipe. The applied flux density depends on the coupling efficiency between the magnetizing assembly and the pipe and on the local wall thickness. Brushes are less efficient than solid steel at coupling magnetic energy into the pipe wall, and shorter brushes provide higher magnetic fields ^[Eiber90]. Long brushes provide better tool flexibility to negotiate pipeline obstructions.

  Bends can cause a decrease in coupling efficiency, which will decrease the applied flux density. If the applied flux density is slightly above saturation, a small decrease in coupling can reduce the applied level to below saturation.
  
  

• Pole spacing: Short pole spacings produce higher magnetization levels and enable the negotiation of tighter bends, but they make signal analysis more difficult. Long pole spacing provides uniform magnetic field and wide areas for sensor placement ^[Eiber91]. Long pole spacing also requires stronger magnets.

  Background Information on Pole Spacing Effects
  
  

• Velocity: All electrical systems, from car alternators to power generation stations, rely on the physical principle that a changing magnetic field passing by an electrical conductor will induce a current in the conductor. An MFL tool moving down a pipeline represents a changing magnetic field, and the pipe is an electrical conductor. As a result, currents are induced in the pipe. Applied magnetization levels decrease as velocity increases, with the largest changes in flux leakage at speeds exceeding 4 to 6 mph ^{[Nestleroth96b]}.

  The Effect of Velocity on Applied Fields
  
  

• Remanent Magnetization: Remanent magnetization is the magnetic field remaining in the pipeline after previous inspection. Remanent magnetization can affect the magnetization level of the current inspection, especially when low to medium magnetization levels are used ^{[Nestleroth95b]}. High magnetization levels are often used for corrosion inspections because they reduce the effects of remanent fields, as well as those of stress, material property variations, and velocity.

  Basic Effects of Remanent Magnetization
  

Leakage

When a magnetic field in a pipeline encounters an anomaly such as a metal-loss defect, flux is diverted or leaks. Sensors measure part of the leakage field: the leakage into the interior of the pipe. The leakage field around a defect can resemble the defect, but it usually does not have the same shape. So, the shape of the leakage field is not necessarily a good indicator of the shape of the defect. Also, the location of the defect, for example on the inside pipe wall versus the outside pipe wall, affects leakage.

Examples of flux leakage field for different defect shapes are given in the following link. These examples illustrate some of the difficulties in trying to estimate the geometry of a defect from the leakage field.

Flux leakage from various metal loss defects

Metal-Loss Defects

When an MFL tool encounters a metal-loss defect, flux is diverted. Flux is diverted in the pipe wall, around the defect, and out of the pipe at the inner and outer diameter. The amount of flux that is diverted out of the pipe depends on the geometry of the defect. The primary variables that affect the flux leakage are the ones that define the volume of the metal loss: Depth - the maximum wall thickness that has been removed (by the corrosion process, third parties, etc.)   Length - the axial extent of the defect   Width - the circumferential extent of the defect Other variables that can significantly affect flux leakage include: Sharpness - the shape of the transition from nominal wall thickness to maximum depth (as viewed in an axial-radial plane)   Roundness - the plan shape (as viewed in an axial-circumferential plane)   Orientation - cracks aligned with the applied magnetic field are not detectable while cracks transverse to the field can sometimes be detected, depending on other geometric parameters   Locations of adjacent defects - Proximity of neighboring defects and pits in general corrosion patches affect the flux leakage   Stress and strains - Stresses and strains make a material easier or harder to magnetize, changing the distribution of flux around the defect.

Depth   MFL Signals versus Depth The amplitude or magnitude of an MFL signal is strongly related to defect depth. Defect depth, the maximum wall thickness that has been removed, is usually specified as a percentage of the nominal wall thickness. Quantifying the depth of defects is important for pipeline serviceability calculations using formulae such as B31G [ASME B31G]. The figure at right shows MFL signals measured in the axial direction as a function of depth with length, width and other defect variables constant. The output of a single sensor through the center of series of metal-loss defects shows that flux leakage is proportional to defect depth, keeping all other variables constant. While the relationship between depth and amplitude appear nearly linear, the significant effect of the other variables on signal amplitude negates this supposition. Accurate depth predictions require an understanding of the relationship between signal amplitude and defect depth. They also require an understanding of how other parameters affect amplitude, so that their effects can be accounted for in the analyses.   Radial Flux Leakage Signals as a Function of Depth

Width   Flux Flow Around a Defect Magnetic flux has a tendency to remain in the pipe. So, flux spreads in the circumferential direction, making the flux leakage field more elliptical than the defect. This effect is called blooming. When the path around the defect becomes large, as for defects that are wide (several times the nominal wall thickness), the effects of blooming become less significant and more flux leaks at the center of defect.

MFL Signals versus Width Narrow defects cause less flux leakage than wide ones for defects with the same depth and other geometric parameters. The figure at right shows the output of a single sensor through the center of series of metal-loss defects ranging in width from 0.25 inches to full circumferential extent. As the defect becomes narrow, the flux leakage drops dramatically. The effects of width also depend on defect depth. There is less blooming for shallow defects than for deep defects of the same width.   Radial Flux Leakage Signals as a Function of Width

Length   MFL Signals versus Length The length of the flux leakage field is related to the length of the defect. The figure at right shows the output of a single sensor through the center of series of metal-loss defects ranging in length from 0.25 inches to 6 inches. The figure shows that defect length also affects the amplitude of the flux leakage signal, with longer defects having lower flux leakage values. This is a significant property of flux leakage since longer defects can be a greater threat to pipeline integrity than shorter defects. A simple signal analysis procedure that identifies the highest flux leakage amplitude defects as the most severe would incorrectly classify longer defects as less severe. The variables that most significantly affect the accuracy of length estimation are sharpness and plan shape, which are discussed in next two sections. Depth and width do not as strongly affect length estimation.   Radial Flux Leakage Signals as a Function of Length

Sharpness   MFL Signals versus Sharpness Sharpness is defined as the angle of the transition from nominal wall thickness to maximum depth. The figure at right shows a series of axial MFL signals through four defects with different sharpnesses. The signal amplitude is larger for more gradual defects with the same volume of metal loss, and less for more sharp defects. In addition, the length of the flux leakage field is less for more gradual defects. In general, the length of the flux leakage signal is better related to the average length of the defect than it is to surface length of a metal-loss defect. The average length is defined by Average Length = Cross-Section Area / Depth The difference between average and surface length can be problematic when attempting to correlate field measurements with inspection results.   Radial Flux Leakage Signals as a Function of Sharpness

Metal-Loss Roundness   MFL Signals at Various Defect Roundnesses Defects that are squarish in shape, as sometimes occur near gaps in wrapped coating, can produce flux leakage patterns that have strong signals at the edges and low levels at the center. These can be misinterpreted as two distinct short defects providing inaccurate defect assessment.

Metal-Loss Location The location of an imperfection or defect on the inside or outside surface affects the flux leakage field. Metal-loss anomalies on the inside pipe surface produce stronger signals for the same depth. Many inspection vendors incorporate separate sensor systems to determine the surface on which the anomaly is located.

Complex Metal-Loss Defects   MFL Signals at Various Multiple Defects The proximity of neighboring defects and pits in larger corrosion patches affects the flux leakage. The result can be inaccuracies in the interpretation of the geometry. The two figures at right illustrate the interaction effects. In the first figure, multiple 1-inch long, 1 inch wide, 50 percent deep pits are arranged in various configurations. The pits in the hoop direction (shown in the upper lefthand corner of the figure) have the most interaction, with this pair of pits appearing as a single, wide defect. In contrast, the pits aligned in the axial direction (upper right) are clearly distinguishable. The pits aligned on a diagonal (lower left) have the least amount of interaction. The pits shown in the lower right exhibit a combination of the upper two effects.

MFL Signals at Various Defects within Defects In the second figure, multiple 1-inch long, 1-inch wide, 50 percent deep pits are arranged in various configurations inside a 3-inch long, 3-inch wide, 20 percent deep patch. This configuration resembles that found in many real-world inspection conditions. Identification of the individual pits within the larger corrosion patch is difficult. (Compare these figures to the plot for the 3-inch long, 3-inch wide, 20 percent deep patch shown earlier, under Defect Roundness.) While the pits produce changes to the signal from the patch, analysis is complicated by the overlap of all of the signals. Identifying and quantifying the various defect parameters is quite difficult from these images.

Other Types of Defects   MFL Signals at Various Defects MFL is capable of detecting many different types of defects, including metal loss, dents, and mechanical damage. However, MFL does not reliably detect all of these defect types. Detection depends on the design of the inspection tool and the sophistication of the analysis procedures, as discussed later. MFL signals for metal loss, dents, and mechanical damage are fundamentally different [Davis96, Davis97]. These differences can be seen in the experimental MFL signals shown at right. The signals correspond to the axial component of the MFL field. Overview of MFL Signals for Metal Loss, Dents, and Cold Work Further Developments on the Use of MFL for Mechanical Damage

Other Sources of Flux Leakage Other pipeline anomalies and features produce flux leakage. Girth welds, valves, off-takes, wall thickness changes, sleeves and other pipeline features are detectable using flux leakage.

Other Parameters Affecting Flux Leakage   MFL Signals as a Function of Velocity A number of parameters affect flux leakage. Most of these parameters also affect the applied field, as discussed earlier. The leakage effects are in addition to the applied field effects. The parameters include: Velocity: Currents that are induced in the pipe by the movement of an inspection tool affect the leakage field, typically reducingit. These effects are greatest at low to medium magnetization levels and for shallow defects [Nestleroth96b].   Basic Theory of Velocity Effects The Effect of Velocity on Flux Leakage Fields

MFL Signals as a Function of Stress Stress: Applied and residual stresses affect the magnetization curve, which in turn affects flux leakage. Similarly, plastic strains affect leakage. As expected, these effects are largest in high-pressure lines and where there is significant secondary loading. They can also be significant when sizing defects in or near dents and attachments.   Basic Stress Effects Effects of Stress on Flux Leakage

Remanent Magnetization Effects Remanent Magnetization: Remanent magnetization also affects the flux leakage field, especially when low to medium magnetization levels are used [Nestleroth95b]. As shown at right, signal amplitudes can vary by 10 to 20 percent compared to the values attained from unmagnetized pipe. These effects tend to plateau after several inspections, after which the amplitudes remain relatively constant.   Effects of Remanent Magnetization on Flux Leakage

Measurement Variables The sensor system on an MFL tool measures the flux leakage. The measurement system converts the leakage field into an electrical signal that can be stored and analyzed. All sensor systems filter and average the actual field, and all measured signals include noise. Thus, the measured field and the actual field are not the same. The design of the sensor system has two goals. The first goal is to provide enough information to allow the signal to be analyzed for detection and characterization of defects. The second goal is to produce a manageable amount of information. Often, these two goals conflict: the amount of information needed to detect and characterize all indications may not be manageable. Therefore, engineering compromises are usually necessary.

Sensor Type   Sensor Output The two types of sensors most commonly used in MFL tools are induction coils and Hall elements. Coils measure the rate of change of a magnetic field, while Hall elements measure the actual magnetic field strength. Historically, induction coils have been the most commonly used type of sensor on MFL inspection tools because they do not require a power source. Instead, a voltage is generated in a passive coil of wire or printed circuit as it passes through a changing magnetic field. A recording device measures this voltage, which is proportional to the change in flux density. Since a coil responds to a change in flux density, the output of a coil is a function of the speed at which it is moving. Integration techniques can be used to convert coil measurements to flux density measurements, but the constant component is lost. The constant component is needed to determine the applied magnetic field strength. Newer MFL tools often use Hall elements. Hall elements, coil sensors, measure the magnetic field directly. The most common type of Hall element directly converts the magnetic field level to an output voltage. Field and flux density are related by a constant in air, and the output voltage of a Hall-element is directly proportional to the flux density. Further Information on Sensor Types Sensor Orientation Flux leakage is a vector field. So, it has three unique components that can be measured. Because MFL tools inspect pipe, a cylindrical coordinate system is used, with the components referred to as the axial, radial, and circumferential. In MFL tools, the radial and axial components are most commonly measured. The third component, in the circumferential direction, is rarely used because flux leakage levels are small and the signals are difficult to interpret. Further Information on Sensor Orientation

Circumferential Size   Sensor Position Effects Sensor size directly impacts the resolution of the measurement system. All sensors have an axial length, circumferential width, and radial height, and they provide an average measurement of the flux passing though the sensor. The resolution of a system is defined by the circumferential width of the sensor. The use of narrow sensors improves system resolution by providing more signals for analysis from a given metal-loss region. When the sensor width is on the order of or greater than the width of a defect, flux leakage levels may not be properly measured. In general, accurate characterization of general wall thinning or defects that occur over a large percentage of the pipe circumference is possible using wide sensors spread about the circumference. Damage processes that leave short defects, narrow pits or pinhole defects, require small sensors for accurate defect detection and sizing.   Further Information on the Effects of Sensor Size

Axial Position The position of sensor with respect to the magnets can also affect the measured signal, and it affects the sensitivity of the inspection results to tool velocity. Effects of Axial Sensor Position The location at which measurements are made affects the shape of the measured leakage field. When multiple sets of sensors are used, the sets are sometimes staggered axially to provide 100 percent coverage. Because axial location affects the measured leakage fields, analysis will be more difficult when staggered systems are used. For a static or slow moving system, a sensor located midway between the poles measures a symmetric signal for a symmetric metal-loss region. Away from the midpoint, the measured static signal is asymmetric. As shown above, moving the sensor toward the front or back pole causes the signal peaks to shift up or down. The effects of axial sensor position are a function of inspection velocity, which amplifies and introduces additional sources of asymmetry. Asymmetry is important because it makes interpretation of the inspection log more difficult.   Further Information on the Effects of Axial Sensor Position

Sensor Liftoff   Effects of Sensor Liftoff The separation between the MFL magnetizers and sensors (referred to as) and the steel piping affects the inspection results. Liftoff is caused by internal deposits and/or liners that can be over an inch thick. Liftoff affects both the magnetization level and the signal shape.   Further Information on the Effects of Sensor Liftoff

Recording and Displaying MFL Data

MFL pigs record flux leakage at specified intervals in both the axial and circumferential directions in the pipe. The data interval in the circumferential direction is defined by the number of sensors. Some older MFL tools have sensor spacings of several inches, while the latest generation inspection pigs have an order of magnitude more sensors. A high-resolution 24-inch pig will typically tool will have between 150 and 300 sensors, thus the circumferential data interval be between 0.25 and 0.5 inches.

The axial data recording interval is defined by the data recording system, and is usually between 0.1 and 0.2 inches (2.5 -5.0 mm). Over a billion flux leakage measurements are required for a 100 mile pipe inspection using a pig with 200 sensors and a data recording interval of 0.1 inches.

Typical Data Storage Requirements

The flux leakage data record or "log" must be examined to detect the presence of possible defects. After a possible defect has been found, the log must be further analyzed to characterize the geometry. The detection and sizing process is usually performed manually, although computer automation techniques are beginning to be implemented.

Many display methods have been developed to aid log analysts in the process. Detection starts with visualization of the flux leakage data the over a large area. Once defects are identified additional data display methods are used including strip chart recording and computer generated displays in pseudo color and three dimensions either with wire frames or in color.

Additional Details on Data Visualization

Libraries of Defect Signals

Selected examples have been presented throughout this report to illustrate consequence of a variable or inspection parameter. Addition insight into the nature of flux leakage can be attained when comparing the signals from many defects and inspection variables.

The GRI Pipeline Simulation Facility ^{[Eiber90, Eiber91, Nestleroth95a, Nestleroth96a, Bubenik95b, Bubenik99]} has the equipment and defects sets needed to demonstrate how various parameters affect MFL signals. The MFL test bed vehicle ^{[Nestleroth96a]} was used to collect data from the hundreds of metal loss defects ^[Koenig95b] at the facility. Data from selected tests have been complied into libraries of defect signals. These results illustrate flux leakage for many defect geometries and inspection conditions.

MFL Test Bed Vehicle
Metal Loss Defects Sets

The flux leakage maps in the libraries can be accessed through the links given later in this section. Each link calls up a table of defects for one library. By clicking on a defect number, the defect's MFL signal will be presented in a topographical display. The color scale for the display is fixed within each library and depends on the dynamic range of the signals within that library. For example, the library for the metal-loss detection set has a 4 gauss per color change scale, while the library for characterization set, which has larger defects, has a coarser 10 gauss per color change scale.

The display maps have a grid superimposed on the top of them to aid in measurement of defect width and length. Each shows a photograph of the metal-loss region and a description of the defect geometry. A color scale is also shown, along with notations that list the highest and lowest recorded signal amplitudes.

There is a tutorial available to assist in interpreting MFL signals. This tutorial also reviews basic information on MFL signals and the parameters that affect them:

MFL Tutorial

The defect libraries are:

• Metal loss characterization library - This library contains larger metal-loss defects and defects that produce large flux leakage amplitudes. Many of the defects have sufficient size to require either a pressure reduction or repair in pipelines operating at 72 percent SMYS. Some of the defects would not affect the serviceability of the pipeline but have a geometry that produces larger flux leakage. Other defects have geometric parameters that are close to ones that are severe or produce large signals. The defects range in depth from 20 to 80 percent, and in length and width from 1 to 6 inches. These defects are useful in developing corrosion sizing methods.

Metal-Loss Characterization Library

• Metal loss detection library - This library contains smaller metal loss defects and defects that produce smaller flux leakage amplitudes. Most of these defects have depths less than 20% or lengths less than 1 inch (25.4 mm). Some of these defects are axially long and circumferentially narrow, which produce low flux leakage signals. These defects are useful in establishing detection threshold criteria and small defect characterization.

Metal-Loss Detection Library

• Metal loss interaction library - This library contains an exploratory set of defects to illustrate the effect of compound defect geometries. These defects, when not used for characterization function development, are useful in developing a general understanding on the effect of neighboring defects.

Complex Metal-Loss Library

 

Some of the defects are included in all three libraries so that qualitative comparisons can be made. However, quantitative comparisons should be avoided because either the sensors or the magnetizer configuration for the three libraries are different. Compensation for these variables would have to be applied to ensure direct comparability.

Analysis of Flux Leakage Data

The last step in an MFL inspection is analysis. Analysis is the process of estimating the geometry or severity of a defect (or imperfection) from the measured flux leakage field. The techniques and success of analyzing MFL data depend on the capabilities and limitations of the MFL tool ^{[Johnson96, Roche96, Smith96]} , which are established by design and operational trade-offs. Typical design compromises include selecting a shorter magnet pole spacing to provide better ability to pass through tight bends or larger lift-off (wear) plates to provide longer inspection runs.

The interpretation of MFL signals is difficult because there is not a simple relationship between the signal shape and the defect geometry or severity. Characterization is compounded by inspection variables associated with inspection including flow velocity, remanent magnetization, variations in the steel properties, and operating pressure. The goal of this section is to show the characterization capability of analysis techniques that would be typically used to analyze MFL in-line inspection data.

Performance expectations, like inspection requirements, cover location, detection, and characterization accuracies. Each of these is discussed below.

Location Accuracy

Most MFL vendors report that their tools provide location accuracies to within 3 to 7 feet or within 0.1 to 0.3 percent of the distance from the nearest reference point. Inspection tools determine the location of an indication by odometer measurements from known reference points. So, the location accuracy of a tool depends on both the accuracy of the odometer and the location of the reference points.

One pipeline operator recently reported using magnetic reference markers points every 1.5 miles along a pipeline route. A 1.5-mile spacing and a 0.1 percent inaccuracy gives an expected location accuracy of within 4 feet midway between the markers. There are few reports of location accuracy for actual MFL tools. An advanced tool vendor reported that 97 percent of indications were located within 5 feet of the actual condition.

Accurate pipeline drawings with detailed locations of valves, branch connections, and other pipeline features help improve location accuracy. By setting reference points (for example, magnetic markers) each mile or less, an inspection vendor can tailor the location accuracy of its tool to a required value. On lines with many clearly defined reference points, these accuracies can approach several inches.

No significant theoretical restrictions exist on location accuracy other than odometer inaccuracy. Odometer inaccuracies result wear and slip of the wheels.

Detection Thresholds

In general, the amplitude of a flux leakage field is related to the volume of metal loss. Therefore, the threshold of detection or minimum detectable metal-loss region for MFL tools is related to the length, width, and depth of the region.

Several reports have been published giving thresholds of detection for MFL tools. For conventional tools, vendors state that the smallest detectable corrosion pits have depths between 15 and 20 percent of the wall thickness. ^{[AMF] [Mohr]} Similarly, the smallest detectable pits have lengths and widths that are 80 percent of the wall thickness. For advanced tools, the smallest detectable corrosion pits are reported to be 20 to 40 percent deep for one vendor and 20 to 70 percent deep for another. The 20 percent depth refers to corrosion patches with a length and width equal to three times the pipe wall thickness; the 40 to 70 percent depths refer to pits that are one-third smaller.

 

Detection Threshold

Theoretically, the detection threshold should be a function of the flux leakage amplitude compared to the noise and background signal level. Typical pipeline steels have background noise levels of about 3 gauss, but the noise can be as high as 15 to 20 gauss.

Details on Noise and Background Signals

Detection thresholds depend on the signal-to-noise ratio. A small 10 percent deep defect produces a signal that is larger than typical noise levels, but a small 5 percent defect produces a signal that is lost in the noise. So, detection thresholds of 10 percent are attainable for most pipeline steels. Lower thresholds are only possible in quiet steels, and larger thresholds are likely in noisier steels.

Detection Thresholds for Small Defects

Probability of Detection

Most conventional tool vendors do not publicly show information on expected probabilities of detection levels. ^[Mohr] These data are considered proprietary. When published, a single probability of detection value or confidence level is generally given, rather than both.

One advanced tool vendor reports a confidence level of 80 percent for metal-loss anomalies with a length or width greater than the wall thickness of the pipe. ^[Shannon88] This confidence level includes false calls as well as missed defects. So, the actual confidence level on detection may be higher. Several advanced tool vendors report confidence levels that depend on the size of the metal-loss region; one vendor gives a 40 percent confidence level for a region with a length or width equal to the wall thickness and 95 to 99 percent for a region that is three times larger.

In one published report for an advanced tool, a pipeline operator reported on the results of a trial where a tool was run through a line with 79 metal-loss defects. ^[Jones] These metal-loss regions consisted of corrosion pits ranging in depth from 14 to 61 percent deep and corrosion patches from 11 to 52 percent deep. All metal-loss regions were detected, and no false calls were reported. An advanced tool vendor also reported on a program where 33 indications were investigated. ^[Jackson] All of the indications reported by the tool existed, and there were no false indications.

Theoretically, the probability of detection should be set by the magnitude and spread of leakage signals compared to the background signals. If the leakage field is well above the noise and background level, the probability of detection should be near 100 percent. At or near the noise and background level, the probability of detection should drop significantly.

An important consideration in determining the probability of detection during an actual inspection is the presence of "blind spots" or areas where the pipe is not inspected. Blind spots can occur due to excessive speed, sensors bumping off the pipe wall, deposits inside the pipe, sensor failures, electronic failures, and the capabilities of the inspection log analyst or analysis program. Depending on the capabilities of a tool, the presence of blind spots can strongly impact the true probability of detection.

Characterization of Metal-loss Defects

Once a defect is detected, its signal must be analyzed to determine the defect's potential effect on the operation pipeline. Because there is not a simple and direct transformation between flux leakage and defect geometry, many methods have been developed to interpret MFL signals and characterize the geometry of defects. These methods include template matching, statistical methods, and neural networks ^{[Lord77, Mandayam96, Nestleroth96]} . Each method has had varying degrees of success, and each has its own strengths and weaknesses.

The development of a characterization method using statistical methods illustrates the many of issues associated with characterization functions. The most commonly used method of analyzing MFL signals is to make inferences or calculations based on features of the signals.

To determine realistic estimates of the capability of such methods, we used classical mathematical modeling techniques to develop characterization algorithms. First, features of signal, such as peak amplitude, signal duration and sensors responding, were extracted from the recorded flux leakage response. Then statistical methods were used to establish characterization and compensation algorithms.

Overview of Statistical Analyses of MFL Signals

Depth Accuracy

Some inspection tool vendors report defects by categories or ranges of depth or severity. ^{[AMF] [Vetco]} Severe or "Class 3" defects often have an estimated depth greater than 50 percent of the wall thickness. Moderate or "Class 2" defects have depths between 25 and 50 percent or 30 to 50 percent. Light or "Class 1" defects have depths up to 25 percent or from 15 to 30 percent. When accuracies on the classes are reported, they are typically reported to be within 10 percent of the wall thickness. ^[AMF]

Other tool vendors report an estimated depth, rather than a broad classification of severity. ^{[British Gas]} The reported accuracies are typically ±10 percent of the wall thickness with a confidence level of 80 percent. For some advanced tools, software is used to invert the measured signals, providing a contour map of the signal amplitude. These contour maps may be calibrated to be proportional to the defect depth. The inversion process often uses the same basic amplitude-depth relationships used for conventional-tool analyses.

Calculated Depth Accuracy The statistical analyses performed in this project suggest that depth accuracy of 8 percent of the wall thickness (with 95 percent confidence) is ultimately possible for elliptical defects less than 50% deep. However, we could not obtain an accuracy this high. Accurate depth estimation is possible only when the analyses are appropriately compensated for other geometry variables. The best accuracy obtained in the analyses is ±19 percent (for a 50 percent deep defect and with 95 percent confidence). Most of the error is likely due to the width estimation procedure used in the analyses, although it is not clear that better methods exist. The statistical analyses suggest that defect parameters, such as the width-to-length ratio, are particularly important when estimating depth. If depth predictions are made on amplitude alone - that is, if these other parameters are not taken into account - the accuracy plummets. The magnitude of depth estimation error increases with increasing defect depth. Depth estimation can be improved by compensation for inspection variables, but the impact of inspection variable compensation is small compared to geometry compensation for the range of variables considered in this study. Confidence levels are particularly important in defining accuracies. At lower confidence levels (e.g., 80 percent, a commonly reported confidence level), the accuracy appears much greater. A 95-percent confidence level implies that 19 out 20 defects (95 percent) are reported within the tolerance given. An 80-percent confidence level implies that 16 out of 20 defects are correctly reported. Statistical Analyses of Depth Characterization Accuracy

Width Accuracy   Calculated Width Accuracy Width is not commonly reported by inspection vendors, and when it is, it is typically based on the width of the recorded MFL signal. Most, or all, inspection vendors do not report accuracy of their width estimates. Because width-to-length ratio significantly affects the ability to predict depth, accurate width estimates are important. The statistical analysis performed in this project suggests that width accuracy of ±1.5 to 2 inches (with 95 percent confidence) is possible for defects with widths from 1 to 6 inches. Accuracies as low as ±2 to 4 inches are likely with unsophisticated analysis procedures. As with depth estimation, errors in width estimation are due primarily to defect geometry (and/or permanent local pipe conditions). Statistical Analyses of Width Characterization Accuracy

Length Accuracy   Calculated Length Accuracy The length of individual defects is commonly reported by inspection vendors. Reported accuracies are typically with 0.25 to 0.5 inch with no claim on confidence level. The statistical analysis performed in this project suggests that individual defect length can be estimated quite well without compensation for other features. In fact, an individual defect's length seems to be the geometry characteristic most accurately estimated, at least for individual defects. Methods were developed that provided length estimation errors of approximately ±1 inch (25.4 mm) with 95% confidence. Improvements come at the cost of defect detection capability. The errors in length estimation are due primarily to defect geometry (and/or permanent local pipe conditions) and random error, with the two factors switching relative importance with increasing length. Almost no unexplained length variability is attributable to inspection conditions. The defect geometry effects are especially important when multiple defects are in close proximity to each other. While not explicitly evaluated in this study, the accuracy with which the length of individual defects in close proximity to others is expected to be low. Statistical Analyses of Length Characterization Accuracy

Severity Accuracy Severity criteria typically use length and depth estimates or defect profile estimates to determine whether a repair is needed. Industry accepted code and method such as ASME B31G [ASME] and RSTRENG relate defect geometry to severity. For individual pits with well defined edges, length and depth estimates based on MFL inspections are reasonably accurate and severity predictions should be similarly accurate. For larger corrosion patches, length and width estimation is more difficult. There are significant errors in predicting width, resulting in corresponding errors in depth prediction. Errors in estimating the geometry of a defect are compounded in severity calculations. Characterization accuracy is typically reported for individual defects or for the deepest defect within a pipe joint. Individual defects can be reported as a composite or single defect. The effects of such reporting can be significant, especially when several small defects are reported as one large defect. To the best of the author's knowledge, no inspection company as yet provides accuracy estimates for severity calculations. Understanding the accuracy of such calculations is essential to using the results of an MFL inspection to prioritize defects for excavation and repair. Issues and Insights Current Detection Capabilities MFL can detect metal-loss defects in pipelines with good confidence, but operational considerations restrict its use in some pipelines. These restrictions are not limitations of MFL per se. Rather they result from physical constraints such as reduced port valves, or normal variations in operating conditions such as product flow speed. Most metal-loss regions produce a measurable flux leakage that is detectable with typical MFL tools, even for small imperfections that do not threaten the structural integrity of a pipeline. For very shallow, long, or narrow metal-loss regions, the MFL signal can become hard to detect. Extremely narrow defects (for example, electric resistance seam weld corrosion or stress-corrosion cracks) do not produce measurable signals in typical MFL systems. Also, background noise levels and variations in tool speed and remanent magnetization impact the detection threshold. These operational variations occur, for example, after the MFL tool exits a bend or restriction, at which time the tool speed can be quite high. Current Characterization Capabilities Metal-loss defects can generally be detected with MFL tools, but characterization accuracy is also important. Analyses to determine the maximum safe operating pressure of a pipeline require information on the depth, length, and shape of metal-loss regions. As a result, characterization accuracy plays a strong role in an MFL tool's ability to provide results that can be used to estimate maximum safe operating pressures. The ability of an MFL tool to characterize the depth, shape, and length of a metal-loss region depends on the size of the sensors and the sophistication of the data analysis system. Conventional MFL tools have a limited potential for characterizing defects because they typically use large sensors and manual (noncomputerized) analysis systems. Advanced or high-resolution MFL tools, with small sensors and computerized analysis systems, have the potential for more accurate characterization. The characterization accuracy of most MFL tools is highly variable. Most vendors report sufficiently high accuracy on depth and length predictions of individual defects to make accurate serviceability calculations. However the confidence level of the measurement can mean a significant number of defects will not be properly characterized. For example, many vendors state a depth accuracy of ± 10 percent of wall thickness and a length accuracy of ± 0.5 inches (12mm) with a confidence of 80 percent. That is one out of every 5 defects will be characterized incorrectly. This lack of confidence is due to the inherent problems associated with the prediction of defect geometry. Complex shapes, long and narrow grooves, multiple pits, and inspection variables present analysis problems for either the inspector or computer analyzing the log. As a result, it is difficult for pipeline operators to estimate the maximum safe operating pressure of a pipeline on the basis of current MFL inspection reports [Rust96] . For groups of defects or defects within other defects, it is not likely that an accurate ranking of defect severity can be made using present technology. Improved characterization accuracy of MFL tools would allow pipeline operators to better understand the likely severity of reported anomalies. However, there will be an ultimate limit to characterization accuracy. High characterization accuracy is not always needed. The required accuracy depends on the goal of the inspection and on the number of indications found. On lines with few indications, a high characterization accuracy may not be needed if all indications are independently investigated. Conversely, where access to the line is difficult and on lines with many indications, characterization accuracy may be far more important, especially in critical areas. In addition, the required characterization accuracy depends on the depths of the metal-loss regions found. Inaccuracies in estimating the remaining wall thickness directly impact the estimated severity of a metal-loss anomaly. For deep metal-loss regions, errors in depth strongly affect calculated severity for defects. For shallow regions, errors in depth are less important. Areas for Future Developments Although MFL is the oldest technology for the inspection of pipelines, new developments will propel its use for decades. Restricted Lines An area that will receive significant advances is the continued development of tools that can inspect lines that were not previously inspectable [Pikas96, Scrivner96] . These tools will be able to pass reduced port valves, tight bends, miter bends, or other pipeline features that previously restricted inspection. Current restricted-line tools use large sensors, which limit data analysis capabilities. Future tools will match the sensors and data recording capability of a high resolution MFL tools. However, because of the additional variables associated with restricted lines, the defect sizing accuracy of these tools may be closer to conventional MFL tools. Improved mechanical design is needed to inspect some restricted lines. For example, current MFL tools are limited in the tightest bend they can pass through by the pole spacing used in the system. So, pipeline operators must either replace tight bends or wait for the development of inspection tools with shorter pole spacings. While shorter pole spacings would allow the inspection of tighter bends, the signals from such systems are more difficult to analyze, reduces accuracy. In order to allow inspection of tighter bends with the same accuracy as in current tools, future analysis systems will need to account for the effects of velocity and the actual applied magnetic field. Velocity Control Advancements in tool technology will allow inspections of pipelines where the velocity is extremely high and cannot be reduced or where it varies significantly. Active speed control systems could also allow more accurate detection and characterization of pipeline defects during routine inspections. Speed control involves more than just modifying the drive module to enable the flow bypass. The magnetizer module needs to be redesigned to enable sufficient bypass. One speed control option is to use shorter brushes, thus enabling more flow bypass through the center of the magnetizer module. Shorter brushes restrict the ability of the MFL tool to pass diameter restrictions, and other design compromises meet similar operational limitations and trade-offs between accuracy, cost, and flexibility. So, pipeline operators should consider the actual accuracy attained with such systems before using them. Defection of Small Defects Detection and characterization of small metal-loss regions could be improved by using smaller sensors, as in advanced or high-resolution MFL systems, and by using higher magnetization levels. Small metal-loss regions are usually considered imperfections that do not threaten the structural integrity of a pipeline. Detection of small imperfections could help pipeline operators that wish to identify the onset of corrosion damage. Use of Low Magnetic Field Levels Many MFL inspection systems use extremely high magnetic intensities to reduce the effects of inherent material property variations, residual stresses, and many inspection variables. At lower magnetization levels, these variations impede the detection of smaller metal loss defects and characterization of the defect size. This change in signal, often considered noise since it alters the flux leakage response from metal loss defects, may contain information about other defects in the pipeline. Recent technology development has shown the potential of low field measurements to characterize mechanical damage defects. Such developments will continue and be offered on commercial tools in the near term. The ability of these system reliably detect mechanical damage defects and differentiate harmful and benign damage will need to be demonstrated. Current in-line inspection systems predict the geometry of metal-loss defects, which is in turn used to calculate failure stress. The ability of current systems to accurately characterize complicated defect shapes is limited. A design alternative is to inspect at high and low magnetization levels. A significant difference in relative signal levels could indicate the location of the most severe defects. This approach is still conceptual and would require further development to prove its viability. Circumferential MFL Very narrow axially oriented defects, such as cracks and seam corrosion are rarely detected using current MFL technology. This is a limitation of the implementation and orientation of the magnetizing assembly used on current MFL tools. Rather than the current axial orientation of the magnetizing assembly, a circumferential orientation could be implemented along with novel sensor systems to look for axial defects. Inspection systems using this magnetizer orientation are available for special purposes, and circumferential MFL has the potential to become a widely used inspection technology. This technology could be used for many pipeline defects such as corrosion, cracking, mechanical damage and seam weld defects. A benefit should be increased accuracy of sizing and characterization. A limitations of this technology will be identification of defect type and sizing of certain defects. These limitations could be overcome by combining circumferential MFL data with axial MFL data. A separate application of circumferential MFL could be as a screening tool. The results of a screening inspection could be used to determine the need for in-the-ditch sizing of a few defect locations or a high resolution, defect specific tool such as a crack detection tool. While developing a commercial circumferential MFL inspection tool, performance levels as well as other needs, constraints, and options must be clarified for this inspection technology to gain acceptance by the pipeline operators. . . . A significant development effort for MFL will continue at in-line inspection service companies, universities, and as part of pipeline company and gover