Christian Hauge, David L. Atherton
Queen's University
Kingston, Ont.
Recent research now indicates the large effects of stresses on magnetic flux leakage (MFL) defect signals can be explained in terms of stress-induced changes in line pipe magnetic properties.
MFL is the technology used in most instrumented in-line pipeline inspection devices ("smart pigs").
The effects were previously reported (OGJ, July 6, 1992, p. 81; Oct. 27, 1986, p. 86) after tests conducted on a single 36-in. sample with a machined defect.
The crystals in most steels exhibit preferred alignments caused by rolling and other manufacturing processes. This is termed "crystalline anisotropy" which, when combined with the effects of residual stresses left by pipe fabrication, causes the magnetic properties to be different in different directions. This is known as "magnetic anisotropy."
An anisotropic magnetic material can be magnetized more easily in some directions than others. These directions are known as "easy axes." When a field is applied to a magnetically anisotropic material, the magnetic flux induced in the material may not be exactly in the same direction as the field.
Line pipe steels often exhibit magnetic anisotropy which varies greatly as a result of slight differences in the manufacturing processes used in different mills and irregularities or nonuniform processing in even a single pipe length.
When external stresses resulting from bending or line pressure in pipes are applied, they are superimposed on the residual manufacturing stress and may cause the initial magnetic anisotropy and the preferred easy magnetization axes to change.
These changes in magnetic anisotropy caused by applied stress have two effects on the MFL signals (from such defects as corrosion pits):
1. Stress-induced changes in the bulk magnetic anisotropy alter the amplitude of the MFL signal.
2. When stress is applied to a pipe-wall region containing a defect, the magnitude and direction of the stress in regions near the defect are altered by the presence of the defect.
In particular, there are regions near the defect where the stress is increased significantly from the uniform background stress which would be found if the defect were not present. The stress increases occur particularly on the surface of the pipe wall around corrosion pits which are said to act as local surface stress raisers.
These local variations in stress can cause local changes in the magnetic anisotropy which in turn affect the MFL pattern, particularly just outside the surface of the defect. This causes additional changes in the MFL patterns which are greatest for defects on the nearest side of the pipe wall to the detector.
Previous reports showed examples of stress-induced changes measured with the principal applied stress aligned with the magnetic field generated in the pipe wall by the detector magnetizing circuit. This occurs when the pipe itself is bent.
The examples shown here are for the previously used axial magnetic field but with circumferential tensile stress, such as generated by line pressure.
The different relative alignments of field, applied stress, and magnetic easy axis, which depends on the particular pipe material, cause important differences in the effects of stress on the MFL signal. Understanding these effects is important in evaluating the results of an inspection pig run.
Research reported here was conducted at Oueen's University, Kingston, Ont., and was sponsored by Gas Research Institute, Canadian Natural Sciences & Engineering Research Council, and Pipetronix Ltd., Toronto.
Corrosion monitoring
Magnetic flux leakage is the most cost-effective technique for corrosion monitoring of in-service oil and gas pipelines and has been in common use for more than 20 years.
A typical magnetic inspection tool uses circumferential arrays of permanent magnets placed at both ends of the detector section of the smart inspection pig (Fig. 1 [75602 bytes]).
The front magnets are coupled in series with the rear magnets, creating opposite magnetic poles at the front and rear of the pig. The pipe wall is thus excited to close to saturation magnetic flux density.
Defects, such as corrosion pits or other anomalies, tend to impede the flow of magnetic flux, thus diverting some flux into the air regions adjacent to the pipe wall. The resulting anomalies in the magnetic leakage flux are detected with Hall probes or induction coils.
Inspection uses product flow to propel the inspection tool through the pipe from one compressor or pumping station to the next, which may be as far as 100 km. Modern extra-high definition MFL inspection tools record tens of gigabytes of data.
The challenge is to interpret these data for accurate values of defect types, sizes, and geometries so that fracture mechanics calculations can then be used to estimate maximum allowable operating line pressures (MAOP).
Running conditions influence the defect-induced signals as a result of two main factors. The first is tool velocity (OGJ, Jan. 21, 1990, p. 84) which varies during inspection because the pipeline dips or rises to conform to the local topography or because of changes in product velocity or differential pressure.
As tool velocity increases, the amplitudes of the MFL signals typically decrease due to the eddy currents induced in the pipe wall.1 Relatively simple algorithms derived from experimental measurements can be used to correct for velocity effects when interpreting MFL inspection logs.
The second operational factor affecting MFL defect signals is considerably more complex.
Pipelines are operated with line pressures giving circumferential (hoop) wall stresses of up to 75% of yield strength. It has been found that stresses in the steel walls of a pipeline can alter MFL defect-signal amplitudes by as much as 70%, depending on the relative directions of stresses and magnetic field and on the magnetic properties of the particular pipe.2
In addition, stress generally decreases the background noise appearing on MFL logs, since stress tends to reduce the effect of permeability variations in the pipe wall, which tend to obscure the defect-induced MFL signals.3
Pipeline operators need to inspect in-service lines at normal operating pressures in order to minimize losses arising from reduced product throughput. It is therefore important to be able to correct for the effects of line-pressure stresses on MFL signals in order to ensure that they can be interpreted to identify and size defects accurately.
Furthermore, the effects of residual stresses or stresses due to bending on unstable terrain such as permafrost, river, or sea beds should also be considered.
Perpendicular stresses
Most of the previously reported work on the effects of stress on MFL signals has been based on the application of uniaxial stress and magnetic field,3-6 but it is also necessary to consider the effects of stress which is perpendicular to the magnetic field used to generate MFL signals.
The objective of work described here, therefore, was to investigate circumferential (perpendicular) stress effects on axially excited MFL patterns. This would simulate the responses from an MFL inspection tool in a section of pipeline with internal gas or liquid pressure.
The test rig built for this work consisted of four parts: a compact stressing mechanism, a magnetizing unit, a scanning and data acquisition system, and a representative section of line pipe.
Fig. 2 [ 24278 bytes ] shows a cross sectional outline of the apparatus.
The API X-70 pipe, which was destined for use in the Gas Research Institute's pipeline simulation facility flow loop at Battelle, Columbus, was 24 in. (610 mm) OD, 0.360-in. (9 mm) W.T.
The pipe section and the hydraulic stressing mechanism, which used an inner spool piece sealed to the test pipe by compressed elastomer O-rings in order to minimize axial stress, are shown in the center of Fig. 2 [ 24278 bytes].
Line pressure was simulated by controlled hydraulic pressure in the cylindrical annulus between the spool piece and test pipe. The magnetic field, generated by the magnetizer, was in the axial direction.
Three pits
The MFL signals were generated at three ball-milled blind holes simulating corrosion pit defects. They were positioned well away from the weld and separated by a minimum circumferential distance of 300 mm. All the pits had depths of 50% W.T. and diameters of 12.5 mm.
The first pit was machined on the inner surface of the pipe wall at the axial center of the pipe. This was intended to illustrate the conditions for a corrosion pit on the far surface of the pipe wall (a far side defect).
The second pit was machined, without applied stress, on the outside surface of the pipe wall (in this case a near-side defect) at the axial center of the pipe.
Since most corrosion pits develop while pipelines are operational, a third pit was machined on the outside surface of the pipe while the hydraulic pressure was at 6.9 MPa (1,000 psi) corresponding approximately to a wall hoop stress of 220 MPa (45% of yield strength). This also simulated a near-side corrosion pit.
The magnetization unit, shown also in Fig. 2 [ 24278 bytes ], provided the exciting axial magnetic field of the MFL detector. The magnetizers were suspended from a wheeled trolley running overhead on an aluminum frame. This enabled movement along the length of the pipe section.
The magnetizers could also be raised or lowered by a jack and rotated about a pivot. For simplicity, the MFL detector (magnetizer and sensor) was positioned on the outside surface of the pipe.
The magnetizers consisted of sets of high strength Neodymium-Iron-Boron (Nd2Fe14B) permanent magnets providing a unidirectional axial field in the pipe wall, shown in the region of a pit in Fig. 3 [38182 bytes].
Different magnetic flux densities were obtained in the pipe section by varying the number of permanent magnets in each pole arm. Changes in the flux density were also made by doubling the cross sectional area of the back-iron flux return path or by using solid, iron flux connectors instead of the customary steel brushes to couple the magnetic flux into the pipe section.
Results, analysis
The MFL patterns at the three different pits were examined, using various magnetic flux densities and circumferential hoop stresses applied perpendicularly to the axial magnetic field.
Figs. 4 [93126 bytes] and 6 [94212 bytes] show results from scans of radial MFL components measured over 40 x 40-mm axial by circumferential areas enclosing the defects. The data are shown as surface plots, which indicate the senses of the fluxes, and also as contour maps on which the pits are indicated by shaded areas.
Fig. 4a [93126 bytes] shows the surface and contour plot of the radial MFL signal for the far side pit at 0 applied pressure and 1.49T nominal flux density. (T = Tesla, a new S.I. unit of magnetic flux density; 1 Tesla = 10 kiloGauss.)
As the flux density is increased, the circumferential profile of the surface plot increases in amplitude and in the spread, defined by the width at half maximum (Fig. 4b [93126 bytes]).
When circumferential stress is generated in the pipe wall by increasing the hydraulic pressure, the spread of the circumferential profile also increases. This circumferential spread increases by 10% when the tensile hoop stress is increased from 0 to 220 MPa and by 14% when the flux density is increased from 0.88T to 1.49T.
This spreading indicates a flux density increase in the volume opposite the far-side defect, tending to divert flux further away from the pit.
As the stress is increased, the easy axis of magnetization is rotated towards the direction of tensile stress, as described below, and more flux is diverted around the pit. This also produces a larger spread.
Fig. 5 [62617 bytes] shows the change in the normalized peak-to-peak value of the MFL signal as tensile circumferential hoop stress is generated in the pipe wall by the internal hydraulic pressure. The stress dependent peak-to-peak variation decreases monotonically as the flux density is increased from 0.88T to 1.49T.
The decrease in MFL peak-to-peak signals with increasing stress is the result of the flattening of the B-H hysteresis loops observed when applying tensile bulk hoop stress perpendicular to the magnetization direction.7 It is caused by rotation of the bulk easy-magnetization axis towards the applied circumferential tensile stress which diverts more of the pipe wall flux around the defects.8
In general for steel, which has increased length on magnetization (positive magnetostriction), tensile stress tends to give an easy axis of magnetization aligned with the tensile stress, whereas compressive stress tends to give a perpendicular easy axis.
Fig. 6a [94212 bytes] shows surface and contour plots of the radial MFL component from the near-side pit machined without applied stress. The general decrease in MFL peak-to-peak signal can again be explained by the effect of bulk stress either in terms of the rotation of bulk easy-magnetization direction or as the result of stress on the hysteresis loop of the pipe material.
Far, near-side differences
There are, however, differences between the far and near-side defect signals obtained with increasing tensile hoop stress.
In addition to an increase in the peak-to-peak amplitude with increasing flux density, a double peak is apparent on both the positive and the negative sides of this near-side pit signal.
Fig. 6b [94212 bytes] shows the corresponding circumferential profiles at 0 applied stress and various flux densities.
The double peaks are quantified by a saddle factor, defined as the average of the local maxima less the local minimum between the double peaks, normalized to the global maximum value.
For the 1.49T data shown in Fig. 6b [94212 bytes], the saddle factor is 7.3%. For uniaxial magnetic field and stress, the localized stress distribution around a near-side pit has been shown previously to increase the saddle factor.9
Since the current results are obtained with mutually perpendicular magnetic field and stress, the saddle factor would be expected to decrease because of the relative rotation of the stress distribution around the pit.
This is indeed observed and, at a flux density of 1.49T, the saddle factor for this near-side pit machined without applied stress decreases to 5.2% with increasing tensile hoop stress.
It has been suggested that stresses where the applied magnetic field is tangential to the pit modulate the changes in the double-peak feature and therefore the saddle factor.10
At the circumferentially opposed points, where the applied magnetic field is tangential to the pit, the hoop stress is 0 and the axial stress is compressive, resulting in an easy axis of magnetization perpendicular to the field.
This gives decreased flux in the vicinity of the defect which causes the relative decrease in the double peak signature.
Fig. 7 [59362 bytes] graphs the tensile hoop stress' dependence of the radial MFL peak-to-peak signals from the near-side pit machined before application of stress. There is a notable decrease in the relative stress dependence at the maximum flux density (1.49T), but this is not monotonic from the lowest flux density to the highest flux density.
At the maximum tensile hoop stress and a flux density of 0.88T, the radial MFL peak-to-peak signal has decreased to 60% of the initial unstressed value. The maximum decrease is found for a flux density of 1.18T, where the radial MFL peak-to-peak signal has decreased to 50% of the unstressed MFL peak-to-peak value.
The minimum change is found at the highest flux density of 1.49T where the signal has only decreased to 80% of the initial unstressed value.
The radial MFL peak-to-peak signal from the pit machined while stress was applied is 10% larger than obtained from the second near-side pit machined prior to the application of stress.
This effect has been observed previously in experiments where the directions of applied stress and magnetic field were parallel.4 It suggests that there are different surface-stress concentrations induced by pits eroded prior to stressing compared to those induced in the vicinity of pits eroded while stress is applied.
The normalized radial MFL peak-to-peak signal stress dependence measured for the near-side pit machined with an applied 220 MPa tensile hoop stress (Fig. 8 [43450 bytes]) is generally similar to that of the near-side pit machined without applied stress, shown in Fig. 7 [59362 bytes].
The largest MFL peak-to-peak stress dependence is again found for the flux density of 1.18T, where the radial MFL peak-to-peak signal has decreased by 50% at a tensile hoop stress of 220 MPa. As the flux density is increased further, a decrease in stress dependence is observed until, at a flux density of 1.49T, the decrease is to 70%.
The saddle factor at 1.49T and zero applied stress is 10.5% and decreases to 9.6% when a 220 MPa tensile hoop stress is applied. The spread of the circumferential profiles decreases by 8% as the tensile hoop stress is increased and increases with increasing pipe wall flux density.
Interpreting MFL logs
In general, increasing pipe wall flux density increases MFL signals. Variations between similar defects are caused by both local variations in magnetic properties and running conditions.
This work showed that the effects of line pressure stress can be very large so that corrections need to be made when interpreting MFL logs to estimate defect severities.
The effect of line pressure, which gives a circumferential hoop stress perpendicular to the MFL detector axial magnetic field, is to decrease the peak-to-peak amplitude of the radial MFL pattern. This decrease with stress depends upon both the flux density and upon the local easy axis of magnetization.
The greatest stress dependence that was found here was at the lowest flux density (0.88T) with the far-side defect signal. This normalized peak-to-peak signal decreased to only 25% of the initial unstressed value at a tensile stress of 220 MPa (45% of yield strength).
The stress dependence of far side MFL peak-to-peak signals decreased as the flux density in the pipe wall was increased, resulting in a smaller decrease in the normalized peak-to-peak signal to 50% of the initial unstressed value, when applying 220-MPa tensile hoop stress at a flux density of 1.49T.
The near-side pit defects showed generally similar effects, although some of the differences are attributed to differences in the easy axes on the two surfaces. unlike the far-side defect, the stress dependence did not monotonically depend on the flux density in the pipe wall.
The MFL pattern, notably the double peak and saddle factor, which was more distinct for the pit machined while stress was applied, changed as a function of stress. The saddle factor decreased for both near-side pits, machined with and without applied stress, and correlated with the defect edge positions of local surface compressive stress concentrations, which are oriented perpendicularly to the overall applied tensile hoop stress.
Local surface stress concentrations have been shown earlier to modulate the double peak feature of the radial MFL signal.9 The double peak and saddle feature was not observed for the far side defect because it is considered to result from local surface stress concentrations.
References
1. Atherton, D.L., Jagadish, C., Laursen, P., Storm, V., Ham, F., and Scharfenberger, B., "Pipeline inspection-tool speed alters MFL signals," OGJ, Jan. 21, 1990, p. 84.
2. Hauge, C, "Effects of line pressure on axially excited magnetic flux leakage patterns," thesis, Queen's University, September 1995.
3. Atherton, D.L., "Effects of line pressure on performance of magnetic inspection tools for pipelines," OGJ, Oct. 27, 1986, p. 86.
4. Laursen, P., "Effects of line pressure stress on magnetic flux leakage patterns," thesis, Queen's University, December 1991.
5. Barnes, R., "Comparison of the effects of surface and uniform stresses on magnetic flux leakage patterns," thesis, Queen's University, September 1993.
6. Atherton, D.L., Dhar, A., Hauge, C., and Laursen, P, "Line stresses affect MFL defect indications," OGJ, July 6, 1992, p. 81.
7. Sablik, M.J., Riley, L.A., Burkhardt, G.L., Kwun, H., Cannel, P.Y., Watts, K.T., and Langman, R.A., "Micromagnetic model for biaxial stress effect on magnetic properties," Journal of Magnetism and Magnetic Materials, Vol. 132 (1994), pp. 131-48.
8. Leonard, S., and Atherton, D.L., "Calculations of the effects of anisotropy on magnetic flux leakage detector signals," submitted to IEEE Trans. on Magnetics.
9. Krause, T.W., Little, R.W., Barnes, R., Donaldson, R.M., Ma, B., and Atherton, D.L., "Effect of stress concentrations on magnetic flux leakage signals from blind hole defects in stressed pipeline steel," submitted to Research in Nondestructive Evaluation.
10. Little, R.W., "The design of ultra-high resolution read head sensors for pipeline steel defect detection," thesis, Queen's University, May 1995.
The Authors
Christian Hauge is a research engineer with Pipetronix Ltd. in Toronto. He holds a BS (1991) in electronics and computer engineering from Engineering College of Odense, Denmark, and an MS (1995) in physics from Queen's University, Kingston, Ont.
David L. Atherton is professor of engineering physics at Queen's University and leads its applied magnetics research group. He moved to Canada in 1959 and was employed at Ferranti Packard Electric, Toronto, eventually becoming head of superconductivity research.Atherton joined the faculty of Queen's University in 1971. He is a fellow of the Royal Society of Canada and of the British Institute of Nondestructive Testing and is a registered professional engineer in Ontario.
Copyright 1996 Oil & Gas Journal. All Rights Reserved.