Professional Documents
Culture Documents
x
From Materials Evaluation, Vol. 74, No. 9, pp: 1296-1303.
Copyright © 2016 The American Society for Nondestructive Testing, Inc.
ABSTRACT Introduction
Previous studies commonly did analysis of mag- As an important method for the steel pipe detection in
netic flux leakage (MFL) signals at only one line- industry, magnetic flux leakage testing (ML) has been used
widely for years. The essence of ML is to get 3D discontinuity
scan liftoff and could not give comprehensive information through inverse solving techniques based on the
information about the MFL field. This paper uses information extracted from the magnetic field leakage (MFL).
the finite element method to analyze MFL infor- In general, ML includes three fundamental steps: firstly,
mation of surface cracks on pipelines at multiple magnetically saturating the object by applying an appropriate
magnetization; secondly, obtaining magnetic properties
liftoff values. The valley-valley spacing value of around the surface of the object by using magnetic field
axial components and peak-valley spacing value sensors; thirdly, analyzing magnetic field data to extract
of radial components are analyzed at multiple geometry information (position, length, width, depth, and so
liftoff values. Studies show that signals at multi- on) on the discontinuities. Among these three steps, step two
is of supreme importance in that the magnetic properties
ple liftoff values indicate more information than obtained directly affect the accuracy of the detection results
signals at one liftoff. A technique using a double and the evaluation of the security of the industrial steel pipe.
liftoff compensation algorithm is proposed to Currently, most researchers conduct only axial line scans
measure width more precisely. at a certain liftoff value to capture an MFL signal, which is not
enough to get all the magnetic properties. For example, one
KE YWO R DS: NDT, MFL, crack, FEM, liftoff.
study used the peak-valley spacing value (PVSV) to measure
the discontinuity width at a liftoff value of 1.5 mm. (Man-
dache and Clapham, 2003). Others studied the relationships
between the three-axial components and the discontinuity
width via the dipole model technique at a fixed liftoff value of
1 mm and drew the conclusion that it was most precise to use
the valley-valley spacing value (VVSV) of the axial compo-
nent, Bz, to signify the discontinuity width at liftoff of 1 mm
(Kopp and Willems, 2013). A third study used a particle
swarm optimization algorithm for reconstructing the sizes of
rectangular crack at different liftoff but only selected two liftoff
values, 1 and 1.5 mm, respectively (Zhang et al., 2009).
In fact, the MFL field is a 3D magnetic information field in
the spatial domain containing the axial, radial, and circumfer-
ential components, Bz, By, and Bx, respectively. The MFL field
distribution may have an important relationship with the spa-
tial location parameter liftoff value. An outside work re-
searched the liftoff effect and concluded that the PVSV of the
radial component was proportional to the discontinuity depth
and liftoff value (Uetake and Ito, 1986). Another work re-
searched different sensitivity of the radial component at differ-
* Ph.D., School of Electrical Engineering, Wuhan University, China.
ent liftoff values and concluded that the liftoff value has a
† M.E., School of Electrical Engineering, Wuhan University, China. direct effect on the accuracy of the testing result (Dutta et al.,
‡ Ph.D., Huazhong University of Science and Technology, China. 2009). Another group made a special study of the selection
§ School of Electrical Engineering, Wuhan University, China. of the liftoff value in the alternating current MFL detection
(1) ∇×B =0
Figure 1. The software model of the steel pipe detection system.
(2) ∫∫ L B × dl = µ 0 ∑ I i
Ò
i
L Y
With the aid of finite element simulation software, a model By, radial component
of the steel pipe detection system was constructed and is D Discontinuity Bz, axial component
shown in Figure 1. In this model, helmholtz coils were se- W Bx, circumferential component
lected as the magnetic excitation device (Zhang et al., 2007).
The magnetization of the steel pipe was adjusted by regulating Magnetization
the supply current value of the helmholtz coils. The model pa- direction
Z
rameters of the helmholtz coils and those of the steel pipe are
as follows: the coil radius and the coil spacing distance were
both 150 mm; the thickness and width of the coils were 6 and
Steel pipe X
20 mm, respectively; the wall thickness, external radius, and
length of the pipe were 8, 89, and 400 mm, respectively.
To analyze the relationship between the MFL information Figure 2. The three-direction component of the 3D magnetic flux
and the discontinuity size, a typical kind of circumferential leakage field.
14
12
Amplitude
10
PVSV (mm)
6
Liftoff = 0.4 mm
Liftoff = 1 mm
4 Liftoff = 2 mm
Liftoff = 3 mm
Liftoff = 4 mm
Liftoff = 5 mm
2 Liftoff = 6 mm
(b) Position (mm) Liftoff = 7 mm
Liftoff = 8 mm
0
Figure 3. Typical By, Bz signal waveform and definitions: (a) peak-valley 1 2 3 4 5 6 7
spacing value (PVSV); and (b) valley-valley spacing value (VVSV). Width (mm)
Liftoff = 7 mm
Liftoff = 8 mm
ent liftoff value is shown in Figure 5.
10 Figure 5 shows that when the liftoff value is small, such as
8 at 0.4 and 1 mm, the PVSV is almost a horizontal straight line
with little increase in amplitude; but, when the liftoff value is
6
great, like 6, 7, and 8 mm, the wave looks like an approximate
4 diagonal line. It implies that the depth has a slight influence on
2 the difference value between the PVSV and the discontinuity
0
width, and the smaller the liftoff value, the smaller influence
1 2 3 4 5 6 7 the depth has on the PVSV. In Figure 5, PVSV is closer to the
Depth (mm) true width value 2 mm at a depth value of 1 mm compared to a
depth value of 2 mm. Therefore, the measurement result of
Figure 5. Relationship between peak-valley spacing value (PVSV) width would be more accurate with a small discontinuity
and the depth at different liftoff values.
depth. However, in the actual testing, the discontinuity depth,
as an unknown parameter to be measured, could not be
viewed as a criterion for the accuracy of the testing result. In
fact, since the depth affects the PVSV slightly, especially at a
small liftoff value, the testing error caused by the depth could
be ignored within the range of the errors permitted. Actually,
if the depth and the liftoff value are large enough and the error
caused by the depth could not be permitted, a compensation
algorithm should be presented based on the liftoff value and
the peak amplitude because the peak value reflects the depth
of the discontinuity. Further study is shown in the discussion
section.
the straight line. However, the change of the slope and the in- When the liftoff is smaller than 1 mm, the measured PVSV
tercept in Table 1 has no obvious regularity. Figure 8 and Table 2 can be construed as the discontinuity width, while if the liftoff
show that when the width ranges from 1 to 8 mm, the slope of the is big enough, the difference value between the measured
line decreases slightly, while the intercept increases approximately PVSV and the discontinuity width cannot be ignored. So, a
by an equal difference. Therefore, the width has a relative small in- compensation algorithm was presented to get the PVSV at a
fluence on the slope of the straight line, but a great influence on small liftoff value such as 0.4 or 0 mm because the discontinu-
the intercept. The size of the intercept can reflect certain infor- ity width is approximately equal to the intercept of the fitting
mation about the discontinuity width. straight line, referring to discussion section.
8
smaller than 1 mm, two troughs appear symmetrically on both
6 sides of the crest, noted as the negative side lobes, as shown in
Width = 1 mm Figure 9. The simulation research shows that there is a certain
Width = 2 mm
4 Width = 3 mm
relationship between the VVSV for short of the two negative
Width = 4 mm side lobes and the liftoff value. A study of this relationship will
2 Width = 5 mm
contribute to sizing the discontinuity. In addition, the first de-
Width = 6 mm
0
Width = 7 mm rivative of the component Bz also serves as an important ele-
0 1 2 3 4 5 6 7 8 ment for analyzing the discontinuity size.
Liftoff (mm)
Relationship between the Valley-valley Spacing Value of the
Figure 8. Relationship between peak-valley spacing value (PVSV) Axial Component and the Width
and liftoff at different widths.
Based on the preceding analysis, the PVSV of the radial
component, By, followed a linear relationship with the dis-
TABLE 1 continuity width. Likewise, this section explores the rela-
The slopes and intercepts of the fitting lines with different depths tionship between the VVSV of the axial component, Bz, and
in Figure 7 the discontinuity width. The depth was set as 7 mm. The
Depth k B width ranged from 1 to 7 mm. Figure 10 shows the relation-
1 mm 0.964 1.308 ship between the VVSV and the discontinuity at different
2 mm 1.056 1.423 liftoff values.
3 mm 1.126 1.424
4 mm 1.164 1.478
5 mm 1.198 1.502 0.15
6 mm 1.225 1.520
0.14
7 mm 1.250 1.539
0.13
0.12
TABLE 2 0.11
Bz (mm)
The slopes and intercepts of the fitting lines with different widths
0.1
in Figure 8
0.09
Width k B
1 mm 1.264 1.033 0.08
2 mm 1.229 1.730 0.07
3 mm 1.146 2.645
4 mm 1.068 3.590 0.06
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
5 mm 0.990 4.601
Position (mm)
6 mm 0.918 5.622
7 mm 0.846 6.695
Figure 9. The axial component (Bz) and its negative side lobes at
small liftoff values.
2 mm 4.074 5.892
20
3 mm 3.773 7.563
15 4 mm 3.686 8.591
10 Liftoff = 0.4 mm 5 mm 3.589 9.761
Liftoff = 1 mm
5 Liftoff = 2 mm
6 mm 3.543 10.756
Liftoff = 3 mm 7 mm 3.450 11.880
0 Liftoff = 4 mm
Liftoff = 5 mm
Liftoff = 6 mm
–5 Liftoff = 7 mm
Liftoff = 8 mm
–10
1 2 3 4 5 6 7
width at a small liftoff value, the discontinuity length and
Width (mm) depth were set as 8 and 1 mm, respectively, the width ranging
from 1 to 7 mm. The wave data of the VVSV changing with
Figure 10. Relationship between valley-valley spacing value (VVSV) the discontinuity width at different liftoff values are shown in
and the width at different liftoff values. Figure 11. Because the negative side lobe disappeared when
the liftoff value was greater than 6 mm, the liftoff was selected
Figure 10 indicates that the VVSV of the axial component, from 0 to 6 mm. Table 3 indicates the slopes and intercepts of
Bz, follows a linear relationship with the discontinuity width the fitting straight lines in Figure 11.
when the liftoff value remains within a small range, as is shown In Figure 11, as the width increases, the slope of the line
by the six lines in the lower part of Figure 10. However, there decreases slightly, while the intercept increases approximately
exists a huge gap between the VVSV and the width value. At in an arithmetic sequence.
the liftoff value of 0.4 mm, the width is 1 mm, while the VVSV
is 5 mm. The difference between the two values is 4 mm. The The First-order Derivative of the Axial Component
difference value increases as the liftoff value increases. More- Figure 12 shows the waves of the first-order derivative of Bz at
over, the linearity of the three curves in the top is the smallest, different liftoff values, which have a crest and trough similar to
because the negative side lobe disappears when the liftoff the radial component, By. Figure 13 shows the relationship be-
value is greater than 6 mm. tween the PVSV of the first-order derivative of Bz and the
liftoff. Here, the discontinuity width is 2 mm and the depth is
Relationship between the Valley-valley Spacing Value of the 7 mm. In order to make a comparison, another line indicating
Axial Component and the Liftoff Value the relationship between the PVSV of By and the liftoff was
In order to identify the linear relationship between the VVSV added in Figure 13. From Figure 13, it can be known that,
and the liftoff value and its correlation with the discontinuity when the liftoff value is small (for liftoff T < 1 mm), the PVSV
35 0.04
Liftoff = 0.4 mm
Liftoff = 1 mm
0.03 Liftoff = 2 mm
30 Liftoff = 3 mm
Liftoff = 4 mm
0.02 Liftoff = 5 mm
Derivative of Bz
25 Liftoff = 6 mm
VVSV (mm)
Liftoff = 7 mm
0.01 Liftoff = 8 mm
20
0
15
–0.01
Width = 1 mm
10 Width = 2 mm –0.02
Width = 3 mm
Width = 4 mm
5 Width = 5 mm –0.03
Width = 6 mm
Width = 7 mm –0.04
0 0 5 10 15 20 25 30 35 40
0 1 2 3 4 5 6
Liftoff (mm) Position (mm)
Figure 11. Relationship between valley-valley spacing value (VVSV) Figure 12. The waves of the first-order derivative of Bz at different
and the liftoff at different widths values. liftoff values.
8
data compensation algorithm to get the ML data at small
7
liftoff according to the data at larger liftoff values. As is pre-
6 sented in the preceding section, the PVSV of By and the liftoff
5 value have an approximately linear relationship, which can be
4 denoted as a straight line. The slope of the line can be ob-
3 tained by two great liftoff values and the corresponding PVSV,
2
and then the intercept value is the PVSV at liftoff of 0 mm.
This intercept value is the measured discontinuity width, as
1
0.5 1 2 3 4 5 6 7 8 shown in Figure 14.
Liftoff (mm)
0.4 domain, and there exists a certain mechanism between its dis-
tribution law and the discontinuity size. It is not enough to an-
0.2 alyze the signals generated by a limited variety of discontinuity
models through FEM. Further studies will focus on the analy-
0
0 1 0
sis of signal characteristics of various discontinuity models and
2 3 2 1 the experimental data.
4 5 4 3
5
Liftoff (mm) 6
7 7 6 Depth (mm)
ACKNOWLEDGMENTS
Figure 15. Reference map of the relationship between By amplitude This paper was financially supported jointly by the National Natural Science
and depth at different liftoff values. Foundation of China (NNSFC) and the Open Foundation of the Key Labo-
ratory of Nondestructive Testing, Ministry of Education, grant numbers
51475194 and ZD2012290072, respectively.
REFERENCES
Dutta, S.M., F.H. Ghorbel, and R.K. Stanley, “Simulation and Analysis of 3-D
15 Magnetic Flux Leakage,” IEEE Transactions on Magnetics, Vol. 45, No. 4,
2009, pp. 1966–1972.
PVSV (mm)