Measurement xxx (2011) xxxxxx

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

A simple forward direct problem solver for eddy current

non-destructive inspection of aluminum plates using uniform eld probes
A. Lopes Ribeiro a,b,, H. Geirinhas Ramos a,b, O. Postolache a
a
Instituto de Telecomunicaes, I.S.T., Av. Rovisco Pais, 1049-001 Lisboa, Portugal
b
Instituto Superior Tcnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: In this paper we show that a simple algorithm used to model the eddy current inspection of
Received 14 November 2010 an aluminum plate can be used to preview the acquired voltage signals. Thus, the algo-
Received in revised form 11 March 2011 rithm is suitable to work as a forward problem solver to determine the expected measure-
Accepted 11 March 2011
ment signal obtained with a uniform excitation eld probe including a giant
Available online xxxx
magnetoresistor sensor. The algorithm is based on a conformal transformation and is able
to preview the shape of the electrical current lines when a metallic plate with a supercial
Keywords:
straight crack is subject to a sinusoidal excitation eld with constant amplitude and orien-
Eddy currents
Non-destructive testing
tation in a bounded zone around the sensor element. Simple and fast algorithms as the one
Uniform eld probes presented in this paper are of paramount importance for testing inversion algorithms used
Giant magnetoresistors to characterize defects in metallic plates.
2011 Elsevier Ltd. All rights reserved.

1. Introduction rely on the geometrical conguration of the excitation

Different methods of non-destructive testing are cur-
rently used in industry. These methods rely on several
physical phenomena. Methods based on ultrasound propa-
gation [1], on electromagnetic induction [2] and on liquid
penetrants [3] are usually employed to test non-ferromag-
netic materials. When a time-varying magnetic eld is ap-
plied to a conductor an electromotive force is induced and
eddy currents appear inside it. The nature of these currents
depends on the electromagnetic properties of the material,
such as the conductivity, and on the object geometry. Thus,
the eddy current method may be applied to search for
material inhomogeneities, aws, cracks, or any other kind
of material defects that causes perturbations of the eddy
currents [4].
Different authors studied the eddy current phenomena
applied to the non-destructive testing of metallic materi-
als. The main differences between the applied methods
Corresponding author at: Instituto de Telecomunicaes, Av. Rovisco
Pais, 1049-001 Lisboa, Portugal.
E-mail address: arturlr@ist.utl.pt (A.L. Ribeiro).
coils, on the time-varying excitation elds, and on the type
of the sensor elements under use.
The use of circular excitation coils is quite common.
When the coil axis is perpendicular to a metallic structure
and the axisymmetry is preserved, the magnetic eld can
be determined analytically [5]. An excitation coil with rect-
angular cross-section and capable of producing a uniform
eld was proposed to inspect weld aws [6]. The same
geometry was recently proposed to detect the edge of de-
fects [7].
Another important difference between eddy current
methods relies on the time variation of the excitation
elds. Some authors [8,9] use sinusoidal excitation to get
the benets of working in a harmonic regime, but others
prefer to investigate the pulsed excitation [7] working in
the time-domain regime.
The third experimental difference between eddy cur-
rent methods relies on the sensor element. In the simpler
cases, the excitation coil is used as a sensing element and
the coil impedance is measured [5,8]. Other setups use
special pickup coils [7], with different mounting arrange-
ments. Some authors use single absolute pickup coils, but

0263-2241/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.measurement.2011.03.029

Please cite this article in press as: A. L. Ribeiro et al., A simple forward direct problem solver for eddy current non-destructive inspection of
aluminum plates using uniform eld probes, Measurement (2011), doi:10.1016/j.measurement.2011.03.029
2 A.L. Ribeiro et al. / Measurement xxx (2011) xxxxxx
others prefer differential arrangements to subtract the
voltage induced by the primary excitation eld. The main
drawback of the use of excitation coils is the necessity of
using higher frequencies, because the induced voltages
are proportional o the magnetic ux time derivative. More
advanced apparatus for eddy current inspection use high
sensitivity devices to sense the magnetic eld directly.
These devices are usually Hall sensors [9] or giant magne-
toresistors (GMR) as is the case presented in this work.
To apply the eddy current method, there must be a
probe (excitation coil and magnetic eld sensor) which ap-
plies a primary excitation eld and reads the secondary
eld produced by the eddy currents. This probe is driven
either manually by an operator [10] or automatically by a
mechanical positioning system.
Although it is very simple to implement inexpensive
measurement systems based on eddy current methods,
the data they produce are quite difcult to interpret. Thus,
the development of automated processes of data analysis is
of great importance.
This paper is a contribution to this area of study as it
presents a fast algorithm to simulate straight cracks when
a uniform electric eld is imposed in the crack region. To
evaluate its performance the theoretical results are com-
pared with the experimental data.
The paper, including an explanation of the constant
eld probe, is divided into four sections. After this intro-
duction, the experimental setup is presented in Section 2,
which is a central issue in this work. In Section 3 we
describe the conformal mapping algorithm and nally in
Section 4 we conclude about the algorithm ability to be
used as a problem solver in our future work.
2. Description of the experimental system
The experimental setup in use in our laboratory is rep-
resented in Fig. 1. It includes a positioning system with a
motion controller which drives an eddy current probe
(ECP) over the surface of an aluminum plate where linear
cracks were articially machined. The complete system
runs under the control of a LabView program via a modular
(a)
Fig. 1. (a) System block diagram. (b) Probe with the GMR sensor.
Please cite this article in press as: A. L. Ribeiro et al., A simple forward direct problem solver for eddy current non-destructive inspection of
aluminum plates using uniform eld probes, Measurement (2011), doi:10.1016/j.measurement.2011.03.029
NI-PXI system, which includes a data acquisition module
to acquire the output signals from the ECP, an arbitrary
function generator module and a serial RS232 interface to
command the motion controller. The ECP excitation cur-
rent is provided by a transadmittance generator, driven
by the NI-PXI arbitrary function generator.
2.1. The eddy current probe
The eddy current probe is composed of a rectangular
cross-section coil and a giant magnetoresistor (GMR) sen-
sor on the bottom plane, as depicted in Fig. 1a. This coil
shape was chosen to obtain a spatially uniform magnetic
eld on a limited area of the surface of the aluminum plate
that is being scanned. The giant magnetoresistor sensor,
provided by Non Volatile Electronics, is composed of four
giant magnetoresistors mounted in a bridge conguration
and measures the magnetic eld component parallel to
the plate and perpendicular to the excitation coil axis. This
giant magnetoresistor sensor must be polarized by a con-
stant magnetic eld in order to work in a linear region of
its characteristic. This is achieved using a small permanent
magnet attached to the probe. The GMR sensor presents a
sensitivity of 3.7 mV/Oe for each volt applied by the bridge
power supply. The ECP was excited with a sinusoidal cur-
rent of Iex = 200 mA at the frequency f = 1 kHz. Fig. 1b de-
picts the ECP. The coil square cross-section presents a
lower surface in the close proximity of the plate. The wires
of this bottom surface carry a current which is directed
along the y-direction. The primary magnetic eld under
this surface is directed along the x-direction. The GMR sen-
sitive axis is directed along Oy and, in theory, the magnetic
eld along Oy is of zero amplitude, on the condition that
the plate is homogeneous. If the plate presents defects, a
component Hy of the magnetic eld will be detected.
2.2. The experimental method
The method was experimentally tested by scanning a
2 mm thick aluminum plate, where a crack with 1 cm of
length was machined. The area scanned by the probe was
z
ECP
O y
GMR Aluminum plate
(b)
 A.L. Ribeiro et al. / Measurement xxx (2011) xxxxxx 3

such that the eddy currents being perturbed by the pres-

ence of the crack were inside the constant excitation eld
at all times. This condition imposes that the probe dimen-
sions must be scaled to the length of the crack to be de-
tected, and is a necessary condition to match the
experimental to the simulation results.
Due to the high sensitivity of the GMR sensor, the mag-
netic eld component to be measured must be orthogonal
to the direction of the excitation eld to avoid sensor sat-
uration since the primary eld is several orders of magni-
tude greater then the secondary eld. In our probe the
eld component under measurement is parallel to the alu-
minum plate. Under these conditions, and for an ideal
geometry, the GMR sensor only detects the perturbations
of the eddy current magnetic eld in one direction.

3. Conformal mapping and modeling results

Fig. 2. Lines of current deviating from a linear crack (solid) and
In this section we use a conformal mapping to preview equipotential lines (dotted).
how the current lines deviate from a crack, and the Biot
Savart law is used to preview the resulting magnetic eld
components. In the domain of non-destructive testing the
conformal mapping was used before to model the mag-
netic ux leakage from a crack in a steel specimen [11].

3.1. Conformal mapping

To determine the geometric conguration of the current

lines in the presence of a linear crack of length 2L we may
use a conformal transformation [12]. Considering the Z
complex plane, with Z x jy, and the complex potential
P, with P u jv , the following transformation allows us
to determine both the scalar potential u = u(x, y) and the
ux function v = v(x, y) as functions of position:
h i
P u jv K Z cos a  jZ 2  L2 1=2 sin a 1

The consideration of the conformal transformation is

possible because, in a limited area under the excitation
coil, the applied magnetic eld is uniform and the electric Fig. 3. Current density vector deviating from a linear crack.
eld Ei produced by induction can be considered as the gra-
dient of a scalar potential, Ei = ru(x, y), in that limited
region. currents. Thus, we must calculate the magnetic eld from
If the lines of current are perpendicular to the crack the difference between the currents represented in Fig. 3
direction the angle a is a = p/2. To obtain the lines of cur- and the unperturbed currents without crack. This differ-
rent we must take v = const. and invert (1) to determine the ence is represented in the following Fig. 4.
corresponding curve in the Z plane. To determine the
equipotentials, the condition u = const. must be taken. 3.2. Components of the magnetic eld
Fig. 2 depicts these curves.
The ux density is the complex conjugate of the deriv- The determination of the magnetic eld originated by
ative of P. In our case the ux density must be interpreted the current distribution represented in Fig. 4 is performed
as a current density. Taking a = p/2, and being r the mate- using the BiotSavart law [13], which for continuous med-
rial conductivity: ia takes the form:
 " # Z Z Z
dP Z 1 Jr0  R
r  rK j 2 2 1=2 Jx jJy 2 Hr dv 3
4p R2
dZ  Z  L V

The graphical representation of this current density can
be seen in Fig. 3.
We wish to determine the perturbation on the magnetic
eld originated by the corresponding perturbation on the
In (3), r represents the point where the eld is calculated,
0 0
and r the point where the element of current J(r ) is lo-
0
cated inside dv. The vector R = r  r connects the middle-
point of dv to the point where the eld is calculated. In

Please cite this article in press as: A. L. Ribeiro et al., A simple forward direct problem solver for eddy current non-destructive inspection of
aluminum plates using uniform eld probes, Measurement (2011), doi:10.1016/j.measurement.2011.03.029
4 A.L. Ribeiro et al. / Measurement xxx (2011) xxxxxx

Fig. 4. Perturbation of the current density due to the crack.

our calculation the current is supposed to lie in a plane at

the plate mid-thickness, and the points where the eld is
calculated are supposed to lie in a plane above and parallel
to the plate, where the GMR sensor is supposed to travel.
We used an iterative approach in our calculations. Thus,
the pair (m, n) references the position of each current ele-
ment in the plate and the pair (k, l) references the position
of each point where the eld is evaluated. The discrete ver-
sion of Eq. (3) takes the form:
k;l
1 X Jm; n  Rm;n
Hk; l Dv 4 Fig. 6. Surface plots of the components of the magnetic eld in the (x, y)
4p m;n R2 plane. (a) Hx, (b) Hy.

The eld map was determined and the results are de-
picted in Fig. 5.
In Fig. 6 we plot the surfaces Hx(x, y) and Hy(x, y), repre-
senting the two components of the in-plane magnetic eld.
Unfortunately the Hx component is not measurable exper-
imentally, because it superimposed to the excitation eld.

Fig. 7. Surface plot of the measurement of the Hy component using the

ECP probe.

4. Experimental results and crack detection

Fig. 7 represents the voltage measured with the ECP.

The excitation frequency was set to fex = 1 kHz. The
Fig. 5. Components of the magnetic eld in the (x, y) plane. sampling frequency was fS = 400 kHz and the number of

Please cite this article in press as: A. L. Ribeiro et al., A simple forward direct problem solver for eddy current non-destructive inspection of
aluminum plates using uniform eld probes, Measurement (2011), doi:10.1016/j.measurement.2011.03.029
 A.L. Ribeiro et al. / Measurement xxx (2011) xxxxxx
samples to acquire the amplitude and phase of the GMR
output signal was set to N = 105, which corresponds to a to-
tal of 250 periods of the excitation for each point under
measurement. A total of 50  50 points separated by
1 mm of distance were measured over an aluminum plate
with 2 mm of thickness where a crack with 1 cm of length
and 0.5 mm wide had been cut. The gap between the probe
and the plate was set to approximately d = 1 mm.
The resemblance between the pictures obtained by sim-
ulation and experimentally is remarkable, and shows how
useful the simulation work can be to preview the experi-
mental results. The offset of 40 mV, visible in Fig. 6, is pro-
duced by a very small amount of the excitation eld that is
detected by the GMR sensor.
The method to detect the crack is based on the steepest
descent method, which seeks for the correct value of the
parameters K, a and L in the transformation (1), the dis-
tance d between the plate surface and the plane where
the GMR sensor detector travels, and the translation (dx,
dy) to correctly locate the center of the model crack with
the experimental one. The function to minimize was the
energy DE of the difference between the experimental data
Exp(i, j) represented in Fig. 7 and the model data Model(i, j)
represented in Fig. 6b,
X 2
DE Expi; j  Modeli; j 5 [1] M. Kubinyi, A. Docekal, H.G. Ramos, A.L. Ribeiro, Signal processing for
i;j
After minimization of the signal energy difference the
measured values were a = 1.55 rad, L = 5.2 mm and d =
0.64 mm. A translation (dx, dy) = (1.5, 0.0) mm was also
obtained to center the two data images.
The small differences between the experimental and
the real values of the crack parameters may be explained
by the approximations in the physical model.
The main approximation is related to the metallic plate
thickness. The mathematical conformal mapping model as-
sumes that the electric currents are conned to a plane
surface, but the real plate presents a thickness of two mil-
limeters. Being the excitation frequency in this experi-
ment, fex = 1 kHz, the current density distributes across
the whole thickness, introducing an error in the magnetic
eld calculations.
5. Conclusions and future work
This paper reports the results obtained using a confor-
mal transformation to preview the spatial variation of the
voltage detected by an eddy current probe when an alumi-
num plate with a straight crack is scanned.
The simulation work was followed by a real experiment.
In this plate a crack with 1 cm of length and 0.5 mm wide
had been machined.
The good agreement in the surface plots of the results
obtained by simulation and experimentally allows us to
Please cite this article in press as: A. L. Ribeiro et al., A simple forward direct problem solver for eddy current non-destructive inspection of
aluminum plates using uniform eld probes, Measurement (2011), doi:10.1016/j.measurement.2011.03.029
5
conclude that the proposed algorithm is adequate. Due to
its simplicity, it constitutes a very good tool to test the
inversion algorithms that will be the subject of our future
work. Moreover, the achieved results point out a direction
to explore for the inversion problem solution. By using a
uniform eld probe, and subtracting from the detected sig-
nal image the values corresponding to a material without
defects, we obtain a nal image that can be considered as
representing the magnetic eld map resulting from the
current perturbation distribution.
Thus, our future work will consist in an inversion prob-
lem trial, determining the distribution of dipolar current
eddies from the magnetic eld perturbation map, knowing
the magnetic eld produced by a single eddy.
Acknowledgments
This work was supported in part by the Portuguese
Science and Technology Foundation Project PTDC/EEA-
ELC/67719/2006 and in part by the Instituto de Telecomu-
nicaes Project CLASSE. This support is gratefully
acknowledged.
References
non-contact NDE, Przeglad Elektrotechniczny (Electrical Review), R.
86 Nr 1/2010.
[2] A. Bernieri, L. Ferrigno, M. Laracca, M. Molinara, Crack shape
reconstruction in eddy current testing using machine learning
systems for regression, IEEE Transactions on Instrumentation and
Measurement 57 (9) (2008).
[3] I.G. Palmer, The use of acoustic-emission techniques and liquid
penetrants to detect surface cracks, Non-Destructive Testing 7 (1)
(1974) 2527.
[4] A. Lopes Ribeiro, Francisco Alegria, O. Postolache, H. Geirinhas
Ramos, Eddy current inspection of a duralumin plate, in: Proc.
I2MTC-IEEE International Instrumentation and Measurement
Technology Conference, Singapore, pp. 1367-1371, May, 2009.
[5] C.V. Dodd, W.E. Deeds, Analytical solutions to eddy-current probe-
coil problems, Journal of Applied Physics 39 (6) (1968) 28292838.
[6] K. Koyama, H. Hoshikawa, N. Taniyama, Investigation of eddy current
testing of weld zone by uniform eddy current probe, in: 15th World
Conference on Non-Destructive Testing, Rome, 2000.
[7] Y. He, F. Luo, M. Pan, X. Hu, B. Liu, J. Gao, Defect edge identication
with rectangular pulsed eddy current sensor based on transient
response signals, NDT&E International 43 (2010) 409415.
[8] J.C. Moulder, E. Uzal, J.H. Rose, Thickness and conductivity of metallic
layers from eddy current measurements, Review of Scientic
Instruments 63 (6) (1992) 34553465.
[9] Y. Li, T. Theodoulidis, G.Y. Tian, Magnetic eld-based eddy-current
modeling for multilayered specimens, IEEE Transactions on
Magnetics 43 (11) (2007) 40104015.
[10] R. Smid, A. Docekal, M. Kreidl, Automated classication of eddy
current signatures during manual inspection, NDT&E International
38 (2005) 462470.
[11] J.R. Bowler, N. Bowler, Evaluation of the magnetic eld near a crack
with application to magnetic particle inspection, Journal of Physics
D: Applied Physics 35 (2002) 22372242.
[12] R.V. Churchill, J.W. Brown, Complex Variables and Applications, fth
ed., McGraw-Hill, 1990.
[13] Martin A. Plonus, Applied Electromagnetics, McGraw-Hill, New York,
1978.