Excerpt from the Proceedings of the COMSOL Conference 2010 Paris

Using COMSOL-Multiphysics in an Eddy Current

Non-Destructive Testing Context
L. Santandrea, Y. Le Bihan
Laboratoire de Genie Electrique de Paris, CNRS, Supelec, Univ. Paris-Sud 11, UPMC
*Laurent Santandrea: LGEP 11 rue Joliot Curie, 91192 Gif sur Yvette , santandrea@lgep.supelec.fr

Abstract: EDDY current testing (ECT) is widely function of coil position, and to compare the
used to check the integrity of electrically computed results to the experimental ones.
conducting parts and notably to detect flaws. It is
based on the interaction between a probe and the
part under testing. The finite element method
(FEM) is well fitted to the modeling of these ECT probe
kinds of problems because of its large flexibility Magnetic core
which allows to deal with complex probe and
Defect
part configurations. In this context, the
benchmark problems TEAM Workshop 15-1 and y
JSAEM 2-5 have been considered to validate a
COMSOL-Multiphysics 3D-resolution using a
x
3D electromagnetic formulation with Whitney Coil Conductive part
edge elements. A good agreement is obtained
between the numerical results and the
experimental ones. Figure 1. Problem geometry

Keywords: Eddy current, non-destructive 2. Use of COMSOL Multiphysics in ECT

testing, edge elements, harmonic quasi-static
fields. The release of the software used for the study is
the 3.5 release.
1. Introduction
2.1 Model definition
A typical geometry ECT problem is constituted
of a conducting domain in which exists a defect The frequency is sufficiently low so that the
(Figure 1), a probe (coil driven by an excitation displacement currents can be neglected (electro-
current density which may include a magnetic technical hypothesis). Since the exciting coil is
core) and the surrounding air.. The benchmark driven by a sinusoidal current, the time harmonic
problems TEAM (Testing Electro-magnetic approach is used. The presence of the defect
Analysis Methods) Workshop 15–1 and JSAEM implies a 3D FEM resolution:
(Japan Society of Applied Electromagnetic and
Mechanics) 2-5 have been considered in order AC/DC Module -> quasi-Statics Magnetics ->
to validate these approaches. The forward Induction currents (EMQA)
problem typically consists in the determination
 j      A     
 1
 
of the probe impedance variation. The real and 2
o r o  r 1   A  J 0
imaginary parts of the probe impedance are
determined by using the magnetic energy and the
power losses, respectively. Both are deduced This Finite Element Formulation results from the
from the FEM results. Maxwell equations. It uses Whitney edge
elements. The scalar unknown is then the
A circular air-cored coil is moved along the circulation of the magnetic vector potential along
length of a rectangular slot in a conductive plate. the element edges of the mesh.
Both frequency and coil lift-off are fixed. The
objective is to compute the change of the
impedance of the coil (comparatively to its value
over an unflawed portion of the plate) as a
2.1 Geometry definition Taking into account the high number of turn of a
coil, this one is assumed as a medium having an
The predefined forms with CSG (Constitutive uniform vector current density J0 (Figure 3). The
Solid Geometry) approach are used to define the resistance of the coil wire is then not taken into
geometry in the software. account in the modeling. This field is defined in
the volume domain corresponding to the coil. Its
Geometric parameters JSAEM TEAM norm is constant through the coil volume and
Coil inner radius (mm) 0.6 6.15 defined by:
Coil outer radius (mm) 1.6 12.4
Coil length (mm) 0.8 6.15
Lift off (mm) 0.25 0.88
Plate length (mm) 25 100 Nb _ coil _ turn * I
Plate width (mm) 15 50 J0 
Plate thickness (mm) 1.25 12.22 h _ coil * (r _ coil _ out  r _ coil _ in)
Cracks length (mm) 10 12.6
Cracks depth (mm) 0.75 0.5
Cracks width (mm) 0.21 0.28 With Nb_coil_turn the coil number of turns,
h_coil, the coil length, r_coil_out the coil outer
A typical physical parameter of this kind of radius, , r_coil_in the coil inner radius.
problem is the skin depth  defined by:

1 1
 
 f
2
With  the magnetic permeability, the 
electrical conductivity,  the pulsation, f the
frequency
Figure 3. Coil representation left: reality, right: model
for FEM calculation.
For the JSAEM problem the skin depth (1.30
mm) is greater than the plate thickness. In this The J0 vector is then expressed in the (x, y,
case, to avoid numerical problems due to the z) coordinate system by the following formula:
boundary conditions, it is necessary to add an air
domain under the plate.
J0 z J0 y
J 0  [0, , ]
y2  z2 y2  z2

With x the vertical component of the coordinate

system and with a coordinate system reference
located at the center of the coil (Figure 4).

Figure 2. JSAEM geometry

The depth of the surrounding air is considered to

be equal at three times the length of the coil.

2.2 Physical properties

physical parameters JSAEM TEAM

Core relative permeability 1. 1.
Coil number of turns 140 3790
Intensity (A) 0.8 6.15
Frequency (Hz) 1.5e5 900
Plate conductivity (S/m) 1e6 30.6e6 Figure 4. Orientation of the coil for J0 definition
The defect is supposed to be an insulating crack
and this domain has the same physical properties
than an air domain.

2.3 Boundary conditions setting

In all the boundaries of the study domain, the

magnetic induction B is supposed tangent and
then the magnetic insulation condition (nxA=0)
is imposed (default condition in COMSOL-
Figure 6. Example of 3D mesh
Multiphysics).
The geometry is meshed at each displacement
step of the probe. To avoid numerical noises due
2.4 Mesh generation
the mesh effect during the probe displacement,
The quality of the mesh is crucial for a finite the variation of impedance is computed at each
element simulation. It allows to obtain an displacement step with the same mesh.
accurate solution with a reduced number of
elements and so a reasonable computational
time. The mesh is generated with tetrahedral
elements. A fine mesh is required in the region
where the variation of the fields is important. As
an example we have to refine the mesh around
the coil or in the skin depth in the plate since the
magnetic induction field varies significantly in
these regions. In fact to take into account the Figure 7. Planar visualization of the 3d mesh around
field variation in the skin depth, three layers of the coil
elements are necessary if first order elements are
used. 2.4 Resolution

A dual core computer with 4 Gb RAM memory

is used. The mesh has about 100 000 second
order tetrahedral edge elements. “Spooles”
solver is used to solve the linear system. For one
position the computing time is less than five
minutes.

2.5 Impedance variation computation

The impedance variation Z is a complex

number. The real part is computed with the Joule
Losses (JL) in the conductive media and the
imaginary part is computed with the magnetic
energy (WM) in the whole meshed domain:

 JL Z
 Re( Z )  I 2  I Re(Z )
2

Figure 5. Norm of the magnetic induction field


visualization.

  1 J 2  J 2 d
  f 
c
  WM Z
Im(Z )   I 2 Im(Z )

2
I
 1

    B f  B d
 
2 2

c
Jf ,J and Bf ,B are the current density field and
the magnetic induction field respectively
calculated with and without the defect and are
both software post-treatment quantities.

3. Scripting consideration

For the computing process, a script program is

used. A loop allows calling a function which
computes the impedance at each displacement
step.
In this function as an example, the magnetic
energy and the Joules losses are directly
evaluated using the function ‘postint’ as it is
described in the script bellow:
Figure 8. Map of the eddy currents in the plate
without defect
% Computation of the Energy
%  CALCUL DE b2/

WM=postint(xfem,'(normH_emqa*normH_emqa)*
(4*pi*1e-7)', 'unit','W*m')

% Computation of the Joule losses

% ...  CALCUL DE j2/

if (okdefect==0) % no defect

Joule_Losses=postint(xfem,'(normJ_emqa*no
rmJ_emqa)/(1e6)','unit','A*A/m',...
'dl',[2,3]) % Defect+ plate domain

elseif (okdefect==1) %with defect

Joule_Losses=postint(xfem,'(normJ_emqa*no
rmJ_emqa)/(1e6)', 'unit','A*A/m',
'dl',[2])%  Only in the plate domain

Zr = PJ Figure 9. Map of the eddy currents in the plate with a

Zi = 2*pi*Fq*WM defect
deltaZr = Zr-Zr0
.
deltaZi = Zi-Zi0

4. Results

The circular coil is displaced along the y-axis

(Figure 1). The impedance variation is computed
as a function of the position of the coil center.
Quantitatively, Figures 8 and 9 show a good
behavior of the current density. Quantitatively
the Figures 10 and 11 show a good agreement of
the impedance variation results in comparison
with the results obtain in references [1] and [2]. Figure 10. JSAEM 2-5 – Variation of the probe
resistance (bottom) and reactance (top) versus the
probe position along the crack length
Figure 11. Team-Workshop 15 - Variation of the
probe resistance (bottom) and reactance (top) versus
the probe position along the crack length

5. Conclusions

This paper presents a validation of an ECT

3D problem with the software COMSOL
Multiphysics 3.5. The use of the software to
solve this kind of problem is detailed and
explained. Benchmark numerical results are
presented.

6. References

1. M. Bensetti and all, Adaptive Mesh

Refinement and Probe Signal Calculation in
Eddy Current NDT by Complementary
Formulations, IEEE trans mag, Volume 44
number 6, 1646-1649 (june 2008)

2. Y Choua and all, Specific development on a

Finite Element Tool for thin crack modeling in
EC testing, ECNDT conference 2006, Poster 145

7. Acknowledgements

Acknowledgments to Carlos Furtado and

Zhen Zhu, students of the IFIPS school of
University Paris-Sud for their contributions to
this work during their internship at LGEP.