The 10th INTERNATIONAL SYMPOSIUM ON ADVANCED TOPICS IN ELECTRICAL ENGINEERING

March 23-25, 2017

Bucharest, Romania

Mathematical Modeling of Eddy Current

Non-Destructive Testing
E.S. APOSTOL1 , Adrian NEDELCU 1 , Dan V. DANIEL 2 , Ionel CHIRIŢĂ1 , Nicolae TĂNASE1,2
1
INCDIE ICPE-CA, ECCE Department, Splaiul Unirii, Nr. 313, District 3, 030138, Bucharest, Romania,
2
University POLITEHNICA of Bucharest, Splaiul Independentei nr. 313, District 6, RO-060042, Bucharest, Romania
dan@lmn.pub.ro, nicolae.tanase@icpe-ca.ro, simona.apostol@icpe-ca.ro, ionel.chirita@icpe-ca.ro, cristinel.ilie@icpe-ca.ro

Abstract—This paper proposes analytical and numerical mod-

els for the measurement process of Eddy-Current (EC) non-
destructive testing. EC inspection represents an essential method
for the electromagnetic nondestructive evaluation (NDE) of
cracks in conductive materials, with its main applications being
found in the examination of aircraft, particle accelerators, and
other engineering constructions. The method is based on the
detection of the magnetic field produced by eddy currents induced
in the specimen being tested. The presence of a crack disturbs
the flow of the eddy currents, thus producing a magnetic field
perturbation dependent on the position and shape of the defect
itself. Variations in the electrical conductivity and magnetic
permeability of the test object material, and the presence of
defects in the object causes a change in eddy current and
a corresponding change in phase and amplitude that can be Fig. 1: EC non-destructive testing [6]
detected by measuring the impedance changes in the magnetic
field generating coil, which is a telltale sign of the presence
of defects. The nature of the method is complex and there
is therefore a need for deeper understanding that may be
gained from mathematical models. Such models can have several
objectives as for example procedure and equipment optimization
II. A NALYTICAL MODEL
or understanding of the method capability and reliability.
Keywords—eddy current testing, mathematical modeling, ana-
lytical model,numerical model

I. I NTRODUCTION
The geometry of the analytical model is the one presented in
EC non-destructive testing is widely used in metal industry
Fig. 2. The geometrical model consists of one copper winding
and science in order to evaluate the properties of materials
placed above a conductive plate at a distance l. This winding
without causing damage. This testing method is used in the
is circulated by an alternating current [2].
maintenance of aircrafts, of particle accelerators etc. for:
crack detection, conductivity measurements, coating evalu-
ations, material thickness [1]. The phenomenon behind the
eddy testing consists in an alternating current flowing through
a coil at a chosen frequency, current that generates a time
varying magnetic field around the coil as in Fig. 1. When
the coil is placed close to an electrically conductive material,
eddy currents are induced in the material. If there is a defect
in the conductive material, the eddy currents circulation is
disturbed and the magnetic coupling between the coil and the
tested material is modified and a defect signal can be read by
measuring the coil impedance variation [2], [1].
Mathematical models [3], [4] of the physical mechanism
of EC inspection are necessary not only for a deeper under- Fig. 2: Single turn coil
standing of the process but also for good results regarding
equipments design [5].
This paper proposes a numerical model of the measurement
process of EC non-destructive testing which will be validated The analytical model starts from the Maxwell equations
by an analytical solution. from which the diffusion equation is obtained for the magnetic

c
978-1-5090-5160-1/17/$31.00 2017 IEEE

469
vector potential A. Considering that m = 1, solution of (5) is:
√ √
rotH = J, G(z) = Aexp( sz) + Bexp(− sz), (8)
∂B
rotE = − , where s = iωµσ + 1.
∂t
Although equation (6) becomes a Bessel equation under
divB = 0, ∂A m = ±1 condition, [2], [7] consider an integration of m for
B = rotA, =⇒ ∇2 A = µσ (1)
∂t values from 0 to ∞.
divA = 0, The correct solution of (3) is:
J = σE,
 √ √ 
B = µH. A(r, z) = [CJ1 (r) + DN1 (r)] Aexp( sz) + Bexp(− sz) .
(9)
In order to analytically solve this equation for our problem,
For regions I, II and IV , (2) has the following solution:
the computational domain must be separated into four regions,
as in Fig. 2: first region above the winding, second region
between the winding and the conductive plate, third region A(r, z) = [CJ1 (r) + DN1 (r)] [Aexp(z) + Bexp(−z)] ,
belongs to the plate and the fourth region is below the plate (10)
[7]. The A, B, C and D coefficients must be determined. In
The next step in solving (1) is to write it in cylindrical order for the model to have physical consistency, we assume
coordinates for those four mentioned regions. For the first, that the potential must have a limited value in the area of
second and fourth regions, the equations are equal to zero interest and must vanish towards infinity.
since the conductivity of air is considered 0. For the third Thus, since the second kind Bessel function does not
region, which belong to the plate, the conductivity is different converge, D coefficients will be eliminated.
from zero so the resulting equations are as follows: The following functions are obtained for the four regions:
• For I, II and IV :

∂ 2 A 1 ∂A ∂ 2 A A A(I) (r, z) = J1 (r)B1 exp(−z),

+ + − 2 = 0, (2)
∂r2 r ∂r ∂z 2 r A(II) (r, z) = J1 (r) [A2 exp(z) + B2 exp(−z)] ,
 √ √  (11)
• For III: A(III) (r, z) = J1 (r) A3 exp( sz) + B3 exp(− sz) ,
∂ 2 A 1 ∂A ∂ 2 A A A(IV ) (r, z) = J1 (r)A4 exp(z).
+ + − 2 = iωµσA. (3)
∂r2 r ∂r ∂z 2 r Continuity conditions are imposed on the magnetic vector
These equations can be solved in similar way for the four potential and its derivative.
regions, so we will show the procedure only for the third
region. Solving (3) by the method of variable separation, yields A(I) (r, l) = A(II) (r, l);
 
two equations: ∂A(I) ∂A(II)
(r, z) = (r, z) − µIδ(r − r0 ),
∂z z=l ∂z z=l
∂2A A(II) (r, 0) = A(III) (r, 0);
− iωµσA = m2 A,
∂z 2 ∂A(II)

∂A(III)

(12)
∂ 2 A 1 ∂A A (4) (r, z) = (r, z) ,
+ − 2 = −m2 A, ∂z z=0 ∂z z=0
∂r2 r ∂r r
A(r, z) = F (r)G(z). A(III) (r, −d) = A(II) (r, −d);
 
∂A(III) ∂A(IV )
∂2G
(r, z) = (r, z) .
+ G −m2 − iωµσ = 0, (5) ∂z z=−d ∂z z=−d
∂z 2
From the above continuity conditions a system of six equations
  with six unknowns is obtained:
∂2F 1 ∂F 2 1
+ +F m − 2 =0 (6)
∂r2 r ∂r r
| {z } B1 exp(−l) = A2 exp(l) + B2 exp(−l),
Bessel equation for m = ±1
− B1 exp(−l) = A2 exp(l) − B2 exp(−l) − µIr0 J1 (r0 ),
In (6), we see that for values of m = ±1, this becomes
a Bessel equation. But since the mathematical model has to A2 + B 2 = A3 + B 3 ,
√
have physical consistency, only the positive value must be con- µ [A2 − B2 ] = µ s [A3 − B3 ] ,
√ √
sidered, otherwise negative values of the material resistance A3 exp(− sd) + B3 exp( sd) = A4 exp(−d),
would be obtained. Solution of (6) is: √
s
[A3 exp(−sd) − B3 exp(sd)] = A4 exp(−d).
µ
F (r) = CJ1 (r) + DN1 (r), (7) (13)
where J1 is the first kind Bessel function and N1 is the From system (13) the coefficients B1 , A2 , B2 , A3 , B3 , C4
second kind Bessel function. are determined.

470
 In the above, the current was considered to flow in a single
turn coil, but in practice the N -windings coils as in Fig. 3 are
used. After computing the coefficients, we can now determine
the magnetic vector potential and the impedance of the entire
coil by the principle of superposition, considering the linearity
of Maxwell’s equations. Thus, considering the case of the coil

Fig. 4: Domain physics and boundary conditions

where the first factor in the square brackets depends only on

Fig. 3: Coil of N -windings the problem parameters such as coil size and shape, frequency,
conductivity and it is called the "defect sensitivity factor" [7]
from Fig. 3, the magnetic vector potential A is determined and the second factor, V ol · α22 , depends on the geometry and
using the relations obtained above for single turn coil. orientation of the defect.
The expression for the total magnetic vector potential is:
n
III. N UMERICAL MODEL
X
A(r, z)total = A(ri , zi ), (14) A. Conceptual model
i=1
The considered model is the one from Fig. 3 with the
and the general expression for the coil’s impedance is:
following geometric parameters:
V
Zcoil = . (15) I = 0.5A,
I
r1 = 6 · 10−3 m,
From (14) and (15), we have:
r2 = 7.4 · 10−3 m,
i2πωN
Z l2 Z r2 l1 = 0.4 · 10−3 m,
V (r, z) = rA(r, z)drdz,
(l2 − l1 )(r2 − r1 ) l1 r1 l2 = 9.6 · 10−3 m,
(19)
i2πωN
Z l2 Z r 2 N = 124, φwire = 0.35 · 10−3 m,
Zcoil = rA(I−II) (r, z)drdz,
(l2 − l1 )(r2 − r1 ) l1 r1 d = 5 · 10−3 m,
(16) µr Cu = µr air = µrplate = 1,
where σCu = 5.9 · 107 S/m,
σplate = 7.2 · 107 S/m.
A(I−II) (r, z) = A(I) (r, z) + A(II) (r, z). (17)
The EC testing is a time harmonic regime problem charac-
A. Defects terized by: magnetic flux law, electromagnetic induction law,
Using the vector potential in the metal plate given by (11) Ampere law and laws of materials, as in (20).
for the third region, the effect of a defect can be determined.
According to [7] and [8], a defect may be represented by
the sum of a "current defect" and a "magnetic defect". The ∇B = 0,
current defect is caused by an abrupt change in the metal’s ∂B
∇×E=− ,
conductivity, and the magnetic defect is the result of an abrupt ∂t
∇ × H = J, (20)
change in its magnetic permeability.
Since the magnetic defects appear only in the case of B = B(H) → B = µH + Br ,
ferromagnetic materials, for the problem described above, only J = σ(E) → J = σE + Ji .
the current defects will be considered.
The impedance change due to a current defect can be The electromagnetic modeling was made with the Finite
calculated by the current defect equation [7]: Element Method, using Comsol software program.
"  2 # Figure 4 indicates the computational domain and its bound-
′ 3 2 A ary conditions.
Z = σω · V olα22 , (18)
2 I Figure 5 indicates the materials used for numerical model.

471
 Fig. 7: Induced current density streamline

Fig. 5: Materials

Fig. 8: Applied and induced current

TABLE I: Analytical(AN) and numerical(NUM) values of the

Fig. 6: Current density streamline inductance L[H] of the coil with and without plate
No plate With plate
Frequency [Hz] AN NUM AN NUM
100 7.957e-5 7.8214e-5 7.827e-5 7.638e-5
The magnetic field source is the current density J (Fig. 6), 500 6.135e-5 6.0096e-5 5.472e-5 5.3569e-5
which has the following expressions on the domain of the coil: 1000 3.572e-5 3.4930e-5 3.302e-5 3.248e-5

p
Jex = −J0 · y/ (x2 + z 2 ) A/m2 ,
p From Table I and Fig.10 one can see that there is good
Jey = J0 · x/ (x2 + z 2 ) A/m2 , (21)
agreement between numerical and analytical results, with an
Jez = 0 A/m2 . acceptable small relative error (Fig. 11). The analytical and
The current induced in the conductive plate (Fig. 7 and 8) numerical results were compared only for the case of the coil
will have a different pattern in a plate with or without defect, with conductive plate, but since the applied equations are the
as can be seen in Fig. 9 . same form Section II, the same small relative error is expected
for the other case of the coil with no
conductive plate. The
B. Results relative error is determined by x−x x
0
, where x represents the
A modification in the pattern of eddy currents determines analytical value and x0 the numerical value.
the modification of the inductance L of the coil, which results In literature (see [7], [8])it is considered the possibility
in a modification of the coil’s total impedance according to of an analytical approach of the defect by (18). In order to
equation (22), as can be seen in Table I. see whether or or not this is possible, we have considered
some cases of different volumes defects as in Fig. 12 located
at 0.37mm below the surface, within the conductive plate
Z = R + iXL ,
(22) in accordance with the penetration depth for the considered
XL = 2πf L. frequencies.

472
 (a) Eddy currents in no defect conductive plate
(b) Eddy currents in conductive plate with defect
Fig. 9: Eddy currents pattern in the presence of a defect
Fig. 10: Analytical and numerical results vs. frequency
Considering the results plotted in Fig. 13, we conclude that
there can not be applied any analytical relation that would
indicate a linear dependence of the inductance of the coil on
the volume of the defect as suggested by (18).
IV. C ONCLUSIONS
This paper presents the analytical and numerical mathemat-
ical model of the eddy testing non-destructive procedure. The
473
Fig. 11: Relative error of analytical and numerical results
Fig. 12: Capture of the coil and the plate with a defect
Fig. 13: Inductance of the coil vs. volume of the defect
analytical and numerical results are compared, for the case of
the coil with no defect plate, and one can see good agreement
between these. The influence of defects on the inductance of
the coil can not be approached from an analytical point of
view since the contribution of a flaw to the chance of the
coils inductance depends on various parameters of the defect
such as: volume, geometric parameters, position.
 R EFERENCES
[1] P. Xu, Eddy Current Testing Probe Composed of Double Uneven Step
Distributing Coils for Crack Detection. PhD thesis, Harbin Institute of
Technology, China, 2008.
[2] L. Heinzle, “Eddy current testing using the bode 100,” 2011. OmicronLab.
[3] M. Morozov, G. Rubinacci, A. Tamburrino, and S. Ventre, “Numerical
models of volumetric insulating cracks in eddy-current testing with
experimental validation,” IEEE Transactions on Magnetics, vol. 42, no. 5,
pp. 1568–1576, 2006.
[4] E. Weststrate, M. Steinback, N. M. Rensing, and T. Tiernan, “Comsol
multiphysics modeling for design optimization of eddy current crack
detectors,” Proceedings of the COMSOL Conference 2010 Boston, 2010.
474
[5] S. Harzallah and M. Chabaat, “Eddy current sensor modeling for nonde-
structive evaluation of stress intensity factor,” Proceedings of 2014 AASRI
Conference, vol. 9, pp. 57–63, 2010.
[6] T. Nelligan and C. Calderwood, “Introduction to eddy current testing,”
2017.
[7] C. Dodd and al, “Some eddy-current problems and their integral solu-
tions,” 1969. Oak Ridge National Laboratory.
[8] M. L. Burrows, A Theory of Eddy Current Flaw Detection. PhD thesis,
University Microfilms, 1964.