Separation of conductivity and distance measurements

for eddy current nondestructive inspection of graphite

composite materials
Isabelle Dufour, Dominique Placko

To cite this version:

Isabelle Dufour, Dominique Placko. Separation of conductivity and distance measurements for eddy
current nondestructive inspection of graphite composite materials. Journal de Physique III, EDP
Sciences, 1993, 3 (6), pp.1065-1074. �10.1051/jp3:1993107�. �jpa-00248979�

HAL Id: jpa-00248979

https://hal.archives-ouvertes.fr/jpa-00248979
Submitted on 1 Jan 1993

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
J. Phys. III France 3 (1993) 1065-1074 TUNE 1993, PAGE 1065

Classification
Physics Absn.acts
07.55 06.30L

Separation of conductivity and distance measurements for eddy

current nondestructive inspection of graphite composite
materials

Isabelle Dufour and Dominique Placko

Laboratoire d'Electricit6, Signaux et Robotique, 61 avenue du Pr6sident Wilson, 94235 Cachan

Cedex, France

(Received I June 199?, iei'i.led 9 Noiembei /992, accepted ii Januaij, 1993)

Abstract. This article deals with

a the
process based on the principle of eddy
study of current

sensors for the nondestructive graphite composite plates. This

evaluation of research has been
carried out in the Laboratoire d'Electricit6 Signaux et Robotique by the team working on data-
collecting sensors for robotics in collaboration with Aerospatiale. Eddy current sensors are

characterized by their impedance, which varies when a conducting material is approached in their
sensitive area. For a given sensor, the output signal depends directly on the electrical and
geometrical properties of the object. In the case discussed here, the interesting data are the distance
between the sensor and the object, and its local conductivity. In order to invert the relationships
between the sensor signal and the properties of the material, an external parametrical model has
been developed. A scanning of the surface with a sensor designed for good spatial resolution

measurements gives two accurate maps of the useful data.

1. Eddy current sensors principle [il.

An eddy current sensor is made of open

an magnetic core on which an excitation coil is driven
by a high frequency sine wave magnetic field is thus created, defining a
signal. A variable
sensitive area in front of the sensor, which essentially depends on the geometry of the magnetic
core. The approach of a target I-e-, a conducting material, which in our case is a graphite
composite into this area modifies the path of the magnetic field (consequently the value of
the inductance) and induces losses. The signal is then obtained through the measurement of
real and imaginary parts of the winding impedance it depends on the characteristics of the

sensor (geometry, distance between sensor and target, frequency of excitation, etc. and on the
target properties such as conductivity «, permeability v, geometry. In our application, because
of the anisotropy of the graphite composite materials, we chose an eddy current sensor which
presents a symetry of revolution : a ferrite cup core.
1066 JOURNAL DE PHYSIQUE III N° 6

2. Measurement and modelJing.

2. I NORMALIzED IMPEDANCE PLANE [2-4]. In order to evaluate the sensor

efficiency with
respect to a given target, we can describe sensor impedance variations in the complex plane
Z~ =
R~ + jX~. However, a difficulty in this approach is the growth of the reactance with
frequency and the existence of sensor losses when there is no target in front of it
(Zo Ro + jXo).
=
To overcome difficulty, this we
prefer to study the variations of normalized
impedance (Z~~ R~~ + jX~~), which are
=
only due to the presence of the target. Thus, we can

describe the graphic function X~~ f(R~~), =

where :

R~ Ro X~
R~~= and X~~=-
Xo Xo

2.2 EXPERIMENTAL RESULTS. The experimental setup is presented in figure I. The sensor is
placed over a sample of graphite composite, the impedance measurements are
made with an

impedance analyser (HP 4192A) and the treatment with a computer.

' f'i~

*~,, <a
,,,,
i# ,
«,
",

Fig. I. Experimental setup.

We present in figure 2 some experimental results obtained for several excitation frequencies
and distances between the sensor and a sample of graphite composite placed in front of it. The
following remarks may be noted about this figure

. for a given distance, the curve representing the function X~~ =

f(R~~ ) can be assimilated
to a circle when the frequency is made to vary. This circle joins the point (0, 1) for low
frequencies (R~ =
Ro and X~ =
Xo because the eddy currents account for very little) and the
N° 6 EDDY CURRENTS CONDUCTIVITY AND DISTANCE MEASUREMENTS 1067

point (0, a for very high frequencies (the field does not penetrate the sample : losses in the
sample are very small)
. for a given frequency, the curve representing the function X~~ =
f(R~~) is almost a line
passing through the point (0, 1) when the distance sample-sensor is made to vary.

Normafiscdwmwwplan©
i

Fmq=lADlz

o.8
~ mq=2AD~z

0.6

xun

0.4

Di©m0.2mm
~~
Di©m0mm

0 0.2 0.4 0.6 0.8

Rcn

Fig. 2. Experimental results obtained in the normalized impedance plane with frequency [I MHz,
9 MHz] and distance [0 mm, 2 mm].

Note that points depend on the sensor-target

measurement distance and on a factor which is
proportional to product the
of the excitation frequency by the sample conductivity. In fact,

same measurement points should be obtained by increasing the conddctivity of the sample
instead of the excitation frequency (in the same proportions). This property readily appears in
the Maxwell equations.

2.3 EXTERNAL PARAMETRICAL

MODELLING [5]. An analytical model of the relations
between the
sensor signal and the properties of the target has been developed. This model is
based on an electrical diagram constituted by a transformer in which the primary and the
secondary circuits respectively symbolize the sensor and the target (Fig. 3). The load
impedance is linked to the complex power dissipated in the target. A very simple interpretation
is that the real impedance represents the resistance of the loop in which eddy currents are

induced (Re will be consequently proportional to I/«), and the imaginary term accounts for the
leakage inductance of this circuit. The coupling coefficient k, is linked to the distance between
the sensor and the target : it decreases when the distance increases. These parameters are

obtained by identification with experimental results.

When there is no target in front of the sensor I-e-, the coupling factor k is zero the
impedance of the transformer is Zo =
Ro + jXo. In the presence of target, this impedance
JOURNAL DE PHYSIQUE »> T 3, N'6. JUNE lQ93 3Q
1068 JOURNAL DE PHYSIQUE III N° 6

Ro
~
/N

'Lo Li Re + jIm

Fig. 3. Extemal parametrical model.

becomes Z~, as described by the following equation

~2 ~~ ~, ~
2

~~ ~° ~~~° ~° ~
Re + jLj w + j Im

So, obtaining :

k~ Lj w
Re
~~~
Re~ + (Lj +
Im)~
w

and

k~ Lj w
(Lj w + Im)
X~~ =

Re ~
+ (Ljw +Im) ~

we get

k~ Im 2 k~ 2 k~ Im~
l~ ~~
~
2 Re
~ ~
~~
~
2 4 Re~
~

and

Lj w + Im
X~~ =
I R~~.

. On the assumption that Im, Re and Lj

depend (for a given sample) on the distance
do not
between the sensor and the
we target,
a for a given frequency in the
obtain normalized
line
impedance plane when the coupling coefficient changes. The slope of the line is
(Lj w + Im)/Re. Note that, if Im is negligible in respect to Lj w or proportional to w, we find
again that the slope is proportional to the product frequency conductivity (Fig. 4).
. On the assumption that Im/Re is constant (for a given sample), we obtain a circle in the
normalised impedance plane for a given coupling coefficient k, when the frequency changes.
The centres of circles are on the line connecting (O, I) and (- Im/Re, 0). Moreover, these
circles intercept the ordinate axis at (O, I k~) (Fig. 4).
It can be seen that this electrical model correlates sufficiently with the experiments.
Therefore, we can discuss identification of parameters.

2.4 IDENTIFICATION. For a given distance, we identify the curve obtained in the normalized
impedance plane to a circle with the least square method. By means of this circle equation, we

get the values of the ratio Im/Re and of the coupling coefficient k.
N° 6 EDDY CURRENTS CONDUCTIVITY AND DISTANCE MEASUREMENTS 1069

Normafised Plane

0.6 '

'

~"
'
fl
0.4

0.2 1-k' fl

0
°
-J~n/Re

Rcn

Fig. 4. -Normalized impedance plane analysis when model parameters are constant.

Results are given in figures 5a and 5b for two different ranges of frequencies. We see that the
ratio Im/Re is not constant but increases with frequency, and depends on coupling coefficient.
We find out that it varies linearly versus I k~.

Nonndhed HwWmce plane IhvRe versus 1-k'

l I

0.8

0.6

Xcn IdWe
0.4

0.2

o o

o o.3 o o.3
Rcn i-ka
a)
Fig. 5.-a) Identification frequency [I MHz,
results
witJ1 4 MHz] and distance [0 mm, 2mm].
b) Identification results withfrequency [5 MHz, 9 MHz] and distance [0 mm, 2 mm]. c) Identification
results with frequency [I MHz, 9 MHz] and distance [0 mm, 2 mm].
1070 JOURNAL DE PHYSIQUE III N° 6

Nonndhed kp&mce plbre IhvlRe Venus I-k'

I

0.8
!
I
0.6 :
0.4 Xcn

0.2

o .3

cn i-ka

b)

Nonndhed itnpwhnce plane IhvRe versus 1-k'

l

0.8 f
I
t
~
"
o o
o
o.3

cn -ka
N° 6 EDDY CURRENTS: CONDUCTIVITY AND DISTANCE MEASUREMENTS 1071

Distance estimation. We can estimate the sensor-target distance in calculating the value

a
I k~ defined as the ordinate of the intersection point of the circle and the ordinate axis
=

(Fig. 6). point corresponds to the simpler case of use of eddy current
This sensors (there is no
losses and the magnetic field is tangent to the surface of the target). This case can be studied
with the electrical-image method which gives in some particular geometries of sensors an
analytical relation connecting a to the distance [I].
Conductivity estimation. As it can be seen previously, conductivity is coupled to the
slope of the line obtained when k varies. It is simpler to boil down to the case of
k I
=
the parametrization of this circle towards « can be effected by introducing the angle
(Fig. 6).

Notafions
I

I
,

I
,

M
~
I
0.4 ,
,
I
,

0.2

Rcn

Fig. 6. Definition of alpha and theta used for the model inversion.

Since R decreases more rapidly when frequency

low than high, it is is more interesting to
study variations of I/R versus frequency improve the linearity of
in order to the curve (Fig. 7).
In order to obtain the values of distance and conductivity, we made a first order development
of a
and I/R about the point defined by freq fo, dist do and « «o. = = =
Thus, we get the
following equations :

a = ao + a, dist and I/R

= ao + a, freq.

Thus the distance value is readily calculable

a a~
dist =

a,
1072 JOURNAL DE PHYSIQUE III N° 6

Alpha v©rSUS distance I*h©ta v©rSUS *eqU«CY

0.8 Freq=[3ADb;5ADb] Dhm[0.4mm;lmm]

0.6 *
.
+
+
1/0(1/°)
~
0.4 *
~
S '
+ ,
'
*
+ +
.
+
~ ~

0 0
0 0.5 1.5 2 0 5 10

Distance (mm) FreqU«CY @flh)

Fig. 7. -Graphs representing the relationships between alpha and distance and between theta and
frequency.

Taking into account that R is a function of the product frequency conductivity, we can calculate
the relative variation of conductivity

« «~ l/R a~ f~. a,

«o ai fo

3.2 RESULTS. Experimental tests have been performed by scanning a sensor over a graphite
composite plate which has been impacted. In order to test the method, the plate has been
voluntary tilted. The results of this scanning are presented below (Fig. 8) : the raw information
(real and imaginary parts of the sensor signal) is coded according to a colour scale (the whole
range of variations is coded linearly with 16 colours). We can see the drastic effects of a small
distance variation (about I mm) on the normalized impedance.
Our inversion method applied to the previous signals gives the results presented below
(Fig. 9). Useful information is now, respectively, the distance between the sensor and the
sample, and the local conductivity for each measurement point. With this excitation frequency
(4 MHz), the skin depth is about 2.5 mm, then the local conductivity estimated concems the
whole thickness of the samplp.
Owing to this method, we can simultaneously know if the impact has affected the surface of
the sample and estimate the intemal extent of the material deterioration. The degree of gravity
can be appreciated by studying the geometrical extensions and the conductivity variation
range.

4. Conclusion.

This signal
processing allows us to know simultaneously the distance sensor-target and the
local conductivity : the main
target interest of such a method is to have a control over
conductivity independent of the parasitic height variations during the inspection. This process
can be used in the nondestructive inspection of graphite composite materials for impact
N° 6 EDDY CURRENTS: CONDUCTIVITY AND DISTANCE MEASUREMENTS 1073

Rcn Xcn

Fig. 8. Maps of R~~ and X~~ obtained on a

graphite composite sample with impact, frequency =
4 MHz.

Distance Reldlive

Mnimvm: 1.67mm Minimum: -6.2%

Marimvm: 2.93mm Maximum: +8.9%

Fig. 9. Maps of the distance and relative variation of the local conductivity of the graphite composite
sample withimpact, frequency =
4 MHz.

detection for in-service materials or for the detection of fibre ratio variations after the
composite making. It should be noted that the method application area is greater than graphite
composite materials : this treatment can be applied to other conducting samples, like metallic
samples, provided that the frequency is changed. Some extensions of this model can be
obtained by taking into account both dependence of Im/Re versus distance and conductivity (or
frequency) for large ranges of conductivity measurements, or for systems with a variable or

multi excitation frequency (multifrequency could be used for estimate the local conductivity
versus the depth).
1074 JOURNAL DE PHYSIQUE III N° 6

References

ii PLACKO D., Contribution h la conception de capteurs inductifs pour la robotique industRelle, Rapport
de synthbse pour l'habilitation h diriger des recherches en sciences (22 mai 1990).
[2] HAGEMAiER D. J., Eddy current impedance plane analysis, Mater. Eval. 41 (1983) 211-218.
[3] VERNON S., The universal impedance diagram of ferrite pot core eddy current transducer, IEEE
Trans. Magn. 25, n 3 (may 1989).
[4] VALLEAU A. R., Eddy current non destructive testing of graphite composite materials, Mater. Eval.
48 (1990) 230-239.
[5] PLACKO D., SANTANDER E., Capteur de mesure de foibles £paisseurs h courants de Foucault,
Congrbs MESUCORA91, Paris, session n 15, pp. 3-17.