Non-destructive inspection of

carbon fibre-reinforced plastics

using eddy current methods
M.P. DE GOEJEand K.E.D. WAPENAAR

(TNO Plastics and Rubber Research Institute, The Netherlands)

Received 21 January 1991; accepted in revised form 26 November 1991

The possibilities and limitations of eddy current methods for the inspection
of carbon fibre-reinforced plastics have been investigated. The electrical
properties of these composites lead to modifications of current methods
applied to metals; in particular the required frequencies are much higher,
1-30 MHz for weave and 10-500 MHz for unidirectional reinforcements. A
proper design of the probes is essential. Single turn coils are adequate, and it
is shown how a higher sensitivity and suppression of the lift-off effect can be
obtained. Fibre orientations can be detected conveniently with dedicated
probes using a polar scan technique. The eddy currents are insensitive to
delaminations. Only for unidirectional reinforcements and extensive delami-
nation, over 20%, is the effect large enough to show up in C-scan images. In
contrast, fibre fracture is readily detected for unidirectional and weave
reinforcements, with a lower limit of about 8% fracture. As a result, eddy
current methods are useful to establish the type of defect when a composite
is damaged. This is confirmed by comparing eddy current and ultrasonic
measurements on composites with impact damage.

Key words: composite materials; eddy current; non-destructive testing;

ultrasonic inspection; conductivity; carbon fibre; impact damage

Carbon fibre-reinforced plastic-matrix (CFRP)compos- meaning of the results for the functional properties of
ites attract increasing attention for use in load-bearing composite components will be a prerequisite for their
components, particularly in aerospace. These materials application.
combine a low specific weight with excellent mechani-
cal properties and durability, especially when full use is Inspection techniques based on the principles of
made of the anisotropy in their properties. ultrasonic measurements are well developed and give
valuable information on delaminations, voids and
The rapid increase in the availability of improved inclusions. These techniques have the advantage that
matrix materials, e.g., the high performance thermo- they are suited for field application and single-sided
plastics such as polyetheretherketone and polyetherim- inspection. Other techniques, mostly for laboratory use
ide, and the progress that is made in the development only, include radiography, thermography, acoustic
of fast and reliable processing techniques, will widen emission and vibration tests (see, for example, Refer-
the scope even further. ences 1-3).
With these developments there is an increasing demand Eddy current methods find widespread use for the
for non-destructive inspection methods to ensure that detection of cracks and corrosion in metal structures.
the material meets the requirements, both initially and As in the case of the ultrasonic methods field applica-
in-service. For example, airworthiness requirements tion and single-sided inspection are feasible, with the
for civil aircraft metal structures are well defined by additional advantage that no couplant material is
certification authorities, but are lacking for the same required. Although the application of these methods to
structures made of composites because there is less CFRP composites was first reported 15 years ago 4, they
experience with these materials f. This situation will are still considered to be in the early stages of develop-
probably endure for many years in view of the ment 1"3. It has been shown that eddy current examina-
emerging new composites. The availability of inspec- tion reveals broken fibres, fibre orientation, and fibre
tion techniques and a proper understanding of the volume fraction4"5. Moreover, it has been claimed
0010-4361/92/030147-11©1992 Butterworth-Heineman n Ltd
COMPOSITES . VOLUME 23 . NUMBER 3. MAY 1992 147
recently that delaminations and barely visible impact where Zo is the impedance of the probe in air
damage can be detected as well I. (Zo=io~Li), and to is the angular frequency ((o=2:tf,
withfthe measuring frequency).
The present investigation was undertaken to gain a
better understanding of the possibilities and the limita- Fig. 2 shows the locus of Z/Zo in the complex plane
tions of eddy current methods for the inspection of representation for several values of R/(ooL2) and k.
CFRP composites. The principles of the methods are Although this treatment is an oversimplification, it
explained, and the values of important parameters-- shows that to observe variations in the resistivity of the
the measuring frequency and the probe dimensions-- material good coupling and a proper selection of the
are derived. The requirements for the measuring measuring frequencies are prerequisites. Variations in
equipment are determined by the electrical properties the resistance are indicative of defects, such as cracks
of the composites, which are quite different to those of or improper weldings in metal consiructions.
metals. Several types of probes are described. It is
shown that better results are obtained with specially In a more rigorous treatment the shielding of the
magnetic field by the conducting material has to be
designed probes, e.g., to detect fibre orientation or to
compensate for the lift-off effect. Finally, measure- taken into account. The penetration depth, or skin
depth, is defined as the distance in an isotropic, con-
ments are reported on samples with model defects, and
ductive material where the magnetic field is reduced by
with impact damage. These measurements are
63%:
compared with ultrasonic inspections of the same
samples. a = V'-(2/(t01ao))
where la is the magnetic permeability and o the conduc-
Principles and important parameters tivity of the material. The impedance of a coil above a
conducting plane has been derived by Dodd and
Eddy current methods are based on the induction of Deeds ¢'. Fig. 3 shows the results of their calculations as
eddy currents in a conductor by an applied alternating the impedance loci with r/6 and I/r as parameters, r is
magnetic field. The induced currents generate a the radius of the coil and I the distance between the coil
magnetic field opposing the applied field. The resulting and the plane.
magnetic field therefore depends on the local conduc-
tivity of the inspected material. More specific conclusions follow from these calcula-
tions. The penetration depth should be large enough to
A representation of the situation with an equivalent detect defects throughout the material thickness, unless
electronic circuit is frequently used to model the the detection of skin defects is required. The ratio r#3
response of eddy current measurements. Fig. 1 shows should have a value between about 1.4 and 5 (see also
the basic circuit of two coupled coils, representing the Reference 7) to observe sufficient variation in the
probe, L~, and the current loop in the material with impedance when the conductivity of the material
inductance L2 and resistance R respectively. The probe changes due to the presence of defects. At a given
and the material are coupled by the mutual inductance value of 6, about the thickness of the material to be
M, with:
M = kV'(LiL2)
where k is the coupling factor which depends on the
spacing between the probe and the material and on the
design of the probe. Its value varies between 0 and 1.
The normalized impedance of the coil L l in the
presence of the conductive material is:
Z _ 1 k2
Zo l-iR/(toL2)
0.5 1
f M t~

j 2
0 I
0.5

-Im 1~0}

Fig. 2 Complex plane representation of the normalized imped-

Fig. 1 Basic equivalent circuit for the eddy current coil-probe ance (ZIZo) of the circuit of Fig. 1, for three values of the
(L 1)-conductive material (L2,R) coupling coupling factor k

148 COMPOSITES • M A Y 1992

 the magnitude of the local conductivity and the measur-
ing frequency. Relations for the shielding effectiveness
have been presented in the literature s.`' . The reduction
in the magnetic field, IB/BoI, by the insertion of an
isotropic, conductive material of thickness d between
the coils is approximately given by'~:

2.08

where z is the distance between the coils, and the

3.51 assumptions are r/b> 10, r>>d and d/b< X/2.
This equation shows that the magnetic shielding,
expressed as -201og I B/B.I, will increase linearly with
log(o) and log(to). The advantage of this two coil
configuration is the absence of the lift-off effect.
However, it will rarely be possible in practice to
approach both sides of the material.
0.
For the configuration of Fig. 4(c) the normalized
impedance has the same characteristics as for the single
coil. This configuration is introduced because the
[. I 1
0.1 0.2
principle of a separate detection coil can be used for
measuring techniques with increased sensitivity and
-Im (Z~) suppression of the lift-off effect. These techniques will
be discussed later on.
Fig. 3 Complex plane representation of the normalized imped-
ance (Z/Z o) of a coil with radius r at a distance/from a conduc-
tive plane, according to Dodd and Deedss. b is the penetration Conductivity of CFRPcomposites
depth of the magnetic field
Eddy current methods can only be applied to materials
with a sufficiently high conductivity. In contrast to
inspected, this limits the size of the coil, and thus the
metals, CFRPcomposites show inhomogeneous and
spatial resolution of the eddy current method. Defects
anisotropic electrical properties. Moreover, the con-
much smaller than the thickness of the material cannot
ductivity is much lower. This is shown in Table 1,
be detected. The ratio I/r determines the coupling
where conductivity values of metals and composites are
between the coil and the plane, and should be as small
compared. For unidirectional composites the conduc-
as possible. A variation in/, e.g., due to surface rough-
tivity in the fibre direction is a factor of about 1(~)0
ness, causes a variation in the impedance along the
lower than that of metals, while the conductivity in the
dotted lines in Fig. 3. This variation is called the
cross direction is a further factor of about 100 lower. It
'lift-off' effect, and it should not be confused with the
was shown in the preceding section, in the equation for
effect of a variation in the conductivity of the material.
the skin depth, that the product too has to be within a
It must be emphasized that these calculations are
rather narrow range of values. A reduction in o by a
performed for isotropic conductors. CFRP composites,
factor of 1000 or more requires an increase in the
however, are anisotropic, so one ought to incorporate
measuring frequency by the same factor to observe
the anisotropic nature of the composites in the calcula-
comparable eddy currents in composites and metals.
tion. The calculations presented here are therefore only
Consequently, the present available eddy current
a brief indication of the influence of the different
systems are not suited for measurements on compos-
parameters on the eddy current technique.
ites. The required frequencies lie in the 1 to 500 MHz
So far the measuring probe has consisted of a single range. The large anisotropy in the conductivity, in
coil. However, it is also feasible to generate and detect particular of unidirectional composites, suggests that
the magnetic field in separate coils, see Fig. 4. In the eddy current methods are ideally suited to inspect for
configuration shown in Fig. 4(b) the two coils are the correct orientation of the fibres. The intrinsic
separated by the material to be inspected. The material inhomogeneity of composites, due to variations in
shields the magnetic field to some extent, dependent on fibre-fibre contacts and in the fibre volume fraction,
may cause difficulties in distinguishing acceptable
variations from defects.

EXPERIMENTAL
IIIIIIIIIIIIi'IIIIII " ~ " ~
Probes and measuring equipment
a b ~ c
The resonance frequency of the eddy current probes
should be well above the required measuring frequen-
cies that can range up to 500 MHz. In most cases single
Fig. 4 Different methods to generate and detect eddy currents, turn coils are adequate. The diameter of the coils is a
see text compromise between the required penetration depth

COMPOSITES. MAY 1992 149

Table 1. Specific conductivity of metals and CFRP

Material Conductivity Direction of measurement Reference

(10 4 S m -1)

Copper 5900 10
Aluminium 3500 10
I ro n 1000 10
Graphite 13 10
Carbon 3 10
Carbon fibre 4--17 11
Unidirectional 0.9-1.5 Parallel 4
CFRP 0.01-0.2 Perpendicular 4
2 Parallel 5
0.01 Perpendicular 5
0.13 Parallel 12
0.03 Perpendicular 12
3.6 Parallel 13
Satin weave CFRP 0.1-1 5

and in-plane resolution: r>6 (see Fig. 3), i.e., r>d, shown in Fig. 6. Measurements with and without
with d the thickness of the composite. Diameters of 5 to lift-off compensation are compared in Fig. 7.
10 mm were used, which limit the resolution to about
the same values. The resolution refers to the accurate It is concluded that lift-off compensation is possible for
location of the defect, and not to the size of the defect. eddy current measurements on composites. However,
The sensitivity of the measurements can be increased the correct tuning of the compensation coil is more
when separate coils are used for the generation and the involved than for metals, because there is much more
detection of the applied magnetic field. For detection variation in the average conductivity between samples.
two identical coils, a receiving and a compensating coil, Elliptical coils were used to detect the fibre orientation.
are connected in series in such a way that the induced Three possible configurations are shown in Fig. 8. The
voltages almost balance out. Small variations in the configurations where the generator and the detector
induced voltage in one of the coils disturb the balance, coil are not aligned (a and c in Fig. 8) were found to be
which can be detected with high sensitivity (differential the more sensitive, see later on.
measurement). Two possible configurations of the The equipment used to generate and detect the eddy
compensated probe are shown in Fig. 5, Configuration currents were: a spectrum analyser with tracking
b in Fig. 5 has also been used to study the suppression generator, Advantest TR4131/E and TR4153A, with a
of the lift-off effect. frequency range of 100 kHz to 2 GHz; and a signal
It can be concluded from Fig. 3 that lift-off and changes generator with vector analyser, Rohde & Schwarz SMS
in the conductivity of the test material have different and ZPV-E1/E2, with a frequency range of 100 kHz to
effects on the complex impedance of the system, 1 GHz.
namely a shift along the lift-off line and along the The eddy current measurements were compared with
impedance locus, respectively. By changing the ultrasonic inspections (C-scan). Apart from current flat
position of the compensation coil the impedance locus and focused ceramic transducers, Harisonic 1 40708--T
measured with the detection coil can be shifted to a of 12.7 mm diameter, focused transducers of 15 mm
position in the complex plane where lift-off mainly diameter were used that have been developed at the
causes a change in the phase angle, whereas changes in TNO Plastics and Rubber Research Institute on the
the conductivity mainly result in changes in the ampli- basis of the piezo-electric polymer polyvinylidene
tude of the impedance vector. This is schematically fluoride (PVDF).This material has the advantage that it
can be easily shaped to construct transducers with
a a C special features such as a line focus or an elongated
IIIlll~lllllllllllllll IIIIIII1~1111111111 I/l~llllllllll focus. The signals are generated and detected with a
a Sonic model FTS Mark IV ultrasonic flaw scope.

r t Samples
Two sets of composite samples were tested. The first
r t c
11111111111111111111 set was prepared by the Composites Group of TNO
d Plastics and Rubber Research Institute, and consisted
Fig. 5 Possible configurations for a compensated eddy current of carbon fibre/epoxy prepregs kindly provided by Ten
probe, t, r and c stand for transmitting, receiving and compen- Cate. Unidirectional (UD), CD 553 8448, and satin
sating coil, respectively weave (sw), CD 282 42, reinforcements were obtained.

150 COMPOSITES . MAY 1992

 Im

s' ' ' ~ ' s 0

at-'~
l--
i---~s..i

Re

50 mm
I I

Position
Fig. 7 Comparison between scans with the probe shown in
Fig. 5(b) on a carbon fibre weave laminate (sample D2, see
Table 2), with ( - - ) and without (..... ) tuning of the compensat-
ing coil for lift-off compensation. The amplitude of the induced
signal, V, is shown vs. the position along the scan. The distance
between the probe and the sample is increased from 0 to 9 mm
in steps of 0.11 ram. The dip in the signal about halfway across
the scan is caused by an artificial defect, see text. Measuring
frequency: 2 MHz

I1//11/1/111///11111111/11/I , , ; ' , , ..... ;;. . . . . ,;,........7,.............

a b c
Fig. 8 Schematic representation of the probes used to detect
the fibre orientations in carbon fibre composites. Every probe
consists of a transmitting and a receiving coil
cl ~ I
D1 by co-laminating pieces of Teflon foil, 5 × 5,
Fig. 6 Schematic representation of lift-off compensation with 10 × 10 and 25 x 25 mm 2 in size, between the second
the probe shown in Fig. 5(b). O is the origin of the complex and the third layer, and between the ninth and the
plane, and C is the position of the vector for the signal picked up tenth layer. Broken fibres were caused in sample D2 by
by the receiving coil in series with the compensating coil, in the making cuts of 10 and 25 mm length in layers 2 and 7,
absence of a conductive material. Moving the probe towards a
conductive material shifts this vector along the lift-off line V to and cuts of 25 mm length running through two subse-
point P. The dotted semi-circle is the impedance locus of the quent layers, i.e., 2 and 3, and 7 and 8.
conductive material-receiving coil combination. A change in the
conductivity of the material, or in the measuring frequency, The second set of samples was kindly provided by
shifts the vector from point P along the impedance locus. several producers of composite parts. Three of these
Tuning of the compensating coil shifts point C and the semi- samples, coded 11, I2 and 13, were used to study the
circle along the compensation line cl. It can be seen that correct detectability of impact damage. They were damaged by
tuning leads to a situation where the amplitude of the signal
(OP) is mainly affected by changes in the conductivity, and not
a dropping weight with a 12.5 mm diameter spherical
by li•off impact surface. During the impact the samples were
supported on a ring with inner and outer diameters of
The number and the orientation of the layers, the 50 and 65 mm, respectively. The impact energy was
overall thickness, the density, and the fibre volume determined by measuring the velocity of the weight just
before the impact.
content of the composite samples are given in Table 2.
The densities have been determined in a gradient Samples I 1 and 12 are sandwich structures with a core
column. The fibre content (fc) is calculated with: of a 6.8 mm thick honeycomb material, and skins of
carbon fibre weave laminates of 0.5 and 2 mm thick-
p - - (Or
/c- ness, respectively. Sample 13 is a 7.2 mm thick carbon
pf- p, fibre weave laminate.
where p, p, and pf are the densities of the composite,
the resin and the fibre, respectively. The densities of RESULTS AND DISCUSSION
the resin and the fibre are 1.20 and 1.76 g cm-3, Induction of eddy currents
respectively.
In non-magnetic materials high frequency magnetic
Two samples, coded D1 and D2, were provided with fields are shielded by the induction of eddy currents.
artificial defects. Delaminations were caused in sample The frequency range in which eddy currents are

COMPOSITES. MAY 1992 151

Table 2. Carbon fibre/epoxy composites studied

Code Reinforcement No of Stacking Thickness Density Fibre volume

type layers sequence (mm) (g cm -3) content (%)

1 UD 7 0/0 1.12 1.487 51

2 UD 20 0/0 3.19 1.448 44
3 UD 7 010190(3 x )/0/0 1.07 1.475 49
4 UD 7 0190 1.07 1.489 52
5 UD 7 01451451901-451-45/0 1.05 1.470 48
6 SW 3 0190 1.04 1.377 32
7 SW 10 0/90 3.05 1.45 45
8 SW 11 01901451-45 3.08 1.475 49
9 SW 13 0190 3.13 1.509 55
10 SW 16 0190 4.66 1.448 44
11 SW 15 0/90 3.90 1.539 61
12 SW 15 0/90 5.17 1.413 38
D1 UD 17 0145/90/-45 2.4
D2 SW 14 0•90 3.4

UD--unidirectional; SW--satin weave

induced has been determined by measuring the shield- one direction, show an intermediate conductivity. The
ing characteristics of the composites in the configura- value for sample 3 is about 0.1 × 104 S m-~.
tion of Fig. 4(b). The results of the measurements on
samples 1-7 (see Table 2) are shown in Fig. 9. It is interesting to compare the shielding of the samples
3 and 4. Both samples have four layers in one direction
It is seen that the shielding effectiveness increases with and three layers in the cross direction. In sample 4 the
the measuring frequency, as expected. This proves that orientation of the subsequent layers alternates, and the
eddy currents are induced in the CFRPcomposites like shielding is higher than that in sample 3 where the
in metals. However, for about the same material orientations are grouped together. This proves that the
thickness and carbon fibre content there are large impedance between the layers limits the eddy currents
differences in the shielding at a constant frequency. in these samples. The number of interfaces between
The samples with satin weave reinforcement show the layers with cross orientation is two in sample 3, and six
highest shielding. From the shielding at 10 MHz the in sample 4, leading to conductivity about three times
conductivity is calculated to be about 1 to 2 × 104 higher in sample 4, see Fig. 9.
S m-I. The unidirectional samples with all the layers
having the same orientation show hardly any shielding, It is concluded that eddy currents can be induced in
even at 1 GHz. This indicates a conductivity of less CFRP composites. The required frequencies are deter-
than 0.01 x 10 4 S m -1. These conductivity values agree mined by the conductivity, which for unidirectional
with literature data, see Table 1. The unidirectional laminates strongly depends on the number of interfaces
samples 3-5, with the fibre orientation in more than between layers with cross orientation. The conductivity
varies between less than 0.01 x 104 S m-I (fibres all
80 parallel), and about 2 × 104 S m-I for a satin weave
I
- - Satin weave / ('infinite' number of interfaces). The frequency ranges
- - Unidirectional // I I for 1-3 mm thick samples are 1-30 MHz for weaves
and 10--500 MHz for unidirectional laminates.
60 / /

/ ,~/
¢" 7 ~// ]r
Fibre orientation
o .0 Probes with elliptical coils, schematically shown in
," .' / s / Fig. 8, were used to determine the fibre orientation in
several samples. The probes were rotated over 360°
< ." ..',C// while the amplitude of the signal induced in the
detector coil is monitored. The results are shown in
Fig. 10. The probe with the two coils aligned does not
give a clear picture of the fibre orientation. The mea-
surements on sample 1, where all the fibres have the
0.1 I 10 100 1000 same orientation, merely show that this sample does
Frequency (MHz) not shield the magnetic field. The measurements on the
other samples show that there is more shielding, that is,
Fig. 9 Shielding of the magnetic field vs. the frequency for the
carbon fibre composites 1-7 of Table 2. Higher shielding means a lower amplitude, when the long axis of the coils is
higher conductivity of the sample. Transmission m e a s u r e m e n t s parallel to the fibres. It is remarkable that even for the
according to Fig. 4(b) samples with the satin weave reinforcement there is a

152 COMPOSITES . MAY 1992

visible angular dependence of the shielding. It appears tion. Fig. 11 shows the locus of measurements with an
that the conductivity is still not completely homoge- elliptical compensated probe (long axis 47 ram, short
neous in the plane of these samples, despite the axis 4 mm, measuring configuration like in Fig. 5(b))'on
'infinite' number of contacts between the fibres. samples 3-5, presented in polar coordinates. Turning
the probe over 360° (or 180° , which for reasons of
A clear picture of the fibre orientation is obtained symmetry gives all the information) gives a closed
when the two coils are next to each other. In this curve. These curves become very complex when the
configuration the fibres are part of an induction loop number of layers with different orientations increase.
formed by the two coils, and the elliptical shape is in The characteristic shape suggests that they could be
fact of minor importance. The amplitude of the signal used as 'fingerprint' to identify the composition of
should scale with the amount of fibres with the orienta- laminated parts by comparison with a standard. See,
tion under observation, in particular for the transmis- for example, the difference in the locus between
sion measurements where it is not affected by the samples 3 and 4 (Fig. 11) that is not found by merely
distance between the fibres and the coils. measuring the angular dependence of the amplitude
This is roughly confirmed by the measurements. In (Fig. 10). However, more research is necessary to
principle it should be possible to determine the stacking establish the reproducibility of the fingerprint, and to
sequence of the layers by varying the penetration depth enable the interpretation of deviations from the
of the magnetic field, i.e., by using more and higher expected shape.
measuring frequencies. This has to be investigated
further. The small additional lobes observed at angles
halfway between the angles of the fibres are probably
Sample homogeneity
caused by the geometry of the probe. In that case they Samples 3-12 (Table 2) were scanned with a
could be suppressed by changing the shape of the coils. 5 x 5 mm 2 eddy current probe in transmission. Typical
The phase angle of the induced signal has also been examples of the variation of the amplitude vs. the
considered as a means to characterize the fibre orienta- position on the sample are shown in Fig. 12. Two types
of inhomogeneities are observed: an irregular response
1 for samples 3-5 and more gradual changes in samples
11 and 12.
The inhomogeneities in the samples with unidirectional
reinforcement can be explained by the strong depen-
dence of the local conductivity on the interlaminar
fibre-fibre contacts. They reflect the statistical nature

,iiili
. . . ..... of these contacts. The observed spatial resolution of the
response, about 10 mm, is caused by the resolution of
the probe. It is clear that the detection of defects in
such samples will be hampered by these intrinsic
inhomogeneities.
The samples with satin weave reinforcement show a
more regular response. The gradual changes in the
observed signal for samples 11 and 12 are probably due
S to an uneven distribution of the pressure during the
lamination.
Because the effect of the interlaminar fibre-fibre
contacts on the in-plane conductivity is less for weaves,
it is plausible that the electrical contact between
crossing fibres in the weave, which is determined by the
lamination pressure and the fibre content, determines
the conductivity. The fact that the three samples that
8 show the higher variation in the signal--samples 6, 11
and 12--also have the lower fibre contents supports
this explanation.
It is concluded that the variation in the eddy current
signal due to intrinsic inhomogeneities of the compos-
ites is limited for weaves (about 3%), but can be very
large for unidirectional laminates (about 20%).

Fig. 10 A n g u l a r dependence of the amplitude of the eddy Delamination a n d broken fibres

current signals, measured by rotating the probes shown in From the results and discussion so far it can be con-
Fig. 8 over 360 ° (polar scan), s h o w i n g the fibre orientations of
the samples 1,3-5, 7 and 8, see Table 2. The different lines cluded that delaminations will be very difficult to
correspond to probes a ( - - ) , b (..... ) and c ( ..... ). Measuring visualize with eddy current methods. Eddy currents in
frequency: 50 MHz composites reinforced with weaves hardly depend on

C O M P O S I T E S . M A Y 1992 153
 Sample 3 Sample q Sample 5

~ 0o/180 o
/270o
q5°/22S°
70°

135o1315°

001180°
900•270 °
Fig. 11 Representation in polar coordinates of the amplitude and the phase of eddy current signals, measured by rotating an elliptical
compensated probe (seetext) over 360° for samples 3-5, see Table 2. Measuring frequency: 50 MHz

11 :u

l/ : , ,~ tl t

9 7 .... \?, ......

........ " •........ " "s"

"O

SOmm

Position
Fig. 12 Typical examples of the variation in the eddy current signal for transmission scans on samples 3-12 (see Table 2), shown as
the amplitude, V, of the induced signal vs. the position on the sample. The measuring frequency was matched with the conductivity of
the samples, to give about the same level of shielding (see Fig. 9). For the successive samples the frequencies were: 131,42, 51, 50, 4.2,
3.5, 3.3, 2.5, 2.3 and 2.7 MHz

interlaminar contacts. For unidirectional reinforce- more sensitive to broken fibres than to delaminations.
ments this dependence is strong, but intrinsic variations This means that for situations where delaminations can
in the interlaminar contacts may well obscure the effect be tolerated, but fibre breakage can not, e.g., where
of a delamination. components are only to be subjected to tensile stress,
the eddy current methods indicate the degree of
A comparison between an eddy current and an ultra- severity of the damage. 5 The claim that delaminations
sonic inspection of sample D1 (see Table 2) confirmed can be visualized as well is not supported by these
that, although the artificial delaminations caused by findings.
co-lamination of pieces of Teflon foil are clearly
detected with the ultrasonic method, they go unnoticed
with the eddy current method. It proves that the Impact damage
sensitivity of the latter method is not sufficient to detect
one delamination on a total of 16 interfaces. From the The results of the impact tests on the samples [1-I3 are
observed intrinsic inhomogeneity of unidirectionally shown in Table 3. The eddy current measurements
reinforced composites it can be expected that delamina- were performed with the lift-off compensated probe to
tions become visible when about 20% of the interfacial prevent irregularities of the surface, such as dents,
contacts are lost. Fig. 13 shows an eddy current inspec- causing misleading signals. The ultrasonic measure-
tion of sample D2 (see Table 2). The 25 mm cuts ments were performed with two focused transducers in
running through two subsequent layers are clearly transmission, a ceramic transmitter and a PVDF
revealed. Even a 25 mm cut in one of the 14 layers is receiver. Some of the damages were also inspected
detected. The 10 mm cuts in one layer go unnoticed. It destructively, by sawing right through them, to estab-
can be estimated that broken fibres will be detected lish the presence of broken fibres or delaminations.
when at least 8% of the thickness of the composite is
affected. The impacts on the two sandwich structures, I1 and I2,
give the same type of damages and inspection images.
It is concluded that eddy current methods are much The ultrasonic and eddy current C-scans obtained on

154 COMPOSITES. MAY 1992

 sample I2 are shown in Figs 14 and 15, respectively.
Of particular interest are the results on the impacts
with the lowest energy. In this case there is no visible
damage, none or a barely visible eddy current image,
and a clear spot in the ultrasonic image. Destructive
inspection revealed that this spot is mainly due to
cracks in the honeycomb core. The skin was hardly
weakened because there were only a few broken fibres.
At the higher impact energies the eddy current image
clearly indicates that the skin is damaged by fibre
fracturing, which is confirmed by destructive inspec-
tion. Delaminations within the skin were not found.
10 ran"
The ultrasonic and eddy current C-scan images of
sample I3 are shown in Figs 16 and 17, respectively.
From the eddy current images it can be concluded that
significant fibre fracturing occurs for impact energies
higher than 11 J. The damages detected with the
ultrasonic measurements at lower impact energies
should be mainly delaminations. This was confirmed by
destructive inspection of the 27.3 J impact damage.
With a dye-penetrant technique extensive delamination
was found, in addition to fibre fracture that originated
from a crack at the surface.
The eddy current images show various characteristic
features, see Figs 15 and 17. A halo is most commonly
observed. This could be caused by a relatively higher
conductivity in the centre of the defect, indicating that
fewer fibres are broken underneath the falling weight
than in its surroundings. However, it is more likely that
the eddy currents are less disturbed if the probe is right
above the centre of the failure, because the currents are
then just flowing around the failure. Another explana-
tion is that the tuning of the lift-off compensation
Fig. 13 Eddy current C-scan of sample D2 (see Table 2). The causes the measured amplitude of the signal to go
two dark bars correspond with 25 m m cuts through t w o of the
14 layers of the laminate, the greyish bar with a 25 mm cut
through a minimum when the conductivity decreases,
through 1 layer. Transmission measurement at 20 MHz see Fig. 6. The shape of the images can mostly be
described as a rounded-off square, which is caused by
the orientations of the fibres. It is not straightforward
how the size of the image should be correlated with the
size of the defect, because it is mainly determined by
the geometry of the probe. This is clearly demonstrated
in Fig. 13, where probes of about 5 mm diameter

2.8J 9.1J

5.6J 17.5J

Fig. 14 Ultrasonic C-scan inspection in transmission of sample

12, that was subjected to four impacts (see Table 3). A 7 MHz Fig. 15 Eddy current C-scan inspection of sample 12, compare
ceramic transmitter, focal length 36 mm, and a 8.6 MHz PVDF Fig. 14. A lift-off compensated probe was used at 2 MHz. One
receiver, focal length 39 mm, were used. One division is 20 mm division is 20 mm

C O M P O S I T E S . M A Y 1992 155
Table 3. Impact damage* detected on t w o sandwich structures with a core of a 6.8 mm thick
honeycomb material, and skins of carbon fibre weave laminates of 0.5 (11) and 2 mm (12) thickness, and
on a 7.2 mm thick carbon fibre weave laminate (13)

Code Impact Eddy current Ultrasonic image Visible damage

energy (J) image spot ~ (mm)

I1 0.69 nv 11-15 nv
1.33 halo 16-20 dent, bf
2.39 halo 20-26 hole
12 2.83 bv halo 18-22 nv, (bf)
5.59 halo 37 dent
9.12 halo 33 dent, (bf)
17.5 halo 41 hole
13 5.72 nv 15 bv dent
10.7 bv dark spot 22 dent, cm
27.3 dark spot, halo 29 dent, cm, (bf, d)
37.5 halo 33 dent, cm
41.7 light spot 41 dent, cm

* nv--not visible; by--barely visible; bf--broken fibres; cm--cracked matrix; d--delaminations. Remarks in parentheses refer to
destructive inspection

41.7J 10.7J

37.5J
27.3J

5.7J

L. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fig. 16 Ultrasonic C-scan inspection in transmission of sample 13, that was subjected to five impacts (see Table 3). The same condi-
tions as in Fig. 14 apply. The dark spot in the lower right corner is caused by a fastener. One division is 20 m m

depict a sharp cut through a weave by a bar of about CONCLUSIONS

10 mm width. In contrast, ultrasonic inspection with
focused transducers provides an accurate picture of the Eddy currents are induced in carbon fibre-reinforced
size of the defects (mainly delaminations). Moreover, it plastics, so inspection methods based on this phenome-
is relatively easy to build up a three-dimensional image non are feasible. The measuring conditions and the
of a defect with ultrasonic reflection measurements by properties and defects that can be detected are deter-
selectively mapping the reflections from a number of mined by the (high frequency) electrical properties of
depths in the sample. With a specially designed long- the material. Typical is the large variation in the
focus PVDF transducer, focal length 150 mm, it was anisotropy of the conductivity for composites with the
found that the delaminations in sample 13 extend same fibre volume fraction but different fibre orienta-
conically through the thickness, with the top on the tions, which leads to a required range of measuring
impacted side. frequencies of 1-500 MHz.

156 COMPOSITES. MAY 1992

 I
I0.7J
37.5J

41.7J

27.3J

37.5J
5.7J

i !

Fig. 17 Eddy current C-scan inspection of sample 13, compare Fig. 16• A lift-off compensated probe was used at I MHz. One division is
20 m m

Fibre orientation and fibre fracture can conveniently be Composites--TECQ 87, Guildford, UK, 22-24 September 1987
detected with eddy current methods, in contrast to edited by J. Herriot (Butterworth, Sevenoaks, 1987) pp 3-10
4 Owston, C.N. "Eddy current methods for the examination of
delaminations. The sensitivity of the methods relies on carbon fibre reinforced epoxy resins" Mater Eva134 (1976)
a proper design of the probes, e.g., by introducing pp 237-250
balancing, lift-off compensation or orientation• 5 Vernon, S.N. 'Parametric eddy current defect depth model and
its application to graphite epoxy" NDT lnt 22 No 3 (1989)
Eddy current methods are non-destructive, and field- pp 139-148
applicable in view of the relatively simple construction 6 Dodd, C.V. and Deeds, W.E. "Analytical solutions to eddy
of the probes and the possibility of single-sided inspec- current probe-coil problems" J Appl Phys 39 No 6 (1968)
pp 2829-2838
tion. They provide valuable information on the compo- 7 Bayer, H.E. (Ed) 'Eddy current inspection" Metals Handbook.
sition of carbon fibre-reinforced plastics, in particular 8th edition (American Society for Metals. Metals Park. OH.
the orientations of the fibres and on the severity of USA, 1976)
service-incurred damage, i.e., whether the tensile 8 Moser, J.R. 'Low-frequency shielding of a circular loop electro-
strength is affected by fibre fracture• magnetic field source" IEEE Trans EMC 9 (1967) pp 6--18
9 Bannister, P.R. 'Theoretical expressions for predicting shielding
effectiveness for the plane shield case' IEEE Trans EMC 10
ACKNOWLEDGEMENTS No 1 (1968) pp 2-7 and 'Further notes for predicting shielding
effectiveness for the plane shield case" IEEE Trans EMC I I
We would like to thank G•T. van Heck for the con- No 2 (1969) pp 50--53
struction of the eddy current probes and for performing 10 Hodgman, C.D. (Ed) Handbook of Chemistry and Physics. 43rd
edition (Chemical Rubber Publishing Company, Cleveland. OH.
most of the measurements, and P.W.A. Stijnman for USA, 1961)
the ultrasonic measurements• 11 Altmann, O., Winter, L., Riieker, D. and Schr'oder, G. "ZerstO-
rungsfreie Priifung yon Delaminationen in Bohrungcn von
This investigation was supported by the Dutch Ministry CFK-Laminaten mit Wirbelstrom" Kunststoffe 75 No 6 (1985)
of Economic Affairs, under the Innovation-Oriented pp 346-349
Programme on Polymer Composites and Special 12 Prakash, R. and Owston, C.N. 'Eddy-current method for the
Polymers (IOP-PCBP). determination of lay-up order in cross-plied cfrp laminates"
Composites 7 No 2 (1976) pp 88-92
13 Volpe, V. 'Estimation of electrical conductivity and electromag-
REFERENCES netic shielding characteristics of graphite/epoxy laminates" J
Composite Mater 14 (1980) pp 189-198
I Mahoon, A. "The role of non-destructive testing in the airworthi-
ness certification of civil aircraft composite structures' Compos-
ites 19 No 3 (1988)pp 229-235
A U THORS
2 Scott, I.G. and Scala, C.M. 'A review of non-destructive testing The authors are with TNO Plastics and Rubber
of composite materials" ND TInt 15 (1982) pp 75-86
3 Edwards, G.R. "The non-destructive testing of welds in continu-
Research Institute, PO Box 6031,2600 JA Delft, The
ous fibre reinforced thermoplastics" in Composites Evaluation, Netherlands• Enquiries should be addressed to Dr de
Proc 2nd Conf on Testing, Evaluation and Quality Control of Goeje.

C O M P O S I T E S . M A Y 1992 157