You are on page 1of 9

Ultrasonic attenuation as a measure

of void c o n t e n t in carbon-fibre
reinforced plastics
D. E. W. Stone, B. Clarke

The authors describe a technique for determination of the void content of carbon-fibre
reinforced plastics (cfrp) from its effect on the attenuation of ultrasound. They de-
scribe briefly the fabrication of cfrp laminates and the defects which cause an increase
in ultrasonic attentuation. The article concentrates on the effect of voids and shows
that void content and ultrasonic attenuation are directly related. This attenuation in-
creases with frequency. The authors establish a close correlation between ultrasonic
attenuation and interlaminar shear strength. Finally, the authors describe difficulties
in the application of the technique and suggest where further investigation is needed.

Carbon fibre reinforced plastics (cfrp) has been developed then the resin is extruded in the direction o f fibre alignment.
over the last decade to provide a range of materials which If, however, woven glass release and bleeder cloths are plac-
combine high stiffness and high strength with low density. ed between one of the platens and the pre-preg then the sur-
This combination of properties makes cffp particularly plus resin can also be extracted normal to the panel, passing
attractive as an alternative to aluminium alloys for many through the release cloth to the bleeder cloth. This method
aerospace applications. 1,2 It is also finding increasing applic- has been found to produce a higher quality of panel but
ation in other fields such as the textile industry 3 as cfrp the use of a release cloth results in a serrated surface which,
components will have much lower inertia, allowing improved as will be seen later, could cause errors when using ultra-
performance in high-speed reciprocating machinery. For sonics to determine void content.
commercial applications there are basically two types of Many practical components exhibit factors such as double
fibre available, both of which have a diameter of 6 - 8 / a m . curvature and are not readily produced in a simple press and
The high-modulus (hm) fibres have a tensile modulus of an autoclave is used instead (Fig. 1). Once again bleeder
about 480 GN m 2 and a tensile strength of about 1600 cloths are employed but the specimen is placed inside a
MN m -z, whereas the high-tensile-strength (ht) fibres have a rubber bag and the appropriate compaction pressure applied
modulus of about 280 GN m -2 and a tensile strength of by increasing the pressure Po in the autoclave, the bag read-
about 2800 MN m -2. A suffix s is used to indicate that the fly conforming to the contours of the specimen. In addition,
fibre has been surface treated to improve the strength of however, the pressure Pi inside the bag may also be adjusted
the fibre-matrix bond. Although the fibre can be supplied so that the compaction pressure is (Po-Pi). Thus by ad-
dry and subsequently impregnated with the resin matrix justing Po and Pi at the same time it is possible to maintain
material in a mould or in some other wet lay-up process, it the required compaction pressure while reducing Pi, if need
is convenient for the manufacture of many structural com- be below atmospheric pressure. This gives a greater degree
ponents to use previously impregnated (pre-preg) sheets. of control over the process and assists in expelling the sol-
These are sheets of partially cured resin containing a uniform vents which are often present in pre-preg. A subsequent in-
distribution of unidirectionally aligned fibres. A surplus of crease in Pi to a level significantly above atmospheric pres-
resin is present and this is bled off when the pre-preg sheets sure results in most of the remaining volatile elements going
are laid-up at the required orentation and subjected to an back into solution to give a void-free laminate. This part of
appropriate compaction pressure and temperature cycle to the process was developed by Hawker Siddeley Aviation Ltd
form a composite of predetermined fibre volume fraction. and details of the procedure are contained in the relevant
All the work described in this paper was done on panels application for a patent.
fabricated from pre-preg material with hts fibre in an
ERLA 4617/DDM resin matrix giving a composite of A comprehensive investigation aimed at optimizing the pro-
nominally 60% volume fraction. cedure was carried out by Ewins et al. 4 By varying Pi while
holding (Po-Pi) constant they were also able to produce
reasonably uniform panels with a predetermined void con-
Fabrication methods
tent. From the point of view of ultrasonic inspection, how-
The majority of panels for laboratory test purposes are fab- ever, the autoclave technique introduced a further difficulty
ricated in a simple press with heated platens and the stage because the surface of the resultant panels, as well as bearing
during the heating cycle at which the compaction pressure the usual serrated appearance caused by the imprint of the
is applied has a strong effect on the quality of the resultant release cloth, can also become convoluted. These convolutions
material. Another important factor is the way in which the are aligned with the fibres in the top laminate and can be
surplus resin is bled off. If simple polished platens coated quite coarse. Wavelengths of about 4 mm and a peak-to-
with release agent are used (to give a good surface finish), trough amplitude of 0.5 mm are typical when using 0.25 mm
David Stone and B. Clarke are at the structure department of the thick pre-preg, but the use of thinner pre-pregs of improved
Royal Aircraft Establishment, Farnborough, Hants, UK quality is expected to eliminate this effect.

NON-DESTRUCTIVE TESTING. JUNE 1975 137


crease m attenuation towards the corners and latcJ investi-
gations showed this to be associated with a wild content ot
/ Autoclavepressurevessel "~ about 4%. At first sight it would appear that both panels
exhibit a border of higher attenuation material, but this i~
rather misleading in the case of the autoclave specimen he-
~,\\\\ cause the method of applying pressure through a rubber bag
results in a chamfered edge about 5 mm wide. Sin~e !t is m~l
normal to the ultrasonic bemn the chamfered edge regiun in-
lexible membrane Membrane sealant troduces severe reflection and refraction losses giving the
appearance of high attenuation o f the C-scan. In order ~,~
demonstrate that the poor quality corners are a function ,,i
the manufacturing process, rather than the direction oF the
plies, a scan o f a hot-press panel with 0 ° 90 ° orientatiun ~s
shown (Fig.4). It can be seen that the same effect is obtained
Release a n ~ : f CFRP leminote / PI
at the corners of the panel. The white squares are deliberately
induced simulated delaminations.
cloths

Fig.1 Composite laminates can be formed and cured in an


autoclave

Material quality and ultrasonic t e s t i n g


A number o f different types o f defect can inadvertently be
introduced into cfrp components during the fabrication pro-
cess and it is not yet entirely clear to what extent each type
of defect will prove detrimental to the structural performance
of the component. The general question o f fabrication de-
fects and non-destructive methods for detecting and quantify-
ing them has been discussed elsewhere s by one o f the authors
and it was concluded that, whilst most o f the established ndt
techniques have some application to cfrp, ultrasonic inspec-
tion is the most generally useful inspection technique at
present.
Although there are a number of ways in which ultrasonics
can be applied to cfrp the most suitable method for produc-
tion inspection of sheet material is to use a conventional
pulse-echo or pulsed through-transmission system with a C-
scan presentation. In our work a 10 MHz probe was focused Fig.2 An ultrasonic C-scan of a cfrp panel made in a hot press shows
on the top surface and gated on the echo from a glass reflec- the alignment of the fibres (0° 45 °) and high attenuation at the edges.
The panel is 3mm thick of type hts fibre with E R L A 4617 resin,
tor plate so that the pulse passed through the specimen twice The attenuation levels shown are 4--8, 8--12 and 12--16 dB
(Figs 2, 3). A quantized display was used, so the various
attenuation levels are presented as finite changes to tone
density on a plan view of the specimen. The darkest regions
are of lowest attenuation and there is of course very little
attenuation in the absence of a specimen, which accounts
for the black border outlining the specimen.
One specimen was fabricated in a hot press with the fibres
aligned at 0 ° -+ 45 ° to the long axis o f the panel (Fig.2)
while the other was fabricated in an autoclave and the fibres
are aligned at 0 ° and 90 ° to the plate axes (Fig.3). A similar
bleeder cloth arrangement and the optimum curing cycle was
used in both cases. The laminate directions are clearly reveal-
ed in both cases and at the centre of the panels the attenuation
levels are very similar indicating that there is little difference ii iii l
in quality in this region. Fig.3 An ultrasonic C-scan of an autoclave-cured cfrp panel shows
less increase in attenuation at the edges. The 0 - 90 ° cross ply of
A marked difference is, however, observable at the edges and the fibres shows clearly for this 2mm thick composite of type I IS
corners of the panels. The hot-press panel shows a sharp in- fibre and E R L A 4617 resin. The attenuation levels are as before

138 NON-DESTRUCTIVE TESTING. JUNE 1975


The significance of the observed changes in attenuation level
must now be considered. There will be surface losses which
are dependent on the surface texture as discussed later, but
within the material itself the main factors influencing atten-
uation are the presence of voids; delaminations, state of cure
of the resin, fibre volume fraction; condition of the fibre-
matrix interface. Other features such as foreign inclusions

)
would also scatter the ultrasonic beam and increase the
attenuation, but the types o f inclusion likely to be met in
practice would be detectable by radiography. Attention
will later be concentrated on the effect of voids but let us
first consider the other factors noted above.
Delamination is the separation or lack of bond between in-
dividual laminates and delaminations will produce a high
degree of attenuation; if large enough, they may reduce the
transmitted signal to zero. It should, however, be possible
to distinguish between voids and significant delaminations
by using a pulse-echo system and observing the oscilloscope
(A-scan) presentation, because a delamination will produce
a separate back echo in between those from the front and
back surfaces. To distinguish between the individual echoes
from the front and back surfaces and from a delamination it
is necessary to use a focused probe of reasonably high fre-
quency (/> 10 MHz). It is in fact possible to distinguish
quite readily between the top surface echo and the echo
from a delamination (Fig.5). Delaminations as close as
0.25 mm below the surface have been successfully detected.
It is difficult to separate the effect of the state of cure of Fig,5 An ultrasonic p ulse-echo technique reveals detamination
the resin from that of the voids because incorrect cure will in a panel. The upper oscillogram shows the normal back-wall echo
while the lower shows the reflection from the delamination

often also result in the production of voids. Remember,


however, that the converse is not true; incorrect fabrication
procedures can result in badly voided material even though
the resin is fully cured. Also, provided that the resin in the
pre-preg is up to standard (which would be checked by
coupon tests), it is unlikely that incomplete cure would
occur in practice without this being revealed by monitoring
the autoclave temperature.
Fibre volume fraction does not appear to have a very strong
effect on attenuation, and it will be shown later that this
effect is small compared to that of voids and need only be
taken into account when obtaining precise measurements
of void content. In practice the mean value of the volume
fraction, taken over the volume of material through which
the ultrasonic beam passes, does not vary very much from
point to point within panels fabricated from pre-preg.
Various forms of surface treatment have been devised for
carbon fibres aimed at improving the bond between the
fibre and the matrix, and hence at increasing the interla-
minar shear strengths and transverse tensile strength of the
composite. No evidence has, however, yet been produced
on the effect of the surface treatment on the ultrasonic
attenuation, although it seems likely that a failed or weak
bond would increase the degree of internal scattering and
hence the attenuation.
Now a large number of cfrp specimens are of reasonably
constant volume fraction, contain no significant delamin-
ations and are fully cured, yet they still exhibit quite a
Fig.4 An ultrasonic C-scan of a 3 m m thick hot-press panel w i t h
wide variation in ultrasonic attenuation. This is due to the
0 90 ° crossply shows the high attenuation at the edges associated
with the hot press process. T w o delaminations appear in the scan presence of voids which scatter the ultrasonic beam. The
of 12 ply E R L A 4 6 1 7 composite w i t h unground surfaces. The diameter of tile voids is, of course, in general only about
attenuation levels are 4--7, 7 - 1 0 , 10--13 dB 30/am, which is an order of magnitude less than the wave-

NON-DESTRUCTIVE TESTING. JUNE 1975 139


lengths of the ultrasonic beam so that individual voids can- diameter of the parallel ultrasonic beams. For these reason~
not be detected, although it has been suggested 6 that ul- it was decided to employ a simple through-transmission
trasonic spectroscopy could provide a means for distin- system with parallel probes.
guishing between voids of different diameter. For a given A standard commercial immersion-tank system was used with
frequency, however, the attenuation is strongly dependent tile probes mounted 100 mm apart in a calliper, which was
on the void content and this paper will show how ultrasonic carefully adjusted to ensure that the axes o f the two probes
attenuation measurements can provide a quantitative non- were aligned so that the iransmitted signal was a maximum
destructive method for determining void contents in cfrp The exact distance between the ,~robes was not considered
laminates. It is important that this should be done, be- important because parallel bea: ~s were produced and the
cause it has already been established 7 - 9 that the presence attenuation of the water is low compared to that o f the cfrp,
of voids will significantly reduce the interlaminar shear but to ensure maximum reproducibility the same separation
strength (ils), the compressive strength and the transverse was used throughout. The transmitter, receiver and display
flexural strength of cfrp. Hancox 1o has suggested that units employed were the standard modules from a Series 10
the ils is very sensitive to 1% voids, while above 5% voids Ultrasonoscope except that a calibrated step attenuator
it is the compressive strength that is most affected. In most replaced the standard variable attenuator.
of the above tests the correlations quoted used void content
figures obtained by destructive means such as resin burn- With the panel located mid-way between the probes one or the
off or acid digestion. 11 These are essentially rather labor- probes was used for pulse-echo testing and the calliper unit
ious point by point measurements giving an average value adjusted to maximize the back echo. This ensured that lhe
over a finite volume of material and whilst they can give ultrasonic beana was normal to the plane of the panel. The
some guidance as to the variation of the void content a- instrument was then switched back to through transmission
cross the plank, they say nothing about the uniformity of and the attenuator adjusted to bring the trace corresponding
the mechanical test specimen. Furthermore being destruc- to the first received pulse on to some fixed point on the os-
tive they can only be performed on a specimen after it has cilloscope screen. The probe carriage was then moved until
been tested mechanically, or on a region o f the plank the calliper was clear of the specimen and there was only water
immediately adjacent to the test specimen. It is clear between the probes. The attenuator was again adjusted to
therefore that a more detailed investigation of the effect of bring the oscilloscope trace back to its original reference
voids on mechanical performance requires an accurate non- point. In this way the attenuation A T was measured directly
destructive determination o f the void content at all points in as the difference between the two attenuator settings. Al-
a plank. This would allow mechanical test specimens to be though the attenuator was calibrated in steps of 0.5 dB. inter-
cut in such a way that there was a uniform, known void polation using the oscilloscope display enabled attenuations
content within the gauge length. In section 6 correlations to be measured with a sensitivity of about + 0.2 dB.
are demonstrated between the ultrasonic attenuation and Attenuation readings were then taken using this procedure on
both the void content and the ils. the selected areas of the various panels. Three different sets
of parallel probes were used in turn, having centre frequencies
of 2.5, 5 and 7 MHz. The panels were then cut up in such a
Test procedure way that two types of specimen could be cut from each o f the
A series of 2 mm thick panels o f unidirectional lay-up and selected regions: a standard L2 interlaminar shear strength test
having different void contents was produced by varying the specimen 12 x 10 x 2 ram; and a specimen of minimum di-
autoclave procedure, as described earlier. Both surfaces of mensions 12 x 5 x 2 mm for destructive 11 evaluation of
the panels were ground so that all the panels were of con- volume fraction and void content.
stant thickness and constant surface finish. Each panel in
turn was scanned ultrasonically using a 10 MHz focused
probe having a focal spot size of about 1 m m diameter. The
use of a focused probe gave improved resolution which en- Theory
abled small local changes in attenuation to be displayed on Consider a simple through-transmission system using a pair of
the quantized C-scan record. By examination o f the records probes each capable of producing a parallel ultrasonic beam.
for each panel it was possible to identify areas which exhibit- If the axes of the two probes are aligned in a calliper and the
ed uniform attenuation at various different levels; an area of assembly immersed in water then a short pulse leaving one
at least 15 x 12 mm was required for the subsequent mechan- probe will be detected by the other. If we assume a simple
ical testing. The selected areas were then outlined with plane wave solution then the magnitude o f the pulse leaving
masking tape which made them easy to identify during the transmitter may be measured in terms o f the maximum
subsequent ultrasonic testing. values of particle displacement ~o, the acoustic pressure
It would have been possible to continue using a focused Po or the acoustic intensity I o. These are related by p,, =
probe and glass reflector plate system (or two focused probes pvco~o and
for through transmission) to obtain attentuation measure~
ments when investigating the correlation with void content.
This would, however, have necessitated maintaining constant
probe-to-specimen and specimen-to-reflector plate distances 2pv
in all tests, and cotdd have introduced errors when specimens
of different thicknesses were compared. It would also have where p = tile density of the transmitting medium, v = tile
made comparison with probes used elsewhere more difficult. velocity of longitudinal waves in the medium, Z=pc = the
Furthermore there is little point, at this stage, in improving acoustic impedance o f the medium, co = angular velocity
the resolution when the void content values themselves are = 27rf. Suffix o indicates that the quantities refer to the
obtained as a nrean value over an area rather larger than the initial transmitted pulse.

140 NON-DESTRUCTIVE TESTING . JUNE 1975


After suffering attenuation the corresponding values for Combining the equations for total attenuation due to the
the pulse arriving at the receiver are ~, p and I and the total specimen and for these separate attenuations, the total
attenuation may be expressed in decibels as attenuation is

~o Po Io Vl
A = 20 log10 ~ - - = 20 lOglo P - - = 10 lOglo I- - 20 log ~-2 = AT = Af + At + Ab

Now since the piezoelectric elements in the probes produce Now, provided that the void content is not sufficient to
a voltage V proportional to the particle displacement or to alter Z c significantly, Af and Ab are independent o f the
the acoustic pressure, the attenuation may be defined as condition of the panel, apart from its surface finish, and
A = 20 loglo Vo/K Let the received voltage after the pulse would be expected to be constant for panels of different
has been transmitted through a simple water path be V1 thicknesses. The extent to which this assumption is valid
and let the received voltage with a specimen present between will be discussed later. At, however, will increase linearly
the probes be II2. The total attenuation caused by the pre- with the panel thickness and with the absorption coefficient
sence of the specimen will then be AT = 20 loglo V1/V2. a. The latter is governed by the internal condition o f the
cfrp material, especially the void content, and it will later
If the specimen is in the form of a fiat panel aligned normal
be shown that there is a marked frequency effect.
to the ultrasonic beam, this total attenuation will be com-
posed o f three parts: losses at the front surface (water/cfrp), It is convenient to express the results as attenuation per
attenuation taking place as the pulse passes through the unit thickness, (dB mm "l) in the form AT/t = As/t + a, where
thickness t of the panel, losses at the back surface (cfrp/ A s = Af + A b.
water). Now if the surfaces of the cfrp panel are perfectly The theory of wave propagation in anisotropic media has
flat then the standard expression for the transmission co- been considered at some length by Musgrave 13 and the par-
efficient at an interface may be used ticular problems of propagation in cfrp have been discussed
by Reynolds and Wilkinson.14,1s For the purposes of this
Pt 2Z2 paper we need only consider longitudinal waves whose dir-
k.t - ection of propagation is normal to the plane o f the panel
Pi Zl + 22
(Fig.6). Being a unidirectional composite the specimen ex-
hibits transverse symmetry and possesses five independent
where Pi is the incident acoustic pressure, Pt is the trans- elastic stiffnesses cl l, Cl2, Cl3, c33 and c44 , where 1, 2 and
mitted acoustic pressure on passing from a medium o f 3 refer to the x, y and z axes respectively. The z axis lies in
acoustic impedance Z1 to a medium o f acoustic impedance the direction of fibre alignment whilst the x and y axes are
z2. arbitrary orthogonal axes in the plane of isotropy. For
Thus, if the suffices c and w are used to indicate cfrp and simplicity we shall take the x axis to be normal to the plane
water respectively, then at the front surface the incident of the panel. We are therefore concerned only with the
acoustic pressure is Pw and the acoustic pressure transmitted velocity of longitudinal waves whose normal lies in the x
into the cfrp is Pc. Similarly at the back surface the incident direction; this velocity vc is given by
acoustic pressure is Pc and the acoustic pressure trnasmitted
back into the water is Pw where CI1

P c

Pc _ 2Zc and p'(' - 2Zw


pw Zw+Z~ p~ &+Zw
Results and discussion
Thus the front surface losses are All the tests were performed on 2mm thick panels o f uni-
directional hts fibre in an ERLA 4617 resin matrix. The
Zw + Zc volume fraction was found to vary very little amongst the
Af = 20 lOglo various samples and may be taken as 60% • 1% throughout.
2Zc Accurate density measurements were required for the mea-
surements o f void content and a value of Pc = 1548 kg m 3
In fact there is a gain in acoustic pressure on passing from a
less dense to a denser medium, so that this term will be
negative.
The back surface losses are

z,v+&
A b = 20 lOglo . . . . Fibre direction :
2Zw z/(3)

If the surfaces are not fiat, however, scattering o f the beam


will take place and the above losses may be increased.
Finally the transmission losses on passing through a speci-
men of absorption coefficient a and thickness t are given F ig.6 Directions are chosen arbitarilly to describe the anisotropic
by A t = c~ t. properties of the fibre reinforced composites

NON-DESTRUCTIVE T E S T I N G . J U N E 1975 141


was found for samples with a void content of virtually zero. samples. To check on this, a sample of high-quality material
It is difficult to obtain very accurate values of void content was carefully ground down to form a series o f parallel steps
using the acid digestion method, however, and all values are increasing from 0.3--1.8 mm. Using 7MHz probes a linear
only quoted to the mearest ½%. Before considering the ex- increase in attenuation with thickness was obtained (Fig.8 !.
perimentally determined ultrasonic attenuation results it is This indicates that the attenuation at zero thickness (ie the
helpful to consider the theoretical surface losses, and to do surface loss) is 2.2 dB and also shows that for this sample
this it is necessary to know the velocity o f propagation of the attenuation coefficient was 0.68 dB mm "1 corresponding
longitudinal waves normal to the plane of the panel. Mea- to 3.55 dB at 2ram thickness. Now the lowest attenuation
surements of velocity on a thicker sample of the same material obtained on 2ram thick samples at this frequency was
gave a value of this velocity vc = 2980 m s] which is consis- 3.2 dB. It is reasonable to suppose therefore that the latter
tent with the results quoted by Dean and Turner ~6 on a sample was of effectively zero void content whilst the stepped
similar material. They quote Pc = 1560 kg m -3 and Cll = sample, although of high quality, nonetheless contained a
13.7 GN m -2, giving vc = 2960 m f l . Substitution o f these very small percentage of voids, insertion of the single point
acoustic impedance values in the equations derived earlier for the zero-void sample predicts an attenuation coefficient
shows that the theoretical total surface loss As = (Ar + Ab) of 0.50 db mm q as found.
is 2.66 dB (Table 1). Now it would be convenient if the surface losses remained
The experimentally determined variation of ultrasonic atten- essentially constant allowing c~, the attenuation per unit
uation with void content Vv is known and it may be seen thickness, to be presented directly against the void content
how for good specimens with effectively zero void content giving a calibration for panels of any known thickness.
the attenuation is relatively independent of frequency hav- Accurate determination of these surface losses is clearly
ing a constant value of about 3 dB (Fig.7). This is very close more important for thinner material, when these losses
to the suggested theoretical surface loss and it is reasonable account for a high proportion of the total attenuation. We
to suppose that the internal losses at are very low for these have shown how At and Ab are functions o f the acoustic
impedance of the composite Z c = pcVc. Thus the surface
losses will change if ~'c, the velocity o f wave propagation
52
through the composite were to alter significantly. Now
3oi experimental observation is has in fact shown that the
presence of voids will decrease this velocity: as a guide tests
28 on the authors' material showed that a void content of 5<;
reduced the velocity by about 6%. Thus 5% voids would
26
reduce vc to 2800 m f l , while at the same time reducing
24 the overall density of the composite Pc from 1548--1486
kg m-3. The effect of both these changes is to reduce Z~
22 to 4160 x 103 kg m -2 s~1 , which has the effect o f reducing
2O / A s to 2.22 dB. This is a comparatively small change when
compared with the attenuation values for voided material
5 MHz
but a correction has been made for this effect before replot-
_g ring the data.
The effec~ of decreasing the fibre volume fraction from (~0(~;
to 50% may be shown theoretically, to reduce vc by about
654 whilst at the same time reducing Pc to 1498 kg nr 3
t2}- Thus Z c is reduced to about 4187 x 10.3 kg m 2 s4 which
io) may be seen to be very comparable with the reduction caused
by 5% voids: the effect on A s is therefore again quite small,
i
MHz It has in fact been proposed that the variation m velocity

! Stepped specimen of
i good quality but not zero V v /
3 2 ~- e= 068dBmm -I v ~ / / : ~

0 _L _L__L__± I 1 1 L 30 × / /
0 05 tO 15 20 25 30 35 40 45
Void content Vv [ % ]
Known zero VV
Fig.7 The total attenuation of ultrasound by cfrp increases w i t h 28 [ / / / ~=050dBmm-'
void content w i t h greater increases for higher frequencies

<
24
Table 1. Relative properties of cfrp and water
22
Velocity Density A c o u s t i c impedance
Material [m s4 ] [kg m -3] 103 kg m 2 sq ] !
9 n ~ I t I I ,i • 1. . .t . •
0 02 04 06 08 I0 12 14 16 [8 2¢
CFRP v~=2980 P c = 1548 Z c =4613 Thickness [mm]

Water v,,, = 1483 Pw = 1000 Z w = 1483 Fig.8 The attenuation of ultrasound by cfrp increases linearly w i t h
thickness from a fixed m i n i m u m value

142 NON-DESTRUCTIVE T E S T I N G . JUNE 1975


with void content could itself be used as a method o f de- ed are the centre frequencies of each of the probes as measur-
termining void content. However, although the percentage ed using continuous wave excitation. The highest frequency
changes in velocity are quite large the transit times involved probe had in fact a nominal centre frequency o f 10 MHz.
when measuring velocities in typical laminates are themselves Smooth curves were drawn through the experimental points
small. Take, for example, the measurement o f the time delay and it is interesting to examine the relationship between v,,
between the echoes from the front and back faces of 2mm and oe, because there is at present no established theoretical
thick laminate. Assuming the same velocity o f 2980 m s4 , basis on which this relationship could be predicted. Lloyd t9
the transit time for a specimen with no voids would be has considered the effect o f a finite number of scattering
1.342/as, whilst a specimen with 1% voids would have a centres and has attempted to construct a mathematical model
velocity o f about 2944 m s"l giving a transit time o f to describe the frequency dependence o f grain-boundary
1.359/as, ie there is a change in transit time of 17 ns. scattering in metals. As yet, however, there has been no applic-
ation o f this model to cfrp, although Brown 6 has demon-
Whilst precision equipment such as the NPL ultrasonic test
strated that there is an increase in ultrasonic back-scattering
facility 17 can measure transit times to an accuracy as high as
when fatigue damage has occurred and he attributed this to
-+ 1 ns, most commercial test sets are not capable o f the re-
the nucleation and growth of distributed scattering centres.
quired accuracy. Also if simple through transmission is used
He also stated that this resulted in a reduction in the high
then the temperature of the water bath is very important. A
frequency response relative to the low frequency. To see if
I°C rise in temperature from 20 ° to 21°C will increase the
there is a simple relationship between Vv and ~ the experimen-
velocity of compressional waves in water from 1 4 8 3 - 1 4 8 5 . 5
tal relationship was approximated by parabolae o f the form
m s4 , which for a 100mm long water path means a change in
c~= k ~ . It may be seen that this gives quite a fair approxi-
transit time of 113 ns. Thus for general inspection purposes
mation to the observed relationship although there is some
the use of attenuation measurements is to be preferred, al-
discrepancy below about 1% Vv for the 5 and 7 MHz curves,
though clearly there are situations where the velocity method
whilst the curves for 2.5 MHz begin to deviate above 3% Vv.
would be more appropriate,
The way in which this constant k varies with frequency may be
Where the attenuation coefficient is plotted against void shown to be exponential and the three theoretical curves
content there is of course some scatter, arising largely from may be expressed as ~ = 0 . 0 7 9 4 f 1"27 V~v.
the acid digestion method of measuring the void content,
Now if tile size o f individual voids remained constant and
but a family of curves has been clearly established and it is
their number simply increased with void content then it
shown how the attenuation losses caused by the presence o f
might be possible to base a hypothesis on a linearly increas-
voids increase with frequency (Fig.9). The frequencies quot-
ing number of distributed scattering centres. For planks
manufactured with pre-preg, however, experimental observa-
tion shows that the total volume of voids may roughly be
13 - ' Experimentol results divided into two types. Up to about 1.5% Vv the voids tend
! --~-- Al"rempted curve f i t
using porobole a =kVv2
7 MHz to be spherical, with the void diameter varying from about
12
5tzm to 20/Ira. These voids are due to the various volatile
elements present and there is some evidence that the size of
individual voids increases with Vv. Beyond about 1.5% Vv
interlaminar w)ids caused by air trapped between the lamin-
ates start to predominate; these are flattened and elongated
,o b and tend to be significantly larger than the volatile-induced
voids. This two stage process immediately suggests that a
-~ 9 bi-linear relationship might be an alternative way o f describing
'E
E the variation of attenuation with void content. The experi-
Q3
8 mental data was therefore re-plotted and bi-linear functions
fitted (Fig. 10). Once again there is a certain amount o f un-
7~ certainty at low values of Iv, but overall the bi-linear relation-
ship appears to give a rather better fit to the experimental
results. The slopes k I of the initial parts of the function can
only be determined rather arbitrarily, but the slopes k 2 of
the secondary parts of the function are much more clearly
<t
defined. It is interesting to note that k2 = 0.87f I .ss.
k :0 251 /
4 /
Attenuation measurement for quality assurance
The previous section has demonstrated that ultrasonic atten-
uation measurements can be used to give a good measure of
the void content in a particular material. It is likely that
similar, although not identical, calibrations would be obtained
on cfrp laminates employing other resin systems and this is
clearly a factor that must be investigated. Equally important,
0 [ I I I I I I I I however, is the need to investigate other parameters that can
0 05 tO 15 20 25 30 55 40 vary in a system employing a given fibre and resin and that
Void content Vv [ % ] could also affect the attenuation. These other factors are
Fig.9 T h e c o e f f i c i e n t o f a t t e n u a t i o n of u l t r a s o u n d b y c f r p is
discussed earlier and, whilst it seems probable that the varia-
related to the v o i d c o n t e n t in a regular r e l a t i o n s h i p tion in fibre volume fraction normally exhibited by panels

NON-DESTRUCTIVE TESTING. JUNE 1975 143


L 7 MHz / o
tile attenuation measurements are nlore accurate than the w)id
13 content figures the results of the ds tests are plotted agains~
c~ and only spot values of K, are shown. It should be noted
that little scatter is exhibited, despite the fact that the atlenuu-
t2
tion measurements give a mean value through lhe thickness.
whilst the shear strength is likely to be particularly suscep.
II tible to defects lying on the neutral axis. In order lo estimate
the accuracy of the technique for high quality material a
I0 number o f ils specimens were prepared from samples having
a values less than 2.5dB mm ~ . This data was plotted in (Fig.12).
It should be noted that each of the three points at an ils value
9 5 MHz
7 of about 120 MNm -2 is in fact the mean of nine tests, having a
E
E scatter in each case of less than 1.5 MNm 2. The overall scatter
co 8
13 in this region is only +- 3 MNm 2. It will of course he necessary
~t
to investigate the effect of voids on other nlechanical tx opc~-
7 ties, particularly the transverse tensile strength, bel\~re an
a=
acceptable void content can be specified for a particular mater-
o
ill. but once this has been done it would appear thai ultrasonic
g 6
attenuation will provide a satisfactory means of measuring
g
it.
a 5
No general correlation of attenuation with ils is possible at
present, however, because factors such as fibre surface
4 treatment are known to have a large effect on file ils whilst,
as already stated, its effect on attenuation is as yet tmknown.
3 Once again it should prove possible to present the relation-
5 MHz ship in tile form of a family of curves covering say a r;mge
2
of specified surface treatments.

Conclusions
It has been shown that ultrasonic inspection provides a
_ 1 3 1 powerful inspection tool for cfrp laminates with both void-
O0 I 2 3

Void content Vv {%]


120 (

Fig. lO The relationship between void content and attenuation


coefficient can be described by a bilinear function 118

116

fabricated fronl pre-preg material is unlikely to cause signific-


ant errors, the effect of incomplete cure of the resin or varia- 114
tions in fibre surface treatment are as yet largely unquantified.
Even if these factors prove to be significant and of practical
112 F xX
importance it should still prove possible to quantify their
x x x
effect, and to replace the single curve of attenuation versus '5
Vv (for a given frequency) by an appropriate family of curves. I lO !-

The results obtained so far, however, were all obtained on


laminates fabricated from a given pre-preg material and taken 108

to a fully cured state. As far as possible only Vv was varied x


x x x ×
and it is therefore reasonable to assume, in the absence of 106
co
further information, that the attenuation changes demon- ×
c
strated are due only to this variation in Vv. Furthermorethe E IO4
o
pre-preg material used is representative of currently available
commercial material, using fibres with a standard surface
treatment, so that the specimens tested varied from samples 102 x×
of the best cfrp that can be fabricated from commercial pre-
preg, to samples of the same material degraded only by the I00 × ×
presence of voids.
Now if ultrasonic methods are to be used to inspect such a 98 Void content Vv
materials for quality control then it is necessary to specify 05 IO 125 115 XX,j
an acceptable attenuation level and to do this the effect of 96 .... r L ~q--~-r-~ 1 ~- H l
the voids on the mechanical strength properties must be in- 05 075 I0 t25 15 175 20 22,~ 25
vestigated. The variation of one important property, the inter- Alffenu~tion coefflcDent ~[dB mm-I]
laminar shear strength (ils), has been measured 4 in parallel Fig. 11 The shear strength of cfrp is related to the void content
with the present work (Fig. 11). Because it is considered that and therefore to the attenuation coefficient

144 NON-DESTRUCTIVE TESTING . JUNE 1975


ed areas and delaminations readily revealed and quantified.
Ultrasonic attenuation measurements on constant thickness
panels of unidirectional material have been correlated with
void content and a simple b&linear relationship postulated. ( standard deviotlon I48 MNm-’
variation coefficient I 23%)
For one particular fibre-resin system in a fully cured state
a good correlation has also been demonstrated to exist
between ultrasonic attenuation and interlaminar shear
strength.
It will be necessary for a number of other parameters to be
investigated before these results can be fully utilised. In
particular it is necessary to establish the applicability of
these relationships to angle-ply material, and to investigate
the effect of different resin systems and fibre surface
treatments.

Acknowledgements
The authors would like to thank Mr R. Childs for providing
samples of deliberately voided material and for performing
the destructive evaluation of volume fraction and void
content. They would also like to thank the Ministry of
40 -
Defence (Procurement Executive) for permission to pub- Void content V, [%I
lish this paper which is Crown Copyright. 0 15 20 25 30 35 40
I
30-l ’ ’ ,’ I I I I I
0 2 4 6 8 IO 12 I4 I6
Attenuotlon coefflctent aIdBmm~‘l
References
Fig.12 Within broad limits the shear strength of high-quality
1 Ham, A. C. Potential aerospace structural application of
cfrp, The Plastics Institute, International Conference on cfrp is linearly related to void content and attenuation coefficient
Carbon Fibres, their Composites and Applications, Paper
43 (1971)
2 Harris, G. Z. Composite Materials and their applications in
aeronautics, De lngenieur (1970) (21) Ll-LIS, also RAE TR Lucas, R. Carbon Fibres - the versatile performer
69091 (1969) Chartered Mechanical Engineer 20 (1973 j (4) 6 2-67
Ewins, P. D., Childs, R., Clarke, B. C. RAE Report to be
published
Stone, D. E. W. Some problems in the nondestructive testing
of carbon fibre reinforced plastics, RAE TR 74080 (1974)
Symbols Brown, A. F. Materials testing by ultrasonic spectroscopy
Autoclave pressures inside and Ultrasonics 11 (September 1973) (5), 202-210
9, P,
Avco Corporation Systems Division. Effect of voids on
outside the rubber bag
mechanical properties of graphite fibre composites
AVSD-0166-71-RR AD127236 (1970)
P Acoustic pressure
8 Anderson, R. T., Delacy, T. .I. Application of nondestructive
l Particle displacement testing for advanced composites, Aerospace-AFML Conference
on NDT of plastic/composite structures, Dayton, Ohio (1969)
I Acoustic intensity 9 Purslow, D. RAE TR. To be published
Density of the transmitting medium 10 Hancox, N. L. AERE Harwell, Private communication
P
11 Ewins, P. D., Childs, R. The determination of content by
w = 2rlf Angular velocity volume of fibre resin and voids in carbon fibre reinforced
plastics, RAE TR 72082 (1972)
Z Specific acoustic impedance 12 Sturgeon, J. Specimens and test Methods for carbon fibre
V Velocity of longitudinal waves in a reinforced plastics: Recomendations by government laboratories,
RAE TR 71026 (1971)
medium 13 Musgrave On the propagation of elastic waves in aeolotropic
V Voltage at the transducer mcdia,Proc Ro)] Sot A 226 (1954) 339-366
14 Wilkinson, S. J., Reynolds, W. N. The propagation of ultra-
A Attenuation sonic waves in carbon fibre reinforced plastics. J. Phljsics II
7 (1974) (1) 50-57
AT Total Attenuation 15 Reynolds, W. N., Wilkinson, S. J. The propagation of
Front surface losses ultrasonic waves in cfrp laminates, Ulrrasorlics 12 (1974) (3)
Af
109-I 14
Ab Back surface losses 16 Dean, G. D., Turner, P. The elastic properties of carbon fibres
and their composites, Composites 4 (1973) (4) 174-180
A,=Af +Ab Total surface losses 17 Markham, M. F. Measurement of the elastic constants of fibre
Q Attenuation coefficient composites by ultrasonics Composites I (1970) (3) 145-149
18 Krautkramer, J., Krautkramer, H. Ultrasonics testing of materials.
cij Elastic stiffness coefficient Springer-Verlag. Berlin/Heidelberg, (1969) 460
19 Lloyd, E. A. Wide Band Ultrasonic techniques, Paper 2,
V” Void content Symposium on the future of ultrasonic spectroscopy Ed.
k Constant in the parabolic relationship Reunolds, P. M., British Non-Ferrous Metals Research
Association, London (1971)
kl, k2 Constants in the bi-linear relationship 20 Rowland, E. Hawker Siddeley Aviation Ltd, Private
Communication

NON-DESTRUCTIVE TESTING. JUNE 1975 145

You might also like