See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233607588

Defect types and non-destructive testing techniques for

composites and bonded joints

Article  in  Materials Science and Technology · May 1989

DOI: 10.1179/026708389790222366

CITATIONS READS

68 2,785

2 authors:

Peter Cawley R. D. Adams

Imperial College London University of Bristol
383 PUBLICATIONS   15,961 CITATIONS    226 PUBLICATIONS   10,313 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NDT of Adhesive Joints View project

University of Bristol Innovative Approaches to Composite Structures, 1995-1998 View project

All content following this page was uploaded by R. D. Adams on 07 April 2015.

The user has requested enhancement of the downloaded file.

Journal oj Sound and Vibration (1988) 122(2),299-316

THE MECHANICS OF THE COIN-TAP METHOD

OF NON-DESTRUCTIVE TESTING

P. CAWLEY

Department oj Mechanical Engineering, Imperial College oj Science and Technology,

London SW7 2BX, England
AND

R. D. ADAMS

Department oj Mechanical Engineering, University oj Bristol, Bristol BS8 1 TR, England

(Received 5 February 1987, and in revised Jorm 23 June 1987)

The physical basis of the coin-tap method of non-destructive inspection has been
investigated. It has been shown that the characteristics of the force input to a structure
during a tap are changed by the presence of a defect such as a disbond or delamination
. beneath the surface of the structure, the impact duration increasing and the peak force
. decreasing as the defect becomes larger. The spectra of the force-time histories demonstrate
that the rate of reduction of force amplitude with frequency increases as the defect becomes
larger. It has been shown that defects such as disbonds and delaminations may be modelled
as springs, the stiffness of which is that of the layer(s) above the defect. The change in
the impact characteristics as this spring stiffness is varied has been investigated theoretically
on both a rigid structure and on thick and thin beams. On a rigid aluminium structure, a
spring stiffness of 100 MN/m would readily be detectable, whereas on the beams the
sensitivity of the test is reduced due to an increased variation in the impact characteristics
with position over good areas of the structure. Initial experimental results indicate that a
spring stiffness of around 30 MN/m is at the limit of detectability. In an aluminium
structure, this corresponds to a defect 1 mm deep and 6 mm in diameter. The method has
the advantage over higher frequency ultrasonic testing that no coupling fluid is required
between the transducer and the structure. Initial results also suggest that the method is
more reliable on thin structures than the mechanical impedance technique.

1. INTRODUCTION

The coin-tap test is one of the oldest methods of non-destructive testing and it is regularly
used for testing laminated structures, honeycomb constructions and bonded joints. The
test requires an operator to tap each point of the structure to be inspected with a coin,
and to listen to the resulting sound radiated by the structure. It is found that defective
regions sound "dead", and can therefore be identified. Until recently, the test has remained
largely subjective and there has been considerable uncertainty about the physical principles
behind it. However, it has been widely used, and Hagemaier and Fassbender [1] found
that it could detect more types of defect in honeycomb constructions than any other
technique except neutron radiography.
It should be stressed that this test is quite different from the "wheel-tap" test which
was traditionally carried out to detect cracked railway wheels and defective pottery and
glassware [2], though the testing technique and subjective interpretation of the sound
produced is similar in both cases. The wheel-tap test is a global test which investigates
the whole component from a tap applied at a single point, the difference between good
299
0022.460X/88/080299 + 18 $03.00/0 © 1988 Academic Press Limited
300 P. CAWLEY AND R. D. ADAMS

and defective components being detected from changes in the natural frequencies and
damping. The coin-tap test will find defects only in the region of the tap, so it is necessary
to tap each part of the structure under investigation.
The sound produced when a structure is tapped is mainly at the frequencies of the
major structural modes of vibration. These modes are structural properties which are
essentially independent of the position of excitation. Therefore, if the same pulse is
applied to a good area and to an adjacent defective area, the sound produced must be
very similar. The difference in the sounds produced when good and defective areas are
tapped must therefore be due to a change in the force input.
When a structure is struck with a hammer, the characteristics of the impact are dependent
on the local impedance of the structure and on the hammer used. Damage such as an
adhesive disbond results in a local decrease in the structural stiffness, and hence a change
in the nature of the impact. This is illustrated in Figure I(a) which shows measured
force-time histories from taps on good and disbonded regions of a honeycomb structure
with carbon fibre reinforced plastic (CFRP) skins. The impact over the good area is more
intense and of shorter duration than that on the damaged area.
The difference in the sound produced by the two taps may be explained by studying
the frequency content of the two force pulses. This is achieved by carrying out a Fourier
transform of the force-time records to produce the ~odulus spectra shown in Figure
I(b). The amplitude of the force input to the damaged area falls off rapidly with increasing
frequency, while the impact on the sound area has a much lower rate of decrease of force
with frequency. This means that the impact on the defective area will not excite the higher

50.--------------------,
(0)

40

e- 30

400 600 1600

Time (f1s)
50r----------------,
(b)
~ 40r-_-,~
>-
~ 30
:0
~ 20
ai
~
~ 10
&
o 400 600 1200 t600 2000
Frequency (Hz)

Figure 1. Force-time histories of impacts on good and disbonded areas of CFRP skinned honeycomb
structure; (b) spectra of time histories shown in Figure Hal.
 COIN-TAP NON-DESTRUCTIVE TESTING METHOD 301
structural modes as strongly as the impact on the good zone. Therefore the sound produced
does not contain the higher frequencies and the structure sounds duller.
These findings, which were originally reported in reference [2], show that it is possible
to produce a version of the coin-tap test which depends on the measurement of the force
input to the test structure during the tap, rather than on the measurement of the sound
produced by the tap. This has the advantage that the measurements are insensitive to
background noise and it is readily possible to produce an instrument which will apply a
controlled pulse to a structure via an instrumented hammer and compare either the time
history or frequency spectrum of the measured pulse with a standard obtained on a
structure which is known to be good.
In order to determine the utility of such an instrument, it is necessary to define fhe
sensitivity of the test. The method has some similarities with the mechanical impedance
method of inspection [2-6]. A study of that technique [3, 6] showed that it is most sensitive
to defects between a thin top layer and a stiff base structure. It was also found that the
reliability of the impedance method was severely reduced when it was used on relatively
flexible structures. It is therefore of interest to investigate whether the coin-tap test suffers
from similar limitations. This paper reports an investigation of the characteristics of the
coin-tap test.

2. MODEL OF TEST
The work on the mechanical impedance method reported in reference [3] showed that
a defect such as a disbond or delamination may be modelled as a spring whose stiffness
is that of the layer(s) above the defect. In the absence of a defect, the spring stiffness is
infinite and, as the defect becomes larger or moves closer to the surface, the spring stiffness
is reduced. The technique effectively measures the impedance at the top of the spring,
so the presence of a defect causes the impedance to be reduced. In this analysis it is
assumed that the stiffness of the contact between the transducer and the structure is
infinite. In practice, a dry contact is used between the impedance transducer and the
structure. This contact has a finite stiffness given by Hertzian contact theory [7] so the
effective stiffness over a defect is the contact stiffness in series with the defect stiffness,
while over a good area the effective stiffness is the contact stiffness alone, as shown in
Figure 2.
In the impedance method the point impedance of the structure is measured with a
constant static load between the transducer and the structure. However, in the coin-tap
Defect
-d
(e)

~TOP

(b)

Figure 2. (a) Typical defect; (b) model of defect for analysis of the tap test. kd is defect stiffness; k< is contact
stiffness between the tapping head and the structure.
302 P. CAWLEY AND R. D. ADAMS

test the striker effectively hits the top of the spring with a certain velocity and is retarded,
the contact force varying as shown in Figure l(a). According to Hertzian contact theory,
the contact stiffness is load dependent and so it would be expected that in order to model
the coin-tap test, the stiffness, kn shown in Figure 2 would have to be non-linear. However,
the contact area when a spherical striker hits a flat surface is extremely small and it is
likely that highly localized plastic deformation OCcurs. Hertzian contact theory, in which
perfectly elastic behaviour is assumed, would then not apply. It was therefore decided
to investigate the contact behaviour in order to determine whether a linear analysis would
be satisfactory.

3. DETERMINATION OF CONTACT STIFFNESS

When a mass, m, travelling at velocity, v, strikes a grounded spring of stiffness, k, it
may readily be shown that the impact duration, 7, is given by
7= 1T/W, (1)
where w = (k/ m )1/2. The compressive force in the spring, F, is given by
F= (kv/ w) sin wt, (2)

t
The mass will rebound at speed v, and so, from Newton's second law,
T

Fdt=2mv. (3)

It is also readily shown that the spectrum of the force pulse obtained by taking the Fourier
transform of equation (2) has a minimum at a frequency given by 1·5/7 Hz.
The impact of a spherical-tipped striker on a rigid body was investigated by using the
system shown in Figure 3. The striker comprised the core of a solenoid with a force gauge
(Bruel and Kjaer (B&K) type 8200) attached to its lower end, a·n aluminium alloy tip
with radius of curvature approximately 15 mm being joined to the force gauge via a stud
contact. The motion of the striker was monitored by an accelerometer (B&K type 8303)
attached to the top of the solenoid core. The total mass of the striker was 65 gm. When
the solenoid was not energized, the striker was retained at a fixed position in the solenoid
coil by the action of a spring. The solenoid was mounted in a cylindrical holder which

Accelerometer

B 8 K 2032
Chor<;je
Coil Solenoid omplifiers. Spectrum
Core Anolyzer
Volto<;je
pulser

Profiled alumjnium--4I~
head .......- " " ' - - - " - - - - ,

HP 9816
Computer

Figure 3. Schematic diagram of apparatus.

 COIN-TAP NON·DESTRUCTIVE TESTING METHOD 303
was rested on the structure to be tested, and so maintained a fixed distance between the
structure surface and the striker in its rest position.
When the solenoid was energized by a voltage pulse, the striker was accelerated towards
the surface. The length of the pulse and the distance between the structure surface and
the striker in its rest position were adjusted so that the striker was in free flight (apart
- from the action of the retaining spring) at the moment of impact. The impact velocity
could be varied by changing the duration of the voltage pulse applied to the solenoid.
The retaining spring ensured that the striker did not return to the surface of the structure
and so produce a double impact.
The signals from the force gauge and accelerometer were digitized and stored by a
B&K type 2032 spectrum analyzer which can store 2048 points on each of two channels,
the sample rate used giving an interval between successive points of 15·3 IJ.s. The analyzer
was triggered when the force signal exceeded a pre-set level, the pre-trigger facility being
used to ensure that the acceleration of the striker from rest up to the impact was recorded.
The digitized signals from the analyzer were passed to an HP 9816 computer for further
processing.
The acceleration signal was numerically integrated to give the velocity of the striker at
impact, a typical velocity being 0·25 m/s. (This is the velocity which would be produced
by free-fall over a distance of 3 mm.) The force-time history and corresponding modulus
spectrum obtained from a test on one of the flat faces of a solid aluminium alloy cylinder
215 mm in diameter, 145 mm long and weighing 14·2 kg are shown in Figures 4(a) and
4(b) respectively. (The points captured by the analyzer are joined by straight lines on
Figure 4(a).) The duration of the force pulse is lower than that shown on Figure l(a) for
an impact on a honeycomb structure with CFRP skins because the contact stiffness

240r------------..,
(e)

180

~
~ 120
&

300

60r------------------,
,.;
~
'" 40
E
:0
~
ci
:; 20
u
&

o 2 4 6 8
Frequency (k Hz)

Figure 4. (a) Force-time history of impact on large aluminium block; (b) spectrum of pulse shown in
Figure 4(a).
304 P. CAWLEY AND R. Do ADAMS

between the striker and the aluminium surface is higher than that obtained on the
thin-skinned honeycomb component.
Numerical integration of the measured force with respect to time gave a value of
0·030 N s, compared with a value of 0·034 N s for 2mv (2 X striker mass x striker velocity
before impact). This result is in good agreement with equation (3) which governs the
impact of a striker on a massive body. The probable true beginning and end portions of
the force pulse are shown dotted on Figure 4(a), the alternative indicated by the measured
data being due to the finite sample rate employed. The estimated duration of the pulse
shown in Figure 4(a) is 0·24 m s so from equation (I), the frequency, w, is 13 X 103 rad/s.
Hence, the effective contact stiffness, k( = w 2 m), is II MN/m.
The peak force calculated from equation (2) by using the values of k and W Mrived
above, together with the measured value of the impact velocity, v, is 220 N. This compares
with the measured peak force of 198 N shown in Figure 4(a). Hence, the adoption of a
linear model of the contact stiffness gives good agreement between the predicted and
measured results.
The duration of the impact, T, shown in Figure 4(a) is 0·24 m s. Hence, if the system
behaviour follows that of a mass striking a linear spring, the spectrum of the force should
have a minimum at a frequency of I'S/T=62S0 Hz. Inspection of Figure 4(b) reveals a
minimum at 6400 Hz, giving further confirmation that a linear representation of the contact
behaviour is satisfactory.
Hence, subsequent predictions in this paper have been made on the assumption of a
linear contact stiffness between the aluminium striker and a flat aluminium surface of
11 MN/m, this value being appropriate at striker velocities of about 0·25 m/s. Some tests
were also carried out at an impact velocity of 0·4 mis, a contact stiffness of 13 MN/m
being calculated in this case. Hertzian contact theory predicts that contact stiffness
increases with contact force so, since higher impact velocities produce larger contact
forces, some increase in the effective contact stiffness as the impact velocity increases is
to be expected. However, the errors incurred by assuming a linear contact stiffness are
not significant for the purposes of the predictions made in this paper.

4. THEORETICAL INVESTIGATION OF THE INFLUENCE

OF DEFECTS ON IMPACT BEHAVIOUR
4.1. METHOD OF ANALYSIS
In order to determine the potential of the coin-tap test in non-destructive testing, it is
necessary not only to investigate how the impact force changes in the presence of defects,
but also to determine the extent to which it varies with position in a sound structure. If
there is little change in the characteristics of the impact with position in a good structure,
it wiII be possible reliably to detect defects which produce a small change in the measured
impact. Conversely, if there is significant variation with position in a good structure, the
reliable identification of small defects wiII be more difficult. This variability over a good
structure is responsible for the reduction in reliability of the impedance method on thin
structures [3, 5].
Since the aim of the analysis was to obtain an indication of the sensitivity of the test
and to determine whether the reliability of the test is likely to be affected by structural
stiffness, it was decided to analyze a simple system comprising a rigid mass striking a
beam. Goldsmith [8] has shown that the difference, a, between the displacement of the
mass and the deflection of the beam at the contact point at time t is

1
a=vt--
moo
I' I'
dt 1 L [XI (cW
Fdt--
pA ; wjB
I'0
F(T) sinw;(t-T) dT, (4)
 COIN-TAP NON-DESTRUCTIVE TESTING METHOD 305
where m is the mass of the striker, F is the contact force between the striker and the
beam, v is the initial velocity of the striker, p, A are the beam density and cross-sectional
area respectively, Wi is the natural frequency of the ith mode of the beam, B = J~ dx, X;
L is the length of the beam, Xi is the mode shape of the ith mode of the beam, x is the
position along the beam, and c is the value of x at the position of impact.
Goldsmith [8] termed a the "approach" of the striker and the beam. For a linear
contact stiffness, k, the approach is given by
a = FI k. (5)
The time history of the approach of the striker and the beam, and hence of the contact
force, F, was obtained by numerical integration of equation (4) with a time step much
smaller than the duration of the impact. (For a typical impact duration of 200 IJ-s, a tim'e
step of 7'5 IJ-S was used.) The number of modes included in the summation of equation
(4) was chosen so that no significant difference in the results was obtained with the
addition of extra modes. With the thick beam described below, 15 modes were used,
while with the thin beam, 33 modes were employed.
4.2. RESULTS ON RIGID STRUCTURE
When a rigid structure is tested, the impact does not excite any significant vibration
of the structure and the test may be modelled as a mass striking a grounded spring. When
no damage is present, the spring stiffness is the contact stiffness between the striker and
the structure, the stiffness being decreased by the presence of defects as described above.
Figure 5(a) shows the predicted time histories of the contact force between a striker of
mass 65 gm travelling at 0·25 mls and a rigid aluminium block, as the defect stiffness,
k", in Figure 2 is reduced from infinity (which corresponds to a sound structure, giving

240r----------------,
(0)

t80
kd=tO MN/m

Figure 5. Predicted force-time histories of impacts on rigid aluminium structure with different defect
stiffnesses; (b) corresponding spectra.
306 P. CAWLEY AND R. D. ADAMS

an overall stiffness at the defect site equal to the contact stiffness) to 100,50 and 10 MN/m.
The contact stiffness, kn in Figure 2 was assumed to be 11 MN/m which was the value
obtained in the tests described above.
It is clear from Figure 5(a) that a defect stiffness of 100 MN/m produces a significant
change in both the peak force and impact duration, the change increasing as the defect
stiffness is reduced. This progression is also seen in the modulus spectra of the force-time
histories which are shown in Figure 5(b). Hence, if the structure is rigid, a defect stiffness
of 100 MN/m should be readily detectable.

4.3. RESULTS ON THICK DEAM

The analysis of section 4.1 was then used to study the impact of a striker on a free-free
beam 300 mm long, 50 mm wide and 12·6 mm deep. The mass of the striker was 65 gm
and its initial velocity was 0·25 m/s. Figure 6(a) shows the time history of the impact
force when the impact position is 10, 50, 90 and 150 mm along the beam. In each case,
the peak force and impact duration are lower than those predicted for an impact on a
rigid block. This is because the beam moves away from the striker after the impact,
whereas the rigid block remains stationary. There is also a significant variation in both
the impact duration and peak force with position along the beam, the largest changes
occurring when the impact is close to the end of the beam.
Figure 6(b) shows the predicted effect of defects of stiffness 100,50 and 10 MN/m at
the middle of the beam, together with the envelope of the predictions for good areas at
different points along the beam. Both the peak force and impact duration for a test on
a defect of stiffness 100 MN/m at the middle of the beam fall within the range of values
predicted for impacts at different positions along a sound beam. The peak force is,
however, lower than that predicted for an impact at the middle of the beam with no
defect present, and the impact duration is longer. When the defect stiffness is reduced to
50 MN/m, the impact duration is longer than that predicted for any position in the
absence of a defect, but the peak force remains higher than that for an impact 10 mm
from the end of the beam with no defect present. When the defect stiffness is reduced to
10 MN/m, the peak force is lower than for an impact at any position on an undamaged
beam, and the impact duration is considerably longer.
Figure 6( c) shows predictions corresponding to those of Figure 6(b), but with the defect
10 mm from the end of the beam. The peak forces predicted for impacts above defects
of stiffness 100 and 50 MN/m 10 mm from the end of the beam are lower than those
obtained for impacts at any position in the absence of a defect, but the impact durations
are similar to those obtained for impacts towards the middle of a sound beam. Again,
when the defect stiffness is reduced to 10 MN/m, the peak force is lower than for an
impact at any position on an undamaged beam, and the impact duration is considerably
longer.
Figure 6(d) shows the modulus spectra of the force-time histories shown in Figure
6(a). The differences in the values of the spectra at zero frequency correspond to the
different mean force levels shown in the time histories (the differences appear smaller on
the spectra because they are plotted on a log scale). Figure 6(e) shows the modulus
spectra of the force-time histories shown in Figure 6(b). The spectrum corresponding to
an impact above a defect at the middle of the beam with a stiffness of 100 MN/m falls
within the range of those for impacts at different positions on a good beam. When the
defect stiffness is reduced to 50 MN/m, the rate of decrease in the level of the spectrum
with frequency is higher than that predicted for impacts on a sound structure at most
positions along the beam. However, the shape of the spectrum predicted for an impact
50 mm from the end of a sound beam is very similar to that for an impact above a defect
 COIN-TAP'NON-DESTRUCTIVE TESTING METIlOD 307

(0) 30r- _

20

10

180 (b)
0

~ 120 k4 =loo MN/m

50MN/m 30
"
~
& Vl0MN/m
~

"" ~

"", ~ 20
~
0
a:i
~
., 10
(c) ~
&

O'--_ _---''-_ _---L ---l. ...JU

(f)
30 -

400
Time(fLs) 20

10

o 2 4 6
frequency (kHz)

Figure 6. (a) Predicted force-time histories of impacts at points x mm along a 300 mm long, 12·6 mm thick
aluminium beam; (b) predicted force-time histories of impacts above defects of dilIerent stiffnesses at the
middle (x = 150) of the beam of Figure 6(a); - - , envelope of predictions for good areas; - - -, defecti\'e area
(x;: 150); (c) force-time histories corresponding to those of Figure 6(b) but with the defect 10 mm from the
end of the beam (x;: 10); - - , envelope of predictions for good areas; - - -, defective area (x;: 10); (d) spectra
of time histories shown in Figure 6(a); (c) spectra of time histories shown in Figure 6(b); - - , envelope of
predictions for good areas; - - -, defective area (x;: 150); (f) spectra of time histories shown in Figure 6(c);
- - , em'elope of predictions for good areas; - - -, defective area (x;: 10).

of stiffness 50 MN/m at the middle of the beam. When the defect stiffness is reduced to
10 MN/m, there are clear differences between the spectra of an impact above the defect
and those of impacts at any position on a sound beam. Figure 6(f) shows the modulus
spectra corresponding to the force-time histories of Figure 6(c). The shapes of the spectra
predicted for defects of stillness 100 and 50 MN/m 10 mm from the end of the beam are
very similar to those for impacts at the middle of a soundbeam._ Again, when the defect
stiffness is reduced to 10 MN/m, there are clear differences between the spectra of an
impact above -the defect and those of impacts at any position on a sound beam.
 308 P. CAWLEY AND R. D. ADAMS

100,--------------------,
(a)

30,--------------------,

o 2 4 6 8
Frequency (kHz)

Figure 7. (a) Predicted force-time histories of impacts at different points on 300 mm long, 3 mm thick
aluminium beam with different defect stiffness; (b) corresponding spectra to Figure 7(a). - , Good area;
- - -, defective area.

4.4. RESULTS ON THIN BEAM

Figures 7(a) and 7(b) show impact force-time histories and modulus spectra correspond-
ing to those of Figure 6, but with the beam thickness reduced to 3 mm. (In this instance,
in order to improve the clarity of the diagram, the case of a defect stiffness of 100 MN/ m
is not shown.) All the time histories of Figure 7(a) show peak forces considerably lower
than those predicted for the thicker beam. This is because the beam mass is now only
122 gm compared with a striker mass of 65 gm. The beam therefore accelerates more
rapidly away from the striker and so reduces both the impact duration and the peak
force. It is also noticeable that the force-time histories tend to become asymmetrical with
respect to the peak, the duration after the peak being significantly longer than the rise
time to the peak. The phenomenon is particularly marked in the case of an impact 90 mm
from the end of the beam. This "tail" on the impact force is due to the vibration of the
beam which is more severe with the thinner beam, both because the vibration amplitude
in a particular mode produced by a given force is increased as the stiffness of the beam
is reduced, and also because more modes come into the frequency range which is excited
. by the impact.
With the thinner beam, the impact duration for a test above a defect of stiffness
50 MN/m at the middle of the beam isjust within the range of those predicted for impacts
at different positions on a sound beam, while the peak force is slightly lower than any
of those predicted for impacts on a sound beam. Similarly, a defect of stiffness 50 MN/m
 COIN-TAP NON.DESTRUCTIVE TESTING METHOD 309
10 mm from the end of the beam gives an impact with lower peak force than is predicted
for impacts at any point on a sound beam. The impact duration is, however, within the
range of those predicted for a sound beam..
The spectra of Figure 7(b) show that the impact above a defect of stiffness 50 MN/m
at the middle of the beam produces a pulse with a more rapid rate of decrease of force
intensity with frequency than those at most positions on a sound beam. However, .the
shape of the spectrum predicted for an impact 50 mm from the end of a sound beam is
very similar to that for an impact above a defect of stiffness 50 MN/m at the middle of
the beam. The short duration of the impacts 10 mm from the end of the beam, both
without damage and with a defect of stiffness 50 MN/m, means that the corresponding
spectra show lower rates of decrease of force amplitude with frequency than the speccra
from impacts towards the middle of a sound beam.

5. DEVELOPMENT OF A PASS-FAIL CRITERION

The results presented above show that it is possible to detect defects by comparing
either the time history of the impact force or the corresponding spectrum with signals
from a structure which is known to be sound. If the test is to be carried out automatically,
it is necessary to decide on an objective criterion which can be used to distinguish defective
areas from regions of sound structure.
The results of the previous section show that damage produces a decrease in the peak
force, an increase in impact duration and a reduction in the high frequency content of
the force pulse compared with the results of a test on a good structure at the same position.
Therefore, if results from each position of interest on a good structure are stored, and
test results are compared with those from the corresponding position on the good structure,
changes in any of these quantities can be related to damage. However, if the structure
to be tested is large, this procedure would require a large amount of data storage and
the test procedure would become complicated.
The test would be more practical if a single indicator of a good structure could be
defined, deviations from this indicating the presence of damage. This can readily be done
if the structure is rigid since there is then no change in the impact characteristics with
position on a good structure. However, as the structure becomes more flexible, there are
variations across a good structure as shown in Figures 6 and 7. Therefore if, for example,
it was decided to classify all regions giving an impact duration of more than a certain
value as defective, then either some good areas would be classified as defective, or defects
with a high stiffness would not be found. It is therefore necessary to find a criterion which
will detect defects with a high stiffness without classifying any good areas as defective.
There are many possible ways of comparing the time histories and/or spectra of two
impacts. Three obvious candidates are comparisons of peak force, impact duration and
rate of decrease of force with frequency in the middle of the frequency range shown in
Figures 5(b), 6(d), (e), (f) and 7(b). However, all these possibilities have serious
drawbacks.
If peak force were to be used, it would be essential to ensure very accurate control of
the velocity of the hammer immediately prior to impact, otherwise a standard velocity
impact on a good region could be confused with a higher velocity impact over a defect.
In practice, it is likely to be difficult to provide sufficiently accurate control, particularly
if the hammer is to operate in different orientations.
Impact duration becomes particularly difficult to measure if the force pulse has a "tail"
such as those shown on Figure 7(a). In these instances, the amplitude of the pulse could
310 P. CAWLEY AND R. D. ADAMS

be very close to the system noise level for a significant proportion of the overall pulse
length, which would lead to major measurement uncertainties.
Similarly, in the example of Figure 5(b), ifit was decided to monitor the rate of decrease
of force with frequency at, for example, a frequency of 5 kHz, the defects with stiffnesses
of 100 and 50 MN/m would show an increased rate of decrease of force with frequency
compared with the sound structure, as would be expected. However, the defect with a
stiffness of 10 MN/m would show an increase of force amplitude with frequency and, if
the inspection frequency was slightly higher, the slope of the spectrum could be the same
as on a good structure.
All the possibilities discussed above seek to detect the presence of a defect by using
only a very: small proportion of the data which is available if the time history of the
impact force is measured. Not only does this discard potentially valuable information;
but it makes the result very sensitive to random noise on critical data points. Several
possible methods for overcoming these difficulties have been investigated, the most
satisfactory one to be devised being illustrated in Figure 8. This involves obtaining the
modulus spectrum of the force pulse which, with the advent of dedicated Fourier transform
micro-chips, is a simple and very rapid procedure. The areas A and B under the logarithmic
graph of the spectrum shown in Figure 8(a) are then calculated. The ratio, R, defined by
R=B/(A+B) (6)
is then computed. Impacts over defects have lower high frequency force content than
impacts over good areas, and so the value of R is reduced.
The frequency, Ji" at which the boundary between regions A and B is set is arbitrary,
but 30 to 50% of the full scale frequency,/m:i.. when the first minimum in the spectrum
of an impact over a good region occurs close to the full scale frequency has been shown
to be a satisfactory choiCe. The effect on the calculation of. changes in the impact force
is accounted for by normalizing the results so that the zero frequency'level on the two

o fb
Frequency lIinear scale)

IT---- (b)

iII-
~
--i

Frequency (linear scale)

Figure 8. (a) Areas A and B used in comparison of test spectra with standard; (b) normalization procedure
us'ed in comparison; - . standard spectrum, S; - - -, measured spectrum, T; - - - , normalized measured
spectrum, r.
 COIN·TAP NON·DESTRUCTIVE TESTING METHOD 311
spectra to be compared is the same. Thus, if curve S in Figure 8(b) is a standard curve
from a good area, and curve T is the result of a test, curve T is shifted to T' before the
calculation is performed. The minimum force level to be included in the calculation is
30 dB below the value of the spectrum at zero frequency. In practice, this range is sufficient
to include most of the measured data, and avoids any problems with random noise on
the force signal.
Table 1 shows the values of R computed from the theoretical spectra of impacts on a
rigid structure and on 12·6 and 3 mm thick beams, which were discussed in section 4.
The ranges of R on good areas of the beams which are quoted in Table 1 show the
maximum and minimum values which were calculated for impacts at positions between
10 and 150 mm from the end of the beam (total beam length was 300 mm). No range ,is
shown for the rigid structure since in this instance, in the absence of defects, there is no
variation with position.

TABLE 1
Variation of ratio, R, with defect stiffness and position in rigid structure and 12·6 and 3 mm
thick beams (fm .. = 8 kHz)

R(f~ = 0'3Im3x)

12·6mm beam 3 mm beam

Defect Distance of Rigid structure (range of R in (range of R in
stiffness, defect from end (good area, good areas, good areas,
kJ (MN/m) of beam (mm) R =0'57) 0·58-0'63) 0·65-0'71)

100 150 0·55 0·59 0·69

50 150 0·52 0·57 ,0·68
10 150 0·40 0·46 0·62
50 10 0·60 0·71
10 10 0·51 0·67

The results of Table 1 indicate that on a rigid structure a defect of stiffness 100 MN/m
would be detected reliably, while in the middle of the thick beam a defect of stillness
50 MN/m is at the margin of detectability. If a defect of stiffness 50 MN/m is close to
the end of the thick beam the tap test will not detect it reliably since the value of R is
within the range of those obtained on good areas of the beam. On the thin beam the
sensitivity is reduced further, the margin of detectability being a defect stiffness of between
50 and 10 MN/m when the defect is away from the end of the beam, and if the defect
is close to the end a defect of stillness 10 MN/m is just outside the range for reliable
detection.
It should be emphasized that the defect stiffnesses quoted above are the highest defect
stillnesses which can be detected using the result from a single good area anywhere on
the beam as a standard for comparison, and subject to the condition that no good areas
are to be identified as defective. If different standards were used for test points in the
middle and at the ends of the beam, the test sensitivity, particularly on the thin beam,
would be greatly improved.
For a circular defect of diameter, d, at depth, h, if the edges of the defect are assumed
to be clamped, then the defect stiffness, k d , is given by

(7)
312 P. CAWLEY AND R. D. ADAMS

where D = Eh /{12(l- v )}, in which E, v are Young's modulus and the Poisson ratio,
3 2

respectively. Hence, a stiffness of 100 MN/m corresponds to a defect 3·6 mm in diameter

at a depth of 1 mm, or 19 mm in diameter at a depth 00 mm, while a stiffness of 50 MN/m
corresponds to a defect 5 mm in diameter at a depth of 1 mm, or 26 mm in diameter at
a depth of 3 mm.
The predictions quoted here have all been on free-free beams. The case of simply
supported ends has also been investigated; in this instance the ends of the beam are more
rigid than the middle and so the peak force during an impact at the end of the beam
tends to be higher than that during a test in the middle of the beam, the difference being
particularly noticeable when the beam is thin. However, the predicted test sensitivity was
very similar to that obtained on the free-free beam.

6. EXPERIMENTAL INVESTIGATION
The initial experimental investigation of the method was carried out on beam specimens
similar to those considered in the theoretical analysis. Two aluminium beams 300 mm
long and 50 mm wide were used, one being 12·6 mm thick and the other 3 mm thick.
Tests were carried out at 20 mm intervals between the midpoints of the beams and points
10 mm from the ends, with use of a system similar to that shown in Figure 3, the test
block used in the experiments described in section 3 being replaced by one of the beams
which was supported on a wooden laboratory bench.
The effect of disbonds or delaminations was then simulated by milling slots in the
beams, each slot being across the full width of the beam and of depth such that the
bottom of the slot was 1·5 mm from the surface of the beam on which the tests were
carried out. Slots 30, 20 and 10 mm long were incorporated on each beam, the minimum
distance between the slots being 45 mm; none of the slots was less than 55 mm from the
ends of the beam. The type of slot used is shown in Figure 9(a).
. Figure 9(a) shows force-time histories measured in a typical good area, and at the
centre of each slot on the 12·6 mm thick beam, the data points being joined by straight
lines. The durations of the impacts on both the 20 and 30 mm long defects are significantly
longer than that on the good area, and the peak forces are lower. However, on the 10 mm
long slot, the duration is very similar to that on the good area, while the peak force is
considerably reduced. The test was repeated several times so the result was not due to
variations in the velocity of the striker. Further tests showed that the variation in peak
force over good areas of the beam after the slots were milled was much greater than that
measured in the earlier tests on the virgin beam, though there was little variation in impact
duration, or in the shape of the spectrum and hence the value of the ratio, R, discussed
in section 5. It seems likely that these results, which were also seen on the 3 mm thick
beam, are analogous to the increased variability of the impact characteristics with position
shown in the theoretical results on the thin beam reported in section 4. The milling of
the slots greatly reduced the flexural stiffness of the beam, and so some increase in
variability would be expected.
The modulus spectra corresponding to the time histories of Figure 9(a) which are
plotted in Figure 9(b) show that the force amplitudes of the impacts above the 20 and
30 mm long defects decrease more rapidly with frequency than that of the impact on the
good area while the rate of decrease shown by the impact on the 10 mm defect is very
similar to that over the good area. Table 2 shows the variations in the ratio R defined
by equation (6) along each of the beams before the defects were introduced, together
with the values obtained from the tests above each of the defects. The theoretical static
 COIN.TAP NON-DESTRUCTIVE TESTING METHOD 313

60<0,':",::8:
Slots milled acral' (0 )

.. :J ~:
180

~ _ 12'6mm
~ 120 10mm
defect L-10,20,30 mm
If
20 mm deflct

400 500

60

..:
e
~ 40
E
:s
1;
rD
~ 20
~
If

0 2 'I 6 8
Frequency ( kHz)

Figure 9. (a) Measured force-time histories of impacts on 12·6 mm thick beam; (b) corresponding spectra.

stiffness at the centre of the plate above each slot, with the edges of the slot parallel to
the beam width assumed to be clamped, is shown as k d in the table.
It appears from these results that the 10 mm slot, which corresponds to a stiffness of
30 MN/m, is close to the margin of reliable detectability on both the 12·6 and 3 mm thick
beams. The theoretical analysis suggested that a defect of stiffness 50 MN/m would be
at the margin of detectability on the thicker beam so, given that it is very difficult to
define the margin precisely, the experimental and theoretical results are in good agreement.

TABLE 2
Measured values of ratio, R, over simulated defects (milled slots) in 12·6 and 3 mm thick
beams, together with range of R values measured on sound beams; kd is the theoretical
stiffness at the centre of the plate above the milled slot, with the edges of the slot parallel to
the beam width assumed to be clamped (fmax = 8 kHz)

12·6mm beam 3mm beam

(range of R (range of R
Defect length Defect stiffness over good structure, over good structure,
(mm) kd (MN/m) 0'59-0'63) 0'65-0·71)

30 3·3 0'35 0·51

20 7'5 0·46 0·57
10 30·0 0·57 0·62
314 P. CAWLEY AND R. D. ADAMS

180
Slot milled across
fUllbe~TOP
12·6 mm •
ct. 1
(0)

omm

z 30 mm
-; 120 Edge
~
&

60
(b)
9 mm from centre
~
;:- 40
~
:e
0
ai
~ 20
'"~
&

0 2 4 6
Frequency (kHz)

Figure 10. (a) Measured force-time histories at different positions within 30 mm long defect in 12·6 mm thick
beam; (b) corresponding spectra.

The theoretical results suggested that the test would be less sensitive on the thin beam
because of an increase in the variation of the results at different positions along a good
beam. However, the results of Table 2 show that the increased variability along the thinner
beam was relatively small. This is probably because in the theory the beam was assumed
to be free, whereas in the experimental investigation it was resting on a wooden bench
which would tend to restrict the vibration of the beam, and so reduce the variation seen
with position. The experimental configuration is probably closer to that which would be
used in NOT since it is unlikely that a structure as flexible as the 3 mm thick beam would
be unsupported.
The stiffness of the layer above a defect varies with position over the defect and so a
change in the test results with position would be expected. Figure 10 shows the force-time
histories and the corresponding modulus spectra obtained in· tests at different distances
from the centre of a slot 30 mm long on the 12·6 mm thick beam. (The result at the centre
of the slot is different from that shown in Figure 9 because the thickness of the layer
above the slot was 1 mm compared with 1· 5 mm in the earlier tests, the depth of the slot
having been increased for another investigation.) As expected, the duration of the impact
increases, and the peak force decreases, as the test position moves towards the centre of
the defect; corresponding changes being observed in the spectra.

7. DISCUSSION AND CONCLUSIONS

The physical basis of the coin-tap test has been demonstrated, the change in the sound
produced when a defective area is tapped being due to a change in the force transmitted
 COIN-TAP NON·DESTRUCTIVE TESTING METHOD 315
to the structure by the tap. It has been shown that this process may be modelled by
representing the defect as a spring whose stiffness is that of the layer(s) above the defect.
The change in the characteristics of an impact over a defect compared with one over a
good area of structure increases as the (linear) spring stiffness is reduced, so the method
is most sensitive to defects close to the surface of the structure. The test will detect only
defects such as disbonds and delaminations which reduce the effective stiffness of the
structure in a direction normal to the structure surface. The method is unlikely to detect
transverse cracks (i.e., cracks running normal to the surface).
The sensitivity of the test is determined by the contact stiffness between the striker and
the structure. The test will therefore be more sensitive on an aluminium structure than
on a component fabricated from a composite material such as carbon fibre reinforced
plastic since the contact stiffness is higher on the aluminium. The contact stiffness is also
the determining factor of the sensitivity of the mechanical impedance method of non-
destructive inspection [3], and initial tests indicate that the sensitivity of the two methods
on stiff structures is very similar. However, tests with the impedance method on thin
structures [3,5,6] indicate that the reliability of the impedance test is severely reduced
on more flexible structures, even when t"hey are resting on a solid base such as a laboratory
bench. However, the tests reported here indicate that the fall in the sensitivity of the
coin-tap test is much less severe. Further tests are being carried out to quantify this effect,
and the results will be presented in a future paper.
The energy input to the structure during the tap is very small. For example, in the tests
reported here, the velocity of the striker immediately prior to impact was 0·25 m/s. (This
velocity would be attained in free fall from a height of 3 mm.) The mass of the striker
was 0·065 kg so the kinetic energy of the striker was approximately 2 mJ. This is much
lower than the energy required to propagate impact damage in a fibre reinforced plastic.
The chief advantage of the method over higher frequency ultrasonic testing is that no
couplant is needed between the structure and the tapping head. The technique is therefore
particularly attractive for use in the field, and on honeycomb structures with thin skins
which may be porous. The test simply requires the time history of the force input to the
test structure to be captured, and this can be done via transducer in the tapping head so
no transducer need be attached to the structure. Also, in contrast to the subjective version
of the test, it is not necessary to monitor the sound radiated by the structure so the test
is not sensitive to background noise. The proposed method for distinguishing defective
areas of structure from sound regions requires the Fourier transform of the input force
to be computed and compared with a standard from a good area of structure; with modern
electronics this can readily be done in around 100 ms, so if the hammer position is
controlled by an automatic scanning mechanism, a testing rate of 5-10 positions per
second can be achieved.

REFERENCES
1. D. HAGEMAIER and R. FASSBENDER 1978 SAMPE Quarterly 36-58. Non·destructive testing
of adhesively bonded structure.
2. R. D. ADAMS and P. CAWLEY 1985 in Research Techniques in Non-destructive Testing (R. S.
Sharpe, editor) 303-360. Vibration techniques in non-destructive testing.
3. P. CAWLEY 1984 NDT International 17, 59-65. The impedance method of non-destructive
inspection.
4. P. CAWLEY 1985 Journal ofComposite Structures 3, 215-228. The operation of NDT instruments
based on the impedance method. .
5. P. CAWLEY and D. NGUYEN 1988 Journal of Mechanical Systems and Signal Processing in press.
The use of the impedance method of non-destructive testing on honeycomb structures.
 316 P. CAWLEY AND R. D. ADAMS
6. P. CAWLEY 1987 NDT International. 20, 209-215. The sensitivity of the impedance method of
non-destructive testing.
7. S. P. TIMOSHENKO and J. N. GOODIER 1951 71leory of Elasticity. New York: McGraw-Hili,
second edition.
8. W. GOLDSMITH 1960 Impact. London: Edward Arnold.

View publication stats