Disbond Detection in Bonded Aluminum Joints Using Lamb Wave Amplitude and Time-of-Flight

7 Pages·1992·0.14 MB·English

by  Sun, Keun J.

DISBOND DETECTION IN BONDED ALUMINUM JOINTS USING LAMB WAVE
AMPLITUDE AND TIME-OF-FLIGHT
Keun J. Sun
Department of Physics
College of William and Mary
Williamsburg, VA 23187
Patrick H. Johnston
NASA-Langley Research Center
Hampton, VA 23681
INTRODUCTION
In recent years, there was a need of developing efficientn(cid:0) ondestructive integrity
assessment techniques for large area laminate structures, such asd(cid:0) etections of disbond,
crack, and corrosion in fuselage of an aircraft. Together witht(cid:0) he improving tomography
and computer technologies, progress has been made in many fieldsi(cid:0) n NDE towards a faster
inspection.
Ultrasonically, Lamb wave is considered to be a candidate forl(cid:0) arge area inspections
based on its capability of propagating a relatively long distancei(cid:0) n thin plates and its media-
thickness-dependent propagation properties [1-2]. Moreover, theo(cid:0) ccurrence of disbonds,
corrosion, and even cracks often results in reduction of effectivet(cid:0) hickness of a laminate.
The idea is to assess the condition of a structure by sensing ther(cid:0) esponse of propagating
Lamb waves to these flaws over long path length [3-4]. A serieso(cid:0) f tests in the sequence of
disbond, corrosion, and crack have been done on various types ofs(cid:0) pecimen to investigate
the feasibility of this approach. This paper will present some oft(cid:0) he test results for disbond
detection on aluminum lap splice joints.
MEASUREMENTS AND TEST RESULTS
Laboratory specimens were made of aluminum sheets of 1 mm int(cid:0) hickness. Lap
splice joint and doubler are the two geometries of structuresp(cid:0) rimary interest [4]. The width
of adhesive-bonded area in a lap splice joint or a doubler wast(cid:0) ypically 5 cm, and the
thickness of adhesive layer(mostly, epoxy) was approximately 200m(cid:0) icrometers or less.
Lap joints both with and without rivet holes were fabricated ino(cid:0) rder to see the effects of
rivet rows on wave propagation. For testing, various sizes,s(cid:0) hapes, and locations of
disbonds in the interface of aluminum sheets were built in byl(cid:0) eaving the designated areas
free of epoxy when the sheets were adhered.
To propagate Lamb waves, a pair of piezoelectric transducers wasp(cid:0) laced on top of
the aluminum specimen and was separated at a distance, whichc(cid:0) overs the whole bonded
region of a lap joint or a doubler. Water was the couplantb(cid:0) etween transducer and aluminum
plate. Pulsed, pitch-catch method was utilized for amplitude andt(cid:0) ime-of-flight
measurements. Low-order Lamb modes, excited at a frequency in ther(cid:0) ange from 1 to 2
MHz, propagated across the bonded area with directionp(cid:0) erpendicular to the length of the
bond. During the testing, an automated scanner carried thet(cid:0) ransducer-pair moving in
parallel to the length of the bond. At each location of thet(cid:0) ransducer-pair, amplitudes of the
two predominant signals, the lowest-order symmetric Lamb wave (S0m(cid:0) ode), which was
the first arrival, and antisymmetric mode (A0 mode), werem(cid:0) onitored by peak detectors.
Time-of-flight (T0) of waves was obtained through a pulsed-phase-locked-loopc(cid:0) ircuitry in
terms of frequency [5]. The percentage of change in frequencyi(cid:0) ndicates the percentage of
change in T0. At the end of test, amplitude and time-of-flight as a functiono(cid:0) f transducers’
position were plotted and used to locate disbonds. Scanning ratec(cid:0) ould be adjusted
depending on the smoothness of the surface. The ultimate limit oft(cid:0) ime interval between
acquisitions of two data points is approximately 60 microseconds,w(cid:0) hich is based on a 20
cm separation distance between transducers, and the velocity oft(cid:0) he slower A0 mode is
approximately 3.0 mm/m s in the working frequency range. Generally, a round tripo(cid:0) f
scanning was enough to average out fluctuations in magnitude ofa(cid:0) mplitude resulted from
the movement of transducers. In all of our measurements ond(cid:0) ifferent specimens, data was
repeatable with less than 10% uncertainties. A block diagrami(cid:0) llustrating the setup for the
measurement is displayed in Figure 1.
In order to assure bond quality of the fabricated specimens and tod(cid:0) etermine the
actual size of any built-in disbonds, several of the laboratorys(cid:0) amples were also inspected by
a standard ultrasonic test c-scan performed in a water bath with a3(cid:0) .5 MHz or a 10 MHz, 0.5
in diameter immersion transducer. Data was taken at approximatelye(cid:0) very 2.3 mm.
Adhesive tapes were used to prevent water from penetrating intot(cid:0) he interface and epoxy
layer, which may create some artifacts due to the scattering ofw(cid:0) aves by the edges of tapes.
computer
scanner
(amp. , freq.)
(x, y)
voltmeter frequency counter
peak detector
signal generator oscilloscope
pre-amplifier (pulsed phase
locked loop)
low-pass filter

Figure 1. Block diagram of setup for Lamb wave measurement.
)
s
t
i
n
u

.
b
r
a
(

e
d
u
t
i
l
p
m
A
unbonded area bonded area

Figure 2. Curve shown is the amplitude variation of S0 mode as af(cid:0) unction of transducer-
pair(cid:213)s position. The image is obtained with UT c-scan for ad(cid:0) oubler. The dark (cid:210)I(cid:211) shaped
area is the designated unbonded area. Amplitude increases whilew(cid:0) ave propagates across
the disbond area.
However, these artifacts can be recognized easily. The UT c-scanr(cid:0) esults were also used to
compare with those obtained with Lamb waves techniqueq(cid:0) uantitatively.
Curve shown in Figure 2 is the amplitude variation of Lamb wavesv(cid:0) s. transducer-
pair position taken on a doubler. This fabricated specimen has a"(cid:0) I"-shaped and all-way-
through disbond as illustrated in the image of UT c-scan (bottomg(cid:0) raph in Figure 2). As
displayed, significant amplitude increase was observed when wavesp(cid:0) assed through areas
with disbond, and its increased magnitude was proportional to thep(cid:0) ropagation path through
disbond. Lamb waves are in-plane waves. Their amplitudes signifyt(cid:0) he integrated result of
interactions of waves with material and structure over their path. (cid:0)Therefore, location of a
disbond and percentage of areas with disbond(s) in the path ofw(cid:0) aves can be estimated with
comparison method. However, the estimation may become misleadingw(cid:0) hen there are
multi-site disbonds. In this regard, measurement of time off(cid:0) light would give additional
information, since difference in wave velocity in bonded area andu(cid:0) nbonded area has been
observed.
Similar results were obtained for embedded disbonds. Figure 3e(cid:0) xhibits changes of
amplitude when the transducer pair was moved in parallel to asw(cid:0) ell as in perpendicular to
several doublers. In the former case, amplitude remainedr(cid:0) elatively constant until waves hit
the disbonds. In the latter, signal level is relatively high whent(cid:0) he path of wave is totally
within single layer areas, and relatively low when the path isc(cid:0) ompletely within bonded
areas. As a matter of fact, the observed time-of-flight of wavesi(cid:0) s slightly different in the
two areas. And, it is believed that waves propagated in ad(cid:0) ifferent mode in each area.
Again, amplitude increased whenever there was a through ore(cid:0) mbedded disbond in the path
of wave beam.
The amplitude increase of sound wave in disbond area can bea(cid:0) ttributed to less
energy transferred to the bottom layer of a doubler. Thisi(cid:0) nterpretation became more evident
Side view
Scanning across A-B-C-D
A B C D
Top view
A B C D
Scanning along A-A’
A’
bonded area unbonded area
Figure 3. Geometry of a multi-doubler specimen is shown on thel(cid:0) eft. Amplitude variations
are exhibited on the right with the scanning directions asi(cid:0) ndicated respectively.
when the same measurements were performed on a lap splice joint. F(cid:0) or a lap joint, the
bondline is the only mechanical connection between the two plates,a(cid:0) nd the amount of wave
energy passing from one plate to the other is thus heavilyd(cid:0) ependent of bond quality. A
disbond decreases the energy propagating in upper platet(cid:0) ransferred to bottom plate, and
results in a reduced amplitude picked up by a receiver transducerp(cid:0) laced on it, which is what
we observed. Figure 4 shows the results of measurement on ana(cid:0) luminum lap joint. Again,
data was taken when the transducer-pair moved in parallel to thel(cid:0) ong dimension of the joint,
with one transducer placed on each plate. Disbonds withd(cid:0) imensions 2 cm x 2 cm, 2 cm x 3
cm, 3 cm x 2 cm were built in for test. As can be seen,c(cid:0) orresponding to four disbonds,
there are four valley-like minima shown in the curve whosel(cid:0) ocations are coincident with the
xxxx bonded area unbonded area
)
s
t
i
n
u
b.
r
a
(
e
d
u
t
i
l
p
m
A
0 35
xxxx Distance (cm)
Figure 4. Amplitude variation of S0 mode as the transducer-pairs(cid:0) cans in parallel to the long
dimension of a lap splice joint. Locations of the minima in thec(cid:0) urve are coincident with
those of the built-in disbonds.
Transmitter (T) Receiver (R1) YYYY
Receiver (R2)
.
XXXX
Amplitude of R1 as a function of X
T - R1
Image of UT-c scan unbonded area
Amplitude of R2 as a function of x rivet holes
T - R2

Figure 5. Curve shown at the top is amplitude as a function oft(cid:0) ransducer-pair(cid:213)s position
when both transducers are placed on the top plate of a lap joint. D(cid:0) isplayed in the middle of
the figure is UT c-scan image of the specimen with disbond asi(cid:0) ndicated. Curve on the
bottom is also the amplitude variation but when the transducersa(cid:0) re on the different plates.
centers of fabricated disbonds. The amplitude decreases with as(cid:0) lope instead of a sharp drop
to the minimum at each site because of the finite size of thed(cid:0) efects and the sound beam.
Samples of lap splice joint with rivet were also tested and thea(cid:0) mplitude variation
pattern is more involved due to the scattering of waves by columno(cid:0) f rivets. The typical
diameter of a rivet is 0.7 cm, and that of a transducer used ino(cid:0) ur measurements is 1.27 cm.
A portion of the transmitting wave would be scattered to otherd(cid:0) irections when the
transducer-pair is aligned with the column of rivets, whichr(cid:0) esults in a significant decrease in
received amplitude. Therefore, a periodic up-and-down change ina(cid:0) mplitude is observed
when transducer-pair is scanned along a joint with evenly-spacedr(cid:0) ivet columns. This
periodic change adds some complexity in data interpretation fort(cid:0) he disbond detection.
Fortunately, the response of wave to a disbond of size larger than1(cid:0) cm in diameter is quite
pronounced, and can be recognized. As a matter of fact, thed(cid:0) isappearance of periodicity in
amplitude variation can be used to determine the existence ofd(cid:0) efects. This approach was
used in analyzing data collected from measurements engaged on lapj(cid:0) oints in the skin of a
Boeing 747 aircraft. Results were fairly consistent with thoseo(cid:0) btained by using other
techniques and by visual inspection after this particular sectiono(cid:0) f lap joint was removed
from the aircraft and torn apart. Figure 5 exhibits the resultso(cid:0) f scanning on a laboratory-
fabricated specimen. This epoxy-bonded sample has three rows off(cid:0) asteners. The round
black dots shown in the UT c-scan image indicate the positions off(cid:0) asteners. Curve shown
below the image is amplitude variation of the lowest orders(cid:0) ymmetric (S0) mode as a
function of position of transducer-pair when each of them is ond(cid:0) ifferent plates. The peaks
represent the maximum wave energy propagating between rivetc(cid:0) olumns. As discussed
above, disbond would prevent transfer of wave energy betweenp(cid:0) lates, which has resulted
in a flat line in the curve meaning minimum energy is received. T(cid:0) he small peaks located at
the positions of rivet columns are the result of diffraction ofw(cid:0) aves by rivet column, and
whose magnitude is quite dependent of the bond condition in thea(cid:0) rea surrounding the rivets
and the distance between transducers. For comparison, curved(cid:0) isplayed above the c-scan
image is the amplitude changes when both transducer are placed ont(cid:0) he upper plates. As can
be seen, a larger amplitude reveals the existence of disbond,w(cid:0) hich is similar to what has
been observed for doublers (figures 2 and 3). Amplitudev(cid:0) ariations of the lowest
antisymmetric (A0) mode were also measured and displayed similarb(cid:0) ehavior to those of S0
mode. However, A0 mode seems more sensitive to unevenness int(cid:0) hickness of bondline.
This could be due to the much smaller wavelength of thism(cid:0) ode.
In general, velocity of Lamb wave is not only frequency dependentb(cid:0) ut also
thickness dependent. To the propagation of Lamb waves, a disbondr(cid:0) epresents a relatively
large decrease in effective thickness of the media, which couldr(cid:0) esult in change of wave
mode and/or change of velocity. A pulsed-phase-locked-loop wase(cid:0) mployed to monitor the
change of velocity. This instrument compares the phase of itsp(cid:0) ulsed output signal (which is
sent to transmitting transducer) with that of the returned signal((cid:0) from the receiving
transducer). Phase difference of the two signals varies with thec(cid:0) hange of sound velocity
propagating in the medium when distance between transducers isf(cid:0) ixed. Before the scanning
a certain phase difference is chosen and locked. During thes(cid:0) canning, the loop responds to
the sound velocity change by adjusting its output signal frequency((cid:0) called reference
frequency) in order to keep this phase different constant as itw(cid:0) as locked. Therefore, a
reading from a frequency counter would reveal the information ofv(cid:0) elocity changes. In fact,
it can be proved that the percentage of increase in referencef(cid:0) requency is the percentage of
decrease in time-of -flight. Figure 6 displays the change inr(cid:0) eference frequency for the
specimen with disbond shown in figure 5 in the case when twot(cid:0) ransducers were placed on
different plates. As can be seen, reference frequency decreasesi(cid:0) n the area where there is a
1.70
)
z
H
1.66
M
(

y
c
n 1.62
e
u
q
e
r 1.58
F

e
c
n
e
r 1.54
e
f
e
R
1.50
0 5 10 15 20 25 30
Distance (cm)

Figure 6. Curve shown is the variation of time-of-flight as af(cid:0) unction of transducers(cid:213)
location. A decreased reference frequency represents the longert(cid:0) ime-of-flight caused by a
disbond.
disbond in the wave propagation path, and the magnitude of changei(cid:0) s proportional to the
dimension of disbond in the direction of propagation. For thiss(cid:0) pecimen, disbond causes
longer time-of-flight, which indicates a slower wave velocity. T(cid:0) he small kinks appeared in
the relatively flat portion of the curve are found to locate ate(cid:0) dges of rivet columns, and are
ascribed to the interference effect of waves.
DISCUSSION
With the described measurements and results, it is demonstratedt(cid:0) hat Lamb wave has
promising potential for detection of disbonds, at least, in at(cid:0) wo-layered structure. Although
most of the tests were done on laboratory-fabricated specimens,f(cid:0) ield test on aircraft panel
also showed reasonably good results. One of the advantage ofu(cid:0) tilizing Lamb wave is its
capability of assessing the condition of layered structure over al(cid:0) ong path length. The
provided results are the integrated information of its path. Ifd(cid:0) etailed information at each
locations between the transducers is not crucial for ana(cid:0) ssessment, then this is an approach
much more efficient than conventional point-by-point ultrasonicm(cid:0) easurements. Especially
for a specimen geometry, such as that of a lap joint, ao(cid:0) ne-dimensional scan should provide
the necessary information for disbond evaluation. Otherwise, as(cid:0) econd scan in the other
dimension can be performed and would give the exact location ofd(cid:0) isbond.
Disbond detection for structures having more than two layers hasn(cid:0) ot been tested
intensively yet. In theory, if Lamb wave can be generated in am(cid:0) ulti-layered structure, a
disbond occurred in any one of the interfaces should be able to bed(cid:0) etected. However, in
this case, the wave energy distribution may become an intriguingp(cid:0) roblem and eventually
determine what modes can be generated with measurable amplitudes,b(cid:0) ecause the particle
displacement is a function of depth from the surface and thisp(cid:0) roperty of Lamb wave may
become critical when media thickness is not much smaller than thew(cid:0) avelength.
In summary, it is feasible using Lamb wave for a large aread(cid:0) isbond assessment.
Relatively simple amplitude and time-of-flight measurements on lapj(cid:0) oint type structures
have demonstrated this capability although there are many i(cid:0) mprovements can be done in
terms of increasing the inspection speed and setup for them(cid:0) easurements.
ACKNOWLEDGEMENTS
The authors would like to thank Richard Churray and Ken Hodges fort(cid:0) heir
specimen preparation and assistance in measurements.
REFERENCES
1. D. C. Worlton, J. Appl. Phys., 32, 967 (1961)
2. T. R. Meeker, and A. H. Meitzler, in Physical Acoustics, vol. I, part A, 111, edited by
R. N. Thurston (Academic Press, Inc.(cid:0)1964)
3. K. J. Sun and D. Kishoni, IEEE 1991 Ultrasonics(cid:0)Symposium Proceedings, 859 (1991)
4. K. J. Sun and P.H. Johnston, IEEE 1992 Ultrasonics(cid:0)Symposium Proceedings, 763
(1992)
5. J. S. Heyman and E. J. Chern, Journal of Testing(cid:0)and Evaluation, vol. 10, No. 5, 202
(1982)