Acoustic emission based tensile characteristics of sandwich composites
Amilcar Quispitupaa, Basir Shafiqb,*, Frederick Justa, David Serranoa
aDepartment of Mechanical Engineering, University of Puerto Rico, Box 9045, Mayagu¨ez, PR 00681, USA
bDepartment of General Engineering, University of Puerto Rico, Box 9044, Mayagu¨ez, PR 00681, USA
Received 6 June 2003; accepted 6 November 2003
Available online 9 April 2004
Abstract
Sandwich composite static and fatigue testing results indicated the predominant failure to be the core damage followed by interfacial
debonding, resin cracking and fiber rupture. Under static testing, crack was observed to initiate in the core and ensue planar propagation near
the interface with the facesheets; whereas, onset of crack initiation in the facesheets served as a precursor to the catastrophic failure. Multiple
failure initiation and propagation sites in the core and intermittent interfacial debonding were consistently observed under fatigue. An
acoustic emission based stiffness reduction model is presented that seems to accurately identify the extent of damage in sandwich composites
subjected to fatigue loading conditions.
q 2004 Elsevier Ltd. All rights reserved.
Keywords: D. Acoustic emission; Sandwich composites; B. Fatigue
1. Introduction
It is generally a difficult task to detect damage before the
onset of catastrophic failure in sandwich composites. Major
impacts may produce audible sounds and visible damage on
the surface of the composite, whereas, routine minor
impacts degrading the softer core or the interface between
the core and the facesheets (FS) may go unnoticed [1,2]. In
marine ship hulls, sandwich structures are normally
designed to ensure that the core failure precedes FS failure
to prevent water ingress. Difficulty arises as conventional
non-destructive evaluation (NDE) techniques have limited
capability to perform dynamic inspection on the interior of
the sandwich composites. Whereas, acoustic emission (AE)
technique permits continuous damage inspection, classification
and identification of modes of failure in real time,
which is critical for taking preventive measures [3,4].
While, AE technique has extensively been used in damage
detection and failure analysis of homogeneous metals,
concrete and laminated composites, etc. [3–9]; scarcely any
sandwich composite applications are available in the
literature, especially related to fatigue crack growth (FCG)
behavior [1,2,10,11].
AE technique, though useful, requires significant preliminary
analysis and calibration for each system of material
involved, geometry and type of loading to distinguish
among various types of damage and failure mechanisms.
Threshold frequencies that are material dependent need to
be accurately set to filter out spurious noises without
interfering with the useful data [3,8,12,13]. Furthermore,
AE occurrences need to be carefully related to the
microscopic and macroscopic deformations occurring in
the material. Nevertheless, the payoff in terms of accurate
damage detection far outweighs the few drawbacks
associated with the use of AE technique, as evidence by
its increasing popularity in various fields of engineering.
Presence of various isotropic and orthotropic constituents
renders the failure modes in sandwich composites quite
complex [1,2,10,11]. For example, an important feature of the
fatigue in (sandwich) composite materials is the phenomenon
of multiple cracks; rather than single crack propagation as
observed in most isotropic materials [1,5,14]. Although
sandwich composites are generally designed to carry flexural
loads, ship hull components inevitably undergo tensile loading
conditions. No study has, to date addressed the important
aspect of tensile static or fatigue performance of sandwich
composites.This paper, therefore, is intended to fill some of the
gap that exists in the literature related to FCG characteristics
and application of AE technique to sandwich composites.
1359-8368/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesb.2003.11.012
Composites: Part B 35 (2004) 563–571
www.elsevier.com/locate/compositesb
* Corresponding author. Tel.: þ1-787-832-4040x2094; fax: þ1-787-
265-3816.
E-mail address: abasir@uprm.edu (B. Shafiq).
2. Experimental setup
The sandwich composites used in this study were made
of bi-directional [08/908] carbon fiber FS with urethane filled
kraft paper honeycomb core whereas resin was used to bond
FS to the core, as shown in Fig. 1. VARTM technique was
employed to fabricate the specimens in-house [15]. Three
types of specimen geometries were considered, namely,
rectangular, dogbone (DB) and single edge notch (SEN), as
shown in Fig. 2. However, rectangular specimens were used
only for preliminary analysis and system calibration.
Aluminum plates were glued and bolted to the specimen
ends in a special arrangement as shown in Fig. 3 to avoid
crushing of the specimen at the grips. Additionally, sound
filtering tabs were used between the aluminum plates and
the grips to minimize signals originating from friction at
the grips. All tests were performed on a servo-hydraulic
testing machine attached to a data acquisition system, an
eight channel AE setup, and a digital microscope. A
minimum of five specimens of each type were tested for a
given set of loading parameters.
Static tensile testing was performed on various types of
specimens shown in Fig. 2 under displacement and load
control rates of 2 mm/min and 222 N/min, respectively.
Tension–tension fatigue tests were performed on SEN
specimens only, at load ratio of 0.1 and at a frequency of
1 Hz. Testing yielded unlimited life and/or negligible AE
activity below 80% stress level. Therefore, the fatigue
results reported in this paper correspond only to the tests
performed between 80 and 95% of the ultimate static load.
Acoustic wave speeds that are required for source
location were calculated based on ASTM E976 standard
[16]. For the sandwich composite used, values of wave
speeds were found to be 4347 m/s (with 42 m/s standard
deviation) and 2754 m/s (with 79 m/s standard deviation), in
the longitudinal and through the thickness direction,
respectively. Wave speeds did not agree well with published
results (of other sandwich composite systems) as AE
parameters are very sensitive to the material type and
geometry [1,5].
Accurate interpretation of the AE results demands a
classification strategy to sort out a large amount of data
collected. Detection of damage, crack initiation and growth
in the current study was primarily based on a simultaneous
analysis of AE events, energy, time, amplitude and source
location as they yielded information pertinent to damage
classification (such as core failure, resin cracking, interfacial
failure and fiber rupture).
3. Results and discussion
3.1. Static tensile test results
The loads in the sandwich composites are primarily
carried by the high stiffness carbon fiber FS, whereas an
order of magnitude softer core, serves to enhance toughness
while keeping it lightweight and the bonding agent is
responsible for the maintenance of two-phase action of the
composite. The mean load–displacement curves shown in
Fig. 4 for various types of specimens depict an apparent
linear and reversible behavior leading up to the catastrophic
failure. However, it must be realized that this load–
displacement behavior primarily reflects the load carrying
capability of the FS. AE analysis provides evidence of core
and interfacial failure long before any indication of FS
cracking, therefore, curves shown in Fig. 4 are not
reversible.
Fig. 4 also indicates a gradual decrease in the load
carrying capability and stretching to failure of different
types of specimens due to the presence of induced defects,
such as, the hole and notches. As expected, solid DB
specimens achieved highest ultimate static strength followed
by DB with a hole, SEN and DB with a hole and two
notches, respectively. Furthermore, notched specimen
analysis indicated a relatively low fracture resistance, but
significantly higher stiffness to weight ratio for the sandwich
composites when compared with carbon fiber laminated
composites of similar dimensions [7].
The initial portion of the load–displacement curve up to
the yield point of the weakest constituent of the sandwich
composite (i.e. the core) yielded very quite AE region for all
specimen types, perhaps associated with the incubation
period. Thereafter, data suggested a sequential progression
of AE activity as the test proceeded. An incremental
enhancement in the level of AE activity was observed
Fig. 1. Sandwich composite constituents.
564 A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571
Fig. 3. Sandwich specimens in a test setup showing acoustic sensors and special grip arrangement.
Fig. 2. Samples geometries: solid DB, DB-central hole, DB-central hole and two notches, SEN and rectangular (R: semicircular radius, r: hole radius).
A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571 565
(going from solid DB, DB with a hole, DB with two notches
to SEN specimen) as the fabricated defects became more
pronounced. Fig. 5 (events vs amplitude) and Fig. 6 (energy
vs amplitude) show typical AE activity as it relates
to amplitude of occurrence with the corresponding
classification of failure. Fig. 7 shows time vs amplitude
plots for static and fatigue testing along with the overall
percentage of AE activity (damage) incurred in various
constituent of the sandwich composite during the time
indicated.
In specimens with fabricated defects, crack initiation was
localized near points of high stress concentration. According
to AE results (such as Figs. 5 and 6 and source location)
core failure invariably initiated near the interface with the
FS at the lowest energy level and dominated the earlier part
of testing by gradually propagating along the interface and
through the thickness direction. Core failure was followed
by resin cracking and subsequent interfacial failure, as
reflected by AE events and energy occurrences shown in
Figs. 5 and 6. Post-test analysis of various specimen types
indicated that the resin cracking was not wide spread, as
most of the core-FS separation was caused by a planar core
failure (in the plane of specimen surface) near the FS. Onset
of fiber rupture immediately led to catastrophic failure of the
solid DB specimens. Whereas, some short-term fiber rupture
AE activity preceded catastrophic failure in notched specimens.
Somewhat similar results have been reported for
laminated composites [3,4,7–9]. Catastrophic failure
always occurred in mode I as shown in Fig. 8 for a typical
specimen.
Based on AE analysis of the test results, damage was
classified in various constituents of the sandwich composite
as a function of AE amplitude and energy and
presented in Table 1. The level of AE amplitude and energy
Fig. 4. Mean load–displacement curves of solid DB, DB-central hole, DB-central hole and two notches and SEN specimens.
Fig. 5. Events vs amplitude of a typical specimen at a given instant in time showing highest number of events corresponding to core.
566 A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571
Fig. 6. Energy vs amplitude of a typical specimen at a given instant in time showing highest energy release corresponding to fiber rupture.
Fig. 7. Percentage of overall AE activity for static testing between 12 and 40 s and between 61,959 and 309,790 cycles for fatigue testing of typical specimens.
A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571 567
were found to be independent of the specimen geometry or
loading type for the sandwich composite used. To confirm
AE damage classification results, various preliminary tests
were terminated at certain AE amplitude and energy levels,
specimens were carefully removed, dissected and analyzed
microscopically to confirm AE sequence of failure given in
Table 1. This classification agreed well with the sequence
reported in the literature, however, not necessarily with the
amplitude or energy cutoffs, as AE parameters are quite
sensitive to the material type [1,8,9].
AE figures represent dynamic, transient data that is
updated continuously throughout the duration of the test.
For example, the typical data shown in Figs. 5 and 6 is
captured at an arbitrary instant in time during the test that
changed in the very next moment. Therefore, care must be
exercised in interpreting AE graphical presentations. What
is significant is the cumulative AE activity during various
time intervals as it forms the basis for life prediction
modeling. Data shown in Fig. 7 (amplitude vs time)
provides an overall AE statistics during the indicated
time; it can be seen that the core damage activity occurred
83% of the time whereas fiber breakage consumed only
0.7% of the typical static testing time.
3.2. Tensile fatigue characteristics
Events and energy vs amplitude analysis (such as shown
in Figs. 5 and 6) indicated a high level of AE activity
corresponding to distinct crack initiation sites in the core
and the interface near the notch tip during the initial stage of
the fatigue test that lasted for about 2000 cycles. Bursts of
AE activity persistent at irregular intervals pointed to
Fig. 8. Fracture surface of a typical specimen under static testing.
Table 1
Sequence of failure in sandwich composites with corresponding amplitude
and energy ranges
Failure mode AE amplitude (dB) Range of energy of AE
Core damage 45–60 0–25
Interface failure 60–80 3–219
Resin cracking 80–90 88–374
Fiber rupture Above 90 347–13,568
568 A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571
additional core and interfacial crack initiation sites.
Distinctly different cracks were observed to initiate and
propagate on each side of the specimen as seen in Fig. 9.
Sequence of failure, such as core damage, resin cracking,
interfacial failure and fiber rupture was observed to occur at
amplitudes and energies similar to those obtained under
static testing and shown in Table 1. Somewhat similar
failure sequence has been reported for flexural fatigue tests
on sandwich composites, however, unlike flexural FCG,
significant fiber rupture never took place until catastrophic
failure under tensile fatigue. Furthermore, static tensile and
flexural fatigue tests yield a continuous growth of a single
crack, whereas, tensile fatigue indicated presence of
multiple crack fronts and periodic FCG with long
intermittent dormant intervals as evidenced by AE results.
It is suspected that flexural fatigue tests perhaps underwent
similar discontinuous crack growth, however, such analysis
was not performed in the studies reported [1,10,11].
At 80% stress level, AE and post-test analysis indicated
that the mode I crack growth in the core and interface
dominated the initial stage(s) of the test. Whereas, in the
subsequent stages, mode II crack primarily propagated in
the core along the interface and parallel to the applied load.
The literature does not list a clear endurance limit for
sandwich composites, however, flexural fatigue life has
been reported for as low as 60% of the static load [10]. In the
current study, insignificant failure/AE activity took place
below 80% of stress level, which is a reflection of the
difference between flexural and tensile stress distribution
along various constituents of the sandwich composite.
Furthermore, multiple crack fronts observed under tensile
fatigue served to disperse energy, therefore, stagnating
crack tip(s) advancement.
The AE activity was consistently observed to diminish
within a few thousand cycles of the test. Therefore, after
failing to reach catastrophic failure at 80% stress level in
several specimens, a new testing methodology was
devised. By increasing the stress level during the test,
AE activity at the crack tips was expected to resume. Stress
levels were periodically increased by 5% after the crack
remained dormant for approximately 50,000 to 300,000
cycles. At the end of each stress level, the specimen was
removed from the testing machine, its cracking pattern was
observed microscopically and marked with luminescent
dye penetrant for post-test analysis. AE and post-test
analysis indicated mode I failure to occur primarily during
the initial portion of the first stage (up to 3.9 mm FS crack
length) or during the final portion of the last stage (just
before the catastrophic failure). Catastrophic failure always
occurred in mode I as shown in Fig. 8. FCG during
Fig. 9. Modes I and II crack growth under fatigue testing.
A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571 569
intermediate stages was dominated by intermittent mode II
crack growth in the core and the interface as evidenced by
typical AE statistics shown in Fig. 7 for the cycles
indicated at 90% stress level. Some specimens also
exhibited significant FS crack growth in the loading
direction (up to 15 mm) in the intermediate stress levels,
as shown in Fig. 9.
Curiously enough, high AE activity at the beginning of
each incrementing stress level was observed to wane within
a few thousand cycles. This was probably caused by energy
dispersion due to initiation and propagation of multiple
crack fronts, especially in the loading direction. In the final
stage when the stress level was raised to 95% of the ultimate
static strength, significant fiber rupture led to severe
stiffness reduction and catastrophic failure of the specimen
with little warning or surface damage. Comparison with
reported flexural fatigue results indicates that the sandwich
composites offer better fatigue resistance under tensile
loading conditions [10,11].
4. Stiffness reduction model for tensile fatigue
Tensile FCG with multiple crack fronts which
perhaps resulted from large difference in the
elastic properties of the constituents of sandwich composite
ðEFSqEinterface . EcoreÞ; presence of relatively weak
interface and inherent manufacturing flaws, did not yield
data suitable for the calculation of conventional CGR
(da=dN vs Dk) or S–N curves. Therefore, to accommodate
the complex FCG characteristics of sandwich composites,
parameters of failure had to be modified. A useful parameter
that could indicate damage or stiffness reduction is the
extent of damage (number of cracks and weighted sum of
crack lengths in each constituent of the sandwich composite),
which could be represented as
D1 ¼ X
n
i¼1
EiaiðuÞ ð1Þ
where Ei is the modulus of elasticity of the constituent
and aiðuÞ is the corresponding weighted crack length as a
function of crack angle. However, it would be a tedious
process to accurately measure the length and orientation
of each crack even with the aid of AE. A more
manageable stiffness reduction parameter can be based
on the overall AE activity during the fatigue test. The
underlying assumption is that the extent of damage (or
cracking) in each constituent of the sandwich composite is
directly proportional to the AE activity in that constituent.
This is a reasonable assumption as there cannot be any
AE activity unless damage is accumulating in that
particular constituent of the sandwich composite during
fatigue testing. Furthermore, Kaiser’s effect would ensure
that loading/unloading process does not introduce repetitive
AE activity [9].
The methodology for the acquisition of this stiffness
reduction parameter is based on the percentage of total AE
energy released by each constituent, which is proportional
to the percentage of damage. This percentage (of energy) in
turn is determined from overall number of events associated
with each constituent of the sandwich composite within a
given number of cycles, such as shown in Fig. 7. The weight
factor for each constituent, Xi; is calculated from the static
test results as the sum of the product of total number of
events for each constituent failure and respective energy
level, divided by the total energy. Weight factor indicates
the importance of a constituent in maintaining the integrity
of the sandwich composite. For example, the substantially
higher value of Xfiber as compared to Xcore (Table 2)
indicates that the sandwich composite can remain intact
even after the core fails, whereas catastrophic failure is
invariably observed to be a consequence of fiber rupture.
The quantification of events and energy for both static and
fatigue tests was based upon the cumulative effect of events
vs amplitude and energy vs amplitude data over time (Figs. 5
and 6) along with AE damage classification given in Table 1.
Stiffness degradation D1ðtÞ under fatigue testing can be
presented as
D1ðtÞ¼½D1ðtÞX1þ½D2ðtÞX2þ½D3ðtÞX3¼X
n
i¼1
DiðtÞXi ð2Þ
where D1ðtÞ;D2ðtÞ;D3ðtÞ reflect cumulative percent damage
in the core, interface and the fiber for a given time interval,
respectively; and x1;x2;x3 are the corresponding weight
factors. Notice that interfacial failure and resin cracking are
combined in Table 1 to acquire D2 (the interfacial damage
component).
This model provides an important fatigue stiffness
reduction parameter that can be quite useful in the
remaining lifetime prediction of sandwich composites as a
function of time. A few examples shown in Table 2 with
data taken from actual static and fatigue tests clearly
indicate significantly higher value of stiffness reduction
corresponding to slight increase in the percentage of fiber
rupture.
5. Conclusions
AE yielded very accurate information about the extent
and location of damage in various constituents of sandwich
Table 2
An example of stiffness reduction model applied to actual fatigue data
D1 D2 D3 D1
Test 1 96.6 3.2 0.2 4.05
Test 2 93.9 6.1 0 4.09
Test 3 90 5 5 8.11
Note: The values of x1; x2 and x3 used in above calculations are 3.67,
10.67 and 85.7%, respectively.
570 A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571
composites. AE and post-test analysis indicated core failure
to be the predominant damage mechanism followed by
interfacial failure, whereas, fiber rupture triggered the onset
of catastrophic failure. Fatigue results indicate that
sandwich composites offer better resistance against tensile
loads as compared to flexural loading condition. Multiple
crack fronts in various constituents of the sandwich
composite prohibited calculation of conventional FCG and
lifetime parameters. However, AE based stiffness reduction
model is presented that seems to capture the essence of
material degradation in sandwich composites under fatigue
loading conditions.
The authors wish to acknowledge the authorities at ONRComposites’
for Marine Structures division for their
financial support for this work. Special thanks are due to
Dr Yapa Rajapakse, the ONR program manager for his
unrelenting support and guidance.
References
[1] Burman M. Fatigue behavior initiation and propagation in sandwich
structures. Royal Institute of Technology. Report No. 98-29. ISSN
0280-4646, Sweden.
[2] Avery J, Sankar B. Compressive failure of sandwich beams with
debonded face sheets. J Compos Mater 2000;34(14):1177–99.
[3] Bakukas J, Prosser W, Johnson W. Monitoring damage growth in
titanium matrix composites using acoustic emission. J Compos Mater
1994;28(4):305–28.
[4] Daniel IM. Experimentation and modeling of composite materials.
Exp Mech 1999;39(1):1–19.
[5] Harris D, Dunegan H. Continuous monitoring of fatigue crack growth
by acoustic emission technique. Third SESA International Congress
on Experimental Mechanics, Los Angeles; 1974.
[6] Li F, Li Z. Acoustic emission monitoring of fracture of fiberreinforced
concrete in tension. ACI Mater J 2000;97(6):629–36.
[7] Prosser WH, Jackson KE, Kellas S, Smith BT, Mckeon J, and
Friedman A. Advanced waveform-based acoustic emission detection
of matrix cracking in composites. Mater. Eval 1995;1052–8.
[8] Zheng G, Buckley M, Kister G, Fernando F. Blind deconvolution of
acoustic emission signals for damage identification in composites.
J Am Inst Aeronaut Astronaut 2001;39(6):1198–205.
[9] Vargas A. Acoustic emission for quality control in composites. TR-
103-58-6/85, Princeton Junction, NJ: Physical Acoustic Corporation;
1985.
[10] Mahi A, Farooq M, Sahraoui S. Mechanical behavior of sandwich
composite material under cyclic fatigue. New Trends Fract Fatigue,
Metz, France 2002;.
[11] Kanny K, Mahfuz H, Thomas T, Jeelani S. Effects of loading
frequency on the fatigue behavior of sandwich composites. Center for
Advanced Materials (Internet Report), Tuskegee University; 2002.
[12] Huang M, Jiang L, Liaw PK, Brooks CR, Seelev R, and Klarstrom
DL. Using acoustic emission in fatigue and fracture materials
research. JOM 1998;50(11):1–14.
[13] English L. Listen and learn, AE testing composites. Phys Acoust Corp
1987; TR103-75-6/87.
[14] Caprino G. Predicting fatigue life of composite laminates subjected to
tension–tension fatigue. J Compos Mater 2000;34(16):1334–55.
[15] Vaidya U, Abraham A, Bhide S. Affordable processing of thick
section and integral multi-functional composites. Composites, Part A
2001;32(8):1133–42.
[16] Nondestructive Testing Handbook, 2nd ed. American Society for
Nondestructive Testing, 5; 1987.
A. Quispitupa et al. / Composites: Part B 35 (2004) 563–571 571