Composite Structures 108 (2014) 267–276

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Impact damage identification in composite laminates using vibration

testing
Marco A. Pérez a,c,⇑, Lluís Gil a,c, Sergio Oller b,c
a
Laboratory for the Technological Innovation of Structures and Materials (LITEM), Colon 11, TR45, Terrassa, 08222 Barcelona, Spain
b
International Center for Numerical Methods in Engineering (CIMNE), Gran Capità s/n, Edifici C1, 08034 Barcelona, Spain
c
Departament de Resistència de Materials i Estructures a l’Enginyeria, Universitat Politècnica de Catalunya - BarcelonaTech, Spain

a r t i c l e i n f o a b s t r a c t

Article history: Due to the problems arising from impact damage in composite laminates, there is a need to develop fast,
Available online 25 September 2013 accurate, cost-effective and non-destructive testing methods to identify this type of damage at an early
stage and thus enhance the service life of composite structures. This paper presents the results of an
Keywords: extensive experimental campaign conducted to investigate the feasibility of using vibration-based meth-
Composite laminates ods to identify damages sustained by composite laminates due to low-velocity impacts. The experimental
Damage identification programme included an evaluation of impact damage resistance and tolerance according to ASTM test
Low-velocity impact
methods, characterisation of induced damage by ultrasonic testing and quantification of the effects on
Non-destructive evaluation
Experimental modal analysis
the vibration response. The damage identification involved the detection, localisation, quantification
Compression after impact and estimation of the remaining bearing capacity. Four damage indicators based on modal parameters
were assessed by comparing pristine and damaged states. The results allowed for conclusions to be
drawn regarding the capability and suitability of each damage indicator, including its ability to detect
impact-induced damage, its precision in determining the location of damage, its sensitivity regarding
damage extent and pertinent correlations with residual bearing capacity.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction (NDT) methods is required to identify this type of damage at an

early stage and thus enhance the service life of composite
Laminated composite materials offer engineers advantages that structures.
are especially appealing for structural applications; however, Over the years, due to the significance of this problem, several
underlying problems arising from impact damage have greatly hin- non-destructive damage assessment techniques have been used
dered their widespread application, particularly in structures for impact damage detection in composites, including acoustic
prone to impacts during service. The susceptibility of laminated emission, radiography, shearography, thermography and ultrason-
composites to impact damage is due to the material’s lack of plastic ics, whose development continue to date Aymerich and Meili [3],
deformation, low interlaminar strength and laminated construc- Schilling et al. [4], Růžek et al. [5], De Rosa et al. [6], Hung et al.
tion to reduce the anisotropic nature of the plies Reid and Zho [1]. [7], De Angelis et al. [8], Goidescu et al. [9], and Bull et al. [10].
The damage induced by a low-velocity impact is a mixture of An overview of the advantages and disadvantages of some of the
three main failure modes: matrix cracking, delamination and fibre currently available NDT methods are presented in Garnier et al.
breakage, among which delamination the most severe because it [11]. All of these NDT methods have allowed for considerable pro-
may severely degrade the stiffness and strength of composite gress in the structural integrity assessment of composite compo-
structures. The situation has been proved critical when the dam- nents. However, almost all of these techniques require that the
aged surface of an impact point is not representative of the internal vicinity of the damage site be known in advance and that the dam-
damage induced, commonly referred to as barely visible impact aged region of the structure being inspected is readily accessible. It
damage (BVID), leading to a collapse under unacceptable compres- should be noted that some of the techniques are also impractical
sive load levels Hodgkinson [2]. For these reasons, the develop- for applications under in-service conditions, thus requiring with-
ment of fast, accurate, cost-effective and non-destructive testing drawing the component or structure being assessed from service
Doebling et al. [12]. In addition, these techniques provide local
information and no indication of residual structural performance.
⇑ Corresponding author at: Laboratory for the Technological Innovation of
Over the past three decades, numerous researchers have de-
Structures and Materials (LITEM), Colon 11, TR45, Terrassa, 08222 Barcelona, Spain.
Tel.: +34 937398289. voted their efforts to developing so-called vibration-based NDT
E-mail address: [email protected] (M.A. Pérez). methods Doebling et al. [12, Sohn et al. [13], Carden [14], Yan

0263-8223/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2013.09.025
268 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276

et al. [15], Fan and Qiao [16] and Kim et al. [17]. The principle consolidating a composite specimen in an autoclave. By doing so,
underlying these methods is that a vibration response depends the influence of other failure modes induced during the impact
on the physical properties of a structure (mass, damping and event, such as fibre breakage, is not considered. Despite the exten-
stiffness); therefore, changes that occur in physical properties sive studies of vibration analysis on damaged laminated compos-
due to damage can result in detectable variations in the vibration ites, the literature reviewed concentrates primarily on Levels 1
response, which can serve as an indicator of structural integrity and 2 of damage identification Polimeno and Meo [21], Aymerich
Doebling et al. [12]. A considerable number of contributions in and Staszewski [22], Ooijevaar et al. [23], Wei et al. [24], Kessler
the field of vibration-based NDT have been made – Doebling et al. [25], Frieden et al. [27], Fan and Qiao [28] and Niemann
et al. [12], Sohn et al. [13], Carden [14], Yan et al. [15], Fan and Qiao et al. [29], i.e., detection and localisation, respectively, whereas
[16]. These methods can be broadly classified into two approaches: the quantification of damage and estimation of the residual bear-
model-based methods, i.e., methods that require numerical mod- ing capacity are in a relatively immature stage.
els, and response-based methods, i.e., methods that use only exper- The aim of the present study was to investigate the feasibility of
imental data to identify damage. The second approach includes using vibration-based methods to identify damages on composite
methods based on the time domain, frequency domain and modal laminates resulting from low-velocity impacts. In contrast with
domain, the last of which has received the most attention Fan and previous publications addressing the detection of damage in
Qiao [16], Kim et al. [17], Radzieński et al. [18]. With respect to this composites, the present work focussed on real damage induced
group, one can find in the literature modal parameter-based dam- by low-velocity impacts and sought to address the four levels of
age identification methods, such as natural frequency-based meth- identification mentioned previously. This paper mainly presents
ods, mode shape-based methods and curvature mode shape-based the results of an extensive experimental campaign carried out on
methods, among others. Another type of damage-identification a set of 48 carbon fibre-reinforced composite laminated specimens.
method common in the open literature Carden [14] defines four Composite coupons were manufactured, and the impact damage
levels of damage identification: determination of the presence of resistance and tolerance were evaluated using ASTM standard test
damage in a structure (Level 1), determination of the geometric methods. The effect of impact-induced damage on the vibration re-
location of damage (Level 2), quantification of the severity of dam- sponse was analysed in terms of four damage indicators computed
age (Level 3) and prediction of the remaining service life of a struc- from modal parameters. The results allowed for conclusions to be
ture (Level 4). Detailed reviews on vibration-based NDT methods drawn regarding the capability and suitability of each damage indi-
for structural damage identification can be found in Doebling cator presented to identify impact damage.
et al. [12], Sohn et al. [13], Carden [14], Yan et al. [15], Fan and Qiao In the following section, the mechanical background of the
[16]. experimental modal analysis performed for modal parameter esti-
The concept of using vibration-based NDT as the basis for damage mation and the damage indicators that will be assessed are briefly
identification in laminated composites is not a new one. Experimen- presented. Section three discusses the methodology and details of
tal evidence has shown that the presence of delamination changes the experimental test procedures. The results are discussed in sec-
the strength as well as the vibration response of composite struc- tion four. Finally, the conclusions of the study are presented in the
tures Zou et al. [19] and Della and Shu [20]. Using this approach, last section.
numerous researchers have reported analytical, numerical and
experimental results regarding the effect of delamination on the 2. Theoretical background
vibrational response of laminated composites Zou et al. [19], Della
and Shu [20], Polimeno and Meo [21], Aymerich and Staszewski 2.1. Identification of modal parameters
[22], Ooijevaar et al. [23], Wei et al. [24], Kessler et al. [25], Yam
[26], Frieden et al. [27], Fan and Qiao [28], Niemann et al. [29], Araújo The proposed methods for damage identification examine mod-
dos Santos et al. [30] and Frieden et al. [31]. Due to the complexity of al domain data through modal analysis, in which time domain data
the physical phenomena involved, the analytical formulations pro- are mapped onto the frequency domain. The modal parameters are
posed are restricted to particular cases Della and Shu [20]. On the then extracted from the so-called frequency response functions
other hand, the effectiveness of the model-based methods is depen- (FRFs). The structural response X(x) is directly related to the sys-
dent on the accuracy of the numerical model used. This fact is of ut- tem forcing function F(x) through the quantity H(x) as follows:
most importance in impact damage detection in laminated
XðxÞ
composites because induced damage is a complex mixture of multi- HðxÞ ¼ ð1Þ
ple failure modes interacting with each other Pérez et al. [32]. FðxÞ
Several response-based methods have been proposed in the lit- where X(x) and F(x) are the n-vectors of Fourier transformed re-
erature as alternatives or supplements to the model-based meth- sponses and force inputs, respectively, x is the frequency variable
ods Polimeno and Meo [21], Aymerich and Staszewski [22], and H(x) is the FRF n  n matrix, where n is the number of test de-
Ooijevaar et al. [23]. To date, the majority of these studies have fo- grees-of-freedom (DoFs) of the structure. To estimate the modal
cused on measuring the vibration response before and after dam- parameters from FRF measurements, a curve-fitting method is used.
aging a composite, analysing the data in both the frequency and Assuming that stationarity, linearity and reciprocity are valid for a
modal domains. These studies provide a basic understanding of test structure, the FRF matrix can be analytically represented in
the influence of damage -especially delaminations- on vibration re- terms of the modal parameters of the structure.1 Each FRF matrix
sponse and seek to demonstrate the feasibility of using measurable component can be written in partial fraction form as follows:
changes in vibrational characteristics to identify damage in lami-
m 
X 
nated composites. However, the sensitivity and measurability of Q i / /t
i i Q i /i /ti
HðxÞpq ¼ þ ð2Þ
shifts in vibration response due to induced damage is a point of i¼1
jx  k i jx  ki
controversy among many researchers Kim et al. [17], Wei et al.
[24] and Kessler et al. [25]. It is also notable that a large number
of experimental works are based on the analysis of the influence
of an isolated, artificially induced damage Ooijevaar et al. [23], 1
The formulations used are thoroughly described in references Piersol and Paez
Wei et al. [24], Kessler et al. [25], Yam [26]. It is common practice [33] and will not be elaborated here except for the main aspects, which are briefly
to induce delamination by inserting a polyimide film before described subsequently.
 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276 269

where p q are the measured and excited DoFs, i is the mode indica- be expected to be identified. Modal Assurance Criterion (MAC) is
tor, m is the number of modes and ⁄ denotes the complex conjugate. a commonly used method for comparing two sets of mode shapes
The numerator is the residue of the FRF, which is a function of the Piersol and Paez [33] and Allemang [34]. MAC provides a measure
product of the modal scaling factor Qi and the modal deformations of consistency (degree of linearity) between two sets of modal vec-
/i for mode i. The denominator gives the modal frequency x and tors. The computation of MAC between dm modal vectors esti-
the complex pole ki, which is related to the modal frequency xi mated from a damaged state and m modal vectors estimated
and modal damping ri as follows: from the pristine state results in a dm  m MAC matrix. However,
in practice, it is usually difficult to interpret the MAC matrix for
ki ¼ ri þ jxi and ki ¼ ri  jxi ð3Þ
damage identification. Thus, PrMAC is proposed in this work as a
It should be noted that the FRF matrix component H(x) results from simpler damage indicator. PrMAC is a modification of MAC that at-
a summation of terms, each term due to the contribution of a single tempts to provide a measure of the degree of correlation between
mode of vibration. It is straightforward to show that because two sets of modal vectors by transforming the MAC matrix infor-
denominators are functions of modal frequencies, they cause peaks mation into a scalar value. It is defined as follows:
in an FRF when x = xi. Furthermore, the numerator is related to the d t 
Y
m  /  / 2
peak magnitude. Some FRF measurements (i.e., components of the PrMACðd U; UÞ ¼ i i
ð4Þ
d /t  d /i /ti  /i
FRF matrix) of an impacted and pristine laminated specimen are i¼1 i
shown in Fig. 1.
where dU and U are the two sets of m modal vectors and d/ and /
2.2. Damage indicators are estimated from the damaged and pristine states, respectively.
PrMAC is essentially the product of the diagonal terms of the
Impact damage was identified by studying four damage indica- MAC matrix. The PrMAC value is bounded between zero and one,
tors, which compare modal parameters from the damaged and where a value of one signifies perfect consistency and a low value
undamaged state. The analysis involved considering changes in indicates discordance between the sets of modal vectors. Therefore,
modal frequencies, changes in mode shapes and changes in mode PrMAC is also a possible damage indicator.
shape curvatures. The extent of the irregularity of mode shapes depends on the
location and extent of induced damage. Mode shapes contain local
2.2.1. Frequency shifts information, i.e., provide spatial information, which enables them
Modal frequency shifts have been extensively used in the liter- to be used as damage localisation indicators. Because the scalar va-
ature as a damage indicator for composites. It is generally known lue of the preceding damage indicator does not allow for damage
that delamination decreases stiffness and, in turn, decreases modal localisation, a DoF correlation criterion must be used for that pur-
frequencies (see Fig. 1). However, frequency shifts have significant pose. The previously published Enhanced Coordinate Modal Assur-
practical limitations because the low sensitivity of modal fre- ance Criterion (ECoMAC) is an extension of the Coordinate Modal
quency shifts to damage requires either very precise measure- Assurance Criterion (CoMAC) Allemang [34] that attempts to iden-
ments or large levels of damage. Moreover, modal frequencies tify what measured DoFs contribute negatively to a low value of
are a global property, and thus, they cannot provide spatial infor- MAC Piersol and Paez [33]. In this work, a modification of the ECo-
mation. As a result, their use is often limited to identifying only MAC is proposed by including a weighted factor in terms of a mod-
the presence of damage. Nonetheless, in the context of this work, al frequency squared ratio. It is defined as follows:
modal frequency shifts were examined due to their simplicity P x2i d 
cmwm d x2
 /p  /p 
i¼1
and were used to establish a damage detection threshold to com- d
WECoMACð U; U; X; XÞp ¼d i
i i
ð5Þ
pare the results obtained to those of previously published 2m
investigations. The so-called WECoMAC is computed for each (measurement)
DoF p over a set of mode pairs, damaged versus pristine, in which
2.2.2. Mode shape changes the two modal vectors d/ and / in each mode pair i represent the
It is known that delaminations cause irregularities in mode same mode shape. WECoMAC provides a spatial correlation from
shapes. These distortions depend on the size and location of the the weighted average of differences between the components of
damage. Thus, by comparing the similarity between the mode two modal vectors, which enables it to be used indirectly as a use-
shapes from the damaged and pristine states, impact damage can ful damage localisation and extent indicator. The criterion is for-
mulated under the assumption that the absolute differences
between the components of two modal vectors are the highest in
the region of damage and negligibly small outside this region.
1000
The weighted factor is included to minimise the masking effect
on results caused by the unaltered modal frequencies. Unlike the
previous criterion, a low value indicates a high DoF correlation be-
Log (m/s2)/N

tween the two states. It should be noted that in the formulation of

100
the above criteria it is assumed that there is a match for each mod-
al vector in the two sets (pristine and damaged) and that the modal
vectors are numbered accordingly. One must bear this fact in mind
10
whenever induced damage may cause the positions of the modal
Damaged frequencies and vectors to be swapped, which can be easily veri-
Pristine
fied using MAC.
1
0.0 2.0k 4.0k 6.0k 8.0k 10.0k 2.2.3. Curvature mode shape changes
Frequency (Hz) The abovementioned criteria for the identification of damage
are based directly on measured mode shapes. An alternative to
Fig. 1. Inertance FRFs plots comparison of a 50 J impacted and pristine laminated
specimen for range of 0 ± 10 kHz. This figure qualitatively highlights the shifts in using mode shapes to obtain spatial information is using mode
resonance frequencies. shape derivatives such as curvatures Doebling et al. [12], Sohn
270 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276

et al. [13], Carden [14], Fan and Qiao [16], Radzienski et al. [18] and inspection to assess the grade of compaction and to discard the pres-
Hamey et al. [35]. It is known that impact-induced damage affects ence of defects, porosities or delaminations.
the laminate flexural stiffness, which in turn affects the mode
shapes and its derivatives. As will be shown below, the experimen- 3.2. Experimental modal analysis
tal results demonstrate that curvature mode shapes changes are
localised in the region of impact damage, which also enables them To obtain the modal parameters to compute the damage indica-
to be used as a damage localisation and extent indicator. tors, modal testing of the pristine and damaged composite lami-
The local curvature values jqi for mode i and DoF q are nates was performed. Tests were performed under free boundary
computed from the displacement mode shape using the central conditions by suspending the coupons vertically. The vibration sig-
difference approximation as follows: nals (excitation and response) were measured and recorded in the
form of time series and processed into inertance FRF data. The iner-
/ðqþ1Þi  2/qi þ /ðq1Þi
jqi ð/i Þ ¼ /00qi  2
ð6Þ tance is defined as the ratio of the acceleration measured to the
l force applied. Vibration measurements were performed using a
where l is the length between the measured DoFs. It is worth noting single-reference roving hammer test. A mono-axial accelerometer4
that both subscript p in Eq. (5) and subscript q in Eq. (6) denote was attached to a single degree of freedom (DoF) reference point on
measured DoFs. However, whereas p represents the coordinate of the top surface of the laminate, whereas the miniature transducer
the modal vector, q + 1, q and q  1 represent the adjacent coordi- hammer5 roved around, exciting the specimen at 25 measuring DoFs,
nates in the defined direction over the discretised surface domain. 5 evenly distributed in the direction of the width and 5 evenly dis-
To compare the local curvature between the damaged and pris- tributed in the direction of the length (see Fig. 5). Because the delam-
tine states, a curvature damage factor (CDF) is used. The local ination mode can introduce non-linearities Aymerich and Staszewski
directional curvature damage factors are expressed as follows: [22], an effort was made to minimise them by using mini hammer
    transducer to create a low-level input excitation. In addition to the
Xm
x2i d jxqi  jxqi  Xm
x2i d jyqi  jyqi  transverse modes of vibration, in-plane modes were also estimated
Djxq ¼   and D jy
¼   ð7Þ
i¼1
d x2 
i
jxqi  q
i¼1
d x2 
i jyqi  by exciting the specimen at 5 evenly distributed locations on each
lateral side. Both the applied excitation and the measured response6
and the local CDF is defined as follows: were perpendicular to the coupon. Signals were averaged 3 times for
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2 each measurement point, and the test frequency band was up to
Djq ¼ Djxq þ Djyq ð8Þ 20 kHz with a resolution of 3.125 Hz.

The computation over a set of measured DoFs provides a curvature 3.3. Drop-weight impact test
damage vector whose coefficients describe the local variation of the
curvature on the discretised surface domain. An ASTM7 D7136 standard test method ASTM-D7136 [36] was
The procedure for estimating the modal parameters from the followed to determine the damage resistance of the laminated com-
measured FRFs and the abovementioned formulation for comput- posite specimens subjected to a drop-weight impact event. The test
ing the different damage indicators were implemented in a spe- procedure consisted in releasing a weight from a certain height,
cially written Matlab code. which determined the incident kinetic energy. A detailed description
of the test conducted can be found in reference Pérez et al. [32].
3. Experimental procedure A total of 44 coupons were impacted with incident energy (EI)
levels ranging from 6.6 to 70 J, in intervals of 5 J, with a minimum
The objective of this study was to quantify and measure the of 2 and maximum of 4 specimens designated for each impact
changes in the measured structural vibration response of damaged energy. The lower limit was determined by the minimum height
composite laminates due to low-velocity impact. The experimental condition imposed by the standard test method. The upper limit
programme involved three main stages: the experimental modal was defined by considering a hypothetical energy level below the
analysis performed before and after damaging the composite lam- penetration and perforation thresholds. The 4 remaining speci-
inates, the drop-weight impact test, in which damage is induced in mens were reserved to ensure the repeatability of modal testing
a controlled manner, and finally the compression after impact test and to measure the compressive strength of the pristine laminates.
to assess the residual bearing capacity of composite specimens. After the impact event, the interlaminar damage onset was esti-
mated and the extent of induced delaminations was measured by
3.1. Test specimens using ultrasonic phased array testing equipment.8
In the next section, the capability and suitability of each
The experimental programme was executed for a total of 48 damage indicator is compared in terms of the absorbed energy.
monolithic composite plates specimens measuring 150  100  The absorbed energy results from the sum of the contributions of
5.2 mm3. In the manufacturing process, commercially available uni- non-conservative forces and the energy dissipated due to the fail-
directional prepreg laminae composed of carbon fibres2 (volume- ure mechanisms. Experimental data regarding the absorbed energy
fraction of 55.2%) embedded in a resin epoxy matrix3 were used. The were estimated from the difference between the incident and
quasi-isotropic laminates plates were composed of 40 unidirectional rebound kinetic energy by measuring the respective velocities of
laminae with a balanced and symmetric stacking sequence [45°/0°/ the impactor, as described in Pérez et al. [32].
45°/90°]5S, resulting in a nominal thickness of approximately
5.2 mm with a uniform cross-section over the entire surface. The lam-
inated plate layup was defined such that the 0° fibre orientation was
aligned with the lengthwise dimension. Laminate curing was per-
formed following a standard autoclave procedure. After the curing
4
Accelerometer Bruel and Kjaer 4518–003.
process, each coupon was examined using non-destructive ultrasonic 5
Hammer Bruel and Kjaer 8204.
6
Data acquisition system Bruel and Kjaer 3050-B-6/0.
2 7
Grafil TR30S 3 K. American Society for Testing and Materials.
3 8
Resin epoxy HSC Epikote 4652. OmniScan MX with standard phased array probe 5 MHz linear array 64 elements.
 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276 271

Fig. 2. Scheme of the compressive residual strength support fixture with specimen in place (left). Coupon size and test boundary conditions (dimensions in mm) according to
ASTM D7137 standard test method (right).

3.4. Compression after impact testing 4. Results and discussion

The compression after impact (CAI) test determines the com- The feasibility of using vibration-based NDT for impact damage
pressive residual strength of flat rectangular composite plates identification in laminated composites is assessed by comparing
previously subjected to an impact event. The CAI test was per- the computed damage proposed indicators for each induced
formed using the support fixture sketched in Fig. 2, which meets damage level. Moreover, attention will be focussed on the relation
the requirements of the standard test method ASTM D7137 out- between the damage indicators and the laminate residual load-
lined in ASTM-D7137 [37]. The compressive test was carried bearing capacity.
out in a multi-piece support fixture, which utilises adjustable
guide plates with knife edges that provide no restraint to local 4.1. Frequency shifts
out-of-plane rotation, support the specimen edges and prevent
buckling when the coupon is end-loaded. The top and bottom Fig. 3 shows the percentage frequency changes as a function of
supports provide no clamp-up but provide some rotational re- the absorbed energy of the first 12 transverse modes and first 3 in-
straint due to the fixture geometry. The support fixture was care- plane modes. The minimum number of modes estimated for each
fully aligned to minimise loading eccentricities and induced test specimen limits the number of transverse modes analysed to
specimen bending. 12, whereas the test frequency band limits the number of in-plane
The coupon was aligned with the lengthwise dimension and modes because these modes are in the range of 13–18 kHz. Each
loaded with a cross-head displacement rate of 1.25 mm/min until mode is labelled according to its position in the frequency spec-
a maximum load was reached; the load was then reduced to a trum. The nomenclature (a, b) identifies each different mode type,
magnitude approximately 25% below the maximum. The com- where a and b denote the number of nodal lines parallel to the
pressive load was applied using a high-performance hydraulic crosswise direction and lengthwise direction, respectively. The
actuator.9 To ensure that the applied compressive load did not gen- graphs present indications that refer to the interlaminar damage
erate any bending moment on the plates, the strain distribution onset (IDO) and the so-called barely visible impact damage (BVID).
was monitored during the test. Three strain gauge10 pairs were The former is determined by the ultrasound inspection results. IDO
bonded back-to-back on each side of the coupon as shown in was estimated to be greater than 8.34 J of absorbed energy
Fig. 2. As indicated in the standard test method, a rapid divergence (EI = 15 J). The latter corresponds -according to industry standards-
of the strain measurements on opposite faces of the specimen or a to the impact energy level that produces an indentation depths of
rapid increase in percent bending is indicative of onset panel insta- 0.3 mm after relaxation. BVID was estimated to be greater than
bility, and consequently, plate buckling affects the results. In this 16.84 J of absorbed energy (EI=30 J).
study, the average of the two back-to-back strain measurements The data plotted represent mean values with standard devia-
was maintained below 5% Pérez et al. [38]. Force, cross-head dis- tion bars for each energy level. Overall, the results show that
placement and strain data were recorded11 at a sampling rate of the vibration response is correlated with the damage severity,
500 Hz. as other researchers have already proved. In spite of the devia-
CAI tests were performed on 4 pristine coupons as well as on a tions, the results show a clear decrease as the extent of induced
series of damaged specimens at various impact energy levels as de- damage increases, as expected, in agreement with the results re-
scribed above. It should be noted that the validity of the estimated ported in previous studies Kessler et al. [25], Frieden et al.
residual strength depends on the mode of failure observed, as dis- [27,31]. It is noted that not all modes are equally affected by
cussed in the next section. the same damage level. The most significant variations are ob-
served in the 4th, 8th, 9th, 10th and 11th transverse modes and
9
in the 23rd and 24th in-plane modes. These results clearly dem-
250 kN capacity MTS actuator model 244.31.
10
Strain gauge HBM 1-LY41-6/350.
onstrate that if damage is located in a region that develops a high
11
Data acquisition system HBM MGCplus. level of potential energy for a particular mode, it will have a
272 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276

1 IDO 1 IDO 1 IDO

0 0 0
-1 -1 -1

(%)
(%)

(%)
BVID BVID
-2 -2 -2 BVID
-3 -3 -3
-4 -4 -4
1st - Mode (1 1) 2nd - Mode (2 0) 3rd - Mode (2 1)
-5 -5 -5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

1 IDO 1 1 IDO
0 0 0
-1 -1 -1

(%)
(%)
(%)

BVID IDO BVID

-2 -2 -2 BVID
-3 -3 -3
-4 -4 -4
4th - Mode (0 2) 5th - Mode (3 0) 6th - Mode (1 2)
-5 -5 -5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

1 1 1 IDO
0 0 0
-1 IDO -1 IDO -1
(%)

(%)

(%)
BVID
-2 -2 BVID -2 BVID
-3 -3 -3
-4 -4 -4
7th - Mode (3 1) 8th - Mode (2 2) 9th - Mode (4 0)
-5 -5 -5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

1 IDO 1 BVID 1 IDO

0 0 0
-1 -1 IDO -1
(%)

(%)

(%)

-2 BVID -2 -2 BVID
-3 -3 -3
-4 -4 -4
10th - Mode (0 3) 11th - Mode (3 2) 12th - Mode (1 3)
-5 -5 -5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
2 2
IDO 2 IDO IDO
1 1
1
0 0
0
-1 -1 -1
-2
(%)

-2
(%)

(%)

BVID -2
-3 BVID -3 BVID
-3
-4 -4 -4
-5 -5 -5
-6 -6 -6
-7 17th - Mode in-plane bending -7 23rd - Mode in-plane shear -7 24th - Mode in-plane compression
-8 -8 -8
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Absorbed energy (J) Absorbed energy (J) Absorbed energy (J)

Fig. 3. Percentage frequency changes as a function of the absorbed energy of the first 12 transverse modes and first 3 in-plane modes. IDO refers to the interlaminar damage
onset and BVID to the so-called barely visible impact damage.

strong effect on the mode’s natural frequency. The converse is extent of damage higher than the BVID and IDO levels. It can also
also true when damage is located on a nodal line. Test repeatabil- be observed that the differences are larger for higher modal fre-
ity was assessed by comparing the results of pristine specimens quencies, where the modes are associated with local responses.
at two different times, showing that the error was lower than Nonetheless, the maximum change is approximately 5% for trans-
approximately ±0.5%. versal modes and 7% for in-plane modes. This difference is clearly
As a first estimation, the magnitude of changes does not begin higher than the frequency resolution of the experimental test
to be significant until 22.28 J of absorbed energy (EI = 35 J), an (lower than approximately ±0.23%). For intermediate damage lev-
 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276 273

els, the differences may be obscured by the changes caused by 26.26 J of absorbed energy (EI = 40 J), a damage level clearly higher
environmental and/or test conditions. than the IDO and BVID level. Below this level, the damage location
cannot be discerned.
4.2. Mode shape changes
4.3. Curvature mode shape changes
To quantify the mode shape changes, PrMAC was first computed
for the set of specimens by comparing the modal vectors estimated Fig. 6 shows the average CDF coefficients over the discretised
from damaged and pristine states. Fig. 4 shows the PrMAC coeffi- surface domain for each level of incident and absorbed energy.
cients corresponding to the transverse mode correlation as a func- The CDF displays a large peak along the impact location, and most
tion of the absorbed energy. The data plotted represent the mean notable variations are observed in the region of maximum degra-
values with standard deviation bars for each energy level. Despite dation, because the bending stiffness at the vicinity of the damage
the dispersion in the results, the PrMAC coefficients show a clear is reduced, thus increasing the magnitude of the mode curvatures,
decreasing trend with an increasing amount of induced damage, as expected. A growing effect of local curvature is also observed as
with a marked decrease of approximately 11.1 J, a value higher impact energy increases. From the analysis of the results, the
than the IDO level but lower than the BVID level. The dashed area detection and localisation thresholds were set to over 14.26 J of ab-
denotes the detection range of the frequency shifts, illustrating the sorbed energy (EI = 25 J), a level that is higher than the IDO level
difference in the identification thresholds between the two dam- but lower than the BVID level.
age indicators. It is remarkable that in the initial region below The capability and sensitivity of the different damage indicators
the IDO level -corresponding to the pristine and nearly pristine were first compared in terms of the absorbed energy and the pro-
states- the PrMAC values differ from unity. These results are due jected delaminated area. Fig. 7 shows a comparison between the
to the inherent experimental error. This criterion presents an identification thresholds of each damage indicator and the experi-
important limitation regarding the number of modes over which mental prediction of the projected delaminated area measured
PRMAC is computed: for a large number of modes, the criterion using C-scan ultrasonic images, as a function of the absorbed en-
is invalid because coefficients may decrease to nearly zero; in con- ergy. The experimental data plotted represent the mean values
trast, the results become inconclusive when using a small number with standard deviation bars for each energy level. The filled areas
of modes. Therefore, the minimum set of mode pairs was estab- correspond to the range of identification of each damage indicator
lished to be 6, and the corresponding results are plotted in Fig. 4. according to the results presented above. An in-depth comparison
The above results prove that the criteria considered are poten- between the capability and sensitivity of the different damage
tial indicators of the presence and severity of induced impact dam- indicators and their correlation with the residual strength is fur-
age. Damage was quantified based on a comparison between ther provided at the end of this section.
adjacent energy levels, thus requiring a pattern or a context for
data comparison. 4.4. Residual load-bearing capacity
As detailed previously, the computation of the WECoMAC over a
set of measured DoFs provides a spatial correlation of the vibration Based on the results discussed up to this point, it is feasible to
response. In Fig. 5, the averaged coefficients of WECoMAC are rep- address the identification of damage at the first three defined lev-
resented for each level of incident and absorbed energy. In the els: detection, localisation and quantification. The fourth level, cor-
computation, a set of 12 transverse mode pairs was considered. responding to the prediction of the remaining performance, is
The sequence of the measured DoFs corresponds to the data acqui- highly important as well. Thus, the results obtained regarding com-
sition order, according to the top-left sketch in Fig. 5. The results pression after impact testing are compared with the results
are show to be correlated with damage severity, particularly the obtained for the damage indicators assessed.
coefficients of the measured DoFs located in the vicinity of the im- Fig. 8 collects the results of the superposition of the identifica-
pact point (location 13). Closer examination reveals the abnormal tion thresholds of each damage indicator computed from modal
behaviour of DoF 25, which is attributable to the effect of the accel- parameters, which were estimated for each coupon in its pristine
erometer, particularly, to the influence of the stiffness of the wire and damaged state; the results of the estimation of the interlami-
connection; thus, the use of a non-contact accelerometer such as nar damage onset determined by ultrasonic inspection; as well as
a laser Doppler vibrometer is preferable. From the analysis of the the results of the compressive residual strength of the composite
results, the detection and localisation thresholds were set to over plates specimens, which are directly related to the absorbed
energy. The CAI data plotted represent the mean values with
1.0 standard deviation bars for each energy level. The general trend
0.9 indicates that the compressive residual strength notably decreases
as the impact energy increases, as expected, in good agreement
0.8
IDO with data reported the literature. The strength results show an
0.7
abrupt decrease after the delamination onset, estimated to be
0.6 greater than 8.34 J of absorbed energy. CAI tests were performed
PrMAC

0.5 on impacted coupons as well as on unimpacted coupons. However,

0.4 it should be noted that the results obtained for the pristine and
0.3 nearly pristine (i.e., EA < 5J) specimens must be discarded because
BVID the failure modes observed when testing these particular speci-
0.2
mens were undesirable according to the standard test method
0.1
ASTM-D7137 [37]. Fig. 9 presents the acceptable and unacceptable
0.0
compressive residual strength failure modes observed for different
0 10 20 30 40 50 60 70 levels of impact energy. Unacceptable failure modes are those re-
Absorbed energy (J) lated to load introduction by the support fixture, local edge support
Fig. 4. Average PrMAC coefficients corresponding to the transverse modes corre-
conditions or specimen instability (Fig. 9(a, c and e)). During the
lation, as a function of the absorbed energy. Dashed area denotes the detection tests, the back-to-back strain measurements remained linear and
range from frequency shifts. did not indicate panel instability. In accordance with the standard
274 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276

1 2 0.3 0.3

WECoMAC
WECoMAC
6 3
11 7 4 EI = 6.6 J EI = 10 J
12 8 0.2 0.2
16 13 9 10 5 EA = 2.71 J EA = 5.45 J
21 17 14
22 18 15 0.1 0.1
23 19
24 20
25 0.0 0.0
5 10 15 20 25 5 10 15 20 25
0.3 0.3 0.3

WECoMAC

WECoMAC
WECoMAC

IDO EI = 15 J EI = 20 J EI = 25 J
0.2 0.2 0.2
EA = 8.34 J EA = 11.11 J EA = 14.26 J
0.1 0.1 0.1

0.0 0.0 0.0

5 10 15 20 25 5 10 15 20 25 5 10 15 20 25
0.3 0.3 0.3
WECoMAC

WECoMAC

WECoMAC
BVID EI = 30 J EI = 35 J EI = 40 J
0.2 0.2 0.2
EA = 16.84 J EA = 22.28 J EA = 26.26 J
0.1 0.1 0.1

0.0 0.0 0.0

5 10 15 20 25 5 10 15 20 25 5 10 15 20 25
0.3 0.3 0.3

WECoMAC
WECoMAC
WECoMAC

EI = 45 J EI = 50 J EI = 55 J
0.2 0.2 0.2
EA = 31.33 J EA = 36.14 J EA = 40.56 J
0.1 0.1 0.1

0.0 0.0 0.0

5 10 15 20 25 5 10 15 20 25 5 10 15 20 25
0.3 0.3 0.3
WECoMAC

WECoMAC

WECoMAC
EI = 60 J EI = 65 J EI = 70 J
0.2 0.2 0.2
EA = 51.91 J EA = 58.05 J EA = 64.83 J
0.1 0.1 0.1

0.0 0.0 0.0

5 10 15 20 25 5 10 15 20 25 5 10 15 20 25
Measured DoFs Measured DoFs Measured DoFs

Fig. 5. Average WECoMAC coefficients computed over a set of 25 measured DoFs for each level of incident (EI) and absorbed (EA) energy. The top left sketch represents the
DoFs data acquisition order and its distribution over the surface of the test specimen.

EI = 6.6 J EI = 10 J EI = 15 J IDO
EA = 2.71 J EA = 5.45 J EA = 8.34 J

EI = 20 J EI = 25 J EI = 30 J BVID
EA = 11.11 J EA = 14.26 J EA = 16.84 J

EI = 35 J EI = 40 J EI = 45 J
EA = 22.28 J EA = 26.26 J EA = 31.33 J

EI = 50 J EI = 55 J EI = 60 J
EA = 36.14 J EA = 40.56 J EA = 51.91 J

EI = 65 J EI = 70 J
EA = 58.05 J EA = 64.83 J

Fig. 6. Average curvature damage factor (CDF) coefficients over the discretised surface domain for each level of incident (EI) and absorbed (EA) energy.
 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276 275

1.0 test method, valid strength results were only reported when
1500
appropriate failure modes were observed, such as the formation
of kink bands due to microbuckling instability (Fig. 9(d, f and g)).
Projected delaminated area (mm2)

BVID PrMAC
1250 0.8 Therefore, in the present work, the measurement of the material
CDF compressive strength of undamaged specimens has not been feasi-
1000 ble due to the high incidence of undesirable failure modes. The
Frequency shifts 0.6 maximum strength value was obtained for coupons impacted at

PrMAC
10 J (EA = 5.45 J), whose energy level is below the IDO level.
750 WECoMAC
The graph shown in Fig. 8 provides experimental evidence of
0.4
the capability and suitability of the different damage indicators as-
500 sessed for impact damage identification regarding the residual
IDO load-bearing capacity of the composite laminates examined. The
0.2
250
WECoMAC criterion provides spatial correlation, which ideally
enables it to be used as an indicator of damage localisation and
extent. However, studies have proved that high levels of damage
0.0
0 are needed for the criterion to be a useful indicator. It has been ob-
0 10 20 30 40 50 60 70
served that a weighted factor in terms of a modal frequency
Absorbed energy (J) squared ratio reduces the detection threshold with respect to the
Fig. 7. Comparison between the identification thresholds of each damage indicator previously published ECoMAC criterion; nevertheless, the capabil-
and the experimental prediction of the projected delaminated area as a function of ity of the WECoMAC criterion for identification is far from the
the absorbed energy. delamination threshold.
Damage identification from frequency shifts has the appealing
450 1.0 feature that it is relatively simple to measure, because it can be
Interlaminar damage estimated from one location. Nonetheless, identification is limited
400 to the detection of the presence of damage, and results have
PrMAC 0.8 proved that high levels of degradation are also needed to detect
350 damage with confidence, as concluded by other authors.
CDF
CAI Strength (MPa)

In spite of the inaccuracy of the estimated experimental mode

300
Frequency shifts 0.6 shapes, several damage indicators based on these mode shapes
PrMAC

250 have shown promising results. CDF is shown to be the most suit-
WECoMAC able for localisation, even if the damage level is below the BVID le-
200 0.4 vel. As mentioned previously, it has been observed that a weighted
BVID factor in terms of a modal frequency squared ratio notably mini-
150
mises the masking effect on results caused by the unaltered modes.
100 0.2 However, with respect to the identification threshold, the residual
strength exhibits a decrease from the maximum value of approxi-
50 mately 32%.
0.0 Finally, the PrMAC criterion is clearly shown to be the most
0 capable and suitable among the different damage indicators as-
0 10 20 30 40 50 60 70
Absorbed energy (J) sessed, providing an identification threshold close to the IDO level.
However, as was the case above, at this level, the residual strength
Fig. 8. Comparison between the identification thresholds of each damage indicator,
exhibits a decrease of approximately 27%.
the interlaminar damage onset and the average compressive residual strength as a
function of the absorbed energy.

EI = 6.6 J EI = 15 J EI = 20 J EI = 25 J EI = 65 J
Pristine Pristine
EA= 2.71 J EA = 8.34 J EA = 11.11 J EA = 14.26 J EA = 58.05 J

(a) (b) (c) (d) (e) (f) (g)

Fig. 9. Acceptable (d, f and g) and unacceptable (a, b, c and e) compressive residual strength failure modes observed for different levels of incident (EI) and absorbed (EA)
energies. (a, c and e) Failure by compression-shear; (b) failure related to load introduction by the support fixture and local edge support conditions; (d, f and g) kink-band
formation due to fibre instability (microbuckling).
276 M.A. Pérez et al. / Composite Structures 108 (2014) 267–276

5. Conclusions [8] De Angelis G, Meo M, Almond DP, Pickering SG, Angioni SL. A new technique to
detect defect size and depth in composite structures using digital
shearography and unconstrained optimization. NDT E Int 2012;45:91–6.
The results of an extensive experimental campaign conducted on [9] Goidescu C, Welemane H, Garnier C, Fazzini M, Brault R, Péronnet E, et al.
a set of composite laminates specimens were presented and Damage investigation in CFRP composites using full-field measurement
techniques: Combination of digital image stereo-correlation, infrared
discussed, including the evaluation of impact damage resistance,
thermography and X-ray tomography. Compos B Eng 2013;48:95–105.
the characterisation of induced damage, the quantification of the ef- [10] Bull D, Helfen L, Sinclair I, Spearing S, Baumbach T. A comparison of multi-
fects on the vibration response and the evaluation of impact damage scale 3D X-ray tomographic inspection techniques for assessing carbon fibre
composite impact damage. Compos Sci Technol 2013;75:55–61.
tolerance. Also, four damage indicators based on modal parameters
[11] Garnier C, Pastor M-L, Eyma F, Lorrain B. The detection of aeronautical defects
were assessed. The characteristics examined include the indicators’ in situ on composite structures using Non Destructive Testing. Compos Struct
ability to detect impact-induced damage, their precision in deter- 2011;93:1328–36.
mining the location of damage, their sensitivity regarding the extent [12] Doebling SW, Farrar CR, Prime MB, Shevitz DW. Damage identification and
health monitoring of structural and mechanical systems from changes in their
of damage and correlations with residual bearing capacity. vibration characteristics: a literature review. Los Alamos National Laboratory
Experimental evidence indicates that impact-induced damages report, (LA-13070-MS); 1996.
result in detectable changes in the vibration response of coupons. [13] Sohn H, Farrar CR, Hemez FM, Shunk DD, Stinemates DW, Nadler BR. Others, a
review of structural health monitoring literature: 1996-2001. Los Alamos
Overall, the relative changes in mode shapes are greater than those National Laboratory report, (LA-13976-MS); 2004.
observed for natural frequencies. The use of the proposed damage [14] Carden EP. Vibration based condition monitoring: a review. Struct Health
indicators shows promising results because it provides an identifi- Monit 2004;3:355–77.
[15] Yan Y, Cheng L, Wu Z, Yam L. Development in vibration-based structural
cation threshold close to the IDO level and below the BVID level. damage detection technique. Mech Syst Signal Process 2007;21:2198–211.
However, the earliest identification corresponds to a reduction in [16] Fan W, Qiao P. Vibration-based damage identification methods: a review and
strength of approximately 27%. comparative study. Struct Health Monit 2011;10:83–111.
[17] Kim J-T, Ryu Y-S, Cho H-M, Stubbs N. Damage identification in beam-type
Additionally, one must be aware that it is difficult to draw any
structures: frequency-based method vs mode-shape-based method. Eng Struct
general conclusion in relation to the indicators’ applicability to gen- 2003;25:57–67.
eral composite structures for at least two reasons. First, the number [18] Radzieński M, Krawczuk M, Palacz M. Improvement of damage detection
methods based on experimental modal parameters. Mech Syst Signal Process
of measured DoFs and estimated modes determines the capability
2011;25:2169–90.
of damage indicators. Hence, it is important to further investigate [19] Zou Y, Tong L, Steven GP. Vibration-based model-dependent damage
the effect of the number of measured DoFs and modes on the dam- (delamination) identification and health monitoring for composite
age identification results. Next, on the specimens tested, the extent structures: a review. J Sound Vib 2000;230:357–78.
[20] Della CN, Shu D. Vibration of delaminated composite laminates: a review. Appl
of induced damage was significant relative to the coupon surface Mech Rev 2007;60:1–20.
area. On a large structure, changes in the vibration response might [21] Polimeno U, Meo M. Detecting barely visible impact damage detection on
go undetected. Therefore, further research should focus on the aircraft composites structures. Compos Struct 2009;91:398–402.
[22] Aymerich F, Staszewski WJ. Impact damage detection in composite laminates
assessment of the proposed indicators for large structures. using nonlinear acoustics. Compos Part A Appl Sci Manuf 2010;41:1084–92.
Subject to these limitations, these findings will contribute to the [23] Ooijevaar T, Loendersloot R, Warnet L, de Boer A, Akkerman R. Vibration based
development of new vibration-based NDT techniques, that address structural health monitoring of a composite T-beam. Compos Struct
2010;92:2007–15.
the quantification of damage and estimation of the residual bear- [24] Wei Z, Yam LH, Cheng L. Detection of internal delamination in multi-layer
ing capacity of composite structures. composites using wavelet packets combined with modal parameter analysis.
Compos Struct 2004;64:377–87.
[25] Kessler SS, Spearing SM, Atalla MJ, Cesnik CE, Soutis C. Damage detection in
Acknowledgements
composite materials using frequency response methods. Compos B Eng
2002;33:87–95.
This work has been supported by the Spanish Ministry of Sci- [26] Yam LH. Nondestructive detection of internal delamination by vibration-based
method for composite plates. J Compos Mater 2004;38:2183–98.
ence and Innovation through Grant No. UNPC08-4E-007, RECOMP
[27] Frieden J, Cugnoni J, Botsis J, Gmür T. Vibration-based characterization of
Project, Ref. BIA2005-06952 and DELCOM Project Ref. MAT2008- impact induced delamination in composite plates using embedded FBG
02232/MAT., by the Spanish Ministry of Public Works (Project sensors and numerical modelling. Compos B Eng 2011;42:607–13.
‘‘Retrofitting and reinforcement of reinforced concrete structures [28] Fan W, Qiao P. A 2-D continuous wavelet transform of mode shape data for
damage detection of plate structures. Int J Solids Struct 2009;46:4379–95.
with composite materials. Numerical and experimental develop- [29] Niemann H, Morlier J, Shahdin A, Gourinat Y. Damage localization using
ments applied to joint of bars and composites anchorage pro- experimental modal parameters and topology optimization. Mech Syst Signal
posal’’), by AIRBUS (Spain) through the Contract No. PBSO-13-06 Process 2010;24:636–52.
[30] Araújo dos Santos JV, Mota Soares CM, Mota Soares CA, Maia NMM. Structural
FEMCOM and ACCIONA Infraestructuras (Spain) through the Pro- damage identification in laminated structures using FRF data. Compos Struct
ject SPHERA. All these support are gratefully acknowledged. Spe- 2005;67:239–49.
cial thanks also to Roger Serra for useful discussions. [31] Frieden J, Cugnoni J, Botsis J, Gmür T. Low energy impact damage monitoring
of composites using dynamic strain signals from FBG sensors – part II: damage
identification. Compos Struct 2012;94:593–600.
References [32] Pérez MA, Martínez X, Oller S, Gil L, Rastellini F, Flores F. Impact damage
prediction in carbon fiber-reinforced laminated composite using the matrix-
[1] Reid SR, Zho G. Impact behaviour of fibre-reinforced composite materials and reinforced mixing theory. Compos Struct 2013;104:239–48.
structures. CRC Press; 2000. [33] Piersol A, Paez T. Harris’ shock and vibration handbook. McGraw-Hill
[2] Hodgkinson JM. Mechanical testing of advanced fibre composites. CRC Press; Professional; 2009.
2000. [34] Allemang RJ. The modal assurance criterion: twenty years of use and abuse.
[3] Aymerich F, Meili S. Ultrasonic evaluation of matrix damage in impacted Sound Vib 2003;37:14–23.
composite laminates. Compos B Eng 2000;31:1–6. [35] Hamey CS, Lestari W, Qiao P, Song G. Experimental damage identification of
[4] Schilling PJ, Karedla BR, Tatiparthi AK, Verges MA, Herrington PD. X-ray carbon/epoxy composite beams using curvature mode shapes. Struct Health
computed microtomography of internal damage in fiber reinforced polymer Monit 2004;3:333–53.
matrix composites. Compos Sci Technol 2005;65:2071–8. [36] ASTM-D7136. Standard test method for measuring the damage resistance of a
[5] Růžek R, Lohonka R, Jironč J. Ultrasonic C-Scan and shearography NDI fibre-reinforced polymer matrix composite to a drop-weight impact event,
techniques evaluation of impact defects identification. NDT E Int Tech. rep. West Conshohocken (PA, USA): American Society for Testing and
2006;39:132–42. Materials (ASTM); 2007.
[6] De Rosa IM, Santulli C, Sarasini F. Acoustic emission for monitoring the [37] ASTM-D7137. Standard test method for compressive residual strength properties
mechanical behaviour of natural fibre composites: A literature review. Compos of damaged polymer matrix composite plates, Tech. rep. West Conshohocken
Part A Appl Sci Manuf 2009;40:1456–69. (PA, USA): American Society for Testing and Materials (ASTM); 2007.
[7] Hung Y, Chen Y, Ng S, Liu L, Huang Y, Luk B, et al. Review and comparison of [38] Pérez MA, Gil L, Oller S. Evaluación del daño por impacto en laminados de
shearography and active thermography for nondestructive evaluation. Mater material compuesto mediante la respuesta dinámica. CIMNE; 2012 [in
Sci Eng R Rep 2009;64:73–112. Spanish].