International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

International Journal of Adhesion & Adhesives

journal homepage: www.elsevier.com/locate/ijadhadh

Adherend thickness effect on the tensile fracture toughness of a

structural adhesive using an optical data acquisition method
R.D.S.G. Campilho a,b,n, D.C. Moura c,d, M.D. Banea c, L.F.M. da Silva c
a
Departamento de Engenharia Mecânica, Instituto Superior de Engenharia do Porto, Instituto Politécnico do Porto, Rua Dr. António Bernardino de Almeida,
431, 4200-072 Porto, Portugal
b
Faculdade de Ciências Económicas, Sociais e da Empresa, Universidade Lusófona do Porto, Rua Augusto Rosa, 24, 4000-098 Porto, Portugal
c
Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugal
d
INEGI—Instituto de Engenharia Mecânica e Gestão Industrial, Rua Dr. Roberto Frias, 400, 4200-465 Porto, Portugal

art ic l e i nf o a b s t r a c t

Adhesive bonding is nowadays a serious candidate to replace methods such as fastening or riveting,
Keywords: because of attractive mechanical properties. As a result, adhesives are being increasingly used in
Aluminium and alloys industries such as the automotive, aerospace and construction. Thus, it is highly important to predict the
Experimental testing strength of bonded joints to assess the feasibility of joining during the fabrication process of components
Joint design (e.g. due to complex geometries) or for repairing purposes. This work studies the tensile behaviour of
Fracture adhesive joints between aluminium adherends considering different values of adherend thickness (h)
and the double-cantilever beam (DCB) test. The experimental work consists of the definition of the
tensile fracture toughness (GIC) for the different joint configurations. A conventional fracture character-
ization method was used, together with a J-integral approach, that take into account the plasticity effects
occurring in the adhesive layer. An optical measurement method is used for the evaluation of crack tip
opening and adherends rotation at the crack tip during the test, supported by a Matlabs sub-routine for
the automated extraction of these quantities. As output of this work, a comparative evaluation between
bonded systems with different values of adherend thickness is carried out and complete fracture data is
provided in tension for the subsequent strength prediction of joints with identical conditions.
& 2014 Elsevier Ltd. All rights reserved.

1. Introduction assessing the feasibility of joining during the fabrication process of

components (e.g. due to complex geometries) or joining as a repair
In order to increase the efficiency of structures and reduce their method. A large number of predictive techniques are currently
weight, adhesively-bonded joints can be used over fastening, available, ranging from analytical to numerical, using different
riveting or other traditional joining methods. Different joint criteria to infer the onset of material degradation, damage or even
configurations can be applied in structures, such as single-lap, complete failure. Initially, the prediction was performed by theo-
double-lap, stepped and scarf. Many other configurations exist, retical studies as those of Volkersen [4] or Goland and Reissner [5],
with specific advantages, like the joggle-lap joints, applied for which had a lot of embedded simplifying assumptions, by compar-
instance to join fuselage parts in aircraft, reinforcement doublers, ing current stresses with the allowable material strengths. Many
peel joints and L-section joints. Adhesives are progressively being improvements were then introduced, such as the assumption of
applied in structures of several branches of engineering, which elasto-perfectly plastic adhesive proposed by Hart-Smith [6], but
also makes certification-related issues more important, especially these analyses usually suffered from the non-consideration of the
in industries such as aeronautics [1]. Because of this, it is highly material ductility, which is highly relevant because of stress
important the availability of robust predictive techniques that can gradients. Fracture mechanics-based methods took the fracture
reliably be used as design tools, to allow minimization of experi- toughness of materials as the leading parameter for material
mentation during the design stages of products and structures selection. These methods included more simple energetic or
without compromising the design process [2,3]. This will allow stress-intensity fracture techniques that required the existence of
an initial flaw in the materials [7]. More recent numerical
n
techniques, such as cohesive zone models (CZM), combine stress
Corresponding author at: Departamento de Engenharia Mecânica, Instituto
Superior de Engenharia do Porto, Instituto Politécnico do Porto, 431, 4200-072
criteria to account for damage initiation with energetic, e.g.
Porto, Portugal. Tel.: þ 351939526892; fax: þ 351228321159. fracture toughness, data to estimate damage propagation [8]. This
E-mail address: raulcampilho@gmail.com (R.D.S.G. Campilho). allows to consider the distinct ductility of adhesives and to gain

0143-7496/$ - see front matter & 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
2 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎

accuracy in the predictions. All of these fracture toughness- peeling arms. Internal constraint effects are tA and specimen width
dependent analyses rely on an accurate measurement of GIC and (B), which also influence the size of the yield zone and stress and
GIIC (shear toughness). CZM in particular can accurately predict strain distributions in the adhesive. As an example, the increase
damage growth in structures if the fracture laws are correctly of tA can change the yielding conditions from small-scale to full
estimated [9]. These fracture laws are based on the values of plastic [25]. In the work of Giannis et al. [26], two different
cohesive strength in tension and shear, tn0 and ts0, respectively, sealants for aircraft fuel tanks were tested by a modified peel test,
and also GIC and GIIC [10]. These parameters that cannot be directly and it was shown for both adhesives that tA largely increased
related with the material properties measured as bulk, since they the peel resistance between 2 and 4 mm. Because of these issues,
account for constraint effects (in the case of adhesive joints, the taking into account of these constraints requires a precise simula-
constraints are caused by the adherends). Although these para- tion of the states of deformation of both the adhesive and
meters do not have a clear physical significance, they are able to surrounding structure. One of the methods that allow this to be
accurately reproduce the behaviour of the materials in a macro modelled is CZM, by considering the adhesive layer modelled by
scale point of view that is quite accurate [2]. The estimation of cohesive elements [27].
these fracture parameters is generally accomplished by perform- Most of the published work addressing the influence of h
ing pure tension or shear tests. Regarding GIC, the DCB test is examines directly its influence on the failure loads, either static
the most suitable, due to the test simplicity and accuracy [11]. The [28] or fatigue [29] and, in general, increasing h has shown to
typical GIC estimation methods are based on linear-elastic fracture improve the strength. Some studies associated this behaviour to
mechanics (LEFM) and require the continuous measurement of the the reduction of peel and shear peak stress at the overlap edges of
crack length (a) during the test. However, GIC of adhesives with bonded structures [30]. This actually occurs, but other phenomena
large scale plasticity is not accurately characterized with LEFM are also on the basis of these differences in strength. Actually, a
methods since the assumed stress fields at the crack tip vicinity few studied showed variations of GIC by modification of the
are not accurate [12]. More recently, methods that do not require structures thickness. In the work of Mangalgiri et al. [31], sym-
the measurement of a were developed, based on equivalent cracks metric and unsymmetric DCB specimens were experimentally
and including the plasticity effects around the crack tip [13]. As it tested with different values of h (by considering 8, 16 or 24 plies
was described by Suo et al. [14], in the presence of large-scale of carbon–fibre adherends). The static tests showed a large
plasticity, J-integral solutions can also be employed for accurate improvement of GIC between composites with 8 and 16 plies.
results. The J-integral is a relatively straight-forward technique, Devitt et al. [32] equally used the DCB test to investigate this effect,
provided that the analytical solution for a given test specimen and found a 9% increase in the value of GIC of bonded joints made
exists for the determination of GIC or GIIC. The most prominent of glass-epoxy composites by duplicating the number of plies of
example is the DCB specimen, for which J-integral solutions are the adherends. From these studies, it is clear that the differences
available, either for loading by pure bending moments [15] or the take place at relatively low h values. Since most bonded joints are
standardized and moment-free tensile loading [16]. It is also made between thin adherends/sheets, the understanding of how h
possible to estimate the tensile CZM law. Carlberger and Stigh affects the fracture toughness is highly relevant.
[17] computed the CZM laws of adhesive layers in tension and In this work, the value of GIC of adhesive joints between
shear using the DCB and end-notched flexure (ENF) tests, respec- aluminium adherends is studied, considering different values of
tively, considering 0.1 rtA r1.6 mm (tA is the adhesive thickness). h and the DCB test. A conventional fracture characterization
The rotation of the adherends was measured by an incremental method was used, in comparison with a J-integral approach, that
shaft encoder and the crack tip opening by two linear variable take into account for the plasticity effects occurring in the
differential transducers (LVDT). Ji et al. [18] studied the influence adhesive layer. An optical measurement method is used for the
of tA in DCB joints on tn0 and GIC for a brittle epoxy adhesive. evaluation of crack tip opening and adherends rotation at the
GIC was measured by a direct technique. For the measurement of crack tip during the test, supported by a Matlabs sub-routine for
the adherends rotation, two digital inclinometers with a 0.011 the automated extraction of these quantities.
precision were attached at the free end of each adherend. The
normal displacement at the crack tip was measured by a charge-
coupled device (CCD) camera.
Regarding the fracture parameters of adhesives, different
studies showed that these are similar as bulk and in a joint for 2. Experimental work
brittle adhesives, since the yield zone ahead the crack tip is
practically nonexistent [19]. Contrarily, when speaking about 2.1. Characterization of the materials
moderately-to-highly ductile adhesives, the fracture parameters
are not invariant to the joint geometry [20,21]. This occurs because The adherends were cut from a high strength aluminium alloy
of the different degree of restriction to the development of this sheet (AA6082 T651) by precision disc cutting. This material was
yield zone within the adhesive layer and also premature adhesive characterized in bulk tension in previous works by the authors
fracture because of excessive adherend yielding near the adhesive [33,34] using dogbone specimens and the following mechanical
layer. This brings issues about the transfer of small specimen (test) properties were obtained: Young0 s modulus (E) of 70.07 7
results to real-life and complex structures. Actually, the majority of 0.83 GPa, tensile yield stress (sy) of 261.67 77.65 MPa, tensile
industrial structures bonded by adhesives consist of thin sheets failure strength (sf) of 32470.16 MPa and tensile failure strain
between 1 and 3 mm thick, whilst the fracture behaviour of (εf) of 21.7074.24%. The two-component polyurethane adhesive
adhesives is mostly characterized within considerably thicker SikaForces 7888, selected for this work, was formerly tested in the
joints [22]. Pardoen et al. [23] discussed the two types of work of Neto et al. [35]. The bulk specimens were tested in a
constraint effects that affect the fracture toughness of adhesive servo-hydraulic machine to obtain E, sf and εf. The DCB test was
layers in bonded assemblies: external and internal constraint selected to obtain GIC and the ENF test was used for GIIC. The
effects. External effects deal with the varying states of deformation collected data of the adhesive is summarized in Table 1. To be
of the adherends, which alter the stress and strain distributions of noted that the obtained values of GIC and GIIC were obtained in
the adhesive layer. For instance, Wang et al. [24] found that GIC of partial adhesive failure conditions, which is important to notice
adhesives measured in peel tests varies upon the thickness of the when comparing GIC with the results obtained in this work.

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

Table 1 it gives
Properties of the adhesive SikaForces 7888 [35]. !
6P 2 2aeq 2 1
Property
GIC ¼ þ : ð1Þ
B2 h h Ef
2 5G
Young0 s modulus, E [GPa] 1.89 7 0.81 While detailed explanations are presented elsewhere [13], a
Tensile yield strength, sy [MPa] 13.20 7 4.83
Tensile failure strength, sf [MPa] 28.60 72.0
brief explanation of the parameters is given: aeq is an equivalent
Tensile failure strain, εf [%] 43.0 70.6 crack length estimated from the current specimen compliance
Critical energy release rate in tension, GIC [N/mm] 0.7023 and taking into consideration the damage zone, Ef is a corrected
Critical energy release rate in shear, GIIC [N/mm] 8.72 flexural modulus to account for phenomena affecting the P–δ
curve, such as stress concentrations at the crack tip and stiffness
variability between specimens, and G is the shear modulus of the
adherends [37].

3.2. Direct method to define GIC and the CZM law

The method followed in this work allows obtaining the cohe-

sive law of the adhesive by the simultaneous measurement of the
J-integral and δn [16]. The J-integral applies to the non-linear
elastic behaviour of materials, but it remains valid in the presence
of a plastic but monotonically-applied loading, as it is the case of
Fig. 1. Geometry of the DCB specimens.
the cohesive separation and plastic dissipation in adhesive layers
[18]. Based on the fundamental expression for J defined by Rice
2.2. Joint geometries
[38], it is possible to present a closed-form solution for the tensile
energy release rate (GI) from the concept of energetic force and
Fig. 1 represents the geometry of the DCB specimens. The
also the beam theory for the DCB specimen, as [39]:
dimensions of the specimens are: total length L ¼160 mm, initial
crack length a0 E40 mm, h¼ 1, 2, 3 or 4 mm, B¼ 25 mm and ðP u aÞ2
GI ¼ 12 þ P u θo ð2Þ
tA ¼ 1 mm. The bonding process consisted in grit blasting with 3
Ea h
corundum sand, debris cleaning with acetone and assembly in a
steel mould for the correct alignment between the adherends. or
To guarantee a uniform tA value, calibrated spacers were inserted GI ¼ P u θ p ; ð3Þ
between the adherends. The required sharp pre-crack was assured
using a 0.1 mm thick razor blade between calibrated bars at the where Pu represents the applied load per unit width at the
adherends edges, Ea the Young0 s modulus of the adherends and
crack tip. After applying the adhesive and assembling the speci-
mens, these were given a full cure before testing. The spacers were θp the relative rotation of the adherends at the loading line (Fig. 2).
Expression (2) was used in the present work, considering θo
removed and the adherends sides were sprayed with white brittle
paint, to allow an easy identification of a, and a printed scale was instead of θp, to gain accuracy in the measurements, since the
optical method is inherently more precise. The J-integral is defined
glued in both adherends to aid the a measurement or input data
for the digital correlation technique. Twenty-four specimens were along an arbitrary path encircling the start of the adhesive layer,
tested (six for each configuration) at room temperature and 1 mm/ giving [18]:
min in an electro-mechanical testing machine (Shimadzu AG-X Z δnc
100) with a load cell of 100 kN. Each test was fully documented Gn ¼ t n ðδn Þdδn : ð4Þ
0
using a 18 MPixel digital camera with no zoom and fixed focal
distance to approximately 100 mm. This procedure allowed The value of δnc corresponds to the crack-tip end-opening at
obtaining the crack tip opening (δn) and rotation (θo), necessary failure of the cohesive law, while tn is the current normal traction.
for the J-integral method. The correlation of the mentioned GIC is the value of GI when the crack initiates propagation, and is
parameters with the load–displacement (P–δ) data was done by given by the steady-state value of GI in the GI–δn plot [18]. The
the time elapsed since the beginning of each test. tn(δn) plot is obtained by differentiation of Eqs. (2) and (3) with
respect to δn

∂GI
3. Estimation of GIC t n ðδn Þ ¼ : ð5Þ
∂δn

It is known that, when the adhesives are ductile, as it occurs

with the Sikaforces 7888, LEFM methods are inaccurate [36], even 3.2.1. Optical method for the parameter measurement
though considering techniques with correction factors to account The process detailed here is an improvement of the work
for plasticity (e.g. ASTM D3433-99:2005 and BS 7991:2001). Thus, developed by the authors in reference [40]. For calculating δn and
the compliance-based beam method (CBBM), which accounts for θo for a given image, the optical method requires the identification of
the damage zone ahead of the crack tip, and the J-integral, were eight points (Fig. 3): two points (p3 and p4) for measuring the current
considered for the present study. tA value at the crack tip (tACT) during loading in image units (pixels),
two points (p7 and p8) identifying a line segment in the image for
3.1. Compliance-based beam method which the length (d) is known in real world units (mm), two points
(p1 and p5) on the top specimen and two points (p2 and p6) on the
The CBBM was initially selected to measure GIC of the adhesive bottom specimen for computing θo.
[13]. It is a relatively straightforward but robust method, based on
an equivalent crack, and it only depends on the specimen0 s 3.2.1.1. Points identification. All eight points are manually identified
compliance during the test. Applied to the DCB test specimen, in the first picture of a trial using an in-house software tool. The

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
4 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Fig. 2. DCB specimen under loading, with description of the analysis parameters.

Fig. 3. Points taken by the optical method to measure δn and θo.

identification of the points is aided by the ruler attached to the Calculating γ for all the pixels of I results in a matrix where the
specimens, which helps finding their correct locations. In addition, maximum absolute value yields the location of the region in I that
each point between p1 and p6 is printed with a distinct colour has the highest correlation with t and, thus, the most likely
(although this is not perceptible in Fig. 3). Using the location of the location of pi in the next image. This is done for every one of the
points in the first picture, the points of the following pictures are eight points identified in the first image. After successfully
automatically identified using a computer algorithm implemented in identifying all the points of the second image, new templates are
Matlabs. Basically, for each point pi, a rectangular region centred in pi computed from the second image to search for the eight points in
is extracted from the first image forming a template (t). This template the third image, and so on until processing all images.
describes the image pattern that surrounds the point and is used for
locating the point in the next image. This is done by finding the
position (u,v) in the next image (I) that has the highest normalized 3.2.1.2. Computation of δn. The value of tACT in real world units
cross-correlation with the template. The normalized cross-correlation (mm) is calculated as follows
is a measure of similarity between two images that is invariant  
p  p 
to linear changes in the pixel intensities and that quantifies the t A CT ¼ d 3 4
; ð8Þ
p p 
7 8
correlation between two images/regions [41]. This measure of
similarity was chosen due to its low computational requirements, Assuming that the lens distortion is negligible, which is valid for
which is a critical factor given the high resolution of the images, and the central area of pictures acquired with modern CCD cameras
because changes in rotation and scale of the specimens are expected [42]. A length of d ¼15 mm was used for all trials (illustrated in
to be small between two consecutive acquisitions (gapped by 5 s). To Fig. 3). The pixel size was on average 0.021 mm and, thus, the
take advantage of the colour information, the colour space of the estimated maximum error of the image acquisition process is
images (and consequently, of the templates) was transformed to the 70.011 mm. Finally, δn can be defined as
CIELAB colour space. The CIELAB system represents the value of a pixel
by three components, L, a and b, where L represents luminosity and a δn ¼ t A CT t A ; ð9Þ
and b define colour. Since points p1 to p6 are differentiated by their
where tA is the theoretical design value of 1 mm. Since tA can show
colour, only the a and b components are used when detecting points.
small variations due to the fabrication process, an adjustment to δn
The normalized cross-correlation (γ) of template t with image I at the
is also applied to make δn ¼0 at the beginning of the test (detailed
position (u,v) of image I for the colour component c is defined as:
in Section 4).
∑x;y ½Iðx; y; cÞ I u;v;c   ½tðx  u; y  v; cÞ t c 
γ ðu; v; cÞ ¼  0:5 ;
∑x;y ½Iðx; y; cÞ  I u;v;c 2  ∑x;y ½tðx  u; y v; cÞ  t c 2 3.2.1.3. Computation of θo. θo is calculated as the angle between
ð6Þ the tangents to the horizontal curves of the 2 scales closest to the
adhesive, measured at the crack tip (Fig. 4). The curvature of the
where I(x,y,c) is the intensity of the colour component c of the pixel (x, top adherend is first computed by fitting a quadratic function to
y) of image I; t(x,y,c) is the intensity of the colour component c of the points p1, p3 and p5. The first derivative of the quadratic function at
pixel (x,y) of the template t; I u;v;c is the average intensity of the colour p3 yields the slope of the top curve (mtop) at the crack tip, which is
!
component c of the region of image I centred at pixel (u,v) and with then used to define a direction vector v top ¼ (1,mtop). The same
the same size as t, and t c is the average intensity of the colour process is repeated for points p2, p4 and p6, yielding the slope of
component c for the template t. Finally, the normalized cross- the tangent to the bottom curve at the crack tip (mbottom) and its
!
correlation for a single pixel taking into account the colour direction vector v bottom ¼(1,mbottom). Finally, θo is obtained by
components a and b is defined as: measuring the angle between the two vectors:
0 1
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! !
v top  v bottom C
γ ðu; vÞ ¼ γ ðu; v; aÞ2 þ γ ðu; v; bÞ2 : ð7Þ
θ0 ¼ arccosB@! ! A:
 ð10Þ
 v top  v bottom 

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5

Fig. 4. Calculation of θo. Quadratic functions were fitted to points p1, p3, p5 and p2, p4, p6, representing the curvature of the top and bottom specimen, respectively, while the
straight lines show the tangents to the curves at the crack tip (corresponding to 10 mm in the scales).

1.6 1.6

1.2 1.2

GIC [N/mm]
GI [N/mm]

0.8 0.8

0.4 0.4

0 0
30 40 50 60 70 0 1 2 3 4
aeq [mm] h [mm]

Fig. 5. Evolution of GI with aeq for a test specimen with h ¼3 mm. Fig. 6. Average values and deviation of GIC as a function of h by the CBBM.

4. Results 0.06

After having performed the tests on the DCB specimens with 0.04
θo [rad]

varying values of tA, no signs of plasticity were found in the

adherends, although for the specimens with tA ¼ 1 mm the tensile
0.02
displacements of the loading points during crack propagation was
significant.
0
0 50 100 150 200 250 300
testing time [s]
4.1. Evaluation of GIC by the CBBM
Raw curve Adjusted curve Polinomial (Raw curve)

GIC was initially estimated by the CBBM, which is a well known Fig. 7. Plot of θo-testing time for a specimen with h ¼ 4 mm: raw curve, polynomial
method for this purpose, and allows accounting for the plasticity approximation and adjusted polynomial curve.
of materials. By using this method, it was possible to obtain
the R-curve for each specimen, relating GI with aeq, as the crack specimens with h¼2 mm (considering average values) was of
progressed. Fig. 5 gives an example of an experimental R-curve 13.8%, 48.3% to h¼3 mm and 56.7% to h¼4 mm.
obtained by this technique for a specimen with h ¼3 mm, clearly
showing the attainment of a steady-state value of GI. For this 4.2. Evaluation of GIC by the J-integral
specimen, a0 was measured at 31.50 mm. The corresponding value
of aeq by the CBBM was 38.57 mm, calculated directly from the P–δ GIC was calculated by Eq. (2), in which θo was estimated as
data by the first drop of P in the P–δ curve. This difference described in Section 3.2.1.3 and plotted with the time elapsed
occurred since aeq accounts for the damage zone [13]. Fig. 6 since the test initiation, with one data point every 5 s. Fig. 7 gives
compares the GIC (N/mm) values of the specimens with varying an example of the evolution of θo for a selected test specimen
values of h, including the standard deviation for each batch of (with h¼ 4 mm). This specimen is also used in the following
specimens. An increasing trend for GIC was found as a function of figures as being representative of the tests. Shown in the graphic
h, starting from the specimens with h¼ 1 mm, with a steeper are the raw curve, the 4th degree fitting curve and the corrected
increase up to 3 mm and stabilization above this value. Fig. 6 also polynomial and final curve, adjusted to make θo(testing
shows some deviation between specimens of identical conditions, time ¼0) ¼ 0. It should be mentioned that the raw curve of each
but perfectly within reported deviations under identical testing specimen was adjusted by the most suited polynomial function
conditions [40]. Actually, this scatter is related to experimental between 3rd and 6th degrees, by choosing the best correlation
phenomena such as fabrication issues or small measurement factor, R (this also applies to the forthcoming fitting data). This
errors or geometry deviations. However, the reliability of the polynomial adjustment is required to smooth the raw data and
observed tendency with h obviously cannot be questioned, or remove experimental measurement scatter, but also to cancel any
solely attributed to experimental scatter, on account of the large eventual misalignment between glued scales in both adherends.
improvement visible in Fig. 5. In fact, for h¼ 1 mm, the value of The following step consisted on estimating the curve relating δn
GIC ¼ 0.688 70.153 N/mm was found. The improvement to the with the testing time, to determine the cohesive law by Eq. (5).

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
6 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎

0.16 40

0.12 30
δn [mm]

tn [MPa]
0.08 20

0.04 10

0 0
0 50 100 150 200 250 300 0 0.02 0.04 0.06 0.08 0.1
testing time [s]
δ n [mm]
Raw curve Adjusted curve Polinomial (Raw curve) Estimated CZM law Trapezoidal aprox.
Fig. 8. Plot of δn-testing time for a specimen with h ¼4 mm: raw curve, polynomial Fig. 11. Estimated tn–δn law for one test specimen and approximation.
approximation and adjusted polynomial curve.

1.6 in reference [40], where an average value of 1.182 N/mm was

found in DCB specimens of natural fibre composites with h ¼
1.2
5 mm. The tn–δn law for the test specimen followed in this section
GI [N/mm]

is presented in Fig. 11, together with a possible trapezoidal

0.8
approximation, particularly suited to model CZM laws for ductile
0.4
adhesives [43]. The pertinent parameters for this specimen were
t 0n ¼25.1 MPa and δnc ¼0.0692 mm.
0
0 0.02 0.04 0.06 0.08 4.3. Discussion
δ n [mm]
Experimental law Polinomial (Experimental law) The comparison between both of the proposed methods, the
Fig. 9. Experimental Gn–δn law for one specimen and polynomial fitting. CBBM and J-integral, showed that under the present testing
conditions both methods agree quite well, although the J-integral
results are slightly higher. Actually, the difference between average
1.6 values was of 13.5%, 12.3%, 5.3% and 1.6% for increasing h values
between 1 and 4 mm. Despite this small variation, which is
1.2 probably due to statistical reasons, the observed trend was in
GIC [N/mm]

high correlation, and in both cases a stabilization of GIC seems to

0.8
occur at h ¼3 mm. At least, a much lesser effect exists from
h¼3 mm than for smaller values of h. This increase of GIC is
0.4
reported in the literature because of the stress field variations
ahead of the crack tip being dependent on the joint geometry,
0
0 1 2 3 4 which highly influences the shape and size of the damage zone,
h [mm] and the local yield stress as well [44]. As it was discussed in
previous works [45], thicker adherends provide an elevation of
Fig. 10. Average values and deviation of GIC as a function of h by the J-integral.
peel stresses further within the joint, shifting the loading condi-
tions from peeling to cleavage, and giving a larger length for the
Fig. 8 shows the δn-testing time plot for a specimen, more damage zone. These findings are corroborated in the work of Azari
specifically the three curves of Fig. 7. Due to scaling difficulties, et al. [46], regarding the adherend stiffness influence on the
the raw curve in the figure is already translated such that fatigue failure of bonded joints, which proved by finite elements
δn(testing time ¼ 0) ¼0. GIC for each specimen was then defined that the plastic zone in adhesive joints between steel adherends
by the respective GI–δn curve, as the steady-state value of GI [17]. was consistently higher than identical joints between aluminium
Fig. 9 follows the same specimen and shows the experimental GI– adherends during the entire damage uptake process up to crack
δn law and the corresponding 6th degree polynomial. This curve is initiation. Pardoen et al. [23] developed an analytical model to
representative of all specimens, and shows three regions: (1) slow study the constraint effects in adhesive joint fracture, considering
increase of GI with δn, but with increasing growth rate of GI (up to steel adherends and two epoxy adhesives, and reported a sig-
δn E0.02 mm), (2) linear increase of GI (for this specimen between nificant increase of the adhesive plastic dissipation in the fully
0.02 mm r δn r0.04 mm) and (3) gradual attainment of a steady- plastic regime by increasing the adherends thickness. The reported
state value of GI (δn E 0.075 mm). For this specimen, the measured dependence of GIC with h was also addressed in joints with
value of GIC was 1.21 N/mm. The GIC results by applying this composite adherends, in which h is defined by the number of
procedure for all tested specimens are shown in Fig. 10. The GIC plies, and results were found to be consistent with those of this
evolution with h is consistent with that shown in Fig. 6. The work. Mangalgiri et al. [31] justified this tendency with the plastic
deviation is larger, though, and whose justification lies on the zone and stress distributions ahead of the debond tip. Actually, the
experimental process to obtain GIC, which relies on a number of plastic zone was bigger in length across the adhesive layer with
measured parameters and approximation functions, which are increasing number of composite plies (and thus, increasing h).
difficult to adjust to the experimental data [40]. While for the Also, thicker adherends used a larger amount of the input energy
specimens with h ¼1 mm, a value of GIC ¼0.781 70.146 N/mm was to the specimen to develop a lengthier plastic zone, thus leaving
obtained, improvements of 12.6%, 37.7% and 40.2% were attained less available energy for damage growth [47]. On account of this,
by increasing h up to 4 mm. These results reinforce the previous higher values of GIC can be expected for joints with higher degrees
assumption on the stabilization of GIC for a given value of h (in this of restraint (i.e., stiffer or thicker adherends). On the other hand,
case of GIC ¼1.095 70.195 N/mm for h ¼4 mm). This steady-state because of the damage zone length limitation by the adhesive
value of GIC can be compared with previous results by the authors ductility, it is also noted that tough adhesives are particularly

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7

prone to these effects, while brittle adhesives are not. The [9] Campilho RDSG, de Moura MFSF, Ramantani DA, Morais JJL, Domingues JJMS.
consideration of adherends with different stiffness provides a Buckling behaviour of carbon-epoxy adhesively-bonded scarf repairs. J Adhes
Sci Technol 2009;23:1493–513.
similar effect to the variation of h, as stiffer adherends give a [10] Campilho RDSG, Banea MD, Neto JABP, da Silva LFM. Modelling of single-lap
higher restriction to the adhesive layer deformation. Under this joints using cohesive zone models: effect of the cohesive parameters on the
scope, Bell and Kinloch [44] tested DCB specimens with alumi- output of the simulations. J Adhes 2012;88:513–33.
[11] Yoshihara H. Simple estimation of critical stress intensity factors of wood by
nium, steel and carbon-fibre adherends, and obtained an increas- tests with double cantilever beam and three-point end-notched flexure.
ing trend of GIC with the adherend stiffness. Contradicting results Holzforschung 2007;61:182–9.
were found by Choupani [48], but in this case the differences were [12] Wang SS. Fracture mechanics for delamination problems in composite
materials. J Compos Mater 1983;17:210–23.
given by a modification of the failure mechanism for the joints [13] de Moura MFSF, Campilho RDSG, Gonçalves JPM. Crack equivalent concept
with stiffer adherends (steel compared to aluminium). With applied to the fracture characterization of bonded joints under pure mode I
increasing values of h, eventually the plastic zone reaches a loading. Compos Sci Technol 2008;68:2224–30.
[14] Suo Z, Bao G, Fan B. Delamination R-curve phenomena due to damage. J Mech
maximum, i.e., a value above which it does not increase any more,
Phys Solids 1992;40:1–16.
justifying the stabilization of GIC. The plastic zone assumption [15] Sorensen BF, Jacobsen TK. Characterizing delamination of fibre composites by
implies that brittle adhesives are marginally affected, if so, by this mixed mode cohesive laws. Compos Sci Technol 2009;69:445–56.
parameter, and this was confirmed in the work of Fernlund and [16] Zhu Y, Liechti KM, Ravi-Chandar K. Direct extraction of rate-dependent
traction-separation laws for polyurea/steel interfaces. Int J Solids Struct
Spelt [49], whose results on DCB with aluminium adherends up to 2009;46:31–51.
12.7 mm thick and a brittle adhesive did not show any variations [17] Carlberger T, Stigh U. Influence of layer thickness on cohesive properties of an
beyond the statistically related scatter. epoxy-based adhesive-an experimental study. J Adhes 2010;86:814–33.
[18] Ji G, Ouyang Z, Li G, Ibekwe S, Pang SS. Effects of adhesive thickness on global
and local mode-I interfacial fracture of bonded joints. Int J Solids Struct.
2010;47:2445–58.
5. Conclusions [19] Chen Z, Adams RD, da Silva LFM. Prediction of crack initiation and propagation
of adhesive lap joints using an energy failure criterion. Eng Fract Mech
2011;78:990–1007.
This work aimed to analyse the influence of h on the measured
[20] Kinloch AJ, Shaw SJ. A fracture mechanics approach to the failure of structural
value of GIC of a ductile adhesive within a pure tensile test as it is the joints. In: Kinloch AJ, editor. Developments in adhesives—2. London: Applied
DCB test. Different values of h were considered, between 1 and Science Publishers; 1981. p. 83.
4 mm. Two techniques were used for GIC, the CBBM and the J- [21] da Silva LFM, de Magalhães FACRG, Chaves FJP, de Moura MFSF. Mode II
fracture toughness of a brittle and a ductile adhesive as a function of the
integral. For the J-integral methodology, a relatively time-consuming adhesive thickness. J Adhes 2010;86:891–905.
approach was required, which involved evaluating by an optical [22] Ashcroft IA, Shenoy V, Critchlow GW, Crocombe AD. A comparison of the
method θo and δn at the crack tip during the test, followed by prediction of fatigue damage and crack growth in adhesively bonded joints
using fracture mechanics and damage mechanics progressive damage meth-
polynomial fitting and differentiation. The trends between both ods. J Adhes 2010;86:1203–30.
methods were consistent, although the GIC values obtained by the [23] Pardoen T, Ferracin T, Landis CM, Delannay F. Constraint effects in adhesive
J-integral were slightly bigger than for the CBBM (between 1.6 and joint fracture. J Mech Phys Solids 2005;53:1951–83.
[24] Wang RX, Sinclair AN, Spelt JK. Strength of adhesive joints with adherend
13.5%, depending on h). Disregarding the data reduction method, an yielding: II. Peel experiments and failure criteria. J Adhes, 79; 2003; 49–66.
increasing trend of GIC with h was found, suggesting that GIC is not a [25] Ikeda T, Yamashita A, Lee D, Miyazaki N. Failure of a ductile adhesive layer
material parameter, but a geometry-dependent quantity instead. constrained by hard adherends. J Eng Mater Technol 2000;122:80–5.
[26] Giannis S, Adams RD, Clark LJ, Taylor MA. The use of a modified peel specimen
The increase of GIC was bigger for the smaller h values, eventually to assess the peel resistance of aircraft fuel tank sealants. Int J Adhes Adhes
attaining a steady-state value for a given h value, as it sounded by 2008;28:158–75.
comparing results for specimens with h¼3 and 4 mm. This result is [27] Ferracin T, Landis CM, Delannay F, Pardoen T. On the determination of the
cohesive zone properties of an adhesive layer from the analysis of the wedge
highly relevant since structural bonding usually falls within small
peel test. Int J Solids Struct 2003;40:2889–904.
values of h, and was considered to be due to an increasing degree of [28] da Silva LFM, Carbas RJC, Critchlow GW, Figueiredo MAV, Brown K. Effect of
adherend restraining for bigger h values, as a larger region is loaded material, geometry, surface treatment and environment on the shear strength
ahead of the crack tip. The J-integral enabled obtaining the tensile of single lap joints. Int J Adhes Adhes 2009;29:621–32.
[29] Keller T, Vallée T. Adhesively bonded lap joints from pultruded GFRP profiles. Part
CZM law of the adhesive. The CZM curves showed the large plasticity I: Stress–strain analysis and failure modes. Composites Part B 2005;36:331–40.
of the polyurethane adhesive. As output of this work, GIC data was [30] Pinto AMG, Campilho RDSG, Mendes IR, APM Baptista. Numerical and
given for the strength prediction of bonded joints. experimental analysis of balanced and unbalanced adhesive single-lap joints
between aluminium adherends, J Adhes. doi: 10.1080/00218464.2013.773258.
[31] Mangalgiri PD, Johnson WS, Everett Jr RA. Effect of adherend thickness and
mixed mode loading on debond growth in adhesively bonded composite
Acknowledgments joints. J Adhes 1987;23:263–88.
[32] Devitt DF, Schaperv RA, Bradley WL. A method for determining the mode I
The authors would like to thank Sikas for supplying the delamination fracture toughness of elastic and viscoelastic composite materi-
als. J Compos Mater 1980;14:270–85.
adhesive. [33] Campilho RDSG, Banea MD, Pinto AMG, da Silva LFM, de Jesus AMP. Strength
prediction of single-and double-lap joints by standard and extended finite
element modelling. Int J Adhes Adhes 2011;31:363–72.
References
[34] Pinto AMG, Campilho RDSG, Mendes IR, Aires SM, Baptista APM. Effect of hole
drilling at the overlap on the strength of single-lap joints. Int J Adhes Adhes
[1] da Silva LFM, Öchsner A, Adams RD, editors. Heidelberg: Springer; 2011. 2011;31:380–7.
[2] da Silva LFM, Campilho RDSG. Advances in numerical modelling of adhesive [35] Neto JABP, Campilho RDSG, da Silva LFM. Parametric study of adhesive joints
joints. Heidelberg: Springer; 2011. with composites. Int J Adhes Adhes 2012;37:96–101.
[3] Campilho RDSG, Banea MD, Chaves FJP, da Silva LFM. eXtended finite element [36] Giovanola JH, Finnie I. A review of the use of the J integral as a fracture
method for fracture characterization of adhesive joints in pure mode I. parameter. Solid Mech Arch 1984;9:197–225.
Comput Mater Sci 2011;50:1543–9. [37] Banea MD, da Silva LFM, Campilho RDSG. Effect of temperature on tensile
[4] Volkersen O. Die nietkraftoerteilung in zubeanspruchten nietverbindungen strength and mode I fracture toughness of a high temperature epoxy adhesive.
konstanten loschonquerschnitten. Luftfahrtforschung 1938;15:41–7. J Adhes Sci Technol 2012;26:939–53.
[5] Goland M, Reissner E. The stresses in cemented joints. J Appl Mech 1944;66: [38] Rice JR. A path independent integral and the approximate analysis of strain
17–27. concentration by notches and cracks. J Appl Mech 1968;35:379–86.
[6] Hart-Smith LJ. Adhesive bonded single lap joints. NASA Contract Rep [39] Banea MD, da Silva LFM, Campilho RDSG. Temperature dependence of the
1973;112235. fracture toughness of adhesively bonded joints. J Adhes SciTechnol
[7] Chai H. Shear fracture. Int J Fract 1988;37:137–59. 2010;24:2011–26.
[8] Campilho RDSG, de Moura MFSF, Barreto AMJP, Morais JJL, Domingues JJMS. [40] Campilho RDSG, Moura DC, Gonçalves DJS, da Silva JFMG, Banea MD, da Silva
Fracture behaviour of damaged wood beams repaired with an adhesively- LFM. Fracture toughness determination of adhesive and co-cured joints in
bonded composite patch. Composites Part A, 40; 2009; 852–9. natural fibre composites. Composites Part B 2013;50:120–6.

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i
8 R.D.S.G. Campilho et al. / International Journal of Adhesion & Adhesives ∎ (∎∎∎∎) ∎∎∎–∎∎∎

[41] Lewis JP. Fast template matching, vision interface 95, Canadian Image Processing [46] Azari S, Ameli A, Datla NV, Papini M, Spelt JK. Effect of substrate modulus
and Pattern Recognition Society, Quebec City, Canada, May 15-19, 1995, p. 120–123. on the fatigue behaviour of adhesively bonded joints. Mater Sci Eng, A
[42] Zhang Z. Camera calibration. In: Emerging topics in computer vision, G 2012;534:594–602.
Medioni S Kang, editors. Prentice Hall Professional Technical Reference, [47] Azari S, Ameli A, Papini M, Spelt JK. Adherend thickness influence on fatigue
2004, ch. 2. behaviour and fatigue failure prediction of adhesively bonded joints. Compo-
[43] Campilho RDSG, de Moura MFSF, Barreto AMJP, Morais JJL, Domingues JJMS. sites Part A 2013;48:181–91.
Experimental and numerical evaluation of composite repairs on wood beams [48] Choupani N. Mixed-mode cohesive fracture of adhesive joints: experimental
damaged by cross-graining. Constr Build Mater 2010;24:531–7. and numerical studies. Eng Fract Mech 2008;75:4363–82.
[44] Bell AJ, Kinloch AJ. The effect of the substrate material on the value of the [49] Fernlund G, Spelt JK. Mixed-mode fracture characterization of adhesive joints.
adhesive fracture energy, Gc, J. Mater Sci Lett 1997;16:1450–3. Compos Sci Technol 1994;54:441–9.
[45] Blackman BRK, Kinloch AJ, Paraschi M. The effect of the substrate material on
the value of the adhesive fracture energy, Gc: further considerations. J Mater
Sci Lett 2001;20:265–7.

Please cite this article as: Campilho RDSG, et al. Adherend thickness effect on the tensile fracture toughness of a structural adhesive
using an optical data acquisition method. Int J Adhes Adhes (2014), http://dx.doi.org/10.1016/j.ijadhadh.2014.01.015i