Mechanical behaviour evaluation of adhesively bonded joints of

aircraft structures
Sérgio Nunes Correia
sergio.n.correia@tecnico.ulisboa.pt

Instituto Superior Técnico, Lisboa, Portugal

October 2016

Abstract
The structural integrity of some structures can be related to the strength of their unions. Over the
years, adhesively bonded joints have been often chosen over several mechanical systems based joints,
as welding or riveting. Among others, the reduction in stress concentrations, the ability of producing
smooth surfaces, with no discontinuities and the reduced weight penalties are some of the factors that
make this type of joints so attractive. However, when compared with these other joining techniques,
it can be verified that there are practically no accepted guidelines for predicting the strength of an
adhesive joint, being the design of some joints still based in personal experience and in the so-called
rules of thumb. Recently, some numerical models have been proposed to estimate the failure load of
bonded joints. The main problem is related with the fact that, for most of these methods to be applied,
some experimental tests must be performed to identify some material models, which is not always
convenient. With this in mind, an experimental study was performed, based in the Single Lap Joint
from the ASTM D 1002 standard method, with the objective of determining the best surface treatment,
for aluminum-to-aluminum joints, using two types of adhesives. Furthermore, a computational model,
using the Finite Element Method, has been implement in order to predict the adhesive behavior, taking
only the stress-strain curve of the adhesive as an input.

Keywords: Adhesive, Adhesively Bonded Joint, Single Lap Joint, Surface Treatment, ASTM D 1002,
FEM

1. Introduction ence and rules of thumb.

With the advances in manufacturing techniques, Recently, some computational methods have
several improvements have been verified in struc- been developed to modulate adhesively bonded
tural adhesives. As a direct result of this improve- joints. Nonetheless, most of these methods are
ments, the use of bonded joints has supplemented or based in damage mechanics, meaning that several
replaced the use of mechanical-based joints in sev- experimental tests need to be performed in order to
eral aeronautical structures. Between other advan- enquire about the damage models of the adhesive.
tages, adhesively bonded joints allow a better load In this way, the development of simple and accu-
distribution, increase in the service life, reduced ma- rate techniques to predict the failure of this type of
chining cost and reduced complexity [1]. joints, with the minimum number of experimental
Despite the many advantages, bonded joints also tests, would be a great improvement to the indus-
suffer from a significant number of limitations. try.
Among others, one can identify the bad adhesion Several studies and analysis show that the bond
to some substrates, weak resistance to cleavage strength of an adhesively bonded joint depends on
stresses or the degradation due to exposure to hos- many factors, such as the bonded area, the envi-
tile environments [2]. However, in contrast to more ronment in which the structure will be used, the
conventional joining techniques, the major concern service temperature, the type of surface treatment
related with adhesively bonded joints is that there applied to the adherents and, of course, the type of
exist practically no generally accepted guidelines to adhesive used. With all this in mind, it is easy to
predict the joint strength. The aerospace industry, see that all these factors are strictly related to air-
which has been developing the technology, is still craft security, meaning that a very thoughtful joint
designing joints mainly based in previous experi- design and adhesive choice are essential.

1
 Said so, the main motivation of performing this Due to the fact that adhesives are not rigid, non
work is, firstly, to study and experimentally evalu- uniform strain is verified across the bondline, de-
ate the behavior of adhesively bonded joints when veloping, consequently, a non uniform shear stress
subjected to different types of surface treatments distribution. Moreover, since the loads are not ini-
and when bonded with different adhesives, specially tially in the same line of action, a bending moment
in terms of adherence. Secondly, it is of great in- will be developed in the joint, leading to the devel-
terest to develop a computational model capable opment of peel stresses in the adhesive layer.
of predicting the failure of a bonded joint, assum- Over the years, several methods have been de-
ing perfect conditions of adhesion between adhesive veloped to solve, analytically, both these shear
and adherents and having only as information the and peel stress distributions. The first analytical
stress-strain curve of the adhesive. method to solve the SLJ stress distribution was de-
This work has been developed in coopera- veloped by Volkersen [4], in 1938. Volkersen was
tion with TAP Maintenance & Engineering (TAP the first to consider the non uniform strain in the
M&E). Actually, the adhesive used to repair the bondline. Despite, this method presents some limi-
engine nacelle structure is being discontinued. In tations, related with the fact that the bending mo-
this way, knowing about the possibility of perform- ment effects were negleted. Only an expression for
ing some types of anodizing in their complex, TAP the shear stress distribution was obtained, being
M&E engineers are trying to find a combination of presented in Equation 2, for equal adherents. λ is
a new adhesive and surface treatment, that can be given by Equation 3. The adhesive was considered
performed indoors and ensure a joint strength at perfectly elastic.
least as good as the old one.
With this in mind, two principal objectives are " #
P λ cosh λx
proposed in this work, being one experimental and τ (x) = (2)
the other numerical. These are: 2 sinh λ L2

• Identify, experimentally, an optimum combina- r

tion of adhesive and surface treatment that can 2Ga
λ= (3)
be performed inside TAP M&E complex. Eta t

• Develop a computational model that can pre- Where P is the applied load, E is the adherent
dict the behavior and strength of bonded elastic modulus, t represents the adherent thickness
joints. and G is the elastic shear modulus. When accom-
panied by a subscript a, the parameters are referent
The Single Lap Joint (SLJ), present in Figure 1, to the adhesive.
is the type of bonded joint most used in an aircraft,
In 1944, Goland and Reissner [5] were the first
specially due to the ease of produce, inspect and
to consider the effects related with the bending
repair [3]. Said so, this work will be oriented to the
and transverse force that occurs due to the non-
study of the SLJ.
collinearity of the applied load. The solution for
the shear stress distribution is given by Equation 4,
being k the bending moment factor.

" #
P cosh δx
c
τ (x) = (1 + 3k)δ + 3(1 − k) (4)
8c sinh δ

Figure 1: Single Lap Joint [3].

The expression to the bending moment factor,
k, is presented in Equation 5, being u2 given by
2. Literature Review Equation 6.
2.1. Analytical Methods
Starting to analyze the SLJ, a project parameter cosh u2 c
can be defined. This parameter is the apparent k= √ (5)
cosh u2 c + 2 2 sinh u2 c
shear stress, τ , present in the adhesive layer, and
is given by Equation 1. r r
3(1 − υ 2 ) 1
P
P u2 = 2
(6)
τ= (1) 2
t E t
bL
Where P represents the applied load, b the SLJ For the peel stress, the expression is given in
width and L the bond overlap. Equation 7.

2
 Hart-Smith developed two solutions for the peel
  and shear stress distributions: the first one, assum-
Pt 2k 0
σ(x) = [ R2 λ + λk cosh λ cos λ × ing a purely elastic model and, as previously re-
∆c2 2
    ferred, a second one using an elastic-perfectly plas-
λx λx tic model.
× cosh cos +
c c For the first solution, distributions are presented
(7)
in Equations 9 and 10, for shear and peel stresses,
 
k
+ R1 λ2 + λk 0 sinh λ sin λ × respectively. The equations are valid for equal ad-
2
    herents.
λx λx
× sinh sin ]
c c
τ (s) = A2 cosh 2λ0 s (9)
The parameters of the peel stress distribution
equation are given in Equations 8.
0.25 σc (s) = A cos χs cosh χs + B sin χs sinh χs (10)
6Ea c4

λ= (8a)
ta Et3 Where −c ≤ s ≤ c. Constants A2 , A and B are
r obtained using Equations 11.
kc P
k0 = 3(1 − υ 2 ) (8b)
t E
6(1 − ν 2 )M0
 
R1 = cosh λ sin λ + sinh λ cos λ (8c) G
A2 2λ0 sinh 2λ0 c = P+ (11a)
Etta kb t
R2 = sinh λ cos λ − cosh λ sin λ (8d)
1 Ec M0 [sin χc − cos χc]
∆ = [sin 2λ + sinh 2λ] (8e) A=− (11b)
2 ta Dχ2 eχc
Ec M0 [sin χc + cos χc]
As well as Volkersen, Goland and Reissner con- B= (11c)
ta Dχ2 eχc
sidered the adhesive as perfectly elastic.
Hart-Smith [6], in 1973, was the first to consider
Considering the plasticity, the shear stress dis-
the plastic deformation in the adhesive, in addition
tribution is given by Equation 12. Also, the solu-
to the elastic response. If the adhesive plasticity is
tion for the shear strain distribution is presented in
allowed, the analytical joint strength prediction will
Equation 13. The origin for ζ is defined at s = d/2,
be higher, since a failure strain criteria can be used
being only valid in the range [d/2, c]. d represents
instead of a stress based criteria.
the length of the elastic range.
To characterise the adhesive behaviour, Hart-
Smith used an elastic-perfectly plastic model - Fig-
τ = A2 cosh 2λ0 s + +C2 (12)
ure 2.

γ = A3 ζ 2 + B 3 ζ + C 3 (13)
The constants A2 , B2 , A3 , B3 and C3 , as well as
the unknown ratio d/l are found by solving a system
of equations that arises from the application of the
BC defined.
2.2. Numerical Analysis - The Finite Element
Method
The use of the Finite Element Method (FEM) to
study the behavior of bonded joints has been quite
used in the academic world, specially due to the
Figure 2: Effect of plasticity in the adhesive layer great number of commercial software available and
according to Hart-Smith [6]. their ease of use. The computational analysis de-
veloped during this work were all performed using
In Figure 2, there are three different cases, that ANSYS Parametric Design Language (APDL). The
are obtained by increasing the applied load, P . In version of the software used is ANSYS 14.0.
short, for Case 1 the adhesive is in the elastic range. As well as in every finite element (FE) software,
For Case 2, the adhesive starts to yield, since the there are several essential parameters that will influ-
yield limit has been overcome. Finally, for Case 3, ence the quality and accuracy of the results. Among
the adhesive fails, due to the fact that the shear others, should be emphasized the type of FE used
strain achieves its ultimate value. and the mesh density (MD) [7]. One typical choice

3
for adhesively bonded joints are continuum solid
elements, suited for non-linear problems involving
plasticity and large deformations. Regarding the
MD, it do not need to be equally refined across
all the geometry. Due to the complex stress state
present in the adhesive layer, a more refined mesh
is recommended in this location [8].
A 2D analysis in plane strain conditions is almost Figure 3: On the left, singular point D of the SLJ
preferred for the most authors [9, 10] and will be with different degrees of rounding: a) sharp corner;
adopted during the development of this work. b) small rounding; c) medium rounding; d) large
The FEM presents two main problems when ap- rounding (please note that the thickness of the ad-
plied in the analysis of bonded joints: first, the ad- herent, in this Figure, is 3.2 mm). On the right,
herents are to big when compared with the very thin shear stress distribution for each of the cases. [12].
adhesive layer, wish can cause mesh problems and
very slow models. Second, if the non-filleted ge-
ometry of the adhesive is considered, singularities is the assumption that the structure fails as soon
appear in the perfectly sharp corners of the bonded as the failure criteria is verified at some point, i.e.,
joint. it assumes that the material presents a very brittle
As stated by Yue and Wahab [11], a stress sin- behavior. This is not always true, specially when
gularity is detected by divergence of peak stress analyzing very ductile materials, like adhesives [13].
with the mesh refinement. The most common way Some other methods have been proposed in
of handling this problem is to use the stresses ob- recent years, like the Cohesive Zone Modelling
tained along the center-line of the adhesive, being (CZM). However, to be applicable, more experimen-
the convergence in this points achieved for a certain tal tests than the bulk test are necessary, which is
mesh density. This method is defined as the Cen- not always convenient [13, 14, 15, 16]. Also, some
terline Method and can be used to compare compu- of the models consists in changing the input pa-
tational results with the classical solutions. How- rameters based in experimental observations. This
ever, a problem arises from this simplification: the still demands the realization of experimental tests,
stresses used are average stresses, not representing which is what one wants to avoid.
the variation of shear and peel stresses across the A new method has been arising lately, being
adhesive thickness. based in the local application of a continuum me-
The other way of handling singularities implies chanics failure criterion. This method is named
changing the geometry of the SLJ. In practice, the Progressive Failure Method (PFM) and has been
adherent corners are not perfectly sharp, so a small widely used in the past few years to calculate the
rounding can be added Figure 3. Also, during failure load of composite materials, with a very
bonding, there is always some adhesive that flows to good degree of success [17, 18, 19, 20]. More re-
the sides, creating a small fillet. Several researchers cently, this method started to be implemented in
even state that the fillet is essential to obtain accu- the analysis of some adhesive joints [2, 21].
rate results, advising that it shall be built, with an Essentially, the PFM involves three main stages
extra piece of adhesive, during the specimen man- that act in a cyclic way, being them the stress and
ufacturing. Combined with the rounding, the fillet strain analysis, the failure analysis, using an appro-
modulation prevents the appearance of strong sin- priate failure criterion and, finally, the application
gularities during the analysis. of the material degradation rule to the failed ele-
Regarding the failure load calculation, the first ments.
step consists in defining an appropriate and effective To better understand how the analysis is per-
method for the analysis to be performed. formed, a flowchart of the code is highlighted in
The most common method to evaluate material Figure 4.
failure is using continuum mechanics. Essentially, Some good results have been obtained with the
approaches based on continuous mechanics use the PFM [21, 2]. In addition of predicting the failure
maximum values of stress or strain distributions load, the PFM also allows to made a prediction of
against a chosen failure criterion, which uses the the failure path.
critical values of the material. Some advantages re-
lated with this kind of methods are the simplicity 2.3. Adherence and Types of Failure
of use and the fact that little information is needed All the considerations presented in Subsections 2.1
for the criteria, due to the dependency on one or and 2.2 were made assuming a perfect adherence be-
few parameters. tween the adhesive and adherents, with the failure
This methods present a great limitation, which occurring in the adhesive layer. However, in real

4
 Figure 5: Failure types in a bonded joint: cohesive
failure on the top left; adhesive failure on the top
right; adhesive-cohesive failure on the bottom left;
adherent failure on the bottom right [22].

adhesive’s data-sheets, while the last two are the

ones that can be performed in TAP M&E complex.
When the surface treatments are not so efficient,
adhesion and permanence time can be increased by
primers or adhesion promoters [25]. These type of
materials work in a similar way to improve adhe-
sion, adding a new, usually organic, layer at the
Figure 4: Flowchart for the Progressive Failure
interface between adhesive and substrate.
Model, adapted from [21].
3. Methods and Materials
3.1. Methods
joints, this is not always the case and four types As stated, the SLJ will be the main focus of this
of failure modes can be identified (Figure 5), being work. The usual test for this type of joints is the
them [22]: ASTM D 1002 [26], consisting essentially in the
steps described.
• Cohesive Failure - it is characterized by failure
of the adhesive itself; • The specimen must be placed in the grips of
the testing machine.;
• Adhesive Failure it is characterized by a failure
of the joint at the adhesive/adherent interface • The load shall be applied immediately to the
and is typically caused by inadequate surface specimen. The crosshead speed must be 1.3
preparation, chemically and/or mechanically; mm/min;

• Adhesive-Cohesive Failure - the failure presents • Apply the load to failure.

the previous two types simultaneously; The overlap used and the geometry of the speci-
• Adherent failure - it is characterized by fail- men can be seen in Figure 6. The thickness of the
ure of the adherent instead of the adhesive. In metal, t, is 1.6 mm, being the adhesive thickness
metals, this occurs when the adherent yields. dependent on the type of adhesive used.

Adhesive failures are intimately related with two

main factors: the surface treatment of the adherents
and the use of primers or adhesion promoters.
A surface treatment is applied in a material to
provide the best possible bonding conditions. There
are several types of surface treatments for metals
which can be classified as either micro-mechanical,
macro-mechanical, chemical or a combination of
these [23]. In this work, only chemical surface treat-
ments will be tested, being them the Phosporic Acid Figure 6: SLJ Specimen geometry.
Anodyzing (PAA), the Chromic Acid Anodyzing
(CAA), the Sulfuric Acid Anodyzing (SAA) and the From the realization of this test, one obtain the
Boric-Sulfuric Acid Anodyzing (BSAA) [24]. The failure load of the joint, Pf . To obtain the apparent
first two types are the treatments advised in the failure shear stress from Pf , Equation 1 can be used

5
3.2. Materials 3.3. Specimens Manufacturing
The only adherent tested is the aluminum 2024-T3 The specimens used in the experimental tests were
alloy. Since the main interest of the work is to study built in TAP M&E facilities. The manufacturing
the adhesive behavior, one must guarantee that the process had two main phases: Surface Treatment
adherents are in the elastic region. This was veri- and Bonding.
fied using a simple analytical method proposed by To apply the Surface Treatment, it is necessary
Adams and Wake [1]. to firstly clean the adherents surface. This is essen-
Regarding the adhesives, two types were tested tial to remove any type of grease or impurity from
in this work: the Henkel Loctite EA 9658 Aero and the aluminum. Immediately after, the aluminum
the 3M Scotch-Weld AF 163-2. Both are considered is activated by acid action, being submerged in a
structural adhesives, manufactured in film form and bath that promotes the creation of micro-pores in
widely used in the aerospace industry. These two the surface. Next, the plates are submerged in an
adhesives are tested in order to replace an old one anodize solution bath for 45 minutes, at a temper-
- the Hysol EA 9689, also produced by Henkel Loc- ature between the 24 and 29 o C, with a direct cur-
tite. rent applied on them of 15 V. The electric current
Also, each adhesive is advised to be used with a passing through the anodize solution converts the
primer, in order to improve their adherence to the aluminum surface in a form of oxide, increasing the
substrates. The Scotch-Weld EC-3924B must be contact area. Finally, a surface finish is applied.
used with the adhesive AF 163-2, while Henkel Loc- This process is used typically to increase corrosion
tite EA 9258.1 AERO is advised to be used along resistance and promote, even more, adherence of
with the EA 9658 AERO adhesive. paints and adhesives.
For both adhesives, the process of bonding is
Due to the impossibility of performing the bulk
quite similar, being described next. All the process
tensile tests to determine the stress-strain curves of
was performed in a clean room, under controlled
both adhesives, this properties were obtained fol-
environment.
lowing the work done by de Freitas and Sinke [27],
Regarding the Bonding phase, it starts with the
for the AF 163-2. For the EA 9658 AERO, the me-
application of the respective primer and cure. Af-
chanical properties of a similar adhesive were used.
ter curing, the adhesive can be applied in the de-
The adhesive chosen had a similar apparent failure
sired area. To prevent the stress concentrations,
shear stress and had the same curing and storage
and to obtain more reliable results, an adhesive fil-
conditions. Plus, both are designed to metal and
let is added manually to each side of the bond layer.
honeycomb bonding, indicated that it might be a
Finally, the adhesive is cured The specimens are in-
good approximation.
serted inside a peel-ply bag in order to create a vac-
With the analysis of de Freitas and Sinke [27] uum environment. The cure method was different
work and following the formulation presented by for each adhesive. Due to the high cure temper-
Hart-Smith [6], the adhesives were modelled as ature, the EA 9658 AERO specimens were cured
elastic-perfectly plastic. with thermal blankets, while the AF 163-2 speci-
All the mechanical properties used are presented mens were cured in the drying oven.
in Table 1.
3.4. Number and type of specimens
Two types of SLJ were tested, one representing a
Table 1: Resume of the properties used. new joint (NJ), while the other was representing a
Parameter AF 163-2 EA 9658 2024-T3 repaired joint (RJ). The surfaces of the adherents
E [GPa] 2.04 2.62 72.40 were different in the RJ: one substrate was prepared
ν 0.34 0.34 0.33 with a type of anodizing, while the other was com-
pletely clean, without any kind of treatment. Both
σy [MPa] 45.70 56.00 347.00
types of joints were combined with the two adhe-
εy [%] 2.24 2.14 - sives, as well as with the two types of anodizing.
εp [%] 5.38 6.00 - The ASTM D 1002 advices to perform at least 5
G [GPa] 0.76 0.98 - tests for each type of specimen studied. To increase
τy [MPa] 26.38 32.33 - the reliability of the results, 7 tests were conducted
for each one. Altogether, 56 tests were performed.
γy [%] 3.88 3.71 -
In Table 2 are presented all the combinations used.
γp [%] 9.32 10.39 -
τf [MPa] 39.30 40.00 - 4. Results and Discussion
The results have been divided in three main subsec-
t [mm] 0.24 0.60 1.60
tions: analytical results, experimental results and
computational results. To simplify the reading,

6
 The results obtained are presented in Table 3.
Table 2: Tests performed for each specimen.
The errors are relative to the data-sheets values
Joint Type given in Table 1.
Adhesive Treatment NJ RJ
AF 163-2 BSAA 7 7 Table 3: Relative error for each solution.
AF 163-2 SAA 7 7 AF 163-2 EA 9658 AERO
EA 9658 AERO BSAA 7 7 Solution Pf [N] rel [%] Pf [N] rel [%]
EA 9658 AERO SAA 7 7 Volk 5622 54.22 8513 31.90
GR 3677 70.06 6207 50.34
when conclusions are identical for both adhesives, HS (Elast) 3627 70.47 5826 53.39
graphical plots are only presented for the AF 163-2. HS (Plast) 6957 43.35 11962 4.30
Numerical values will be presented for both adhe- LSA 9153 25.47 11422 8.62
sives.
4.1. Analytical Results Most of these predictions present really high rel-
A theoretical analysis of the SLJ was the first step of ative errors, specially all the values obtained using
this work. Using the classical solutions presented, the perfectly elastic classical solutions. The plastic
with an applied load of 6000 N, the distributions formulation proposed by HS is much more accurate
presented in Figure 7 are obtained. The curves than the other classical solutions, due to the afore-
are quite similar for the EA 9658 AERO, exception mentioned consideration of the adhesive plasticity.
made for the plasticized area, which is lower. Also, despite the simplicity, the LSA proved to be a
method with a very good accuracy. In fact, for the
EA 9658 AERO adhesive, the relative error is only
8.62 %.
4.2. Experimental Results
Prior to the tests, measures to the adhesive thick-
ness and fillet length must be performed for each
specimen - Table 4. This values are quite impor-
tant model computationally the geometry of the
SLJ As it can be seen, there are some differences
between the theoretical and measured thickness val-
ues, which could lead to errors.

Figure 7: Shear and Peel Stress distributions for the Table 4: Average measures of ta and lf il
AF 163-2 adhesive. Adhesive ta [mm] lf il [mm]
Analyzing Figure 7, one can see that Volkersen EA 9658 AERO 0.425 ± 0.045 7.082 ± 0.913
solution is clearly different from the other two, due AF 163-2 0.263 ± 0.053 7.468± 0.911
to the non consideration of the bending moment
in the joint. Furthermore, as stated, the peel and The results obtained, for the apparent failure
shear peaks occur in the borders of the adhesive, not shear load of each specimen, are presented in Figure
satisfying the shear stress free boundary condition 8.
[28]. Also, When the adhesive is assumed elastic- Analyzing Figure 8, it is clear that, when applied
perfectly plastic, Hart-Smith method shows how the to NJ specimens, the AF 163-2 presents a better me-
adhesive borders start to plasticize in shear. chanical behavior when compared with the EA 9658
With these distributions, a first estimation of the AERO. The best results are obtained for the SAA
failure load can be obtained. These values are cal- NJ, bonded with the mentioned AF 163-2, with a
culated using an incremental load approach, com- value of τf = 24.96 MPa. This value represents an
bined with an adequate failure criteria. Essentially, improvement of 8% relatively to the old adhesive,
a maximum shear or peel stress criterion was chosen meaning that the goal of finding a solution at least
for all the classical solutions, except for the plastic as good as the old one was achieved.
formulation of Hart-Smith, where a maximum shear However, when applied to RJ, the mechanical be-
strain was selected. In addition, specially due to its havior of the AF 163-2 is really bad, being surpassed
simplicity, the failure load obtained by the Limit by the EA 9658 AERO. Even so, due to the fact that
State Analysis is also calculated for each adhesive this kind of test was not performed by the manufac-
[29]. turer, it is impossible to infer if the EA 9658 AERO

7
 Figure 10: Stress distributions, obtained computa-
tionally for G1 and G2, using both adhesives.
Figure 8: Apparent failure shear stress for both ad-
hesives.
Table 5: Failure loads obtained.
AF 163-2 EA 9658
brings any kind of improvement in bonding RJ rel- Pf [N] G1 G2 G1 G2
atively to the Hysol EA 9689. Furthermore, due to Datasheet 12281 12500
the low values obtained, the security of a repaired PFM 5151 8971 6093 11142
joint cannot be guarantee, and further tests shall Rel. Error [%] 58.06 26.95 51.26 10.86
be performed.

4.3. Computational Results 26.95% and 10.86% for adhesives AF 163-2 and EA
For the computational analysis, the FEM was used. 9658 AERO respectively, the results are considered
The type of FE chosen was the PLANE82, defined quite acceptable.
by by eight nodes, having two degrees of freedom In Figure 11, the computational and experimen-
at each node, having plasticity and large deflection tal load-displacement curves, for the AF 163-2 ad-
strain capabilities. The BC used to obtain stress hesive, are compared. The differences between the
distributions can be seen in Figure 9. For the fail- curves are evident, being only similar for small dis-
ure load and the load-displacement curves, the dis- placements. This discrepancy may be explained by
tributed load must be replaced by a BC of displace- the fact that, since the applied surface treatments
ment, since the experimental tests are performed are not the advised, that the specimens presents
with a constant displacement rate. some loss of adherence, presenting less resistance to
loading than expected. If compared with the curve
obtained for a SLJ treated with CAA [30] (green
curve in Figure 11), it can be seen that the slopes
are quite similar, which corroborates the hypothesis
of adherence loss.

Figure 9: Boundary conditions of the SLJ [21].

Also, as stated in Subsection 2.2, the SLJ shall be

modelled with the adhesive fillet, as well as with the
rounded adherent corners. Two different geometries
were then considered in this analysis: the normal
SLJ (Geometry 1, or G1) and the SLJ modelled as
previously stated (Geometry 2 or G2).
Using the Centerline Method, the stress distribu- Figure 11: Comparison between load-displacement
tions for G1 and G2 were obtained in Figure 10, curves for the AF 163-2.
highlighting the great reduction in in the maximum
values of shear and peel stresses in both adhesives. Finally, the failure of the SLJ is compared in
This indicate that, in principle, the failure loads will Figure 12 and several resemblances can be found.
be higher for G2. Numbers 1 and 2 represents similar zones of the
Applying the PFM described in Subsection 2.2, fracture. Zone 1 represents the location of the fillet
the failure loads presented in Table 5 were obtained. when failure occurs. As it can be seen, the major
These values must be used to validate the model. part of the fillet stays bonded to the respective ad-
With the results presented in Table 5, it is clear herent, with the crack occurring near the overlap
that Geometry 2 must be the one used to calculate area. Zone 2 represents some adhesive that stays
the failure load of the SLJ. With relative errors of bonded in the face of the adherent, showing once

8
 to obtain the correct mechanical properties of each.
This would certainly help to a better and more ac-
curate modulation of the adhesive layer. Plus, the
ASTM D 1002 should be performed once again, this
time with the respective advised surface treatments.
Furthermore, the model must be applied to some
other bonded structures. The ANSYS APDL code
was developed in order to be applied to any bonded
geometry, so this would be a very interesting study.

References
Figure 12: Comparison between experimental (on [1] Robert D. Adams and William C. Wake. Struc-
top) and computational (on bottom) failures for a tural Adhesive Joints in Engineering. Chap-
EA 9658 AERO specimen. man and Hall, 2nd edition, 1997. ISBN:978-0-
412-70920-3.

again concordance between both results. [2] Jun Cui. Finite element modeling of adhesive
With all the results presented, one can conclude failure with adherend yielding. Master’s thesis,
that the adhesive behavior, for the SLJ, has been University of Toronto, 2001.
reproduced computationally with a relatively good
degree of accuracy. [3] Ahmet Calik. Effect of adherend shape on
stress concentration reduction of adhesively
5. Conclusions bonded single lap joint. Engineering Review,
Regarding the first objective, a new combination of 36(1):29–34, January 2016.
adhesive and surface treatment, at least as good as
the old one was found, has indeed been found. SAA [4] Olaf Volkersen. Nietkraftverteiligung in zug-
NJ, bonded with AF 163-2, managed to achieve a beanspruchten nietverbindungen mit konstan-
result 8% better than the old combination. How- ten laschenquerschnitten. Luftfahrtforschung,
ever, it presented a very poor behavior when applied pages 15–41, 1938.
to RJ, whereby the appliance of this adhesive in this
type of joints can jeopardize the safety of the struc- [5] Goland and Reissner. The stresses in cemented
ture. Despite the improvement revealed by the EA joints. Journal of Applied Mechanics, 11:A17–
9658 AERO in RJ, it is not possible to guarantee A27, March 1944.
the safety of this type of joints.
Said so, and even knowing that the results ob- [6] Ll John Hart-Smith. Adhesive-bonded single-
tained by the ASTM D 1002 are quite significant, lap joints. Technical Report NASA CR-112236,
some other tests shall be performed to ensure the National Aeronautics and Space Administra-
quality of the bonding. One must note that, de- tion, jan 1973.
spite being used in aluminum-to-aluminum joints,
[7] ] L. F. M. da Silva and R. D. S. G. Campilho.
the adhesive chosen will also be applied in honey-
Advances in numerical modelling of adhesive
comb sandwich panels, so the behavior of both AF
joints. SpringerBriefs in Computational Me-
163-2 and EA 9658 AERO must also be tested in
chanics, 2012. DOI: 10.1007/978-3-642-23608.
this type of structures.
Relatively to the second part of the work, it was [8] Yinhuan Yang. Stress analysis and failure pre-
completed with a good degree of success. Using just diction of adhesively bonded ingle-lap lami-
the stress-strain curve of both adhesives, it was pos- nates joints subjected to the tensile loading.
sible to model the adhesive behavior. There were, IJSH International Journal of Smart Home,
however, some results that were not totally in ac- 11:19–28, 2015.
cordance with the observations, specially the curves
from the ASTM D 1002 test, which can be related [9] R.D. Adams and R. Davies. The Mechanics of
with the use of different surface treatments from the Adhesion, chapter Strength of lap shear joints.
ones advised by the manufacturers. This may cause Elsevier, Amsterdam, 2002.
a lower adherence between adhesive and substrates,
compromising the curves that would be obtained for [10] R.D. Adams and N.A. Peppiatt. Effects of
perfect adherence. poissons ratio strains in adherend on stresses
With this in mind, some more experimental tests of an idealized lap joint. The Journal of Strain
should also be performed for validation. Firstly, the Analysis for Engineering Design, 8(2):134–139,
bulk test should be performed for both adhesives, April 1973. DOI: 10.1243/03093247V082134.

9
[11] Tongyan Yue and M. Abdel Wahab. Finite Journal of Materials Science and Engineering,
element analysis of stress singularity in par- 3(2):21–28, 2015. DOI:10.12691/ajmse-3-2-1.
tial slip and gross sliding regimes in fret-
ting wear. Wear, 321:53–63, December 2014. [21] Yinhuan Yang. Stress analysis and failure pre-
doi:10.1080/10618560701678647. diction of adhesively bonded single-lap lam-
inates joints subjected to the tensile load-
[12] X. Zhao, R.D. Adams, and L.F.M. da Silva. ing. International Journal of Smart Home,
Single lap joints with rounded adherend 9(11):19–28, 2016. ISSN: 1975-4094 IJSH.
corners: Stress and strain analysis. Jour-
nal of Adhesion Science and Technology, [22] John Tomblin, Waruna Seneviratne, Paulo Es-
25(8):819–836, February 2011. DOI: cobar, and Yap Yoon-Khian. Shear stress-
10.1163/016942410X520871. strain data for structural adhesives. Final
Report DOT/FAA/AR-02/97, Wichita State
[13] D.C. Noorman. Cohesive zone modelling in University, Wichita, KS 67260, Nov 2002.
adhesively bonded joints: Analysis on crack
[23] D. S. Cobb, M. A. Vargas, T. A. Fridrich, and
propagation in adhesives and adherends. Mas-
M. R. Bouschlicher. Metal surface treatment:
ter’s thesis, Delft University of Technology, Sep
Characterization and effect on composite-to-
2014.
metal bond strength. Operative Dentistry,
[14] Mohammed Waseem H.S. and Kiran Kumar 25:427–433, 2000.
N. Finite element modeling for delamination
[24] Austin Vaillancourt and Travis Abele. Ad-
analysis of double cantilever beam specimen.
hesive technology: Surface preparation tech-
International Journal of Mechanical Engineer-
niques on aluminum. Technical report, Worces-
ing, 1(5):23–33, September 2014.
ter Polytechnic Institute, 2006.
[15] R. D. S. G. Campilho, M. D. Banea, A. M. G.
[25] Edward M. Petrie. Handbook of Adhesives and
Pinto, L. F. M. da Silva, and A M P de Je-
Sealants. The McGraw-Hill Companies, Inc.,
sus. Strength prediction o fsingle - and double
second edition, 2007. ISBN: 9780071479165.
- lap joints by standard and extended finite
element modelling. International Journal of [26] ASTM Standards. Standard Test Method for
Adhesion & Adhesives, 31:363–372, February Apparent Shear Strength of Single-Lap-Joint
2011. doi:10.1080/10618560701678647. Adhesively Bonded Metal Specimens by Ten-
sion Loading (metal-to-metal), Oct 1999.
[16] Fiamegkou Ch. Eleni. Development of
improved, multi-functional, nano-structured [27] Sofia Teixeira de Freitas and Jos Sinke. Tem-
polymer based adhesives with applications in perature effect on the static behaviour of
the bonding of composite components and the adhesively-bonded metal skin to composite
repair of engineering structures with composite stiffner. ABCM International Congress of Me-
patches. Master’s thesis, University of Patras, chanical Engineering, 23, December 2015.
2015.
[28] J. A. Harris and R. D. Adams. Strength predic-
[17] Mahmood Shokrieh. Failure of laminated com- tion of bonded single lap joints by non-linear
posite pinned connections. Master’s thesis, finite element methods. International Journal
Iran University of Science and Technology, Jul of Adhesion and Adhesives, 4(2):65–78, April
1991. 1984. DOI: 10.1016/0143-7496(84)90103-9.

[18] Michael Maletta. The development of a pro- [29] A. D. Crocombe. Global yielding as failure cri-
gressive failure model of a fiber-reinforced com- teria for bonded joints. International Journal
posite lamina. Master’s thesis, Grand Valley of Adhesion and Adhesives, 9(3), 1989.
State University, Apr 2008.
[30] Miguel Pereira Palmares. Strength of hybrid
[19] Konstantinos N. Anyfantis and Nicholas G. laminates aluminium carbon-fibre joints with
Tsouvalis. Post buckling progressive failure different lay-up configurations. Master’s thesis,
analysis of composite laminated stiffened pan- Faculdade de Engenharia da Universidade do
els. Applied COmposite Materials, February Porto, July 2016.
2011. DOI 10.1007/s10443-011-9191-1.

[20] Khudhayer J. Jadee. Progressive failure anal-

ysis and failure map into plain weave glass fi-
bre reinforced polymer bolted joint. American

10