Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Analysis of The Dynamic

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

International Journal of Adhesion & Adhesives 118 (2022) 103219

Contents lists available at ScienceDirect

International Journal of Adhesion and Adhesives


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

Numerical analysis of the dynamic behaviour of adhesive joints: A review


L.D.C. Ramalho a, *, Isidro J. Sánchez-Arce a, Diogo C. Gonçalves a, J. Belinha a, b,
R.D.S.G. Campilho a, b
a
INEGI, Institute of Mechanical Engineering, Rua Dr. Roberto Frias, 4200-465, Porto, Portugal
b
School of Engineering, Polytechnic of Porto, ISEP-IPP, Rua Dr. António Bernardino de Almeida, 431, 4200-072, Porto, Portugal

A R T I C L E I N F O A B S T R A C T

Keywords: Adhesive joints are being increasingly used in various industries, including the automotive or the wind turbines
Adhesive joints industries. Such increasing interest is a direct result of its high structural efficiency and also the product of its
Dynamic behaviour related scientific research. Therefore, the state-of-the-art on adhesive joints is significantly expanding. The
Failure
current work aims to discuss the most recent works dedicated to the numerical analysis of the dynamic behaviour
Fatigue
of adhesive joints. Dynamic behaviour was divided into three separate fields: fatigue, variable strain rate and
Impact
Modal analysis impact, and modal analysis. It was found that Cohesive Zone Models are a popular approach to study fatigue,
variable strain rates and impact. Additionally, the available literature focused on fatigue and impact is more
extensive than the one focused on modal analysis. Overall, it was found that the available research on the nu­
merical analysis of the dynamic behaviour of adhesive joints is increasing at a solid rate, and many geometrical
and material variations have been tested numerically. With this review designers and researchers of adhesive
joints should be able to choose the most suitable numerical technique for their specific dynamic analysis.

1. Introduction namely: fatigue, discussed in Section 2; variable strain rate and impact,
discussed in Section 3; and Modal Analysis, discussed in Section 5. There
Due to environmental restrictions, which are increasingly tighter, are several numerical approaches that can be used to analyse each dy­
various transport industries seek lightweight materials, such as namic problem. Continuum mechanics constitute the simplest approach,
aluminium and composites. One the other hand, these industries they allow the calculation of stresses and strains, being important to
frequently use different materials in the different parts that have to be identify critical areas. The stress and/or strain determined using a
joined. Under these conditions, adhesives offer some advantages over continuum mechanics approach can also be used to predictions the life,
other joining methods. The behaviour of adhesive joints in static or in fatigue, or strength, in impact, of joints. However, since adhesive
quasi-static loading rates has been extensively studied over the years as joints generally have stress singularities at the adhesive/adherent in­
several reviews show [1,2]. However, their behaviour in dynamic terfaces these predictions can be mesh dependent. In fatigue, fracture
loading conditions has been studied less extensively. This occurs mechanics can also be used. With this approach, it is possible to deter­
because dynamic analysis is more complex than static analysis, partic­ mine crack growth and predict the life of joints. This approach is not
ularly impact problems which require explicit integration. In recent affected by the singularity at the adhesive/adherent interface, but it
years, this trend is changing, as the applications of adhesive joints in­ normally requires an initial crack. Damage mechanics is another
crease and their behaviour during fatigue and impact require further possible approach, and it has been used in numerical analysis of impact.
studies to develop better applications. Consequentially, the modal This approach is similar to the continuum mechanics approach, but the
behaviour of adhesive joints also needs to be studied to ensure that their regions of the joint subjected to higher stresses/strains are progressively
working conditions are outside the resonant frequencies and to assess degraded, according to a material law, decreasing their ability to carry
their damping capabilities. load, and in extreme cases leading to failure. The main drawback of this
This work aims to provide an extensive review of the most recent approach is the same as the continuum mechanics drawback, the stress
works dedicated to the numerical dynamic analysis of adhesive joints. singularity at the adhesive/adherent interfaces. CZM is a popular
Dynamic analysis was divided into three main fields in this work, approach in both fatigue and impact, combining continuum mechanics

* Corresponding author.
E-mail addresses: lramalho@inegi.up.pt (L.D.C. Ramalho), job@isep.ipp.pt (J. Belinha), rds@isep.ipp.pt (R.D.S.G. Campilho).

https://doi.org/10.1016/j.ijadhadh.2022.103219
Received 3 June 2022; Accepted 15 June 2022
Available online 22 June 2022
0143-7496/© 2022 Elsevier Ltd. All rights reserved.
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

and fracture mechanics concepts. This approach is not affected by the the bolt preload increased the joint’s life and the bonded-bolted joints
stress singularity at the adhesive/adherent interfaces, but some cohesive have a longer life than the bolted joints.
properties are dependent on the thickness of the adhesive layer, which The SWT model was also used by Shahani et al. [12] to predict the
can increase the amount of characterization tests needed. Since modal fatigue life of single lap joints (SLJ), but in this work the fatigue criterion
analysis is an eigenvalue problem, it differs from the other two. In modal was considered in both the adherents and the adhesive. The authors
analysis, the eigenvalues are the natural frequencies, and the eigen­ calibrated the model parameters using common SLJ and then used those
vectors are the mode shapes. A detailed discussion regarding the current parameters to predict the fatigue life of multi-material SLJ with several
state-of-the-art of each topic is presented in Section 5. The conclusions, different adherent thicknesses. The predictions showed a maximum
in Section 6, present a general overview of all the fields of dynamic error of 25% for a load level of 60% of the static failure load. The pre­
analysis covered by this review. dictions were always lower than the experimental life, which is rec­
ommended for safe design. Pereira et al. [13] used the Morrow’s
2. Fatigue modified Manson-Coffin (MMC) and the SWT models to predict the fa­
tigue life of welded-bonded SLJ. An important aspect of this work is that,
The use of numerical methods to study fatigue in adhesive joints can for the fatigue criteria, the stress and strain fields were considered,
be traced back to the 1970s, as shows the work of Mostovoy and Ripling instead of a single critical point. Using either model, the life predictions
[3]. Since then, this field has gathered considerable interest and many were close to the experimental values, but the SWT presented better
scientific reports and papers regarding the subject have been published. predictions.
The 2013 review by Pascoe et al. [4] provides a good summary of many Wu et al. [14] proposed a combined approach, using continuum
of the most significant works published between the first works of the mechanics to determine crack initiation and fracture mechanics to
1970s and 2013. Thus, the present review will mostly focus on works determine fatigue crack growth after initiation. The initiation criterion
published after 2013, with a few exceptions. Now, for fracture numerical was based on the strain in critical places using the
analysis of adhesive joints, three different approaches can be identified: Manson-Coffin-Basquin model (MCB), and the crack propagation crite­
continuum mechanics, fracture mechanics and CZM. rion is discussed in Section 2.2. The authors used this model to predict
the life of SLJ with different overlap lengths and fillets, as well as joints
with and without gaps. These predictions compared well to the experi­
2.1. Continuum mechanics
mental strength and showed that, for shorter overlap lengths, it can take
more cycles to initiate the crack than to grow it until failure.
Using continuum mechanics to study fatigue problems numerically is
In addition to those models, Theories of Critical Distance (TCD) can
the simplest approach. Therefore, it can be useful to perform first esti­
also be used. This classification was first proposed by Taylor [15], who
mations, but for works requiring more precision fracture mechanics or
identified several models that tried to predict fatigue life and static
CZM are generally used. The continuum mechanics approach consists of
strength based on the stress at a fixed critical distance, hence the name.
estimating fatigue life base on the stresses and/or strains in adhesive,
Castro Sousa et al. [16] proposed the use of a TCD to determine the
joints without considering any damage evolution or crack propagation
fatigue life of Arcan joints. In the proposed approach, the critical point
criterion. Continuum mechanics can also be useful just to determine the
was determined as the point in the adhesive mid-thickness plane where
regions of the joint that have the highest stress concentrations, or to
the maximum principal stress is highest. The parameter used to deter­
observe how the stresses and strains are distributed along the joint.
mine failure was a modified version of the effective stress that takes into
Continuum mechanics criteria are most commonly used for metals.
account the hydrostatic stress. To estimate these parameters, experi­
However, some authors have also used them for adhesive joints, but they
ments to Arcan joints under pure mode I and pure mode II loading were
are not as common as CZM. There is a myriad of models to predict fa­
performed to obtain an S–N curve, and then linear elastic numerical
tigue life, of which some only consider stress or strain, while others
simulations were performed to relate that curve to the effective stress at
consider both or energy. Esmaeili et al. [5] performed an extensive
the critical point, getting a σe-N curve. The σe-N curve, obtained based on
comparison of several models, namely the Smith-Watson-Topper (SWT)
the pure mode I and pure mode II experiments, related well with the
model [6], the Kandil, Brown and Miller (KBM) model [7], the Glinka
experimental fatigue life of Arcan joints under mixed mode loading at
model [8], the Fatemi–Socie model [9], the Crossland model [10] and
45◦ . Castro Sousa et al. [17] later adapted this approach to predict the
the Jahed–Varvani model [11]. The authors tested these models in
fatigue life of SLJ with two different overlap lengths and adherent
bolted and bonded-bolted DLJ, considering the fatigue criteria in the
thicknesses, achieving good predictions for all the tested joints, as shown
aluminium adherents. Overall, the Crossland model, which is a
in Fig. 2. However, to predict the life of SLJ a parameter M (which did
stress-based model, presented the life predictions closer to the experi­
not seem to change significantly for the joints tested) was introduced,
mental life, as can be observed in Fig. 1. It was also found that increasing

Fig. 1. Predicted fatigue life with several models against the experimental fatigue life for bolted (a) and bonded-bolted (b) specimens with tightening torque equals
to 5 Nm [5].

2
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 2. Comparison of the fatigue life predictions using M = 0.424 (a) and M = 0.99 (b), where I, II and III are joints with different geometric parameters [17].

but more extensive testing was considered necessary, with different that takes into account the mode mixity. After determining the param­
adhesives, adherents, and geometric parameters, to determine how this eters of this Paris’ law, it was possible to estimate the fatigue life of SLJ.
parameter is affected by these geometrical variations. The disadvantage of this approach is that it does not consider crack
Schneider et al. [18] used a TCD based on the effective stress to initiation, which can be a considerable part of the fatigue life. The
predict the fatigue life of SLJ and Scarf Joints (SJ) at different temper­ previously mentioned work of Wu et al. [14] was able to address this
atures ranging from − 35 ◦ C to 80 ◦ C. These predictions were based on issue by using a combination of fracture mechanics and continuum
the S–N curves of SJ with different angles and Thick Adherent Shear Test mechanics, as discussed in the previous section. Fig. 3 shows that, for the
(TAST). With the SLJ fatigue life predictions it was shown that the shorter overlap, the initiation life is more relevant than the crack growth
critical distance increases as the temperature increases. For all tested life, while for the longer overlap the crack growth life is more relevant.
joints it was also observed that their life decreases with increases in In 2021, Wu et al. [25] also used the J-integral to determine the SERR of
temperature. This study was further expanded by Beber et al. [19], bonded-riveted joints. With this calculation, it was found that, as the
which compared the predictions obtained using different material crack propagates in the adhesive and it gets closer to the rivet, the SERR
models for the adhesive, namely von-Mises, linear Drucker-Prager and decreases. On the other hand, in an adhesively bonded joint the SERR
linear-elastic. It was found that the life predictions were closer to the always increases as the crack propagates.
experimental life when an elastic-plastic material model was used. Be­ Another way to determine the SERR is using the Virtual Crack
tween the different models, the Drucker-Prager model provided the best Closure Technique (VCCT), as performed by Pascoe et al. [26]. These
prediction for the SLJ, and the von Mises criterion provided the best authors experimentally and numerically tested the effects of the adhe­
prediction for the SJ with a 56.6◦ angle. Later, Beber et al. [20] used the sive thickness in Double Cantilever Beam (DCB) tests under mode I fa­
TCD with an elastic-plastic material model to predict the life of SLJ. tigue loading. Using the VCCT to calculate the SERR, it was observed
Initially, the critical parameters were obtained by performing bulk that, as the crack grows, the SERR decreases for a constant imposed
experimental tests of the adhesive notched with different notch lengths displacement. Moreover, the adhesive thickness did not affect the SERR
and locations. Afterwards, with the aid of numerical simulations the values in a significant way. The numerical model also showed that the
intersection of the equivalent stress of those differently notched speci­ plastic zone increases with the adhesive thickness, which in turn in­
mens was determined. This data the used to predict the fatigue life of SLJ creases the dissipated plastic energy, but this effect decreases as the
with the same adhesive. These predictions showed some variation, albeit adhesive thickness increases.
small, depending on which bulk specimen was used as a reference to The SIF is less commonly used than the SERR, since it is more diffi­
make them. However, the predictions were similar to the experimental cult to calculate. However, Lai and Pan [27] used the SIF to predict the
life of the SLJ. life of SLJ with multi-material adherents and an initial crack, where the
SIF was directly obtained from Abaqus®. The predictions were under the
experimentally observed life of the joints. Also using Abaqus®, Samaei
2.2. Fracture mechanics
et al. [28] studied the SIF in bolted and bonded-bolted SLJ with different
bolt preloads and an initial crack in the bolt hole. It was shown that
Fracture mechanics are ideal to calculate fatigue life based on the
increasing the preload and using the bonded-bolted joints resulted in
fatigue crack growth, since they are focused on studying cracks. Fatigue
lower SIF, even as the crack propagates. This was proposed as the reason
crack growth can be generally determined using the Stress Intensity
why longer lives were observed experimentally for the bonded-bolted
Factor (SIF) [21] or its energy equivalent, the Strain Energy Release Rate
joints with the highest bolt preload.
(SERR) [22].
Another possibility in fracture mechanics is the used of the Gener­
The SERR is more commonly used that SIF, and there are several
alized Stress intensity Factor (GSIF), also known as Intensity of Singular
ways to determine the SERR. Chen et al. [23] determined the fatigue life
Stress Fields (ISSF). The ISSF allows the study of the singularities arising
of U-joints by calculating the SERR using the J-integral with the aid of
from sharp multi-material corners, which are observed in many adhesive
the Finite Element Method (FEM) and relating it to the SERR of SLJ and
joints, such as SLJ, as studied by Tang et al. [29]. In their work, it was
T-peel joints. The predictions were slightly over the experimental life in
shown that the fatigue crack initiation of SLJ with thick adhesive layers
the high-cycle region, but very close in the low-cycle region. Later, Chen
occurred at around the same ISSF range, for two different adhesive
et al. [24] also used the SERR to predict the fatigue crack growth. This
thicknesses, as seen in Fig. 4. This finding indicates that, when using the
was done by performing an extensive experimental investigation of
ISSF to determine fatigue life, the ISSF is independent from the adhesive
Compact Mixed Mode specimens, under several mode-mixities, from
thickness, but a wider range of adhesive thicknesses would have to be
pure mode I to pure mode II. Then, the experimental investigation was
tested to assess the universality of this finding.
combined with numerical simulations, to allow the characterization of
the fatigue crack growth behaviour using an adaptation of Paris’ law

3
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 3. Comparison of the fatigue life predictions for a joint with a 5 mm (a) and a 25.4 mm overlap (b) [14].

expected on account of their ductility. The predictions were good for all
humidity conditions, but they were made only for the joints used to
calibrate the model.
In some works, triangular CZM have been used to determine the
fatigue life of joints used to estimate the cohesive parameters, such as
DCB and End-Notched Flexure (ENF). This type of study is important to
ensure that the cohesive parameters have been correctly determined.
Ebadi-Rajoli et al. [31] used triangular CZM to predict the fatigue crack
growth rate of DCB, ENF and mixed mode bending (MMB) joints with
composite adherents, achieving results comparable to the experimen­
tally obtained Paris’ law curve. Akhavan-Safar et al. [32] also used
triangular CZM in pure mode I DCB test, and Monteiro et al. [33] and
Akhavan-Safar et al. [34] in pure mode II conditions with ENF speci­
mens. Additionally, Rocha et al. [35] used the same type of cohesive law
in mixed-mode loading conditions, using SLB tests.
De Moura and Gonçalves [36] used triangular CZM to simulate the
Fig. 4. Number of cycles to crack initiation against the ISSF for SLJ with two fatigue crack growth of a DCB joint, using a modified Paris law, pro­
different adhesive thicknesses [29]. posed by Pirondi and Moroni [37], to degrade the cohesive elements as
fatigue accumulates. This model was not compared against experimental
2.3. Cohesive Zone Models results, but the purely numerical study was able to correctly identify the
effect of different load ratios and modified Paris’ law parameters. Higher
The review about static strength prediction of adhesive joints by load ratios and lower Paris’ law parameters lead to a longer fatigue life.
Ramalho et al. [1] showed that CZM were the most commonly used The same authors applied a similar model to mode II fatigue loading and
approach to predict joint strength. Regarding fatigue life predictions, the tested it on ENF joints [38], achieving similar conclusions. Later, the
same is shown to be true. CZM are a hybrid technique based on the authors extended this approach to mixed-mode loading conditions [39]
introduction of special purpose (cohesive elements) in the models at using Single-Leg Bending (SLB) joints. This model was shown to
specified fracture paths to promote automatic and self-similar fatigue compare well with the Paris Law used as the input, when considering the
damage growth. With this purpose, this technique uses continuum me­ crack growth per number of cycles. Moreira et al. [40], and de Moura
chanics and fracture mechanics concepts. Continuum mechanics is used and Moreira [41], later used this model to predict the fatigue life of
to initiate damage in un-cracked structures, while crack propagation is several composite bonded repairs, namely Single and Double Strap
based on fracture concepts, typically with Paris-like laws, Thus, CZM Joints (SSJ and DSJ), one and two stepped joints, and SJ. It was
actually uses concepts from the approaches of Sections 2.1 and 2.2 to concluded that the one-stepped joint had the shortest life, because it
provide a more efficient solution. CZM define a cohesive law that relates developed considerable early damage, while the SJ with a reinforcement
a certain stress and displacement with damage, which goes from 0 (un­ offered the best fatigue life. However, it is important to note that these
damaged) to 1 (complete failure). As damage increases the mechanical results are purely numerical.
properties of the material decrease. There are several types of cohesive Hosseini-Toudeshky et al. [42] performed a comparison between
law that can be used, classified according to their shape, but the most CZM and an approach based on the VCCT to evaluate the fatigue crack
commonly used is the triangular law, since it is implemented on com­ growth in an adhesive joint under mixed mode loading. Their work
mercial FEM software, such as Abaqus®. However, the triangular law is showed that the VCCT and CZM predicted a similar crack evolution, but
not the most appropriate for significantly ductile materials, as many the CZM was 240 times faster computationally. These predictions were
modern adhesives are. To simulate fatigue behaviour of adhesive joints not compared to experiments, but Khoramishad et al. [43] performed
using CZM, a cycle jumping strategy is usually employed. This approach that comparison with SLJ and laminated doublers in bending, achieving
consists of applying a constant load to the joint and consider that the a good correlation between the predicted fatigue life and the experi­
joint was subjected to a given number of cycles with that load and mental fatigue life. Belnoue et al. [44] determined the fatigue life of DLJ
degrading the mechanical properties accordingly. Alternatively, each with different adhesive thicknesses by using cohesive elements at the
cycle can be simulated individually. However, since this is a very adherent/adhesive interface, achieving good predictions. Ibrahim et al.
computationally heavy approach, it is not often seen. [45] used triangular CZM to predict the fatigue crack growth rate of SLJ.
A comparison between different cohesive law shapes was performed These authors predicted the crack size after 2000 cycles and 4000 cycles
by Costa et al. [30], who used CZM to predict the fatigue life of DCB with good accuracy, i.e., a 5.2% error in the first case and a 4.6% error in
degraded by humidity. The authors concluded that the trapezoidal law the second case.
was the most appropriate for the two adhesives tested, which was CZM have also been used in several works to assess the crack

4
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

retardation effects of several solutions. Floros and Tserpes [46] inves­ probabilistic life predictions: first the Fatigue Damage Growth (FDG)
tigated the crack retardation effect of introducing bolts into composite simulator, then the Uncertainty Quantification (UQ), and finally the
adhesive joints using triangular CZM. The numerical solution showed confidence bounds of the fatigue life prediction. The first module uses
that the bonded-bolted joints are damaged faster, i.e., in fewer cycles, the mapping to simulate the FDG in the adhesive; the second module
until damage reaches the bolted zone, which slows down the damage. employs Bayesian inference to calibrate the fatigue model parameters by
This effect can be observed in Fig. 5. The overall results showed that the quantifying uncertainties in the parameters using experimental data;
bonded/bolted joints have slower damage propagation than the bonded and the final module uses those parameter samples in the FDG simulator
joints. Tserpes and Floros [47] also tested this model against experi­ and attributes a confidence interval for each parameter sample. The
mental values for adhesively bonded joints, showing that the crack model was successfully validated by calibrating the parameters using
growth predicted by the model was similar to the experimentally SLJ experiments and then performing life predictions for the LDB, using
observed crack growth. Sachse et al. [48] introduced rivetless nut plates the previously determined parameters. The predictions of the LDB seen
into adhesive joints to assess their crack growth retardation abilities in in Fig. 6 agree well with the experimental data. Two other validation
fatigue. The adhesive was modelled using a trapezoidal CZM, due to the tests were successfully performed but are not described here for the sake
adhesive’s ductility. The numerical results showed that the crack growth of brevity.
rate was reduced by the rivetless nut plate, which was in accordance
with the experimental results for the three different load levels tested. 3. Variable strain rate and impact
Later, Sachse et al. [49] also investigated the crack-arrest capabilities of
bonded-pin joints and bonded-bolted joints. For the bonded-pin joints Impact is a field that has been gaining interest in recent years due to
the numerical model predicted a crack growth similar to the experi­ the increased use of adhesive joints in transportation industries. The
ments, showing that the rivetless nut plates had stronger crack-arrest review of Machado et al. [53] presents a good summary of the numerical
effect than the pin. Regarding the bonded-bolted joints, when the modelling of adhesive joints under impact until 2016. Therefore, the
bolts had no preload, the numerical model predicted a crack growth current review will focus mainly on works published after that year,
similar to the experiments. However, if the bolt was preloaded, some with occasional exceptions, to avoid repetition. Some of the works
corrections were needed to achieve a crack growth similar to the ex­ considered in this section are not dealing with impact, instead they deal
periments, namely considering the effects of the compression introduced with variable strain-rate (non-static) loading conditions, which involve
by the bolt and the friction. Compared to the rivetless nut plates, dynamic behaviour and also require explicit numerical integration. Due
introducing a bolt with a preload had similar crack arrest properties, but to the small number of such works, they do not have a dedicated section
using bolts with no preload resulting in a significantly worse and are included here. Currently, variable strain rate and impact studies
performance. can be divided into three different categories: continuum mechanics,
Triangular CZM have also been used to predict the fatigue life of damage mechanics and CZM.
components used commercial products, not just idealised joints. An
example is the work of Zuo et al. [50], which studied the fatigue life of 3.1. Continuum mechanics
wind turbine blades using a continuum damage model for the blade
components and a triangular CZM for the adhesive joining the blade Continuum mechanics are mainly used to assess the stresses and
components. It was found that damage initiated firstly in the blade strains of adhesive joints. However, some authors also perform strength
components and later in the adhesive, causing debonding. Both types of predictions based on that data. Performing strength predictions with this
damage initiated around the same region of the blade and the adhesive approach presents a significant difficulty, which is the singularity arising
damage was mainly caused by shear stresses, although there was also a from the interface corners present in most adhesive joints. However,
significant amount of peel stress. Hosseini-Toudeshky et al. [51] also predictions with this approach are easier to program and can provide
used triangular CZM to study the fatigue life of another wind turbine good results. Unlike static load conditions, under impact loading the
component, i.e. the blade root joint. In that work, a CZM was used to stress propagates as a wave, which leads to several local stress peaks
assess damage evolution after an overloading event. This initial event along the adhesive. This peculiarity is discussed and demonstrated in
caused a damage of up to 8% of the joint height, which after a month of several continuum mechanics works where a stress analysis is performed
cyclic loading would grow to 40% of the joint height. [54–56], and it can be visualized in Fig. 7. However, failure still usually
A unique approach in this field was recently proposed by Lyathakula occurs at the overlap ends in SLJ and similar joints because the overall
and Yuan [52], who combined Artificial Neural Networks (ANN) and a peak stress still occurs in that region, this can also be observed in Fig. 7.
triangular CZM to predict the fatigue life of adhesive joints. This model Boling and Dongyun [57] performed a stress analysis of SLJ with
was tested in three different types of joints: SLJ, Laminated Doublers in different adherents using a 3D Finite Element Analysis (FEA) consid­
Bending (LDB) and Tapered SLJ (TSLJ). First, the ANN was trained using ering linear-elasticity. The joints always used the same adhesive, but
FEM simulations with the CZM, mapping the damage to the principal three different adherent combinations were studied, namely
strain at the adhesive. Then, three successive modules were used for the composite-composite (CC), composite-aluminium (CA) and

Fig. 5. Predicted fatigue crack growth in the adhesive layer of bonded and bonded-bolted joints [46].

5
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 6. Life prediction for the SLJ (a) and the LDB (b) [52].

more vulnerable to the heat than the sandwich adherent used.


Furthermore, the degradation of mechanical properties due to the heat
had a significant effect on the stress curves. In 2020, Shim et al. [62]
developed a model that takes into account temperature and
rate-dependent material behaviour. In this work, the authors tested SLJ
at several loading and temperature conditions and then compared the
experimental results with the results from the numerical model. It was
shown that increasing the strain rate leads to a higher joint strength,
while the inverse happens when the temperature is increased. The fail­
ure strength predictions made using a strain-based criterion correlated
well with the experimental failure strength for a wide range of strain
rates.

3.2. Damage mechanics

Damage mechanics can be used in to predict the impact strength of


adhesive joints, but it is not as popular as CZM. This technique allows the
simulation of progressive material degradation and ultimate failure
Fig. 7. Stress propagation in the adhesive of a SLJ [56].
through an arbitrary path, unlike CZM, which only allow damage in the
predefined cohesive elements.
composite-titanium (CT). The results showed that the CA and CT joints
Gollins et al. designed a joint to predominately be under mode I
promoted higher stress concentrations and an asymmetric stress distri­
loading and tested it under impact experimentally [63] and numerically
bution at the adhesive mid-thickness line, which explains the experi­
[64]. The developed damage model was able to accurately predict the
mentally observed lower strength of those joints. Quaglini et al. [58]
impact strength when compared to the experimental results for several
numerically tested glass plates with point fixings considering two
different impact velocities and two different adhesives. The comparison
different joining methods, i.e. adhesives and bolts. The work was mainly
with quasi-static loading found that, under impact loading, the joints
focused on the stress, and it showed that the bolts induced large stress
were able to absorb more energy and support higher loads. The differ­
concentrations in the glass, which did not occur when adhesives were
ence in energy absorbed is particularly high as is shown in Fig. 9. Yildiz
used.
et al. [65] used a damage model to observe the strain rate effect on the
The stress singularities observed at the interface corners of adhesive
energy absorbed by an adhesive joint subjected to a shockwave impact.
joints can be mitigated. Valès et al. [59] compared the effects of different
It was found that the energy absorbed was more significant for
adhesive/adherent corner geometries on the reduction of stress singu­
quasi-static loads than for impact loads.
larities. This was done by comparing the stresses in the adhesive, near
A damage model was used by Boling and Dongyun [66] to study the
the interface, of joints with: a straight corner, an inside taper, a cham­
behaviour of composite-titanium SLJ. A 3D analysis was considered, and
fered beak, and a filleted beak. The FEA considered linear-elastic ma­
the damage model was only used for the composite adherents, while the
terial behaviour and plane stress. It showed that the inside taper
adhesive and the titanium used an elastic-plastic material model with no
removed the stress peaks at overlap ends, but created other peaks before
damage. A comparison of the numerical and experimental transverse
it, being overall not very beneficial; while both the chamfered beak and
strains showed a good correlation. Numerically, different adhesive
the filleted beak removed the stress peaks without introducing new ones.
thicknesses were tested, showing that peak compressive and tensile
The differences between the configurations can be visualized in Fig. 8.
stresses decrease as thickness increases. A similar effect was also found
Using a similar FEA, they studied an Arcan specimen, with additional
for overlap length increases. The obtained results also showed that it is
details and experimental tests in Ref. [60]. In summary, this work
important to account for the material properties’ strain rate dependency
showed that the mass of the impactor has a significant influence on the
in impact cases, since otherwise significantly different results will be
stability of the stresses, with lighter weights a significant oscillation was
obtained. Bedon et al. [67] tested an adhesively bonded square hollow
observed, which was also confirmed experimentally.
glass column under different impact velocities. As in the work of Boling
The effect of temperature on the mechanical response of adhesive
and Dongyun [66], the damage model was not applied to the adhesive,
joints is also a field of interest. Rizk et al. [61] developed a 3D ther­
instead it was applied to the glass adherents. The numerical compressive
momechanical model where a DLJ was subjected to heat and then
stress was similar to the one observed experimentally for the different
impacted. The developed model showed that the adhesive was much
impact velocities when considering initially undamaged columns.

6
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 8. Stress in the adhesive near the upper interface at the instant with the highest shear stress for the inside taper (a), chamfered beak (b), filleted beak (c) and
filleted beak with a shoulder thickness (d) [59].

prediction of the peak load, as shown in Fig. 10.

3.3. Cohesive Zone Models

As was observed for fatigue, in Section 2, and static loading condi­


tions, in Ref. [1], CZM are also the most commonly used method to
determine the strength of adhesive joints under impact. Under impact,
CZM are used in a similar manner to static loading, but many of the
properties used to define the cohesive law are strain rate dependent.
Thus, the characterization tests have to be performed at different strain
rates, and the properties determined with static tests are usually not
applicable to impact. Due to this strain rate dependence, it is important
to note how the CZM properties are determined how they are modelled.
Regarding the experimental tests, as in quasi-static characterization
experiments, to determine the mode I dynamic properties DCB are
usually used and to determine mode II properties, ENF tests are the most
commonly adopted. The only difference is the strain rate at which the
tests are performed. Most authors perform tests at two different loading
Fig. 9. Comparison between the energy absorbed by a joint under static and velocities and then perform a logarithmic extrapolation to obtain the
impact loading conditions [64]. properties at the numerical model’s velocity, meaning that the proper­
ties will then remain the same throughout the complete simulation.
However, if the columns were initially damaged the numerical model Some examples of such experimental trials and subsequent numerical
overestimated the stress, mainly due to the simplifications in modelling models can be found in references [69–73], meaning that the properties
the initial damage. will then remain the same throughout the whole simulation. However,
This type of damage model was also applied recently to a bonded some authors have adopted a more complex approach, where a loga­
automotive structure by Hou et al. [68], testing it with two different rithmic law is defined and the cohesive properties are updated at each
joining methods, i.e. adhesive bonding and riveting. The tested structure time-step according to the current strain rate, examples of this are found
was a single hat-shaped T-joint, and the difference between the two in Refs. [74,75]. Generally, the only CZM properties that are affected by
joining methods was observed to be minimal, through experiments and the strain rate are tn and ts, which tend to increase with strain rate in­
numerical results, in terms of energy absorbed and peak load, for various creases, while GIc, GIIc, E and G do not change with strain rate [69–73].
impact velocities. The numerical time-load curves of both joining Numerical investigations into joints used to characterize adhesives
methods were also similar to the experiments, but with a small over are also important, to validate the characterization. Nunes et al. [76]

7
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 10. Time-load response of the adhesively bonded (a) and the riveted (b)automotive structure [68].

numerically tested DCB and ENF specimens at several low displacement Under impact, SLJ are also one of the most commonly studied types
rates, i.e. 150, 15 and 1.5 mm/s, to assess how the strain rate changes of joints. Araújo et al. [83] employed a triangular CZM to evaluate the
with time and as the crack grows. It was found that, for the ENF test, the load bearing capabilities of composite SLJ under impact. They used
strain rate remained stable for the test range generally used to determine cohesive elements for the whole adhesive and for a small region of the
GIIc. However, for the DCB it was shown that the strain rate decreases as adherent near the interface, to simulate delamination. The maximum
the crack propagates for a constant displacement rate, which was also load predicted by the numerical model was similar to the experimental
found by the numerical and experimental investigation of Borges et al. one. However, the triangular cohesive law was not able to take into
[77] and Nunes et al. [78], this is also displayed in Fig. 11. Thus, Nunes account the adhesive’s plasticity and, as a result, the P-δ curves were not
et al. [78] used that knowledge to develop a model where the strain rate similar. Another interesting finding of this work, that was verified
was constant at the crack tip. This goal was achieved by having a vari­ experimentally and numerically, was that under impact the joints sus­
able displacement rate. This numerical model then served as a basis for tained a higher load before failing, when compared to quasi-static
an experimental test with a constant strain rate and variable displace­ loading. Jia et al. [74] also used triangular CZM to evaluate the same
ment rate. type of joint. The CZM used considered the variable strain rate observed
Lißner et al. [79,80] developed a trapezoidal CZM capable of taking in impact situations, and this strain rate dependency and its calibration
into account the load-rate and adhesive thickness. The experiments used were performed with numerical and experimental tests to DCB, ENF and
to characterize the cohesive law showed an increase in dissipated en­ TAST specimens. The strength predictions of the composite SLJ were
ergy, in mode I, mode II and mixed-mode loading, with adhesive similar to the experimental strength when using the strain rate depen­
thickness increases. However, the maximum stress did not increase, dent cohesive parameters. The TAST specimens were also tested
which was verified for high rate (3 m/s), medium rate (10 mm/s) and numerically and experimentally under different temperatures in
quasi-static loadings. The comparison between load rates shows that the Ref. [84], with good a correlation between the numerical and experi­
peak stress tends to increase with the rate increase, in all loading modes, mental results for the different temperatures and impact speeds. How­
while the energy dissipation decreases, also in all loading modes. Later, ever, the model was only verified against the experiments used to
Lißner et al. [81] tested multi-material joints, made with titanium and a calibrate it. Liu et al. [85] experimentally tested SLJ under low strain
CFRP, under three different loading conditions: mode I, using a DCB test, rates and used a numerical model with 2D cohesive elements with a
mode II, using an ENF test, and mixed-mode, using a SLB test. The triangular law. The authors determined the adhesive properties using
experimental results showed a decrease in strength for all loading con­ quasi-static experiments and showed that using them for low strain rates
ditions when a high-rate load is applied (4 m/s), as opposed to a still resulted in good strength predictions. A comparison between three
quasi-static load. However, while in quasi-static loading failure was different adhesives, ranging from brittle to strong and very ductile, was
cohesive, in high-rate loading this was not verified for the DCB and SLB performed by Valente et al. [69] on SLJ with steel adherents, using
experiments where failure occurred in the composite substrate. The triangular CZM. A comparison with experimental work showed that this
strength predictions, using the CZM of reference [80] for the adhesive model was able to predict the maximum load of the joints with a slight
layer, proved to be accurate for all types of loading. Lißner et al. [82] overestimation, but the displacement at that load level was severely
also tested the same type of joints, but using single material adherents underpredicted, whose difference was justified with the experimental
(titanium). Using the same CZM and the CZM parameters from Ref. [80], conditions. In the same work, a comparison between a 3D and a 2D
the joint strength was over-predicted by 16%–20% in high rate loading. plane strain model was performed showing no significant difference
By recalibrating the CZM parameters, the authors were able to find a between them. Furthermore, a purely numerical comparison between
better agreement with the experimental results. different adherents showed that using less stiff adherents results in a

1 1
Fig. 11. Strain rate variation for a 150 mm/s test speed and a 15 s− strain rate (a), and a 15 mm/s test speed and a 10 s− strain rate [78].

8
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

decrease of the maximum load for all three adhesives, but an increase of and not for the SERR. Dos Reis et al. [88] numerically and experimen­
the displacement at maximum load. In a later work [70], the same au­ tally studied SLJ with a graded adhesive, stiffer in the middle and more
thors used the same model to test, solely numerically, different flexible at the ends, using a triangular CZM. The results showed that the
geometrical configurations such as: an outer chamfer, an inner chamfer joint strength was not significantly affected by grading the adhesive, as
and an adhesive fillet. Overall, the adhesive fillet showed to be the most shown in Fig. 13, but the energy absorbed was greater with the graded
advantageous configuration to improve the maximum load the most for adhesive, when compared to using a non-graded stiff or a non-graded
all three adhesives, as it can be seen in Fig. 12. flexible adhesive. This behaviour contrasts with the quasi-static obser­
Ledford and May [86] compared two different modelling approaches vations, where the graded adhesive improved the joint strength and the
of a DLJ tested in a Split-Hopkinson Tension Bar. In both models, the energy absorbed. The authors justified this difference with
author used CZM for the adhesive and continuum elements for the ad­ shockwave-like stress propagation under impact conditions, resulting in
herents and bolts. The joints were bonded-bolted with composite and several stress peaks in the adhesive along the overlap.
aluminium adherents. In one model only the specimen was simulated, Sandwich panels are a composite material of interest and, recently,
while in the other the whole test setup was simulated. The comparison two numerical and experimental works have used these materials as the
between these models showed that the later model reduced oscillations adherents of adhesive joints. Khosravani et al. [89] experimentally and
in the force-displacement curves. numerically tested T-joints with sandwich adherents, which had a
Due to the stress concentrations found at the overlap ends of SLJ,
some authors proposed the use of graded adhesives, or mixed adhesive
joints, to achieve a more uniform stress distribution. Machado et al. [87]
used CZM on mixed adhesive SLJ, with a strong and brittle adhesive at
the centre of the overlap and a flexible adhesive at the overlap ends. Two
different adhesives were tested for each region. A comparison between a
2D plane strain model and a 3D model, both with cohesive elements
simulating the adhesives, showed no significant difference. A mesh de­
pendency study also showed that CZM are mesh independent under
impact loads, which is also a well-documented behaviour under
quasi-static loads. The strength predictions under impact were over­
estimated compared to the experimental data. However, for joints
bonded with a single adhesive, these overestimations were not signifi­
cant. Numerically a strength increase was predicted for the
mixed-adhesive joints, when compared to the single adhesive joints with
the flexible adhesives, but this was not verified experimentally. These Fig. 13. Joint strength, where a low cure temperature resulted in a stiff ad­
discrepancies were associated to the strain dependency of the material hesive, a high cure temperature resulted in a ductile adhesive and the graded
properties, which was only accounted for the tensile and shear strength cure resulted in an adhesive with graded properties [88].

Fig. 12. Strength prediction (Pm) and displacement at that load (δPm) for different outer chamfer angles (α) (a), inner chamfer angles (β) (b) and adhesive fillet angles
(θ) (c) [70].

9
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

honeycomb core. The numerical model was 3D and used cohesive ele­ with a functionally graded adhesive, which was validated against a FEA
ments with a triangular law for the adhesive and shell elements for the that predicted the natural frequencies similarly to the analytical model.
sandwich structure. This model was used to predict the joint strength In this work, the authors also compared aluminium and steel adherents
under a strain rate of about 5000 s− 1. This prediction was below the and found that using aluminium adherents, which are softer but less
experimental strength but not by a large amount. Caliskan and Apalak dense, resulted in higher natural frequencies. Vazirir and
[90] studied adhesively bonded T-joints made of foam-aluminium Nayeb-Hashemi [100] validated the model for composite repair patches
sandwich panels adherents. The foam was modelled considering against a FEA and experiments. The FEA and the analytical model were
hyperelasticity and a damage model, while the aluminium was modelled in good agreement with the experiments.
as elastic-plastic, and the adhesive was modelled with a triangular CZM. These analytical models are generally faster than a FEA. However,
Two different aluminium and foams were tested. It was found that using they are not as versatile and a given model is only applicable to a
a denser foam resulted in higher forces, for the same impact energy, and restricted set of joints, while a FEA, or another discretization technique,
the more ductile aluminium absorbed more energy by deforming plas­ is applicable to any joint. The Finite Difference Method (FDM) was used
tically. A good agreement between the experimental and numerical re­ by Rao and Zhou [101] to perform a modal analysis of Tubular Lap
sults was found. Joints (TLJ). The effects of several parameters were tested. A decrease in
Hybrid joints, jointed with multiple joining methods, are also an the natural frequencies was observed with increases to the adhesive
object of interest under impact conditions. Chen et al. [91] experimen­ thickness. Also, increasing the temperature also decreases the fre­
tally and numerically tested riveted/bonded composite SLJ used in the quencies because the joint becomes softer. Yaman and Sansveren [102]
automotive industry under three different transverse impact loads of numerically and experimentally tested composite SLJ, SSJ and DSJ, with
2.27 m/s, 2.53 m/s and 2.77 m/s. The comparison between rive­ several geometric variations. The FEA correlated well with the experi­
ted/bonded SLJ and riveted SLJ showed that the later were stronger ments for the different joint configurations. It was also shown that fibre
under impact. However, in the tests both types of joints had the same orientation angle and adherent thickness have a significant effect on the
total thickness, which in the riveted/bonded joints meant that the natural frequencies, increasing them as they increase. On the other
thickness of the adherent was reduced, due to the thickness of the ad­ hand, the damping ratio improves with decreases in adherent thickness
hesive. The numerical model used CZM for the adhesive and continuum and increases in overlap length. For DSJ, a similar improvement in
damage models for the titanium rivets and the composite substrates. The damping due to the overlap length was also observed by Gorrepati and
force-displacement predictions compared well with the Rao [103]. Additionally, it was observed that natural frequencies in­
force-displacement observed experimentally for both joint types and the crease slightly with the increase of the overlap length and the decrease
different load rates. of the adhesive thickness.
The work of Silva et al. [92] presented a use of CZM in real world Ahmad Basri et al. [104] used experiments and a FEA to determine
applications. A triangular CZM was used to predict the life of SLJ with the natural frequencies of a composite structure made of a hat and a
different adhesive thicknesses and of an adhesively bonded automotive plate that were adhesively-bonded. Initially, the hat and plate were
structure. The numerical test showed a small difference in the maximum tested separately, and the FEA predictions of the natural frequency
load between the two adhesive thicknesses tested, with the thinnest showed a maximum error of 11%. Afterwards, the model was updated
being the strongest, but the difference was higher in the experimental based on the original test and the error was reduced to a maximum of
results. The automotive structure simulations were able to predict the 4%. Then, the complete structure was tested, and a parametric study was
strength with minimal deviation from the experiments, but the energy performed to assess the influence of several parameters of the numerical
absorbed showed a larger difference between experimental and nu­ model. The natural frequencies of the structure were predicted with a
merical results. maximum error of 3% when using the adjusted parameters. The mode
shapes of the structure predicted by the FEA were also similar to the
4. Modal analysis mode shapes measured experimentally, as shown in Fig. 14.
Apalak at al [105] tested the effect on the natural frequencies of
The accurate prediction of natural vibration frequencies and corre­ several geometrical and mechanical variations to a L-shaped corner
sponding mode shapes is highly relevant to ensure that the working joint. This task was performed only numerically using a 3D FEA. The
conditions of adhesive joints are outside the natural frequencies, which authors found that changes to the adhesive mechanical properties,
could lead to premature failures. Furthermore, adhesives have damping namely the Young’s modulus, Poisson’s ratio, and density, did not alter
properties, reinforcing the importance of these investigations. The 2011 the natural frequencies in a significant manner. The same was also
review of He [2] has a section dedicated to modal analysis, but it is not observed when the adhesive thickness was varied. Oppositely, changes
very extensive. This review aims at presenting a more complete picture to the geometry of the adherents had a significant effect on the natural
of the current state of the investigation of modal analysis in adhesive frequencies, which increased for higher thicknesses and decreased for
joints. shorter joint lengths. The natural frequencies also increased with longer
The first works focused on the modal analysis of adhesive joints support lengths, but not as significantly. Based on the obtained data, an
consisted of analytical models, being the works of Saito and Tani [93], ANN was trained, and a genetic algorithm was used to predict the op­
and Rao and Crocker [94], for SLJ, and the work of Miles and Reinhald timum joint dimensions, which maximized the natural frequency.
[95], for two metal layers bonded with an adhesive, some of the earliest Bidyadhar and Mishra also studied this type of joint [106], corrobo­
research studies. Unlike the two previously mentioned loading cases, in rating the findings of Apalak et al. Additionally, different adherents
modal analysis analytical models are still commonly used, as attested by were tested, finding that stiffer adherents exhibit higher natural fre­
several recent works developing new analytical models, which are quencies, and also that corner angles do not significantly influence the
frequently validated by FEA. Almitani and Othman [96] proposed an natural frequencies. ANN were also used by Gunes et al. [107,108] for
analytical model for SLJ and DLJ, which was validated by FEA. The work SLJ with graded adherents. The ANN was trained using data from FEA by
of Challita et al. [97] is another example of an analytical model vali­ varying mechanical and geometrical properties. It was found that
dated by a FEA, but this model is applicable to SLJ with voids. The decreasing the joint width below 50 mm significantly increased the
natural frequencies predicted by both methods were similar. Wang et al. natural frequencies, but widths above 50 mm did not visibly change the
[98] validated an analytical model to calculate the stress under a har­ natural frequencies. Furthermore, changes to the adhesive mechanical
monic load against a FEA for DLJ. This model predicted the natural and geometrical properties resulted in negligible changes to the natural
frequencies similarly to the FEA, but the stresses of the analytical model frequencies. The most significant effect on the natural frequencies was
were slightly higher. Recently, in 2021, Sindi et al. [99] studied DLJ due to the property gradation of the adherents and their thickness,

10
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

Fig. 14. Vibration mode shapes of the structure tested in Ref. [104].

producing higher natural frequencies. reaching similar conclusions.


He and Oyadiji [109] and He [110] studied the effects on the natural Wang et al. [113] used a layerwise FEM to perform the modal
frequencies of the adhesive mechanical properties using a FEA, showing analysis of a SLJ and a Step-Lap Joint (S-LJ). The natural frequencies
that changes to the Poisson’s ratio generally do not significantly alter the predicted for the SLJ were compared to the natural frequencies pre­
natural frequencies. The same was also observed for realistic changes to dicted by Gunes et al. [107] using a standard FEM, achieving similar
the Young’s modulus, but when using very low Young’s modulus, below predictions, thus validating their layerwise FEM. Then, a parametric
1 GPa, the natural frequencies significantly changed, as can be observed study was performed for the S-LJ, whose main conclusions were: the
in Fig. 15. Kaya et al. [111] observed, with a FEA, that in SLJ thicker adhesive thickness does not significantly change the natural frequencies,
adhesive layers lead to decreases in the natural frequencies. However, but it increases damping; and increasing the overlap length leads to
this increase was not significant. Damm et al. [112] studied with a FEA higher natural frequencies and damping. The FEA of Ko et al. [114] also
the modal response of an adhesively bonded light signal system with a reached similar conclusions regarding the overlap length effects on the
tubular joint and compared it to a welded one. The authors tested two natural frequencies of SLJ.
different adhesives, which showed a relevant difference in terms of Several joining methods were tested by Van Belle et al. [115]
natural frequencies, with the stiffest adhesive having slightly higher experimentally and numerically for SLJ. It was found that, for the ad­
natural frequencies. The welded joint also resulted in slightly higher hesive joints, the natural frequencies predicted by the numerical method
natural frequencies when compared to the bonded joint, but the differ­ were similar to the experimental ones. Du and Lu [116] studied the
ence was not significant. It was also found that a thicker adhesive layer change to the modal behaviour of SLJ with dissimilar adherents due to
and a less stiff adhesive leads to more damping, which is beneficial. fatigue. Their experiments showed a decrease in the natural frequencies
Additionally, simulations with different overlap lengths showed that as the fatigue cycles increased. Experiments with aluminium beams, the
they have a small influence on the damping characteristics. Finally, the adherent material, also showed that this decrease was due to the ad­
authors tested an observation tower, with multiple bonded areas, hesive layer, since the natural frequencies of the beams remained almost
the same even after more than a million cycles. The numerical model for
the undamaged joint predicted natural frequencies similar to the ex­
periments. However, for the damaged joints a significant reduction to
the adhesive stiffness, over 99%, or bonded area, around 80%, was
required to achieve similar frequencies to the ones observed experi­
mentally. Such a significant decrease was not observed in any of the
works referenced in Section 2. He [117] predicted the first 20 natural
frequencies of a SLJ using a FEA. The authors compared their results to
experiments for two different joint widths, achieving a maximum error
of 26.88% for one of the modes. Lin and Ko [118] were able to determine
the natural frequencies of an adhesively bonded layered plate using a
FEA. It was shown that an increase in plate thickness results in increases
to the natural frequencies, which was expected considering that it in­
creases the overall stiffness of the joint.

5. Discussion

5.1. Fatigue

The period covered by this review shows that CZM are the most used
method to numerically study fatigue in adhesive joints. Such observa­
Fig. 15. Changes to the natural frequencies due to adhesive’s Young’s
tion can be justified by the fact that CZM are able to accurately predict
modulus [109].

11
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

crack initiation and crack propagation, translating into accurate life impact loading). While in static loading conditions the effects of
predictions. Another significant advantage of CZM is their mesh inde­ changing the adhesive thickness on the cohesive properties are well
pendence, which is not always verified with the other alternatives, i.e., documented, namely an increase in GIc, GIIc and ts, and a decrease in tn
continuum or fracture mechanics, due to the stress singularities arising with adhesive thickness increases [1], under variable strain rate and
at the adhesive/adherent interfaces. However, CZM also have some impact loads few works exist. Lißner et al. [80] showed variation on the
limitations, such as need for more characterization tests. In turn, these cohesive parameters with adhesive thickness changes in variable strain
tests are also adhesive thickness dependent, due to the constraining ef­ rate loading conditions. However, since these effects are still under­
fects of the adherents on the adhesive layer. This phenomenon is well studied, in the future it would be important to verify if adhesive thick­
documented in static analysis [1], but in fatigue more research is ness changes also affect the cohesive properties, since this is a major
required. Considering that the constraining effects are still present in limitation of CZM in static loading conditions. Another pertinent aspect
fatigue, it is expected that CZM parameters are also dependent on the relating to CZM that is understudied at the moment is the use of different
adhesive thickness in fatigue analysis. Continuum mechanics is not able cohesive laws. While under static loading conditions there is extensive
to predict crack propagation, and fracture mechanics cannot predict literature comparing different cohesive laws, in impact loading condi­
crack initiation. Considering that, the approach proposed by Wu et al. tions the overwhelming majority of works use triangular cohesive laws,
[14] could be an important advancement. Wu et al. approach, which with some exceptions, such as the work of Machado et al. [87], using
combines fracture and continuum mechanics, enables the prediction of trapezoidal laws. Given that triangular laws are less appropriate for
both crack propagation and crack initiation. Even though the life pre­ ductile adhesives, it would be important to increase the amount of
dictions with this approach were accurate, there was no mesh inde­ research devoted to other laws, especially when considering that these
pendence study, so it is not clear if this approach can have that problem adhesives are commonly used.
and further research is needed. ANN could also be a major investigation
focus and the work of Lyathakula and Yuan [52] showed promising 5.3. Modal analysis
results. This type of approach could rely on the enormous amount of
experimental and numerical data published in this field to perform life Contrary to fatigue and impact, in modal analysis analytical studies
predictions. are still common. Even so, as was demonstrated in Section 4, many
Regardless of the numerical method, a lack of works focused on short analytical models proposed in recent years were validated against nu­
cracks is noticeable. In metals it has been shown that short crack growth merical models. This shows that numerical models are currently taken as
is an aspect to consider when performing fatigue analysis. This behav­ the standard to which other analysis methods are compared to. Addi­
iour was first observed by Pearson et al. [119] for aluminium alloys. tionally, the works that compared experimental results to numerical
Short cracks have been defined as cracks of the size of the material’s results also showed the accuracy of these models regarding the predicted
microstructure [120]. At this scale it was observed that cracks propagate natural frequencies.
faster than what is predicted by traditional fatigue theory, such as Paris’ A significant amount of the research dealing with modal analysis is
law. Moreover, at this scale cracks also propagate at stress intensities focused on parametric studies. This is useful to determine how material
ranges below the threshold [121]. The main explanation for this phe­ and geometric parameter changes affect the natural frequencies and
nomenon is a smaller crack closure effect on short cracks [122]. To date, mode shapes. Regardless of the type of joint, it is evident, from these
to the authors’ best knowledge, this phenomenon is yet to be numeri­ studies, that the mechanical properties of the adhesive have a small
cally studied in adhesive joints. In the future, more works devoted to this influence on the natural frequencies of the joints. This is explained by
subject are required to determine if this phenomenon also occurs in the fact that the adhesive layer corresponds to a small fraction of the
adhesives and, if so, to which degree. The work of Pitt et al. [123] joint’s volume and weight, since adhesive layers are, generally, signifi­
proposes the use of a formulation alternative to the Paris’ law, namely cantly thinner than the adherents. It follows that changes to the me­
the Hartman-Schijve, which, in metals, unifies the crack growth chanical properties of the adherents have a significant effect on the
behaviour of short and long cracks. However, this approach was not natural frequencies of the joint. The effects of geometric parameter
applied to short cracks in adhesive joints. changes seem to follow this rule: changes that increase the ratio of ad­
hesive volume/adherent volume tend to decrease the natural fre­
5.2. Variable strain rate and impact quencies, while the inverse results in natural frequency increases.
Additionally, increasing the overlap area, in joints like the SLJ, also
Continuum mechanics can be useful to obtain stresses and strains tends to lead to increases in the natural frequencies.
under variable strain rate and impact loadings, showing a significant Compared to other joining methods, adhesive offers an advantage
difference from the stress distributions observed in static loading con­ where vibration problems are concerned, their damping, since most
ditions. Here the stress propagates as a wave, creating several local adhesives are viscoelastic. In this case, increasing the adhesive thickness
peaks in the adhesive, while in static loading conditions the only stress and the overlap length tends to increase the modal loss factor. However,
peaks in the adhesive are located at the overlap ends. However, con­ since increasing the adhesive thickness also leads to decreases in the
tinuum mechanics are of limited used to perform strength predictions natural frequencies, an analysis of these two factors is always needed for
under impact loading conditions, while using CZM or damage mechanics a given application, to determine the optimal adhesive thickness.
is more advisable. The use of ANN in modal analysis is promising. This technique can be
Damage mechanics can be useful to determine joint strength. How­ used in parametric studies to obtain the optimal parameter combination.
ever, the existence of stress/strain singularities, albeit not as severe as in Instead of testing each parameter individually, as is common in tradi­
static loading conditions, is a limiting factor since stress/strain are used tional parametric studies. This could be particularly useful to determine
to determine material degradation. This can explain the smaller amount the optimal adhesive thickness, since increasing it increases the damp­
of works using damage mechanics when compared to the ones using ing, but decreases the natural frequencies.
CZM.
As in fatigue, in impact CZM are also the most popular numerical 6. Conclusion
approach. This occurs due to similar reasons, particularly their mesh
independence, which is not always verified when using continuum or Overall, the present review shows that, in recent years, there is a
damage mechanics due to the stress singularities in the adhesive/ significant amount of research devoted to the numerical study of the
adherent interfaces (even though these singularities are not as severe as dynamic behaviour of adhesive joints. However, when compared to
in static loading, because of the wave-like way the stress propagates in static loading conditions, as seen in the review of Ramalho et al. [1], the

12
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

numerical study of adhesive joints under dynamic conditions is not as References


well developed, but the increasing number of research articles in recent
years is encouraging and shows that this is a field of study with [1] Ramalho LDC, Campilho RDSG, Belinha J, da Silva LFM. Static strength
prediction of adhesive joints: a review. Int J Adhesion Adhes 2020;96:102451.
increasing interest. In this work, the dynamics behaviour of adhesive [2] He X. A review of finite element analysis of adhesively bonded joints. Int J
joints was divided into three loading cases, whose conclusions can be Adhesion Adhes 2011;31(4):248–64.
drawn: [3] Mostovoy S, Ripling EJ. Flaw tolerance of a number of commercial and
experimental adhesives. Adhes.Sci.and Tech. 1975;9:513–62.
[4] Pascoe JA, Alderliesten RC, Benedictus R. Methods for the prediction of fatigue
- For fatigue there is a previous review by Pascoe et al. [4], which delamination growth in composites and adhesive bonds - a critical review. Eng
shows that until 2013 there was a good amount of research, but the Fract Mech 2013;112(113):72–96.
[5] Esmaeili F, Zehsaz M, Chakherlou TN, Barzegar S. Fatigue life estimation of
current review shows that, since then, this field has significantly double lap simple bolted and hybrid (bolted/bonded) joints using several
expanded. In this field, the most commonly used numerical approach multiaxial fatigue criteria. Mater Des 2015;67:583–95.
is CZM, particularly CZM with triangular cohesive laws. The variety [6] Smith P, Topper KN, Watson TH. A stress–strain function for the fatigue of metals
(stress-strain function for metal fatigue including mean stress effect). J Mater
of joint geometries analysed, from the very common SLJ to wind
1970;5(January 1970):767–78.
turbine root joints, shows the versatility of numerical methods. [7] Kandil FA, Brown MW, Miller KJ. Biaxial low-cycle fatigue failure of 316 stainless
- Regarding variable strain rate and impact, the 2018 review of steel at elevated temperatures. United Kingdom: Metals Society; 1982.
Machado et al. [53] presented the most significant works in the [8] Glinka G, Shen G, Plumtree A. A multiaxial fatigue strain energy density
parameter related to the critical fracture plane. Fatig Fract Eng Mater Struct
impact field up to 2016. The current review shows that, in this field, 1995;18(1):37–46.
CZM also proved to be the most common numerical approach, being [9] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage
triangular CZM the most popular. It was shown that the correct including out-of-phase loading. Fatig Fract Eng Mater Struct 1988;11(3):149–65.
[10] Crossland B. Effect of large hydrostatic pressures on the torsional fatigue strength
characterization of materials is essential for the accuracy of the nu­ of an alloy steel. In: Proceedings of the international conference on fatigue of
merical simulations, since the adhesives’ mechanical properties metals; 1956. p. 138–49.
show significant variation due to strain rate. The use of mechanical [11] Farahani A, Varvani. A new energy-critical plane parameter for fatigue life
assessment of various metallic materials subjected to in-phase and out-of-phase
properties determined with quasi-static tests leads to inaccurate re­ multiaxial fatigue loading conditions. Int J Fatig 2000;22:295–305.
sults. In impact conditions, the stress in adhesive joints propagates as [12] Shahani AR, Pourhosseini SM. The effect of adherent thickness on fatigue life of
a wave, which creates several local peaks. This feature means that adhesively bonded joints. Fatig Fract Eng Mater Struct 2019;42(2):561–71.
[13] Pereira AM, Ferreira JAM, Antunes FV, Bártolo PJ. Assessment of the fatigue life
solutions to reduce the peak stresses at the overlap ends of adhesive of aluminium spot-welded and weld-bonded joints. J Adhes Sci Technol 2014;28
joints, like using graded adhesives, are not as advantageous as in (14–15):1432–50.
quasi-static conditions. From the analysed, works it can be concluded [14] Wu G, et al. Experimental and numerical evaluations on the effects of adhesive
fillet, overlap length and unbonded area in adhesive-bonded joints. Fatig Fract
that joint strength under impact conditions is higher than under
Eng Mater Struct 2020;43(10):2298–311.
quasi-static conditions. [15] Taylor D. The theory of critical distances. Eng Fract Mech 2008;75(7):1696–705.
- Numerical modal analysis of adhesive joints was briefly covered by [16] Castro Sousa F, Akhavan-Safar A, Rakesh G, da Silva LFM. Fatigue life estimation
the 2011 review of He [2]. The current work provides a more com­ of adhesive joints at different mode mixities. 00 J Adhes 2020:1–23. 00.
[17] Sousa FC, Akhavan-Safar A, Goyal R, da Silva LFM. Fatigue life estimation of
plete analysis, also including more recent works that were not single lap adhesive joints using a critical distance criterion: an equivalent notch
available at the time. Modifying adhesive properties do not seem to approach. Mech Mater 2021;153(July 2020):103670.
significantly alter the resonant frequencies, since the adhesive [18] Schneider B, Beber VC, Brede M. Estimation of the lifetime of bonded joints under
cyclic loads at different temperatures. J Adhes 2016;92(7–9):795–817.
seldom contributes to the overall stiffness of the joints, due to the [19] Beber VC, Fernandes PHE, Schneider B, Brede M, Mayer B. Fatigue lifetime
adherents being in general much stiffer that the adhesive. However, prediction of adhesively bonded joints: an investigation of the influence of
adhesives possess damping properties, which are beneficial. This material model and multiaxiality. Int J Adhesion Adhes 2017;78(August):240–7.
[20] Beber VC, Schneider B, Brede M. Efficient critical distance approach to predict the
field is not as widely studied as the other two. Nonetheless, resonant fatigue lifetime of structural adhesive joints. Eng Fract Mech 2019;214(March):
frequencies and damping are an important part of dynamic analysis 365–77.
and should be more studied. [21] Irwin GR. Analysis of stresses and strains near the end of a crack traversing a
plate. J Appl Mech 1957;24(September):351–69.
[22] Griffith AA, Eng M. VI. The phenomena of rupture and flow in solids. Philos Trans
In all loading cases described, numerical work is generally performed R Soc Lond - Ser A Contain Pap a Math or Phys Character Jan. 1920;221
with the FEM, with the exception of a FDM work in the modal analysis (582–593):163 LP–198.
[23] Chen Q, Guo H, Avery K, Su X, Kang H. Fatigue performance and life estimation of
section. This contrasts with the static analysis of adhesive joints, as
automotive adhesive joints using a fracture mechanics approach. Eng Fract Mech
shown in the review of Ramalho et al. [1]. In the static analysis other 2017;172:73–89.
numerical methods have been employed, namely the eXtended FEM [24] Chen Q, Guo H, Avery K, Kang H, Su X. Mixed-mode fatigue crack growth and life
(XFEM) and various meshless methods, including Smoothed Particle prediction of an automotive adhesive bonding system. Eng Fract Mech 2018;189:
439–50.
Hydrodynamics (SPH) and the Radial Point Interpolation Method [25] Wu G, et al. Fatigue behaviors and mechanism-based life evaluation on SPR-
(RPIM). Future works devoted to the dynamic analysis of adhesive joints bonded aluminum joint. Int J Fatig 2021;142(June 2020):105948.
could make use of these alternative numerical methods to assess their [26] Pascoe JA, Zavatta N, Troiani E, Alderliesten RC. The effect of bond-line thickness
on fatigue crack growth rate in adhesively bonded joints. Eng Fract Mech 2020;
viability in these loading conditions. They could be especially useful for 229(February):106959.
problems involving crack propagation, since they do not require [27] Lai WJ, Pan J. Stress intensity factor solutions for adhesive-bonded lap-shear
remeshing which can be computationally expensive. specimens of magnesium and steel sheets with and without kinked cracks for
fatigue life estimations. Eng Fract Mech 2014;131:454–70.
[28] Samaei M, Zehsaz M, Chakherlou TN. Experimental and numerical study of
Acknowledgement fatigue crack growth of aluminum alloy 2024-T3 single lap simple bolted and
hybrid (adhesive/bolted) joints. Eng Fail Anal 2016;59:253–68.
[29] Tang JH, Sridhar I, Srikanth N. Static and fatigue failure analysis of adhesively
The authors truly acknowledge the funding provided by Ministério bonded thick composite single lap joints. Compos Sci Technol 2013;86:18–25.
da Ciência, Tecnologia e Ensino Superior – Fundação para a Ciência e a [30] Costa M, Viana G, Créac’hcadec R, da Silva LFM, Campilho RDSG. A cohesive
Tecnologia (Portugal), under project funding MIT-EXPL/ISF/0084/ zone element for mode I modelling of adhesives degraded by humidity and
fatigue. Int J Fatig 2018;112(March):173–82.
2017, POCI-01-0145-FEDER-028351 and SFRH/BD/147628/2019. [31] Ebadi-Rajoli J, Akhavan-Safar A, Hosseini-Toudeshky H, da Silva LFM.
Additionally, the authors gratefully acknowledge the funding of Project Progressive damage modeling of composite materials subjected to mixed mode
NORTE-01-0145-FEDER-000022 – SciTech – Science and Technology cyclic loading using cohesive zone model. Mech Mater 2020;143(January):
103322.
for Competitive and Sustainable Industries, co-financed by Programa
[32] Akhavan-Safar A, Monteiro J, Carbas R, Marques E, Goyal R, da Silva LFM.
Operacional Regional do Norte (NORTE2020), through Fundo Europeu Tensile fatigue life prediction of adhesively bonded structures based on CZM
de Desenvolvimento Regional (FEDER).

13
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

technique and a modified degradation approach. Proc IME G J Aero Eng 2020; [62] Shim W, et al. Simulating rate- and temperature-dependent behaviors of
234(13):1988–99. adhesives using a nonlinear viscoelastic model. Mech Mater 2020;147(April).
[33] Monteiro J, et al. Mode II modeling of adhesive materials degraded by fatigue [63] Gollins K, Elvin N, Delale F. Characterization of adhesive joints under high-speed
loading using cohesive zone elements. Theor Appl Fract Mech 2019;103(April): normal impact: Part I – experimental studies. Int J Adhesion Adhes 2020;98
102253. (December 2019):102529.
[34] Akhavan-Safar A, Monteiro J, Carbas R, Marques E, Goyal RK, da Silva L. [64] Gollins K, Elvin N, Delale F. Characterization of adhesive joints under high-speed
A modified degradation technique for fatigue life assessment of adhesive normal impact: Part II – numerical studies. Int J Adhesion Adhes 2020;98:
materials subjected to cyclic shear loads. Proc IME C J Mech Eng Sci 2021;235(3): 102530.
550–9. [65] Yildiz S, Andreopoulos Y, Delale F. Mode I characterization of toughened epoxy
[35] Rocha AVM, et al. Numerical analysis of mixed-mode fatigue crack growth of adhesive joints under shock-wave loading. Int J Adhesion Adhes 2019;90
adhesive joints using CZM. Theor Appl Fract Mech 2020;106(December 2019): (February):71–87.
102493. [66] Boling H, Dongyun G. Dynamic analysis of single-lap, adhesively bonded
[36] De Moura MFSF, Gonçalves JPM. Cohesive zone model for high-cycle fatigue of composite-titanium joints subjected to solid projectile impact. J Adhes Sci
adhesively bonded joints under mode i loading. Int J Solid Struct 2014;51(5): Technol 2018;32(11):1176–99.
1123–31. [67] Bedon C, Kalamar R, Eliášová M. Low velocity impact performance investigation
[37] Pirondi A, Moroni F. A progressive damage model for the prediction of fatigue on square hollow glass columns via full-scale experiments and Finite Element
crack growth in bonded joints. J Adhes 2010;86(5–6):501–21. analyses. Compos Struct 2017;182(September):311–25.
[38] De Moura MFSF, Gonçalves JPM. Development of a cohesive zone model for [68] Hou W, Xu X, Sang L, Tong L. Failure of single hat-shaped thin-walled tubular
fatigue/fracture characterization of composite bonded joints under mode II composite T-joints under impact loading. Thin-Walled Struct 2020;154(April):
loading. Int J Adhesion Adhes 2014;54:224–30. 106815.
[39] de Moura MFSF, Gonçalves JPM. Cohesive zone model for high-cycle fatigue of [69] Valente JPA, Campilho RDSG, Marques EAS, Machado JJM, da Silva LFM.
composite bonded joints under mixed-mode I+II loading. Eng Fract Mech 2015; Adhesive joint analysis under tensile impact loads by cohesive zone modelling.
140:31–42. Compos Struct 2019;222(January):110894.
[40] Moreira RDF, de Moura MFSF, Silva FGA, Ramírez FMG, Silva FDR. Numerical [70] Valente JPA, Campilho RDSG, Marques EAS, Machado JJM, da Silva LFM.
comparison of several composite bonded repairs under fatigue loading. Compos Geometrical optimization of adhesive joints under tensile impact loads using
Struct 2020;243(March):112250. cohesive zone modelling. Int J Adhesion Adhes 2020;97(November 2019):
[41] de Moura MFSF, Moreira RDF. Fatigue analysis of composite bonded repairs. 102492.
J Adhes Sci Technol 2017;31(19–20):2164–79. [71] Li Y, Yang Y, Li J, Wang B, Liao Y. Experimental-numerical analysis of failure of
[42] Hosseini-Toudeshky H, Sheibanian F, Ovesy HR, Goodarzi MS. Prediction of adhesively bonded lap joints under transverse impact and different temperatures.
interlaminar fatigue damages in adhesively bonded joints using mixed-mode Int J Impact Eng 2020;140(October 2019):103541.
strain based cohesive zone modeling. Theor Appl Fract Mech 2020;106(April [72] Peres LMC, et al. Geometry and adhesive optimization of single-lap adhesive
2019):102480. joints under impact. J Adhes 2022;98(6):677–703.
[43] Khoramishad H, Crocombe AD, Katnam KB, Ashcroft IA. Predicting fatigue [73] Silva AFMV, Peres LMC, Campilho RDSG, Rocha RJB, Silva FJG. Cohesive zone
damage in adhesively bonded joints using a cohesive zone model. Int J Fatig parametric analysis in the tensile impact strength of tubular adhesive joints. Proc
2010;32(7):1146–58. IME E J Process Mech Eng 2022:095440892210882.
[44] Belnoue JPH, Giannis S, Dawson M, Hallett SR. Cohesive/adhesive failure [74] Jia Z, Yuan G, Feng X, Zou Y. Numerical study on the mechanical behavior of a
interaction in ductile adhesive joints Part II: quasi-static and fatigue analysis of polyurethane adhesive under high strain rate. Compos B Eng 2019;158
double lap-joint specimens subjected to through-thickness compressive loading. (September 2018):131–40.
Int J Adhesion Adhes 2016;68:369–78. [75] Yang Z, Zhu Z, Xia Y, Yang F, Sun Y, Jiang H. Modified cohesive zone model for
[45] Ibrahim GR, Albarbar A, Brethee KF. Progressive failure mechanism of laminated soft adhesive layer considering rate dependence of intrinsic fracture energy. Eng
composites under fatigue loading. J Compos Mater 2021;55(1):137–44. Fract Mech 2021;258(September):108089.
[46] Floros I, Tserpes K. Numerical simulation of quasi-static and fatigue debonding [76] Nunes PDP, et al. Numerical assessment of strain rate in an adhesive layer
growth in adhesively bonded composite joints containing bolts as crack stoppers. throughout double cantilever beam and end notch flexure tests. Proc IME E J
00 J Adhes 2019:1–23. 00. Process Mech Eng 2020;234(5):415–25.
[47] Tserpes K, Floros I. Fatigue crack growth simulation in adhesively bonded [77] Borges CSP, et al. A strain rate dependent cohesive zone element for mode I
composite joints. Fatig Fract Eng Mater Struct 2019;42(7):1430–40. modeling of the fracture behavior of adhesives. Proc IME J Mater Des Appl 2020;
[48] Sachse R, Pickett AK, Essig W, Middendorf P. Experimental and numerical 234(4):610–21.
investigation of the influence of rivetless nut plate joints on fatigue crack growth [78] Nunes PDP, Marques EAS, Carbas RJC, Akhavan-Safar A, da Silva LFM. DCB tests
in adhesively bonded composite joints. Int J Fatig 2017;105:262–75. at constant strain rate using crash-resistant epoxy adhesives: a numerical and
[49] Sachse R, Pickett AK, Middendorf P. Analysis of crack-arrest design features for experimental approach. Proc Inst Mech Eng - Part D J Automob Eng 2020:1–9.
adhesively bonded composite joints: an experimental and numerical study. 2020.
Compos Struct 2021;274(July):114301. [79] Lißner M, Alabort E, Cui H, Pellegrino A, Petrinic N. On the rate dependent
[50] Zuo Y, Montesano J, Singh CV. Assessing progressive failure in long wind turbine behaviour of epoxy adhesive joints: experimental characterisation and modelling
blades under quasi-static and cyclic loads. Renew Energy 2018;119:754–66. of mode I failure. Compos Struct 2018;189(September 2017):286–303.
[51] Hosseini-Toudeshky H, Jahanmardi M, Goodarzi MS. Progressive debonding [80] Lißner M, Alabort E, Cui H, Rito R, Blackman BRK, Petrinic N. Experimental
analysis of composite blade root joint of wind turbines under fatigue loading. characterisation and numerical modelling of the influence of bondline thickness,
Compos Struct 2015;120:417–27. loading rate, and deformation mode on the response of ductile adhesive
[52] Lyathakula KR, Yuan F-G. A probabilistic fatigue life prediction for adhesively interfaces. J Mech Phys Solid 2019;130:349–69.
bonded joints via ANNs-based hybrid model. xxxx Int J Fatig 2021:106352. [81] Lißner M, et al. Multi-material adhesively bonded structures: characterisation and
[53] Machado JJM, Marques EAS, da Silva LFM. Adhesives and adhesive joints under modelling of their rate-dependent performance. Compos B Eng 2020;195(March):
impact loadings: an overview. J Adhes 2018;94(6):421–52. 108077.
[54] Jalali S, Ayatollahi MR, Akhavan-Safar A, da Silva LFM. Effects of impact fatigue [82] Lißner M, Alabort E, Erice B, Cui H, Blackman BRK, Petrinic N. On the dynamic
on residual static strength of adhesively bonded joints. Proc IME J Mater Des Appl response of adhesively bonded structures. Int J Impact Eng 2020;138(June 2019):
2021;235(7):1519–31. 103479.
[55] Yildiz S, Andreopoulos Y, Delale F, Smail A. Adhesively bonded joints under [83] Araújo HAM, Machado JJM, Marques EAS, da Silva LFM. Dynamic behaviour of
quasi-static and shock-wave loadings. Int J Impact Eng 2020;143(September composite adhesive joints for the automotive industry. Compos Struct 2017;171:
2019):103613. 549–61.
[56] Xu Y, Ke Y, Ma X. Finite element analysis of stress wave propagation in adhesive [84] Jia Z, Yuan G, Feng X, Zou Y, Yu J. Shear properties of polyurethane ductile
joints under low speed impact tensile loadings. Macromol Theory Simul 2021;30 adhesive at low temperatures under high strain rate conditions. Compos B Eng
(1):1–11. 2019;156(July 2018):292–302.
[57] He B, Ge D. Dynamic strength of adhesively bonded composite joints with similar [85] Liu PF, Peng XQ, Guo ZY, Leng JX, Jiao L. Viscoelastic bilinear cohesive model
and dissimilar assembled adherends. J Reinforc Plast Compos 2017;36(23): and parameter identification for failure analysis of adhesive composite joints.
1683–92. Compos Struct 2019;224(May):111016.
[58] Quaglini V, Cattaneo S, Biolzi L. Numerical assessment of laminated cantilevered [86] Ledford N, May M. Modeling of multimaterial hybrid joints under high-rate
glass plates with point fixings. Glass Struct. Engg. 2020;5(2):187–204. loading. Proc IME E J Process Mech Eng 2020;234(5):446–53.
[59] Valès B, Marguet S, Créachcadec R, Sohier L, Ferrero JF, Navarro P. Numerical [87] Machado JJM, Marques EAS, Silva MRG, da Silva LFM. Numerical study of impact
study of the local behaviour of adhesive bonds under dynamic loading. J Adhes behaviour of mixed adhesive single lap joints for the automotive industry. Int J
Sci Technol 2016;30(12):1319–39. Adhesion Adhes 2018;84(February):92–100.
[60] Valès B, Marguet S, Créac’hcadec R, Sohier L, Ferrero JF, Navarro P. [88] dos Reis M, Carbas R, Marques E, da Silva L. Functionally graded adhesive joints
Experimental & numerical study of the Tensile/Compression-Shear Arcan test under impact loads. Proc Inst Mech Eng - Part D J Automob Eng 2021;235(13):
under dynamic loading. Int J Adhesion Adhes 2017;78(June):135–47. 3270–81.
[61] Rizk G, et al. Durability of composite assemblies under extreme conditions: [89] Khosravani MR, Anders D, Weinberg K. Influence of strain rate on fracture
thermomechanical damage prediction of a double-lap bonded composite behavior of sandwich composite T-joints. Eur J Mech Solid 2019;78(April):
assembly subject to impact and high temperature. Compos Struct 2019;213 103821.
(January):58–70. [90] Caliskan U, Apalak MK. Low speed impact behaviour of adhesively bonded foam-
core sandwich T-joints. J Adhes Sci Technol 2019;33(3):217–42.

14
L.D.C. Ramalho et al. International Journal of Adhesion and Adhesives 118 (2022) 103219

[91] Chen C, Sun C, Han X, Hu D, Zhou J, Guan Z. The structural response of the [106] Bidyadhar A, Mishra PK. Modelling and prediction of free vibration response of
thermoplastic composite joint subjected to out-of-plane loading. Int J Impact Eng adhesive bonded corner joint structures. Structures 2021;29(May 2020):185–98.
2020;145(August):103691. [107] Gunes R, Apalak MK, Yildirim M. The free vibration analysis and optimal design
[92] Silva NDD, Machado JJM, Marques EAS, Moreira PMGP, da Silva LFM. of an adhesively bonded functionally graded single lap joint. Int J Mech Sci 2007;
Experimental and numerical study of the dynamic response of an adhesively 49(4):479–99.
bonded automotive structure. Proc Inst Mech Eng - Part D J Automob Eng 2020; [108] Gunes R, Kemal Apalak M, Yildirim M, Ozkes I. Free vibration analysis of
234(14):3385–97. adhesively bonded single lap joints with wide and narrow functionally graded
[93] Saito H, Tani H. Vibrations of bonded beams with a single lap adhesive joint. plates. Compos Struct 2010;92(1):1–17.
J Sound Vib 1984;92(2):299–309. [109] He X, Oyadiji SO. Influence of adhesive characteristics on the transverse free
[94] Rao MD, Crocker MJ. Analytical and experimental study of the vibration of vibration of single lap-jointed cantilevered beams. J Mater Process Technol 2001;
bonded beams with a lap joint. J.of Vibrat.and Acoustics,Trasactior 1990;112(4): 119(1–3):366–73.
444–51. [110] He X. Finite element analysis of torsional free vibration of adhesively bonded
[95] Miles RN, Reinhall PG. An analytical model for the vibration of laminated beams single-lap joints. Int J Adhesion Adhes 2014;48:59–66.
including the effects of both shear and thickness deformation in the adhesive [111] Kaya A, Tekelioğlu MS, Findik F. Effects of various parameters on dynamic
layer. J.of Vibrat.and Acoustics,Trasactior 1986;108(1):56–64. characteristics in adhesively bonded joints. Mater Lett 2004;58(27):3451–6.
[96] Almitani KH, Othman R. Analytical solution of the harmonic response of visco- [112] Damm J, Ummenhofer T, Albiez M. Influence of damping properties of adhesively
elastic adhesively bonded single-lap and double-lap joints. Int J Adhesion Adhes bonded joints on the dynamic behaviour of steel structures: numerical
2016;71:55–65. investigations. 00 J Adhes 2020:1–29. 00.
[97] Challita G. Analytical study of the dynamic behavior of a voided adhesively [113] Wang S, Li Y, Xie Z. Free vibration analysis of adhesively bonded lap joints
bonded lap joint under axial harmonic load. Int J Solid Struct 2018;141–142: through layerwise finite element. Compos Struct 2019;223(May):110943.
183–94. [114] Ko TC, Lin CC, Chu RC. Vibration of bonded laminated lap-joint plates using
[98] Wang S, Guo Q, Xie Z. Extended analytical model for interfacial stresses of adhesive interface elements. J Sound Vib 1995;184(4):567–83.
double-lap joints under harmonic loads. Int J Adhesion Adhes 2019;91(February): [115] Van Belle L, et al. Experimental validation of numerical structural dynamic
23–35. models for metal plate joining techniques. JVC/Journal of Vibration and Control
[99] Sindi SA, Othman R, Almitani KH. Theoretical solution for the axial vibration of 2018;24(15):3348–69.
functionally graded double-lap adhesive joints. Math Mech Solid 2021;26(6): [116] Du Y, Shi L. Effect of vibration fatigue on modal properties of single lap adhesive
823–42. joints. Int J Adhesion Adhes 2014;53:72–9.
[100] Vaziri A, Nayeb-Hashemi H. Dynamic response of a repaired composite beam with [117] He X. Numerical and experimental investigations of the dynamic response of
an adhesively bonded patch under a harmonic peeling load. Int J Adhesion Adhes bonded beams with a single-lap joint. Int J Adhesion Adhes 2012;37:79–85.
2006;26(5):314–24. [118] Lin CC, Ko TC. Free vibration of bonded plates. Comput Struct 1997;64(1–4):
[101] Rao MD, Zhou H. Vibration and damping of A bonded tubular lap joint. J Sound 441–52.
Vib 1994;178(5):577–90. [119] Pearson S. Initiation of fatigue cracks in commercial aluminium alloys and the
[102] Yaman M, Şansveren MF. Numerical and experimental vibration analysis of subsequent propagation of very short cracks. Eng Fract Mech 1975;7(2):235–47.
different types of adhesively bonded joints. Structures 2021;34(June):368–80. [120] Hutař P, et al. Short fatigue crack behaviour under low cycle fatigue regime. Int J
[103] Gorrepati KM, Rao MD. Analysis of modal parameters of adhesively bonded Fatig 2017;103:207–15.
double-strap joints by the modal strain energy method. J.of Vibrat.and Acoustics, [121] Rao KTV, Yu W, Ritchie RO. On the behavior of small fatigue cracks in
Trasactior 1996;118(1):28–35. commercial aluminum-lithium alloys. Eng Fract Mech 1988;31(4):623–35.
[104] Ahmad Basri AB, Chae DW, Lee H. Investigation of the dynamic characteristics of [122] Lados DA, Apelian D, Paris PC, Donald JK. Closure mechanisms in Al-Si-Mg cast
a carbon-fiber-reinforced epoxy with adhesive-jointed structure. Compos Struct alloys and long-crack to small-crack corrections. Int J Fatig 2005;27(10–12):
2020;247(May):112499. 1463–72.
[105] Apalak MK, Ekici R, Yildirim M. Optimal design of an adhesively-bonded corner [123] Pitt S, Jones R, Peng D. Characterization of the durability of adhesive bonds. Fatig
joint with single support based on the free vibration analysis. J Adhes Sci Technol Fract Eng Mater Struct 2012;35(11):998–1006.
2006;20(13):1507–28.

15

You might also like