You are on page 1of 6

Damage Modelling of Adhesive Joints

Using the Cohesive Zone Model


Alexandra Wätzig1 , Vinzent Schmid, Ingo Ehrlich
Laboratory for Composite Technology
OTH Regensburg
Regensburg, Germany
1
Email: alexandra.waetzig@st.oth-regensburg.de

Abstract—Adhesive bonding constitutes an aspiring joining and therefore the impossibility of defining a general failure
technology. Nevertheless, the prediction of adhesive failure, nec- criterion, cause challenges for its use. These can be levered out
essary to derive the bonding dimensions, is still a challenging by alternative methods that are premised on the principles of
task due to local stress concentrations. Closed-form analysis
considering the stresses within the adhesive layer are limited fracture and damage mechanics. The most widely used damage
to simple geometries and loading conditions and require several mechanics approach is the cohesive zone model (CZM). It
simplifications. Numerical methods are therefore more appro- describes the damage evolution including the material weaken-
priate in terms of practical applications. In this work a finite ing, crack initiation and crack propagation at artificially intro-
element (FE) model of a single-lap joint according to the standard duced fracture interfaces by applying a softening relationship.
DIN EN 1465 is built up using the software ANSYS Workbench
V16.1. For this simple geometry results from both analytical This work aims to evaluate the use of CZM for adhesive
and empirical investigations are available for, respectively, the joints. To this end a finite element model of a single-lap
verification or validation of the numerical model. A damage
mechanics approach, the cohesive zone model (CZM), is applied.
joint is developed using the software ANSYS Workbench
Necessary material parameters for the implementation of CZM V16.1. Since convergence issues are always relevant for CZM,
and appropriate test methods for their determination are ascer- factors affecting the result’s convergence are examined [5].
tained. Furthermore, factors affecting convergence difficulties, Furthermore, there are several material parameters, which are
which are commonly given at this method, are examined. The necessary for the implementation of CZM. These parameters
overall aim is to evaluate the use of CZM for adhesive joints to
enable its transfer to more complex bonding geometries.
as well as the corresponding tests for their determination are
collated. Thus, the method is set up to be transferred to more
Index Terms—adhesive joints, single-lap joints, cohesive zone complex bonding geometries.
model, fracture mechanics, damage mechanics, ANSYS Work-
bench
II. FAILURE A NALYSIS
I. I NTRODUCTION Fundamentally, there are different strategies to describe the
Adhesive bonding offers several advantages compared to material behaviour or rather the failure mechanisms. They can
competing joining technologies. These include a more uniform be classified in continuum, fracture and damage mechanics [5].
stress distribution across the bonding surface in contrast to for Continuum mechanics deals, as the name implies, with
example riveted joints as well as the opportunity to connect continuums [20]. Thus, discontinuities of the molecular struc-
different materials [21]. A variety of applications benefits from ture as well as those on the microscopic scale are neglected
these potentials. However, industrial applications require a or abstracted, respectively. Instead, the homogenised material
reliable design and strength calculation. For this purpose, a behaviour is used. Continuum mechanical state variables, such
multitude of analytical models investigating the stresses within as stress, strain and temperature, show continuous functions
the adhesive layer had been developed over the last decades as well as continuous and sufficient differentiability. This is
[12]. But since these are limited to simple geometries and only valid for structures without any defects. Furthermore, the
loading conditions, numerical methods are more suitable for interpretation of the results respecting bonded joints is a major
practical applications. challenge due to the presence of stress singularities [14]. They
occur at sharp corners at the ends of the overlap and at stiffness
There are several approaches using the finite element
jumps caused by the different adjacent materials. That results
method for the simulation of adhesive joints [5]. They differ
in a total dependency of the obtained stress value on the mesh
in their fundamental mechanical description as well as in the
density at these points, whereby the stresses rise with a finer
findings which can be attained from it. Approaches based
mesh. Another difficulty is selecting the proper failure criterion
on continuum mechanics consider the stress-strain behaviour.
suitable to the material behaviour.
But even though they allow a much more precise calculation
of the stress distribution within the adhesive layer compared To take defects or precracks into consideration, fracture
to the analytical models, stress singularities at the edges mechanics approaches can be applied [5]. There singularities
scribes the static or fatigue damage at the cohesive process
zone ahead of the crack-tip. Subsequently the model as well
as the cohesive parameters, necessary for its application are
briefly analysed.
A. Model Description
Applying the CZM approach, both crack initiation and
crack propagation can be simulated by the use of continuum
Fig. 1. Fracture Modes: a) Mode I; b) Mode II; c) Mode III [11]. mechanical parameters for the damage onset and fracture
mechanical parameters for the crack growth. To this end,
at crack-tips and re-entrant corners between two bonded fracture interfaces are artificially introduced into structures.
materials are allowed and well accepted. Essentially, fracture This makes it particularly suitable for adhesive joints, where
mechanics describes the macroscopic effects of fracture. They the damage propagation is restricted to defined planes.
are based on the assumption of a continuum already containing The interfaces are modelled by applying softening relation-
a crack or any other defect. Examining the deformation of ships, the traction-separation laws, also named CZM laws
cracks, there are three crack opening types, presented in figure [19]. These define the response between two paired nodes
1 [23]. Mode I describes a symmetric crack opening, where the on the fracture interface comparable to a non-linear spring
crack faces are moving apart with the displacement normal to by relating the crack opening δ to a surface traction σ or τ ,
the crack. Parallel displacement of the crack faces in shear respectively, on the crack faces [8]. The area under this curve
results in mode II. Mode III is defined by out of plane is represented by the value of the fracture resistance Gc . A co-
sliding with relative displacement tangential to the crack front. hesive law is characteristic for a certain material or bimaterial
Generally fracture mechanics can be classified in, inter alia, interface, but independent of the geometry. Hence, it can be
linear elastic fracture mechanics (LEFM) and elasto-plastic applied for arbitrary specimens [5]. The resulting macroscopic
fracture mechanics (EPFM) [11]. LEFM assumes linear elastic reproduction of damage consists of the strength evolution,
material behaviour in the whole region of a cracked body or the weakening and gradual degradation of material properties
rather inelastic processes only in regions of very restricted and finally the complete failure of the bond. Precisely the
areas that can be neglected. The principles of LEFM are detaching process under loading starts with a rising traction
suitable for the majority of adhesively bonded joints except until the cohesive strength is reached. Subsequently the bond
for very ductile adhesives with a large amount of material softens, which leads to a reduced stiffness and strength for the
yielding. case of unloading before complete failure [22]. If the critical
Important parameters of LEFM are the stress intensity separation distance δc is attained, the bond opens and the crack
factor K and the energy release rate G [11]. The K-factor propagates to the next pair of nodes.
characterizes the local conditions at the crack-tip and can The accuracy of the mentioned parameters, especially the
be interpreted as the strength of the crack-tip field [1]. By fracture resistance, is much more important for the overall out-
comparison, G is defined as the rate of energy released by the come than the detailed shape of the cohesive law. Nevertheless
crack growth and thus represents the global behaviour of the there are several possibilities for the shape of CZM laws. The
crack propagation. This relationship is evinced by most commonly used laws are the bilinear and the exponential
dU traction-separation-curve, shown in figure 2. However, the
G= , (1) CZM law shape does not only influence the accuracy of
dA
whereby dU is the amount of energy change and dA is the solution, but also the iterative solving procedure and the
the area increase [23]. The related failure criteria arise from computing time.
the corresponding material-specific critical values, the fracture
toughness Kc and the fracture resistance Gc [5].
Damage mechanics take one step backwards and addition-
ally include microscopic effects, namely the material weak-
ening and the consequent crack initiation. The aim is the
effective modelling of the damage evolution within a material
or structure by levering out the restrictions of both the other
described categories [2]. Thereby it combines the concepts of
continuum and fracture mechanics to describe the damage [5].

III. C OHESIVE Z ONE M ODEL


The cohesive zone model (CZM) is the most widely used Fig. 2. Bilinear and exponential (dashed line) CZM law shapes, modelled
damage mechanics approach for the FE method [5]. It de- according to DA S ILVA [5]
Regardless of the chosen CZM law, convergence issues are
always relevant for the use of CZM [5]. These difficulties are
mainly linked to the damage propagation and instability, the
plasticity of adherends and adhesive, the adherends stiffness,
course meshes and large stress concentrations or singularities.
Composite adherends enlarge the difficulties because of their
orthotropic elasticity. The same applies to brittle adhesives,
which lead to a more abrupt softening process. Fig. 4. ENF test setup, modelled according to DA S ILVA [6].

B. Cohesive Parameters
In order to reasonably use the CZM analysis, attaining
an accurate simulation of the failure process, the cohesive
parameters have to be carefully determined [5]. The adhesive
in the underlying application is the Araldite AW 4859 from
Huntsman Advanced Materials, a two-component epoxy resin
Fig. 5. ELS test setup, modelled according to DA S ILVA [6].
adhesive [15]. Some parameters, such as the cohesive tensile
strength σc , are given in the data sheet. However, particularly
the critical values for the energy release rate G have to be clamped on a linear bearing trolley, sliding free in horizon-
obtained from bonded specimens due to the difference in crack tal direction to annul tensile stresses along the longitudinal
growth in bulk materials and thin adhesive layers. At these, direction [7].
the cracks are restricted to the bond line and exposed to the
strain constraining effects of the adherends.
IV. F INITE E LEMENT A NALYSIS OF A S INGLE -L AP -J OINT
For the implementation of CZM and the investigation of
The presented CZM approach is applied in a finite element
adhesively bonded joints, the fracture resistance in mode I GIc
analysis (FEA). Thereby the adhesive layer is not modelled
as well as in mode II GIIc are needed. For the determination
seperately. Instead a contact region is expanded with a de-
of GIc in structural adhesive joints a standardized test method,
taching condition to simulate the development of damage and
the double-cantilever beam (DCB) test according to ISO 25217
the crack growth through the interface. The investigations
[17], is available. It is suitable for metallic as well as fibre-
are performed using the software ANSYS Workbench V16.1.
composite adherends [4]. The principles of the DCB test are
Since the focus of this work is on the evaluation of the
based on LEFM. The test setup is shown in figure 3. Thereby
methodology, the finite element model is precisely described
both the resistance to crack initiation, either from a non-
in the following.
adhesive insert or a mode I precrack, and the resistance to
crack propagation are analysed. From that, the resistance (R)- A. Finite Element Model
curve of GIc versus the crack length a can also be deduced. The model development is structured in the modules engi-
Yet there is no agreed test standard for the determination neering data, geometry, model and setup. In the engineering
of GIIc in adhesive joints [6]. However, the most popular and data, the material properties are set. The aluminum alloy
considered to be standardized tests for the characterization of AlCu4PbMgMn, with the standardized designation EN AW-
mode II failure are the end-notched flexure (ENF) and the 2007, is chosen for the adherends. The characteristics of
end-loaded split (ELS) test. The test setups of these tests are
schematically depicted in figures 4 and 5.
TABLE I
At the ENF test, which represents a classical 3-point bend- C OHESIVE Z ONE PARAMETERS
ing test, the specimen is centrically loaded. For an alternative Symbol Property Unit Value
design, the 4-ENF test, a 4-point loading is used instead to σc Maximum normal contact stress MPa 40
achieve stable crack growth. In contrast, the ELS specimen is GIc Critical fracture energy for normal
J
425
m2
separation
τc Maximum equivalent tangential MPa 20
contact stress
J
GIIc Critical fracture energy for tangential slip m2
2000
η Artificial damping coefficient s 0.0001
β Tangential slip under compressive - 0
normal contact stress
The parameters are chosen in accordance to the manufacturer’s specifica-
tions [15], recommendations by the software’s representatives [3], [10]
Fig. 3. DCB test setup using load blocks for the application of force, modelled
according to ISO 25217 [17]. and corresponding literature [9].
the cohesive zone are given by establishing a new material
card. Thereby the toolbox for CZM, implemented in the
software, is used. Due to the difficult determination of the
opening displacement at the completion of debonding, the
option fracture-based debonding is favourable in contrast to
separation-distance based debonding. With that the detaching Fig. 7. Mesh of the FE model in the area of the overlap.
behaviour is characterized by the parameters specified in
table I.
0.1 multiplied with the smallest time increment are often
The setting mixed mode is necessary to activate all parame- recommended [3], [10].
ters for both mode I and mode II. The values for both GIc and The geometry, shown in figure 6, is modelled three-
GIIc are estimated according to relevant literature [9]. The dimensional according to the standard DIN EN 1465 which
maximum normal contact stress is equated to the cohesive describes the determination of the lap shear strength of adhe-
tensile strength σc and likewise the maximum equivalent sive joints [16]. The adhesive layer is not modelled separately.
tangential contact stress to the cohesive shear strength τc . Instead, the adhesive joint consists only of the two adherends
Furthermore, there is an option particularly for mixed-mode which are directly located on top of each other and are
debonding which controls tangential slip under compressive connected by a bonded contact. The dimensions are given in
normal contact stress [19]. To prevent that, it is set to zero. As table II. For the implementation of the CZM approach either
mentioned before, the CZM formulation always brings along the pure penalty or the augmented lagrange algorithm, which
convergence difficulties due to the numerical instabilities, are both based on the assignment of a contact stiffness, are
notably when the adhesive bond is close to complete failure thereby permitted [19]. Regarding the mesh, the use of very
[2], [13]. To remedy the difficulties, a numerical solution can uniform elements is recommended for the overlapping area as
be included by applying viscous damping. For this purpose the shown in figure 7 for this study [10]. Here the element type
artificial damping coefficient η is implemented. It influences SOLID186, a hexahedral element with 20 nodes and quadratic
the debonding parameter δv by trial functions, is chosen.
∆t By the use of the contact debonding object of the fracture
δv = δold + · ∆δ (2)
∆t + η tool, the cohesive zone can be introduced for the contact
region applying the predefined cohesive material. Further
with the debonding paramater at the previous substep δold , the
settings, which are not yet implemented in ANSYS V16.1
increment of the debonding paramter in the current substep ∆δ
are carried out by the integration of APDL commands [3]. It
and the time increment ∆t [18]. It has the unit of time and
is advantageous to preset the contact stiffness, videlicet the
shall impede exceeding values of the maximum traction or
gradient of the ascending contact stress. Thereby, particularly
maximum separation respectively the critical fracture energy.
for transient simulations concerning creep or fatigue, the
Indeed, the artificial damping should not impair the overall
Youngs modulus of the adhesive can be taken into account.
behaviour of the structure. To achieve this, it should be smaller
Otherwise the contact stiffness can be adjusted in compliance
than the minimum time step size. Values between 0.001 and
with the maximum numerical stability as it was done for this
study. Furthermore the pinball radius, which describes the
region within nodes are detected for the contact definition, has
to be set [19]. It should be at least twice the critical separation
distance δc .
For the analysis settings, a direct solver was chosen, weak
springs deactivated and large deflection enabled. The latter was
found to be essential for CZM due to the strong deformations
during the debonding process. Because of the high nonlinear-
ity, the load is divided in several load steps. Furthermore a
Fig. 6. Geometry of the FE model. large number of substeps is required [2], [3]. As a boundary
condition a fixed support is attached on one adherend’s end.
TABLE II
A remote displacement at the other adherend’s end inhibits
D IMENSIONS OF THE SINGLE - LAP JOINT all rotations as well as the displacements perpendicular to the
joint’s longitudinal axis, while it loads the joint lengthwise
Dimension Direction Unit Value with a displacement of 0.8 mm.
Length x mm 115
Width y mm 20 B. Results and Discussion
Height z mm 1.5 The deformation during the debonding process shows no
Overlap x mm 12.5 apparent crack initiation before the joint fails completely at
Fig. 8. Force reaction at the remote displacement. Fig. 9. Traction-separation-relation.

a failure displacement of about 0.578 mm. There the bonding Finally, it is necessary to prove the accuracy of the CZM
opens, the adherends’ elastic deformation reverses and from law’s reproduction. For this purpose the contact stress and
then on the adherends slide unhindered along each other until contact gap, which can both be obtained by user defined results
the prescribed displacement is reached. This evinces, that the for a selected node, are plotted against each other. This is
joint fails due to a brittle fracture with a sudden release of the depicted in figure 9 for the performed simulation. It demon-
energy and without a preceding reduction owing to plastic strates the traction-separation-relation and therewith character-
yielding. The reasons for this can be found in the brittle izes the debonding behaviour. It is therefore necessary, that
behaviour of the adhesive and the loading, which is mainly this complies with the predefined CZM law for the main
shear. Therefore it is relatively evenly spread over the bonding features. It shall reproduce the whole debonding consisting of
area and rises very uniformly until the permissible limits are the rising contact stress up to a maximum contact stress, the
reached and the bond opens at once. Before the bonding starts subsequent softening and finally the complete opening. Thus,
to fail, the VON M ISES equivalent stresses within the ad- the calculation of a sufficient number of sampling points is
herends concentrate at the ends of the overlap. This meets the essential. This can be influenced by the analysis settings. For
expectations on the basis of analysed analytical approaches as the case of insufficient CZM law’s reproduction, the number
well as previous numerical investigations following continuum of substeps has to be adjusted. In this case, the adaption led
mechanics. to the demand of a minimum number of substeps of 60 to
The force reaction at the remote displacement can be output 80. Owing to the brittle failure the ascending part is mapped
to determine the failure load. Figure 8 plots its line graph, by many supporting points whereas there are only two points
which shows an almost linear increase of the force without calculated at the descending part of the curve, which is the
any distinguishable crack initiation point until a failure load required minimum. The fragmentation of the calculation into
of approximately 4379 N is reached and the force sharply those large numbers of increments and substeps leads to a high
declines to zero. This confirms the observations from the increase in computation time. This predominates the effect
previously described deformation. The failure load yields the of the low number of required mesh elements compared to
expected scale, but nevertheless deviates significantly from continuum mechanics approaches, where the adhesive layer is
earlier experimental results of about 7000 N for lap shear modelled separately with the consequence of a very fine mesh.
tests of specimens with the given dimensions and material
combination. This is consequently caused by the fact, that the
V. C ONCLUSION AND O UTLOOK
values for the fracture resistance Gc have not been determined
experimentally yet and instead values from literature have been The object of this work was the evaluation of the cohesive
used. zone model for the simulation of adhesively bonded joints.
This damage mechanics approach covers the whole process
The energies released due to failure in mode I and II can
of damage evolution, including the material weakening, the
be displayed and exported if the amounts of the involved
crack initiation and its propagation through the structure. It is
failure modes are of interest. This is executed by inserting user
implemented by expanding a bonded contact with a detach-
defined results for the applied contact elements, which are of
ing condition, the CZM law. This requires certain cohesive
the type CONTA174 in this case. For the investigated single-
parameters. It has been found, that the fracture resistance in
lap joint that results in a released energy of (0.352 ± 0.006) J
mode I GIc can be determined by the standardized DCB test.
for mode I and (15.153 ± 0.116) J for mode II for the whole
In contrast to that, there is no standardized test method for
bonding.
the characterization of mode II failure. However, the two most
popular tests are the ENF test and the ELS test.
Another main issue regarding CZM are the convergence [6] DA S ILVA , L. F. M.; Ö CHSNER , A.; A DAMS , R.: Handbook of Adhesion
difficulties. But there are a few factors, that can influence Technology. Springer, Berlin/Heidelberg, 2011
[7] DE M OURA , M. F.; D OURADO ; N. M.: Mode II Fracture Characteriza-
this iterative process. Smaller step increments can avoid diver- tion of Bonded Joints Using the ELS Test. in: Testing Adhesive Joints:
gence due to unstable crack growth for G surpassing Gc [5]. Best Practices, Eds.: da Silva, L. F. M.; Dillard, D. A.; Blackman, B. R.;
Another assisting method is the implementation of a viscous Adams, R. D., Wiley-VCH, Weinheim, 2012
[8] F ERAREN , P.; J ENSEN , H. M.: Cohesive Zone Modelling of Interface
regularization by the use of an artificial damping coefficient Fracture near Flaws in Adhesive Joints. in: Engineering Fracture Me-
[13]. This value has to be small compared to the smallest chanics, vol. 71 (15) pp. 2125-2142, 2004
time increment for not impairing with the results’ accuracy. [9] F ORS , F.: Master Thesis: Analysis of Metal to Composite Adhesive
Joints in Space Applications. Linköpings universitet, Linköping, 2010
Furthermore, stress concentrations and a distortion of elements [10] G RAF, K.: Seminar Documentation: Bruchmechanische Berech-
can be prohibited by a very uniform mesh and the reduction nungsmethoden. CADFEM GmbH, Grafing, 2015
of the initial stiffness of the cohesive elements. [11] G ROSS , D.; S EELIG , T.: Bruchmechanik. 4th Edition, Springer,
Berlin/Heidelberg, 2011
The reproduction of the given CZM law should be proved by [12] H ABENICHT, G.: Kleben. Grundlagen, Technologien, Anwendungen.
Vieweg+Teubner, Wiesbaden, 2009
the traction-separation-relation, obtained from the output for [13] K HORAMISHAD , H.; C ROCOMBE , A. D.; K ATNAM , K. B.; A SHCROFT,
the contact stress and contact gap. Although the consequent I. A.: Predicting Fatigue Damage in Adhesively Bonded Joints Using a
incremental division of the solution and the necessity of Cohesive Zone Model. in: International Journal of Fatigue, vol. 32 (7),
pp. 1146-1158, 2010
updating the contact’s constraint equations for every step [14] M ITTAL , K. L.: Adhesive Joints: Formation, Characteristics and Test-
causes a largely increased computation time, this is required to ing. 2nd Edition, VSP, Utrecht/Boston, 2002
achieve a veridical representation of the debonding behaviour. [15] N. N.: Data Sheet: Araldite R
AW4859/ Hardener HW4859 – Structural
Adhesives. Huntsman Advanced Materials, Basel, 2011
This could indeed be obstructive for the feasible usage of CZM [16] N. N.: DIN EN 1465: Klebstoffe – Bestimmung der Zugscherfestigkeit
for more complex geometries and real applications. Thus, it von Überlappungsklebungen. Beuth Verlag, Berlin, 2009
has to be analysed how the computation time can still be [17] N. N.: ISO 25217: Adhesives – Determination of the Mode 1 Adhesive
Fracture Energy of Structural Adhesive Joints Using Double Cantilever
diminished. Beam and Tapered Double Cantilever Beam Specimens. Beuth Verlag,
Berlin, 2009
R EFERENCES [18] N. N.: Program Documentation: ANSYS Help for Release 17. ANSYS
[1] A NDERSON , T. L.: Fracture Mechanics: Fundamentals and Applica- Inc., Canonsburg (USA), 2016
tions. 2nd Edition, CRC Press, Boca Raton (USA), 1995 [19] N. N.: Program Documentation: ANSYS Mechanical APDL Theory
[2] A NYFANTIS , K. N.: PhD Thesis: Analysis and Design of Composite- Reference. Ed: Kohnke, P., ANSYS Inc., Canonsburg (USA), 2013
to-Metal Adhesively Bonded Joints. National Technical University of [20] PARISCH , H.: Festkörper-Kontinuumsmechanik. Von den
Athens, School of Naval Architecture and Marine Engineering, Athens, Grundgleichungen zur Lösung mit Finiten Elementen. Teubner,
2012 Stuttgart/Leipzig/Wiesbaden, 2003
[3] BARTHOLD , U.: Seminar Documentation: Berechnung von Klebe- [21] ROMANOS , G.: Fügetechnik – Kleben. in: Handbuch Konstruktion, hg.
verbindungen. CADFEM GmbH, Grafing, 2015 von Steinhilper, R.; Rieg, F., Carl Hanser Verlag, Mnchen, 2012
[4] B LACKMAN , B. R. K.; K INLOCH , A. J.: Fracture tests on structural [22] WAHAB , M. A.: The Mechanics of Adhesives in Composite and Metal
adhesive joints. in: Fracture of Polymers, Composites and Adhesives Joints. Finite Element Analysis with ANSYS. DEStech Publications,
II, Eds.: Moore, D.R.; Pavan, A.; Williams, J.G., ESIS Publication 28, Pennsylvania, 2014
Elsevier, Amsterdam, 2001 [23] W ILLIAMS , J. G.: Introduction to Linear Elastic Fracture Mechanics.
[5] DA S ILVA , L. F. M., C AMPILHO , R. D.: Advances in Numerical Mod- in: Fracture Mechanics Testing Methods for Polymers, Adhesives and
elling of Adhesive Joints. SpringerBriefs in Applied Sciences and Composites, Eds.: Moore, D. R.; Pavan, A.; Williams, J. G., ESIS
Technology, Springer, Berlin/Heidelberg, 2012 Publication 28, Elsevier, Amsterdam, 2001

You might also like