Académique Documents
Professionnel Documents
Culture Documents
Abstract
This paper presents a new model for three-dimensional finite element analysis of adhesive joints. The model considers geometric
and material nonlinearities and uses solid brick elements as well as specially developed interface elements. The interface elements
allow the calculation of stresses at the adherend–adhesive interfaces. The application of the model to a single-lap joint is presented.
The results of a linear elastic analysis highlight the three-dimensional nature of the stresses and stress concentrations at interfaces.
The influence of material nonlinearities on the behavior of the joint is also discussed.
r 2002 Elsevier Science Ltd. All rights reserved.
0143-7496/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 7 4 9 6 ( 0 2 ) 0 0 0 1 5 - 5
358 J.P.M. Gonc-alves et al. / International Journal of Adhesion & Adhesives 22 (2002) 357–365
shear stiffnesses, which are dependent on the thickness elements based on displacements do not satisfy the
of the adhesive layer. Carpenter [10] developed a simple continuity of the stress vector at the interfaces. Cheikh
element that considered the adhesive with no thickness. et al. [22] proposed a technique, based on the finite
Later, the same author [11] developed another element element method formulated in terms of displacements
that considers the adhesive thickness and a linear that enables stress conditions to be imposed. They used
variation of the displacement field across it. Carpenter two-dimensional analysis and applied their technique in
and Barsoum [12] developed two specific elements that the study of the stress distributions at the adherend–
incorporate some assumptions used in several analytical adhesive interfaces.
studies. These elements can be used with two-dimen- In this paper, a new model for three-dimensional finite
sional plane stress or plane strain elements, as well as element analysis of adhesive joints is presented. The
with shell or beam elements. Rao et al. [13] developed a model considers the stress variation across the adhesive
six-node isoparametric element similar to the one thickness as well as geometric and material nonlinea-
presented in [9]. Yadagiri et al. [14] modified that rities. Adherends and adhesive are represented by solid
element to include longitudinal normal stresses and brick elements and the interfaces are modeled by special
viscoleastic analyses. Lin and Lin [15] introduced an interface elements. Interface elements are included in
element that represents both the adhesive layer and the order to allow the calculation of stresses at the
adherends and that is based on Timoshenko’s beam adherend–adhesive interfaces. The application of the
theory. Reddy and Roy [16, 17] developed a special model to a single-lap joint is presented. The results
element for geometric and material nonlinear analyses. highlight the three-dimensional nature of the stresses
All the finite element analyses referred to above are and the influence of material nonlinearities on the
two-dimensional. However, even for a simple adhesive behavior of the joint.
joint such as the single-lap joint, stresses are of three-
dimensional nature. This was investigated by Tsai and
Morton [18] who performed three-dimensional linear 2. Interface finite element
elastic finite element analysis of a single-lap joint in
which boundary conditions account for the geometric An interface finite element for three-dimensional
nonlinear effects. These authors showed that three- analysis has been previously developed [23, 24]. It
dimensional regions exist in the specimen. They showed consists of 18 nodes distributed in two faces as shown
that the adherend and adhesive stress distributions in in Fig. 1. The element has zero thickness and is
the overlap, near the free surface, are quite different compatible with 27 node isoparametric hexahedral
from those occuring in the interior. Pandey and elements available in the ABAQUSs software [25]. Its
Narasimhan [19] also contributed to the identification formulation can be viewed as a contact problem as
of the three-dimensional nature of stresses in adhesively detailed below.
bonded single-lap joints by performing three-dimen-
sional viscoplastic analysis considering material and
geometric nonlinearities. 2.1. Theory
Special elements that model the adhesive layer were
also developed for three-dimensional analyses. Edlund The equilibrium equations of a structural problem
and Klarbring [20] developed an element for geometric involving contact can be formulated as the minimization
nonlinear analysis. They assumed that the adhesive layer of the potential energy subjected to certain kinematic
is of negligible thickness and has linear elastic behavior. constraints. Considering the variational method and
Andruet et al. [21] developed a model for three- using the penalty function method, the potential energy
dimensional analysis of adhesive joints based on shell
and solid elements. The shell elements are used to model
the adherends and the adhesive layer is modeled as a 16
solid element with offset nodes in the mid-planes of the 15
adherends. The element formulation includes geometric 17
7
nonlinearities, and thermal and moisture effects. 14
18 6
In addition to considering the three-dimensional 10 8 13
nature of adhesive joints, the calculation of accurate 11 9 5
stresses at the adherend–adhesive interfaces is of
primary concern. The failure of adhesive joints usually 1 12 4
occurs in the adhesive (cohesive failure) or at the 2
adherend–adhesive interfaces (adhesive failure). More-
3
over, cohesive failure usually occurs near the adherend–
adhesive interface. However, conventional finite Fig. 1. The interface element.
J.P.M. Gonc-alves et al. / International Journal of Adhesion & Adhesives 22 (2002) 357–365 359
vector, K the stiffness matrix, F the nodal forces, Kr ¼ e Ck nk nTk CTk ; ð12Þ
GðUÞ ¼ 0 the kinematic constraint, and e the penalty k¼1
parameter. X
d
The kinematic constraint for a pair (k) of homologous Fr ¼ 'e g k C k nk : ð13Þ
points is [26, 27] k¼1
ðU2i ' U1j Þ ( nk þ gk ¼ 0; ð3Þ It can be stated that Kr and Fr represent the
contribution of the constraints to the global problem.
where U2i ; U1j are the displacement vectors of two Thus, the interface finite element formulation is the
homologous points (i and j) belonging to the two determination of its contribution to the global problem
contacting bodies (1 and 2), nk the unit vector normal to by matrices Kr and Fr :
the contact surface, and gk the initial distance in the
normal direction n between two homologous points (i 2.2. Stresses at the interfaces
and j) belonging to the two contacting bodies.
The equilibrium state is obtained from the minimiza- Contrary to continuum elements where stress–strain
tion of Eq. (1) relative to the displacement field (U). It relationships are used, interface elements are governed
can be shown that, when e approaches infinity, the by stress-relative displacements relationships. The
minimization of the functional pðUÞ
% produces a better vector of relative displacements between two
approximation of GðUÞ ¼ 0: The vectors U2i and U1j can homologous points can be obtained from the displace-
be written ments fields associated with the element faces (top and
U2i ¼ eTi U; ð4Þ bottom)
8 9 8 9 8 9
< ds >
> = > < us >
= < us >
> =
U2j ¼ eTj U; ð5Þ d ¼ dt ¼ ut ' ut ; ð14Þ
: >
> ; > : > ; : >
> ;
where ei and ej are the three columns of the identity dn un top un bot
matrix that allow the calculation of the displacement
where s and t represent the tangential directions and n
vector of the homologous points i and j; respectively.
the normal direction. The displacements in the local
Eq. (3) can be rewritten as
directions (normal and tangential to the contact surface)
ðCTk UÞ ( nk þ gk ¼ 0; ð6Þ can be obtained from the generalized displacements
doing
where 8 9 8 9
< us >
> = <u>
> =
CTk ¼ eTi ' eTj : ð7Þ T
ut ¼ h v ; ð15Þ
: >
> ; : >
> ;
Considering the total number of constraints (d), the un w
functional for the problem being treated can be obtained
as where
h ¼ ½Vs ; Vt ; Vn &: ð16Þ
pðUÞ
% ¼ pðUÞ
Vs ; Vt and Vn are the unit vectors of the local directions
eXd
þ ½ððCTk UÞ ( nk þ gk ÞT ððCTk UÞ ( nk þ gk Þ&: ð8Þ (see Fig. 2) and are obtained from
2 k¼1 8 9
> qx >
>
> >
Minimizing this equation leads to > qx >
> >
>
< qy >
> =
dpðUÞ
% ¼ dpðUÞ Vx ¼ ; ð17Þ
>
> qx >
X
d > >
> >
>
> >
þe ½UT Ck nk nTk CTk þ gk nTk CTk &dU ¼ 0 ð9Þ : qz >
> ;
k¼1 qx
360 J.P.M. Gonc-alves et al. / International Journal of Adhesion & Adhesives 22 (2002) 357–365
Adhesive
Adherend 1.5
0.25
Adherend 1.5
C D
B x
A
1.5 3
2.5 z = 0.176 mm
z = 0.176 mm
z = 6.426 mm z = 6.426 mm
1 2
z = 10.761 mm z = 10.761 mm
z = 11.719 mm 1.5 z = 11.719 mm
σyy /τ average
σxx/τ average
z = 12.324 mm z = 12.324 mm
0.5 1
0.5
0 0
_0.5
_ 0.5 _1
(a) 0 2 4 6 8 10 12 14 (b) 0 2 4 6 8 10 12 14
x (mm)
1.5 2.5
z = 0.176 mm z = 0.176 mm
z = 6.426 mm z = 6.426 mm
1 2
z = 10.761 mm z = 10.761 mm
z = 11.719 mm z = 11.719 mm
xy average
σzz /τ average
z = 12.324 mm z = 12.324 mm
0.5 1.5
τ /τ
0 1
_0.5 0.5
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
(c) x (mm) (d) x (mm)
0.06 0.8
z = 0.176 mm z = 0.176 mm
z = 6.426 mm 0.6 z = 6.426 mm
0.04
z = 10.761 mm z = 10.761 mm
z = 11.719 mm 0.4 z = 11.719 mm
0.02 z = 12.324 mm z = 12.324 mm
0.2
τ yz /τ average
τxz /τ average
0 0
_0.2
_0.02
_0.4
_0.04
_0.6
_0.06 _0.8
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
(e) x (mm) (f) x (mm)
Fig. 6. (a) Normalized stress sxx at the middle of the adhesive thickness. (b) Normalized stress syy at the middle of the adhesive thickness. (c)
Normalized stress szz at the middle of the adhesive thickness. (d) Normalized stress txy at the middle of the adhesive thickness. (e) Normalized stress
txz at the middle of the adhesive thickness. (f) Normalized stress tyz at the middle of the adhesive thickness.
middle of the adhesive layers (see Figs. 6(f) and 7(c)) static force of 2.9 kN. A different yield criterion was
have similar distributions. used for each material. The Von Mises yield criterion
was used for the adherends, and the Raghava et al. [28]
3.2. Elasto-plastic materials parabolic criteria was used for the adhesive. This
criterion can be written as
Taking into consideration the elasto-plastic materials’
ðs1 ' s2 Þ2 þ ðs2 ' s3 Þ2 þ ðs3 ' s1 Þ2
behavior, a material and geometric nonlinear finite
element analysis was performed considering an applied þ 2ðjsc j ' st Þðs1 þ s2 þ s3 Þ ¼ 2jsc jst ; ð25Þ
J.P.M. Gonc-alves et al. / International Journal of Adhesion & Adhesives 22 (2002) 357–365 363
12 8
10 7 z = 0 mm
z = 0 mm
z = 6.25 mm z = 6.25 mm
8 6
z = 10.156 mm z = 10.156 mm
z = 11.719 mm 5 z = 11.719 mm
6
τxy /τ average
yy average
z = 12.5 mm z = 12.5 mm
4 4
σ /τ
2 3
0 2
_2 1
_4 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
(a) x (mm) (b) x (mm)
0.8
z = 0 mm
0.6 z = 6.25 mm
z = 10.156 mm
0.4
z = 11.719 mm
z = 12.5 mm
0.2
τ yz /τ average
0
_0.2
_0.4
_0.6
_0.8
0 2 4 6 8 10 12 14
(c) x (mm)
Fig. 7. (a) Normalized stress syy at the interface y ¼ 0:125 mm. (b) Normalized stress txy at the interface y ¼ 0:125 mm. (c) Normalized stress tyz at
the interface y ¼ 0:125 mm.
where s1 ; s2 and s3 are the principal stresses, and sc and In Figs. 10(a) and (b), the normalized stresses syy and
st are the compression and tensile strengths, respec- txy at the upper interface (y ¼ 0:125 mm) and obtained
tively. In this criterion, a ratio of 1.3 between the at the maximum load (P ¼ 2:9 kN) are presented. The
compression and tensile strengths was assumed. The peak value of the normalized stress syy varies between
progressive plastification of the materials was analyzed 1.8 kN (z ¼ 0:0 mm) and 3.8 kN (z ¼ 12:5 mm) and an
at several stages of the loading history. In Figs. 8(a) and almost constant behavior is observed for the txy
(b) a schematic representation of the plastic zones in the normalized stress. As expected, a general decrease of
adhesive layer is presented for two different load levels. stress levels is observed relative to the linear case (see
The gray element faces show that at least some of the Figs. 7(a) and (b)).
integration points closer to those faces have plastified.
In Fig. 8(a) it can be seen that the plastification starts at
the ends of the adhesive layer (points A and D in Fig. 4). 4. Conclusions
Fig. 8(b) shows that the plastification grows toward the
longitudinal symmetry plane and across the adhesive A new three-dimensional finite element model for
thickness. studying the behavior of adhesive joints is presented.
The maximum normalized stresses syy and txy at the The model includes an interface element previously
upper interface (y ¼ 0:125 mm) for different load levels developed that is compatible with brick solid elements
are presented in Fig. 9. It can be stated that the peak from the ABAQUSs software. The main objective was
normalized stresses decrease as the applied load to calculate the stresses at the interfaces between
increases. Also, for the higher load levels the peak adherends and adhesive, which are considered critical
normalized stresses shift from the symmetry plane to the regions in these structures. An application to single-
free edge. lap joints was performed considering two different
364 J.P.M. Gonc-alves et al. / International Journal of Adhesion & Adhesives 22 (2002) 357–365
σyy/τ average
_2 z = 0 mm
z = 6.25 mm
z = 10.156 mm
_4 z = 11.719 mm
z = 12.5 mm
_6
0 2 4 6 8 10 12 14
(a) x (mm)
z = 0 mm
3 z = 6.25 mm
z = 10.156 mm
z = 11.719 mm
τxy /τ average 2 z = 12.5 mm
12 Fig. 10. (a) Normalized stress syy at the interface y ¼ 0:125 mm for
P ¼ 2:9 kN. (b) Normalized stress txy at the interface y ¼ 0:125 mm for
Max (σ /τ ) P ¼ 2:9 kN.
10 yy average
Max (τ /τ )
Max (normalized stress)
xy average
8
much higher than at the middle of the adhesive layer.
6 This can explain why the interfaces are critical regions
regarding adhesive joints failure.
4 The analysis with elasto-plastic materials’ behavior
z = 10.938 mm
z = 12.5 mm
z = 12.5 mm
z = 0 mm
z = 0 mm
z = 0 mm
z = 0 mm
z = 0 mm
z = 0 mm