You are on page 1of 9

An analytical model for the strength prediction of hybrid (bolted/bonded)

composite joints
Christophe Bois

, Herv Wargnier, Jean-Christophe Wahl, Erwann Le Goff


Universit de Bordeaux, Institut de Mcanique et dIngnierie de Bordeaux, 15 rue Naudet, CS 10207, 33175 Gradignan Cedex, France
a r t i c l e i n f o
Article history:
Available online 7 November 2012
Keywords:
Bolted joints
Hybrid joints
Load transfer
Failure criteria
a b s t r a c t
In some applications, aircraft manufacturers plan to use the interface material (adhesive joint) as a struc-
tural part to increase mechanical joint efciency. The design and dimensioning of hybrid joints would
therefore benet from a time-efcient model allowing designers to test and compare various proposals,
and thus optimize the assemblies. This paper proposes an analytical model as an alternative to complex
nite element models.
The model includes the nonlinear behaviour required to evaluate the bolt load transfer contribution up
to failure. Each expected failure mode is taken into account by using a multi-criteria approach in which
each related criterion is identied from bonded or bolted tests.
The ability of the analytical model to predict load transfer in bolt and adhesive joints is established by
comparing analytical model results to those obtained by nite element analysis. The whole model is val-
idated experimentally in terms of joint efciency prediction for two different hole diameter-to-width
ratios.
2012 Published by Elsevier Ltd.
1. Introduction
Most aeronautic mechanical assemblies made of fasteners
(bolts, rivets, thread insert or bushings) include an adhesive or li-
quid shim layer to ensure sealing and avoid fretting corrosion phe-
nomena. The structural function of this interface material is rarely
studied and almost never taken into account when predicting
assembly strength.
In some applications, aircraft manufacturers plan to use the
interface material (adhesive joint) as a structural part to reduce
the number of fasteners or increase mechanical joint efciency.
In such cases, the assembly loading is transmitted through both
the fasteners and the adhesive material: the joint is dened as a
hybrid joint.
There are two main stages in the dimensioning of hybrid joints.
The rst is to evaluate the contribution of the two parallel load
transfer mechanisms through the fasteners and adhesive joint.
The second deals with the assessment of the corresponding failure
strength. For instance, predicting the strength of a bonded/bolted
single or double lap joint requires various studies of strength con-
sidering the bolt (shear failure), each adherend (bearing, net-
section failure) and the adhesive (cohesive or adhesive failure).
Unfortunately, the use of hybrid joints leads to complex prob-
lems because the contributions of each load transfer mechanism
depend on the nonlinear behaviour of constituents. In an example
presented previously, it was shown that adhesive plasticity leads
to a reduction in the effective stiffness of the adhesive joint and
consequently increases the contribution of the bolt in the assembly
load transfer [1,2]. Some authors have proposed nite element
modelling (FEM) to approach this problem and highlight the inu-
ence of the main parameters [13]. These studies, based on com-
plex models and huge calculation times, show that the
distribution of load transfer between adhesive joint and bolt is
strongly dependent on adhesive behaviour (modulus and non-
elastic strain), adhesive thickness, overlap length and bolt diameter
to joint width ratio. Consequently, a hybrid joint is able to exhibit
greater strength only if design parameters are judiciously chosen.
Thus, the design and dimensioning of hybrid joints would benet
from time-efcient models allowing designers to test and compare
various proposals, and moreover to optimize the assemblies. Barut
and Madenci [4] proposed a semi-analytical 2-D model dedicated
to single-lap hybrid joint. In this model, the adherent and bolt dis-
placements are based on the Mindlin plate and Timoshenko beam
theories, respectively. However, material behaviour of constituents
was assumed linear and no failure criterion was associated to this
model.
In this paper, a semi-analytical 1-D model of a hybrid double lap
joint including a multi-criteria approach is proposed to predict
hybrid joint strength. The model includes nonlinear behaviour of
0263-8223/$ - see front matter 2012 Published by Elsevier Ltd.
http://dx.doi.org/10.1016/j.compstruct.2012.10.022

Corresponding author. Tel.: +33 5 56 84 79 78; fax: +33 5 56 84 58 43.


E-mail addresses: christophe.bois@u-bordeaux1.fr (C. Bois), herve.wargnier@
u-bordeaux1.fr (H. Wargnier), jean-christophe.wahl@u-bordeaux1.fr (J.-C. Wahl),
erwann.le-goff@etu.u-bordeaux1.fr (E. Le Goff).
Composite Structures 97 (2013) 252260
Contents lists available at SciVerse ScienceDirect
Composite Structures
j our nal homepage: www. el sevi er . com/ l ocat e/ compst r uct
both the adhesive joint and the composite material in order esti-
mate the bolt load transfer contribution throughout a quasi-static
loading. The prediction of the adhesive joint failure is based on a
maximum shear strain criterion. The prediction related to bearing
and net-section failure of the composite part is dissociated follow-
ing two different criteria. Bearing failure is approached in the
traditional way using a maximum pressure criterion, whereas
net-section failure is processed with a stress concentration factor
approach adapted to the specic behaviour of composite materials.
The overall model is rst validated in terms of load transfer with a
3-D FEM model, then in terms of strength prediction accuracy with
tensile tests on a hybrid joint with two different geometrical
congurations.
2. Description of semi-analytical model
The study concerns a double lap joint, as illustrated in Fig. 1.
Bolt diameter d, adherent thicknesses t
1
and t
2
and overlap length
L were selected in order to avoid shear failure of the bolt and shear-
out failure of the adherents. Thus, as shown in Fig. 2, expected fail-
ure modes are bearing degradation, net-section failure and adhe-
sive joint failure. The aim of the model is to determine loads
transmitted in the assembly for each of the three failure modes.
Assembly load F is divided into a shear load transmitted by the
adhesive joint and a contact load transmitted by the bolt. However,
the load supported by the net-section is the sum of that transmit-
ted by the adhesive before the bolt, and that transmitted by the
bolt. Thus the model evaluates the adhesive shear stress along
the overlap length in order to determine the load supported by
the net-section.
We use the 1-D model proposed by Paroissien et al. [5,6]. It rep-
resents a hybrid joint consisting of one or several bolts. Pure tensile
load is assumed in the adherents while pure shear load is assumed
in the adhesive joint. Each bolt is modelled by a spring that gener-
ates a force F
i
between the two adherents on the section located at
x = d
i
as shown in Fig. 3. The contribution of bolt number i to the
load transfer is evaluated using the transfer rate s
i
dened as
follows:
F
i
s
i
F k
i
u
1
i
d
i
u
2
i
d
i
1
where k
i
is the stiffness of bolt number i, u
1
i
d
i
and u
2
i
d
i
are the
displacements of the two adherents in the section located at x = d
i
.
If the assembly involves n bolts, the overlap length is divided
into n + 1 portions. In each portion of part 1, the mechanical equi-
librium gives:
dN
1
i
x
dx
T
i
xw 0 2
where N
1
i
x and T
i
(x) are respectively half of the tensile load in
part 1 and the shear stress in the adhesive joint between bolts i
and i + 1, and w is the assembly width.
The mechanical equilibrium is completed by the global equilib-
rium of the joint:
N
1
i
x N
2
i
x
F
2
3
where N
2
i
x is the tensile load in adherent number 2. By
differentiating on each portion the equilibrium Eq. (2) in which
material constitutive equations of the adhesive were introduced,
we obtain:
d
2
N
1
i
x
dx
2

G
t
w
du
1
i
x
dx

du
2
i
x
dx
_ _
0 4
where G is the shear modulus of the adhesive and t the adhesive
thickness. By introducing the constitutive equations of each adher-
ent in Eq. (4) we obtain:
d
2
N
1
i
x
dx
2

G
t
2N
1
i
x
t
1
E
1

N
2
i
x
t
2
E
2
_ _
0 5
where E
1
and E
2
are the Young modulus of adherent number 1 and 2
respectively, and t
1
and t
2
are the thickness of adherent number 1
and 2 respectively. Incorporating Eq. (3) into Eq. (5) the following
second order differential equation is obtained:
Fig. 1. Dimensions and instrumentation of double lap joint.
Fig. 2. Failure modes of hybrid double lap joint.
C. Bois et al. / Composite Structures 97 (2013) 252260 253
d
2
N
1
i
x
dx
2
g
2
N
1
i
x c
F
2
6
with
g
2

G
t
2
t
1
E
1

1
t
2
E
2
_ _
7
and
c
G
t
2
E
2
t
8
The general solution of Eq. (6) is:
N
1
i
x A
i
e
gx
B
i
e
gx

c
g
2
F
2
9
Thus, we obtain 2n + 2 unknowns A
i
and B
i
. The adhesive shear
stress is expressed by incorporating Eq. (9) into Eq. (2):
T
i
x
g
w
A
i
e
gx
B
i
e
gx
10
Adhesive shear stress continuity at each bolt section located at x = d
i
gives n equations:
e
gd
i
A
i
e
gd
i
B
i
e
gd
i
A
i1
e
gd
i
B
i1
0 11
The bolt behaviour dened by Eq. (1) can be expressed as follows
using the constitutive equation of the adhesive and Eq. (10):
s
i
F
2

tg
wG
k
i
A
i
e
gd
i
B
i
e
gd
i
_ _
12
Moreover, the mechanical equilibrium of bolt number i gives:
N
1
i1
d
i
N
1
i
d
i
s
i
F
2
13
In incorporating Eq. (12) into Eq. (13), we obtain n equations:
1
tg
wG
k
i
_ _
e
gd
i
A
i
1
tg
wG
k
i
_ _
e
gd
i
B
i
e
gd
i
A
i1
e
gd
i
B
i1
0 14
Two other equations are obtained by writing the free tensile load at
x = 0 of the adherent and total load at x = L:
A
1
B
1

c
g
2
F
2
15
A
n1
e
gL
B
n1
e
gx
1
c
g
2
_ _
F
2
16
Eqs. (11), (14), (15) and (16) give a system with 2n + 2 equations
and 2n + 2 unknowns. As written above, equations assume a linear
behaviour for the adhesive and bolts. However, the model formula-
tion enables a linear elastoplastic behaviour for the adhesive to be
incorporated in the form of a bilinear law by introducing a plastic
shear modulus G

as follows:
T
i
x
G
t
u
1
i
x u
2
i
x if T
i
x 6 T
p
T
i
x T
p

t
u
1
i
x u
2
i
x tc
p
_ _
if T
i
x > T
p
Gc
p
_
_
_
17
where T
p
is the shear yield stress and c
p
is the shear yield strain. It is
worth noting that it is fundamental to incorporate an elastoplastic
behaviour for the adhesive in order to study a hybrid joint. In fact,
plastic strains generate a loss of adhesive joint stiffness that leads to
an increase in bolt load transfer rate. In order that the general solu-
tion (9) remains valid, based on the assumption that plastic zones
start from the ends of joints, two more portions are introduced,
delineated by abscissa x
p1
and x
p2
, to separate the central elastic
zone (modulus G) and external plastic zones (modulus G

). Thus,
the overlap length is divided into n + 3 portions which gives a sys-
tem with 2n + 6 equations. The adhesive shear stress continuity and
the tensile load continuity at abscissa x
p1
and x
p2
give 4 additional
equations.
The model resolution consists in progressively increasing the
force applied to the joint. Initially, x
p1
= 0 and x
p2
= L, and for each
load increment, x
p1
is increased and x
p2
is decreased in an iterative
scheme until T(x
p1
) = T
p
and T(x
p2
) = T
p
. The iterative scheme also
enables nonlinear bolt stiffness to be incorporated in the model.
Paroissien et al. [5,6] applied this model to metal/metal assem-
blies. The authors emphasize that the ability of the model to predict
load transfer is inherent in identication of the adhesive behaviour
and the bolt stiffness. These two points are developed next.
3. Adhesive behaviour modelling
The adhesive used in this study is 3M 2216 structural epoxy. It
has a nonlinear behaviour with a high ultimate strain. The adhesive
behaviour is modelled by a linear elastoplastic law identied
through a tensile test performed on bulk specimens. The model
and the experimental curve are compared in Fig. 4.
The related shear behaviour is deduced from tensile behaviour
using an isotropic hypothesis and Poisson coefcient equal to
0.32. The yield stress and plastic strain are derived from the Von
Mises yield criterion.
It should be noted that while constitutive laws for materials are
fundamental when estimating the load transfer contribution of the
adhesive joint and the bolt, joint strength prediction also relies on
failure criteria.
Many studies deal with strength prediction of adhesive joints.
Concerning ductile adhesive, the consensus tends to recommend
using a maximum strain criterion rather than stress or energy
based criteria [710]. This type of criterion can be easily combined
with elastoplastic behaviour. The analytical model presented be-
low gives only the shear stress and strain and disregards peeling
stress which strongly inuences the joint fracture in some cong-
urations. However, the double lap joint conguration with ductile
adhesive reduces the effect of piling stress on the joint fracture.
Thus, the adhesive failure criterion can be written as:
c
a

c
xz
c
r
18
Fig. 3. Description of 1-D model of hybrid double lap joint.
254 C. Bois et al. / Composite Structures 97 (2013) 252260
where c
r
is the ultimate shear strain and c
xz
the shear strain applied
to the adhesive. The tensile test used to identify elastoplastic behav-
iour is not pertinent when evaluating ultimate shear strain. Indeed,
the moulding process generates a large volume fraction of porosity
in specimens, which alters the adhesive behaviour when the ulti-
mate load is approached. The ultimate shear strain was identied
from double lap adhesive joint tests. The mean ultimate force ob-
tained for four tests was used as the input of the analytical model
described above where the effect of the bolt was removed by
addressing a very low stiffness k
i
. Thus, the shear strain eld along
the adhesive joint was plotted at the ultimate experimental force.
The ultimate shear strain of the adhesive c
r
was identied as the
maximal shear strain located at the ends of aluminium adherents.
As expected for a ductile adhesive, a high value c
r
= 0.72 was
obtained.
4. Identication of the bolted joint behaviour
4.1. Bolted joint stiffness and bearing degradation of composite
material
Evaluating bolted joint stiffness is a difcult problem. Fig. 5
highlights the different mechanical phenomena involved in the rel-
ative displacement of adherents. Relative displacement results
from the combination of contact crushing, concentrated strains in
adherents around the hole and also shear and bending strains of
the bolt. At the macroscopic scale, bolted joint stiffness can be de-
ned by:
k
F
DL
a
19
where DL
a
, dened in Fig. 5, is the displacement induced by the
bolted joint. This stiffness, as previously dened, can be derived
through modelling or experimentation, provided that the sample
is judiciously instrumented.
From a modelling point of view, some analytical or semi-
analytical formulations are proposed to evaluate the displacement
induced by a bolted joint [11,12]. These formulations are generic so
that different materials and various dimensions can be dealt with
in a single formula. However, these formulae are based on quite
simple models associated to rough assumptions and sometimes
non-generic empirical corrections. The formula proposed by
Nelson et al. [11] was applied to the double lap joint studied
here and the stiffness found was about three times lower than
the value identied below with both FEM and instrumented tests.
This difference is acceptable in a multi-bolt (non-hybrid) joint,
since by introducing the same error on all the bolt stiffnesses,
the distribution of load transfer between bolts predicted by the
model will be reliable. In a hybrid joint the distribution of load
transfer is the result of a competition between adhesive and bolts.
Thus, the stiffness of the bolts must be evaluated more precisely.
Moreover, analytical formulations exclude stiffness decrease
due to bearing damage in composite material [11,13]. In the pres-
ent case, nonlinear material behaviours have to be taken into ac-
count because they lead to a reduction in bolt assembly stiffness
when the load increases. It is therefore preferable to analyse bolt
assembly stiffness through experimental tests and then compare
them to simulations. Using this method it is also possible to ana-
lyse degradation mechanisms involved in stiffness change [1418].
Fig. 6 shows load versus extensometer measurement for a dou-
ble lap joint in accordance with Fig. 1. The composite adherent is
made of T700GC/M21 CFRP with the following stacking sequence:
[90/45/0/45]
2s
. Bolt-hole clearance is 0.02 mm and bolt tighten-
ing corresponds to a torque of 1.3 Nm. The hole diameter-to-width
ratio d/w is 0.21 in order to avoid net-section failure. The curve
plotted in Fig. 6 shows ve stages: (1) load increase before sliding,
(2) sliding with contact setting, (3) quasi-linear loading, (4) bearing
degradation with progressive stiffness decrease, (5) bearing degra-
dation at quasi-constant load.
By subtracting from the extensometer measurement the dis-
placement due to the strain of the aluminium adherent, a good
estimation of the relative displacement DL
a
can be obtained. In or-
der to isolate the bolt behaviour, the friction load and the clearance
displacement were directly identied on the load versus displace-
ment curve, as proposed by McCarthy and Gray [13], and then sub-
tracted from the total load and the displacement measurement
respectively. Following this procedure, the third curve shown in
Fig. 6 was obtained.
Taking bearing degradation into account requires analysing
damage mechanisms involved in this degradation mode. Bearing
damage is associated with a progressive degradation of the com-
posite material [19,20]. The rst damage mechanism is diffuse
damage (matrix cracking, bre decohesion) which has no effect
on load versus displacement curves. The progressive decrease in
stiffness observed during stage (4) is due to bre failure (micro-
buckling) in 0 plies and 45 plies and delamination propagation.
When the composite is conned, the delamination onset threshold
is increased and the delamination propagation rate is slowed.
In order to highlight the effect of tightening torque and clear-
ance, several tests were performed with three tightening torque
values (0 Nm, 1.3 Nm and 3.5 Nm) and two clearance values
(0.02 0.02 mm and 0.15 0.02 mm). The results are gathered in
Fig. 7 and show that the bearing initiation load dened as the start
of stage (4) and the maximum load are almost unchanged when
the bolt torque increases. The slight increase in bearing initiation
load can be attributed to the friction load transmitted between
the two adherents [13]. However, load versus displacement curves
show that when the tightening increases, the transition between
stages (4) and (5) is slightly more gradual. It is signicant that
without any tightening, the composite material is nevertheless
conned between adherents since the compressive in-plan stress
generated by the pin leads to a compressive out-of-plan stress by
the Poisson effect. For comparison, the values obtained with a pure
bearing test are also plotted in Fig. 7. It can be concluded that it is
not necessary to take into account the tightening torque and the
clearance value as long as the bolt ensures composite connement
and clearance is low compared to hole diameter [14]. The bearing
criterion can then be derived from the denition of mean bearing
stress p
a
:
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
2
4
6
8
10
12
14
16
18
20
Strain
S
t
r
e
s
s

(
M
P
a
)
experiment
linear elastoplastic model
Fig. 4. Comparison between model and experimental tensile curves of 3M 2216
adhesive.
C. Bois et al. / Composite Structures 97 (2013) 252260 255
c
b

F
i
p
a
t
1
d
20
where c
b
is the bearing criterion, F
i
is the load applied to bolt num-
ber i, p
a
is the allowable bearing stress and t
1
is the composite
thickness.
Allowable bearing stress depends on design requirements. In
some cases maximum hole deformation is specied and in others
the objective is to evaluate strength margins. Let us now consider
two allowable bearing stresses: one matching a bearing initiation
leading to p
a
= p
b1
= 433 MPa and another matching a bearing
strength leading to p
a
= p
b2
= 1000 MPa. These values are in good
agreement with those found in the literature [21,22].
In a hybrid joint conguration, the prediction of allowable load
alone is of no interest since the load transmitted by the bolt
strongly depends on bolt stiffness. Consequently, with regard to
the load displacement behaviour of the bolt presented in Fig. 6, it
was decided to model the bolt with a bilinear behaviour in order
to take into account the effect of bearing damage. Bolt stiffness
up to bearing initiation (F
i
= 11000 N) is noted k
0
and above this va-
lue, the tangential bolt stiffness is ak
0
, where a represents the rel-
ative stiffness reduction. k
0
and a can be estimated through load
versus joint displacement (DL
a
) curves, one of which is illustrated
in Fig. 6. Values obtained for k
0
are quite scattered. The eight tests
performed lead to a mean value of 102 kN/mm and a standard
deviation of 33 kN/mm. These values will be discussed later with
the help of nite element simulation. The values obtained for a
are less scattered: the mean value is 0.151 and the standard devi-
ation is 0.026.
The method proposed above to nd bolt stiffness and bolt
behaviour in general (including bearing damage) can nevertheless
Fig. 5. Illustration of mechanical phenomena involved in the relative displacement of adherents.
0 0.5 1 1.5 2 2.5 3 3.5 4
0
0.5
1
1.5
2
2.5
3
x 10
4
Displacement (mm)
L
o
a
d

(
N
)
zoom
(3)
(1)-(2)
(4)
(5)
Fig. 6. Load versus extensometer measurement for a double lap joint with d/w = 0.21, bolt torque 1.3 Nm and bolt-hole clearance 0.02 mm.
256 C. Bois et al. / Composite Structures 97 (2013) 252260
be made more generic. In fact, as proposed by Gray and McCarthy
[15], the load versus displacement curve obtained on a control case
can be used to build a pressure versus displacement curve by
dividing the load by the projected hole surface dened by the hole
diameter-to-thickness product. A generic pressure versus displace-
ment curve can then be used for other hole diameter or thickness
as long as the laminate sequence remains the same and the hole
diameter-to-thickness ratio does not change too much.
4.2. Net-section failure of composite adherent
The semi-analytical model previously described gives the total
load undergone by the composite net-section with no information
on stress distribution in this section. However, the stress concen-
tration induced by the hole controls the net-section strength.
Hart-Smith [23,24] proposed a method to estimate the stress con-
centration factor in the net-section of a pin-loaded composite hole
at the laminate scale. The elastic stress concentration factor K
te
can
be calculated by:
K
te

d
w

w
d
0:5 1
d
w
_ _
h if
d
w
P0:25 21
where h is an adimensional parameter which depends on the edge
distance e = L/2:
h
w
e
1 if e < w and h 1 if e Pw 22
Eqs. (21) and (22) are based on an analytical formulation with an
elastic behaviour for the composite part. Thus, this formulation
leads to an underestimation of the composite strength because of
stress gradient effect on composite failure [2527]. The actual stress
concentration factor is lower than the elastic stress concentration
factor because of the stress redistribution that occurs around the
hole due to nonlinear behaviour., Using a great number of bolted
tests with different hole diameter-to-width ratios d/w, Hart-Smith
proposed a linearly corrected stress concentration factor K
tc
dened
by:
1 K
tc
C1 K
te
23
where C is an adimensional parameter lower than 1 identied from
experimental results. Obviously, C depends on composite lay-up,
but in the assembly zone the composite lay-up is sufciently bal-
anced in terms of ply angles to avoid shear-out or cleavage failures.
Thus, the value obtained by Hart-Smith for CFRP quasi-isotropic
laminate, C = 0.25, can be used in most application cases.
The maximum load F
u
ns
that leads to net-section failure can then
be calculated by a maximum stress criterion at the laminate scale:
F
u
ns

t
1
wdr
u
K
tc
24
where r
u
is the ultimate stress of the laminate in the considered
direction and t
1
is the composite thickness.
The net-section strength criterion is then expressed as:
c
ns

K
tc
F
ns
r
u
t
1
wd
25
where F
ns
is the actual load applied to the net-section. The ultimate
stress r
u
is determined with a tensile test on a specimen with a
rectangular section. This test gives r
u
= 600 MPa. The net-section
criterion is validated with a bolted double lap joint test (Fig. 1). A
width of 15 mm (d/w = 0.42) is selected in order to induce net-sec-
tion failure. As for bearing degradation analysis, several tests were
performed with three tightening torque values and two clearance
values. Fig. 8 presents load versus extensometer measurement
curves for three different congurations. As expected, the net-sec-
tion failure is preceded by bearing degradation at the same level
as for d/w = 0.21. The slight increase in ultimate load when tighten-
ing increases can be attributed to the friction load transmitted be-
tween the two adherents. The criterion dened by Eq. (25) gives
an ultimate load of 14350 N, which is in very good agreement with
experimental results.
5. Validation of the analytical model using nite element
simulations
Several studies demonstrate that a 3-D FEM is able to predict
the bolt load transfer rate in a hybrid joint [1,3]. Thus, in this part
analytical model and nite element simulations are compared in
terms of bolt load transfer rate, axial force (stress ux) and adhe-
sive shear strain along overlap length. In the FEM, the composite
part is made of a transversely isotropic ply lay-up with elastic
behaviour. The adhesive is modelled by an elastoplastic behaviour
identical to that used in the analytical model. The bolt and the alu-
minium adherents are modelled by an elastic behaviour since yield
stress is never reached. Material and geometrical parameters are
reported in Fig. 1 and Table 1. The 3-D FEM was performed using
0 0.5 1 1.5 2 2.5 3 3.5 4
0
0.5
1
1.5
2
2.5
3
x 10
4
Torque (Nm)
L
o
a
d

(
N
)
Bearing initiation clearance 0.02 mm
Bearing initiation clearance 0.15 mm
Bearing initiation without confinement
Bearing strength clearance 0.02 mm
Bearing strength clearance 0.15 mm
Bearing strength without confinement
Fig. 7. Inuence of bolt tightening torque on bearing initiation load and bearing
maximum load.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
Displacement extensometer (mm)
L
o
a
d

(
N
)
Bolt torque : none, Clearance : 0.02 mm
Bolt torque : none, Clearance 0.15 mm
Bolt torque : 3.5 Nm, Clearance 0.15 mm
Fig. 8. Load versus extensometer measurement for a bolted double lap joint with d/
w = 0.42.
C. Bois et al. / Composite Structures 97 (2013) 252260 257
the ABAQUS/Implicit software. Only half of the geometry was mod-
elled according to the symmetrical plane of the joint. Bolt-hole
clearance was set to 0.02 mm and the hard contact law available
in ABAQUS was used with a friction coefcient of 0.1. Wedge and
cubic quadratic elements (designated respectively C3D15 and
C3D20R in ABAQUS [28]) were used with one element in the adhe-
sive thickness, three elements in the aluminium adherent thick-
ness and height elements in the composite adherent thickness (1
element per ply). From a mesh sensitivity study we concluded that
a mesh that is rened progressively towards the hole up to
0.26 0.26 mm element size represents a good compromise be-
tween accuracy and computation time.
First, a bolted joint is simulated in order to evaluate the numer-
ical bolt stiffness before bearing initiation and compare it to the
experimental one. The numerical bolt stiffness obtained (127 kN/
mm) is within the range of experimental stiffness obtained (79
150 kN/mm) and quite close to the mean value (102 kN/mm). Gi-
ven the reliability of the numerical bolt stiffness, it was used
subsequently.
From Fig. 9, results obtained on both hybrid and bonded joints
for d/w = 0.21 can be compared. Fig. 9a and b shows the shear
strain and the axial force along the overlap length for a load close
to the ultimate experimental load. The load versus extensometer
measurement is plotted in Fig. 9c. In spite of the peeling stress at
the ends of the joint, the analytical model correctly predicts the
shear strain level in the adhesive. The effect of total load on the
bolt load transfer rate is illustrated in Fig. 10. The role of adhesive
plasticity is highlighted since the bolt load transfer rate ranges
from a few percent to 40%. The results from the analytical model
were found to be in very good agreement with those obtained from
FEM. So the analytical model is able to replace a time-consuming
FEM with a view to applying adhesive and composite failure crite-
ria, as proposed above. The role of bearing degradation introduced
into the analytical model is also highlighted in Fig. 10.
6. Validation of adhesive and composite failure criteria using
hybrid joint test
Adhesive and composite failure criteria were proposed and
identied from basic tests on non-hybrid joints, i.e. bonded or
bolted joints. A validation step on hybrid joints was performed
on double lap shear joints for two different congurations: d/
w = 0.21 and d/w = 0.42. Other parameters used are reported in
Fig. 1 and Table 1. Load versus machine displacement curves are
plotted in Fig. 11. Concerning the conguration d/w = 0.21, as pre-
dicted by the model, failure starts in the adhesive joint where the
shear strain is greatest (near the aluminium adherent free edge).
The adhesive crack propagates along about 60% of the overlap
length generating a load fall greater than 50%. From this moment
the bolt supports practically all the load and the bearing degrada-
tion continues.
Concerning the conguration d/w = 0.42, the model predicts
that the net-section failure criterion reaches 1 just before the adhe-
sive failure criterion. However, in some samples a short crack can
be observed in the adhesive joint before the net-section failure.
In terms of failure scenario, results from the analytical model
and the experiments are broadly in good agreement.
A good way to compare model predictions and experiment
strength on the one hand, and relative performances of different
joint congurations on the other, is to plot joint efciency as a
function of d/w. Joint efciency can be dened by:
eff
F
joint
F
adherent
26
where F
joint
is the strength of the studied joint and F
adherent
is the
tensile strength of the weakest adherent alone (width and thickness
matching adherent dimensions in the joint). In the present case, the
weakest adherent is the composite part and its tensile strength is:
F
adherent
r
u
wt
1
27
The three failure criteria c
a
, c
b
and c
ns
are calculated using the
analytical model after each load increment and the strength of
the studied joint F
joint
is obtained when one of the criteria reaches
1. Thus, the model can be used to identify the failure mode that
leads to joint failure.
Fig. 12 illustrates the model predictions of joint efciency for a
bonded joint, bolted joint and hybrid joint. Experimental results for
the 14 tests performed on bolted joints and 12 tests performed on
hybrid joints are also shown in Fig. 12. Results scattering is rela-
tively low with regard to the many parameters that can affect re-
sults. Some of these parameters, like adhesive joint thickness or
bolt-hole clearance, are themselves subject to scattering.
As discussed in a previous section, for a hybrid joint, the bearing
initiation prediction is not sufcient to predict joint strength de-
ned as maximum load. Therefore, in Fig. 12, two joint efciency
curves are plotted for both bolted and hybrid joints. The rst
Table 1
Material and geometrical parameters used in analytical model and FEM.
Double lap joint dimensions Composite
thickness t
1
(mm)
Aluminium thickness
t
2
(mm)
Adhesive
thickness t (mm)
Overlap length L
(mm)
Bolt diameter
d (mm)
4 (16 plies) 5 0.5 60 6.35
Properties of the unidirectional ply for T700/M21 used
in FEM
E
11
(GPa) E
22
= E
33
(GPa) G
12
= G
13
(GPa) G
23
(GPa) m
12
= m
13
m
23
130.3 7.6 4.75 2.65 0.33 0.43
Homogenised properties for T700/M21 quasi-isotropic
lay-up used in analytical model
E
xx
(GPa) Laminate ultimate
stress r
u
(MPa)
Bearing pressure
p
b1
(MPa)
Bearing pressure
p
b2
(MPa)
48 600 433 1000
Properties for Aluminium alloy 2024 used in FEM and
analytical model
E (GPa) m
68 0.3
Properties for Adhesive 3M 2216 used in FEM E (GPa) m Yield stress r
e
(MPa)
0.38 0.32 5
Properties for Adhesive 3M 2216 used in analytical
model
G (GPa) G

(GPa) Shear yield stress


T
p
(MPa)
Ultimate shear
strain c
r
0.144 0.0136 2.89 0.72
Titanium (Hi-Lok Bolt) properties used in FEM E (GPa) m Yield stress r
e
(MPa)
106 0.3 830
Bolt joint properties used in analytical model Stiffness k
0
(kN/
mm)
Relative stiffness
reduction a
127 0.151
258 C. Bois et al. / Composite Structures 97 (2013) 252260
corresponds to bearing initiation and the second to strength (max-
imum load). With the model, joint strength can be predicted both
precisely and conservatively, bearing in mind that all model
parameters were identied from elementary tests. The underesti-
mation of joint strength for d/w = 0.42 can be attributed to the
overestimation of the stress concentration factor in the net-section
criterion for a hybrid joint. In fact, FEM demonstrates that the
stress concentration factor is reduced when part of the load under-
gone by the net-section is related to bonding instead of to the bolt.
From a design point of view, it is worth noting that hybrid joints
can lead to a signicant increase in mechanical performance since
optimal efciency is about 0.45 for non-hybrid joints whereas it is
0.66 for hybrid joints. Moreover, hybrid conguration extends the
strength domain for a given efciency level, thus increasing design
margins.
0 0.01 0.02 0.03 0.04 0.05 0.06
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
S
h
e
a
r

s
t
r
a
i
n


x
z
Position along x axis (m)
bonded joint analytical (load 32 kN)
bonded joint FEM (load 33.1 kN)
hybrid joint analytical (load 50 kN)
hybrid joint FEM (load 49.3 kN)
0 0.01 0.02 0.03 0.04 0.05 0.06
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
x 10
4
Position along x axis (m)
L
o
a
d

i
n

c
o
m
p
o
s
i
t
e

p
a
r
t

a
l
o
n
g

x

a
x
i
s

(
N
)
bonded joint analytical (load 32 kN)
bonded joint FEM (load 33.1 kN)
hybrid joint analytical (load 50 kN)
hybrid joint FEM (load 49.3 kN)
0 0.05 0.1 0.15 0.2 0.25 0.3
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
x 10
4
Displacement (mm)
L
o
a
d

(
N
)
bonded joint analytical
bonded joint FEM
hybrid joint analytical
hybrid joint FEM
(a)
(b)
(c)
Fig. 9. Comparison between FEM and analytical model for a double lap hybrid and
bonded joint with d/w = 0.21. (a) Shear strain along the overlap length, (b) axial
force in composite adherent along the overlap length, (c) load versus virtual
extensometer measurement curve.
0 1 2 3 4 5
x 10
4
0
5
10
15
20
25
30
35
40
45
Load (N)
B
o
l
t

l
o
a
d

t
r
a
n
s
f
e
r

r
a
t
e

(
%
)
analytical elastic
analytical adhesive elastoplastic
analytical adhesive elastoplastic
and bearing degradation
FEM adhesive elastoplastic
Fig. 10. Bolt load transfer rate versus total load on a double lap hybrid joint with
d/w = 0.21.
0 1 2 3 4 5 6 7
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
x 10
4
Tensile machine displacement (mm)
L
o
a
d

(
N
)
d/w = 0.42
d/w = 0.21
Fig. 11. Load versus machine displacement curves of hybrid joints, bolt torque
1.3 Nm.
C. Bois et al. / Composite Structures 97 (2013) 252260 259
7. Conclusions
The aim of this study is to propose an analytical model associ-
ated with a suitable identication method dedicated to hybrid
(bolted/bonded) joints. The model includes the nonlinear behav-
iour required to evaluate precisely the bolt load transfer contribu-
tion up to failure. Each expected failure mode is taken into account
using a multi-criteria approach which allows each related criterion
to be identied from an elementary (non-hybrid) test.
The ability of the analytical model to predict load transfer in
bolt and adhesive joints is established by comparing analytical
model results to those obtained by nite element analysis. The
whole model is validated experimentally in terms of joint ef-
ciency prediction for two different hole diameter-to-width ratios.
The accuracy of the model relies essentially on bolt stiffness
evaluation. Because of the various and complex mechanical phe-
nomena involved, bolt stiffness identication remains a difcult is-
sue. A combined experimental and numerical approach seems to
be the best way to overcome this difculty.
In design, the model can be used to quickly evaluate the inu-
ence of the numerous geometrical or material parameters involved
in a hybrid joint. The model could be extended to 2D problems by
using semi-analytical or FE formulation to deal with bi-axial load-
ing and multi-rows and multi-lines of bolted joints.
References
[1] Kelly G. Load transfer in hybrid (bonded/bolted) composite single-lap joints.
Compos Struct 2005;69:3543.
[2] Hoang-Ngoc CT, Paroissien E. Simulation of single-lap bonded and hybrid
(bolted/bonded) joints with exible adhesive. Int J Adhes Adhes
2010;30:11729.
[3] Lee Y-H, Lim D-W, Choi J-H, Kweon J-H, Yoon M-K. Failure load evaluation and
prediction of hybrid composite double lap joints. Compos Struct
2010;92:291626.
[4] Barut A, Madenci E. Analysis of boltedbonded composite single-lap joints
under combined in-plane and transverse loading. Compos Struct
2009;88:57994.
[5] Paroissien E. Contribution aux assemblages hybrides (boulonns/colls)
application aux jonctions aronautiques. PhD thesis, Toulouse III University,
France; November 2006. <http://thesesups.ups-tlse.fr/3/1/Paroissien_Eric.pdf>
[access July 2012].
[6] Paroissien E, Sartor M, Huet J. Joints applied to aeronautic parts: analytical
one-dimensional models of a single-lap joint. In: Trends and recent advances
in integrated design and manufacturing in mechanical engineering II. Springer;
2007. p. 95110.
[7] Pires I, Quintino L, Miranda RM. Numerical simulation of mono- and bi-
adhesive aluminium lap joints. J Adhes Sci Technol 2006;20:1936.
[8] Cognard JY, Crachcadec R, Sohier L, Davies P. Analysis of the nonlinear
behavior of adhesives in bonded assemblies-comparison of TAST and Arcan
tests. Int J Adhes Adhes 2008;28:393404.
[9] Dean G, Crocker L, Read B, Wright L. Prediction of deformation and failure of
rubber-toughened adhesive joints. Int J Adhes Adhes 2004;24:295306.
[10] Ban C-S, Lee Y-H, Choi J-H, Kweon J-H. Strength prediction of adhesive joints
using the modied damage zone theory. Compos Struct 2008;86:96100.
[11] Nelson WD, Bunin BL, Hart-Smith LJ. Critical joints in large composite aircraft
structure. NASA CR-3710; 1983.
[12] Huth H. Inuence of fastener exibility on the prediction of load transfer and
fatigue life for multiple-row joints. In: ASTM special technical
publication. Charleston, SC, USA: ASTM; 1986. p. 22150.
[13] McCarthy CT, Gray PJ. An analytical model for the prediction of load
distribution in highly torqued multi-bolt composite joints. Compos Struct
2011;93:28798.
[14] McCarthy MA, Lawlor VP, Stanley WF, McCarthy CT. Bolt-hole clearance effects
and strength criteria in single-bolt, single-lap, composite bolted joints.
Compos Sci Technol 2002;62:141531.
[15] Gray PJ, McCarthy CT. A highly efcient user-dened nite element for load
distribution analysis of large-scale bolted composite structures. Compos Sci
Technol 2011;71:151727.
[16] Irisarri F-X, Laurin F, Carrere N, Maire J-F. Progressive damage and failure of
mechanically fastened joints in CFRP laminates Part I: Rened Finite Element
modelling of single-fastener joints. Compos Struct 2012;94:226977.
[17] Hhne C, Zerbst A-K, Kuhlmann G, Steenbock C, Rolfes R. Progressive damage
analysis of composite bolted joints with liquid shim layers using constant and
continuous degradation models. Compos Struct 2010;92:189200.
[18] Frizzell RM, McCarthy CT, McCarthy MA. Simulating damage and delamination
in bre metal laminate joints using a three-dimensional damage model with
cohesive elements and damage regularisation. Compos Sci Technol
2011;71:122535.
[19] Xiao Y, Ishikawa T. Bearing strength and failure behavior of bolted composite
joints (Part I: Experimental investigation). Compos Sci Technol
2005;65:102231.
[20] Wu PS, Sun CT. Modeling bearing failure initiation in pin-contact of composite
laminates. Mech Mater 1998;29:32535.
[21] Gohorianu G. Interaction entre les dfauts dusinage et la tenue en matage
dassemblages boulonns en carbone/poxy. PhD thesis, Toulouse III
University, France; April 2008. <http://thesesups.ups-tlse.fr/248/1/Gohorianu_
Gina.pdf> [access July 2012].
[22] Kelly G, Hallstrm S. Bearing strength of carbon bre/epoxy laminates: effects
of bolt-hole clearance. Compos Part B: Eng 2004;35:33143.
[23] Hart-Smith LJ. Bolted joint analysis for composite structures. In: Joining and
repair of composites structures. Kansas City, MO; 2004.
[24] Hart-Smith LJ. Bolted joint analyses for composite structures current
empirical methods and future scientic prospects. In: ASTM Special
Technical Publication. Kansas City, MO; 2004. p. 12760.
[25] Whitney JM, Nuismer RJ. Stress fracture criteria for laminated composites
containing stress concentrations. J Compos Mater 1974;8:25365.
[26] Hochard C, Lahellec N, Bordreuil C. A ply scale non-local bre rupture criterion
for CFRP woven ply laminated structures. Compos Struct 2007;80:3216.
[27] Kweon J-H, Ahn H-S, Choi J-H. A new method to determine the characteristic
lengths of composite joints without testing. Compos Struct 2004;66:30515.
[28] ABAQUS. Analysis user manual. Version 6.9. Dassault Systmes; 2010.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
d/w
E
f
f
i
c
i
e
n
c
y
bolted joint tests: bearing initiation
bolted joint tests: strength
hybrid joint tests: strength
bolted joint model: bearing initiation
bolted joint model: bearing strength
bonded joint model
hybrid joint model: bearing initiation
hybrid joint model: strength
Fig. 12. Experimental results and joint efciency versus d/w predicted by analytical
model for bonded joint, bolted joint and hybrid joint.
260 C. Bois et al. / Composite Structures 97 (2013) 252260

You might also like