Vous êtes sur la page 1sur 9

Composite Stmchms 32 (1995) 123-13 1

0 1995 Elsevier Science Limited


Printed in Great Britain. All rights reserved
0263~8223/95/$9.50
0263-8223(95)00059-3

The effect of a spew fillet on adhesive stress


distributions in laminated composite single-lap
joints

M. Y. Tsai
Department of Engineering Science and Mechanics, T/irginiaPolytechnic Institute and State University,
Blacksbutg, VA 24061-0219, USA

&
J. Morton
Structural Materials CenteG DRA, Famborough, Hampshire GU14 6TD, UK

A laminated composite single-lap joint without a spew fillet, subjected to


tensile loading, is investigated experimentally and numerically. By directly
comparing the experimentally- and numerically-determined deformations of
the single-lap joints with and without a fillet, the effect of a spew fillet on
adhesive stress distributions is discussed. Moire interferometry is used to
measure the in-plane surface deformation of the overlap region of the test
specimens. The deformation interactions of the laminated adherends,
adhesive layer and a fillet are documented in the form of orthogonal
components of the displacement fields (u and v). Two-dimensional,
geometrically linear and nonlinear finite element analyses are performed to
simulate the mechanical response of the laminated composite single-lap
joint and the effect of a spew fillet. Experimental and numerical results
indicate that the adhesive shear and peel strain (stress) concentrations can
be reduced greatly by introducing a fillet at the end of the overlan, and
these concentrations are affected by the geometrically no&near
deformation of the single-lap joint.

INTRODUCTION theoretically, numerically and experimentally


for more than half a century,3-7 single-lap joints
Adhesive bonded joints offer advantages over with laminated composite adherends have also
mechanically fastened and riveted joints in time received some attentions-10 due to the increas-
and cost savings, higher strength to weight ing use of laminated composite structures. By
ratios, corrosion and fatigue resistance, crack comparing and evaluating several theoretical
retardance, damping characteristics and so and numerical models, Oplingerl’ pointed out
‘,’ Adhesive bonding of composite structures that factors such as transverse shear deforma-
:zs additional merits in avoiding the drilled tion, transverse normal strain, material
holes (and broken fibres) and reducing stress nonlinear behaviour of adherends and adhe-
concentrations. However, factors such as the sives, viscoelastic behaviour, thermal effect, and
inherent material heterogeneity, residual stres- fracture mechanism should be incorporated in
ses, free-edge effects, and relatively low modeling any configurations of composite
transverse strength and shear stiffness, impose bonded joints. In addition, general reviews
greater complexity to the case of adhesive related to the composite joints have been pro-
bonded composite structures as opposed to vided by several researchers.1z-14
homogenous isotropic structures. Despite the research efforts on composite
Although the problem of the single-lap joint joints, the mechanics of laminated composite
with isotropic adherends has been investigated adhesive joints have not been comprehensively
123
124 M. Y Tsai, J. Morton

treated. The specific problem of a laminated the adherend is 25.4 mm. The material for the
composite single-lap joint with a 45” spew fillet adherend is graphite/epoxy (XAW914C). The
has been investigated and the results have been lay-up is [O/45/-45/0],, with fiber orientation
published.” The objective of this study is to defined in Fig. 2. For the purpose of subsequent
investigate a composite single-lap joint with and numerical analysis, the material properties in
without a spew fillet experimentally and numer- each unidirectional lamina are taken as longitu-
ically and thus understand the role of a spew dinal elastic modulus El = 138 GPa, transverse
fillet in the laminated composite single-lap joint. E2=9.4 GPa, in-plane shear G12=6.7 GPa and
Poisson’s ratio v12=0.32. Note that the sub-
scripts 1 and 2 represent the fiber and matrix
EXPERIMENTAL PROGRAM directions in a single lamina. The two adher-
ends are bonded together with a thin adhesive
A two-dimensional schematic of the single-lap layer (~=0.13 mm). The material for the adhe-
joint is shown in Fig. 1. For the geometric para- sive is epoxy resin (Redux 308A) with elastic
meters, 1 is the length of the outer adherend, 2c modulus E,=3 GPa and Poisson’s ratio
the length of the overlap, t the thickness of the v,=O.31. Two end tabs were used for loading
adherend, q the thickness of the adhesive and applied and easy alignment. Nonlinear behavior
a, a force-eccentricity angle. E, and v, respec- for this adhesive appears in deformations
tively represent the adhesive elastic modulus beyond about 3% shear strain and about 1.5%
and Poisson’s ratio. The resultant force per unit normal strain. Shear plastic deformation occurs
width, T’, is approximately equal to the longitu- after 10% shear strain. A moire grating with a
dinally applied force per unit width, T, when frequency of 1200 lines/mm was replicated on
a, < 1 (usually LX,~0.1). An applied force, F, is the edge surface of the overlap, shown in Fig. 2.
defined as T x b, in which b is the width of the Moire inteferometry is an optical method for
adherend. p is an applied stress which is deno- measuring in-plane surface deformation.15 The
ted as TJt. moire fringe patterns u and v represent the
A laminated composite single-lap specimen horizontal and vertical displacement contours,
without a fillet is shown in Fig. 2 where Z=101.6 respectively. The governing equations for strain
mm, 2c=254 mm, t=2 mm and the width of determination from the displacement fields are,

d-
C d-
C I
I‘
‘1 I 1

9 ! t
t I i

t
t #SF
an
c ,vc
Fig. 1. Geometry and related parameters of a single-lap joint.

1 !: r, I_
25.4 l-101.6 2c-25.4 L101.6 25.4

\ Moire surface grating


Materials:
Adhenxdsz Graphite/f$oxy, (0/45/-451O)zs

Ahsive: Epoxy (Redux 308A)


Fig. 2. Geometry and material of the composite test specimen, and moirC grating replication.
Effect of spew fillets on stress dhributions 125

au 1 aN, grating shown in Fig. 2. Typical fringe patterns


f%=-=- - are shown in Fig. 3 which contains the
ax f ax
horizontal (u) and vertical (v) deformations of
half an overlap for the specimen without a fillet,
av 1 alv, under an applied load, F=3202 N (720 lb).
,y=-=- - (1) In order to provide a direct comparison of
aY f aY
moire fringes, the u-field fringe patterns in
regions, distance 3t away from the end of the
au av aiv, arv,
‘“=y +-=-
ax
1
f
(F+T) overlap (in Fig. 4) and around the end of the
overlap (in Fig. 5) are shown for the specimens
with and without a 45” spew fillet, under the
where f=2400 lines/mm in the system used and same load (F~4448 N or 1000 lb). It is appar-
N, and NY denote the fringe orders in the u and ent in Fig. 4 that the u-field fringe patterns
v fields. The sensitivity of measurement is 0.417 away from the fillet-affected region for both
pm per fringe order. A portable achromatic specimens are almost the same. Both longitudi-
interferometer-l6 was employed in the moire nal normal strain distributions across the
experiment. This system features vibration thickness of the overlap, extracted from these
insentivity which allows the specimen to be tes- moire fringes, are also in a very good agree-
ted in a conventional testing machine. This ment. Good agreement in the deformation of
system also provides simultaneous measurement the far fields for two different specimens (with
of u and v displacements so that the shear and without a fillet) provide confidence for fur-
deformation of the loaded body can be ther comparing deformations in the near fields
resolved. around the end of the overlap for both speci-
The moire technique was used to obtain the mens, which are shown in Fig. 5. The resulting
surface deformations of the adherends and adhesive strain distributions for both specimens
adhesive in the region, located on the edge sur- are shown in Fig. 6.
face of the overlap, covered with the moire

(a) u-field

(b) v-field

Fig. 3. Typical moire fringe patterns: (a) u displacement field (b) v displacement field, for a laminated composite single-
lap joint, without a fillet, under an applied load, F=3202 N (720 lb).
126 M. Y Tsai, J. Morton

u-field with a fillet u-field without a fillet 0 0.05 0.1 0.15 0.2

qw

Fig. 4. Comparisons of u-field moirC fringe patterns and longitudinal strain, cx, distributions at the region, 3t distance
away from the end of the overlap, for the specimens with and without a spew fillet under a similar load (Fz4448 N, or
1000 lb).

FINITE ELEMENT ANALYSIS erated by the presence of +45” layers.


Constant-strain (stress) elements were used for
Geometrically linear and nonlinear two-dimen- the adherends and adhesive. There were two
sional finite element analyses using the elements across the thickness of the adhesive
ABAQUS code were performed to model the layer, except for the spew fillet, and one ele-
deformation of the laminated composite single- ment for each lamina of the laminated
lap joint. Two entire single-lap joint models composite adherends. The material of each
(joints with and without fillet), shown in Fig. 7, lamina is modelled as orthotropic, while the
were analyzed. The geometric boundary condi- material of the adhesive as isotropic. Due to the
tions include a hinge on the center line of the use of two the constant-strain (stress) elements
adherend at one end and a roller at the other. across the thickness of the adhesive, the strain
A force is applied horizontally at the roller end. (stress) values calculated siong the center line
For these end conditions, the resultant force of the adhesive represent approximately the
generated always passes through the center of average values over the thickness. To increase
the overlap, so that antisymmetrical loading and the accuracy of the numerical calculation, the
deformed conditions are achieved. The plane finer meshes were adopted near and at the end
strain condition assumed in these analyses fails of the overlap. The geometry, dimensions and
to capture/represent the effect of in-plane shear material properties of the adherend and adhe-
and bending-twisting coupling which are gen- sive in the numerical analyses are the data
Effect of spew fillets on stress distributions 127

(b)
Fig. 5. Moirt fringe patterns in the region near the end of the overlap: (a) u displacement field, for a laminated
corn posite single-lap joint, without a fillet, under an applied load, F=4337 N (975 lb) (b) u displacement field,” for a
laminated composite single-lap joint, witha 45” spew fillet, under an applied load, F=4448 N(1000 lb).
128 M. Y Tsai, J. Morton

described in the experimental program, and it is


assumed that Es=Ez, Gr3=Gi2 and ~13=~12.

RESULTS AND DISCUSSION


0.5

The global bending deformation of the test


specimen is verified by comparing the results
from experimental measurement and finite ele-
ment analyses. The longitudinal strain response
at points 1 and 2, where back-to-back adherend
surfaces are at positions about 25 t away from
the end of the overlap, are shown in Fig. 8 for
-2.5
various applied loads per unit width (T=F/b)
-3 from the finite element analyses (linear and
I nonlinear) and strain gauge measurement. It is
-3.5
, 1.2 1.4 apparent .that the experimental results are in
good agreement with those from the nonlinear
Fig. 6. Comparisons of adhesive strain cnstrmuttons
determined from moire experiments for the laminated
finite element analysis, rather than from the
composite single-lap joint specimens with and without a linear analysis. Thus the specimen deformed
spew fillet under a similar load (FE 4448 N, or 1000 lb). geometrically nonlinearly and the bending

Ec ,vc

Fig. 7. Finite element model; geometry, boundary conditions and detailed meshes used the composite single-lap joints
with and without a spew fillet.
Effect of spew fillets on stress distributions 129

with a fillet near the end of the overlap carries


the longitudinal normal strains (.sx) unlike the
joint without a fillet. This existence of a, indi-
cates that the fillet does help transfer a part of
0.61 I I
the longitudinal load from the lower adherend
to the upper through shear.
The adhesive strain distributions obtained
from the moire experiment are shown in Fig. 6
for the composite joints with and without a
spew fillet under a load, Fr4448 N. These
strain values are calculated by averaging the
value over the thickness of the adhesive. The E,
components are not shown, because of the diffi-
culty of extracting zXdata from the joint without
a fillet, due to high density of the fringes near
and at the end of the overlap. For the peel
strain (a,,) distributions, the location of tensile
T, load (N/mm) strain on the free surface moves from the end
of the overlap to the end of the spew fillet as a
Fig. 8. Comparisons of longitudinal normal strain
responses at the points near the end of the overlap from
result of the presence of the spew fillet. Note
experimental measurement and finite element analyses that the adhesive peel strain is very sensitive to
(linear and nonlinear models). three-dimensional effects.17 E,, measured from
the free surface does not represent that in the
interior, since E,, is more compressive than that
strains near the end of the overlap predicted in the interior. For the shear strain (yV) dis-
from the linear analysis are larger than that tributions, the maximum shear strain occurs at
from the nonlinear analysis. the end of the overlap for both cases. However,
A detailed comparison of moire u-displace- the value of the maximum shear strain for the
ment field near the end of the overlap for the joint with a fillet is only about 40% of that for
joints with and without a spew fillet is shown in the joint without a fillet. And the 45” spew fillet
Fig. 5. Both joints have the same lay-up and are carries a certain degree of shear. Apparently
under similar loads. The numbers marked on the spew fillet is able to reduce significantly the
the fringe patterns are the u-field fringe order, adhesive shear strain concentration at the end
Nx. These fringe patterns are obtained by opti- of the overlap.
cally reducing the number of v-field fringes as Since the surface deformation recorded from
much as possible, at the center point of the the moire experiment does not fully represent
overlap by introducing a carrier pattern of rota- that in the interior (especially for the peel strain
tion. Thus, the term of au/ay in the calculation component), the two-dimensional plane strain
of shear strain in eqn (1) becomes dominant. finite element analysis, however, does provide
That is, the u displacement gradient in the y the deformation in the interior. Note that the
direction ((aNx/ay)lf) approximately represents adhesive shear strain (yxy) distributions for the
the shear strain component. It is observed that joint without a fillet are insensitive to three-
the aNxli3y for the joint without a spew fillet in dimensional effects,17 while those with a fillet
the adhesive near and at the end of the overlap are not near and at the spew fillet.18 Accord-
is larger than that for the joint with a spew ingly, a comparison of the adhesive shear
fillet. The 45” spew fillet carries a portion of distributions from numerical and experimental
shear strains which are used to transfer the results can be used to confirm the validity of
longitudinal load from the lower adherend to both results. The adhesive shear strain distribu-
the upper. This argument can be confirmed by tions determined from the moire experiment
observing that the fringes for the joint without a and two-dimensional finite element analysis are
fillet, at the upper adherend near the end of the shown in Fig. 9 for joints with and without a
overlap are almost horizontal and parallel, fillet, under Fr4448 N. A very good agreement
while the fringes for the joint with a fillet are for the adhesive shear distributions is found in
inclined. Thus the upper adherend for the joint the joint without a spew fillet, while a partially
130 M. Y Tsai, J. Morton

w wl fillet
I
- 4'
l w/o fillet
MOh
t
I
-
A
-6l . ’ ’ ’ ’ ’
0.6
’ ’
0.8
’ ’
1
’ ’
1.2
0 0.2 0.4 0.6 0.6 1 1.2 1.4 0 0.2 0.4

x/c
XIC

Fig. 9. Comparisons of the adhesive shear strain dis-


tributions, obtained from the moire experiment and the 0.6
nonlinear finite element analysis for the composite joints
with and without a spew fillet, under F=4448 N (1000 lb).
0.4

0.2
_I
disagreement near the end of overlap and at the
a -.__.-. without fillet
. w&h f&t

spew fillet is shown in the joint with a fillet.


This disagreement between on the surface and
in the interior is expected, according to the
three-dimensional analysis.”
-0.4
After validating the geometrically nonlinear
finite element model which represents the stress
-0.6’ ’ ’ ’ ’ ’ ’ ’ ’ ’ ’ ’ ’
(strain) state in the interior, the resulting adhe-
0 0.2 0.4 0.6 0.8 1 1.2
sive strain and stress distributions are shown in
x/c
Fig. 10(a) and (b), respectively, for composite
Fig. 10. (a) Adhesive strain distributions and (b) adhe-
joints with and without a spew fillet, under sive stress distributions, determined from the nonlinear
F=7619 N. In Fig. 10(a) the position of max- finite element analyses for the composite single-lap joints
imum aY moves from the end of the overlap to with and without a spew fillet, under F=7619 N (1713 lb)
or p= 150 MPa (21.7 ksi).
the end of the spew fillet, as observed in moire
experiment, and the value decreases signifi-
cantly due to the presence of the spew fillet.
The spew fillet carries appreciable amounts of stress concentrations vary with the change of an
shear strains. Thus, the spew fillet helps reduce applied load, and are affected by introducing a
the maximum shear strain in the end of the fillet. The maximum adhesive stresses normal-
overlap. E,, especially for the joint with a fillet, ized by the applied stress, p, are plotted against
is not small enough to be neglected in terms of the applied stresses for composite joints with
its contribution to initial failure. For the adhe- and without a 45” spew fillet in Fig. 11, which is
sive stress distributions in Fig. 10(b), the determined from nonlinear finite element analy-
maximum values of oY and zV which are domi- ses. The values of o,, and “.V are plotted at the
nant in failure initiation in the joint without a end of the overlap for the joints with and with-
fillet decrease greatly in the presence of the out a fillet. Note that the sign of z_ is ignored.
spew fillet. The maximum 0x located near and The plotted values of a, are located 1.5 adhe-
inside the end of the overlap for the joint with- sive thicknesses inside the end of the overlap
out a fillet moves to the position outside of the for the joint without a fillet, and outside the end
overlap with a fillet. This maximum o, in the of the overlap for the joint with a fillet. It is
joint with a fillet is comparable to other stresses apparent that the nonlinear deformation greatly
(cry and z~), so that it cannot be excluded in a affects the adhesive stress concentrations for
failure analysis. both joints with and without a fillet. The stress
Due to the nature of geometrically nonlinear concentrations from the linear analysis,
deformation in the single-lap joint, the adhesive (approximately equivalent to those at p=O)
EfSect of spew fillets on stress distributions 131

the financial support under the contract


USDOT/FAA 93-G-064. The authors also thank
Mr F. L. Matthews of the Center for Composite
Materials, Imperial College for providing the
test specimens.

L
. I. ----- Wtthout Fillet
REFERENCES
0.8 - ‘*-. - Wlth Fillet 0.6
_-__ 40?6IeilUrnkmd
-.._ fclrjirm withmkt 1. National Material Advisory Board National Research
*.__ .---_._
0.5 - -_ -.-___ 0.5 Council. Structural adhesives: with emphasis on aero-
-.._ ----__ 1;
----ma__ -------_-____________ space applications. Marcel Dekker Inc. New York,
-A----______
0.4 - --------- _______________~ 0.4 1976.
z 2. Kuno, J. K., Structural adhesives continue to gain
v 0.3

OY foothold in aerospace and industrial use. In Structural
-6- __.__.___________.______._.______.
Adhesives and Bonding, Proc. Structural Adhesives
0.2 - 0.2
Bonding Conf., Technology Conference Associates, El
Segundo, California, 1979.

:Io: 50 100 150 200 250


O.l
3000
3. Volkersen,
spruchten
O., Die nietkraftverteilung
nietverbindungen
laschenquerschnitten,
41-7.
mit
in zugbean-
konstanten
Luftfahrtforschung, 15 (1938)
P (MPa)
4. Goland, M. & Reissner, E., The stresses in cemented
Fig. 11. Normalized maximum adhesive stresses vs joints. J. Appl. Mech., 11 (1944) A17-A27.
applied stresses, p, for the composite single-lap joints with 5. Hart-Smith, L. J., Adhesive-bonded single-lap joints.
and without a spew fillet. NASA. CR-112236, 1973.
6. Oplinger, D. W., Effects of adherend deflections in
single-lap joints. ht. .I. Solids & Struct., 31 (1994)
would be overestimated. The values of c,, and 2565-87.
zV decrease significantly due to the effect of the 7. Tsai, M. Y. & Morton, J., An evaluation of analytical
and numerical solutions to single-lap joint. Znt. J.
spew fillet, while gX has a mild drop. After the Solids & Struct., 31 (1994) 2537-63.
large decrease of the oY and zqxy,the crXand rxy 8. Renton, W. J. & Vinson, J. R., Analysis of adhesively
would become the dominant components for bounded joints between panels of composite materi-
the joint with a fillet, rather than aY. als. J. AppZ. Mech., (1977) 101-6.
9. Mignery, L. A. & Schapery, R. A., Viscoelastic and
nonlinear adherend effects in bonded composite
joints. J. Adhesion, 34 (1991) 17-40.
CONCLUSIONS 10. Tsai, M. Y., Morton, J. & Matthews, F. L. Experi-
mental and numerical studies of a laminated
composite single-lap adhesive joint. J. Comp. Mat., (in
From the experimental and numerical studies, it press).
is concluded that, in the laminated composite 11. Oplinger, D. W., Stress analysis of composite joints.
single-lap joint, the adhesive shear and peel In Proc. Fourth Army Materials Technology Conference
- Advances in Joining Technology, 1975. -
strain (or stress) concentrations, generally
12. Matthews, F. L., Kilty, P. F. & Godwin, E. W., A
occurring at the end of the overlap, are reduced review of the strength of ioints in fiber-reinforced
significantly by introducing a fillet. This reduc- plastics: part 2: adhesively bonded joints. Comp.,
tion of the stress concentration is attributed to (1982) 29-37.
13. Vinson, J. R., Adhesive bonding of polymer compos-
the ability of the spew fillet to carry some of the ites. Polymer Eng. & Sci., 29 (1989) 1325-31.
shear stresses and thus plays a part in transfer- 14. Adams, R. D., Strength predictions for lap joints
ring an element of longitudinal load from one especially with composite adherends: a review. J.
Adhesion, 30 (1989) 219-42.
adherend to the other. The geometrically non- 15. Czarnek, R., Moire interferometry. Structural Testing,
linear deformation greatly affects the adhesive Society of Experimental Mechanics, 30 (1990)
stress (strain) concentrations which, unlike the 195-200.
16. Czarnek, R., High sensitivity moire interferometry
linear deformation, varies with the change of
with compact achromatic interferometer. Optics and
the applied load. Lasers in Engng, 13 (1990) 99-115.
17. Tsai, M. Y. & Morton, J., Three-dimensional defor-
mations in a single-lap joint. J. Strain AnaZysis, 29
ACKNOWLEDGEMENT (1994) 137-45.
18. Tsai, M. Y. & Morton, J., Mechanics of the single-lap
adhesive joint with laminated composite adherends.
The authors would like to express their grati- 17th Annual Meeting of Adhesion Society, Orlando,
tude to the Federal Aviation Administration for Florida, Feb. 1994.

Vous aimerez peut-être aussi