Adhesive-Bonded Composite Joint Analysis with Delaminated

Surface Ply Using Strain-Energy Release Rate

Alireza Chadegani

and Chihdar Yang

Wichita State University, Wichita, Kansas 67260

and
Stanley S. Smeltzer III

NASA Langley Research Center, Hampton, Virginia 23681

DOI: 10.2514/1.C031516
This paper presents an analytical model to determine the strain energy release rate due to an interlaminar crack of
the surface ply in adhesively bonded composite joints subjected to axial tension. Single-lap shear-joint standard test-
specimen geometry with thick bondline is followed for model development. The eld equations are formulated by
using the rst-order shear-deformation theory in laminated plates together with kinematics relations and force
equilibriumconditions. The stress distributions for the adherends andadhesive are determinedafter the appropriate
boundary and loading conditions are applied and the equations for the eld displacements are solved. The system of
second-order differential equations is solved to using the symbolic computation tool Maple 9.52 to provide
displacements elds. The equivalent forces at the tip of the prescribed interlaminar crack are obtained based on
interlaminar stress distributions. The strain energy release rate of the crack is then determined by using the crack
closure method. Finite element analyses using the J integral as well as the crack closure method are performed to
verify the developed analytical model. It has been shown that the results using the analytical method correlate well
with the results fromthe nite element analyses. An attempt is made to predict the failure loads of the joints based on
limited test data from the literature. The effectiveness of the inclusion of bondline thickness is justied when
compared with the results obtained from the previous model in which a thin bondline and uniform adhesive stresses
through the bondline thickness are assumed.
Nomenclature
A
11
, A
U
11
, A
L
11
= in-plane modulus per unit width, N=m
A
55
, A
U
55
, A
L
55
= transverse modulus per unit width, N=m
a = prescribed crack length, m
B
11
, B
U
11
, B
L
11
= coupling modulus per unit width, N
b = virtual crack extension length, m
D
11
, D
U
11
, D
L
11
= exural modulus per unit width, N:m
E
a
= adhesive Youngs modulus, Pa
F
x
, F
y
= reaction force per unit width, N=m
G
a
= adhesive shear modulus, Pa
G
T
= total strain energy release rate, J=m
2
G
Tc
= critical total strain energy release rate, J=m
2
G
xz
= adherend x-z plane shear modulus, Pa
G
I
= mode I strain energy release rate, J=m
2
G
Ic
= critical mode I strain energy release rate, J=m
2
G
II
= mode II strain energy release rate, J=m
2
G
IIc
= critical mode II strain energy release rate, J=m
2
h
2
= adherend ply thickness, m
h
U
, h
L
, h = adherend thickness, m
J = J integral value, J=m
2
k
s
, k
U
s
, k
L
s
= shear correction factor
l
n
= notch length of the specimen, m
l
o
= overlap length before crack is initiated, m
M
C
= equivalent crack-tip moment, N
M
y
, M
U
y
, M
L
y
= bending moment per unit width, N
N
C
= equivalent crack-tip force, N=m
N
x
, N
U
x
, N
L
x
= normal stress resultants per unit width, N=m
P = applied tensile force per unit width, N=m
Q
C
= equivalent crack-tip transverse force, N=m
Q
z
, Q
U
z
, Q
L
z
= transverse shear stress resultant per unit width,
N=m
q
a
= adhesive peel stress, Pa
q
i
= interlaminar peel stress, Pa
T
i
= (i =1; 2; 3), traction components, N=m
2
U = strain energy per unit width, J=m
U
+
= strain energy density, J=m
3
u, u
U
, u
L
= x directional displacement, m
u
a
= adhesive x-directional displacement, m
u
i
= (i =1; 2; 3), displacement components, m
u
o
, u
oU
, u
oL
= midplane x directional displacement, m
u
o
8
, w
o
8
= displacement components at midplane of
adherend portion 8, m
u
+
= bottom-surface x-directional displacement, m
u
+
8
, w
+
8
= displacement components at the bottom
surface of adherend portion 8, m

a
= adhesive Poissons ratio
W = work required to close virtual crack per unit
width, N
w, w
U
, w
L
= z-directional displacement, m
w
a
= adhesive z-directional displacement, m
w
+
= bottom-surface z-directional displacement, m
"
a
,
a
= adhesive strains
"
x
, "
z
,
xz
= adherend strains
= adhesive thickness, m

x
= adhesive normal stress, Pa

a
= adhesive shear stress, Pa

i
= interlaminar shear stress, Pa
,
U
,
L
= bending slope, rad
I. Introduction
A
DVANCED composite materials and adhesive-bonding
technology have been widely applied in aerospace/aircraft
Received 23 May 2011; accepted for publication 8 August 2011.
Copyright 2011 by the American Institute of Aeronautics and Astronautics,
Inc. All rights reserved Copies of this paper may be made for personal or
internal use, on condition that the copier pay the $10.00 per-copy fee to the
Copyright Clearance Center, Inc., 222 RosewoodDrive, Danvers, MA01923;
include the code 0021-8669/12 and $10.00 in correspondence with the CCC.

Graduate Research Assistant, Department of Aerospace Engineering;

currently Ph.D. Candidate and Research Assistant, Department of
Engineering Science and Mechanics, Virginia Polytechnic Institute and
State University; chadegan@vt.edu.

Professor and Airbus Fellow, Department of Aerospace Engineering;

charles.yang@wichita.edu.

Deputy Manager, Ares Project Ofce; stanley.s.smeltzer@nasa.gov.

JOURNAL OF AIRCRAFT
Vol. 49, No. 2, MarchApril 2012
503
structures thanks to their high strength-to-weight ratios and excellent
resistance to corrosion. In many applications, bolted joints have been
replaced by adhesively bonded joints because of the weight penalty
and corrosion problems associated with bolted joints. One of the
major issues in applications of adhesively bonded composite joints is
the prediction of damage and failure mechanisms. A complete
investigation of the failure modes leads to an efcient and durable
joint design.
Delamination always refers to interlaminar failure, which initiates
by a crack in the matrix and may lead to ber separation. Therefore, a
matrix crack occurs within laminates where the bers are parallel to
the load direction and these bers are placed adjacent to an adhesive
bond. Based on the direction of crack propagation, edge or local
delaminations have been categorized. Edge delamination occurs at
the load-free edges of the laminate, while local delamination initiates
froma transverse matrix crack. It has been observedin simple tension
tests of a uniform rectangular-cross-section specimen that delami-
nations begin along the load-free edges and propagate normal to the
load direction [1]. Many research papers have been devoted to the
study of matrix cracking in laminated composite plates, although
most have developed nite element approaches. Moreover, several
researchers have studied the surface-ply delamination failure of
laminated composite plates analytically, while surface-ply failure in
bonded laminated composite joints rarely has been discussed.
Among researchers, Tong et al. [2] and Tong [3] have shown the
failure of adhesively bonded composite joints with embedded
interlaminar cracks based on Williamss mode partitioning model
[4]. Recently, Chadegani et al. [5] have developed an analytical
model based on the fracture-mechanics approach for calculating
strain energy release rate of adhesively bonded composite joints with
interlaminar crack for thin bondlines.
An accurate prediction of a delamination failure mechanism and
substrate failure load are essential in understanding the failure
process as well as reliability of bondedlaminated composite joints. In
a bonded composite joint, failure typically occurs in the surface ply
of the adherend near the stress singularity [1].
Delaminations near free edges, holes, ply drops, and notches have
been noticed in laminated composites in service. In practical
composite structures, delamination is a mixed-mode fracture
process, which includes mode I (opening mode), mode II (shearing
mode), and mode III (tearing mode). The growth process of edge
delaminations and local delamination is often modeled by a fracture-
mechanics-based approach leading to the calculation of the strain
energy release rate, which can be considered for various geometries
and loading conditions. In combination with an appropriate failure
criterion, the strain energy release rate can be used as a means to
predict the failure load of the structure. In a research study conducted
by Yang et al. [6] on adhesively bonded joints, it was concluded that a
fracture-mechanics-based approach would be an effective method
for predicting the load-carrying capacity of bonded joints.
Earlier studies on adhesively bonded joints were directed by
Goland and Reissner [7] Erdogan and Ratwani [8], Hart-Smith [9],
and Adams and Peppiatt [10]. Enormous literature on the analysis of
adhesively bonded joints-can be found, among others, in extensive
reviews by Kutscha [11], Kutscha and Hofer [12], Matthews et al.
[13], Vinson [14], da Silva et al. [15], and Zhao et al. [16]. Tsai and
Morton [17] compared results from a two-dimensional (2-D) geo-
metrically nonlinear nite element analysis with those from the
analytical solutions. Yang and Pang [18] derived an analytical model
that provided the stress distributions of adhesively bonded single-lap
composite joints subjected to axial tension. Huang et al. [19] and
Yang et al. [20] also derived an elasticplastic model for adhesively
bonded single-lap composite joints. Their approaches have included
important capabilities such as the asymmetry of the adherend
laminates and effects due to transverse shear deformation. An
existing crack is usually assumed to be in a joint when conducting a
fracture analysis. Several methods in the literature are available for
calculating the strain energy release rate: nite element analysis and
the complex variable stress potential approach [1]. In a report
published in 2002, Krueger [21] described the virtual crack closure
technique (VCCT), including its history, approach, and applications
in conjunction with nite element analysis. Davidson et al. [22]
published a series of papers that employed the classical plate theory
version of the VCCT to predict the strain energy release rate of
mixed-mode delamination in composite laminates. Acracktip force
method was derived by Park and Sankar [23] to compute the strain
energy release rate in delaminated beams and plates. Kim and Kong
[24] proposed a simplied method for determining the strain energy
release rate of free-edge delamination in composites using the
classical laminated plate theory. Davidson [25] and Yu and Davidson
[26] used three-dimensional (3-D) crack-tip elements to analytically
determine strain energy release rate and mode-mixity for different
types of laminated plates containing delamination. Their results were
compared with 3-Dnite element analyses and the VCCT. Moreover,
Rybicki and Kanninen [27] and Raju [28] used nite element
methods to calculate the strain energy release rate.
Finite element methods play a signicant role in structural
analysis and have been widely used to study the adhesively bonded
composite joint. Wang et al. [29] applied the VCCT to calculate the
strain energy release rate of cracked composite panels with nonlinear
deformation. Wei et al. [30] presented an improved VCCT to
determine the energy release rate using a three-step analysis that was
used also by Murthy and Chamis [31]. Contour integrals to compute
the strain energy release rate have been employed, among others, by
Fernlund et al. [32], Yang et al. [33,34], and Chadegani et al. [5].
Yang et al. [33,34] developed nite element models using the nite
element software ABAQUS [35] to estimate the J integral of an
adhesively bonded joint with a crack. Although nite element
methods are capable of solving problems with various types of
materials and complicated geometrical congurations, analytical
methods offer advantageous performance and solutions, especially
with parametric analyses and optimization. Interlaminar fracture in
composites using the sublaminate approach has been studied, among
others, by Armanios and Reheld [3638] and Reheld et al. [39], in
spite of simplications related to their case studies. A comparative
study of the analytical models can be found in a review paper by da
Silva et al. [40].
As described earlier, the fracture-mechanics-based analysis is
believed to be effective in predicting the load-carrying capacity of an
adhesivelybonded composite joint [6]. It should be emphasize that
the advantage of a simple and robust analytical model motivates the
current study. The objective of the present paper is to develop an
analytical fracture-mechanics-based model that can be used to
determine the strain energy release rate and failure load in an
adhesively bonded single-lap composite joint due to surface-ply
delamination failure. Because of the availability of test data, the
model derivation and failure analysis are based on the ASTMD3165
[41] specimen geometry shown in Fig. 1. Cracks usually start at
locations of high stress concentration. For the ASTM D3165
specimen conguration the critical area is located at the corner of the
adhesive layer, where a continued adherend and a discontinued
adhesive are present. Often times, either the adhesive/adherend inter-
face starts to fail, as in the case of metal adherends, or the adherend
starts to fail, as in the case of composite adherends. Failure of such
Fig. 1 ASTM D3165 [41] specimen geometry and dimensions.
504 CHADEGANI, YANG, AND SMELTZER
joints at the adherend/adhesive interface with either thin or thick
bondlines has been studied by Yang et al. [33,34] and Chadegani
et al. [5] Because focus of the current study is on the surface-ply
failure/delamination of the adherend, the failure is assumed to initiate
at the surface ply adjacent to the adhesive, as shown in Fig. 2.
Thereafter, the crack propagates along the length direction of the
joint until the entire joint unzips.
Linearelastic material properties as well as small displacements
are assumed for both the adhesive and adherends in order to make the
analytical approach feasible. In the following sections, step-by-step
procedure and formulation of the problem using laminated
orthotropic plate theory for the adherend to calculate interlaminar
stress distributions are discussed. The analytical solutions are deter-
mined using the symbolic computation tool Maple 9.52 [42]. Results
fromthe analytical model are veried by nite element analysis using
ABAQUS 6.8-1 [35].
II. Stress and Displacement Model Development
Details of the analytical method for determining the strain energy
release rate of adhesively bonded single-lap composite joints with a
prescribed interlaminar crack is based on the laminated plate theory
version of Irwins virtual crack method [43]. The strain energy
release rate is derived in terms of the forces and moment at the crack
tip: N
C
, Q
C
, and M
C
. These forces and moment at the crack tip are
determined fromthe linearelastic shear and peel-stress distributions
at the interface between the rst and second plies of the adherend
adjacent to the adhesive within the overlap area. Therefore, a
description of the stress and displacement states in the pre- and
postpropagation specimen geometry is required before an estimate of
the strain energy release rate can be obtained. A summary of the
methodology used to derive the equations for determining the
required stress and displacement elds in an adhesively bonded joint
is presented in this section.
An adhesively bonded single-lap joint with the standard geometry
of an ASTM D3165 [41] specimen and an applied tensile load P per
unit width is shown in Fig. 2. The joint is divided into six regions in
the x direction for the convenience in the model development, where
regions 1, 4, and 6 consist of two adherends and a thick adhesive
layer, regions 2 and 5 represent two notches, and region 3 is the area
where surface-ply delamination failure of the laminate is located, as
shown in Fig. 2. Region 4 is the bonded-joint overlap area where the
applied mechanical loads are transferred from one adherend to the
other; this is also the area on which joint strength is typically based.
Overall procedure of the model development can be summarized
as follows:
1) Divide the model into adherend portions according to the
ASTM D3165 [41] specimen geometry, and derive the kinematics
relations for each adherend portion.
2) Derive force and moment equilibriums for each adherend
portion according to sign convention.
3) Derive equations for the adhesive and interlaminar stresses
using the kinematics of adherends.
4) Specify boundary conditions according tothe loading status and
displacement constraints.
5) Find the equivalent crack-tip forces and moment, and propagate
the crack to nd displacement distributions.
6) Apply the crack closure method to calculate the strain energy
release rate.
A. Adherend Formulation
The specimen is divided into 12 adherend portions from left to
right, as shown in Fig. 3. The general formulas for the adherend
portions are the same for all, except for adherend portions 5 and 8.
The displacement elds of the adherend are described by the
laminated orthotropic plate theory.
1. Adherend Portions 14, 6, 7, and 912
Based on the rst-order laminated plate theory, the displacement
elds for adherend portions 14, 6, 7, and 912 can be written as
u =u
o
(x) z(x) (1)
a) The entire joint
b) The overlap area
Fig. 2 Regions used in model derivation.
Fig. 3 Discretized model with 12 adherend portions.
CHADEGANI, YANG, AND SMELTZER 505
w =w(x) (2)
where the superscript o represents the midplane displacement, z is
measured from the midplane of each adherend portion, and is the
corresponding bending slope. Considering the given displacement
functions and the strain-displacement relations yields the normal
strain "
x
and shear strain
xz
for each adherend portion as
"
x
=
@u
@x
=
du
o
(x)
dx
z
d(x)
dx
(3)

xz
=
@w
@x

@u
@z
=
dw(x)
dx
(x) (4)
Therefore, based on the classical laminate theory, the relationships
among stress resultants, strain, and bending slope can be formulated.
Furthermore, using the extensional stiffness [A|, the coupling
stiffness [B|, and the bending stiffness [D| for orthotropic laminates,
the normal stress resultant N
x
, bending moment for unit width M
y
,
and transverse shear stress resultant Q
z
can be obtained as
N
x
=A
11
du
o
(x)
dx
B
11
d(x)
dx
(5)
M
y
=B
11
du
o
(x)
dx
D
11
d(x)
dx
(6)
Q
z
=k
s
A
55
_
(x)
dw(x)
dx
_
(7)
where k
s
is the shear correction factor, and A
ij
, B
ij
, and D
ij
are terms
taken from the common extensional, bending, and extensional-
bending coupling stiffness matrices from the laminated plate theory
as
(A
11
; B
11
; D
11
) =
_
h=2
h=2
Q
(i)
11
[ 1; z; z
2
| dz (8)
A
55
=
_
h=2
h=2
Q
(i)
55
dz (9)
where Q
(i)
11
and Q
(i)
55
represent the normal and transverse shear
stiffness, respectively, of the ith ply, and h is the thickness of the
adherend.
2. Adherend Portion 5
As shown in Fig. 3, adherend portion 5 is the delaminated surface
ply within the overlap area. Because of the small length-to-thickness
ratio of adherend portion 5, which is a single ply, displacement
distributions u and w in the x and z directions, respectively, are
assumed to be independent of z, u
o
5
, and w
o
5
(see Fig. 4), i.e., no
bending slope contributes to u within this portion. This is a
reasonable assumption because of the thickness in this adherend
portion, which is a single ply, small, and traction-free at the top
surface. The normal and shear stress resultants N
5x
and Q
5z
and the
bending moment M
5y
per unit width are related to the strains within
this adherend portion and the constitutive relations of the adhesive
material. Because of small thickness, M
5y
is assumed to be
negligible:
N
5x
=
_
h
2
=2
h
2
=2
Q
11
du
o
5
(x)
dx
dz (10)
M
5y
=
_
h
2
=2
h
2
=2
Q
11
du
o
5
(x)
dx
z dz ~0 (11)
Q
5z
=
_
h
2
=2
h
2
=2
k
s
Q
55
_
dw
o
5
(x)
dx
_
dz (12)
where h
2
is the thickness of adherend portion 5.
3. Adherend Portion 8
Displacement distributions u and w in the x and z directions,
respectively, within adherend portion 8, which is also a single ply, are
assumed to follow a similar approach as discussed for adherend
portion 5. The normal and shear stress resultants and bending
moment per unit width can be related to the displacement elds as
specied in Eqs. (1012).
To correlate the shear and peel stresses at the adherend/adhesive
interface at the bottom of adherend portion 8, the displacements u
o
8
and w
o
8
at the midplane, and the displacements u
+
8
and w
+
8
at the
bottomsurface of adherend portion 8, as shown in Fig. 5, are used by
assuming that
z
and
xz
are uniform within the lower half of
adherend portion 8. Therefore, as shown in Fig. 5, the shear stress and
peel stresses
+
a
and q
+
a
are the same as
xz
and
z
, respectively, within
the lower half of adherend portion 8 and can be represented in terms
of the midplane displacements u
o
8
and w
o
8
and the bottom-surface
displacements u
+
8
and w
+
8
as

+
a
=
xz
[
z=h
2
=2
=G
xz
_
2
h
2
(u
o
8
u
+
8
)
dw
o
8
dx
_
(13)
q
+
a
=
z
[
z=h
2
=2
=
_

xz

xy

yz
E
x
E
y

_
du
o
8
dx
2
_
1
xy

yx
E
x
E
y

_
(w
o
8
w
+
8
)
h
2
(14)
where
xy
and
yx
are the major and minor Poissons ratios in the x-y
plane;
xz
is the major Poissons ratio in the x-z plane;
yz
is the major
Poissons ratio in the y-z plane; E
x
, E
y
, and G
xz
are the moduli of
elasticity in the x and y directions and shear modulus in the x-z plane,
respectively; and can be given as
=
1
xy

yx

yz

zy

zx

xz
2
xy

yz

zx
E
x
E
y
E
z
(15)
where
zy
and
zx
are minor Poissons ratios in the y-z plane and the
x-z plane, respectively, and E
z
is the modulus of elasticity in the z
direction.
The interlaminar shear and peel stresses
i
and q
i
at the upper
surface of adherend portion 8, as shown in Fig. 5, which are the
stresses between adherend portions 7 and 8, are to be used for the
equivalent crack-tip forces calculation and, later on, for the strain
energy release rate determination. Assuming uniform
z
and
xz
within the upper half of adherend portion 8 and a perfect bond
between adherend portions 7 and 8, based on the kinematics and
Fig. 4 Free-body diagram of adherend portion 5. Fig. 5 Free-body diagram of adherend portion 8.
506 CHADEGANI, YANG, AND SMELTZER
constitutive relations,
i
and q
i
are related to the displacement
functions of adherend portions 7 and 8 as
q
i
=
_

xz

xy

yz
E
x
E
y

_
du
o
8
dx

2
h
2
_
1
xy

yx
E
x
E
y

_
(w
7
w
o
8
) (16)

i
=G
xz
_
2
h
2
_
u
o
7

h
1
2

7
u
o
8
_

dw
o
8
dx
_
(17)
where h
1
is the thickness of adherend portion 7.
The bottom surfaces u
+
8
and w
+
8
, shown in Fig. 5, will be solved
later by the equilibrium equations of adherend portions 8 and 9.
B. Adhesive Formulation
The adhesive is assumed tobehave as anelasticisotropic material.
The adhesive displacements in the x direction, u
a
, and the z direction,
w
a
, are assumed to be polynomials of both x and z. To ensure the
continuity of the displacement elds of the adhesive and adherends at
the interfaces (z =h=2), u
a
and w
a
are written as
u
a
=u
o
(x; z)

n
m=1
a
m
u
m
(x; z) =u
o
(x; z)

_
z

2
__
z

2
_
(a
1
a
2
x a
3
z ) (18)
w
a
=w
o
(x; z)

n
m=1
b
m
w
m
(x; z) =w
o
(x; z)

_
z

2
__
z

2
_
(b
1
b
2
x b
3
z ) (19)
where is the adhesive thickness, and u
o
(x; z) and w
o
(x; z) are
different for each region and will be described in the following
sections.
The assumption of terms u
m
and w
m
in the adhesive displacement
elds, which are zero at both upper and lower adhesive/adherend
interfaces, z ==2 and z ==2, can be regarded as the deviation
of the displacement eld from a linear function of z.
It is assumed that the adhesive normal stress in the x direction is
negligible; only peel stress
a;z
and shear stress
a;xz
exist in the
adhesive. Under the assumption of plane-strain condition, the
adhesive peel and shear stresses
a;z
and
a;xz
can be formulated in
terms of the adhesive strains "
a;x
, "
a;z
, and
a;xz
as

a;z
=
E
a
(1
a
)(1 2
a
)
[
a
"
a;x
(1
a
)"
a;z
|
=
E
a
(1
a
)
2
"
a;z
=
E
a
(1
a
)
2
@w
a
@z
(20)

a;xz
=G
a

a;xz
=G
a
_
@u
a
@z

@w
a
@x
_
(21)
where E
a
and G
a
are the adhesive Youngs and shear moduli,
respectively, and
a
is the Poissons ratio of the adhesive.
The adhesive shear and peel stresses at the upper and lower
adherend/adhesive interfaces,
U
a
,
L
a
, q
U
a
, and q
L
a
, as shown in Fig. 6,
are

U
a
=
a;xz
[
z=

2
(22)

L
a
=
a;xz
[
z=

2
(23)
q
U
a
=
a;z
[
z=

2
(24)
q
L
a
=
a;z
[
z=

2
(25)
Strain energy in the adhesive can be written as
U =
1
2
__
(
a;z
"
a;z

a;z

a;z
) dx dz
=
E
a
2(1
a
)
_
=2
=2
__
l
0
_
1
1
a
_
@w
a
@z
_
2

1
2
_
@u
a
@z

@w
a
@x
_
2
_
dx
_
dz (26)
where l is the length of the overlap. Using the principle of minimum
potential energy, the displacement u
a
and w
a
must satisfy the
following equations when the adhesive is under static equilibrium
condition:
@U
@a
m
=

Xu
m
dS m=1; 2; 3; . . . ; n (27)
@U
@b
m
=

Zu
m
dS m=1; 2; 3; . . . ; n (28)
where

X and

Z represent the surface loads in the x and z directions
applied on the adhesive, respectively, and dS covers the periphery of
the adhesive. Because the adhesive has two free surfaces at the left
and right edges in each region, both surface loads

X and

Z are zero at
these two edges. As described previously in Eqs. (18) and (19), u
m
and w
m
are zero at both the upper and lower adherend/adhesive
interfaces in order to satisfy the continuity condition. Therefore,
Eqs. (27) and (28) become
@U
@a
m
=0 m=1; 2; 3; . . . ; n (29)
@U
@b
m
=0 m=1; 2; 3; . . . ; n (30)
1. Adhesive Midplane Displacement-Field Formulation in Regions 1
and 6
Using displacement elds of the upper and lower adherends,
adhesive midplane displacement elds can be written as
u
o
(x; z) =
1
2
u
oU
(x)
1
4
h
U

U
(x)
1
2
u
oL
(x)
1
4
h
L

L
(x)

z
2
[2u
oU
(x) h
U

U
(x) 2u
oL
(x) h
L

L
(x)| (31)
Fig. 6 General free-body diagram and sign convention.
CHADEGANI, YANG, AND SMELTZER 507
w
o
(x; z) =
1
2
w
U
(x)
1
2
w
L
(x)
z

[w
U
(x) w
L
(x)| (32)
where h
U
and h
L
are the thicknesses of the upper and lower
adherends, respectively.
2. Adhesive Midplane Displacement-Field Formulation in Region 3
Using displacement elds of adherend portions 5 and 6 on the top
and bottom interfaces of the adhesive, respectively, the adhesive
midplane displacement elds can be written as
u
o3
(x; z) =
1
2
u
o
5
(x)
1
2
u
o
6
(x)
1
4
h
L

6
(x)
z
2
[2u
o
5
(x)
2u
o
6
(x) h
L

6
(x)| (33)
w
o3
(x; z) =
1
2
w
5
(x)
1
2
w
6
(x)
z

[w
5
(x) w
6
(x)| (34)
3. Adhesive Midplane Displacement-Field Formulation in region 4
In region 4, the adherend portion above the adhesive (namely,
adherend portion 8) is a single ply. While the adherend portion above
the adhesive is modeled using the approach discussed previously, the
adherend portion belowthe adhesive is modeled using the rst-order
shear-deformation theory. In terms of the displacement elds of the
adherend portions above and below the adhesive, the adhesive
midplane displacements can be written as
u
o4
(x; z) =
1
2
u
+
8
(x)
1
2
u
o
9
(x)
1
4
h
L

9
(x)
z
2
[2u
+
8
(x)
2u
o
9
(x) h
L

9
(x)| (35)
w
o5
(x; z) =
1
2
w
+
8
(x)
1
2
w
9
(x)
z

[w
+
8
(x) w
9
(x)| (36)
where u
+
8
and w
+
8
are bottom-surface displacements of the adherend
portion 8 in the x and z directions, respectively, as shown in Fig. 5.
C. Adherend Equilibrium Equations
To establish the equations of equilibrium for each adherend
portion, a free-body diagram of a differential element from the
overlap regions is illustrated in Fig. 6. The upper adherend portion
shown in the gure is the adherend portion immediately above the
adhesive and represents adherend portions 1, 5, 8, and 12 in regions
1, 3, 4, and 6, respectively.
The lower adherend portion in the gure is the lower adherend
under the adhesive in each region. The general equations for force
and moment equilibrium of the adherend portion above the adhesive
are given as
dN
U
x
dx
=
i

U
a
(37)
dM
U
y
dx
=Q
U
z

h
U
2
(
U
a

i
) (38)
dQ
U
z
dx
=q
U
a
q
i
(39)
where
U
1a
and q
U
1a
are the shear and peel stresses on the top surface of
the adhesive in region 1,
i
and q
i
are the interlaminar shear and peel
stresses of the upper adherend portion, and h
U
is the thickness of the
upper adherend portion. It should be noted that
i
and q
i
do not exist
in regions 1 and 6 because the upper adherend portions in these two
regions are modeled as one piece, while the upper adherend portions
in regions 3 and 4 are single-ply, so
i
and q
i
may be introduced.
Three equilibriumequations also can be obtained for the adherend
portion below the adhesive in a similar manner, but without any
stresses at the bottom surface, as
dN
L
x
dx
=
L
a
(40)
dM
L
y
dx
=Q
L
z

h
L
2

L
a
(41)
dQ
L
z
dx
=q
L
a
(42)
where
L
1a
and q
L
1a
are the shear and peel stresses on the bottom
surface of the adhesive in region 1.
Based on the general equilibrium equations for the adherend
portions above and below the adhesive, equations of equilibrium for
each adherend portion can be written and are described in detail
below.
1. Region 1: Adherend Portions 1 and 2
From the free-body diagram in Fig. 6, it can be observed that

i
=0 and q
i
=0, because the top surface of adherend portion 1 is a
traction-free surface. Hence, Eqs. (3742) can be rewritten for
region 1 as
dN
1x
dx
=
U
1a
(43)
dM
1y
dx
=Q
1z

h
2

U
1a
(44)
dQ
1z
dx
=q
U
1a
(45)
dN
2x
dx
=
L
1a
(46)
dM
2y
dx
=Q
2z

h
2

L
1a
(47)
dQ
2z
dx
=q
L
1a
(48)
By substituting the stress resultants from Eqs. (57) and the
adhesive stresses on the top and bottom surfaces from Eqs. (2025),
six coupled second-order ordinary differential equations in terms of
u
o
1
, u
o
2
,
1
,
2
, w
1
, and w
2
are obtained.
2. Region 2: Adherend Portion 3
Region 2 represents a notch in the overall specimen geometry.
Because of the interlaminar failure assumed between adherend
portions 4 and 5, where the surface ply of the upper adherend
adjacent to the adhesive is supposed to continue from the lower right
end of adherend portion 3 and form adherend portion 5, a crack is
initiated between adherend portions 3 and 5, as shown in Fig. 7, as the
free surface. Because of the free surface at the lower right end of
adherend portion 3, the effective thickness and the tensile, bending,
and shear stiffnesses of adherend portion 3 are different from its left
end to its right end. Based on the free-body diagram shown in Fig. 7,
the equations of equilibrium can be written as
N
L
3x
=N
R
3x
(49)
Q
L
3z
=Q
R
3z
(50)
508 CHADEGANI, YANG, AND SMELTZER
M
R
3y
=M
L
3y
N
L
3x
h
2
2
Q
L
3z
L
2
(51)
where superscripts Land Rrefer to the left and right ends of adherend
portion 3, respectively; h
2
is the thickness of a single ply; and L
2
is
the entire length of adherend portion 3. Toobtain u
o
3
,
3
, and w
3
at the
right end of adherend portion 3, the following equations are used,
assuming a small change in their derivatives:
u
o
3
[
x
2
=L
2
=u
o
3
[
x
2
=0

L
2
2
_
du
o
3
dx

x
2
=0

du
o
3
dx

x
2
=L
2
_
(52)

3
[
x
2
=L
2
=
3
[
x
2
=0

L
2
2
_
d
3
dx

x
2
=0

d
3
dx

x
2
=L
2
_
(53)
w
3
[
x
2
=L
2
=w
3
[
x
2
=0

L
2
2
_
dw
3
dx

x
2
=0

dw
3
dx

x
2
=L
2
_
(54)
where the rst-order derivatives of u
o
3
,
3
, and w
3
at the right end,
where x
2
=L
2
can be determined by the following equations, which
are based on equilibrium equations Eqs. (4951). In Eqs. (5557),
the values of u
o
3
,
3
, w
3
, and their derivatives at the left end, where
x
2
=0, are obtained from the right end of adherend portion 1 using
the continuity condition. It should be noted that stiffnesses A
11
, B
11
,
D
11
, and A
55
have different values at the left and right ends due to
delamination of the surface ply at the right end:
_
A
11
du
o
3
dx
_

x
2
=0

_
B
11
d
3
dx
_

x
2
=0
=
_
A
11
du
o
3
dx
_

x
2
=L
2

_
B
11
d
3
dx
_

x
2
=L
2
(55)
_
k
s
A
55
_

3

dw
3
dx
__

x
2
=0
=
_
k
s
A
55
_

3

dw
3
dx
__

x
2
=L
2
(56)
_
B
11
du
o
3
dx
_

x
2
=L
2

_
D
11
d
3
dx
_

x
2
=L
2
=
_
B
11
du
o
3
dx
_

x
2
=0

_
D
11
d
3
dx
_

x
2
=0

h
2
2
__
A
11
du
o
3
dx
_

x
2
=0

_
B
11
d
3
dx
_

x
2
=0
_
L
2
_
k
s
A
55
_

3

dw
3
dx
__

x
2
=0
(57)
3. Region 3: Adherend Portion 4
Because of delamination of the surface ply at the bottom and the
traction-free surface at the top, equations of equilibrium of adherend
portion 4 can be written as
dN
4x
dx
=0 (58)
dM
4y
dx
=Q
4z
(59)
dQ
4z
dx
=0 (60)
Substituting the stress resultants as functions u
o
4
,
4
, and w
4
into
Eqs. (5860) yields three coupled second-order ordinary differential
equations in terms of u
o
4
,
4
, and w
4
:
A
11
d
2
u
o
4
(x)
dx
2
B
11
d
2

4
(x)
dx
2
=0 (61)
B
11
d
2
u
o
4
(x)
dx
2
D
11
d
2

4
(x)
dx
2
=k
s
A
55
_

4
(x)
dw
4
(x)
dx
_
(62)
k
s
A
55
_
d
4
(x)
dx

d
2
w
4
(x)
dx
2
_
=0 (63)
4. Region 3: Adherend Portions 5 and 6
As previously noted, adherend portion 5 contains a single ply and
is separated from adherend portion 4 due to the prescribed
interlaminar crack. The equations of equilibrium of adherend
portions 5 and 6 are similar to those for adherend portions 1 and 2 but
without considering the bending moment in adherend portion 5, as
previously mentioned:
dN
5x
dx
=
U
3a
(64)
dQ
5z
dx
=q
U
3a
(65)
dN
6x
dx
=
L
3a
(66)
dM
6y
dx
=Q
6z

h
2

L
3a
(67)
dQ
6z
dx
=q
L
3a
(68)
Substituting the stress resultants from Eqs. (10) and (12) for N
5x
and Q
5z
and the stress resultants from Eqs. (57) for N
6x
, M
6y
, and
Q
6z
, into Eqs. (6468), ve coupled second-order ordinary
differential equations u
o
5
, w
o
5
, u
o
6
, w
6
, and
6
are obtained.
5. Region 4: Adherend Portions 7, 8, and 9
A similar approach as that used for region 3 for adherend portions
5 and 6 can be applied to region 4, with an additional adherend
thickness at the top and the addition of the interlaminar stresses
between adherend portions 7 and 8. Eight equations can be written
based on normal stress resultants, bending moment, and shear stress
resultants as
dN
7x
dx
=
i
(69)
dM
7y
dx
=Q
7z

h
1
2

i
(70)
dQ
7z
dx
=q
i
(71)
Fig. 7 Free-body diagram of adherend portion 3.
CHADEGANI, YANG, AND SMELTZER 509
dN
8x
dx
=
U
4a

i
(72)
dQ
8z
dx
=q
i
q
U
4a
(73)
dN
9x
dx
=
L
4a
(74)
dM
9y
dx
=Q
9z

h
L
2

L
4a
(75)
dQ
9z
dx
=q
L
4a
(76)
where h
1
and h
L
are thicknesses of adherend portion 7 and adherend
portion 9, respectively. The interlaminar shear and peel stresses
i
and q
i
between adherend portions 7 and 8 are in terms of
displacement elds of adherend portions 7 and 8, as shown in
Eqs. (16) and (17).
Thereby, the displacements at the bottom surface of adherend
portion 8, u
+
8
and w
+
8
, as shown in Fig. 5, can be solved in terms of all
other variables by equating the shear and peel stresses on the bottom
surface of adherend portion 5, as specied in Eqs. (13) and (14), and
their corresponding shear and peel stresses from the top surface of
adhesive in region 4:

+
a
=
U
4a
(77)
q
+
a
=q
U
4a
(78)
6. Region 5: Adherend Portion 10
This region represents a notch in the model. The equations of
equilibrium for adherend portion 5 are the same as those given for
adherend portion 4 but with different adherend stiffnesses.
7. Region 6: Adherend Portions 11 and 12
Because of the similarities between region1 and region 6, the same
equations of equilibrium can be derived for region 6 with
corresponding variables.
III. Solution Methodology
The overall system of governing equations, including all six
regions, contains 31 s-order ordinary differential equations with 31
unknown variables. A total of 62 boundary conditions, summarized
in Appendix, are obtained at the two ends of each region based on
either continuity or applied force conditions. The symbolic solver
Maple 9.52 [42] was used to solve the systemof equations and obtain
the displacement, strain, and stress elds. Governing equations for
each region are parametrically solved rst and then continuity and
boundary conditions are evaluated to solve for constants of
integrations.
IV. Strain Energy Release Rate Calculation
Once the stress, strain, and displacement elds are known in the
adhesively bonded single-lap composite joints, the crack closure
method is applied to estimate the strain energy release rate of the joint
with a prescribed crack. According to linearelastic fracture
mechanics, the energy released due to an extension/propagation of a
crack is equivalent to the work needed to close that extension/
propagation [27], and this is the foundation of the crack closure
method.
A. Analytical Approach
In deriving the expression for the strain energy release rate, a joint
is assumed to have an overlap length l
o
, a notch size L
2
, and a crack
length a, which is located at the interface of the bottomsurface ply of
the upper adherend, as shown in Fig. 8. The displacement of the crack
tip C, after a load is applied, can be determined using the mechanical
model previously described with an overlap region length of
L
4
=l
o
a.
Interlaminar and adhesive stress distributions can be determined
using the solution from the analytical model described previously.
Once the crack propagates a small length b, the previous crack tip C
separates into two points A and B, as shown in Fig. 9. Before the
crack growth of the additional small length b, the surface ply between
C
/
and Cadheres to the upper adherend, where interlaminar shear and
peel stresses exist at the interface, as shown on the left side of Fig. 10.
The interlaminar shear and peel stresses between C
/
and C vanish
after the crack propagates a small length b. The equivalent crack-tip
forces and moment N
C
, M
C
, and Q
C
, corresponding to a small crack
propagation b, are related to the shear and peel stresses between C
/
and C, as shown on the right side of Fig. 10, and can be calculated as
N
C
=
_
b
0

i
dx
4
(79)
M
C
=
_
b
0
q
i
x
4
dx
4
(80)
Q
C
=
_
b
0
q
i
dx
4
(81)
where the interlaminar shear stress
i
and peel stress q
i
are obtained
from the stress model with an overlap length L
4
=l
o
a. Note that
only the bottom surface ply of the upper adherend, adhesive layer,
and lower adherend are shown in Fig. 10 and that the positive
directions of N
C
, M
C
, and Q
C
are dened to be consistent with the
positive directions of displacements u, , and w, respectively.
To determine the relative displacements between points Aand B, a
subsequent stress analysis is performed using a joint with a central
overlap length L
4
=l
o
a b, which simulates the overlap up to
the new crack tip C
/
.
Fig. 8 ASTM D3165 [41] specimen with initial crack of length a.
Fig. 9 ASTM D3165 [41] specimen with a virtual crack extension of
length b.
510 CHADEGANI, YANG, AND SMELTZER
To close the virtual crack propagation (length b), crack-tip forces
are applied on points Aand Bto move themback to the location of the
original crack tip C. Therefore, the total work required to close the
small crack propagation b is
W =
1
2
[N
C
(u
B
u
A
) M
C
(
B

A
) Q
C
(w
B
w
A
)| (82)
For joints with a unit width, the strain energy release rate is dened
as the derivative of energy released from the crack propagation with
respect to the length of the crack propagation as
G
T
=
dU
da
(83)
where U is the strain energy stored in the body. Based on the crack
closure method, the total energy released fromthe crack propagation
is equivalent to the work needed to close the same crack propagation.
The total strain energy release rate G
T
is a summation of mode I,
mode II, and Model III strain energy release rates in a three-
dimensional case. However, under the current plane-strain condition
the mode III fracture does not exist. Therefore, G
T
can be calculated
as Eq. (84) with a virtual crack propagation of length b, where the
equivalent crack-tip moment M
C
and transverse force Q
C
contribute
to mode I strain energy release rate and the equivalent crack-tip in-
plane force N
C
contributes to mode II strain energy release rate:
G
T
=G
I
G
II
=
dU
da

W
b
=
1
2b
[M
C
(
B

A
)
Q
C
(w
B
w
A
)|
1
2b
[N
C
(u
B
u
A
)| (84)
B. Finite Element Approach Using Three-Step VCCT
The strain energy release rate due to the small increase in crack
length is equivalent to the energy per created crack surface area that is
required to close that small crack increment. Therefore, the strain
energy release rate can be computed by nite element models using
the three-step VCCT described below [33,34]. As shown in Fig. 11,
the tip of a crack with an original length a is located at C.
Assuming a virtual crack propagation of length b, the new crack
tip becomes C
/
, and the original crack tip becomes two separate
nodes A and B. If nodes A and B are restrained at the original crack-
tip location, this virtual crack of length b is closed, and the work to
close this virtual crack can be calculated by multiplying the reaction
forces at nodes A and B by the relative displacements of these two
separate nodes to the original crack tip C. The three-step procedures
to calculate the strain energy release rate using nite element
software are described as follows:
1) Build the nite element model with an original crack of length
a, and determine the displacements of crack tip C, u
c
, and w
c
in the x
and z directions, respectively.
2) Propagate the crack with a small length b (usually one element
size); then the original crack tip C becomes two separate nodes.
Record the displacements of both nodes u
A
, w
A
, u
B
, and w
B
.
3) Constrain the two separate nodes so that they have the same
displacements as the original crack tip C, and obtain the reaction
forces F
xA
, F
yA
, F
xB
, and F
yB
.
4) The work needed to close the virtual crack is
W =
1
2
[F
xA
(u
c
u
A
) F
yA
(w
c
w
A
)|
1
2
[F
xB
(u
c
u
B
)
F
yB
(w
c
w
B
)| (85)
The total strain energy release rate is then obtained by
G
T
=
W
b
(86)
In the present study, the commercial nite element software
ABAQUS was used following the three-step procedures described
above. Two-dimensional four-node bilinear quadrilateral plane-
strain elements with reduced integration and hourglass control
(CPE4R) were used in the nite element model for the three-step
VCCTapplication. The convergence of the strain energy release rate
solution using VCCT was studied by sequential renement of the
nite element mesh, as shown in Fig. 12. Hence, the nite element
models with a total of 18,000 elements, which corresponds to 20-
eight elements through the adherend thickness, i.e., four elements per
ply, were used to analyze the joints of seven-ply laminates of T300/
E765 3KPW plain-weave carbon/epoxy (C/Ep) and 7781/E765
satin-weave glass/epoxy (Gl/Ep), both with the same layup sequence
of [0=60= 60=0= 60=60=0]. Detailed view of the deformed
mesh around the crack tip is shown in Fig. 13. Dimensions of the
elements around the crack are 0:0635 0:0635 mm
2
.
Solution of the near crack-tip eld indicates oscillation of the
stresses in the immediate vicinity of the crack tip for cracks located at
the bimaterial interface [21,44]. As a result, the mode-mixity is not
deterministic as crack extension (length b) goes to zero. Thereafter,
strain energy release rate was calculated for various choice of b, as
shown in Fig. 14, to accommodate this issue and b=a =0:2 (or
b =0:254 mm), the minimum value for stable G values, was
selected and was used hereinafter.
C. Finite Element Approach Using J Integral
Finite element models with the J integral calculation were
constructed using ABAQUS to verify the present analytical model.
Two-dimensional eight-node biquadratic planestrain elements with
Fig. 10 Equivalent forces at the crack tip.
Fig. 11 Finite element approach using VCCT.
Fig. 12 Convergence of strain energy release rate vs nite element
mesh renement.
CHADEGANI, YANG, AND SMELTZER 511
reduced integration and hourglass control (CPE8R) were used in the
nite element model to calculate J integral. It should be noted that
midside nodes of those elements sharing the crack tip moved to the
quarter-points of the each element as described in [25].
The J integral is usually used in quasi-static fracture analysis to
characterize the energy release associated with crack propagation.
It is equivalent to the strain energy release rate if the material
response is linearelastic. Considering an arbitrary counter-
clockwise path around the crack tip, as illustrated in Fig. 15, the
J integral is dened as
J =
_

_
U
+
dy T
i
@u
i
@x
ds
_
(87)
where repeated index implies summation, U
+
is the strain energy
density, u
i
(i =1; 2; 3) are the components of the displacement
vector, ds is the incremental length along the contour , and T
i
(i =1; 2; 3) are components of the traction vector. The traction is a
stress vector normal to the contour. In other words, T
i
are the
components of the normal stresses acting at the boundary if a free-
body diagram on the material inside of the contour is constructed.
Several contour integral evaluations are possible at each location
along the crack front. In a nite element model, each evaluation can
be thought of as the virtual motion of a block of material surrounding
the crack tip. Each block is dened by contours, and each contour is a
ring of elements completely surrounding the crack tip or crack front
from one crack face to the opposite crack face. These rings of
elements are dened recursively to surround all previous contours.
ABAQUS/Standard automatically nds the elements that form each
ring from the node sets given as the crack-tip or crack-front
denition. Each contour provides an evaluation of the contour
integral [27].
One should carefully apply Eq. (87) to calculate J integral since
the crack is located at the interface of two different linear materials.
Smelser and Gurtin [45] and Chou [46] have extended this quantity
for bimaterial bodies and have recommended this equation, without a
change, to be applied only when the bondline is straight. Referring to
the deformed congurations of specimen shown in Fig. 16, it is
obvious that bondline is straight; therefore, the extension of the
Eq. (87) can be used.
Theoretically, the J integral should be independent of the domain
used, but the J integral estimated from different rings may vary
because of the approximate nature of the nite element solution.
Strong variation in these estimates, commonly called domain-
dependent or contour-dependent, indicates a need for mesh rene-
ment (provided that the problem is suitable for contour integrals).
Numerical tests suggest that the estimate from the rst ring of
elements abutting the crack front does not provide a high-accuracy
result, so at least two contours are recommended. In the present study
10 contours were used, and the average of last ve contours was
taken as the nal J integral value. This method is quite robust in the
sense that accurate contour integral estimates are usually obtained
even with quite-coarse meshes.
Sharp cracks, where the crack faces lie on top of one another in the
undeformed conguration, are usually modeled using small-strain
assumptions. Focused meshes for the J integral calculation should
normally be used for small-strain fracture-mechanics evaluations.
For linearelastic materials, the linearelastic fracture-mechanics
approach predicts an r
1=2
singularity near the crack tip, where r is
Fig. 13 Detail view of the deformed mesh around the crack tip for
VCCT calculation.
Fig. 14 Convergence of strain energy release rate as a function of
normalized virtual crack extension length.
Fig. 15 Arbitrary contour around crack tip.
Fig. 16 Detail view of the deformed mesh around the crack tip for J
integral calculation.
512 CHADEGANI, YANG, AND SMELTZER
the distance fromthe crack tip. In nite element analyses, forcing the
elements at the crack tip to exhibit an r
1=2
strain singularity greatly
improves accuracy and reduces the need for a high degree of mesh
renement at the crack tip [47]. This r
1=2
singularity can be
produced using an eight-node quadrilateral element by moving the
midside nodes to the quarter-points, as noted by Barsoum [48] and
Henshell and Shaw [49]. Yang et al. [33,34] have demonstrated a
complete step-by-step procedure calculating the J integral using
ABAQUS. Even though the singularity at the crack tip of two
dissimilar materials or orthotropic plates is known to be of the form
1=2 i for two-dimensional problems, the use of quarter-point
singularity elements that produce the classical square-root singu-
larity, r
1=2
, at the crack tip has been shown to yield converged results
for the total strain energy release rate G
T
calculation and a slightly
nonconverged results for the individual modes [50] as described
previously. In the current study, two-dimensional eight-node
nonlinear quadrilateral plane-strain elements were used in the nite
element model for the J integral application.
V. Material Properties, Layup, Dimensions,
and Applied Load
To demonstrate the application of the developed model ASTM
D3165 [41] specimens using T300/E765 3KPWplain-weave carbon/
epoxy (C/Ep) and 7781/E765 satin-weave glass/epoxy (Gl/Ep) with
a ply thickness of 0.25 mm and a quasi-isotropic layup sequence of
[60= 60=0= 60=60=0] as adherends were used. An interlaminar
crack was assumed at one edge of the overlap as described in the
model development section. The engineering constants for C/Ep and
Gl/Ep lamina were as follows.
C/Ep:
E
11
=56:50 GPa; E
22
=55:20 GPa
G
12
=3:86 GPa;
12
=0:076
G1/Ep:
E
11
=26:40 GPa; E
22
=23:60 GPa
G
12
=4:14 GPa;
12
=0:17
For convenience, other mechanical properties of adherends were
assumed as follows.
C/Ep:
E
33
=8:00 GPa; G
13
=G
23
=3:56 GPa

13
=
23
=0:35
G1/Ep:
E
33
=3:74 GPa; G
13
=G
23
=1:66 GPa

13
=
23
=0:16
The shear correction factor k
s
=5=6 was assumed. Three different
paste adhesives, Hysol EA9394, PTM&WES6292, and MGS A100/
B100, all with distinct moduli, were used in the analyses. These
adhesives have the following material properties.
Hysol EA9394:
E =4:24 GPa; =0:45
PTM&W ES6292:
E =1:58 GPa; =0:31
MGS A100/B100:
E =3:51 GPa; =0:34
The joint dimensions of the ASTM D3165 [41] specimen include
the central overlap length l
o
=30 mm, notch size L
2
=1:6 mm, and
adherend lengths outside the central overlap L
1
=L
6
=78:4 mm.
Load P =1:78 kN=m as the minimum load without failure was
applied to calculate the strain energy release rate in this study.
Test data obtained by Tomblin et al. [51] for joint specimens with
various bondline thickness and distinct adherend and adhesive
materials were analyzed and used for joint-failure load prediction.
Results are given in the next section.
VI. Results and Discussion
In this study, ASTM D3165 [41] specimens with an interlaminar
crack were modeled analytically to determine the strain energy
release rate using the methodology described in the previous
sections. The symbolic solver Maple 9.52 [42] was used as the
mathematical tool. The nite element models for the VCCT and J
integral were conducted using ABAQUS 6.8-1 [35] to verify the
analytical results.
Once the stress, strain, and displacement elds were obtained, the
strain energy release rates of the joints were estimated, and the
solutions from the analytical models were compared with nite
element models using the VCCTand the J integral. mode I, mode II,
and total strain energy release rates of an initial surface-ply
delamination, a, of 1.27 mm under a 1.78 kN tensile load for a C/Ep
adherend with Hysol EA9394 adhesive of bondline thicknesses from
0.127 mmto 3.175 mm are shown in Figs. 1719, while Figs. 2022
showthe results with Gl/Ep adherends. The strain energy release rate
values for C/Ep and Gl/Ep are plotted vs bondline thickness to depict
the purpose of the present analytical model in dealing with thick
bondlines. These gures also show the comparisons of strain energy
release rate among the present model, the previous model [5] that
uses assumptions of thin bondline and uniform adhesive stresses
through bondline thickness, and the nite element model using the
VCCT. As expected, strain energy release rate values from different
models show an excellent match for the cases of a thin bondline,
whereas the previous thin-bondline model shows large deviations for
thicker bondlines. As can be seen from Figs. 1722 that the
developed new model correlate reasonably well with the nite
element results using the VCCT and also the J integral for G
I
, G
II
,
and G
T
calculations. From these gures, it can be seen that G
I
, G
II
,
and G
T
increase with the increase in bondline thickness under the
same applied load. If strain energy release rate is critical for joint
failure, the load-carrying capacity of the joint will be lower when the
bondline is thicker. It is also noted that the contribution of G
II
to G
T
is
less than G
I
in all cases.
Even though the purpose of this investigation is to derive an
analytical model to calculate the strain energy release rates of an
adhesively bonded composite joint due to a delamination crack at the
surface ply of the adherend, an attempt was made to predict the joint
strength using the limited available data. The test data used for
analysis were obtained by Tomblin et al. [51] of the National Institute
for Aviation Research at Wichita State University. The specimens
Fig. 17 Mode I strain energy release rate as a function of bondline
thickness with C/Ep adherends and Hysol EA 9394 adhesive.
CHADEGANI, YANG, AND SMELTZER 513
used for strength analysis are those failed under the same delami-
nation mode of the surface ply of the adherend. Ideally, a series of
separate ASTMtests, including the double-cantilever-beamtest [52],
the end-notch exure test [53,54], and the mixed-mode bending test
[55], should be conducted in order to determine the critical strain
energy release rates, G
Ic
and G
IIc
, of the composite laminates with
various mode mixities G
II
=G
T
. The critical total strain energy release
rate G
Tc
=G
Ic
G
IIc
, which is then a function of mode-mixity, can
be used as the failure criterion for the joint specimens analyzed here
in this paper. Other criteria, such as power law criterion, exponential
hackle criterion, bilinear criterion, etc., were investigated by
described by Reeder [56]. Because of the fact that there are no
available data on the fracture toughness, G
Ic
, G
IIc
, etc., of the two
composite laminates, [60= 60=0= 60=60=0] T300/E765 3KPW
plain weave and 7781/E765 stain weave, which were used for the
ASTM D3165 [41] specimens analyzed in this paper, the required
G
Ic
and G
IIc
were backcalculated from the experimental data of
failure load of ASTM D3165 specimens. This was done by applying
the experimental failure load of each of the specimens of the same
laminate with a surface-ply delamination of 1.27 mm at the edge of
the overlap to determine the G
I
and G
II
values using the analytical
model. The average values of the G
I
and G
II
at failure among all
specimens with the same composite adherend, even with different
adhesives, are used as G
Ic
and G
IIc
. In other words, two sets of G
Ic
and G
IIc
were determined, one for T300/E765 (C/Ep) laminate,
G
Ic
=0:95 kJ=m
2
, G
IIc
=0:92 kJ=m
2
, and one for 7781/E765 (Gl/
Ep) laminate, G
Ic
=0:74 kJ=m
2
, G
IIc
=0:55 kJ=m
2
, even though
the specimens analyzed with T300/E765 (C/Ep) laminate were
bonded by three different adhesives. These values are of the same
order as those of similar materials in the literature [57,58]. In the
following failure analysis, the predicted failure loads were
determined based on either G
Ic
, G
IIc
, or G
Tc
, where G
Tc
=
G
Ic
G
IIc
. To demonstrate the effectiveness of the present model,
the failure loads predicted by the present model were compared with
the failure loads predicted by the previous analytical model [5],
which assumes a thin bondline with uniform adhesive stresses
throughout the adhesive thickness. The same method used to
Fig. 18 Mode II strain energy release rate as a function of bondline
thickness with C/Ep adherends and Hysol EA 9394 adhesive.
Fig. 19 Total strain energy release rate as a function of bondline
thickness with C/Ep adherends and Hysol EA 9394 adhesive.
Fig. 20 Mode I strain energy release rate as a function of bondline
thickness with Gl/Ep adherends and Hysol EA 9394 adhesive.
Fig. 21 Mode II strain energy release rate as a function of bondline
thickness with Gl/Ep adherends and Hysol EA 9394 adhesive.
Fig. 22 T strain energy release rate as a function of bondline thickness
with Gl/Ep adherends and Hysol EA 9394 adhesive.
514 CHADEGANI, YANG, AND SMELTZER
determine the G
Ic
and G
IIc
by the present model was used with the
previous analytical model to obtain the G
Ic
and G
IIc
used for the
failure load prediction by the previous model, where G
Ic
=
0:32 kJ=m
2
and G
IIc
=0:53 kJ=m
2
were used for the T300/E765
(C/Ep) laminate and G
Ic
=0:95 kJ=m
2
and G
IIc
=0:62 kJ=m
2
were
used for the 7781/E765 (Gl/Ep) laminate.
Figures 2325 compare failure loads calculated fromthe predicted
method, based on either G
Ic
, G
IIc
, or G
Tc
, respectively, using the
present model and the previous analytical models [5] for the joints of
C/Ep adherends and Hysol EA 9394 adhesive. Similarly, for joints
with Gl/Ep adherends and Hysol EA 9394 adhesive, Figs. 2628
show comparison of the predicted failure loads calculated based on
either G
Ic
, G
IIc
, or G
Tc
, respectively, using the present and previous
analytical models [5] and experimental data [51]. For specimens
made of C/Ep adherend and other two paste adhesives, PTM&W ES
6292 and MGS A100/B100, Figs. 2934 show comparison of the
experimental failure loads and the predicted failure loads calculated
based on either G
Ic
, G
IIc
, or G
Tc
using the present and previous
analytical models [5].
It can be seen fromthe gures that the predicted failure loads based
on G
Ic
using the present analytical model inspecimens made of either
C/Ep or Gl/Ep adherends correlate very well with the experimental
data [51]. Furthermore, for all the joints made of either C/Ep or Gl/Ep
adherends, the predicted failure loads computed based on G
IIc
using
Fig. 23 Failure load prediction based on critical mode I strain energy
release rate with C/Ep adherends and Hysol EA 9394 adhesive.
Fig. 24 Failure load prediction based on critical mode II strain energy
release rate with C/Ep adherends and Hysol EA 9394 adhesive.
Fig. 25 Failure load prediction based on critical total strain energy
release rate with C/Ep adherends and Hysol EA 9394 adhesive.
Fig. 26 Failure load prediction based on critical mode I strain energy
release rate with Gl/Ep adherends and Hysol EA 9394 adhesive.
Fig. 27 Failure load prediction based on critical mode II strain energy
release rate with Gl/Ep adherends and Hysol EA 9394 adhesive.
CHADEGANI, YANG, AND SMELTZER 515
Fig. 28 Failure load prediction based on critical total strain energy
release rate with Gl/Ep adherends and Hysol EA 9394 adhesive.
Fig. 29 Failure load prediction based on critical mode I strain energy
release rate with C/Ep adherends and PTM&W ES 6292 adhesive.
Fig. 30 Failure load prediction based on critical mode II strain energy
release rate with C/Ep adherends and PTM&W ES 6292 adhesive.
Fig. 31 Failure load prediction based on critical total strain energy
release rate with C/Ep adherends and PTM&W ES 6292 adhesive.
Fig. 32 Failure load prediction based on critical mode I strain energy
release rate with C/Ep adherends and MGS A100/B100 adhesive.
Fig. 33 Failure load prediction based on critical mode II strain energy
release rate with C/Ep adherends and MGS A100/B100 adhesive.
516 CHADEGANI, YANG, AND SMELTZER
the current analytical model correlate fairly well with those from the
test data [51], while the present analytical underestimates the failure
loads for specimens with C/Ep adherend and Hysol EA 9394
adhesive, as shown in Fig. 24. Moreover, the current analytical model
gives satisfactory prediction of failure loads based on G
Tc
in all
specimens made of either C/Ep or Gl/Ep adherends. It should be
noted that, in all cases, the present analytical model predicts the trend
of failure loads vs bondline thickness, where a thicker bondline
results in a weaker joint. Because the previous analytical model
cannot calculate well the strain energy release rates for the joints with
thick bondlines, it does not provide good failure load estimates even
though the failure criterion was backcalculated from the test data.
VII. Conclusions
An analytical method was developed to calculate the strain energy
release rate and failure load specimen due to delamination of the
surface ply of the adherend adjacent to the adhesive. The stress and
displacement elds of the adhesively bonded single-lap composite
joint were determined based on the laminated orthotropic plate
theory. The crack closure method was applied effectively in conjunc-
tion with the analytical stress and displacement models in deter-
mining the strain energy release rate. T300/E765 3KPWplain-weave
carbon/epoxy (C/Ep) and 7781/E765 satin-weave glass/epoxy (Gl/
Ep) with a ply thickness of 0.25 mm and a quasi-isotropic layup
sequence of [60= 60=0= 60=60=0] were used as adherends to
evaluate and validate the application of the developed model.
Strain energy release rates obtained from the current analytical
method, the previous model, and the nite element models using the
VCCTand J integral were compared for joints with C/Ep and Gl/Ep
adherends and different paste adhesives. Results from the developed
model correlate well with the nite element models and show great
improvement comparing to the previous model for the cases with
thick bondlines.
Even though the purpose of this investigation is to derive an
analytical model to calculate the strain energy release rates of an
adhesively bonded composite joint due to a delamination crack at the
surface ply of the adherend, an attempt was made to predict the joint
strength using the limited available data. Therefore, use of critical
strain energy release rate as the criterion to predict failure of the joint
was investigated using joints reportedly failed under the same
delamination mode of the surface ply of the adherend. Test data from
the literature were employed to validate the approach. Because of the
fact that there were not available data on the fracture toughness for the
two composite laminates studied here, the required G
Ic
and G
IIc
were
backcalculated from the test data. It was shown that the failure loads
predicted based on either G
Ic
, G
IIc
, or G
Tc
using the present analytical
model correlate fairly well with those from the test data. Despite the
fact that the failure criterion used was backcalculated and averaged
based the experimental data, the previous model does not show as
good a correlation with the experimental data as the present model
does. It can be concluded that the current analytical model is able to
predict the joint-failure loads reasonably well using either G
Ic
or G
Tc
as the failure criterion.
It should be noted that although nite element method is capable of
solving problems with various types of materials and complicated
geometrical congurations, analytical methods offer advantageous
performance and solutions, especially with parametric analyses and
optimization. Moreover, it should be emphasized that presenting a
simple and robust analytical model was the motivation for the current
study.
Appendix: Boundary Conditions
Region 1
u
o
1
[
x
1
=0
=0 (A1)
w
1
[
x
1
=0
=0 (A2)
M
1y
[
x
1
=0
=0 (A3)
N
2x
[
x
1
=0
=
A
U
11
A
U
11
A
L
11
P (A4)
M
2y
[
x
1
=0
=0 (A5)
Q
2z
[
x
1
=0
=0 (A6)
N
2x
[
x
1
=L
1
=0 (A7)
M
2y
[
x
1
=L
1
=0 (A8)
Q
2z
[
x
1
=L
1
=0 (A9)
Region 3
u
o
3
[
x
2
=L
2
=u
o
4
[
x
3
=0
(A10)
w
3
[
x
2
=L
2
=w
4
[
x
3
=0
(A11)

3
[
x
2
=L
2
=
4
[
x
3
=0
(A12)
N
3x
[
x
2
=L
2
=N
4x
[
x
3
=0
(A13)
M
3y
[
x
2
=L
2
=M
4y
[
x
3
=0
(A14)
Q
3z
[
x
2
=L
2
=Q
4z
[
x
3
=0
(A15)
N
5x
[
x
3
=0
=0 (A16)
Fig. 34 Failure load prediction based on critical total strain energy
release rate with C/Ep adherends and MGS A100/B100 adhesive.
CHADEGANI, YANG, AND SMELTZER 517
M
5y
[
x
3
=0
=0 (A17)
Q
5z
[
x
3
=0
=0 (A18)
N
6x
[
x
3
=0
=0 (A19)
M
6y
[
x
3
=0
=0 (A20)
Q
6z
[
x
3
=0
=0 (A21)
u
o
4
[
x
3
=L
3
=u
o
7
[
x
4
=0
(A22)
w
4
[
x
3
=L
3
=w
7
[
x
4
=0
(A23)

4
[
x
3
=L
3
=
7
[
x
4
=0
(A24)
N
4x
[
x
3
=L
3
=N
7x
[
x
4
=0
(A25)
M
4y
[
x
3
=L
3
=M
7y
[
x
4
=0
(A26)
Q
4z
[
x
3
=L
3
=Q
7z
[
x
4
=0
(A27)
u
o
5
[
x
3
=L
3
=u
o
8
[
x
4
=0
(A28)
w
5
[
x
3
=L
3
=w
8
[
x
4
=0
(A29)
N
5x
[
x
3
=L
3
=N
8x
[
x
4
=0
(A30)
Q
5z
[
x
3
=L
3
=Q
8z
[
x
4
=0
(A31)
u
o
6
[
x
3
=L
3
=u
o
9
[
x
4
=0
(A32)
w
6
[
x
3
=L
3
=w
9
[
x
4
=0
(A33)

6
[
x
3
=L
3
=
9
[
x
4
=0
(A34)
N
6x
[
x
3
=L
3
=N
9x
[
x
4
=0
(A35)
M
6y
[
x
2
=L
3
=M
9y
[
x
4
=0
(A36)
Region 4
N
7x
[
x
4
=L
4
=0 (A37)
M
7y
[
x
4
=L
4
=0 (A38)
Q
7z
[
x
4
=L
4
=0 (A39)
N
8x
[
x
4
=L
4
=0 (A40)
Q
8z
[
x
4
=L
4
=0 (A41)
u
o
9
[
x
4
=L
4
=u
o
10
[
x
5
=0
(A42)
w
9
[
x
4
=L
4
=w
10
[
x
5
=0
(A43)

9
[
x
4
=L
4
=
10
[
x
5
=0
(A44)
N
9x
[
x
4
=L
4
=N
10x
[
x
5
=0
(A45)
M
9y
[
x
4
=L
4
=M
10y
[
x
5
=0
(A46)
Q
9z
[
x
4
=L
4
=Q
10z
[
x
5
=0
(A47)
Region 5
u
o
10
[
x
5
=L
5
=u
o
12
[
x
6
=0
(A48)
w
10
[
x
5
=L
5
=w
12
[
x
6
=0
(A49)

10
[
x
5
=L
5
=
12
[
x
6
=0
(A50)
N
10x
[
x
5
=L
5
=N
12x
[
x
6
=0
(A51)
M
10y
[
x
5
=L
5
=M
12y
[
x
6
=0
(A52)
Q
10z
[
x
5
=L
5
=Q
12z
[
x
6
=0
(A53)
Region 6
N
11x
[
x
6
=0
=0 (A54)
M
11y
[
x
6
=0
=0 (A55)
Q
11z
[
x
6
=0
=0 (A56)
518 CHADEGANI, YANG, AND SMELTZER
N
11x
[
x
6
=L
6
=
A
U
11
A
U
11
A
L
11
P (A57)
M
11y
[
x
6
=L
6
=0 (A58)
Q
11z
[
x
6
=L
6
=0 (A59)
N
12x
[
x
6
=L
6
=
A
L
11
A
U
11
A
L
11
P (A60)
M
12y
[
x
6
=L
6
=0 (A61)
Q
12z
[
x
6
=L
6
=0 (A62)
Acknowledgments
This investigation was partially sponsored by Kansas NASA
Experimental Program to Stimulate Competitive Research grant
no. NNX07A027A. Support of the National Science Foundation
under grant nos. EIA-0216178 and EPS-0236913, matching support
from the State of Kansas, and the Wichita State University High
Performance Computing Center is also acknowledged.
References
[1] Armanios, E. A., Analysis of Delamination Related Fracture Processes
in Composites, NASA CR-184603, 1988.
[2] Tong, L., Zhu, X. C., and Steven, G. P., Damage Tolerance of
Adhesively Bonded Composite Single Lap Joints, Key Engineering
Materials, Vols. 145149, 1998, pp. 537542.
doi:10.4028/www.scientic.net/KEM.145-149.537
[3] Tong, L., Failure of Adhesive-Bonded Composite Single Lap Joints
with Embedded Crack, AIAA Journal, Vol. 36, No. 3, 1998, pp. 448
456.
doi:10.2514/2.385
[4] Williams, J. G., On the Calculation of Energy Release Rate for
Cracked Laminates, International Journal of Fracture, Vol. 36, 1988,
pp. 101109.
doi:10.1007/BF00017790
[5] Chadegani, A., Yang, C., and Dan-Jumbo, E., Strain-Energy Release
Rate Analysis of AdhesiveBonded Composite Joints with a Prescribed
Interlaminar Crack, Journal of Aircraft, Vol. 46, No. 1, 2009, pp. 203
215.
doi:10.2514/1.37513
[6] Yang, C., Tomblin, J. S., and Guan, Z., Analytical Modeling of ASTM
Lap Shear Adhesive Specimens, Federal Aviation Administration,
U.S. Department of Transportation, Rept. DOT/FAA/AR02/130,
2003.
[7] Goland, M., and Reissner, E. The Stresses in Cemented Joints,
Journal of Applied Mechanics, Vol. 66, 1944, pp. A17A27.
[8] Erdogan, F., and Ratwani, M. M., Stress Distribution in Bonded
Joints, Journal of Composite Materials, Vol. 5, 1971, pp. 378393.
doi:10.1177/002199837100500308
[9] Hart-Smith, L. J., Adhesive-Bonded Single-Lap Joints, NASA CR-
1973-112236, 1973.
[10] Adams, R. D., and Peppiatt, N. A., Stress Analysis of Adhesively
Bonded Lap Joints, Journal of Strain Analysis for Engineering Design,
Vol. 9, 1974, pp. 185196.
doi:10.1243/03093247V093185
[11] Kutscha, D., Mechanics of Adhesive-Bonded Lap-Type Joints: Survey
and Review, U.S. Air Force Materials Lab Technical Report
No. AFML-TDR-64-298, Wright-Patterson AFB, 1964.
[12] Kutscha, D., and Hofer, K. E., Jr., Feasibility of Joining Advanced
Composite Flight Vehicles, U.S. Air Force, Technical Report
No. AFML-TR-68-391, 1969.
[13] Matthews, F. L., Kilty, P. F., and Goodwin, E. W., A Review of the
Strength of Joints in Fiber-Reinforced Plastics: Part 2. Adhesively
Bonded Joints, Composites, Vol. 13, No. 1, 1982, pp. 2937.
doi:10.1016/0010-4361(82)90168-9
[14] Vinson, J. R., Adhesive Bonding of Polymer Composites, Polymer
Engineering and Science, Vol. 29, No. 19, 1989, pp. 13251331.
doi:10.1002/pen.760291904
[15] da Silva, L. F. M., das Neves, P. J. C., Adams, R. D., and Spelt, J. K.,
Analytical Models of Adhesively Bonded Joints-Part I: Literature
Survey, International Journal of Adhesion and Adhesives, Vol. 29,
2009, pp. 319330.
doi:10.1016/j.ijadhadh.2008.06.005
[16] Zhao, X., Adams, R. D., and da Silva, L. F. M., ANewMethod for The
Determination of Bending Moments in Single Lap Joints, Interna-
tional Journal of Adhesion and Adhesives, Vol. 30, 2010, pp. 6371.
doi:10.1016/j.ijadhadh.2009.09.001
[17] Tsai, M. Y., and Morton, J., An Evaluation of Analytical and
Numerical Solutions to the Single-Lap Joint, International Journal of
Solids and Structures, Vol. 31, 1994, pp. 25372563.
doi:10.1016/0020-7683(94)90036-1
[18] Yang, C., and Pang, S. S., Stress-Strain Analysis of Single-Lap
Composite Joints under Tension, Journal of Engineering Materials
and Technology, Vol. 118, No. 2, 1996, pp. 247255.
doi:10.1115/1.2804896
[19] Huang, H., Yang, C., Tomblin, J. S., and Harter, P., Stress and Failure
Analyses of AdhesiveBonded Composite Joints Using ASTM D3165
Specimens, Journal of Composites Technology and Research, Vol. 24,
No. 2, 2002, pp. 345356.
doi:10.1520/CTR10563J
[20] Yang, C., Huang, H., Tomblin, J. S., and Sun, W., Elastic-Plastic
Model of Adhesive-Bonded Single-Lap Composite Joints, Journal of
Composite Materials, Vol. 38, No. 4, 2004, pp. 293309.
doi:10.1177/0021998304039268
[21] Krueger, R., The Virtual Crack Closure Technique: History, Approach
and Applications, NASA Technical Report No. CR-2002-211628,
ICASE Report No. 2002-10, 2002.
[22] Davidson, B. D., Yu, L., and Hu, H., Delamination of Energy Release
Rate and Mode Mix in ThreeDimensional Layered Structures Using
Plate Theory, International Journal of Fracture, Vol. 105, No. 1, 2000,
pp. 81104.
doi:10.1023/A:1007672131026
[23] Park, O., and Sankar, B. V., Crack-Tip Force Method for Computing
Energy Release Rate in Delaminated Plates, Composite Structures,
Vol. 55, No. 4, 2002, pp. 429434.
doi:10.1016/S0263-8223(01)00170-2
[24] Kim, I.-G., and Kong, C.-D., Generalized Theoretical Analysis
Method for Free-Edge Delaminations in Composite Laminates,
Journal of Materials Science, Vol. 37, No. 9, 2002, pp. 18751880.
doi:10.1023/A:1014903518913
[25] Davidson, B. D., Analytical Determination of Mixed-Mode Energy
Release Rates for Delamination Using a Crack Tip Element, Key
Engineering Materials: Fracture of Composites, Vol. 120121, 1996,
pp. 161180.
[26] Yu, L., and Davidson, B. D., AThree-Dimensional Crack Tip Element
for Energy Release Rate Determination in Layered Elastic Structures,
Journal of Composite Materials, Vol. 35, 2001, pp. 457488.
doi:10.1177/002199801772662118
[27] Rybicki, E. F., and Kanninen, M. F., A Finite Element Calculation of
Stress Intensity Factors by a Modied Crack Closure Integral,
Engineering Fracture Mechanics, Vol. 9, No. 4, 1977, pp. 931938.
doi:10.1016/0013-7944(77)90013-3
[28] Raju, I. S., Calculation of Strain-Energy Release Rates With Higher
Order and Singular Finite Elements, Engineering Fracture Mechanics,
Vol. 28, No. 3, 1987, pp. 251274.
doi:10.1016/0013-7944(87)90220-7
[29] Wang, J. T., Xue, D. Y., Sleight, D. W., and Housner, J. M.,
Computation of Energy Release Rate for Cracked Composite Panels
with Nonlinear Deformation, 36th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics and Materials Conference, Part 4
AIAA Paper 95-1463, 1995, pp. 27132727.
[30] Wei, Y., Yang, T., Wan, Z., and Du, X., New VCCT and Its
Application in Composite Delamination Analysis, Chinese Journal
of Computational Mechanics, Vol. 17, No. 3, 2000, pp. 308312 (in
Chinese).
[31] Murthy, P. L. N., and Chamis, C. C., Composite Interlaminar Fracture
Toughness: ThreeDimensional FiniteElement Modeling for Mixed
Mode I, II, and Fracture, Composite Materials: Testing and Design,
American Society for Testing and Materials, ASTM STP 972,
Philadelphia, 1988, pp. 2340.
[32] Fernlund, G., Papini, M., McCammond, D., and Spelt, J. K., Fracture
Load Prediction for Adhesive Joints, Composites Science and
Technology, Vol. 51, 1994, pp. 587600.
doi:10.1016/0266-3538(94)90091-4
CHADEGANI, YANG, AND SMELTZER 519
[33] Yang, C., Sun, W., Tomblin, J. S., and Smeltzer, S. S., III, A Semi-
Analytical Method for Determining the Strain Energy Release Rate of
Cracks in Adhesively Bonded Single-Lap Composite Joints, Journal
of Composite Materials, Vol. 41, No. 13, 2007, pp. 15791602.
doi:10.1177/0021998306069872
[34] Yang, C., Chadegani, A., and Tomblin, J. S., Strain Energy Release
Rate Determination of Prescribed Cracks in Adhesively Bonded Single-
Lap Composite Joints with Thick Bondlines, Composites, Part B,
Vol. 39, No. 5, 2008, pp. 863873.
doi:10.1016/j.compositesb.2007.10.001
[35] ABAQUS/CAE User Manual and Package, Version 6.8-1, Dassault
Systmes Simulia, Providence, RI, 2008.
[36] Armanios, E. A., and Reheld, L. W., Sublaminate Analysis of
Interlaminar Fracture in Composites: Part I-Analytical Model, Journal
of Composites Technology and Research, Vol. 11, No. 4, 1989, pp. 135
146.
doi:10.1520/CTR10166J
[37] Armanios, E. A., and Reheld, L. W., Interlaminar Analysis of
Laminated Composites Using a Sublaminate Approach, 27th AIAA/
ASME/ASCE/AHS Structures, Structural Dynamics, and Materials
Conference, AIAA Paper 86-0969-CP, Part 1, San Antonio, TX, 1986,
pp. 442452.
[38] Armanios, E. A., and Reheld, L. W., ASimplied Approach to Strain
Energy Release Computations for Interlaminar Fracture of Compo-
sites, Proceedings of the FourthJapanU.S. Conference onComposite
Materials, Technomic, Washington, D.C., 1988, pp. 285296.
[39] Reheld, L. W., Armanios, E. A., and Weinstein, F., Analytical
Modeling of Interlaminar Fracture in Laminated Composites,
Composites 86: Recent Advances in Japan and the United States,
Proceedings of the Third JapanU.S. Conference on Composite
Materials, edited by K. Kawata, S. Umekawa, and A. Kobayashi, Japan
Society for Composite Materials, Tokyo, 1986, pp. 331340.
[40] da Silva, L. F. M., das Neves, P. J. C., Adams, R. D., and Spelt, J. K.,
Analytical Models of Adhesively Bonded JointsPart II: Com-
parative Study, International Journal of Adhesion and Adhesives,
Vol. 29, 2009, pp. 331341.
doi:10.1016/j.ijadhadh.2008.06.007
[41] Strength Properties of Adhesives in Shear by Tension Loading of
Single-Lap-Joint Laminated Assemblies, Annual Book of ASTM
Standards, 2nd ed., Vol. 15.06, Std. ASTM D3165, ASTM
International, W. Conshohocken, PA, 2004.
[42] Maple, Software Package, Ver. 9.52, Waterloo Maple, Ontario, 2005.
[43] Irwin, G. R., Fracture I, Handbuch der Physik, Vol. VI, edited by
S. Flugge, Springer, Berlin, 1958, pp. 558590.
[44] Sun, C. T., and Jih, C. J., On Strain EnergyRelease Rates for Interfacial
Cracks in Bi-Material, Engineering Fracture Mechanics, Vol. 28,
No. 1, 1987, pp. 1320.
doi:10.1016/0013-7944(87)90115-9
[45] Smelser, R. E., and Gurtin, M. E., On the J-Integral for Bi-Material
Bodies, International Journal of Fracture, Vol. 13, 1977, pp. 382384.
[46] Chou, S. J., Note on the Path-Independent Integral for Bi-Material
Bodies, International Journal of Fracture, Vol. 45, 1990, pp. R49
R53.
doi:10.1007/BF00036279
[47] Anderson, T. L., Fracture Mechanics: Fundamentals and Applications,
3rd ed., CRC Press, Boca Raton, FL, 2005.
[48] Barsoum, R. S., On the Use of Iso-Parametric Finite Elements in
Linear Fracture Mechanics, International Journal for Numerical
Methods in Engineering, Vol. 10, No. 1, 1976, pp. 2537.
doi:10.1002/nme.1620100103
[49] Henshell, R. D., and Shaw, K. G., Crack Tip Finite Elements are
Unnecessary, International Journal for Numerical Methods in
Engineering, Vol. 9, No. 3, 1975, pp. 495507.
doi:10.1002/nme.1620090302
[50] Raju, I. S., Crews, J. H., Jr., and Aminpour, M. A., Convergence of
Strain Energy Release Rate Components for Edge-Delaminated
Composite Laminates, Engineering Fracture Mechanics, Vol. 30,
No. 3, 1988, pp. 383396.
doi:10.1016/0013-7944(88)90196-8
[51] Tomblin, J. S., Yang, C., and Harter, P., Investigation of Thick
Bondline Adhesive Joints, Federal Aviation Administration DOT/
FAA/AR-01/33, U.S. Department of Transportation, 2001.
[52] ASTMStandard D5528, Test Method for mode I Interlaminar Fracture
Toughness of Unidirectional Fiber-Reinforced Polymer Matrix
Composites, Annual Book of ASTM Standards, Vol. 15.03, ASTM
International, W. Conshohocken, Philadelphia, 2004.
[53] Russell, A. J., On the Measurement of mode II Interlaminar Fracture
Energies, Defense Research Establishment Pacic Materials Report
No. 82-0, Victoria, BC, Canada, 1982.
[54] Davidson, B. D., and Sun, X., Effects of Friction, Geometry, and
Fixture Compliance on the Perceived Toughness fromThree- and Four-
Point Bend End-Notched Flexure Tests, Journal of Reinforced Plastic
and Composite, Vol. 24, No. 15, 2005, pp. 16111628.
doi:10.1177/0731684405050402
[55] Test Method for Mixed mode Imode II Interlaminar Fracture
Toughness of Unidirectional Fiber Reinforced Polymer Matrix
Composites, Annual Book of ASTM Standards, Vol. 15.03, ASTM
International, Std. ASTM D6671, W. Conshohocken, PA, 2004.
[56] Reeder, J. R., A Bilinear Failure Criterion for Mixed-Mode
Delamination, Composite Materials: Testing and Design, Vol. 11,
ASTM International, ASTM STP 1206, West Conshohocken, PA,
1993, pp. 303322.
[57] Szekrnyes, A., Delamination of Composite Specimens, Ph.D.
Dissertation, Department of Applied Mechanics, Budapest University
of Technology and Economics, Budapest, 2005.
[58] Kerien, N., iliukas, A., and Kerys, A., Inuence of Ply Orientation
on Mode I Interlaminar Fracture Toughness of Woven Carbon and Glass
Composites, Mechanika, Vol. 82, No. 2, 2010, pp. 3136.
520 CHADEGANI, YANG, AND SMELTZER