Professional Documents
Culture Documents
composite materials
Ahmad K. Hussain
1
, Donald F. Adams*
,2
Composite Materials Research Group, University of Wyoming, Laramie, WY 82071, USA
Received 16 November 2002; received in revised form 4 June 2003; accepted 5 June 2003
Abstract
Linear three-dimensional nite element analyses were conducted to evaluate various rail and specimen congurations of the two-
rail shear test procedure. Additional analytical evaluations were performed to investigate the eects of specimen aspect ratio,
material orientation, loading direction, and the rigidity of the rails on stress distributions. Shear stresses were found to be dis-
tributed uniformly over a large region in the specimen gage section for most of the models analyzed, but normally accompanied by
signicant transverse and axial normal stresses. The results indicated that an axial loading applied to a test xture of uniform steel
rails and a test specimen with aspect ratio of 6 was the preferred test procedure. Simple test specimens of trapezoidal and rect-
angular geometries, and specimens with tabs, were chosen for further evaluation. Also, the new Wyoming-Modied Two-Rail Shear
test xture was established based on these analyses.
#2003 Elsevier Ltd. All rights reserved.
Keywords: A. Polymer matrix composites; C. Finite element analysis; B. Mechanical properties; C. Stress concentrations; Shear test methods;
B. Modelling
1. Introduction
The nite element method is a powerful technique for
the numerical solution of a variety of engineering pro-
blems, and has frequently been used in the evaluation of
test methods. The nite element method is capable of
quickly and eciently predicting the response of a vari-
ety of test specimen congurations and material orien-
tations. In contrast, the fabrication and experimental
evaluation of such test specimens is expensive and time
consuming, and hence less practical, during an initial
evaluation of candidate congurations. Apart from
considerable savings of cost and time, nite element
analyses require only minimal amounts of input data,
and the resulting stress and strain distributions can be
determined throughout the entire specimen.
In the present study, linear three-dimensional nite
element analyses were conducted to predict the perfor-
mance of the two-rail shear test method. The eects of
various rail and test specimen congurations, material
orientations, and loading directions on stress distribu-
tions were evaluated. These analytical results were then
used to select preferred specimen congurations and to
obtain design criteria for the new Wyoming-Modied
Two-Rail Shear test xture, a proposed alternative to
the ASTM Standard D 4255 (Method A) Two-Rail
Shear test procedure [1]. The present study was part of a
larger research program to re-evaluate the present
ASTM Standard D 4255 two-rail shear test procedure.
This existing standard incorporates features that are not
adequately understood, as summarized in Ref. [2]. These
include specimen aspect ratio, the use of diagonal load-
ing, and the o-center location of the clamping bolts.
In the ASTM standard test method, a 76152 mm
(36 in) rectangular test specimen is sandwiched
between two pairs of rails by bolting through six
oversized holes in the specimen. An o-axis (diagonal)
tensile load is introduced at the ends of the rails. A
shearing load is thus transmitted into the test specimen
via the rails, and the shear strain response can be
measured by strain gages mounted in the gage section.
A slightly dierent version of this test procedure was
proposed earlier [3]. As outlined by Boller [3], the test
0266-3538/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0266-3538(03)00254-9
Composites Science and Technology 64 (2004) 221238
www.elsevier.com/locate/compscitech
* Corresponding author. Tel.: +1-307-742-8641; fax: +1-307-742-
8682.
E-mail address: wtf@wyomingtestxtures.com (D.F. Adams).
1
Currently Instructor, the MARA Institute of Technology (ITM),
Malasia.
2
Emeritus Professor; President, Wyoming Test Fixtures, Inc.,
Laramie, WY, USA.
specimen was 76 mm (3 in) wide and 152 mm (6 in) long
with 12.7 mm (0.5 in) wide unconstrained central (gage)
region. At free edges of the gage region, 12.7-mm (0.5-
in) radius curves were cut. Specimens with this geometry
provided satisfactory results with greater uniformity in
stress distribution than a rectangular specimen [3]. Two
pairs of rail guides were attached to the rails for lateral
support. A diagonal compressive load was applied to a
ball and socket or spherical seat at each end of the rails.
The reason why a diagonal load was applied was not
given but the self-aligning spherical seat ensured that
each rail transmitted equal proportions of the total
load.
There are several other rail shear versions that have
been introduced as alternatives. Rather than the con-
ventional bolted-uniform-rail, a bonded-tapered-rail has
been used [4,5]. Rail shear test xtures with low friction
rail guides, roller spacers, or knife-edge spacers have
been also cited [2,68]. These mechanisms, which are
normally utilized with the compression loaded two-rail
xture, can increase buckling stability and also improve
alignment [3].
Conventional at specimens of rectangular shape
have been widely used. Specimens with 12.7 mm (0.50
in) radius llets at the edges [3] have been designed to
minimize end eects. A trapezoidal-shaped specimen
has also been suggested [5,9,10], to reduce the magni-
tude of the combined stresses in each corner while
maintaining a large region of uniform shear stress in the
middle. Other specimen geometries may also be ana-
lyzed, such as a rectangular specimen with V-notches at
the free edges (analogous to the Iosipescu shear test
specimen [11]), and at-bottom notched specimens
[12], to name a few. Modication of the existing ASTM
standard rectangular specimen by adding bonded
tabs extending into the transition region, may also be
considered.
Several earlier analytical evaluations of the two-rail
shear test method with ber-reinforced composite lami-
nates have been reported [4,5,7,13,14]. Two-dimensional
linear nite element analyses of the two-rail shear test
method were performed by Garcia et al. [4], to assess
the eects of aspect ratio, axial versus diagonal load
introduction, and uniform versus thickness-tapered rail
congurations. Results indicated an inuence of loading
direction and rail conguration. However, aspect ratio
had a major eect on stress distributions, which also
varied with laminate orientation.
A group of investigators at the Virginia Polytechnic
Institute [7,13] compared four shear test methods,
including the two-rail shear method, using the same
analytical techniques. Their nite element results indi-
cated that several factors inuence the stress distribu-
tions in the gage section of the rail shear test specimen.
Uniform shear stress distributions were produced over a
large area in the gage section, but were often
accompanied by signicant in-plane normal stresses
depending on the stiness of the rails, the method of
load application, and the laminate properties [7,13].
Black and Hart-Smith [5] presented their linear nite
element analysis results for the bonded tapered rail spe-
cimen with [ 45] angle-ply and [0/ 45/90]
2S
quasi-iso-
tropic laminates. They approximated a three-
dimensional eect in the three layers (rail, adhesive, and
laminate) using two-dimensional elements, emphasizing
the nonlinear behavior of the adhesive. From their ana-
lysis, they predicted that the existence of stress con-
centrations in the bonded tapered rail specimen would
not be sucient to cause premature failures. However
an earlier analytical investigation [15] of a uniform
thickness two-rail shear xture with [ 45]
ns
angle-ply
laminates found that the shear stress distribution was
very non-uniform and highly irregular, which would
lead to unacceptably low values of experimental shear
strength [15]. It is apparent from these works [5,15] that
shear testing of laminates such as [ 45]
ns
angle-ply and
[0/ 45/90]
ns
quasi-isotropic lay-ups are considered to
be dicult. This diculty is generally associated with
inadequate gripping, as well as a highly nonuniform
stress distribution in the test area. In addition, sub-
stantial local stress concentrations are normally present
and can cause premature failures in specimens of these
material orientations.
Chang and Chen [14] used a three-dimensional analy-
sis with an eight-node isoparametric element to study
the eects of ply orientation on the in-plane shear
strength distribution in laminates. Laminates with var-
ious cross-ply orientations were analyzed and a strong
inuence of the interlaminar stress in lowering the
laminate strength relative to lay-up sequence was
observed.
The denition of the material direction of a unidirec-
tional composite shear specimen, particularly the rail
shear specimen, has been inconsistent in the literature.
For composites in general, the denition mainly
depends on the type of test method, direction of load
introduction, and how the test xture is oriented, e.g.,
the rails in the case of the rail shear test. Some prior
investigators considered the [0] orientation of the rail
shear test specimen to be in the loading direction (par-
allel to the rails), but then this is unlike the Iosipescu
shear specimen where the 0
direction is perpendicular
to the loading direction. Others simply assigned the [0]
and [90] rail shear orientations corresponding to con-
ventional x and y Cartesian coordinates, horizontally
and vertically, respectively. These mixed denitions
sometimes cause confusion, especially when compara-
tive studies between dierent test methods are carried
out.
In the present work, the denition of the rail shear
specimen material orientation will follow the convention
of the Iosipescu shear test method [11]. Thus, a 0
222 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238
unidirectional composite has bers oriented in the x
direction, perpendicular to the rails, and a 90
unidirec-
tional composite has bers oriented in the y direction,
parallel to the rails.
2. Finite element models
Models of 13 basic two-rail shear specimen geometries
and a number of variations (varying the aspect ratio or
notch radius of the specimen, altering the loading
direction, varying the rail thickness, adding a tab mate-
rial, or changing completely the materials of the rails or
the specimen) were generated using MSC/PATRAN
software [16], and analyzed using a previously-devel-
oped nite element package known as WYO3D [17].
These models are listed in Table 1. The basic geometries
are sketched in Fig. 1. Complete model details are given
in Ref. [10].
Two loading directions, axial and o-axis (diagonal)
tension, were simulated. Two composite laminate
orientations, [0]
16
unidirectional, and [90]
16
unidirec-
tional were studied. In addition, a [0
4
/90
4
]
S
cross-ply
laminate was modeled as a special case. Steel was
employed for the rails and carbon/epoxy was the mate-
rial utilized for the test specimen in most of the cases
analyzed.
The steel rails and the composite specimen were
modeled as three-dimensional solids, such as shown in
Fig. 2. For the models incorporating tabs, an additional
solid layer was used. Each individual solid layer repre-
senting the rails, the unidirectional composite specimen,
and the tabs was modeled as a single layer of elements.
For the cross-ply specimen, two layers of elements were
generated; each layer represented one of the orthogonal
material orientations. For simplicity, it was assumed
that the solids were perfectly bonded to each other.
Furthermore, the symmetry of the rail shear specimen
Table 1
Summary of cases for the nite element analysis
Model Geometry Laminate orientation Loading type Angle
a
(
) Aspect ratio
b
R
1. Rectangular:
(a) Rectangular A [0],[90] Axial, diag. 90 6.0/0.75=8
(b) Rectangular B [0],[90] Axial, diag. 90 6.0/0.5=12
2. Rectangular with 0.5 in llets [0],[90] Axial 5.13/0.75=6.85
3. Rectangular with V-notches [0],[90] Axial 45 5.25/0.75=7
4. Rectangular with rounded notches [0],[90] Axial 45 5.47/0.75=7.3
5. Rectangular A with tabs on two sides [0],[90] Axial 90 6.0/0.75=8
6. Rectangular A with tabs on four sides [0],[90] Axial 90 6.0/0.75=8
7. Flat-bottomed notches:
(a) FBN A [0],[90] Axial 30 3.375/0.75=4.5
(b) FBN B [0],[90] Axial 45 4.5/1.0=4.5
8. Flat-bottomed notches with radii:
(a) FBN Radius A (r=0.265 in) [0],[90] Axial 45 6.11/0.75=8.15
(b) FBN Radius B (r=0.354 in) [0],[90] Axial 45 4.65/1.0=4.65
9. Flat-bottomed notches extended
to the rails (FBN Extended)
[0],[90] Axial 45 4.65/1.0=4.65
10. Trapezoidal:
(a) Trapezoidal A [0],[90] Axial, diag. 45 4.5/0.75=6
(b) Trapezoidal B [0/90] Axial 45 4.5/0.75=6
(c) Trapezoidal C [0],[90] Axial 45 4.5/1.0=4.5
(d) Trapezoidal D [0],[90] Axial 45 4.5/0.5=9
(e) Trapezoidal E [0],[90] Axial 60 4.5/0.5=9
(f) Trapezoidal F [0],[90] Axial 30 4.5/0.5=9
(g) Trapezoidal G [0],[90] Axial 30 4.5/0.75=6
(h) Trapezoidal H [0],[90] Axial 30 6.0/0.75=8
(i) Trapezoidal A with titanium rails [0],[90] Axial 45 4.5/0.75=6
(j) Trapezoidal A with rigid rails [0],[90] Axial 45 4.5/0.75=6
(k) Trapezoidal A with glass/epoxy specimen [0],[90] Axial 45 4.5/0.75=6
11. Trapezoidal with thickness-tapered rails [0],[90] Axial 45 4.5/0.5=9
12. Trapezoidal with width-tapered rails [0],[90] Axial 45 4.5/1.5=3
13. Trapezoidal thickness ratios:
(a) Trapezoidal (2:1) [90] Axial 45 4.5/1.0=4.5
(b) Trapezoidal (3:1) [90] Axial 45 4.5/1.0=4.5
(c) Trapezoidal (13:1) [90] Axial 45 4.5/1.0=4.5
a
Angle of the specimen free edges (at top and bottom) measured from vertical.
b
R is the specimen aspect ratio (length of the plate along the gage section/width of the plate between the rails).
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 223
was exploited; hence, only one half was modeled. The
mid-thickness of the specimen was a plane of symmetry.
Unless otherwise specied, the thickness of the uni-
form rails and one-half the thickness of the composite
specimen were 12.7 mm (0.50 in) and 1.02 mm (0.040
in), respectively, noting that the unidirectional compo-
site specimen was actually modeled as one-half of a 16-
ply laminate, assuming each ply to have a thickness of
0.13 mm (0.005 in). Similarly, as a special case, a [0
4
/
90
4
]
S
cross-ply laminate, having a total of 16 plies, was
modeled as a four-ply laminate due to the symmetry
about the mid-thickness. The thickness of the tab layer
in the models was 1.50 mm (0.060 in). To model the
thickness-tapered rails, the rail thickness was varied
from 12.7 mm (0.50 in) to 3.20 mm (0.125 in), and for
the width-tapered rails, the width was varied from 25.4
mm (1.00 in) to 6.4 mm (0.25 in). For the trapezoidal
thickness ratio models (model geometry #13), the steel
rails were replaced by specimen material. As an exam-
ple, for trapezoidal (2:1), the ratio 2:1 was based on the
thickness of the rails (0.080 in) relative to the (con-
stant) thickness of the test specimen (0.040 in).
Fig. 1. FEA model geometries.
Fig. 2. Typical nite element grids of (a) the model geometry #1 rec-
tangular A (plan view) and (b) the model geometry #10, trapezoidal A
(isometric view).
224 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238
In MSC/PATRAN, after the solids were generated,
they were meshed using HEX-8, an eight-node isopara-
metric element. Finer meshes were formed in the regions
between the rails and the specimen and also in the
regions of the specimen free ends where high gradients
of stresses were expected. In WYO3D, boundary and
loading conditions were imposed on each model. To
simulate the actual boundary and loading conditions of
the two-rail shear test, the nodes in the plane of
through-thickness symmetry of the model were con-
strained in the z direction. In addition, for the axial
loading case, nodal forces were applied in the y direc-
tion (parallel to the rails) to a pair of nodes (Nodes A1
and A2, see Fig. 2) while constraining them in the x
direction. For the diagonal loading case (7
o-axis for
the rectangular A and B models, and 5
A
= P=A
12
, the axial stress
1
(the stress component parallel to
the ber direction) and the transverse stress
2
(the
stress component normal to the ber direction) will be
fully presented and discussed here. The out-of-plane
(interlaminar) stress components, i.e.,
13
,
23
, and
3
,
were also available from the WYO3D outputs, but their
Table 2
Material properties of the rails, tabs, and composite laminates used in the present nite element analysis
Property Rails Tabs Unidirectional laminates
Steel Titanium Ti6A14V [4] Aluminum Carbon/epoxy AS4/3501-6 [17] Glass/epoxy S-2/3501-6 [17]
E
11
(GPa) (Msi) 208 (30.0) 114 (16.5) 69.3 (10.0) 142 (20.5) 53.4 (7.74)
E
22
(GPa) (Msi) 208 (30.0) 114 (16.5) 69.3 (10.0) 9.2 (1.33) 16.4 (2.37)
E
33
(GPa) (Msi) 208 (30.0) 114 (16.5) 69.3 (10.0) 9.2 (1.33) 16.4 (2.37)
G
23
(GPa) (Msi) 78.3 (11.3) 42.1 (6.10) 26.1 (3.75) 3.6 (0.52) 5.5 (0.80)
G
13
(GPa) (Msi) 78.3 (11.3) 42.1 (6.10) 26.1 (3.75) 6.1 (0.88) 6.1 (0.88)
G
12
(GPa) (Msi) 78.3 (11.3) 42.1 (6.10) 26.1 (3.75) 6.1 (0.88) 6.1 (0.88)
n
23
0.333 0.342 0.333 0.29 0.32
n
13
0.333 0.342 0.333 0.25 0.27
n
12
0.333 0.342 0.333 0.25 0.27
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 225
magnitudes were small and their presence incon-
sequentially aected the in-plane stress response of the
unidirectional composites.
In general, for all cases analyzed, the in-plane shear
stress was fairly uniformly distributed, and close to the
average applied shear stress, throughout entire test sec-
tion except near the free edges. However, undesirable
axial and transverse normal stresses, both tensile and
compressive, were also present in the gage section, with
the axial stresses being very much higher than the
transverse stresses. The presence of these axial stresses
was not a major concern, however, since the reinforcing
bers were oriented in this direction and thus this com-
ponent of composite strength was high. On the con-
trary, the existence of the transverse stresses can cause a
problem since the transverse strength of unidirectional
composites is typically very much lower than the long-
itudinal strength. Furthermore, a transverse tensile
stress could combine with the shear stress to promote
failure. The existence of a compressive transverse stress
2
() could induce failure by out-of-plane buckling.
The transverse tensile stress was signicantly higher in
[90] specimens than in [0] specimens for the various rec-
tangular models, especially for rectangular A and B.
The transverse tensile strength
2
(+), being the low-
est of all the strengths, [i.e., axial tensile
1
(+) and
compressive
1
() strengths, transverse tensile
2
(+)
and compressive
2
() strengths, and in-plane shear
12
strength], is normally responsible for the so-called rst
ply failure in the specimen. As a comparison, typical
unidirectional composite strength properties for a
widely used carbon/epoxy system [18], with (+) and ()
signs indicating tensile and compressive strengths,
respectively, are:
1
(+)=1890 MPa (274 ksi)
1
()=1930 MPa (280 ksi)
2
(+)=65.5 MPa (9.5 ksi)
s
2
()=269 MPa (39.0 ksi)
12
=119 MPa (17.3 ksi)
Localized stress concentrations were usually predicted
to be present at the following locations; near the top left
and bottom right hand corners (loaded corners), and
the bottom left and top right hand corners (unloaded
corners) of the full gage section, and at the roots of the
notches, if present. These stress concentrations gave an
indication that premature failure might occur at that
particular location rather than in the desired gage sec-
tion region. However, the actual seriousness of these
local eects was dicult to predict since it is possible
that premature localized cracks might give stress relief,
and thus be benecial. This phenomenon has been
observed in 0
orientation for (a) the model geometry #1, rectangular A and (b)
the model geometry #10, trapezoidal A.
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 227
roots. Due to the sharp notches, the rectangular with V-
notches model exhibited the largest magnitudes of the
three stress concentrations, for both the [0] and [90]
orientations, compared to those with the rounded
shapes.
The stress distributions for the rectangular specimens
with tabs on two sides were uniformly distributed over a
smaller region in the test specimen than for the untab-
bed rectangular model. This could be due to the tabs
acting as secondary rails. Hence, the stresses were shif-
ted into the transition regions between the tab layer and
the plate specimen, rather than in the regions between
the rails and the specimen of the untabbed rectangular
model. The stress concentrations in this model geometry
were among the lowest of all the models, however. This
is very favorable, since premature failures caused by
unwanted peak stresses may not take place during test-
ing. Thus, the specimen may achieve its true strength
Fig. 4. Normalized stress distributions along the center of gage section for various model geometries: axially loaded [0] laminate (model comparison).
228 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238
prior to failure. It was probably for this reason that the
tabbed specimens gave consistently high results in the
ASTM round-robin rail shear tests [19].
The trapezoidal A conguration was found to be the
optimum conguration for the two-rail shear specimen
from the present analyses, although the results were not
exceptionally impressive relative to the rectangular
models. By way of comparison, the axially loaded tra-
pezoidal A model provided a more uniform shear stress
distribution across the gage section than any of the
rectangular models. The normalized transverse stress
for the axially loaded [0] and [90] orientations of trape-
zoidal A only varied within 0.10 over a large area
(over more than 50% of the gage section) in the
Fig. 5. Normalized stress distributions along the center of gage section for model geometry #1, rectangular A: axially and diagonally loaded [0] and
[90] laminates (loading comparison).
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 229
specimen center. Although the transverse stress was
higher by a factor of 10 than that of the rectangular
models, it was still too low to cause detrimental eects.
The presence of a tensile transverse stress concen-
tration (1.19) near the unloaded corners for the [0]
orientation, and (1.08) near the loaded corners for the
[90] orientation, could initiate premature cracks. These
cracks would be acceptable if they were in the form of
many microcracks, or could be constrained locally
(without extending into the grip area). Otherwise they
could cause a lower shear strength measurement. The
normalized shear stress concentrations for the axially
loaded [0] and [90] orientations were 2.69 and 1.11,
respectively. Since these peak shear stresses developed in
Fig. 6. Normalized stress distributions along the center of gage section for model geometry #10, trapezoidal A, axially and diagonally loaded [0] and
[90] laminates (loading comparison).
230 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238
the same small area as the highest tensile transverse
stresses, the combined stress eect could speed up the
formation of the cracks.
3.2. Axial versus diagonal loading comparison
Two directions of load introduction were applied to
rectangular A and B model geometries, a 7
o-axis
(diagonal) loading and an axial loading, with both load
lines passing through the center of the specimen. The 7
12
x y
1
(+) x y
2
(+) x y
Rectangular A Axial, [0] 1.25 0.96 0.89 14.81 1.14 1.00 1.62 0.99 1.00
7
, [0] 1.22 0.96 0.79 13.52 1.14 1.00 1.08 0.99 1.00
Axial, [90] 1.38 0.96 0.97 2.67 1.13 0.99 4.28 0.96 1.00
7
, [90] 1.24 0.96 0.94 1.92 1.13 0.99 4.09 0.96 1.00
Rectangular B Axial, [0] 1.31 0.94 0.94 13.73 1.14 1.00 1.45 0.98 1.00
7
, [0] 1.26 0.94 0.83 12.14 1.14 1.00 0.92 0.98 1.00
Rectangular with Axial, [0] 1.31 0.91 0.90 4.77 1.58 0.99 1.21 1.00 1.17
0.5-in llets Axial, [90] 1.21 0.48 1.07 2.48 1.28 1.10 1.27 0.92 1.17
Rectangular with Axial, [0] 2.03 0.00 1.00 7.61 1.50 1.00 1.26 0.71 1.00
V-notches Axial, [90] 1.41 0.96 1.14 2.17 1.13 1.13 1.39 0.71 0.99
Rectangular with Axial, [0] 1.37 0.19 1.00 5.82 1.43 1.00 1.25 1.00 1.10
rounded notches Axial, [90] 1.29 0.91 1.09 2.04 1.13 1.10 1.41 0.51 1.03
Rectangular A with Axial, [0] 1.14 0.43 0.90 4.55 1.26 1.00 0.13 0.56 0.86
tabs on 2 sides Axial, [90] 1.18 0.43 0.86 1.13 1.26 0.71 1.60 0.48 1.00
Flat Bottomed Notch Extended Axial, [0] 2.93 0.96 0.79 2.34 1.96 0.80 1.50 0.92 0.80
Axial, [90] 1.39 0.96 1.21 2.08 1.00 1.18 1.35 0.96 1.21
Trapezoidal A Axial, [0] 2.69 0.94 0.84 2.84 1.73 0.84 1.19 0.87 0.85
5
, [0] 2.65 0.95 0.84 2.54 1.73 0.84 1.16 0.87 0.86
Axial, [90] 1.11 0.87 1.14 1.54 1.00 1.02 1.08 0.79 1.13
5
, [90] 1.19 0.95 1.16 1.50 1.00 1.10 1.20 0.94 1.16
Trapezoidal B Axial, [0] 1.77 0.95 0.84 2.69 1.02 0.97 1.27 1.00 1.17
(Cross-Ply) Axial, [90] 1.24 0.87 0.78 1.89 1.57 0.84 0.43 0.79 0.87
Trapezoidal C Axial, [0] 2.93 0.96 0.79 2.10 1.93 0.81 1.48 0.92 0.80
Axial, [90] 1.46 0.96 1.21 2.30 1.00 1.18 1.41 0.96 1.21
Trapezoidal D Axial, [90] 3.20 0.96 0.90 4.09 1.44 0.90 1.25 0.87 0.91
Axial, [0] 1.36 0.96 1.10 1.69 1.00 1.06 1.33 0.96 1.10
Trapezoidal A with Axial, [0] 2.87 0.95 0.84 4.33 1.73 0.84 1.30 0.87 0.86
Titanium rails Axial, [90] 1.66 0.95 1.16 2.47 1.00 1.13 1.56 0.95 1.16
Trapezoidal A with Axial, [0] 2.55 0.95 0.84 1.91 1.02 0.92 1.10 0.79 0.87
rigid rails Axial, [90] 1.23 0.95 0.56 1.46 1.00 1.17 1.06 0.94 1.16
Trapezoidal A with Axial, [0] 2.21 0.95 0.84 1.59 1.64 0.88 1.23 0.79 0.87
glass/epoxy Axial, [90] 1.39 0.95 0.78 1.39 1.00 1.17 1.10 0.94 1.05
Trapezoidal with Axial, [0] 2.97 0.96 0.90 4.14 1.44 0.90 1.03 1.00 1.10
thickness-tapered rails Axial, [90] 1.42 0.96 1.10 1.78 1.00 1.05 1.44 0.96 1.10
Trapezoidal 2:1 Axial, [90] 7.92 0.96 1.21 22.66 1.00 1.20 6.37 0.96 1.21
Trapezoidal 3:1 Axial, [90] 5.87 0.96 1.21 15.41 1.00 1.20 4.87 0.96 1.21
Trapezoidal 13:1 Axial, [90] 2.12 0.96 1.21 3.57 1.00 1.18 1.97 0.96 1.21
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 235
rigid rails were compared (Fig. 8), where the rigid
rails were modeled as having a modulus twice as high as
steel. It was found that for the [0] and [90] orientations,
the rigid rails provided slightly more uniform shear
stress distributions and the lowest normalized axial and
transverse stresses across the center of the gage section.
Most of the stress concentrations were the lowest for the
rigid rails, while the titanium rails gave the highest stress
concentrations in all cases.
A glass/epoxy unidirectional composite specimen with
[0] and [90] orientations, and a carbon/epoxy [0
4
/90
4
]
S
cross-ply laminate, were analyzed using the trapezoidal
A geometry. In brief, the analytical results indicated
that the trapezoidal specimen should work well with
both glass/epoxy and carbon/epoxy (Figs. 9 and 10),
and with both unidirectional and cross-ply orientations.
The stress distribution in the cross-ply laminate did dif-
fer somewhat from that in the unidirectional laminates.
This could be due to the interply eects between the 0
and 90
uni-
directional carbon/epoxy test specimens. Two addi-
tional analyses were also performed, for a carbon/epoxy
cross-ply laminate and for a unidirectional glass/epoxy
laminate. Although other factors such as stacking
sequence, type of composite (chopped, woven, con-
tinuous ber, etc.), and type of ber and matrix mate-
rial, are expected to have some inuence also, the
following summary of results and conclusions of the
present analysis is expected to be generally applicable.
That is, the present results can be summarized as
follows:
1. The shear stress states were fairly uniform
throughout the test specimen for most of the
model geometries, especially for the trapezoidal
A model, except in small areas near the free edges
and in the transition region close to the rails.
2. Stress concentrations exist at corners and
notch roots, and may cause premature failure.
These stresses are inuenced by geometric
discontinuities.
3. Free edges and the complex stresses at the
interface between the rails and the specimen give
rise to stress concentrations near the corners of
the gage section. However, the dissimilar mate-
rial properties of the specimen and the rails do not.
Although varying the specimen conguration and
rail stiness can minimize these stress concentra-
tions, they have not been totally eliminated.
4. The presence of compressive transverse stresses
may cause out-of-plane buckling failures, and
high tensile transverse stresses may combine with
the shear stresses to cause premature failure. The
severity of these stresses was particularly notice-
able in [90] rectangular specimens, which may
make this material orientation unsuitable for
testing. It is also relatively easy to accidentally
break a [90] specimen while installing it in the
xture.
5. The trapezoidal A model has been chosen as the
optimum test specimen conguration because the
normalized shear stress distribution is uniform
throughout almost 80% of the gage section.
Furthermore, the normalized transverse stress is
quite small in the vicinity of the gage section. The
rectangular A and the rectangular with tabs on
two sides models are worthy of consideration
also. The trapezoidal and the rectangular speci-
mens have an advantage of being easier to fab-
ricate than those with notches, and the tabbed
specimen provides reinforcement in the grip sec-
tion to resist possible crushing failure from the
xture.
6. The angle of the trapezoidal geometry does aect
the stress states. The shear stress distribution is
more uniform for a trapezoid with a smaller
acute angle, but the local stress concentrations
are larger.
7. There may be an optimum aspect ratio for each
type of specimen geometry, which could also
depend on the laminate orientation. Governing
factors in selecting an aspect ratio include the
stress states produced, the accessibility of the
gage section, and the overall size of the test
specimen.
8. O-axis (diagonal) loading very slightly improves
the overall stress states in the rectangular models
and in the trapezoidal A model. However, due to
a lack of detailed study to predict the optimum
loading angle, there is not yet enough justica-
tion for it to be considered as the preferred
loading method.
9. The material stiness and thickness of the rails
inuence the stress states, with more rigid rails
being superior. Tapering the rails does not
favorably inuence the stress distributions or
reduce the stress concentrations.
236 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238
In most of the nite element analysis cases, the rail
components were modeled as matching the rails used in
the standard ASTM D 4255 xture, but without bolts.
The test specimen was also modeled without holes. A
signicant simplication was made by assuming a per-
fect bond between a specimen and the rails, which
explains the absence of the bolts and the holes. In an
actual ASTM D 4255 test however, bolting rather than
bonding is used. Close examination of the present nite
element analyses will reveal that they did in fact closely
represent the actual test condition imposed by the newly
designed Wyoming-modied test xture and its test
specimen, shown in Fig. 11 [2]. Thus, most of the ana-
lytical results should predict the actual response more
accurately for the newly designed xture than the exist-
ing ASTM xture.
Having analyzed various specimen and rail congur-
ations using the nite element analysis, trapezoidal A (a
trapezoidal specimen) was predicted to be the optimum
conguration. Rectangular A (a simple rectangular spe-
cimen) and rectangular with tabs on two sides (a tabbed
rectangular specimen) were also considered favorable
for potential experimental evaluation.
Thus the nite element analysis did serve its purpose,
in predicting the responses of a variety of test specimen
congurations and material orientations, and the eects
of several other parameters. The results also helped the
authors better understand the complexities of the stress
distributions near specimen edges and stress transition
regions. The large region of uniform shear stress
accompanied by insignicant transverse normal stresses
strongly indicated the presence of a pure shear state in
the gage section for almost all of the models. In addi-
tion, by knowing that the shear stress in the center of
the test area was close to the applied shear stress, the
authors became condent that the shear modulus mea-
surements obtained experimentally should be correct,
and correction factors unnecessary.
5. Conclusion
Criteria for use in the design of a new shear test x-
ture were obtained from the detailed nite element ana-
lysis study presented here. A variety of specimen and
rail congurations were analyzed. In addition, several
other parameters that were expected to aect the results
were also studied. Signicant amounts of information
were obtained through this analytical evaluation, and
considerable amounts of time were saved by utilizing
this approach for the initial evaluation of candidate test
specimen geometries.
After several candidate congurations have been
chosen, a nonlinear three-dimensional analysis using
WYO3D [17], or a failure initiation and propagation
analysis using the recently modied WYO2D program
[20], can be performed. As an alternative, experimental
evaluation can now be conducted using only a small
number of dierent congurations. In fact, a concluding
experimental evaluation is often the best approach since
analytical methods do not always accurately predict the
response in an actual test environment.
A large central region of uniform shear stress, along
with a low transverse tensile stress, was obtained for
most of the nite element analysis models. Therefore, it
is strongly believed that the [90] and [0] specimens are
capable of providing reliable shear modulus values.
However, a [0/90] cross-ply specimen should provide an
accurate measure of both the shear modulus and shear
strength of a composite material.
Acknowledgements
The authors express their sincere appreciation to their
fellow members of the Composite Materials Research
Group at the University of Wyoming for their invalu-
able advice and assistance. Special thanks are also
extended to the MARA Institute of Technology (ITM),
Malaysia, and the Federal Aviation Administration,
Washington, DC, for funding this study.
References
[1] ASTM Standard D 4255-83. Standard guide for testing in-plane
shear properties of composite laminates. West Conshohocken
(PA): American Society for Testing and Materials; 1995.
[2] Hussain AK, Adams DF. The Wyoming-modied two-rail shear
test xture for composite materials. Journal of Composites
Technology and Research 1999;21(4):21523.
[3] Boller KH. A method to measure intralaminar shear properties
of composite laminates. Technical report AFML-TR-69-311.
Dayton (OH): Air Force Materials Laboratory; December 1969. Fig. 11. Sketch of the Wyoming-modied two-rail shear test xture.
A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238 237
[4] Garcia R, Weisshaar TA, McWithey RR. An experimental and
analytical investigation of the rail shear-test method as applied to
composite materials. Experimental Mechanics 1980;20:2739.
[5] Black JB Jr., Hart-Smith LJ. The Douglas bonded tapered rail-
shear test specimen for brous composite laminates. In: Proceed-
ings of the 32nd International SAMPE Symposium, April 1987.
p. 36071.
[6] Whitney JM, Daniel IM, Pipes RB. Experimental mechanics of
ber reinforced composite materials. Society for Experimental
Stress Analysis monograph no. 4. Englewood Clis (NJ): Pre-
ntice-Hall; 1982.
[7] Bergner HW, Davis JG, Herakovich CT. Analysis of shear test
methods for composite laminates. Report VPI-E-77-14, Blacks-
burg (VA): Virginia Polytechnic Institute and State University.
April 1977.
[8] Advanced composites design guide; 3rd ed. Wright Patterson Air
Force Base (OH): Air Force Systems Command; 1977. vol. IV.
[9] Chatterjee SN, Adams DF, Oplinger DW. Test methods for
compositesa status report, vol. III. Shear test methods. Report
DOT/FAA/CT-93/17, III. FAA Technical Center, Atlantic City
International Airport (NJ): Federal Aviation Administration;
June 1993.
[10] Hussain AK, Adams DF. An analytical and experimental eval-
uation of the two-rail shear test for composite materials. Report
no. UW-CMRG-R-98-105. Laramie (WY): Composite Materials
Research Group, University of Wyoming; February 1998.
[11] ASTM Standard D 5379-93. Standard test method for shear
properties of composite materials by the V-notched beam
method. West Conshohocken (PA): American Society for Testing
and Materials.
[12] Lewis EQ, Adams DF. An evaluation of composite material
shear test methods. Report no. UW-CMRG-R-91-103, Laramie
(WY): Composite Materials Research Group, University of
Wyoming; May 1991.
[13] Herakovich CT, Bergner HW, Bowles DE. A comparative study
of composite shear specimens using the nite element method. In:
Chamis CC, editor. Test methods and design allowables for
brous composites [ASTM STP 734]. Philadelphia (PA): Amer-
ican Society for Testing and Materials; 1981. p. 12951.
[14] Chang F-K, Chen M-H. The in situ ply shear strength distribu-
tions in graphite/epoxy laminated composite. Journal of Compo-
site Materials 1987;21:70833.
[15] Whitney JM, Stansbarger DL, Howell HB. Analysis of the rail
shear testapplications and limitations. Journal of Composite
Materials 1971;5:2434.
[16] MSC/PATRAN version 6.0. Los Angeles (CA): The Macneal-
Schwendler Corporation; August 1996.
[17] Xie M, Adams DF. A study of compression and shear test methods
for composite materials using a nonlinear nite element analysis.
Report no. UW-CMRG-R-94-102. Laramie (WY): Composite
Materials Research Group, University of Wyoming; June 1994.
[18] Engineers guide to composite materials. Metals Park (OH):
American Society for Metals; 1987.
[19] Lockwood PA. Results of the ASTM round-robin on the rail
shear test for composites. Composite Technology Review 1981;3:
836.
[20] Kotha S, Adams DF. Analytical investigation of crack initiation
and propagation in composite materials. Report no. UW-
CMRG-R-98-119. Laramie (WY): Composite Materials Research
Group, University of Wyoming; December 1997.
238 A.K. Hussain, D.F. Adams / Composites Science and Technology 64 (2004) 221238