Académique Documents
Professionnel Documents
Culture Documents
HANI A. SALIM
Department of Civil Engineering
University of Missouri
Columbia, MO 65211-2200, USA
JEREMY SCHLUSSEL
Department of Civil and Environmental Engineering
West Virginia University
Morgantown, WV 26506-6103, USA
(Received August 10, 2000)
(Revised November 9, 2000)
1151
1152
JULIO F. DAVALOS
ET AL.
In relation to more elaborate methods, the proposed torsional approach can be easily
implemented in the laboratory, and by testing a representative volume of the
material with this method, it is expected to obtain reliable shear moduli data for
structural composites.
KEY WORDS: fiber-reinforced plastic composites, laminates, in-plane and out-ofplane shear moduli, torsion, micro/macro-mechanics, data reduction techniques.
INTRODUCTION
beams, columns, and cellular panels are
increasingly used in structural applications. Structural sections, such as wide-flange,
box, and other shapes, consist typically of arrangements of flat walls. The most common
processing method for FRP structural members is pultrusion, followed by vacuumassisted resin transfer molding (VARTM). The constituents are typically E-glass fiber
systems and vinylester or polyester resins, and the ply materials may include unidirectional
fiber bundles or roving, continuous or chopped randomly-oriented strand mats, and
stitched fabrics with orientations. In the pultrusion process, the sections are not
produced by lamination lay-up, but the arrangement of the constituent materials can be
simulated as a layered system, and the stiffness properties of each panel can be predicted
by lamination theory in terms of the ply stiffnesses computed through micromechanics [1].
Since it is nearly impossible to test the individual plies of a pultruded section, it is
required to determine the material properties at the laminate level through reliable test and
analysis methods.
Shear deformations for pultruded FRP composites can be significant [1] due to the
relatively low shear moduli of the constituents used, and therefore, the modeling of FRP
structural components should account for shear effects. In general, the evaluation of shear
stiffness and strength is a difficult task, and while a number of test methods are available,
as discussed for example by Chiao et al. [2], there can be a considerable discrepancy in the
reported results.
For fiber reinforced composite materials, it is difficult to design a test in which only pure
shear is applied. Generally, there are five acceptable test methods for evaluation of
composite shear properties [3], namely, the Iosipescu test, 45 coupon tension test, the
off-axis tension test, the rail shear test, and the torsion test. The torsion test can induce a
pure shear state of stress, provide average shear properties over a large volume of the
specimen, and minimize localized material and stress concentration effects. For a
rectangular bar under torsion, the cross-section is a plane of elastic symmetry [4], and the
center of twist is assumed to remain constant along the length of the bar. In this state, only
two of the six components of stress are not zero; i.e., xy and xz. Therefore, the shear
moduli Gxy and Gxz can be evaluated from torsion tests, which allow for larger samples to
be tested when compared to other test methods available.
Since most structural composites consist of thick or moderately-thick laminates (0.2500
0.500 or thicker), the number of test methods for determining shear moduli and strengths
is limited. In particular, torsion tests and side-notched (Iosipescu) tests are commonly
used. The torsion tests make use of either cylindrical rods [2,5] or rectangular samples [6],
while the Iosipescu shear test [3] uses a flat specimen with side notches, similar to the
sample proposed by Arcan et al. [7], to develop a state of uniform stress in the gauge
section. The computational procedure of shear moduli from torsion tests and linear elastic
1153
torsion theory for anisotropic bars [8] is well established [9]. Similarly, the Iosipescu test
method has been extensively discussed by various researchers [6,10,11].
The Iosipescu test is the most common test method to compute the in-plane shear
stiffness for composite materials, including pultruded FRP composites. The major
disadvantages of this method are the small effective sections tested, which may not be
representative of the material [11], and the stress concentration at the notched section of
the specimen [12,13]. For pultruded FRP composites [11], it has been observed that there is
significant variability in the results obtained with the Iosipescu test, which can be
attributed to the presence of material defects (resin-rich and fiber-rich zones) over the
small area tested by this method. Other drawbacks of this method are the labor-intense
sample preparation, the large number of required repetitions, and the required fabrication
of special testing fixtures. Moreover, the Iosipescu test is generally used to obtain only
in-plane shear properties.
The torsion test, on the other hand, produces pure shear state of stress in the specimen,
and the analysis of the experimental results can be easily interpreted. The torsion test also
provides average shear properties over a larger volume of the sample minimizing the
localized effect problems encountered in the Iosipescu test. In the literature, however, the
torsion test method has been used mainly for unidirectional laminates, and the procedure
consists of testing laminates with the same material orientation but varying widths [14].
In this paper, we show that this technique grossly underestimates the out-of-plane
shear modulus. A better approach is to test paired samples with two distinct material
orientations.
OBJECTIVES
The objective of this paper is to propose a comprehensive combined experimental and
analytical method to effectively evaluate shear moduli for FRP structural laminates from
torsion. By simulating the pultruded panel as a layered system, the panel shear stiffness
properties are predicted by classical macromechanics and modified transverse shear
deformation theory in terms of the ply stiffnesses and compliances computed through
micromechanics. The experimental program includes testing three distinct pultruded FRP
laminates (unidirectional, angle-ply, and angle/cross-ply) under torsion, and the shear
moduli are evaluated through a new approach using paired samples with material
orientations normal to each other, as opposed to classical methods that can grossly
underestimate the out-of-plane moduli. Torsion solutions by Lekhnitskii [8] and Whitney
[13] are implemented as data-reduction methods to obtain the shear moduli. The test
methods and data-reduction techniques presented in this paper provide practical
guidelines for evaluation of shear moduli by investigators, material producers and
design engineers.
1154
JULIO F. DAVALOS
ET AL.
(a)
(b)
(c)
Figure 1. Material architecture of test samples: (a) unidirectional (Sample U); (b) angle-ply (Sample A) and
(c) angle- and cross-ply (Sample AC).
used to predict the laminate elastic properties. Pultruded FRP shapes are not laminated
structures in a rigorous sense; however, they are produced with material architectures that
can be simulated as laminated configurations [1]. A typical pultruded section may include
the following four types of layers (see examples in Figure 1): (1) a thin layer of randomlyoriented chopped fibers (Veil) placed on the surface of the composite. This is a resin-rich
layer primarily used as a protective coating, and its contribution to the laminate response
can be neglected; (2) continuous or Chopped Strand Mats (CSM) of different weights,
consisting of continuous or chopped randomly-oriented fibers; (3) Stitched Fabrics (SF)
with different angle orientations, and (4) roving layers that contain continuous
unidirectional fiber bundles, which contribute the most to the stiffness and strength of a
section. Each layer is modeled as a homogeneous, linearly elastic, and generally
orthotropic material. The modeling of FRP structural composites as laminates and
prediction of their stiffness properties require accurate ply stiffnesses, which can be
1155
evaluated through micromechanics by estimating the individual ply thickness and the
fiber volume fraction (Vf) of the constituents. Further information on modeling of
pultruded FRP sections and computation of Vf for simulated layer systems can be found
in Davalos et al. [1].
Several formulas of micromechanics of composites have been developed and used over
time [15], and the degree of correlation between experimental data and theoretical
predictions depends on the accuracy of the model used. In this study, we compute the ply
shear stiffnesses for the roving and stitched fabric layers using primarily a micromechanics
model for composites with periodic microstructure [16]; as an illustration, the expression
for the computation of the in-plane shear modulus G12 is
S3
1
G12 Gm Vf
Gm Gm Gf
where Gm and Gf are the shear moduli of the matrix and the fibers, respectively, and S3 is
given as
S3 0:49247 0:47603Vf 0:02748Vf2
In addition, the composite cylinder model developed by Hasin and Rosen [17], which is
based on the self-consistent theory, can provide reasonably accurate stiffness predictions
for roving and SF layers. The formula for G12 given by the composite cylinder model is
G12 Gm
1 Vf Gf 1 Vf Gm
1 Vf Gf 1 Vf Gm
For comparative purposes and because of its popularity, we also compute the elastic
properties of roving and SF layers by the inverse rule of mixtures approach [18]. For
example, the expression for G12 is given by
G12
Gm Gf
1 Vf Gf Vf Gm
Based on the assumption that the material is isotropic in the plane [19], the in-plane
shear stiffness of the CSM layers are estimated from the approximate expressions given by
Hull [20] for randomly oriented composites:
1
1
G12 E1 E2
8
4
where, E1 and E2 are computed from any of the micromechanics models described above.
Using macromechanics, the in-plane shear stiffness of a laminate can be predicted from
the shear compliance (66), which is computed from ply material properties and the inverse
of the extensional stiffness matrix [A] as
Gxy
1
66 h
1156
JULIO F. DAVALOS
ET AL.
where, h is the thickness of the whole laminate. The in-plane shear modulus computed by
Equation (5) is valid for the case of the laminate under in-plane loading. For a laminate
under torsion, the in-plane shear stiffness of the laminate is computed from the twist
compliance coefficient (66), which is dependent on the stacking sequence and is evaluated
from the inverse of the bending stiffness matrix [D] as
Gxy
12
66 h3
Qx
Qy
A55
A45
A45
A44
xz
yz
7
where
N
1
5X
4 3
3
Qij k zk zk1 zk zk1 2
Aij
4 k1
3
h
i, j 4, 5
and zk is the distance from the mid-surface of the kth ply to the middle surface of the
laminate. Ply shear stiffness quantities Q ij k are computed from the individual ply shear
moduli G13 and G23, which are obtained from micromechanics models. Then the out-ofplane modulus of a laminate, Gxz, is
Gxz
A55
h
9
Qy
yz
F45 F44
where
(
Qx
Qy
F44
R55
;
R44 R55 R245
Rij
F55
h=2
h=2
R44
;
R44 R55 R245
xz
yz
)
dz
F45
R45
;
R44 R55 R245
N
9 X
ij k zk zk1 8 z3 z3 1 16z5 z5 1
S
k1 2
k
k1 4
4h2 k1
3 k
h
h
i, j 4, 5
1157
and Sij k denotes the ply transverse shear compliance. Then the out-of-plane shear
modulus of a laminate, Gxz, is
Gxz
F55
h
10
The shear modulus in Equation (10) is based on the integral of weighted shear
compliances and on the assumption that the transverse shear stress distributions are
parabolic and continuous at layer interfaces.
The predicted values from Equations (6) and (10) are representative of the laminate
responses under torsion loading, and they are used in this study as baseline values for
comparisons with experimental data.
E (GPa)
G (GPa)
q (g/cm3)
72.395
5.061
28.841
1.634
0.255
0.30
2.550
1.136
G12 (GPa)
Lamina
t
(mm)
Vf
Periodic
Microstructure
Composite
Cylinder
Inverse Rule
of Mixture
1 oz. CSM
61 yield roving
3/4 oz. CSM
45 SF
62 yield roving
0 /90 SF
0.508a
8.509b
0.381a
0.660a
0.851b
0.660a
0.236
0.410
0.236
0.362
0.627
0.370
4.357
3.536
4.205
3.205
5.852
3.256
4.424
3.652
4.276
3.294
6.219
3.350
3.945
3.907
4.237
3.489
7.138
3.553
1158
JULIO F. DAVALOS
ET AL.
Gxz (GPa)
Sample
Type
Ex
(GPa)
Ey
(GPa)
Equation
(5)
Equation
(6)
Equation
(8)
Equation
(10)
mxy
U
A
AC
30.192
26.366
28.310
10.446
13.259
14.858
3.627
6.412
5.343
3.772
6.247
4.475
2.958
3.813
3.813
2.958
4.199
4.199
0.327
0.396
0.300
values evaluated from micromechanics formulae. Then, using Equations (5), (6), (8) and
(10), the in-plane and out-of-plane shear properties of the laminates (Gxy and Gxz) are
evaluated and are given in Table 3; the predicted values by Equations (6) and (10) are
correlated with experimental results in subsequent sections.
TORSION SOLUTIONS
The first solution for torsion was presented by Coulomb in 1784. About 40 years later,
Navier presented a rigorous solution for torsion of circular shafts. In 1855, St. Venant
developed a torsion solution for rectangular bars that included a warping function. A
detailed discussion of torsion theories is presented by Hsu [23].
Lekhnitskii [8] applied St. Venants formulation for torsion to orthotropic materials and
obtained solutions for circular and rectangular bars with three mutually perpendicular
planes of material symmetry. Davalos et al. [9] applied this approach to orthotropic wood
samples and presented a combined analytical/experimental method to obtain principal
shear moduli. Lekhnitskiis solution can be successfully used for orthotropic samples; for
more complex materials, Whitney [13] presented a solution incorporating transverse shear
deformation for the torsion of rectangular laminated anisotropic composite plates. In the
following two sections, we briefly describe the computational algorithms of Lekhnitskiis
and Whitneys torsion solutions for anisotropic materials.
Lekhnitskiis Solution for Orthotropic Materials
A solution for the torsion of rectangular orthotropic bars in their principal material
directions was presented by Lekhnitskii [8]. In Figure 2, the sample is subjected to a
torque, T, applied about the longitudinal axis x and causing an angle of twist . The
dimensions of the sample are: a width, b thickness, and L length. The torque versus
angle of twist (T/) can be expressed as
T
ab3
Gxy
L
11a
11b
1159
In Equation (11), the torsional stiffness is a function of the two principal shear moduli.
When Equation (11) is reduced to an isotropic case (Gxy Gxz G) and evaluated for a
large number of terms, is denoted by k and expressed as
1
b
1 0:63
k
3
a
By substituting k for in Equation (11), we obtain St. Venants torsion solution for a
rectangular isotropic bar:
T
ab3 1
b
Gxy
1 0:63
a
L 3
12
The most commonly used approach to solve for the shear moduli in Equation (11) is to
test two samples with different cross-sectional dimensions but same material orientation
[4]. A graphical method proposed by Tarnopolskii and Kincis [24] is typically used to
solve for the shear moduli. The problem with this approach is that the out-of-plane shear
modulus, Gxz, is grossly underestimated. For example, Sumsion and Rajapakse [14]
reported Gxz values as much as 90% lower than Gxy for unidirectional transversely
isotropic composites, for which one should expect Gxz Gxy.
A better approach is to pair samples with either the same or different cross-sectional
dimensions but with material orientations normal to each other. Davalos et al. [9] used this
approach for paired orthotropic wood samples manufactured with the same crosssectional dimensions, but with principal material orientations perpendicular to one
another. The corresponding two distinct equations were then solved by a method of
successive substitutions. In this study, we modify the approach proposed by Davalos et al.
[9] to obtain a more efficient solution by the bisection method [25,26]. Pultruded FRP
structural laminates are considered, and to obtain samples of variable widths in the
xz-plane, the pultruded laminates [Figure 3(a)] were bonded flat-wise, and vertical strips
were cut from the bonded assemblies as shown in Figure 3(b).
1160
JULIO F. DAVALOS
(a)
ET AL.
(b)
Figure 3. Sample orientations for: (a) in-plane (xy) and (b) out-of-plane (xz) shear moduli.
D66 1 2= 1 D 66 =D66 tanh =2
13
where D 66 is given by
B D B D16 B11 D16 D16 A11 B11 B16
D 66 D66 16 11 16
D11 A11 B2
11
14
Aij , Bij , and Dij are the coefficients of the inverse of the ABD matrix, and is defined as
2s3
D 66 A44 A55 A245 5
a4
A44
15
where a width of sample (Figure 3). For a symmetric orthotropic sample, we get
A45 B16 D16 0, and Equation (13) can be simplified as
4ab3 Gxy
T
2
tanh
1
2
12
where
"s#
12Gxz
a
b2 Gxy
16
1161
The Lekhnitskiis and Whitneys torsion solutions summarized above are used in
conjunction with torsion tests to evaluate the shear moduli for various laminates. The
application of torsion solutions as data reduction techniques for torsion tests is further
discussed in the Results section.
TEST SAMPLES AND PROCEDURE
The three sample types shown in Figure 1 were tested using a manually operated torsion
machine that permitted accurate measurements of applied torque and angle of twist. In
this section we describe the test specimens and methods used.
Test Samples
The unidirectional samples (U) were cut from a pultruded flat plate of thickness
9.53 mm (3/800 ) [Figure 1(a)], and both the angle-ply samples (A) and angle/cross-ply
samples (AC) were obtained from pultruded wide-flange beams 304.8 304.8 12.7 mm
(1200 1200 1/200 ); thus the thickness of these samples was 12.7 mm(1/200 ), and their fiber
architectures are shown respectively in Figures 1(b) and 1(c). For each of these three
sample-types, specimens of 0.432 m (1700 ) in length were cut in four widths: 25.4 mm (1.000 ),
38.1 mm (1.500 ), 50.8 mm (2.000 ), and 63.5 mm (2.500 ) [e.g., for the U samples, U10, U15, U20
and U25 represent the unidirectional samples with widths of 25.4 mm (1.000 ), 38.1 mm
(1.500 ), 50.8 mm (2.000 ) and 63.5 mm (2.500 ), respectively], with four replications for each
width. The orientation of these samples is described in Figure 3(a), with the width
dimension along the y-axis and the longitudinal dimension along the x-axis.
To obtain samples with larger dimensions than the available material thickness in the
xz-plane [see Figure 3(b)], flat strips were cut from the plate and beam components and
bonded flat-wise with epoxy. The bonded layered specimens were subsequently cut into
vertical strips to obtain samples with the same dimensions as those obtained directly
from the pultruded components; i.e., samples with thickness 9.53 mm (3/800 ) or 12.7 mm
(1/200 ) and width 25.4 mm (1.000 ), 38.1 mm (1.500 ), 50.8 mm (2.000 ) and 63.5 mm (2.500 ). This
approach was used to produce paired samples with same dimensions, but material
orientations normal to each other (see Figure 3). To denote the bonded samples for each
type of laminate, the letter B is added to the corresponding designation; e.g., UB identifies
the unidirectional bonded samples.
Test Procedure
The samples were tested in a manually-operated torsion machine. The torque was
applied through a large wheel at one end and a counterbalance pendulum at the other
end. To increase the accuracy of angle of twist measurement, a circular gear was fitted
concentrically through the specimen at each gage point, and the rotation of the gear
was transferred by a plastic chain to an outside small pulley in order to magnify the
displacement of a transducer (LVDT) suspended from the pulley by a plastic thread.
The angle of twist between two sections sufficiently away from the grips (ASTM D198)
was computed from the displacements of the transducers and the geometry of the
gear-and-pulley mechanism. In this study, the gauge length of 0.229 m (9.0 in) is used to
measure the angle of twist between two sections.
1162
JULIO F. DAVALOS
ET AL.
The torque was applied at a rate of about three degrees per minute, for a total angle of
twist within the material elastic limit of about 10 . Each sample was tested four times, and
the torque versus angle of twist (T/) was obtained through linear regression of the data.
For further details of the experimental procedure, see Schussel [27]. The data reduction
techniques used for various torsion solutions are discussed next.
RESULTS, DATA REDUCTION METHODS AND DISCUSSION
The torsional stiffnesses (T/) for three types of laminated samples were obtained
experimentally to evaluate their shear moduli from torsion solutions. The experimental
program is organized in two parts: (1) unidirectional (U) samples and (2) angle-ply (A)
and angle/cross-ply (AC) samples.
Unidirectional Samples
This phase of the study is concerned with the response of the unidirectional samples
shown in Figure 1(a). Two approaches are used to evaluate the shear moduli Gxy and Gxz
from paired samples. The first approach consists of applying Lekhnitskiis orthotropic
torsion solution to test results of samples of different sizes but same material orientations;
in the second approach, the solutions by Lekhnitskii and Whitney are used in conjunction
with test results for samples with mutually orthogonal material orientations; i.e.,
combining samples U (cut from the pultruded plate) and UB (produced by bonding) as
shown in Figure 3. Further, as an approximation and to show the validity of transverse
isotropy for these samples, the shear moduli is also obtained from St. Venants isotropic
solution (Equation 12). The results for each method used are discussed next.
UNIDIRECTIONAL SAMPLES WITH SAME MATERIAL ORIENTATIONS
A commonly used method to evaluate shear moduli for specially orthotropic materials
(when the geometric and material axes coincide) is by torsion tests of samples with
different cross-sectional dimensions but same material orientation. However, as shown in
this study, this method can grossly underestimate the out-of-plane shear modulus Gxz.
Samples with four different widths were tested in torsion, and the average dimensions and
torsional stiffnesses of the samples, including the bonded samples to be used later, are
given in Table 4.
Following the graphical method by Tarnopolskii and Kincis [24], the results for the
unidirectional samples U with same material orientations are shown in Table 5. Based on
Table 4. Average dimensions and response of U and UB samples.
Sample
U10
U15
U20
U25
UB10
UB15
UB20
UB25
Width (mm)
Thickness (mm)
T//(N-m)
24.638
37.694
50.444
63.221
25.781
37.694
50.419
65.075
9.296
9.373
9.398
9.373
9.296
9.373
9.550
9.042
1163
Gxy (GPa)
Gxz (GPa)
3.647
3.813
3.806
4.020
3.944
3.992
1.269
1.151
1.151
0.848
0.910
0.841
3.868 (3.65%)
1.027 (17.90%)
U10 : U15
U10 : U20
U10 : U25
U15 : U20
U15 : U25
U20 : U25
Average (COV)
transverse isotropy, it is expected that the two shear moduli be approximately equal to
each other, but in this case, the out-of-plane modulus Gxz obtained is only about 27% of
the in-plane modulus Gxy. A similar discrepancy can be observed in the results reported for
unidirectional samples by Sumsion and Rajapakse [14], who provided no explanation for
the out-of-plane modulus being as low as one-tenth of the in-plane modulus. A better
approach is to use paired samples with material orientations normal to each other, as
shown next.
UNIDIRECTIONAL SAMPLES WITH DISTINCT
MATERIAL ORIENTATIONS
To obtain shear moduli from torsion solutions by Lekhnitskii [Equation (11)] and
Whitney [Equation (16)], an effective approach is to determine the (T/) values for at least
two samples with distinct material orientations [9]. In this study, samples with the same
cross-sectional dimensions but orthogonal material orientations were paired (Figure 3),
and using Equations (11) and (16), the in-plane and out-of-plane shear moduli values were
obtained. Each size group consisted of four samples in the xy direction and four samples in
the xz direction to obtain 16 sets of values for Gxy and Gxz. Furthermore, by cross-pairing
each size-group of xy samples with the four size-groups of xz samples, a total of 64 sets of
Gxy and Gxz values were obtained, with a final grand total of 256 sets of values.
Data Reduction Based on Lekhnitskiis Solution
To use Lekhnitskiis solution, Equation (11) was rewritten for each sample orientation as
Gxy
K1 L1
;
a1 b31 1
Gxz
K2 L2
a2 b32 2
17a
where
Ki
Ti
,
i
i 1, 2
17b
17c
1164
JULIO F. DAVALOS
s!#
"
r
1
32a22 Gxy X
1
2a2 Gxy
i
b2 Gxz
2 4 2
1
tanh
i
b2 Gxz
2a2 Gxy
ET AL.
17d
Both Equations (17a) need to be solved simultaneously, and to simplify the computational
effort we divide Equation (17c) by Equation (17d), resulting in
1
h
X
p i
p
1=i4 1 2a1 =i
b1 tanhi
b1 =2a1 1=
3
a2 b2 K1 L1
i1, 3, 5
18
1
X
p
a1 b1 K2 L2
1h
p i
1
2a
=i
b
1=
tanh
i
b
=2a
2
2
2
2
i4
i1, 3, 5
where
Gxz
Gxy
Now, there is only one unknown, , in Equation (18), and by solving for by the bisection
method [25,26], we can obtain the in-plane (Gxy) and out-of-plane (Gxz) shear moduli as:
(
Gxz
a31 b1
4
1
h
X
p i 32
p
1=i4 1 2a1 =i
b1 tanh i
b1 =2a1 1=
K1 L1
19a
i1, 3, 5
Gxy
Gxz
19b
Table 6 presents the results for paired samples of the same size and cross-paired samples
of different sizes. The results of this method support the assumption that an orthotropic
composite material with unidirectional fibers can be modeled as transversely isotropic.
Also, the results indicate that as the width of the sample increases, the corresponding shear
modulus on that plane increases (Table 6). To observe this effect in Table 6, see any row
for Gxy and any column for Gxz.
Data Reduction Based on Whitneys Solution
Whitneys simplified torsion solution [13] is given in Equation (16). By mathematically
rotating the coordinate system (Figure 4), the following system of equations are obtained
for the xy and xz planes:
Table 6. Shear moduli results using Lekhnitskiis solution for U samples.
Width
(mm) xz
Plane
25.781
37.694
50.419
65.075
Gxy (GPa)
Gxz (GPa)
xy Plane
xy Plane
24.638
37.694
50.444
63.221
24.638
37.694
50.444
63.221
2.959
2.936
2.858
2.841
3.250
3.233
3.180
3.169
3.384
3.371
3.332
3.323
3.535
3.524
3.491
3.484
2.683
2.793
3.250
3.366
2.644
2.767
3.225
3.344
2.627
2.755
3.214
3.338
2.611
2.743
3.203
3.327
(a)
1165
(b)
Figure 4. Orientation for: (a) in-plane shear and (b) out-of-plane shear moduli.
s!
p
#
"
3a1 p
T1
1 4a1 b31 Gxy
b1
1
tanh
xy-plane : K1
1 p
1 L1
12
b1
3a1
20a
p s!#
"
3a2 1
T2
1 4a2 b32 Gxy
b2 p
xz-plane : K2
tanh
1 p
2 L2
12
b2
3a2
20b
where
Gxz
Gxy
p
p
K2
L1
a2 b32 1 b2 = 3a2 tanh
3a2 =b2 1=
21
Again, by using the bisection method [26] and solving for the only unknown, , we can
obtain the in-plane (Gxy) and out-of-plane (Gxz) shear moduli as:
Gxy
3K1
p p
p
p
a1 b31 1 b1 = 3a1
1= tanh
3a1 =b1
Gxz Gxy
22a
22b
Table 7 presents the results for paired samples of the same size and cross-paired samples
of different sizes. Similar to the results based on Lekhnitskiis solution, the shear modulus
increases as the corresponding width of the sample increases.
1166
JULIO F. DAVALOS
ET AL.
Width (mm)
xz plane
25.781
37.694
50.419
65.075
Gxy (GPa)
Gxz (GPa)
xy Plane
xy Plane
24.638
37.694
50.444
63.221
24.638
37.694
50.444
63.221
2.877
2.853
2.786
3.199
3.195
3.178
3.131
3.199
3.343
3.330
3.294
3.285
3.501
3.490
3.460
3.453
2.619
2.754
3.215
3.337
2.586
2.728
3.189
3.314
2.571
2.718
3.179
3.310
2.557
2.708
3.169
3.303
Samples
Width
(mm)
a/b
Ratio
Gxy
24.638
2.65
37.694
4.02
50.444
5.37
63.221
6.75
Average (COV)
Gxz
25.781
37.694
50.419
65.075
2.77
4.02
5.28
7.20
Average (COV)
Isotropic
(GPa)
Lekhnitskii
(GPa)
Whitney
(GPa)
Predicted
(GPa)
2.882
3.282
3.344
3.496
3.254 (8.05%)
2.896
3.206
3.351
3.509
3.241 (8.06%)
2.930
3.179
3.316
3.475
3.227 (7.16%)
3.771
Equation (6)
3.034
3.199
3.241
3.372
2.641
3.041
3.227
3.344
2.586
2.734
3.185
3.316
2.958
Equation (10)
3.213 (4.34%)
3.061 (10.06%)
2.958 (11.89%)
1167
analyzed by both Lekhnitskiis and Whitneys solutions. The fiber architectures are shown
in Figure 1(b) (samples A) and Figure 1(c) (samples AC), and the layer and laminate
properties are given in Tables 2 and 3, respectively. The average dimensions and measured
torsional responses (T/) for the as-manufactured and bonded samples are summarized in
Table 9.
Using the sample-pairing technique by testing samples with material orientations
normal to each other (e.g., samples A and AB; samples AC and ACB), as discussed
above for the unidirectional samples, the shear moduli were obtained through data
reduction torsion solutions. The results for Lekhnitskiis and Whitneys solutions
are given, respectively, in Tables 10 and 11. Each value reported in these tables is the
average of 16 results, corresponding to four replications per sample. As observed in
Tables 10 and 11, consistent results are obtained for both Lekhnitskiis and Whitneys
solutions.
(a)
(b)
Figure 5. (a) In-plane shear modulus Gxy for U samples; (b) Out-of-plane shear modulus Gxz for U samples.
1168
JULIO F. DAVALOS
ET AL.
Width
(mm)
Thickness
(mm)
A10
A15
A20
A25
AB10
AB15
AB20
AB25
AC10
AC15
AC20
AC25
ACB10
ACB15
ACB20
ACB25
24.994
37.719
50.470
63.754
25.222
37.744
50.394
63.856
25.070
37.795
50.495
63.830
25.222
37.871
50.571
63.906
12.649
12.573
12.573
12.598
11.786
11.862
11.811
11.735
12.675
12.675
12.675
12.675
12.040
12.141
12.090
12.065
T//(N-m)
311.445
516.752
756.929
1122.146
185.585
308.939
418.072
553.605
261.299
439.600
662.312
911.575
201.557
351.071
476.616
623.885
(COV 3.22%)
(COV 8.75%)
(COV 8.18%)
(COV 9.94%)
(COV 5.63%)
(COV 6.47%)
(COV 11.91%)
(COV 6.91%)
(COV 2.85%)
(COV = 6.93%)
(COV = 7.51%)
(COV = 6.97%)
(COV = 5.78%)
(COV = 10.43%)
(COV = 5.80%)
(COV = 9.57%)
Table 10. Shear moduli results using Lekhnitskiis solution for A and AC samples.
Samples
A
AC
Gxy (GPa)
Gxz (GPa)
xy Plane
xy Plane
Width
(mm)
xz Plane
24.994
37.719
50.470
63.754
24.994
37.719
50.470
63.754
25.222
37.744
50.394
63.856
7.382
7.376
7.446
7.288
6.438
6.449
6.479
6.424
6.462
6.469
6.489
6.454
7.266
7.266
7.280
7.247
3.927
3.933
3.862
4.028
4.016
3.985
3.901
4.055
4.016
3.982
3.900
4.050
3.936
3.936
3.873
4.031
Width
(mm) xz
Plane
xy Plane
xy Plane
25.070
37.795
50.495
63.830
25.070
37.795
50.495
63.830
25.222
37.871
50.571
63.906
5.461
5.434
5.490
5.470
5.076
5.072
5.099
5.091
5.367
5.360
5.379
5.371
5.619
5.611
5.618
5.619
4.278
4.336
4.215
4.258
4.351
4.372
4.241
4.278
4.296
4.342
4.220
4.267
4.252
4.314
4.258
4.252
1169
Samples
A
AC
Gxy (GPa)
Gxz (GPa)
xy Plane
xy Plane
Width
(mm) xz
Plane
24.994
37.719
50.470
63.754
24.994
37.719
50.470
63.754
25.222
37.744
50.394
63.856
6.864
6.840
6.882
6.857
6.243
6.240
6.260
6.212
6.334
6.332
6.345
6.313
7.149
7.142
7.151
7.121
3.865
3.897
3.842
3.875
3.920
3.929
3.864
4.023
3.911
3.924
3.860
4.020
3.842
3.882
3.833
3.997
Width
(mm) xz
Plane
xy Plane
xy Plane
25.070
37.795
50.495
63.830
25.070
37.795
50.495
63.830
25.222
37.871
50.571
63.906
5.197
5.159
5.195
5.175
4.960
4.945
4.964
4.955
5.281
5.266
5.279
5.271
5.549
5.535
5.545
5.538
4.169
4.271
4.173
4.229
4.207
4.293
4.188
4.240
4.156
4.262
4.169
4.224
4.118
4.239
4.153
4.213
Sample
A
Shear
Modulus
Gxy
Sample
Width
(mm)
a/b
Ratio
Lekhnitskii
(GPa)
Whitney
(GPa)
Predicted
(GPa)
24.994
37.719
50.47
63.754
1.98
3.00
4.01
5.06
7.373
6.447
6.469
7.265
6.861
6.238
6.331
7.141
6.247
Equation (6)
6.889
(7.25%)
6.643
(6.48%)
3.974
3.959
3.884
4.041
3.885
3.908
3.850
3.979
3.965
(1.63%)
3.905
(1.40%)
5.463
5.084
5.369
5.617
5.181
4.956
5.275
5.542
5.383
(4.16%)
5.239
(4.63%)
4.294
4.341
4.233
4.264
4.162
4.267
4.171
4.227
4.283
(1.07%)
4.207
(1.17%)
Average (COV)
Gxz
25.222
37.744
50.394
63.856
2.14
3.18
4.27
5.44
Average (COV)
AC
Gxy
25.070
37.795
50.495
63.830
1.98
2.98
3.98
5.04
Average (COV)
Gxz
25.222
37.871
50.571
63.906
Average (COV)
2.09
3.12
4.18
5.30
4.199
Equation (10)
4.475
Equation (6)
4.199
Equation (10)
1170
JULIO F. DAVALOS
ET AL.
(a)
(b)
Figure 6. (a) In-plane shear modulus Gxy for A samples; (b) Out-of-plane shear modulus Gxz for A samples.
solution. Graphical representations of the results are given in Figures 6a, 6b, 7a and 7b for
samples A and AC, respectively.
For the U samples, we observed an asymptotic increase of shear moduli values as the
sample-width increases (Figures 5a and 5b). In contrast to this behavior, there is no clear
trend on the data obtained for the A and AC samples in relation to the sample-width; in
general, the experimental values are randomly distributed around a narrow band, with
relatively larger discrepancies for the A samples (Figures 6a and 6b) than for the AC
samples (Figures 7a and 7b). One reason for the large variability with the A samples can be
attributed to the larger difference between Gxy and Gxz (degree of anisotropy) than for the
AC samples, which are nearly transversely isotropic (similar to the U samples). For both A
and AC samples, the experimental values are higher than the predictions for the in-plane
shear moduli.
1171
(a)
(b)
Figure 7. (a) In-plane shear modulus Gxy for AC samples; (b) Out-of-plane shear modulus Gxz for AC samples.
CONCLUSIONS
It is shown that torsion tests of pultruded FRP composites with rectangular crosssections can be used to determine in-plane and out-of-plane shear moduli values. The
determination of out-of-plane shear moduli from samples with same material orientation
does not lead to accurate results. It is necessary to test samples with two material
orientations, which can be obtained by bonding flat laminates and cutting vertical strips
from the assemblies. Both Lekhnitskiis and Whitneys solutions with paired samples are
recommended to obtain shear moduli values using the torsional stiffness from torsion tests
and the data reduction methods outlined in this study. Consistent results are obtained for
both Lekhnitskiis and Whitneys solutions. The predicted in-plane and out-of-plane shear
1172
JULIO F. DAVALOS
ET AL.
moduli of the laminates are obtained from micro/macromechanics models for the case of
torsional loading [Equations (6) and (10)]. The combined experimental/analytical torsion
solutions presented in this paper can be efficiently used to obtain reliable shear moduli for
composite laminates.
ACKNOWLEDGMENTS
The test samples were generously provided by Creative Pultrusions, Inc., Alum Bank,
PA. Partial financial support for this study was received from NSF-CRCD program.
REFERENCES
1. Davalos, J.F., Salim, H.A., Qiao, P.Z., Lopez-Anido, R. and Barbero, E.J. (1996). Analysis
and design of pultruded FRP shapes under bending. Composites, Part B: Engineering Journal,
27B(3&4): 295305.
2. Chiao, C.C., Moore, R.L. and Chiao, T.T. (1977). Measurement of shear properties of fiber
composites, Part 1: Evaluation of test methods. Composites, 8(3): 161168.
3. Adams, D.F. and Lewis, E.Q. (1995). Current status of composite material shear test method.
SAMPE Journal, 31(6): 3240.
4. Kurtz, R.D. and Sun, C.T. (1990). Composite shear moduli and strengths from torsion of thick
laminates. In: Garbo, S.P. (ed.), Composite Materials: Testing and Design. 3rd Vol., ASTM STP
1095, Philadelphia, PA: ASTM. pp. 508520.
5. Knight, M. (1982). Three-dimensional elastic moduli of graphite/epoxy. Journal of Composite
Materials, 16: 149153.
6. Adams, D.F. and Walrath, D.E. (1987). Further development of the Iosipescu shear test method.
Experimental Mechanics, 27(2): 113119.
7. Arcan, M., Hasin, Z. and Eoloshin, A. (1978). A method to produce uniform plane-stress states
with applications to fiber-reinforced materials. Experimental Mechanics, 18: 141146.
8. Lekhnitskii, S.G. (1963). Theory of Elasticity of an Anisotropic Body. San Francisco, CA:
Holden-Day Inc.,
9. Davalos, J.F., Loferski, J., Holzer, S. and Yadama,V. (1991). Transverse isotropy modeling of
3D glulam timber beams. Journal of Materials in Civil Engineering, 3(2): 125139.
10. Swanson, S.R., Messick, M. and Toombes, G.R. (1985). Comparison of torsion tube and
Iosipescu in-plane shear test results for a carbon fiber-reinforced epoxy composite. Composites,
16(3): 220224.
11. Sonti, S.S. and Barbero, E.J. (1995). Determination of shear properties for RP pultruded
composite. Journal of Reinforced Plastics and Composites, 14: 390400.
12. Arnautov, A. and Bach, T. (1996). Determination of in-plane shear characteristics of composite
materials with [ 45 ] layup. Mechanics of Materials, 32(2): 176182.
13. Whitney, J.M. (1993). Analysis of anisotropic laminated plates subjected to torsional loading.
Composites Engineering, 3(1): 8397.
14. Sumsion, H.T. and Rajapakse, D.S. (1978). Simple torsion test for shear moduli determination
of orthotropic composite. ICCM/2 Proc., Int. Conf. on Composite Materials. New York.
pp. 9941002.
15. Chamis, C.C. (1984). Simplified composites micromechanics equations for strength, fracture
toughness, and enviromental effects. NASA TM-83696. USA.
16. Luciano, R. and Barbero, E.J. (1994). Formulas for the stiffness of composites with periodic
microstructure. Int. J. of Solids and Structures, 31(21): 29332944.
1173
17. Hasin, Z. and Rosen, B.W. (1964). The elastic modeling of fiber-reinforced materials. Journal of
Applied Mechanics, 31: 223232.
18. Jones, R.M. (1975). Mechanics of Composite Materials. NewYork: Hemisphere Publishing Corp.
19. McCullough, R.L. (1971). Concept of Fiber-Resin Composites. NewYork: Marcel Dekker, Inc.
20. Hull, D. (1981). An Introduction to Composite Materials. Cambridge University Press.
21. Vinson, J.R. and Sierakowski, R.L. (1987). The Behavior of Structures Composed of Composite
Materials. Dordrecht, Netherlands: Martinus Nijhoff Publishers.
22. Whitney, J.M. (1990). A modified shear deformation theory for laminated anisotropic plates. In:
Proceedings of the American Society for Composites, 5th Technical Conf. Lancaster, PA:
Technomic. pp. 469478.
23. Hsu, T.C. (1984). Torsion of Reinforced Concrete. New York: Van Nostrand Reinhold Co.,
24. Tarnopolskii, Y.M. and Kincis, T. (1985). Static Test Methods for Composite. New York: Van
Nostrand Reinhold Co.
25. Conte, S.D. and de Boor, C. (1980). Elementary Numerical Analysis - An Algorithmic Approach.
New York, NY: McGraw-Hill, Inc.
26. Qiao, P.Z. and Wang, J.L. (2000). A computer program for shear moduli evaluation of
structural composites from torsion tests: Lekhnitskiis and Whitneys solutions. Internal Report,
ACMSL #2000-2. Advanced Composite Materials and Structures Laboratory, Department of
Civil Engineering, The University of Akron, OH.
27. Schussel, J. (1996). Analytical and experimental evaluation of stiffnesses for FRP laminates.
Master of Science Thesis, Civil Engineering, West Virginia University, Morgantown, WV.
28. Tomblin, J.S. (1994). Compressive strength models for pultruded glass fiber reinforced
composites. PhD Dissertation, Department of Mechanical and Aerospace Engineering, West
Virginia University, Morgantown, WV.