1 s2.0 S0263822306002364 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Composite Structures 80 (2007) 396–408

www.elsevier.com/locate/compstruct

Determining the elastic interlaminar shear modulus


of composite laminates
A. Chan a, W.K. Chiu a,*
, X.L. Liu b

a
Department of Mechanical Engineering, Monash University, Wellington Road, P.O. Box 31, Clayton, Vic. 3800, Australia
b
Cooperative Research Centre for Advanced Composite Structures Limited, 506 Lorimer Street, Fishermans Bend, Vic. 3207, Australia

Available online 17 July 2006

Abstract

An inverse parameter identification technique has been proposed to determine the elastic interlaminar shear modulus of composite
laminates. The technique involves minimising the difference between an experimentally measured and a numerically determined material
response by varying the interlaminar shear modulus in the numerical model. As such, the ‘optimum’ interlaminar shear modulus can be
found by using a relatively simple mechanical test. The inverse parameter identification technique using a modified Iosipescu shear test
and three-point bend test was applied in the current study. Experimental investigations, numerical studies, as well as the development of
a computer program were conducted to implement the proposed technique. An average elastic interlaminar shear modulus of 3.23 GPa
and 2.37 GPa was obtained with the modified Iosipescu shear and three-point bend tests, respectively. These results are extremely prom-
ising as they are close to previously speculated values. It is acknowledged that a discrepancy of approximately 26.6% exists between the
two results, however, a comprehensive explanation of this difference is provided.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Interlaminar shear modulus; Composite materials; Mechanical properties of composite laminates; Shear modulus

1. Introduction essary use of resources. Such trial and error processes are
generally required as there is often insufficient material
The use of composite materials is becoming very popu- data to allow for accurate predictions in the behaviour of
lar in many industries. In particular, the aerospace and the composite structures. In many cases, insufficient
defence industries have been increasing their dependence material data stems from the absence of reliable test meth-
on the application of composite materials. However, usage ods to provide such information. Standard test methods
of these materials is currently limited, to a certain degree, currently exist for most of the in-plane elastic and shear
by excessive costs that can be attributed to expensive raw moduli, and strength parameters of composite materials
materials and relatively complicated manufacturing and [2]. However, test methods for obtaining interlaminar
assembly procedures. Much effort has been made to reduce Young and shear moduli are primitive at best [3]. It is,
the associated costs in these areas. Projects such as afford- therefore, imperative that robust methodologies for deter-
able manufacturing and assembly have been developed at mining the interlaminar material properties of composite
the CRC-ACS [1], which have been successful in introduc- materials be developed.
ing significant cost reductions. A noticeable void in current literature is the lack of a
However, there are still trial and error processes cur- test method for determining the interlaminar shear modu-
rently being employed in the design and manufacture of lus. This is largely attributable to the fact that conventional
composite structures and components, resulting in unnec- methods of direct stress and strain measurements cannot be
easily adapted for the measurement of interlaminar proper-
*
Corresponding author. Tel./fax: +61 3 9905 5595. ties. For example, utilising these conventional methods for
E-mail address: wing.kong.chiu@eng.monash.edu.au (W.K. Chiu). determining the interlaminar shear modulus requires extre-

0263-8223/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2006.05.025
A. Chan et al. / Composite Structures 80 (2007) 396–408 397

mely thick composite coupons to be manufactured, which • finite element (FE) models that can accurately simulate
has proven to be very difficult and costly [4]. the mechanical tests and
The Iosipescu shear test [5,6] is the only standard proce- • a computer program that can efficiently perform the
dure that has been used in attempts to determine the inter- minimisation process.
laminar shear modulus of composite laminates. However,
previous investigations have revealed that a pure shear Each of these components will be described in greater
state, for which the Iosipescu test relies on, may not be detail in later sections.
achievable in the test [7–9]. Mespoulet et al. [10] attempted
to determine the interlaminar shear modulus using a test 2.1. Material properties
similar to that proposed by Post et al. [11]. Two sides of
a composite specimen were adhesively bonded to steel rails, Hexcel F593-18 carbon–epoxy plain weave pre-preg,
and then loaded to shear. The specimens were strain- designated as W3T282-4200 -F593, is to be used as the quan-
gauged on the two free sides. It was found that failure tifying material in the current study. The stiffness material
always occurred through a combination of shear and trans- properties of this material are shown in Table 1.
verse tension, indicating a pure shear mode was not
achieved in the test. Pierron and Grediac [12] proposed a 3. Sensitivity analysis
technique to take advantage of the existence of a mixture
of tension and shear stresses in the Iosipescu shear tests. For the proposed inverse parameter identification tech-
An identification procedure based on the principle of vir- nique to be successful, it is important that the measured
tual work with particular virtual fields was proposed to response in the selected mechanical test is sensitive to
measure the four through-thickness moduli of thick cou- changes in the interlaminar shear modulus while remaining
pons. Promising analytical results were obtained. However, relatively insensitive to changes in the other unknown
experimental application were yet to be implemented. material properties. A sensitivity analysis was previously
It is evident from these investigations that a methodol- conducted by the current authors [13] to determine a suit-
ogy that can efficiently and accurately determine the inter- able mechanical test and the corresponding material
laminar shear modulus of composite laminates in the response for use with the inverse parameter identification
elastic domain is not yet available. The main intention of technique. Seven mechanical tests were considered in that
the current study is to develop such a methodology and study. Their suitability for the proposed technique was
hence provide a solution to this significant problem. determined by considering their sensitivity to changes in
different material properties. From the analysis results, it
2. Inverse parameter identification technique was found that the modified Iosipescu shear test and the
three-point bend test were the most suitable.
The present authors believe that the elastic interlaminar As can be seen in Table 1, there is a significant difference
shear modulus can be determined by applying an inverse between the material properties used in the previous sensi-
parameter identification approach to a customised mechan- tivity analysis and that of the Hexcel F593-18 composite
ical test. This approach involves minimising the difference
between an experimentally measured and numerically pre- Table 1
dicted material response by varying the interlaminar shear Mechanical tests and geometric parameters considered
modulus in the numerical model. As such, the interlaminar Potential test Geometric parameters
shear modulus can be determined using a relatively Constant (mm) Variables (mm)
straightforward mechanical test.
Standard Iosipescu a = 80.0 b = 20.0, 24.0, 28.0, 32.0, 36.0
The steps involved in the approach are illustrated in shear
Fig. 1, with the key components being: Modified Iosipescu a = 80.0 b = 20.0, 24.0, 28.0, 32.0, 36.0
shear
• an experimental process that can be applied to obtain Three-point bend S = 120.0 t = 10.0, 15.0, 20.0, 25.0
Off-axis tensile L = 60.0 h = 7, 10, 12, 16
reliable and repeatable data;

n n
Experiment: Run FE Obtain G13 / G23 are the
Apply load P0 simulation translation ΔFE Is optimum
0
|ΔFE - ΔEXP| < ε? Yes
and obtain with G13 and from interlaminar shear
translation ΔEXP
0
G23 and P0 simulations moduli

n
Change G13
n
and G23 No
accordingly

Fig. 1. Flowchart of the inverse parameter identification technique.


398 A. Chan et al. / Composite Structures 80 (2007) 396–408

material used in the current study. The four most promis- in an FE analysis, as such careful considerations must be
ing mechanical tests from the previous sensitivity analysis taken to ensure that the chosen element does not affect
were reconsidered in the current investigation. The aim is the overall results of the sensitivity analysis.
to test the validity of these tests when the new material The current FE simulations being conducted are linearly
properties are applied. The four tests considered are the static involving 3-dimensional orthotropic material proper-
standard Iosipescu shear test, modified Iosipescu shear test, ties. In ABAQUS/Standard, the two most obvious choices
three-point bend test and the off-axis tensile test. are C3D8 and C3D20R elements. The C3D8 is a first order
element consisting of eight nodes, while the C3D20R is a
3.1. Geometric parameters second-order elements consisting of 20 nodes. Second-
order elements provide higher accuracy than first-order ele-
The geometric parameters of the four mechanical tests ments for problems that do not involve complex contact
are shown in Fig. 2. The location from which the material conditions, impact, or severe element distortions. They
responses were obtained are indicated in these figures. The capture stress concentrations more effectively and are bet-
corresponding values of the geometric parameters that are ter for modelling geometric features. The downside to sec-
varied in the sensitivity analysis are shown in Table 2. ond-order elements is that they are computationally more
expensive than first-order elements [14].
3.2. Element type—validation study Stress measurements and accurate stress distributions
are not a major factor in the current FE simulations as
A validation study was first conducted to test the affects the chosen material responses are all displacements. There-
of using two different types of elements in the FE models. fore, C3D8 elements would be much more attractive than
There are many different element types that could be used C3D20R elements if similar accuracies can be achieved

LOAD LOAD LOAD LOAD

1 1

2 2

LOAD LOAD LOAD LOAD

z z
Thickness: 6.0mm Thickness: 6.0mm
x Material response: Δ z 12 x Material response: Δ z 12

(a) (b)

LOAD 3

1 2

z z
Thickness: 20.0mm Thickness: 20.0mm
Material response: Δ z A y Material response: Δ y 12
x
(c) (d)

Fig. 2. Specimen size and location from which the material response is measured for each of the mechanical tests considered. (a) Standard Iosipescu shear
test, (b) modified Iosipescu shear test, (c) three-point bend test and (d) off-axis tensile test.

Table 2
Material properties of F593-18 plain weave pre-preg
E11 (GPa) E22 (GPa) E33 (GPa) G12 (GPa) G13 (GPa) G23 (GPa) v12 v13 v23
a a
Reference 54.0 54.0 9.0 4.5 3.0 3.0 0.3 0.25 0.25
10% Increase 59.4 59.4 9.9 4.95 3.3 3.3 Remains constant
20% Increase 64.8 64.8 10.8 5.4 3.6 3.6 Remains constant
30% Increase 70.2 70.2 11.7 5.85 3.9 3.9 Remains constant
Previous assumption 80.0 80.0 9.0 4.5 4.0 4.0 0.3 0.25 0.25
a
Assumed value.
A. Chan et al. / Composite Structures 80 (2007) 396–408 399

due to the lower computational requirements of the C3D8 of the elements in the FE mesh was 0.5 · 0.5 mm. This was
elements. Simulations to test the accuracies of the two ele- maintained for the outer rectangular sections of the FE
ment types were conducted using the standard Iosipescu model. Smaller element sizes were then specified for the
shear test and the three-point bend test. six curves forming the two notches of the specimen to gen-
erate a finer mesh for the section between the notches. The
3.2.1. Validation method FE model was then divided into four equally spaced ele-
FE simulations of the standard Iosipescu shear test and ments in the depth direction.
the three-point bend test were conducted using C3D8 and For the three-point bend test, the default size for the
C3D20R elements. The dimensions of the test specimen, front face (x–z-plane) of the elements in the FE mesh
boundary and loading conditions, and the location of the was 1.0 · 1.0 mm. Due to the symmetry of the model in
material response are shown in Fig. 2a and c for the two the x–z-plane, only half of the specimen was generated
mechanical tests. The parameter b in the standard Iosipe- and the required boundary conditions were applied to sim-
scu shear test was chosen to be 20.0 mm, while a thickness ulate this symmetry plane. As such, it was sufficient to
of 10.0 mm was chosen for the three-point bend test. The divide the FE model into five equally spaced elements in
benchmark material properties as outlined in Table 2 were the depth direction.
used for the validation process.
3.2.3. Validation test results
3.2.2. Numerical models The nodal displacements along the x- and z-axes for the
The FE model generated for this validation test is shown two nodes as shown in Fig. 2a were obtained for the stan-
in Figs. 3 and 4 for the two mechanical tests. For the stan- dard Iosipescu test cases. The subsequent differential dis-
dard Iosipescu shear test, the default size for the front face placement along and x- and z-axes were calculated for

Fig. 3. FE mesh used for the standard Iosipescu shear validation tests.

Fig. 4. FE mesh used for the three-point bend validation tests.


400 A. Chan et al. / Composite Structures 80 (2007) 396–408

both simulations to give the x-axis and z-axis material ments be used. Based on the results achieved in the valida-
responses, respectively. These results, together with the tion study, the use of C3D8 elements in the previously
approximate computation time for each simulation, are sensitivity analysis are justified. Further, also based on
shown in Table 3. Similarly, the z-axis material response the validation results, the FE models used in the follow-
was obtained from the three-point bend test FE models ing sensitivity analysis will be generated using C3D8
using the two different element types. The comparison of elements.
results are shown in Tables 4 and 5.
As can be seen in the tabulated results for both mechan- 3.3. Sensitivity analysis—results
ical tests, the difference in the material responses obtained
from the FE models using the C3D8 and C3D20R elements The current sensitivity analysis was conducted on the
are minimal, with the percentage differences being between four mechanical tests as discussed previously. The speci-
0.19% and 0.41%. This clearly indicates that in the current men dimensions are shown in Fig. 2. The FE models were
FE simulations, the use of C3D8 elements does not affect generated to these dimensions, while boundary and loading
the accuracy of the overall results and that it should be conditions were applied to replicate the support and load-
used in the FE simulations conducted for the sensitivity ing shown in these figures. Different loads were imple-
analysis given the substantial savings in computational mented in each mechanical test. A load of 10.0 kN was
resource and analysis time. The simulations employing applied in the two Iosipescu shear tests, while a load of
C3D8 elements are completed in approximately one fifth 1.0 kN was applied to the three-point bend test. For the
of the time required by the simulations employing off-axis tensile test, a stress of 20.0 MPa was used. The
C3D20R elements for the standard Iosipescu shear test same methodology as used in the initial sensitivity analysis
and one third of the time for the three-point bend test. was used in the current analysis. An in-depth description of
the sensitivity analysis process can be found in Chan et al.
3.2.4. Discussion [13]. The FE simulations were conducted using the new
It is believed that the common problems associated with material properties as discussed above in Section 2.1.
the use of C3D8 elements, such as shear-locking, does not Fig. 5 illustrates the sensitivity results for one of the geo-
affect the results of the current FE models as relatively metric cases considered for the standard Iosipescu shear
small loads are applied. If the test specimens are subjected test. The percentage change in the material response corre-
to higher stresses, it is recommended that C3D20R ele- sponding to the percentage changes in the four stiffness
properties (see Table 2) are shown in this figure. As can
be seen, the relationship between the changes in the
Table 3 mechanical response and the changes in the stiffness prop-
Comparison of results obtained from the Iosipescu shear validation test erties is quite linear. This allows for a comparisons to be
using C3D8 and C3D20R elements made after a 30.0% change in the stiffness properties.
C3D20R C3D8 Difference (%) The percentage changes of the material responses corre-
Number of elements 27,200 27,200 – sponding to a 30% change in the four stiffness properties
Number of nodes 13,2149 35,055 73.47 for each of the geometric cases considered for the four
x-axis material response 24.46 lm 24.56 lm 0.41 mechanical tests are shown in Fig. 6.
z-axis material response 363.03 lm 362.25 lm 0.21
Approximate analysis time 1067.0 s 197.0 s 81.54
3.4. Discussion

From the results of the validation study and the cur-


Table 4
Comparison of results obtained from the three-point bend validation test rent sensitivity analysis, the following conclusions were
using C3D8 and C3D20R elements drawn:
C3D20R C3D8 Difference (%)
• The use of C3D8 elements is valid and appropriate in
Number of elements 7500 7500 –
Number of nodes 37,231 9966 73.23 the current simulations, as little difference is shown in
z-axis material response 573.34 lm 572.27 lm 0.19 the material response obtained from FE simulations
Approximate analysis time 163.0 s 54.0 s 66.87 conducted using C3D8 and C3D20R elements.

Table 5
Experimental loading conditions
Mechanical test Control type Load type Maximum load Maximum average shear stress Load rate
(kN) (MPa) (kN/min)
Modified Iosipescu shear Load Compression 6.0 52.6 1.0
Three-point bend Load Compression 12.0 31.5 2.0
A. Chan et al. / Composite Structures 80 (2007) 396–408 401

30.0

Percentage change in material response (%)


25.0 G13, G23
E1,E2
E3
20.0
G12

15.0

10.0

5.0

0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0
Percentage change in stiffness property (%)

Fig. 5. Sensitivity of the z-axis material response to changes in the four stiffness properties for the standard Iosipescu shear test, b = 20 mm.

• The variations in the material properties used in the two Iosipescu shear test should be used as it eliminates the
sensitivity analyses did not significantly affect the sensi- complexities of having to manufacture the delicate notch
tivity of the four mechanical tests considered. (See Chan in the specimens. It is acknowledge that the stress distri-
et al., 2004 for a comparison of results.) bution would be different between the original and mod-
• Similar sensitivities are achieved in the standard and ified versions of the Iosipescu shear test. However, this
modified Iosipescu shear tests. As such, the modified factor does not affect the suitability of the modified

30.0 30.0
Percentage change in z-axis material

Percentage change in z-axis material

25.62 25.93

25.0 23.58 25.0 23.50


21.88 21.66
20.48 20.20
20.0 20.0
response (%)

response (%)

G13, G23 G13, G23


15.0 15.0
E1, E2 E1, E2
E3 E3
10.0 G12 10.0 G12

5.0 5.0

0.0 0.0
16.0 20.0 24.0 28.0 32.0 36.0 16.0 20.0 24.0 28.0 32.0 36.0
b (mm) b (mm)
(a) (b)
25.0 25.0
Percentage change in z-axis material

Percentage change in z-axis material

20.0 20.0
response (%)

response (%)

15.0 15.0
G13, G23 G13, G23
E1, E2 E1, E2 11.78
10.58
10.0 E3 10.0 E3
8.88
G12 8.37 G12

5.71 5.55
5.0 5.0

2.98
2.33

0.0 0.0
5.0 10.0 15.0 20.0 25.0 30.0 4.0 7.0 10.0 13.0 16.0 19.0
t (mm) (mm)
(c) (d)

Fig. 6. Sensitivity of the material response to a 30% change in the four stiffness properties for the four potential mechanical tests. (a) Standard Iosipescu
shear test, (b) modified Iosipescu shear test, (c) three-point bend test and (d) off-axis tensile test.
402 A. Chan et al. / Composite Structures 80 (2007) 396–408

Iosipescu shear test as the success of the technique is three-point bend test so that all specimens can be cut from
only dependent on the chosen displacement material the same composite block.
response.
• Both the three-point bend test and off-axis tensile tests 4.1. Experimental method
are equally suitable for the proposed technique. Given
that the in-plane Young’s modulus is known for the The tests were conducted on an Instron 4204 universal
material under investigation, their high sensitivity to this testing machine with a 10 kN load cell. The load was
stiffness property is acceptable in both cases. applied to the test specimens at a set loading rate until a
• For the three point bend test, a minimum thickness of specified maximum load was reached. The loading condi-
approximately 20.0 mm should be selected, while an tions used for the two mechanical tests are shown in Table
off-axis angle of greater than 13 if required for the 5. The continuous loading and displacement measurements
off-axis tensile test to be suitable. were logged into a data acquisition computer. A schematic
of the experimental set-up for the three-point bend tests is
Based on these conclusions, a decision was made to illustrated in Fig. 9. A similar set-up was employed for the
apply the inverse parameter identification technique to modified Iosipescu shear test.
the modified Iosipescu shear test and the three-point bend Each experiment was conduced six times to ensure that a
test to determine the elastic interlaminar shear modulus of repeatable, and hence reliable, test procedure is developed.
the Hexcel F-593-18 carbon-composite pre-preg. Due to The specimens were laid-up with Hexcel F593-18 carbon–
a foreseeable complicated process in fabricating test epoxy plain weave pre-preg as [0/90]86, and machined to
specimens, the off-axis tensile test was not selected at this the dimensions shown in Figs. 7 and 8 for the modified
stage. Iosipescu and three-point bend tests, respectively.
Obtaining the required displacement data from the mod-
4. Elastic interlaminar shear modulus ified Iosipescu shear test proved to be considerably difficult.
As illustrated in Fig. 7, the relative displacement of two
The inverse parameter identification technique, using given points on the specimen surface was required for the
both the modified Iosipescu shear and three-point bend duration of the loading. Traditional data acquisition
tests, was applied to determine the elastic interlaminar instrumentations were found to be inadequate in obtaining
shear modulus. The combination of experimental method, the required data. Custom-made clips, as shown in
numerical method and computational method is discussed Fig. 10a, were designed to be used in conjunction with an
in detail in the following sections. Figs. 7 and 8 illustrate Instron extensometer to capture the required displace-
the finalised geometric parameters of the two chosen ments. The clips serve as an extension to the extensometer
mechanical tests. To simplify the fabrication process, a and allow the whole assembly to be attached to the test
decision was made to use a thickness of 19.0 mm for the specimen. Careful considerations of the clip size and design

Fig. 7. Design parameters of the modified Iosipescu shear test. Material response: Dz12 = z1  z2.

LOAD

A
z

Fig. 8. Design parameters of the three-point bend test. Material response: dA.
A. Chan et al. / Composite Structures 80 (2007) 396–408 403

Force
Instron 4204 Output (V)
Control Instron Control Data Acq. Data Acq.
& three-point
Computer Panel Box Computer
bend rig

Probe 1 Probe 2

Fibre
Optic
Sensors
Displacement 2 (V)

Displacement 1 (V)

Fig. 9. Schematic of three-point bend test.

Fig. 10a. Custom-made Iosipescu test clips. Fig. 11a. Holder for the fibre optic sensors.

Fig. 10b. Data acquisition assembly. Fig. 11b. Three-point bend test rig setup.

had to be made in order for it to fit into the Iosipescu shear Data acquisition for the three-point bend test was more
test rig without any of the components interfering with straightforward. The maximum deflection at the mid-span
each other throughout the entire testing process. Fig. 10b of the specimen under load was obtained with a fibre optic
illustrates the assembly of the clips, specimen and test rig. displacement sensor. A special rig attachment was designed

Fig. 12a. FE model of the modified Iosipescu shear test specimen. Front and side views.
404 A. Chan et al. / Composite Structures 80 (2007) 396–408

and manufactured for holding the fibre optic sensor in the viding the required repeatability. The placement of the
correct location (Fig. 11a). This allowed the fibre optic sen- sensor holder relative to the three-point bend test rig is
sor to be in the same position for each specimen, thus pro- shown in Fig. 11b.

Fig. 12b. FE model of the three-point bend test specimen. Front and side views.

PASCAL PROGRAM

Read user file for node &


filename information

Read in experimental
data

Read output file and store the


required nodel displacements

BATCH PROGRAM
Have all of the nodal
NO
displacements been found?
Run FE job in
START

HKS.Abaqus and
generate output file YES
Generate
error file
Read in current material
Run Pascal properties of FE model
Program

Apply the 'secant method' to


determine a new G13/G23
value using the experimental
and numerical data

Does a program
NO termination or error
file exist? Is the percentage change of
the new G12/G23 less than a
specified error-function?

YES

Group the YES NO


END

answer/results
files.

Update material
properties of FE
model
Generate program
termination file

Generate new
input file

Fig. 13. Flowchart of the minimisation program.


A. Chan et al. / Composite Structures 80 (2007) 396–408 405

4.2. Finite element models where DFE and DEXP represents the predicted and measured
displacements, respectively.
FE model were generated to accurately represent the The following nonlinear equation needs to be solved to
various aspects of the actual mechanical test, which minimise the above error function in terms of the interlam-
includes load and boundary conditions, specimen dimen- inar shear modulus:
sions, and known material properties. The stiffness proper-
de
ties shown in Table 1 was applied to the FE models. ¼0 ð2Þ
EDS.FeMap v.8.3 was used as the pre-processor for con- dG13
structing the FE models, and HKS.Abaqus Standard where G13 is the interlaminar shear modulus to be
v.6.4 was used as the FE solver for the analysis. determined.
Mesh density of the FE models can significantly affect It should be noted that Eq. (2) is nonlinear since the dis-
the overall performance of the proposed technique. Accu- placements are related to the interlaminar shear modulus in
rate predictions of the model response cannot be achieved a nonlinear fashion. In the present work, Eq. (2) is solved
with an overly coarse mesh. However, an overly fine mesh iteratively using the secant method [15].
will place an excessively high demand on computing The major functions of the minimisation program have
resources, limiting the overall efficiency of the technique. been written using the Pascal programming language. A
A mesh dependency analysis was conducted for both the batch program has also been written to control the numer-
modified Iosipescu shear test and the three-point bend test ical simulations and the operation of the Pascal program.
to establish the most suitable mesh densities. The final FE These are outlined in the flowchart shown in Fig.13.
meshes are shown in Fig. 12a and 12b. A constant element
size of 0.5 · 0.5 · 1.0 mm was implemented for the modi-
5. Results and discussion
fied Iosipescu test FE model. Whereas a variable mesh size
was adopted for the three-point bend test FE model, allow-
The results obtained from applying the inverse parame-
ing the same accuracy of a much denser model to be
ter identification technique using the modified Iosipescu
achieved while significantly reducing the computational
shear and the three-point bend tests are shown in Figs.
time. Due to symmetry, only one half of the modified Iosip-
14 and 15, respectively. The material responses from all five
escu test specimen and a quarter of the three-point bend
specimens in each of the two mechanical tests are plotted
test specimen were modelled.
against the average shear stress, which is defined as the
applied force over the cross-sectional area of the specimen.
4.3. Computer program (minimisation process)
This data was acquired at a sampling rate of 2.0 Hz, which
was deemed sufficient for material responses at the slow
The core of the inverse parameter identification tech-
rates of load application (1 kN/min and 2 kN/min). The
nique is a computer program that performs the minimisa-
limited spread in the specimen’s material responses in both
tion procedure. The program is used to minimise the
the modified Iosipescu shear test and the three-point bend
following error function:
test indicates that good repeatability have been achieved in
e ¼ jDFE  DEXP j ð1Þ both cases. The consistency of the results also demonstrates

0.25

Specimen 1
0.20
Specimen 2
Material response (mm)

Specimen 3
0.15 Specimen 4
Specimen 5

0.10

0.05

0.00
0.0 10.0 20.0 30.0 40.0 50.0 60.0

-0.05
Average shear stress (MPa)

Fig. 14. Modified Iosipescu shear test experimental results: applied shear stress vs. z-translation between the specified nodes.
406 A. Chan et al. / Composite Structures 80 (2007) 396–408

that the custom-made experimental equipment is robust determining the elastic interlaminar shear modulus. Similar
and can be relied upon for future experiments. nonlinear behaviour is not observed for the three-point
A closer examination of the modified Iosipescu shear bend tests as the interlaminar shear is not the most signif-
test results reveals that the response is slightly nonlinear icant mode of deformation in this case.
from the beginning, and the nonlinearity becomes more The minimisation program was applied using the mate-
obvious when the applied average shear stress reaches rial response obtained from both the modified Iosipescu
approximately 30.0 MPa. This kind of response can be shear test and the three-point bend test at stress levels of
explained by considering the viscoelastic behaviour in the 10.0 MPa, 15.0 MPa, and 20.0 MPa to determine the elas-
interlaminar shear of polymer composites. The interlami- tic interlaminar shear modulus. These results are shown in
nar shear modulus of a composite laminate is a matrix- Figs. 15–17.
dominant property. It has been stated repeatedly that poly- Given the nonlinearity of the material response, the
meric materials can behave in a viscoelastic manner, even modulus values calculated from the modified Iosipescu
at the room temperatures [16,17]. Therefore, this suggests shear tests decreases as the load increases. Whereas, the
that an average shear stress level of less than 30.0 MPa values calculated from the three-point bend tests remain
should be applied to the modified Iosipescu specimens for relatively constant as the loads are varied. The average

0.20

0.00
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0

-0.20
Material response (mm)

-0.40

-0.60

Specimen 1
-0.80
Specimen 2

-1.00 Specimen 3
Specimen 4
-1.20 Specimen 5

-1.40
Average shear stress (MPa)

Fig. 15. Three-point bend test experimental results: applied shear stress vs. mid-span deflection.

4.0

3.5
3.240
3.296
Interlaminar shear modulus (GPa)

3.147
3.0
(Ave=3.22)

2.5

10MPa
2.0 15MPa
3.466

3.412
3.379

3.376
3.331

20MPa
3.281

3.281
3.266

3.205
3.199

3.149
3.068

3.049
3.014

2.941

1.5 Ave (10MPa)


Ave (15MPa)
Ave (20MPa)
1.0

0.5

0.0
1 2 3 4 5
Specimen number

Fig. 16. Interlaminar shear modulus values as calculated by the minimisation program using the three-point bend test.
A. Chan et al. / Composite Structures 80 (2007) 396–408 407

3.0

2.5
Interlaminar shear modulus (GPa) 2.373

2.0

10MPa
1.5 15MPa
20MPa

2.433
2.431
2.427
2.404

2.404

2.404

2.355
2.353
2.348
2.340
2.339

2.340

2.332

2.332

2.332
Ave (10MPa)
1.0 Ave (15MPa)
Ave (20MPa)

0.5

0.0
1 2 3 4 5
Specimen numbers

Fig. 17. Interlaminar shear modulus values as calculated by the minimisation program using the modified Iosipescu shear test.

elastic interlaminar shear modulus values from the five mentally measured and numerically determined material
modified Iosipescu shear tests conducted is approximately response by varying the interlaminar shear modulus in
3.23 GPa, while the average from the three-point bend tests the numerical model. As a consequence, the ‘optimum’
is approximately 2.37 GPa. interlaminar shear modulus can be determined using a rel-
It is noted that there is a difference of approximately atively simple mechanical test. The process involves a com-
26.6% between the interlaminar shear moduli determined bination of experimental investigations, numerical studies,
by the two test methods. This discrepancy can be attributed and the development of a computer program to automate
to several factors. The sensitivity analysis has shown that the iteration process required. In the current investigation,
the three-point bend test is most sensitive to changes in the proposed technique was applied to determine the elastic
the in-plane Young’s modulus, E11, hence slight uncertain- interlaminar shear modulus of the Hexcel F593-18 carbon–
ties in this material property can have a significant impact epoxy pre-preg composite material.
on the interlaminar shear modulus obtained from the test. A sensitivity analysis was initially conducted in the cur-
For example, a 5.0% error in the Young’s modulus will rent investigation on four mechanical tests. From a valida-
affect the interlaminar shear modulus by approximately tion analysis, it was found that the use of C3D8 elements is
9.0%. Another factor affecting the accuracy of the results valid in the current FE models as sufficiently small loads
would be the testing equipment used to measure the mid- are applied to the specimens. Based on the results of the
span deflection. There are inadequacies in the use of the sensitivity analysis, the modified Iosipescu shear and the
Philtec non-contact fibre optic sensors, mainly due to the three-point bend tests were chosen for use with the inverse
fact that any misalignment of the sensor can affect the mea- parameter identification technique.
sured material response significantly. The inverse parameter identification technique was then
Despite inherent possibility of errors in the three-point applied using both the modified Iosipescu shear and three-
bend test, the results obtained from the this mechanical test point bend tests. Data from a series of experimental tests
serves as a good comparison for the interlaminar shear were used in conjunction with FE analyses results through
modulus values obtained from the modified Iosipescu shear a customised computer program to determine the elastic
test. It enhances the degree of confidence in the present interlaminar shear modulus of the composite material
inverse parameter identification approach when the modi- under investigation. This value was found to be 3.23 GPa
fied Iosipescu shear test is implemented is reasonably close and 2.37 GPa from using the modified Iosipescu shear
moduli are obtained. and the three-point bend tests, respectively. These results
were found to be extremely promising as they are within
6. Conclusion an acceptable range of previously speculated values.
A significant contribution has been made in showing
An inverse parameter identification technique has been that the modified Iosipescu shear test can be used to deter-
proposed to determine the elastic interlaminar shear mine the elastic interlaminar shear modulus. The modified
modulus of composite laminates. The basis behind the Iosipescu shear test omits the notches that exist in the stan-
technique is to minimise the difference between an experi- dard Iosipescu test specimens, consequently a much simply
408 A. Chan et al. / Composite Structures 80 (2007) 396–408

preparation is required for the modified Iosipescu shear [7] Morton J, Ho H, Tsai MY. An evaluation of the Iosipescu specimen
test as the process of fabricating the delicate notches is for composite materials shear property measurement. J Compos
Mater 1992;26(5):708–50.
no longer required. [8] Zhou G, Green ER, Morrison C. In-plane and interlaminar shear
The overall outcome of the results suggests that the pro- properties of carbon/epoxy laminates. Compos Sci Technol
posed inverse parameter identification technique can be 1995;55:187–93.
applied to determine the elastic interlaminar shear modulus [9] Lee S, Munro M. Evaluation of testing techniques for the Iosipescu
of composite laminates. The extension of this technique to shear test for advanced composite materials. J Compos Mater
1990;24:419–40.
determine the viscoelastic interlaminar shear modulus is [10] Mespoulet S et al. Design, development, and implementation of test
currently in progress. methods for determination of through thickness properties of
laminated composites. Plast Rubber Compos 2000;29(9):496–502.
[11] Post D, Dai FL, Guo Y, Ifju P. Interlaminar shear moduli of cross-
References ply laminates: An experimental analysis. J Compos Mater
1989;23(3):264–79.
[1] Baker AA et al. An affordable methodology for replacing metallic [12] Pierron F, Grediac M. Identification of the through-thickness moduli
aircraft panels with advanced composites. Compos Part A Appl Sci of thick composites from whole-field measurements using the
Manuf 2002;33(5):687–96. Iosipescu fixture: theory and simulations. Compos Part A
[2] Staab GH. Laminar composites. Oxford: Butterworth–Heinemann; 2000;31:309–18.
1999. p. 102–39. [13] Chan A, Liu XL, Chiu WK. Sensitivity analysis of potential tests for
[3] Hantranft DD. Analysis and experimental characterization of out-of- determining the interlaminar shear modulus of fibre reinforced
plane properties of woven 3-D reinforced composites. PhD University composites. Compos Struct 2004;66:109–14.
of Delaware; 1997. [14] ABAQUS Version 6.5—Analysis: User’s Manual. Vol. 4, Elements,
[4] Walrath DE, Adams DF. The Iosipescu shear test as applied to Chapter 14.1.1.
composite materials. Exp Mech 1983;23(1):105–10. [15] Kreyszig E. Advanced engineering mathematics. 8th ed. John Wiley
[5] Iosipescu N. New accurate procedure for single shear testing of & Sons, Inc.; 1999. p. 845–6.
metals. J Mater 1967;2(3):547–66. [16] Crawford RJ. Plastics engineering. 3rd ed. Butterworth–Heinemann;
[6] Standard test method for shear properties of composite materials by 1998.
the V-notched beam method. ASTM Designation: D 5379/ D 5379M [17] Mills NJ. Plastics: microstructure, properties, and applications.
– 93. second ed. Halsted Press; 1993.

You might also like