Вы находитесь на странице: 1из 9

Composites: Part A 38 (2007) 14931501

www.elsevier.com/locate/compositesa

A method for the approximation of non-uniform ber misalignment


in textile composites using picture frame test
A.S. Milani

a,*

, J.A. Nemes b, R.C. Abeyaratne a, G.A. Holzapfel

Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, USA


b
Department of Mechanical Engineering, McGill University, Montreal, Canada
Department of Solid Mechanics, School of Engineering Sciences, Royal Institute of Technology (KTH), Stockholm, Sweden
Received 23 August 2006; received in revised form 2 January 2007; accepted 7 January 2007

Abstract
Due to the complexity of woven structures, the assumption of perfectly aligned bers for some textile composites is unrealistic. In
more sophisticated material models, therefore, possible ber misalignment is accounted for. On the other hand, non-uniformity of
the misalignment distribution in a fabric may become a second but important problem. This paper presents an inverse methodology from
which a reliable approximation of the non-uniform misalignment state in a woven fabric may be made. Basically, the approximation
requires a representative constitutive model and a set of picture frame tests where ber misalignment plays a key role. Uniaxial and
bias-extension tests are also used to identify the constitutive model parameters independently. The detail procedure is shown for a typical
2 2 twill weave fabric as an illustrative example. Results are discussed and compared to other approaches to reveal the benets and
limitations of the proposed method.
 2007 Elsevier Ltd. All rights reserved.
Keywords: A. Fabrics/textiles; C. Computational modeling; D. Mechanical testing; Fiber misalignment

1. Introduction
The picture frame test has originally been designed as an
alternative for the bias-extension test to obtain a closer
deformation to the pure trellis mode. Unfortunately, however, it has repeatedly been reported that during the picture
frame test, signicant axial forces can be induced in the
direction of tows and this may result in much larger shear
stiness as compared to the bias-extension test [15]. Refs.
[25] connect this eect to the misalignment of bers in the
test conguration and consider the real deformation mode
to be a combination of the trellis and ber extension
modes. In the presence of any misalignment, ber tows
adopt a dierent angle from the pure trellis deformation

*
Corresponding author. Present address: Room 3-173, MIT 77,
Massachusetts Ave. Cambridge, MA 02139, USA. Tel.: +1 617 452
2727; fax: +1 617 258 6156.
E-mail address: milani@mit.edu (A.S. Milani).

1359-835X/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2007.01.008

and, thus, may mismatch the frame motion and stretch


on either side of the crossover points. A rather severe constraint (clamping condition) along the tow directions can
be a further cause of the resulting high tow axial forces
in a picture frame test [1]. Previously, the study [6] had also
concluded that the yarn extensibility assumption can play
an important role in the analysis of the fabrics in dierent
deformation modes.
1.1. Background on ber misalignment in textile composites
To verify the eect of ber misalignment in producing
large forces in the picture frame test conguration, paper
[2] documents tests on specimens that were deliberately
cut at a misalignment of 05. Their results showed that
a 5 misalignment caused high force values similar to what
they had obtained from the original specimens, presumably, with no misalignment. As a result, they concluded
that a misalignment in the range of 5 can be present in

1494

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

the tested plain weave fabric. Using a similar approach, the


study [3] goes further and veries the eect of ber misalignment at a microscopic level for a satin weave fabric.
In that work, the 2.5 deliberately misaligned specimens
were observed to yield much lower weave amplitude for
the tows under tension, and higher for the tows under compression, as compared to bias-extension test specimens.
Accordingly, it was concluded that the excessive load during the picture frame tests is due to the misalignment eect,
estimated in the order of 1 in the original specimens. Note
that the above reported angle ranges can be considered as
an average and the misalignment state may vary from specimen to specimen or from one testing condition to another
(e.g., a more restrictive clamping condition may result in a
higher average misalignment).
For modeling ber misalignments at a continuum level,
there is not much research reported in the textile composites literature. There are some numerical (e.g. [7]) and
analytical (e.g. [8]) models that successfully account for
undulation of ber yarns during a fabric deformation,
however an ideal ber alignment is assumed. Xue et al.
[9] introduced a non-orthogonal constitutive model where
the response of a plain weave fabric was modeled with
respect to the current (deformed) conguration by updating structural directions during the deformation. The
model could also accommodate arbitrary misalignment
angles by applying an additional rotational transformation
to a local ber coordinate system; however some diculties
were observed with respect to an over-estimation of the
misalignment degree that the model should use to reproduce the actual picture frame test data. Milani et al. [10],
using a reference conguration, also reports the eect of
ber misalignment in the predictability of a constitutive
model in the picture frame test. Nevertheless, in both cases,
constitutive models assume a uniform ber misalignment
distribution throughout the fabric; an assumption that
may or may not be realistic depending on the material handling and/or the fabric manufacturing process. Fig. 1

shows an example of a textile composite with apparent


reinforcement irregularities. The detailed modeling of
irregularities in ber reinforcement distributions in a given
fabric can be very complex and generally requires an extensive experimental procedure.
1.2. Objective and methodology
The main goal of this paper is to present a straightforward methodology for the approximation of the non-uniform state of ber misalignment in textile composites in a
macro level. The methodology may then be used to
enhance the predictability of constitutive models, and possibly the numerical codes that use them as material laws.
To show the procedure, as an illustrative example, material
parameters of a hyperelastic constitutive model for a typical 2 2 twill fabric are rst identied. This is done using
test data from uniaxial extension and bias-extension experiments along with an inverse identication method that
accounts for the data variance rising from test non-repeatabilities (Sections 2 and 3). Next, in Section 4, the identied
parameters are used in the constitutive model to reproduce
a set of picture frame test data using arbitrary constant
misalignment angles. In doing so, a range of possible misalignment in the test conguration is extracted. Finally, a
non-uniform state of ber misalignment is approximated
using the results of the previous stage. In dierent sections,
similarities and dierences between the results of the proposed approach and those reported for a plain weave fabric
using a uniform distribution, as well as those reported for
soft biological tissues using a recent analytical approach,
are pointed out. Accounting for such ber misalignment
can provide more accurate constitutive models for use in
simulating material response during forming operations,
such as stamping.
2. Constitutive model
The constitutive model chosen is based on the exponential model proposed in [11,12], and a modication in [13]
with regard to ber-resin interactions. The model in the
in-plane principal directions 1 and 2 is written as [13]
Si

Fig. 1. Example of a woven composite with non-uniform curvy ber


alignment.

1 2
2
2
ki  k2
i kj c1 c2 kj mI 4 I 6  2
k2i
n
o
m
2
2a2i k 1 I 4  1 expk 2 I 4  1  I 1  3
2
n
o
m
2
2
2bi k 1 I 6  1 expk 2 I 6  1  I 1  3 ;
2

where cyclic indices refer to i ! j ! i    = 1 ! 2 ! 1   


(no summation); Si (i = 1,2) are the second PiolaKirchho
principal stresses; k1 and k2 are the related principal
stretches; ai and bi (i = 1, 2) are the cosines of the weft
and warp directions with the corresponding principal directions; c1 and c2 are resin material parameters; k1 and k2 are
ber material parameters; m is a ber-resin interaction
parameter. More closely, c1 and c2 are related to the

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

material properties of the resin material and basically correspond to the molecular chain network characteristics of
the corresponding polymer system. The quantity k1 denotes
a stress-like parameter that relates to the stiness of the
ber material used in the warp and weft family of yarns,
and k2 is a dimensionless parameter that controls the
non-linearity of the ber response under tension, while m
is to account for the ber-resin interface characteristics
such as friction.
At a given temperature and loading rate, the ve
unknown parameters are to be identied by an inverse
method. I4 and I6 are ber invariants that correspond to
the squares of the stretches in weft and warp directions,
respectively [12]. The physically motivated constraint,
namely, the energy can only be stored in stretched bers,
is expressed by the constraint that I4 and I6 must be greater
or equal to one [11]. For a mechanically and geometrically
balanced condition, ai = bi, and I4 = I6. Note that the
stretch value corresponding to the change of fabric thickness k3 can be calculated using the incompressibility
assumption.

1495

Fig. 2. Three repeats of the uniaxial extension test in the warp direction at
a temperature of 170 C and a displacement rate of 162 mm/s (each
curve shows the forcestretch relationship of the test during loading;
displacement at the specimen top edge is d, and the bottom edge is
clamped).

3. Test data used for identication of model parameters


Three repeats of uniaxial extension and bias-extension
tests are conducted at 170 C and a displacement rate of
162 mm/s (see Figs. 2 and 3). In the bias-extension tests,
the variation of the average ber angle / with the loading
direction is monitored to calculate the transverse stretch
values. All tests were on a twill 2 2 fabric (TWINTEX)
performed by the Polymer and Composite Group at IMINRC.1 The in-plane fabric area (length width) in the uniaxial extension tests is 222.25 88.9 mm2, whereas in the
bias-extension tests it is 306.0 88.9 mm2. The fabric thickness is 1 mm.
Following the same inverse identication strategy, as
introduced in [13], the uniaxial test data are used to determine k1 and k2 from a least square regression method. They
are then xed to obtain the c1, c2, and m parameters from the
bias-extension data. Throughout the identication process,
test non-repeatabilities are incorporated using signal-tonoise weights [13] obtained in each test. Hence, the parameters are found to be: k1 = 14.39 MPa; k2 = 1456.87;
c1 = 0.045 MPa; c2 = 0.012 MPa; m = 0.067 MPa.
4. Identication of ber misalignment using the picture
frame test
Having identied the parameters of the constitutive
model from the uniaxial extension and bias-extension tests,
an independent picture frame test is used to identify a
range of possible ber misalignments in the fabric qualitatively and quantitatively. Subsequently, an approximation
of the non-uniform state of ber misalignment may be pos1

Industrial Materials Institute, National Research Council Canada.

Fig. 3. Three repeats of the bias-extension test along with the average ber
angle variation at a temperature of 170 C and a displacement rate of
162 mm/s (each curve shows one repeat of the test during loading;
displacement at the specimen top edge is d, and the bottom edge is
clamped; the ber angle measurements in the three repeats were almost
identical).

sible. The proposed procedure is shown in the following


sections.
4.1. Qualitative identication
General kinematic relationships of the picture frame test
in a reference conguration are presented in the Appendix.
Recall Fig. A1 and consider an arbitrary

 unit
 vector
 in the
T
12 coordinate system as a cos a p4 sin a p4
that corresponds to an initial weft misalignment angle a.
Such misalignment can be due to a combined eect of
inherent random curviness of ber tows, mishandling (misplacement) of the test material inside the frame, preconditioning of the fabric (e.g., by loading and unloading) before
the actual measurement takes place, etc. The invariant I4
can then be expressed as I 4 k21 a21 k22 a22 [12], where a1
and a2 denote the components of the unit vector a. By
using Eq. (A7) in the relation for I4 we get (c is the shear
angle)

1496

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

To simply further, tan a XX 21 , and hence Eq. (4) can be


re-written in a closed form as
 c
c
c

1 sin 2 tan a cos 2
5
a tan
c
c  :
2
cos 2 tan a sin 2
For a given a in Eq. (5), it can be checked that the current
misalignment angle diminishes to 0 when c approches 90.
This variation is non-linear, as shown in Fig. 5.
4.2. Quantitative identication

Fig. 4. Theoretical variation of the ber stretch with deformation for


dierent misalignment angles a in the reference conguration (compare
with Fig. A1).

I 4 1 sin c sin 2a:

Similarly, for a warp misalignment angle b, it can be


written
I 6 1 sin c sin 2b:

It is worth noting that the stretch in the ber direction (i.e.,


the square root of the corresponding invariant) increases
monotonically as the frame moves forward. This is illustrated in Fig. 4 for the weft family of bers. Furthermore,
as addressed earlier, some material models for textile composites are developed in the current (deformed) conguration. In that case, it is important to replace the initial
misalignment angle a by the weft misalignment angle a*,
which is related to the current conguration. The angle
a* may be found by calculating the slope of a material
point on the frame bar using Eq. (A1) (see also Fig. A1).
It follows that

c  x2 X 1 sin 2c X 2 cos 2c
:
tan a

2
x1 X 1 cos 2c X 2 sin 2c

Fig. 5. Theoretical variation of the ber misalignment angle a* in the


current conguration with the shear angle c (compare with Fig. A1). Each
curve corresponds to a given ber misalignment a in the reference
conguration.

4.2.1. Picture frame test data


Three repeats of the picture frame test were conducted
on the same twill fabric by IMI-NRC at 170 C with a
crosshead rate of about 162 mm/min. The bar length was
selected as Lbar = 152.4 mm with an eective (inner) fabric
test area of 139.7 mm 139.7 mm. The measured force
displacement data were corrected for the load acquired
by the unloaded frame [4]. The resulting (net) forcedisplacement results are shown in Fig. 6. When measured data
in Figs. 3 and 6 are compared, it is noticed that in comparison to the bias-extension test, the picture frame test exhibits load values of much higher magnitude, even though the
eective shearing area in both tests is similar. As addressed
earlier in the introduction section, this dierence is believed
to be due to the contribution of tensile loading of high
modulus bers in the picture frame test conguration.
Fiber yarns inside the picture frame apparatus, in contrast
to the bias-extension test, are directly constrained on either
ber end and so even a small amount of misalignment may
yield a signicant contribution to the ber stretching and,
therefore, to a large global load value.
Remark. The picture frame test is known to be highly
non-repeatable, one major source of that being the ber
misalignment [2,9]. To parameterize the non-repeatability
of the test, the signal-to-noise (S/N) statistical measure
may be used as [13]
 Ave 2
Y
S=N 10Log i
;
6
ri

Fig. 6. Three repeats of the picture frame test at a temperature of 170 C


and a displacement rate of 162 mm/s (each curve shows the forcestretch
relationship during loading; displacement at the frame tip is d, and the
bottom tip is xed).

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

1497

where Y Ave
and ri are the mean response and standard devii
ation of the sample at the ith measurement point, respectively. The ensuing variation is shown in Fig. 7. Negative
S/N values indicate highly non-repeatable measurements
in the conducted tests. While the highest S/N ratios occur
at the later stages of deformation, relatively more stable response can be seen in the intermediate stage where the S/N
ratios are less varying. This stage corresponds to the situation in which the initial curvy bers are straightened and the
ber lock-up has not yet occurred [3].

The shear angle c may be directly obtained from the measured displacement values via the test kinematics, Eq. (A6).
For a balanced mode, a = 90  b < 90, Eq. (8) reduces
to
!

L2in T sin 2c
p c c1 sec4 c c2 2 sec4 c  1
F
2Lbar sin 4  2

4.2.2. Forcedisplacement equation


For predicting the picture frame test data, the paper [14]
introduced a method by means of the principle of virtual
work. In this so-called power method, the stress power
induced in the material is assumed to be equal to the rate
of work done by the forces acting on the bars of the picture
frame apparatus. Employing this method in the reference
conguration, by assuming a uniform plane stress deformation (S3 = 0) throughout the sheet, it follows that

From the above closed form, it can be checked that ideally


when a = 90  b = 0 (i.e., no misalignment), the F response decouples from the high ber moduli and the magnitude of force decreases considerably.

F d_ S 1 E_ 1 S 2 E_ 2 V ;

where E_ i are the principal values of the components of the


GreenLagrange strain rate tensor, as given in the Appendix, and V is the deforming material volume in the reference conguration, i.e. V L2in T (here Lin = 139.7 mm).
The initial thickness of the specimen is denoted by T.
Eq. (7) can then be used in conjunction with Eqs. (1)(3)
and Eqs. (A6)(A8) in order to nd a closed form expression for the force response. Thus,
!

L2in T sin 2c

 c1 sec4 c c2 2 sec4 c  1
F
2Lbar sin p4  2c
m
sin c sec2 c2 sec2 c 1sin 2a sin 2b
2
k 1 sin2 2a expk 2 sin2 c sin2 2a

k 1 sin2 2b expk 2 sin2 c sin2 2b
8

m sin c sec2 c2 sec2 c 1 sin 2a



2k 1 sin2 2a expk 2 sin2 c sin2 2a :

4.2.3. Reproducing force response with arbitrary uniform


ber misalignments
Next, Eq. (9) is plotted with the model parameters identied in Section 3 by using arbitrary degrees of misalignment. Results are shown in Fig. 8. It is realized that ber
misalignment, even though small, can signicantly aect
the material response during the rhomboid deformation
of the frame. For instance, only 0.5 dierence of misalignment is enough to cause a signicant divergence in
response. Clearly, without taking into account the misalignment eect (i.e., when a = 0), the constitutive model
is incapable of reproducing most of the experimental
points. Previously, from experimental results, it was concluded that the misalignment eect in the picture frame test
is unavoidable for the same fabric [4]. From the prediction
results in Fig. 8, a misalignment range of 1.55 seems to be
enough to capture individual data points in the entire range
of deformation. The estimated misalignment range for the
twill weave fabric in this work is supported by the range
reported in [2] for a plain weave fabric (i.e., 15).

where a and b are related to an arbitrary weft and warp


ber misalignment angle in the reference conguration.
700

50
S/N

40

Average force

30

Force (N)

500

20

400

10
300

200

-10

100
0
00

S/N (dB)

600

-20
10

19
20

29

3940 49

58 60 68

78 80

-30

Displacement (mm)
Fig. 7. Average response and signal-to-noise (S/N) ratio variation in the
picture frame test (the average values are calculated from Fig. 6).

Fig. 8. Model prediction using arbitrarily chosen misalignment angles


with uniform distribution (the discrete points are test data from Fig. 6;
solid curves are force predictions using Eq. (9) with constant a angles;
variations of force with displacement and shear angle are both illustrated).

1498

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

Fig. 9. An equivalent presentation of non-uniform ber misalignment at a given shear angle c (for simplicity only one family of bers is shown). ai refer to
misalignment angles of individual ber yarns; aeective is the eective misalignment angle in the equivalent representation with uniform distribution.

It is important to notice that the produced curves in


Fig. 8 are based on a state of uniform-misalignment within
the specimens; i.e. constant-misalignment contours. As
seen, dierent parts of some datasets are captured by dierent constant-misalignment contours. Practically speaking,
the uniform state may only be reasonable for specimens
with slightly initial ber curviness. For specimens with an
excessive initial curvy pattern, the straightening of bers
may occur randomly. Hence, during deformation, some
bers are rst straightened and go under tension before
others. In this case, the less curvy bers primarily govern
the response until the others take eect. In turn, the eective (average) ber misalignment angle can vary with the
deformation (see Fig. 9 for a schematic). This may, in turn,
be the motivation for an approximation of the non-uniform misalignment state to enable the constitutive model
to capture the material behavior more closely.
Remark. An interesting link between Figs. 7 and 8 can
be seen. In Fig. 8, it is realized that towards the end of
the deformation, when most bers are straightened, the
majority of data converge to a narrow misalignment range
(around 1.51.6). When Fig. 7 is recalled, the latter conclusion could be linked to the S/N distribution calculated
directly from the test data: the S/N repeatability measure
rises towards the end of deformation. Similar features of
the S/N weights were previously observed [13] for other
deformation modes, namely, the bias-extension and uniaxial extension modes.

7080 mm). Consequently, the following form for the variation of the eective misalignment angle may be taken
aeffective af a0  af ean c ;

where the angles a0 and af are related to the lower and


upper bounds of the misalignment, respectively, a = a0,
when c = 0, and a @ af when c ! 90 and an is attributable
to the extent of curviness in the specimen: the curvier the
initial pattern of the bers in a specimen, the larger the variation of the eective misalignment during the deformation.
As shown in Fig. 10a, Eq. (10) can accommodate larger

800
Test 1
Test 2
Test 3
Model

700
600
500

F (N)

4.2.4. Approximation of a non-uniform ber misalignment


state
For modeling the variation of eective ber misalignment with deformation, an approximate form may be
obtained based on the results of the previous section. From
Fig. 8, it is realized that the eective misalignment angle
starts o with a large range and drops down to a narrower
range as the deformation proceeds. This regime seems to be
exponential because, given a test repeat, for the same
amount of deformation (e.g., 10 mm crosshead displacement), the change of the eective misalignment angle in
an earlier stage of the test is much larger (e.g., 2 between
2030 mm) as compared to a later stage (e.g., 0.1 between

10

400
300
200
100
0
0

20

40

60

80

100

d (mm)
Fig. 10. (a) Variation of an approximated eective misalignment, and (b)
model prediction for each individual data set using a non-uniform
misalignment state (the discrete points are test data from Fig. 6; solid
curves are force predictions using Eq. (10) in (9)).

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

5. Conclusions

1.035
1.03
Fiber stretch

1499

A methodology is presented for introducing the possible


eect of non-uniform ber misalignment in a constitutive
approach. The method may be summarized as follows:

1.025
1.02
1.015
Test 1
Test 2
Test 3
Unifrom 1.5 deg.
Unifrom 5 deg.

1.01
1.005
1
0

20

40
Shear angle ()

60

80

Fig. 11. Calculated ber stretch in the conducted picture frame tests using
a non-uniform misalignment state (the dashed and solid lines show the
asymptotes).

misalignment variation in the earlier stages of deformation


(a ! a0), and less variation towards the end of deformation
(a ! af) where the misaligned angles converge. Ideally,
provided all ber tows are initially straightened and aligned
in parallel, it can be written a = a0 = af, and the model reduces to its original form for a uniform misalignment state.
To evaluate the viability of the above approximated
form, Eq. (9), with the same material parameters but
substituting Eq. (10) for a, is plotted in Fig. 10b. The an
exponent for each of the three data sets (Test 1, Test 2,
and Test 3) is found to be 3.81, 3.87, and 4.58, respectively.
Using the non-uniform misalignment state, the variation of
the test data can be captured for all data sets, by at least
0.995%. Finally, Eqs. (2) and (10) can be combined to
arrive at a modied invariant that one may need to use
in a constitutive approach, i.e.
I 4 k2fiber 1 sin c sin 2af a0  af ean c :

11

For the three repeats of the picture frame test, a plot of the
ber stretch using Eq. (11) is shown in Fig. 11. According
to this gure, the range of induced ber stretch in the picture frame test is comparable to that of the uniaxial extension (2%), as seen in Fig. 2.
It is worth adding that there are some similarities
between a recent work [15] and the present approach. In
[15] it is shown that to account for the non-uniform distribution of collagen bers in arterial walls, a generalized
structural tensor should be used to modify the induced
ber strain (or stretch) in a constitutive model. Accordingly, for transversely isotropic materials, in [15] a new dispersion parameter is introduced and treated as a structure
parameter. Here for a woven fabric, according to Eq. (11),
a similar modication seems to be necessary to account
for the eect of non-uniform E-glass ber distribution.
The dierence seems to be that for woven fabrics more
structural parameters are needed for the inclusion in a constitutive equation in order to capture the true material
response.

 Choose a representative constitutive model.


 Obtain unknown model parameters from sets of uniaxial and bias-extension tests, including the test nonrepeatibilites.
 Use picture frame tests to estimate a range of ber misalignment by reproducing the forcedisplacement data
points with the model using a uniform distribution
assumption; it is to note that the picture frame test as
it is currently carried out with rigid constraints may
not be suitable for the purpose of characterizing the
shear behavior of the fabric itself because the response
can be signicantly aected by the misalignment noise
factor.
 Approximate a non-uniform eective misalignment variation, similar to that in Eq. (10), based on the results of
the previous stage.
For the TWINTEX twill weave fabric used in this
work, the majority of picture frame test data towards
the end of the deformation, when the bers are straightened, can be predicted by a common level of misalignment
(around 1.5). On the other hand, the overall prediction
results, particularly in the early stages of deformation,
emphasize the possibility of a non-uniform misalignment
state within some specimens. This may be related to the
initially randomly distributed curviness of bers. Due to
this eect, some bers may come into tension before others and so the eective misalignment can vary from one
stage of deformation to another. Although the precise
form of the misalignment distribution inside specimens is
complex and remains to be studied from an extensive statistical and experimental procedure, the approximated
form seems to represent the ensuing variation in the eective ber misalignment, which one may need to employ in
a constitutive approach. The application of the same or a
similar misalignment model for other types of fabrics and
constitutive laws would be interesting. It is believed that
the account of a non-uniform ber misalignment becomes
more signicant as the weave structure becomes more
complex. Finally, having identied a constitutive model
and a ber misalignment state from basic standard tests,
it would be desirable to study the eect of misalignment
in a real forming process where the clamp condition
can be more complex than that in the picture frame
conguration.
Acknowledgements
Contributions of Drs. M. Bureau, X-T. Pham, G. Lebrun, R. Diraddo and J. Denault from IMI-NRC in supervising the experimental work and providing useful

1500

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501

Fig. A1. Coordinate systems employed in the picture frame test analysis (a and 90  b can be considered as the initial weft and warp misalignment angles,
respectively) [10].

discussions are greatly acknowledged. The rst author


would also like to thank the Natural Sciences and
Engineering Research Council of Canada for nancial
support.

Appendix. Picture frame test kinematics in the reference


conguration
For the picture frame test, the material X1  X2 and spatial x1  x2 Cartesian coordinate systems can be taken as
shown in Fig. A1. Accordingly, assuming a continuum
manifold and material incompressibility, the relationship
between a material point P in the reference conguration
and its one-to-one mapping p = v(P) in the current conguration can be written as [16]
8
x1 X 1 cos 2c X 2 sin 2c ;
>
>
<
A1
x2 X 1 sin 2c X 2 cos 2c ;
>
>
:
x 3 k3 X 3 :
The through the thickness stretch ratio k3 in the above relation can be found using the incompressibility constraint.
Using Eq. (A1), the deformation gradient F reads in matrix
notation
2
3
cos 2c sin 2c 0


6
7
oxi
c
c
7
F
A2
6
4 sin 2 cos 2 0 5;
oX j
0
0
k3
and the incompressibility condition det[F] = 1 gives
k3 = sec c [16]. Next, the right CauchyGreen tensor C,
the GreenLagrange strain tensor E, and its rate E_ are
given in the matrix representations [10]
2
3
1
sin c
0
6
7
T
C F F 6
1
0 7
A3
4 sin c
5;
2
0
0
sec c

sin c

1
16
7
0
0 5;
E C  I 4 sin c
2
2
0
0
tan2 c
2
3
0
cos c
0
7
_ c_ 6
cos
c
0
0
E
4
5:
2
2
0
0
2 tan c sec c

A4

A5

According to the four-linkage mechanism of the picture


frame apparatus (assuming rigid bars), if the displacement
at the end of the frame is denoted by d, and the length of
each bar by Lbar, one can relate the displacement measurement d to the shear angle c as [16]
!
p
p
d 2Lbar
c  2 arccos
:
A6
2
2Lbar
_ in
In addition, by calculating the eigenvalues of [C] and E
Eqs. (A3) and (A5), respectively, the in-plane squares of the
principal stretches and the rate of the GreenLagrange
principal strains are given by
k21 1  sin c;

k22 1 sin c;

A7

and
c_
E_ 1 E_ 2 cos c:
2

A8

Note that the subscripts in Eqs. (A7) and (A8) refer to the
principal directions 1 and 2, as shown in Fig. A1.
References
[1] Sharma SB, Sutclie MPF, Chang SH. Characterisation of material
properties for draping of dry woven composite material. Compos
Part A 2003;34:116775.
[2] Chen J, Lussier DS, Cao J, Peng XQ. Materials characterization
methods and material models for stamping of plain woven composites. Int J Forming Process 2001;4:26983.
[3] Chang SH, Sharma SB, Sutclie MPF. Microscopic investigation of
tow geometry of dry satin weave fabric during deformation. Compos
Sci Technol 2003;63:99111.

A.S. Milani et al. / Composites: Part A 38 (2007) 14931501


[4] Lebrun G, Bureau MN, Denault J. Evaluation of bias-extension and
picture frame test methods for the measurement of intraply shear
properties of PP/glass commingled fabrics. Compos Struct 2003;
61:34152.
[5] Potter K. Bias extension measurements on cross-plied unidirectional
prepreg. Compos Part A 2002;33:6373.
[6] Mohammed U, Lekakou C, Dong L, Bader MG. Shear deformation
and micromechanics of woven fabrics. Compos Part A 2000;31:
229308.
[7] Page J, Wang J. Prediction of shear force using 3D non-linear FEM
analyses for a plain weave carbon fabric in a bias extension state.
Finite Elem Anal Des 2002;38:75564.
[8] Nguyen M, Herszberg I, Paton R. The shear properties of woven
carbon fabric. Compos Struct 1999;47:76779.
[9] Xue P, Peng XQ, Cao J. A non-orthogonal constitutive model
for characterizing woven composites. Compos Part A 2003;34:
18393.
[10] Milani AS, Nemes JA, Pham XT, Lebrun G. The eect of ber
misalignment on parameter determination using picture frame test.
Society of Manufacturing Engineering, Technical Paper, ID:
TP03PUB236, 2003. p. 10.

1501

[11] Holzapfel GA, Gasser TC, Ogden RW. A new constitutive framework for arterial wall mechanics and a comparative study of material
models. J Elasticity 2000;61:148.
[12] Holzapfel GA, Gasser TC. A viscoelastic model for ber-reinforced
composites at nite strains: Continuum basis, computational aspects
and applications. Comp Meth Appl Mech Eng 2001;190:4379430.
[13] Milani AS, Nemes JA. An intelligent inverse method for characterization of textile reinforced thermoplastic composites using a hyperelastic constitutive model. Compos Sci Technol 2004;64:156576.
Bradaigh CM. A
[14] McGuinness GB, Canavan RA, Nestor TA, O
picture-frame inraply shearing test for unidirectional and fabric
reinforced composite melts. In: Proceedings of the ASME materials
division, international mechanical engineering congress and exposition, vol. 69-2. San Francisco; 1995. p. 110718.
[15] Gasser TC, Ogden RW, Holzapfel GA. Hyperelastic modelling of
arterial layers with distributed collagen bre orientations. J Royal
Society Interface 2006;3:1535.
Bradaigh CM. Development of rheological
[16] McGuinness GB, O
models for forming ows and picture-frame shear testing of fabric
reinforced thermoplastic sheets. J Non-Newton Fluid Mechan
1997;73:128.

Вам также может понравиться