Professional Documents
Culture Documents
1 s2.0 096195269400106J Main PDF
1 s2.0 096195269400106J Main PDF
1 s2.0 096195269400106J Main PDF
1995
Copyright fs: 1995 Elsewer Science Ltd
PrInted in Great Britain. All rights reserved
0961.9526/95 $9.50,
IH)
Pergamon
O%l-9526(94) 00106-5
THERMAL
MADE
CONDUCTIVITY
OF COMPOSITE MATERIALS
FROM PLAIN WEAVES AND 3-D WEAVES
Yasser Gowayed and Jhy-Cherng Hwang
1994)
Abstract-In
this research activity, thermal conductivity of textile composites made from plain
weaves and 3-D XYZ weaves is quantified. Plain weave composites are made from E-glass,
KevlaP and AS4 graphite fibers and epoxy resin. 3-D woven composites are manufactured from
Toho graphite fibers and epoxy resin. The effect of fiber type and fiber volume fraction on the
thermal conductivity of textile composites is investigated.
The fabric Geometry Model is adapted to calculate the thermal conductivity of textile
composites. Results using this tool are compared with experimental data and predictions using the
Graphical Integrated Numerical Analysis. The analytical approach proved to be a good engineering tool to predict the thermal conductivity of textile composites.
INTRODUCTION
WORK
Test apparatus
The thermal conductivity chamber used in this study is a k-Matic 75 thermal conductivity instrument manufactured by Holometrix, Inc. The design of this instrument is in
accordance with ASTM Test Method C518 (1985) for steady state heat flux measurements
and thermal transmission properties by means of a heat flow meter.
1177
1178
Guiderod
Movable base
plate
The thermal conductivity measurement for different composite samples is carried out
by placing the sample between the two surface plates of known temperature. Because of
the temperature difference, heat flows through the sample from the hot side to the cold
side. The quantity of heat flow through the sample is measured by the thermal transducer.
Due to the large size of the heat transducer, the specimen dimensions should be at least
20 x 20 cm to minimize heat loss. The size of the transducer ensures the properties of the
composites over the area under consideration. The large size of the sample allows only the
measurement of heat conduction through the thickness of the sample. In-plane thermal
conductivities are not measured.
The instrument is first calibrated using a specimen with known thermal conductivity.
The thermal conductivity k is measured according to the formula:
(1)
where q = heat flow through the sample measured by the thermal transducer (HF),
A = area of the sample, AT = temperature difference (TC2 - TCl) across sample
thickness Ax.
Fig. 2. Micrographic image of balanced plain weave E-glass/epoxy composite with fiber volume
fraction of 34.6%.
Fig. 3. Micrographic image of AScl/epoxy composite with fiber volume fraction of 26.8%.
1179
1180
Fig. 4. Micrographic image of Kevlar-49/epoxy composite with fiber volume fraction of 42.7%.
1181
Materials
Plain weaves. The characteristics of the balanced plain weave fabrics utilized in this
study are as follows:
??
??
??
2 ends/cm 12k E-glass yarns manufactured by PPG Industries. The E-glass fiber
thermal conductivity in the longitudinal and transverse directions, as calculated
from unidirectional test results, is 0.50 (W/mC). Composites tested have total
fiber volume fractions of: 34.6%, 42.6%, 47.0%, 49.7% and 55.7%.
5 ends/cm 3k AS4 graphite yarns manufactured by BP Chemicals. The thermal
conductivities of the fiber (Dasgupta and Agrawal, 1992) are 8.4 (W/mC) along
the fiber direction and 0.84(W/mC) in the transverse direction. Composites
tested have total fiber volume fractions of: 26.8V0, 37.5%, 47.1%, 54.0% and
55.7%.
12 ends/cm lk Kevlar-49@ fibers manufactured by DuPont. The thermal conductivities of the fiber (Kawabata, 1988) are 3.34 (W/mC) along the fiber direction
and 0.212 (W/mC) in the transverse direction. Composites tested have total
fiber volume fractions of: 42.7%, 49.7%, 60.7070, 63.5% and 64.5%.
Epoxy resin. Ciba-Geigy Araldite epoxy resin and HY 956 hardener are used as the
matrix. The mixing is carried out at room temperature with a ratio of 4 : 1 resin : hardener.
The measured thermal conductivity for pure matrix is 0.196 (W/mC).
Composite manufacturing. Fabrics were manually placed in a 38 x 30 cm mold and
impregnated with epoxy resin. The composite plates were cured in a compression molding
machine at 80C for two hours and then post cured at 120C for one and a half hours.
Composite plates were then cut to 25 x 25 cm for thermal conductivity testing.
Micrographical images of balanced plain weaves E-glass/epoxy, AS4/epoxy, Kevlar49/epoxy composites are shown in Figs 2, 3 and 4, respectively. A micrographic image of
a 3-D XYZ graphite/epoxy composite is shown in Fig. 5.
THEORETICAL ANALYSIS
The Fabric Geometry Model , or the Stiffness Averaging Technique, was originally
developed (Kregers and Melbardis, 1978; Whyte, 1986; Pastore and Gowayed, 1994) to
predict the elastic properties of textile composites. The basic idea behind the Fabric
Geometry Model is to treat the fibers and matrix as a set of composite rods having various
spatial orientations. Each composite rod represents a reinforcing system. The local stiffness
tensor for each of these rods is calculated and rotated in space to fit the global composite
axes. The global stiffness tensors of all the composite rods (i.e. reinforcing systems) are
then superimposed with respect to their relative volume fraction to form the composite
stiffness tensor.
In the current research activity, this approach is adapted to predict the thermal
conductivity of textile composites. The analysis steps are illustrated in Fig. 6.
The first step in the analysis is to calculate the local thermal conductivity tensor for
each reinforcing system. The local conductivity tensor could be represented as follows:
K local
=
k 11
k,,
k,,
k21
k 22
k23
ka
ku I
[ k 31
where k, = symmetric conductivity tensor in the i and j directions (i, j = 1,2, 3).
(2)
1182
Direction cosines
Fig. 6. Analysis steps for conductivity predictions using a Modified Fabric Geometry Model.
A large number of unidirectional micro-level thermal analysis models, used to calculate Kj,j, do exist in archived literature (Bruggeman, 1935; Springer and Tsai, 1967;
Behrens, 1968; Hashin, 1979). The authors utilized the model developed by Behrens (1968)
due to the large experimental verification data available in literature to support this model.
The next step is to transform the local thermal conductivity tensor Klocalfor each
reinforcing system to fit the directions of the composite global axes. This is done utilizing
the transformation relation:
K global
T- Kloca,T
(3)
where Kgtobat = global thermal conductivity tensor for the composite rod, T = transformation tensor of direction cosines. T could be represented as:
T =
1;
1;
13
rnt
rni
rn:
n:
n: I
[ nT
(4)
The formulation of the transformation matrix depends on the direction cosines li, mi
and n, (i = 1,2, 3). These direction cosines may be observed as components of unit basis
vectors associated with the fiber axis. A pseudo-code is presented in Pastore and Gowayed
(1994) to calculate Ii, mi and Q.
Finally, the composite thermal conductivity matrix K,somposite
is calculated by superimposing the global conductivity tensor for each reinforcing system Kgloba,with respect to
its relative volume fraction as follows:
K composite
= !
i=O
Kglobal,iei
(5)
where
Plain weaves
Tables l-3 and Figs 7-9 show out-of-plane thermal conductivity test results and
standard deviations, predictions using the modified FGM, detailed in the previous section,
1183
Prediction
Experiment
34.6
42.6
47.0
49.7
55.7
0.269
0.299
0.313
0.337
0.378
f
+
f
+
f
Mod. FGM
GINA
0.268
0.289
0.30
0.307
0.327
0.214
0.295
0.308
0.316
0.336
0.002
0.005
0.005
0.009
0.032
Volum(+fjaction
0
26.8
37.5
47.1
54.0
55.7
0.288
0.336
0.373
0.441
0.453
+
f
+
+
+
Mod. FGM
GINA
0.291
0.338
0.387
0.426
0.437
0.304
0.353
0.41 I
0.456
0.469
0.027
0.028
0.028
0.01
0.005
Experiment
42.7
49.7
60.7
63.5
64.5
0.236
0.246
0.260
0.268
0.272
-Mod.
*-_*
0.45
+
+
f
f
+
Mod. FGM
GINA
0.264
0.276
0.293
0.298
0.299
0.268
0.28
0.299
0.304
0.306
0.001
0.01 I
0.023
0.018
0.013
??Experiment
FGM
GINA
25.0
35.0
45.0
55.0
Fig. 7. Experimental results and analytical predictions for out-of-plane thermal conductivity vs
volume fraction for plain weave E-glass/epoxy composite.
1184
0.50
W
-GINA
??Experiment
Mod. FGM
0.45
o^
5
$
o&l
i
E
0.35
{
h
L
0
0.30
0
0.25
0.20 L
i !5.()
4j.o
35.0
d.0
Fig. 8. Experimental results and analytical predictions for out-of-plane thermal conductivity vs
volume fraction for plain weave AWepoxy composite.
??Experiment
W
Mod. FGM
-GINA
0.20
25.0
35.0
45.0
-55.0
Fig. 9. Experimental results and analytical predictions for out-of-plane thermal conductivity vs
volume fraction for plain weave Kevlar-49/epoxy composite.
and results from the Graphical Integrated Numerical Analysis (GINA) (Gowayed et al.,
1995) for E-glass/epoxy, AS4/epoxy and Kevlar-49/epoxy test specimens detailed in
section on materials, respectively.
The crimp angles used in the modified FGM and GINA approaches as taken from
the micrographical images for E-glass, AS4 carbon and Kevlar-49 balanced plain weaves
are 2.8, 5.0 and 11.3, respectively. The number of finite element divisions for GINA
is 16 subcells.
Tables l-3 and Figs 7-9 show that the out-of-plane thermal conductivity of plain
weaves increases with the increase of fiber volume fraction. The degree of this increase
could also be attributed to fabric crimp angle, value of fiber thermal conductivities, ratio
1185
Experiment
28.77
29.69
38.0
0.3 17 rtr0.024
0.396 + 0.019
0.43 + 0.041
Mod. FGM
GINA
0.267
0.394
0.544
0.274
0.398
0.480
of fiber thermal conductivity along the fiber axis k,,, to that transverse to the fiber axis
k f,t, and ratio of fiber thermal conductivities (kf,,, kf,,) to matrix thermal conductivity.
The AS4/epoxy composite exhibits a pronounced increase in thermal conductivity (57%)
with the increase in fiber volume fraction between 26.8% and 55.7%. This is mainly
attributed to the high kf,, and the 5.0 crimp angle of this fabric. On the other hand,
Kevlar-49/epoxy composites exhibited the least increase in thermal conductivity although
they had the highest crimp angle of 11.3. This is due to the low thermal conductivity in
the transverse fiber direction.
For all practical purposes, GINA and FGM provide a very good estimate for out-ofplane thermal conductivity of the plain weave composite. For E-glass/epoxy and AS4/
epoxy composites, the results are within the experimental error limits. For the Kevlar/
epoxy composite, the percentage of error is around 12%.
A non-linear increase in the thermal conductivity is observed with the increase of
fiber volume fraction of plain weaves. This could be attributed to the high compaction
of fibers at high volume fractions resulting in increased yarn packing factor. Moreover
at high fiber volume fractions, yarns from different fabric layers could touch and form
a heat flow passage reducing the adverse effect of fiber/matrix interface on thermal
conductivity values. Both the modified FGM and GINA do not account for non-linearity
because the micro-level thermal analysis available in literature, and utilized in these
models, do not consider such phenomena. The E-glass/epoxy and the AS4/epoxy plain
weave composites exhibited a higher degree of nonlinearity than the Kevlar-49/epoxy
plain weave composites.
3-D XYZ weaves
Table 4 provides out-of-plane thermal conductivity test results, predictions using the
modified FGM and results from the Graphical Integrated Numerical Analysis (GINA)
(Gowayed et al., 1995) for 3-D XYZ Toho grahite/epoxy weaves. The input for modified
FGM and GINA is taken from the micro-graphical images. The number of finite element
divisions for GINA is 16 subcells.
Results in Table 4 indicate that the out-of-plane thermal conductivity of the graphite/
epoxy composites increases with the increase of the fiber volume fraction in the out-ofplane direction. This is attributed to the high thermal conductivity along the graphite fiber
axis as compared with the thermal conductivity transverse to the fiber direction and the
thermal conductivity of the matrix. Consequently, increasing the volume fraction of
fibers oriented in the out-of-plane direction (V,,) increased the thermal conductivity of the
composite.
As calculated from Table 4, the average error in predictions using GINA is around
8% and the average error for modified FGM predictions for the same samples is 14%.
CONCLUSIONS
1186