Materials and Design 31 (2010) 837–849

Contents lists available at ScienceDirect

Materials and Design

journal homepage: www.elsevier.com/locate/matdes

A study on a possible correlation between thermal conductivity and wear

resistance of particulate filled polymer composites
Amar Patnaik a,*, Md Abdulla a, Alok Satapathy b, Sandhyarani Biswas b, Bhabani K. Satapathy c
a
Mechanical Engineering Department, NIT, Hamirpur 177 005, India
b
Mechanical Engineering Department, NIT, Rourkela 769 008, India
c
Centre for Polymer Science and Engineering, IIT, Delhi 110 016, India

a r t i c l e i n f o a b s t r a c t

Article history: In this study, thermal conductivity of particulate filled polymer composites is investigated experimen-
Received 9 June 2009 tally by guarded heat flow meter method in accordance with ASTM E 1530. The objective of this research
Accepted 28 July 2009 is to attain a better understanding of global and local thermal behavior of a composite system through
Available online 6 August 2009
basics of thermal properties. The results obtained from this study are compared with an existing theoret-
ical model for validation purposes of pine bark reinforced epoxy based resin composites with and with
Keywords: out particulate filler content. To this end, a correlation has been made between the thermal conductivity
Polymer composite
and specific wear resistance of the tested samples, highlighting the importance of the insulating material.
Thermal conductivity
Particulates
The relationship between control factors and performance outputs are established by means of non-lin-
Specific wear rate ear regression analysis resulting in a valid mathematical model. Finally, a popular evolutionary approach
known as genetic algorithm (GA) is used to generalize the method of finding out optimal factor settings
for minimization of thermal conductivity and specific wear rate.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction
Commonly used polymer materials such as polyethylene (PE),
polypropylene (PP), polyvinyl chloride (PVC) or polyamide are
good electrical and thermal insulators. Recent applications of poly-
mers as heat sinks in electronic packaging require new composites
with relatively high thermal conductivity. Improved thermal con-
ductivity in polymers may be achieved either by molecular orien-
tation or by the addition of conductive fillers. During the last few
decades, however, in the microelectronic systems, great effect
has been contributed to improving higher integration density, fas-
ter performance, miniaturization of electronic devices and lower
cost [1,2]. Therefore, the power density in the electronic devices
is becoming larger and larger. That turns the researchers’ focus
to be placed on the thermal conductivity in order to get heat-dis-
sipating composites. The prediction of thermal conductivity of
composites comprises a significant portion of the heat transfer lit-
erature. Many reports concerning about the thermal conductivity
of polymer composites, associated with various thermal conduc-
tive models or equations for predicting the thermal conductivity,
have been published. They are either theoretically based or are
empirical which means to include one or more experimentally
determined (or empirical) parameters. A good overview has been
* Corresponding author. Tel.: +91 9437188085.
E-mail address: amar_mech@sify.com (A. Patnaik).
given by Progelhof [3]. Procter and Solc [4] used Nielsen model
as a prediction to investigate the thermal conductivity of several
types of polymer composites filled with different fillers, and con-
firmed its applicability. Nagai [5] found Bruggeman model for
Al2O3/epoxy system and a modified form of Bruggeman model
for AlN/epoxy system are both good prediction theories for thermal
conductivity. Bollampally and Wong [6] found that Agari model
predicts better than Maxwell model for thermal conductivity of
SiO2/epoxy composite at high percentage of filler.
However, it is not always possible to manufacture polymer ob-
jects of a desired shape with a prescribed orientation value and in
the desired direction. Hence a more practical method to increase
the thermal conductivity of a polymer is needed. The addition of
thermally conductive particles or short fibers to the polymers dur-
ing the injection molding process seems to be a good method to
obtain conductive polymers. Such filled polymers with higher
thermal conductivities than unfilled ones are becoming a more
important area of study because of the wide range of applications,
e.g., in electronic packaging in applications with decreasing geo-
metric dimensions and increasing output of power, like in com-
puter chips or in electronic packaging [7,8].
In this study, thermal conductivity of particulates filled fiber
reinforced polymer composites is investigated by two different
manufacturing methods. The results clearly show that the in-
creased filler concentration leads to higher thermal conductivity
that can be correlated to the reduced dry sliding wear rate of the

0261-3069/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2009.07.046
838 A. Patnaik et al. / Materials and Design 31 (2010) 837–849
composites. Also, it is observed that the intrinsic thermal conduc-
tivity of fillers is less important than filler–matrix bonding, as the
latter affects strongly the thermal conductivity of the composite,
hence the resistance to wear. At the end, an optimization technique
is applied to addresses the relationship between thermal conduc-
tivity and the specific wear resistance of the worn composites un-
der different operating conditions.
2. Materials
2.1. Matrix material
Unsaturated isophthalic polyester resin (Elastic modulus
3.25 GPa, density 1.35 g/cc) is the matrix material. Two percent
of cobalt nephthalate (as accelerator) is mixed thoroughly in iso-
phthalic polyester resin and then 2% methyl-ethyl-ketone-perox-
ide (MEKP) as hardener is mixed in the resin prior to
reinforcement. Epoxy LY 556 resin, chemically belonging to the
‘epoxide’ family is used as the matrix material. Its common name
is Bisphenol A Diglycidyl Ether. Vinyl ester resins (EPN-1138) are
addition products of various epoxide resins and unsaturated
monocarboxylic acids, most commonly methacrylic acid. The low
temperature curing epoxy resin (Araldite LY 556) and correspond-
ing hardener (HY951) are mixed in a ratio of 10:1 by weight as rec-
ommended. The polyester, epoxy and Vinyl ester resins and the
respective hardeners are supplied by Ciba Geigy India Ltd.
2.2. Fiber materials
In this work, three different types of fibers are taken for exper-
imental purposes. Out of which two conventional fibers (woven
roving E-glass fibers and carbon fiber) and one natural fiber (pine
bark) have been used as the reinforcing material in all the compos-
ites. Natural fibers are becoming potential alternatives for glass fi-
ber reinforced composites in many applications [9].
2.3. Filler materials
Two conventional fillers: aluminium and copper powder and
one industrial waste: cement by-pass dust (CBPD) are chosen to
be used as particulate fillers in pre-determined proportions in var-
ious composites prepared for this investigation. While, aluminium
and copper have conventionally been used in many composite
applications, the industrial waste cement by-pass dust (CBPD)
can be considered as non-conventional materials for use as fillers
in polymer composites. It is a fine powdery material similar in
appearance to Portland cement. It is generated during the calcining
process in the kiln. As the raw materials are heated in the kiln, dust
particles are produced and then carried out with the exhaust gases
at the upper end of the kiln. These gases are cooled and the accom-
panying dust particles are captured by efficient dust collection sys-
tems. Lime (CaO) constitutes about 40% of CBPD composition.
Other compounds include SiO2, Al2O3, Fe2O3, K2O, Na2O and Cl.
2.4. Composite preparation
E-glass fiber/carbon fiber are reinforced with respective resin
also added particulate filler materials to prepare samples with dif-
ferent wt%. All these filler materials are thoroughly dispersed in the
resin by mechanical stirring. The resins used for this study were vi-
nyl ester, polyester and epoxy. Test specimens are prepared using
0%, 4%, and 8% of filler by weight percentages are taken in this work
to determine conductivity properties in three different directions
(i.e. longitudinal, transverse and through-the-thickness) and spe-
cific wear rate of the composites. E-glass fiber/carbon fiber/pine
Table 1
List of thermal conductivity test sample directions.
Effect of different materials and manufacturing methods
Compression molding/hand lay-up techniques
Glass fiber/carbon fiber reinforced Pine bark reinforced
composites composites
L T Th L T Th
L: Longitudinal direction, T: transverse directions and Th: through thickness.
bark is reinforced in three different resins to prepare the composite
with and without filler content by two different manufacturing
techniques such as conventional hand-lay-up technique and com-
pression molding process, respectively. Herein, the thermal con-
ductivity measurement in the transverse direction means that a
sample is prepared in such a way that heat flow direction is trans-
verse to primary fiber orientation of the sample; similarly, the
thermal conductivity measurement in the longitudinal direction
means heat flowing along the fiber direction; and the thermal con-
ductivity measurement in the thickness direction means heat flow-
ing through-the-thickness direction. The thermal conductivity of
an anisotropic composite material depends on the resin nature, fi-
ber type and architecture, fiber weight fraction, manufacturing
technique, direction of heat flow and operating temperature, lead-
ing to a high degree of complexity.
2.5. Thermal conductivity test matrix
The research proceeded as outlined in Table 1. The test matrix
has included various material parameters and process parameters,
leading to a spectrum of composite samples for their thermal con-
ductivities along 0° and 90° fiber orientations and through-the-
thickness direction. It should be noted that sample preparation is
the most challenging part of the entire process. Many tools are
used specially for sample preparation in addition, attention to
environmental and health safety issues are also needed. The test
samples are prepared in the form of 10 cm rectangular plate using
compression molding and hand lay-up methods. Ideally all sam-
ples for three-dimensional measurements should be cut from same
block by diamond cutter. Technically this poses a great difficulty in
manufacturing such a block in terms of proper alignment of fabric
and wet out.
Typically, FRP laminate has a thickness of 3–8 mm. A 10 cm
rectangular plate is cut from this type of laminate to measure its
thermal conductivity with heat flowing through-the-thickness
direction. Both the above discussed approaches are used to prepare
samples in the present study. As a result, we have successfully pre-
pared a couple of samples that allow for thermal conductivity mea-
surements in transverse, longitudinal, and through-the-thickness
directions.
3. Experimental approach for thermal conductivity
measurements
There are various test methods available for thermal
conductivity measurement and each method depends mainly on
configuration of a material and job requirement. One principal
experimental method for determinations of thermal conductivity
is discussed below:
3.1. Guarded heat flow meter
This method is based on two dimensional steady state tech-
niques and is used to measure and compare thermal properties
of materials under controlled conditions and their ability to
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849
maintain required thermal conductance levels [9]. The specimen
and a heat flux transducer (HFT) are sandwiched between two flat
plates controlled at different temperatures, to produce a heat flow
through the stack. A cylindrical guard surrounds the test stack and
is maintained at a uniform mean temperature of the two plates, in
order to minimize the lateral leak of heat. At steady state, the dif-
ference in temperature between the surfaces contacting the speci-
men is measured with temperature sensors embedded in the
surfaces, together with the electrical output of the HFT. The output
voltage is proportional to the heat flow through the specimen, HFT
and the interfaces between the specimen and the apparatus. The
coefficient of thermal conductivity can be obtained by prior cali-
bration of system with the specimens of know thermal
conductivity.
3.2. Experimental set-up
A guarded heat flow meter method has been developed for ther-
mal conductivity measurements. This is achieved by using a ther-
mal conductivity testing system Unitherm model 2022 from
ANTER Corp., Pittsburgh, PA. This unit is supplied with a mid range
flux module covering a thermal resistance range from 0.01 to
0.05 m2 K/W and is able to measure the thermal conductivity of
materials in the range of 0.1–40 W/m K in terms of test standard
ASTM E 1530. The materials that can be tested include metals,
ceramics, polymers, composites and glass. The test samples need
to be prepared in a form of 2-in. diameter circular discs with their
thickness depending on the materials’ thermal conductivity. The
thermal conductivity machine is supplied with three different sets
of calibration samples span the operating range (R) from 0.0005 to
0.01, 0.002 to 0.02 and 0.01 to 0.05 m2 K/W. These samples are
tested for conductivity values and compared with the given values
by the manufacturer for calibration purpose.
A schematic picture shown in Fig. 1 represents the system func-
tioning in detail. The assembly is a stack of parts with different
functionalities. The heater on top and bottom helps to maintain
steady state heat transfer through the sample, two polished sur-
faces on top and bottom the sample transfer heat from top and bot-
tom of heaters with reduced thermal resistance through surface. A
reference calorimeter is placed under the lower plate, which acts as
a heat flux transducer. The heat sink at the bottom avoids excessive
temperature from the system. The sample is compressed in be-
Pneumatic load
Top heater
Upper plate (TU ) Guard heater
Test sample
Lower plate (Tm )
Heat flux transducer
Reference
Calorimeter
(TL)
Bottom heater
Heat sink
Fig. 1. Schematic model showing the system arrangement in Unitherm 2022.
839
tween the polished surfaces, each controlled at different tempera-
tures, using pneumatic load. The pressure is maintained at 10 psi
using pressurized air supply. As an option, coolant water is circu-
lated through heat sink. A circular low temperature heat insulation
ring is wrapped around the lower stack of the assembly to restrain
heat flow to outside atmosphere. The entire system is maintained
in a thermally insulated glass chamber. Unitherm is completely
automatic while testing as the apparatus is completely controlled
by a computer. The test system is hooked to the computer by
means of an USB cable.
Considering the operation of the system, the sample to be tested
is prepared into a flat surface on both sides and thermal compound
is applied on the sample to reduce thermal resistance caused due
to surface roughness. Then the prepared sample is placed in be-
tween two polished surfaces and a pneumatic pressure of 10 psi
is applied on the top portion of the stack. The sample can be tested
in the temperature range from 50 °C to 300 °C. For steady state
heat transfer, the user can divide the testing into different zones,
called as set point temperatures. In this thesis the testing was di-
vided into three temperature zones i.e. 60, 80, and 100 °C. For high-
er thermal conductivity materials, a difference of 50° is
recommended for the machine. At every set point temperature,
the system checks for steady state heat flow through the sample
and thermal conductivity is measured.
3.3. Operation principle of Unitherm 2022
By definition thermal conductivity means ‘‘The material prop-
erty that describes the rate at which heat flows with in a body
for a given temperature change.” For one-dimensional heat con-
duction the formula can be given as
T1  T2
Q ¼ KA ð1Þ
x
where Q is the heat flux (W), K is the thermal conductivity (W/m K),
A is the cross-sectional area (m2), T1T2 is the difference in temper-
ature (K), x is the thickness of the sample (m). The thermal resis-
tance of a sample can be given as
T1  T2
R¼ ð2Þ
Q =A
where R is the resistance of the sample between hot and cold sur-
faces (m2 K/W). From Eqs. (1) and (2) we can derive that
x
K¼ ð3Þ
R
In Unitherm 2022 the heat flux transducer measures the Q value
and the temperature difference can be obtained between the upper
plate and lower plate. Thus the thermal resistance can be calcu-
lated between the upper and lower surfaces. Giving the input value
of thickness and taking the known cross-sectional area, the ther-
mal conductivity of the samples can be calculated using Eq. (3).
3.4. Dry sliding wear test apparatus
To evaluate the performance of these composites under dry
sliding condition, wear tests are carried out in a pin-on-disc type
friction and wear monitoring test rig (supplied by DUCOM) as
per ASTM G 99. The schematic diagram of pin-on-disc set-up is
shown in Fig. 2. The counter body is a disc made of hardened
ground steel (EN-32, hardness 72 HRC, surface roughness
0.6 lRa). The specimen is held stationary and the disc is rotated
while a normal force is applied through a lever mechanism. A ser-
ies of test are conducted with three sliding velocities of 210, 261
and 314 cm/s under three different normal loading of 10 N, 20 N
and 30 N. The material loss from the composite surface is
840 A. Patnaik et al. / Materials and Design 31 (2010) 837–849

Fig. 2. The schematic diagram of pin-on-disc set-up.

measured using a precision electronic balance with accuracy B(2)

±0.1 mg and the specific wear rate (mm3/N m) is then expressed
on ‘volume loss’ basis as
W s ¼ Dm=qtV s  F n ð4Þ (3,4) (6,7)
where Dm is the mass loss in the test duration (g), q is the density
of the composite (g/mm3), t is the test duration (s), Vs is the sliding D(9) (10) (12) (13)
velocity (m/s), and FN is the average normal load (N).
The specific wear rate is defined as the volume loss of the spec- (8,11)
A(1) C(5)
imen per unit sliding distance per unit applied normal load.
Fig. 3. Linear graphs for L27 array.
3.5. Taguchi experimental design

Taguchi design of experiment is a powerful analysis tool for
modeling and analyzing the influence of control factors on perfor-
mance output. The most important stage in the design of experi-
ment lies in the selection of the control factors. Therefore,
initially a large number of factors are included so that non-signif-
icant variables can be identified at earliest opportunity. In the pres-
ent work, the impact of five such parameters are studied using L27
(313) orthogonal design. The operating conditions under which ero-
sion tests are carried out are given in Table 2.
Four parameters viz., sliding velocity, normal load, filler content
and sliding distance, each at three levels, are considered in this
study. Four parameters each at three levels would require 34 = 81
runs in a full factorial experiment. Whereas, Taguchi’s factorial
experiment approach reduces it to 27 runs only offering a great
advantage.
The experimental observations are transformed into a signal-to-
noise (S/N) ratio. There are several S/N ratios available depending
on the type of characteristics. The S/N ratio for minimum erosion
rate coming under smaller is better characteristic, which can be
calculated as logarithmic transformation of the loss function as
shown below.
where ‘n’ the number of observations, and y the observed data.
‘‘Lower is better” (LB) characteristic, with the above S/N ratio trans-
formation, is suitable for minimizations of erosion rate. The stan-
dard linear graph by Glen [10] as shown in Fig. 3, is used to
assign the factors and interactions to various columns of the orthog-
onal array.
The plan of the experiments is as follows: the first column was
assigned to sliding velocity (A), the second column to normal load
(B), the fifth column to filler content (C) and the ninth column to
sliding distance (D), the third and fourth column are assigned to
(A  B)1 and (A  B)2, respectively to estimate interaction between
sliding velocity (A) and normal load (B), the sixth and seventh col-
umn are assigned to (B  C)1 and (B  C)2, respectively, to estimate
interaction between the normal load (B) and filler content (C), the
eighth and eleventh column are assigned to (A  C)1 and (A  C)2,
respectively, to estimate interaction between the sliding velocity
(A) and filler content (C). The remaining columns are assigned to
error columns, respectively.
4. Experimental results and discussion

S 1 X 2  4.1. Neat resin castings and resin castings with fillers

Smaller is the better characteristic : ¼ 10 log y ð5Þ
N n
The thermal conductivity of a composite depends on many
parameters including: (1) fiber orientation; (2) moisture content;
Table 2
(3) distribution of moisture; (4) void percentage, (5) resin type
Levels for various control factors. and (6) filler percentage. The parameters of major influence on
thermal conductivity are fiber weight percentage and conductivity
Control factor Level Units
properties of both resin and fiber. In the present study, thermal
I II III conductivities of different resins like vinyl ester, polyester and
A: Sliding velocity 210 261 314 cm/s epoxy are determined as shown in Table 3 and the results are in
B: Normal load 10 20 30 N agreement with the literature [11]. From this analysis neat epoxy
C: Filler content 0 4 8 wt%
shows better thermal conductivity as compared with other two
D: Sliding distance 3000 4000 5000 m
resins. Hence, epoxy resin is chosen for further study.
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849 841

Table 3 0.45
Thermal conductivity values of different resins.
0.40

Thermal conductivity (W/m-K)

Sample Thickness Thermal conductivity (W/m K)
(mm) 0.35
Test-1 Test-2 Test-3
(60 °C) (80 °C) (100 °C) 0.30
Neat polyester resin 0.567 0.196 0.215 0.240
0.25
Neat epoxy resin 0.743 0.221 0.237 0.263
Vinyl ester resin 0.687 0.185 0.190 0.213 0.20
0.15 4 wt% copper filled epoxy composite

0.10 Neat Epoxy resin

Fillers like aluminium, copper powder and cement by-pass dust 8 wt% of copper filled epoxy composite
are directly mixed with the epoxy resin and the effect of addition of 0.05
these fillers on thermal conductivity is determined. Different sam- 0.00
ples with 4 wt%, 8 wt% of aluminium, and similarly for copper 60 70 80 90 100
powder and CBPD fillers are mixed in the epoxy resin for testing
and results are listed in Table 4 and graphically plotted in Fig. 4a, Temperature (Degree-C)
b and c. Fig. 4b. Thermal conductivity variations with temperatures of copper filled epoxy.
The results show that the thermal conductivity values of vinyl
ester, polyester, and epoxy have nearly the same values. As the
vinyl ester sample has little air voids, it is showing lower thermal 0.3

Thermal conductivity (W/m-K)

conductivity than the other two resins (Table 3). Addition of alu-
minium particulate under three different percentage (0 wt%, 0.25
4 wt% and 8 wt%) in the epoxy matrix composite significantly in-
creases the thermal conductivity as shown in Fig. 4a. For example, 0.2
4 wt% and 8 wt%, respectively result in 64% and 45% increase in
0.15
thermal conductivity over neat epoxy resin for temperatures rang-
ing from 60 to 100 °C. Addition of 4 wt% and 8 wt% of copper pow-
0.1 Neat Epoxy resin

0.05 4wt% CBPD filled epoxy composite

Table 4
8 wt% CBPD filled epoxy composite
Thermal conductivity of epoxy resin with different particulate fillers.
0
Sample Thickness Thermal conductivity (W/ 60 70 80 90 100
(mm) m K)
Temperature (Degree-C)
Test-1 Test-2 Test-3
(60 °C) (80 °C) (100 °C) Fig. 4c. Thermal conductivity variations with temperatures of CBPD filled epoxy.

1. Neat epoxy resin 0.743 0.221 0.237 0.263

2. 4 wt% of aluminium filled 2.114 0.347 0.359 0.351
epoxy composite der in neat epoxy resin resulted in increase of conductivity values
3. 8 wt% of aluminium filled 3.764 0.489 0.534 0.514 by 56% and 59%, respectively (Fig. 4b). However, for CBPD particle
epoxy composite reinforced with epoxy resin composite, slight increase in thermal
4. 4 wt% of copper powder filled 2.734 0.393 0.412 0.422
conductivity is noticed at the temperature segment of 60 °C when
epoxy composite
5. 8 wt% of copper powder filled 3.987 0.372 0.396 0.393 compared to the neat epoxy resin; but the thermal conductivity al-
epoxy composite most remains unchanged at the temperature range of 80–100 °C
6. 4 wt% of CBPD filled epoxy 2.152 0.239 0.197 0.190 This could have happened because addition of CBPD particle makes
composite the sample more brittle and hairline cracks are formed in the sam-
7. 8 wt% of CBPD filled epoxy 3.634 0.227 0.228 0.229
ple under heat and pressure. These cracks create air voids thus
composite
reducing the heat capacity and corresponding thermal conductiv-
ity. These results are tabulated in Table 4 and plotted in Fig. 4c.

0.6 4.2. E-glass fiber reinforced epoxy resin composites

Thermal conductivity (W/m-K)

0.5 E-glass fiber (50 wt%) reinforced with epoxy resin composite is
manufactured in two different methodologies i.e. by hand-lay-up
0.4 technique and compression molding method. The thermal conduc-
tivity values are obtained by testing in through-the-thickness
0.3 direction as shown in Table 5. For glass/epoxy composites, manu-
facturing method has little effect on thermal conductivity. Experi-
0.2 mental results show that compression molding method samples
Neat Epoxy resin
have higher thermal conductivity than the hand-lay-up technique.
0.1 4wt% Aluminium filled composite
The potential cause for this variation in thermal conductivity might
8wt% Aluminium filled composite
0
be due to air voids. Two potential reasons are suggested for forma-
60 70 80 90 100 tion of air voids: (1) stitching closely packs the fabric layers, leav-
ing small air pockets between the layers, and applied resin might
Temperature (Degree-C)
not seep through the air gaps fully, these air pockets hinder full
Fig. 4a. Thermal conductivity variations with temperature of aluminium filled conductivity; and (2) during composite fabrication by hand-lay-
epoxy. up technique the formation of holes can not be neglected and
842 A. Patnaik et al. / Materials and Design 31 (2010) 837–849

Table 5
Thermal conductivity of E-glass fiber/epoxy resin composites.

Sample Manufacturing method Thickness (mm) Thermal conductivity (W/m K)

Test-1 (60 °C) Test-2 (80 °C) Test-3 (100 °C)
E-glass fiber + epoxy resin composite Hand-lay-up technique 5.663 0.304 0.334 0.369
E-glass fiber + epoxy resin composite Compression molding machine 5.225 0.372 0.389 0.412

Table 6 Table 7
Thermal conductivity of E-glass fiber/epoxy resin composites. Thermal conductivity of carbon fiber/epoxy resin composites.

Heat flow Thickness Thermal conductivity (W/m K) Heat flow Thickness Thermal conductivity (W/m K)
direction (mm) direction (mm)
Test-1 Test-2 Test-3 Test-1 Test-2 Test-3
(60 °C) (80 °C) (100 °C) (60 °C) (80 °C) (100 °C)
Transverse 3.228 0.279 0.297 0.334 Carbon fiber reinforced epoxy resin composite for 40% fiber loading
direction Through-the-thickness 3.772 0.245 0.262 0.283
Longitudinal 4.300 0.246 0.271 0.301 Transverse direction 4.325 0.457 0.500 0.541
direction Longitudinal direction 5.176 0.787 0.965 1.231
Through-the- 5.663 0.320 0.340 0.357
Carbon fiber reinforced epoxy resin composite for 60% fiber loading
thickness
Through-the-thickness 4.165 0.312 0.320 0.337
Transverse direction 5.179 0.429 0.447 0.463
Longitudinal direction 5.664 1.367 1.941 2.227

may even formation of a thin interfacial gap between the fiber and
resin due to inadequate interfacial compatibility acting like air void
and in turn reducing the thermal conductivity. These factors poten-
tially change thermal conductivity by around ±0.05 W/m K. For all
samples thermal conductivity increases slightly with increased in
test temperature.
4.3. E-glass reinforced epoxy resin composite in all the three directions
The thermal conductivity values of E-glass/epoxy resin compos-
ite blocks in longitudinal, transverse and through-the-thickness
directions are listed in Table 6. It is observed that the thermal con-
ductivity in through-the-thickness direction is higher than that in
the transverse direction, which in turn is higher than the conduc-
tivity in longitudinal direction. The conductivity value is supposed
to be higher in longitudinal direction. However, in this case, bi-ax-
ial fabric might have led to this anomaly. Further more, the sample
preparation also influences the results. After cutting the sample in
0 and 90 directions the fiber would be in the direction of heat flow
and these fibers form a corrugated shape on the top and bottom
surfaces. According to the sample requirements of the test facility,
these surfaces are supposed to be smooth with a tolerance of
±0.005 inch, which is difficult to achieve with the available re-
sources. However, the thermal conductivity value of the E-glass/
epoxy resin block showed no significant change with respect to
the direction of fiber. The fiber weight fraction for this block is
determined as 49.8%. The low thermal conductivity of balanced
bi-axial E-glass fabric in 0 and 90 directions (i.e. in horizontal
plane) resulted in heat conduction to be nearly uniform in both
longitudinal and transverse directions.
4.4. Carbon fiber reinforced epoxy resin composites
isotropic; (3) thermal conductivity of carbon composites through-
the-thickness direction is close to that of neat resin, indicating that
thermal property of the resin plays a major role in this direction;
(4) carbon composites result in lower thermal conductivity values
in through-the-thickness direction than glass fiber composite. This
observation is in accordance with the results published in the liter-
ature [12] and (5) as shown in Table 7, through-the-thickness ther-
mal conductivity of carbon/epoxy resin composite for 40 wt% is
consistent with that of carbon/epoxy resin composite for 60 wt%.
Carbon fiber is orthotropic in nature and the conductivity of the
carbon fiber along the axis would be higher than in other directions
because of the basal plane formation along the axis. These basal
planes are closely packed in the order of angstrom units, which
helps in transferring heat quickly and by virtue of its property car-
bon fibers contracts upon heat due to negative expansion coeffi-
cient further reducing the distance between the basal planes.
Thus thermal conductivity increased linearly in longitudinal direc-
tion in the measured temperature range of 60–100 °C. In trans-
verse direction, conductivity would be much lower than that of
in longitudinal direction due to lack of basal planes. Moreover, car-
bon sheets would be sized to have good compatibility with the re-
sin, and this sizing causes an insulating layer and thus hinders heat
flow in case of through-the-thickness direction. This glossy sizing
over the carbon tow might be the reason for lower conductivity
of carbon composite in through-the-thickness direction than that
of a glass fiber composite. To test the effect of fiber weight fraction
of a composite on thermal conductivity, another carbon/epoxy re-
sin block was made with higher fiber compactness and tested. The
results are also included in Table 7. The results are compared
graphically in Fig. 5.

The thermal conductivity values of carbon fiber/epoxy resin
composite materials in longitudinal, transverse and through-the-
thickness directions are listed in Table 7. The following observa-
tions are made from the test results of carbon fiber/epoxy resin
composite material: (1) the thermal conductivities of carbon com-
posites are the highest along the fiber direction, medium in the
transverse direction and lowest in through-the- thickness direc-
tion; (2) carbon composites have significantly different thermal
conductivity values in three directions – strongly anisotropic. This
is distinctively different from E-glass composite materials for
which thermal conductivities in three directions are close – almost
4.5. Natural fiber reinforced epoxy resin composites
The thermal conductivity of natural fiber reinforced epoxy com-
posites have an extremely low thermal conductivity as shown in
Table 8 and also shown in Fig. 6, much lower than neat epoxy,
demonstrating excellent thermal insulation characteristics. Thus
after studying all the three combinations i.e. glass fiber/epoxy re-
sin, carbon fiber/epoxy and pine bark/epoxy resin composite, pine
bark reinforced epoxy resin composite shows excellent thermal
insulation characteristics as compared to other two combination
of composites. Therefore, in this study pine bark and epoxy resin
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849 843

2.5

Through thickness for 40wt%

Thermal conductivity (W/m-K)

2
Transverse section for 40wt%

Longitudinal section for 40wt%

1.5
Through thickness for 60wt%

Transverse section for 60wt%

1
Longitudinal section for 60wt%

0.5

0
60 70 80 90 100
Temperature (Degree-C)
Fig. 5. Thermal conductivity variations for carbon fiber/epoxy resin composites.

Table 8 ature predicts through-the-thickness direction the effective ther-

Thermal conductivity of natural fiber/epoxy resin composites. mal conductivity, while thermal conductivity along the fiber
Heat flow Thickness Thermal conductivity (W/m K) direction in a composite can be predicted using rule of mixtures.
direction (mm) Experimental results obtained for the pine bark/epoxy resin sam-
Test-1 Test-2 Test-3
(60 °C) (80 °C) (100 °C)
ples are validated for transverse thermal conductivity and longitu-
dinal thermal conductivity. Thermal conductivities transverse and
Natural fiber reinforced epoxy resin composite for 40% fiber loading
Neat epoxy resin 0.743 0.221 0.237 0.263
longitudinal directions are verified using the theoretical models
Through-the-thickness 3.442 0.123 0.128 0.136 (inverse rule of mixture and rule of mixture)for two phase compos-
Transverse direction 4.128 0.157 0.161 0.155 ites but by incorporation of particulate filler material (cement
Longitudinal direction 4.176 0.175 0.190 0.208 by-pass dust) into two phase composite, the thermal conductivity
Natural fiber reinforced epoxy resin composite for 60% fiber loading can be calculated with the help of Hamilton–Crosser formula [13]
Neat epoxy resin 0.743 0.221 0.237 0.263 are described in Table 9. In this study, thermal conductivity consid-
Through-the-thickness 3.853 0.130 0.142 0.198
ered for validation purpose is the value obtained at a temperature
Transverse direction 4.469 0.177 0.185 0.189
Longitudinal direction 5.102 0.202 0.216 0.228 of 100 °C. The details of thermal conductivity about the samples
are given in Table 10 along with fiber and matrix volume fractions,
respectively. Comparisons between experimental and theoretical
is chosen as fiber and matrix phase and cement by-pass dust is values are shown in Table 11. The obtained experimental results
taken as a filler material for further study. are in good agreement (10%) with the theoretical models.

4.6. Validation of model predictions

4.7. Analysis of experimental results
Analytical or numerical models help to predict the properties of
a material without conducting any experiments. However, these From Table 12, the overall mean for the S/N ratio of the specific
models have to be extensively validated with experimental data wear rate and thermal conductivity are found to be 83.87 dB and
before adopting them in practice on a large scale. Most of the liter- 17.50 dB. Fig. 7a and b shows graphically the effect of the four

0.3
Thermal conductivity (W /m-K )

Neat epoxy resin

0.25
Through thickness for 40%

0.2 Transverse section for 40%

Longitudinal section for 40%

0.15
Through thickness for 60%

0.1 Transverse section for 60%

Longitudinal section for 60%

0.05

0
60 70 80 90 100
Temperature (Degree-C)
Fig. 6. Thermal conductivity variations for natural fiber/epoxy resin composites.
844 A. Patnaik et al. / Materials and Design 31 (2010) 837–849

Table 9
Theoretical models for transverse/longitudinal effective thermal conductivity.

Sl. no Author Model Notes

Transverse effective thermal conductivity
1. Inverse rule of mixture V Derived from rule of mixtures to calculate for transverse thermal conductivity
1
KT ¼ K ff þ VK mm
Does not consider voids, fabric overlap
Longitudinal effective thermal conductivity
2. Rule of mixture K1 ¼ Kf V f þ KmVm For longitudinal effective thermal conductivity

Thermal conductivity of the particulate filled pine bark polyester composites

3. Hamilton–Crosser formula K c ¼ K M K F þðn1ÞK M ðn1Þf ðK M K F Þ
K F þðn1ÞK M þf ðK M K F Þ

Table 10 Before any attempt is made to use this simple model as a predictor
Thermal conductivity of the samples using theoretical model. for the measures of performance, the possible interactions be-
tween the control factors must be considered. Thus factorial design
Composites Kf (along the Kf (along Km Vf (%) Vm (%)
axis of fiber) transverse incorporates a simple means of testing for the presence of the
direction) interaction effects.
E-glass fiber/epoxy resin 0.7 0.6 0.263 50 50 Analysis of the result leads to the conclusion that factor combi-
Carbon fiber/epoxy resin 8 0.54 0.263 50 50 nation of A3, B2, C1 and D3 gives minimum specific wear rate as
Pine bark/epoxy resin 0.3 0.36 0.263 50 50 shown in Fig. 7a and for minimum thermal conductivity the levels
CBPD/epoxy resin 0.74 0.63 0.263 7 93 of factor settings are shown in Fig. 7b such as A3, B3, C3 and D1. The
Pinebark/CBPD/epoxy resin 0.5 0.5 0.263 36.4 63.4
interaction graphs are shown in Fig. 8a and b for specific wear rate
Pinebark/CBPD/epoxy resin 0.5 0.5 0.263 26.1 73.9
and thermal conductivity, respectively.

control factors on specific wear rate and thermal conductivity of 4.8. Surface morphology
the particulate filled pine bark reinforced epoxy resin composites.
The analysis is made using the popular software specifically used The surfaces of the specimens are examined directly by scan-
for design of experiment applications known as MINITAB 15. ning electron microscope JEOL JSM-6480LV. Fig. 9 presents the

Table 11
Comparison between experimental values and theoretical values.

Method/model Manufacturing techniques Pine bark/epoxy resin composite Epoxy resin/CBPD composite Pine bark/CBPD/epoxy resin composite
Experimental Hand lay-up techniques 0.272 0.268 0.196
Inverse rule of mixture Hand lay-up techniques 0.281 0.293 –
Hamilton–Crosser formula Hand lay-up techniques – – 0.206
Error (%) 3.20% 10.23% 4.85%

Table 12
Experimental design using L27 orthogonal array.

Sl No. Sliding velocity, Normal load, Filler content, Sliding distance, Wear rate, S/N ratio (dB) Thermal conductivity, S/N ratio (dB)
A (cm/s) B (N) C (%) D (m) Ws (mm3/N m) K (W/m K)
1 210 10 0 3000 0.0006736 63.432 0.128 17.8558
2 210 10 4 4000 0.0005250 65.597 0.158 16.0269
3 210 10 8 5000 0.0000946 80.482 0.176 15.0897
4 210 20 0 4000 0.0000282 90.995 0.139 17.1397
5 210 20 4 5000 0.0001414 76.991 0.125 18.0618
6 210 20 8 3000 0.0002300 72.765 0.109 19.2515
7 210 30 0 5000 0.0002403 72.385 0.127 17.9239
8 210 30 4 3000 0.0000953 80.418 0.114 18.8619
9 210 30 8 4000 0.0001906 74.398 0.120 18.4164
10 261 10 0 4000 0.0000600 84.437 0.128 17.8558
11 261 10 4 5000 0.0000082 101.72 0.131 17.6546
12 261 10 8 3000 0.0002183 73.219 0.130 17.7211
13 261 20 0 5000 0.0000034 109.37 0.132 17.5885
14 261 20 4 3000 0.0000618 84.180 0.138 17.2024
15 261 20 8 4000 0.0002167 73.283 0.154 16.2496
16 261 30 0 3000 0.0000331 89.603 0.138 17.2024
17 261 30 4 4000 0.0000479 86.393 0.153 16.3062
18 261 30 8 5000 0.0002501 72.038 0.133 17.5230
19 314 10 0 5000 0.0000017 115.39 0.146 16.7129
20 314 10 4 3000 0.0001476 76.618 0.127 17.9239
21 314 10 8 4000 0.0002893 70.773 0.157 16.0820
22 314 20 0 3000 0.0000098 100.17 0.140 17.0774
23 314 20 4 4000 0.0000021 113.55 0.135 17.3933
24 314 20 8 5000 0.0004016 67.924 0.118 18.5624
25 314 30 0 4000 0.0000128 97.856 0.116 18.7108
26 314 30 4 5000 0.0000116 98.711 0.122 18.2728
27 314 30 8 3000 0.0002576 71.781 0.111 17.8558
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849 845

Main Effects Plot for SN ratios

Data Means
A B
90

85

80
Mean of SN ratios

75

210 261 314 10 20 30

C D
90

85

80

75

0 4 8 3000 4000 5000

Fig. 7a. Effect of control factors on specific wear rate.

Main Effects Plot for SN ratios

Data Means

A B
18.00

17.75

17.50

17.25
Mean of SN ratios

17.00
210 261 314 10 20 30
C D
18.00

17.75

17.50

17.25

17.00
0 4 8 3000 4000 5000

Fig. 7b. Effect of control factors on thermal conductivity.

scanning electron microscope of with and without particulate
filled pine bark–epoxy composite worn surfaces under various test
conditions. To correlate the wear data better, SEM graphs pre-
sented in Fig. 9a and b to pertaining to a load of 10 N and at differ-
ent sliding distances for Pine bark reinforced epoxy composites are
considered. Fig. 9a shows the wear surfaces of unfilled composite
at a load of 10 N, 3000 m sliding distance and Fig. 9b shows the
wear surface of unfilled composite of load 20 N, 4000 m sliding dis-
tance, respectively. The large patches formed on the worn surface
of composite are consisted of fine worn particles of polymer matrix
and damage fibers accumulated together by the thermo-mechani-
cal processes at a load of 20 N. Meanwhile, the patches appeared
on the rubbing surface are likely to shield the block from damage
by the hard asperities on the counter face surface and are thus
helpful in reducing wear. At higher sliding distance (3000 m), pho-
tomicrograph shown in Fig. 9c and d depict severe damage to the
matrix, more pine bark breakage and some of the pine bark are
pulled-out from the matrix surface (Table 4, experiment no. 6).
The worn surface is deformed and the pine bark had been torn
from epoxy matrix when the load is 20 N in Fig. 9d.
From Fig. 9e shows the material is removed from the sample
mostly by the adhesion mode. It can be seen that there is a plastic
flow of the matrix material in the sliding direction. It is under-
standable that as the sliding proceeds, the resin soften due to fric-
tional heat generation. As a result, the cement by-pass dust
particles and the pine bark easily tear the matrix and gradually
get aligned along the sliding direction. These particles by virtue
of their size, shape, brittleness and high hardness influence and
modify the wear behavior of the composites. Further, crack forma-
tion and propagation can be seen Fig. 9f at the matrix fiber inter-
846 A. Patnaik et al. / Materials and Design 31 (2010) 837–849

Interaction Plot for SN ratios

Data Means
105
C
0
100 4
8
SN ratios 95

90

85

80

75

70

210 261 314

A
Fig. 8a. Interaction graph between A  C for specific wear rate.

Interaction Plot for SN ratios

Data Means
19.0
B
10
18.5 20
30

18.0
SN ratios

17.5

17.0

16.5

16.0
210 261 314
A
Fig. 8b. Interaction graph between A  B for thermal conductivity.

face, which also contribute to material loss over the test duration.
Extensive pine bark fragmentation indicated that the adhesive
strength between fiber, matrix and filler material became weak
while the high frictional heat accumulated in the contacting
surface.
4.9. ANOVA and the effects of factors
In order to understand a concrete visualization of impact of var-
ious factors and their interactions, it is desirable to develop analy-
sis of variance (ANOVA) table to find out the order of significant
factors as well as interactions. Tables 13a and 13b show the results
of the ANOVA with the specific wear rate. This analysis was under-
taken for a level of confidence of significance of 5%. The last column
of the table indicates that the main effects are highly significant
(all have very small p-values) [14].
From Table 13a, one can observe that the filler content
(p = 0.049), sliding velocity (p = 0.107) and sliding distance
(p = 0.370) have great influence on specific wear rate. The interac-
tion between sliding velocity  filler content (p = 0.288) shows sig-
nificance of contribution on the wear rate and the factor normal
load (p = 0.564) and sliding velocity  normal load (p = 0.969)
and normal load  filler content (p = 0.743) present less signifi-
cance of contribution on specific wear rate.
Similarly, From Table 13b, one can observe that the filler con-
tent (p = 0.082), sliding distance (p = 0.135) and sliding distance
(p = 0.433) have great influence on thermal conductivity. The inter-
action between sliding velocity  filler content (p = 0.288) shows
significance of contribution on the thermal conductivity and the
factor normal load (p = 0.564), sliding velocity  normal load
(p = 0.969) and normal load  filler content (p = 0.743) present less
significance of contribution on thermal conductivity.
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849 847

Fig. 9. SEM micrographs of the worm pine bark-epoxy composites filled with cement by-pass dust.

Table 13a 5. Confirmation experiment

ANOVA table for specific wear rate.

Source DF Seq SS Adj SS Adj MS F P The confirmation experiment is the final test in the design of
experiment process. The purpose of the confirmation experiment
A 2 1084.3 1084.3 542.2 3.33 0.107
B 2 205.8 205.8 102.9 0.63 0.564 is to validate the conclusions drawn during the analysis phase.
C 2 1698.1 1698.1 849.1 5.21 0.049 The confirmation experiment is performed by conducting a new
D 2 384.7 384.7 192.4 1.18 0.370 set of factor settings A2C1D3 to predict the specific wear rate. The
AB 4 80.5 80.5 20.1 0.12 0.969 estimated S/N ratio for specific wear rate can be calculated with
AC 4 1044.6 1044.6 261.2 1.60 0.288
the help of following prediction equation:
BC 4 320.8 320.8 80.2 0.49 0.743
Error 6 977.9 977.9 163.0    
Total 26 5796.8 g 1 ¼ T þ A2  T þ C 1  T
h     i  
þ A2 C 1  T  A2  T  C 1  T þ D3  T ð6Þ

where g  1 is the predicted average, T is overall experimental average

Table 13b and A2 ; C 1 and D3 is the mean response for factors and interactions
ANOVA table for thermal conductivity.
at designated levels.
Source DF Seq SS Adj SS Adj MS F P By combining like terms, the equation reduces to
A 2 1.2213 1.2213 0.6106 0.96 0.433
B 2 4.9578 4.9578 2.4789 3.91 0.994 g 1 ¼ A2 C 1 þ D3  T ð7Þ
C 2 0.0081 0.0081 0.0041 0.01 0.082
D 2 3.6108 3.6108 1.8054 2.85 0.135 A new combination of factor levels A2, C1 and D3 are used to pre-
AB 4 8.6402 8.6402 2.1600 3.41 0.088 dict specific wear rate through prediction equation and it is found
AC 4 3.5943 3.5943 0.8986 1.42 0.334
to be g
 1 ¼ 74:97 dB. For each performance measure, an experiment
BC 4 0.7075 0.7075 0.1769 0.28 0.881
Error 6 3.7993 3.7993 0.6332 is conducted for a different factors combination and compared
with the result obtained from the predictive equation as shown
Total 26 26.5393
in Tables 14a and 14b.
848 A. Patnaik et al. / Materials and Design 31 (2010) 837–849

Table 14a Y Ws ¼ K 0 þ K 1  A þ K 2  C þ K 3  D þ K 4  A  C ð10Þ

Results of the confirmation experiments for wear rate.
YK ¼ K0 þ K1  A þ K2  C þ K3  D þ K4  A  C ð11Þ
Optimal control parameters
Prediction Experimental Here, Yws and Yk is the performance output terms and Ki (i = 0,
1, . . . , 4) are the model constants. The constant are calculated by
Level A2C1D3 A2C1D3
S/N ratio for specific wear rate (dB) 98.9352 95.8286 using non-linear regression analysis method with the help of MINI-
TAB 14 software and the following relations were obtained. The
calculated coefficients from MINITAB software are substituted in
Eqs. (10) and (11) and following relations were obtained.

Table 14b W s ¼ 1:618  1:541  A  1:497  C  0:237  D

Results of the confirmation experiments for thermal conductivity.
þ 2:012  A  C r 2 ¼ 0:69 ð12Þ
Optimal control parameters
K ¼ 0:884  0:152  A  0:271  C þ 0:118  D
Prediction Experimental
þ 0:158  A  C r 2 ¼ 0:99 ð13Þ
Level A1C3D3 A1C3D3
S/N ratio for thermal conductivity (dB) 18.3418 17.6594
The higher correlation coefficients (r2) confirm the suitability of
the used model and the correctness of the calculated constants. In
this study, a weighting method is used for the optimization of the
Similarly, a prediction equation is developed for estimating S/N process with multi-machining performance outputs. Since the spe-
ratio of thermal conductivity as given in Eq. (8). cific wear rate and thermal conductivity are the two different ob-
    jects, in order to overcome the large differences in numerical
g 2 ¼ T þ A1  T þ C 3  T values between the objects, the function corresponds to every
h     i   machining performance output is normalized first. A weighting
þ A1 C 3  T  A1  T  C 3  T þ D3  T ð8Þ method is adopted to normalize the performance output, such as
specific wear rate and thermal conductivity to a single object. Here,
where g  2 is the predicted average, T is overall experimental average the resultant weighted objective function to be maximized is:
and A1 ; C 3 and D3 is the mean response for factors and interactions
at designated levels. Maximize Z ¼ ðw1  1=f1 þ w2  1=f2 Þð1 þ K  CÞ ð14Þ

g 2 ¼ A1 C 3 þ D3  T
A new experimental set-up with factor levels at A1C3 and D3 is
considered to predict the S/N ratio for thermal conductivity and
is found to be g  2 ¼ 18:3418 dB. For each performance measure,
an experiment was conducted for a different factor combination
and compared with the result obtained from the predictive equa-
tion as shown in Tables 14a and 14b. The resulting model seems
to be capable of predicting specific wear rate and thermal conduc-
tivity to a reasonable accuracy. An error of 3.14% for the S/N ratio of
specific wear rate and 3.72% for the S/N ratio thermal conductivity
is observed. However, the errors can be further reduced if the num-
ber of measurements is increased. This validates the development
of the mathematical model for predicting the measures of perfor-
mance based on knowledge of the input parameters.
6. Multi-objective optimization of control parameters
Machining settings that satisfy multiple objectives of minimiza-
tion of wear rate and minimization of thermal conductivity need to
be determined. The mathematical model suggested here is in the
following form.
ð9Þ where f1 is the normalized function for specific wear rate, f2 is nor-
malized function for thermal conductivity, C is violation coefficient,
K is a penalty parameter, usually the value is 10 and Z is the perfor-
mance output.
Subjected to constraints:
Amin 6 A 6 Amax ð15Þ
C min 6 C 6 C max ð16Þ
Dmin 6 D 6 Dmax ð17Þ
where w1 and w2 are the weighting factors for normalized specific
wear rate and thermal conductivity functions used in the objective
function of optimization process. The weighting factors are selected
in such a manner that their sum is equal to one. A higher weighting
factor for an objective indicates more emphasis on it. The min and
max in Eqs. (15)–(17) represent lowest and highest control factor
settings, respectively (Table 2).
Genetic algorithm (GA) was used to obtain the optimum
machining parameters for multi-objective outputs by using the
several combinations of the weight. The larger the weighting fac-
tor, the greater the improvement in the machining performance
outputs. The computational algorithm was implemented in C++

Table 15
Optimum operating conditions for multi-performance with different weighting factors.

Control factors and Optimum operating conditions

performance measures
Case-1: Case-2: Case-3: Case-4: Case-5:
(w1 = 0.50 and w2 = 0.50) (w1 = 0.40 and w2 = 0.60) (w1 = 0.60 and w2 = 0.40) (w1 = 0.20 and w2 = 0.80) (w1 = 0.80 and w2 = 0.20)
A: Sliding velocity (cm/s) 0.715 0.692 0.696 0.993 0.871
C: Filler content (%) 0.579 0.629 0.807 0.957 0.752
D: Sliding distance (m) 0.137 0.151 0.073 0.061 0.076
Ws: Specific wear rate 0.44989 2.35884 0.45017 0.55270 0.44988
(mm3/N m)
K: Thermal conductivity 0.69998 0.69494 0.65687 0.63106 0.66027
(W/m K)
Total out put performance 2.06393 1.03295 1.94178 1.62955 1.86867
 A. Patnaik et al. / Materials and Design 31 (2010) 837–849
code. Genetic algorithms (GAs) are mathematical optimization
techniques that simulate a natural evolution process. They are
based on the Darwinian theory, in which the fittest species sur-
vives and propagate while the less successful tend to disappear.
The concept of genetic algorithm is based on the evolution process
and was introduced by Holland [15]. Genetic algorithm mainly de-
pends on the following types of operators: reproduction, crossover,
and mutation. Reproduction is accomplished by copying the best
individuals from one generation to the next, in what is often called
an earliest strategy. The best solution is monotonically improving
from one generation to the next. The selected parents are submit-
ted to the crossover operator to produce one or two children. The
crossover is carried out with an assigned probability, which is gen-
erally rather high. If a number randomly sampled is inferior to the
probability, the crossover is performed. The genetic mutation
introduces diversity in the population by an occasional random
replacement of the individuals. The mutation is performed based
on an assigned probability. A random number is used to determine
if a new individual will be produced to substitute the one gener-
ated by crossover. The mutation procedure consists of replacing
one of the decision variable values of an individual, while keeping
the remaining variables unchanged. The replaced variable is ran-
domly chosen, and its new value is calculated by randomly sam-
pling within its specific range. A weighting factor is used to
provide importance to the performance measures as per require-
ment of the decision maker. In genetic optimization, population
size, probability of crossover and mutation are set at 50%, 75%
and 5%, respectively. Number of generation is varied till the output
is converged. Table 15 shows the optimum conditions of the con-
trol factors with optimum performance output that suggest best
combination of set of input control factors for multi-performance
outputs with different combinations of the weighting factors. In
this study case-1 gives optimal performance with maximization
of specific wear rate and thermal conductivity under equal impor-
tance of the weighting factors (w1 = 0.55 and w2 = 0.55). Case-1 is
recommended because it gives minimum specific wear rate and
minimum thermal conductivity.
7. Conclusions
Based on the experimental and analytical work presented
above, it can be concluded as follows:
1. With the addition of 4 wt% and 8 wt% of aluminium powder, the
thermal conductivity neat epoxy resin improves by 64% and
45%, respectively. Similarly, addition of 4 wt% and 8 wt% of cop-
per powder is found to have resulted in an increase of conduc-
tivity by 56% and 59%, respectively. It is also observed that by
the addition of industrial waste like cement by-pass dust with
neat epoxy resin, the thermal conductivity value improves by
849
about 92% (for 4 wt% addition) and by about 97% (for 8 wt%
addition).
2. The thermal conductivity along longitudinal direction in case of
carbon/epoxy resin composite is almost twice the conductivity
in transverse and three times greater than through-the-thick-
ness direction. But in case of natural fiber reinforced epoxy
composite there is no significant change in thermal conductiv-
ity values as compared to those in case of glass fiber/carbon
fiber reinforced epoxy composites.
3. Experimental results obtained from pine bark/Epoxy resin sam-
ple in longitudinal and through-the-thickness direction are in
good agreement with the values obtained from theoretical
models. In longitudinal direction, rules of mixture showed an
upper bound value and in through-the-thickness direction
inverse rule of mixture showed a lower bound experimental
data.
4. SEM micrographs for pine bark reinforced epoxy composite
show fairly good bonding with the matrix. The addition of
cement by-pass dust in the pine bark-epoxy composite further
enhances the capability of material and improves the structural
integrity of material under sliding wear condition.
References
[1] Kumbhat N, Raj PM, Pucha RV, Atmur S. Proc Electron Compon Technol Conf
2005;2:1364.
[2] Siu B. In: IEEE reg. 10 annu. int. conf. proc. TENCON 234; 1995.
[3] Progelhof RC, Throne JL, Ruetsch RR. Methods for predicting the thermal
conductivity of composite systems: a review. Polym Eng Sci 1976;16:
615–25.
[4] Procter P, Solc J. Improved thermal conductivity in microelectronic
encapsulants. IEEE Trans Compon Hybrids Manuf Technol 1991;14(4):708–13.
[5] Nagai Y, Lai GC. Thermal conductivity of epoxy resin filled with particulate
aluminum nitride powder. J Ceram Soc Jpn 1997;105(3):197–200.
[6] Bollampally R, Wong CP. Comparative study of thermally conductive fillers for
use in liquid encapsulants for electronic packaging. IEEE Trans Adv Pack
1999;22(1):54–9.
[7] Wong CP, Bollampally RS. Thermally conductivity, elastic modulus, and
coefficient of thermal expansion of polymer composites filled with ceramic
particles for electronic packaging. J Appl Polym Sci 1999;74(14):3396–403.
[8] Lu X, Xu G. Thermally conductive polymer composites for electronic
packaging. J Appl Polym Sci 1997;65(13):2733–8.
[9] Joshi VS, Drzal TL, Mohanty KA, Arora S. Are natural fiber composites
environmentally superior to glass fiber reinforced composites. Composites
Part A 2004;35:371–6.
[10] Glen SP. Taguchi methods: a hands on approach. New York: Addison-Wesley;
1993.
[11] Callister Jr DW. Material science and engineering an introduction. John Wiley
and Sons, Inc.; 1990.
[12] Gowayed Y, Jhy-Cherng H. Thermal conductivity of composite materials made
from plain weaves and 3-D weaves. Compos Eng 1995;5(9):1177–86.
[13] Jackel M. Thermal properties of polymer/particle composites at low
temperatures. Cryogenics 1995;35:713–6.
[14] Montgomery DC. Design and analysis of experiments. 5th ed. John wiley &
Sons, Inc.; 2005. p. 363–82.
[15] Holland JH. Adaptation in natural and artificial systems. Ann Arbor (MI): The
University of Michigan Press; 1975.