Bull. Mater. Sci., Vol. 36, No. 4, August 2013, pp. 575–583.


c Indian Academy of Sciences.

Optimization of mechanical properties of non-woven short sisal

fibre-reinforced vinyl ester composite using factorial design
and GA method

S VELUMANI∗ , P NAVANEETHAKRISHNAN† , S JAYABAL†† and D S ROBINSON SMART#

Department of Mechanical Engineering, Velalar College of Engineering and Technology, Erode 638 012, India
† Department of Mechanical Engineering, Kongu Engineering College, Perundurai 638 052, India
†† Department of Mechanical Engineering, A.C. College of Engineering and Technology, Karaikudi 630 004, India
# Department of Mechanical Engineering, Karunya University, Coimbatore 641 114, India

MS received 11 May 2011; revised 24 May 2012

Abstract. This work presents a systematic approach to evaluate and study the effect of process parameters on
tensile, flexural and impact strength of untreated short sisal fibre-reinforced vinyl ester polymer-based composites
and predicts the optimum properties of random natural fibre-reinforced composites. The natural fibre of sisal at
lengths of 10, 30 and 50 mm and vinyl ester resin at loadings of 15, 30 and 45 (wt%) were prepared. The composite
panel was then fabricated using hand lay method in cold process of size 180 × 160 mm2 . Samples were then cut from
the panel and subjected to mechanical properties testing such as tensile, flexural and impact strengths. The average
tensile strength ranges between 27·1 and 43·9 MPa. The flexural strength ranged between 26·9 and 49·5 MPa and
the impact strength ranged between 16 and 93 J/m. The strength values were optimized using factorial design and
genetic algorithm (GA) method. The predicted optimum process parameter values are in good agreement with the
experimental results.

Keywords. Sisal fibre; composites; polymer matrix; mechanical properties.

1. Introduction articles such as purses, wall hangings, table mats, etc. The
use of sisal fibre as a reinforcement in composites has raised
In the field of composites, the fibre-reinforced matrices were great interest and expectations among the material scientists
initially developed using man-made fibres such as carbon, and engineers (Zhang and Horrocks 2003). Among the vari-
aramid, glass, etc, due to the advantage of their high tensile ous natural fibres, sisal fibres possess moderately high spe-
modulus (Joly et al 1996). But in the recent past, a good deal cific strength and stiffness and hence, can be used as a rein-
of work has been dedicated to natural fibres with the aim forcing material in polymeric resin matrices to make useful
of replacing man-made fibres (Woodhams et al 1984). There structural composite materials (Pavithran et al 1987). Sisal
are many reasons to select these natural fibres as reinforce- fibre is extracted from the plant Agave sisalana and is avai-
ment. They have easy renewability, availability throughout lable in abundant quantity in southern parts of India, north
the world and good biodegradability. Relating to the basic and south America, Africa and West Indies. The incorpora-
properties, these fibres have a specific weight far less than tion of sisal fibres into plastics and elastomers to obtain cost
glass and also their tensile modulus is higher than that of the reduction and reinforcements have been reported by various
aramid fibres (Sotton and Ferrari 1989; Joseph et al 1999). researchers.
In recent years, polymer composites reinforced with short Sisal fibre-reinforced urea formaldehyde matrix compo-
natural fibres have gained wide importance because of the site was investigated for impact strength, flexural strength
advantages they impart in processing and their low cost and wear resistance (Zhong et al 2007). There has been
coupled with high strength (Cruz-Ramos and Carlos 1986). enhancement in the tensile and flexural properties with
Although sisal fibre is one of the most extensively used increase in the carbon fibre content in the unsaturated vinyl
natural fibres, a large quantity of this economic and rene- ester-based sisal/carbon fibre reinforced hybrid composites
wable resource is still not fully utilized. Now, sisal fibre is (Noorunnisa Khanam et al 2010). With minimum number of
mainly used as ropes in the marine and agricultural indus- experiments, process variables can be effectively optimized
tries. Other applications of sisal fibres include twines, cords, using RSM method (Yadav et al 2007). The higher value of
upholstery, padding, mat making, fishing nets and fancy mechanical properties were obtained by developing a model
using regression analysis and the higher value of mechani-
cal properties were found, when coir fibre length was at its
∗ Author for correspondence ([email protected]) maximum length (Jayabal et al 2011).

575
576 S Velumani et al

In the present work, untreated sisal composite has been fibre content of produced composite laminates. The pure
prepared using chopped fibre lengths of 10, 30 and 50 mm resin sample without fibre reinforcement acted as reference
and fibre contents 15, 30 and 45. Unsaturated vinyl ester resin for the experiments. Before going for testing, the samples
was used as matrix for preparing the composites with di- were conditioned for at least 24 h at 23 ◦ C and 50% relative
fferent combinations of fibre length and fibre content (wt%). humidity.
The composite sheets were produced as single layer com-
posite. Mathematical modeling was developed. Mechanical
properties such as tensile, flexural and impact strength were 3.1a Tensile strength test: The tensile strength properties
studied for these sisal composites. The response variables were determined according to ASTM 638-08. The length,
were optimized by factorial design and genetic algorithm width and thickness of each sample were ∼165, 25 and
technique. 3 mm. For this testing, the composite sheets were cut into
number of specimens. The tensile specimens were tested in a
Tenso meter tensile testing machine (Kudale Instruments (P)
2. Experimental Ltd., India) with a preload of 50 N and having a displacement
speed of 2 mm/min. The gauge length was 135 mm. Five
2.1 Material specimens were tested for each combination of fibre para-
meters. The calculated results were the average values
In the composite production, thermosetting resin mixture and of sample responses from the five specimens. The tested
a fibre were used. It was a cold-setting laminating resin sys- mechanical property values for the nine combinations of fibre
tem consisting of vinyl ester resin, MEKP catalyst, cobalt parameters are shown in table 1. The tensile strength of all
naphthenate promoter (6%) and dimethyl aniline accelera- the nine combinations are shown in figure 1.
tor. All these items were supplied by M/s Covai Seenu and
company, Coimbatore. Sisal fibres used in this study were
received from APK Industries, Erode. The sisal fibres were Table 1. Mechanical properties of composite sheets.
of good quality and their diameters varied substantially in the
range 0·080–0·250 mm. Identification f1 f2 ts fs is

A1 10 15 27·1 26·9 16
2.2 Composite fabrication A2 30 15 37·7 41·1 64
A3 50 15 42·8 47·5 89
To prepare the composite, vinyl ester resin was used as the A4 10 30 31·0 33·1 34·7
matrix material. The chopped sisal fibre composite sheets A5 30 30 41·0 44·5 75
were manufactured by simple hand lay-up process in a A6 50 30 45·1 49·5 91·7
A7 10 45 34·5 38·7 47·5
mould at laboratory temperature (cold process). For easier
A8 30 45 42·0 46·3 85
removal of the composite panel from the mould, the frame A9 50 45 43·9 47·8 93
and the board were treated with a parting compound, i.e.
poly vinyl acetate. Releasing agent was applied to the sur-
faces before moulding. Chopped non-woven sisal fibres were
pre-impregnated with the matrix material consisting of vinyl
ester resin, accelerator, catalyst and promoter in the ratio
of 1:0·020:0·020:0·016. The impregnated fibres and resin
were poured in the mould frame of size 180 × 160 mm2 .
Nine composite laminates with different combinations of
fibre length (10, 30 and 50 mm) and fibre content (15, 30
and 45 wt%) were designed and produced. The pouring was
continued till the thickness reached 3 mm and then pressed
heavily for 3 h before removal. Then, the composite sheet
was removed from the mould and cured at room tempera-
ture for 24 h. After the curing process, test samples were cut
according to the sizes of ASTM standards.

3. Mechanical tests

3.1 Characteristics of composites

The composites were investigated for their tensile, flexu-

ral and impact properties for the pre-fixed fibre length and Figure 1. Effect of fibre parameters on tensile strength.
 Optimization of mechanical properties of sisal–vinyl ester composite 577

3.1b Flexural strength test: The rectangular test pieces of on Geeth Impact testing machine, India and the machine
127 × 12·7 × 3 mm3 dimensions for flexural test were cut capacity is 20 J. For each testing procedure, five specimens
from the prepared randomly oriented fibre composites. Flexu- were tested. All the results shown in table 1 under the column
ral test was conducted as per ASTM D790. The test was of impact strength are taken as the average reading of the five
conducted on the spring testing machine. The specimen was measured samples from each run. The impact strength values
freely supported by a beam and the load was applied in the attained are displayed in figure 3.
middle of the specimen. The tests were carried out at a tem-
perature of 27 ◦ C and at a relative humidity of 50%. For sta-
tistical purposes, a total of five samples were tested for each 4. Factorial design method
run and quoted in table 1. Comparisons of flexural strength
properties of different combinations are shown in figure 2. Factorial designs were used first by John Bennet Lawes and
Joseph Henry Gilbert of Rothamsted Experimental Station
in 19th century. Factorial design is a systematic method that
3.1c Impact strength test: For the Izod impact strength investigates the effects of two or more factors on the out-
test, rectangular unnotched specimens (80 × 10 × 3 mm3 ) put response process. Each complete trial or replication of
were cut from the produced laminate sheet and tested accord- the experiment takes into account all the possible combina-
ing to ASTM D 256 standards. Impact strength was measured tions of the varying levels of these factors. Effective factorial
design ensures that the least number of experiment runs are
conducted to generate the maximum amount of information
about how input variables affect the output of a process.
In this case, two factors (fibre length and fibre content)
are the parameters with three levels (−1, 0, 1) in each factor
(fibre length of 10, 30 and 50 mm and fibre content of 15, 30
and 45%), so totally 3 × 3 = 9 runs (experiments) were con-
ducted for studying the influence on the response of tensile,
flexural and impact strengths.
The main effect of a factor is essentially the ‘average’
change in the output response as that factor goes from ‘−’
to ‘+’. Mathematically, this is the average of two numbers:
(i) the change in output when the factor goes from low to
high level as the other factor stays low and (ii) the change
in output when the factor goes from low to high level as the
other factor stays high.
It must be illustrated that apart from ‘main effects’, factors
can like-wise result in interaction effects. Interaction effects

Figure 2. Effect of fibre parameters on flexural strength.

Figure 4. Optimum fibre parameter values obtained from facto-

Figure 3. Effect of fibre parameters on impact strength. rial design method for ts.
578 S Velumani et al

are changes in the process output caused by two or more process most, so that improvements (or corrective actions)
factors that are interacting with each other. Large interactive may be attained (figures 4–6).
effects can make the main effects insignificant, such that it
becomes more important for the engineer to pay attention to
the interaction of the involved factors than to investigate them 5. Non-linear regression analysis
individually.
The running of factorial combinations and the mathema- The mechanical properties such as tensile, flexural and
tical interpretation of the output responses of the process to impact strength were modeled using second-order polyno-
such combinations is the essence of factorial experiments. mial curve fitting equations using Design Expert statisti-
It allows an engineer to understand which factors affect the cal software (Jayabal and Natarajan 2011). Equations (1)–
(3) are the developed nonlinear regression models of tensile
strength (ts), flexural strength (fs) and impact strength (is),
respectively.

ts = 11·622 + 0·909 f 1 + 0·575 f 2 − 0·005 f 1

× f 2 − 0·007 f 12 − 0·004 f 22 , (1)

f s = 7·16 + 1·179 f 1 + 0·74 f 2 − 0·009 f 1

× f 2 − 0·008 f 12 − 0·004 f 22 , (2)

is = − 41·82 + 4·05 f 1 + 1·68 f 2 − 0·023 f 1

× f 2 − 0·031 f 12 − 0·006 f 22 . (3)

6. Analysis of variance

ANOVA technique can be used to investigate any number

of factors which are hypothesized or said to influence the
dependent variable. This design consisted of two factors,
Figure 5. Optimum fibre parameter values obtained from facto- each at three levels.
rial design method for fs.

6.1 ANOVA for tensile strength test

The list of degrees of freedom, sum of squares and mean

square for tensile strength model are given in table 2. The
fit was exact and R 2 value is 0·9974. The model F-value of
229·266 implies that the model is significant. There is only
0·05% chance that a “model F-value” of this large value
could occur due to noise. Values of “probability > F” less
than 0·0500 indicate that the model terms are significant. In
this case, f 1 , f 2 , f 1 × f 2 , f 12 are significant model terms. Valu-
es greater than 0·1000 indicate that the model terms are not
significant. So, f 22 is not significant.

6.2 ANOVA for flexural strength test

The list of degrees of freedom, sum of squares and mean

square for flexural strength model are given in table 3. The
fit was exact and R 2 value is 0·998. The model F-value of
257·779 implies that the model is significant. There is only
0·04% chance that a “model F-value” of this largeness could
occur due to noise. Values of “probability > F” less than
Figure 6. Optimum fibre parameter values obtained from facto- 0·0500 indicate that the model terms are significant. In this
rial design method for is. case, f 1 , f 2 , f 1 × f 2 , f 12 are significant model terms. Values
 Optimization of mechanical properties of sisal–vinyl ester composite 579

Table 2. ANOVA for tensile strength model.

Response 1 Tensile strength

ANOVA for response surface quadratic model
Analysis of variance table (partial sum of squares − type III)
p-value
Sum of Mean
Source squares df square F value Prob > F Remarks

Model 311·527 5 62·3054 229·266 0·0005 Significant

f1 256·107 1 256·1067 942·402 <0·0001
f2 27·307 1 27·3067 100·481 0·0021
f1 × f2 9·923 1 9·9225 36·5121 0·0091
f 12 16·056 1 16·0556 59·080 0·0046
f 22 2·136 1 2·1356 7·858 0·0677
Residual 0·815 3 0·2718
Corr. total 312·344 8

Table 3. ANOVA for flexural strength model.

Response 2 Flexural strength

ANOVA for response surface quadratic model
Analysis of variance table (partial sum of squares − type III)
p-value
Sum of Mean
Source squares df square F value Prob > F Remarks

Model 461·974 5 92·395 257·779 0·0004 Significant

f1 354·202 1 354·202 988·214 <0·0001
f2 49·882 1 49·882 139·168 0·0013
f1 × f2 33·063 1 33·063 92·244 0·0024
f 12 22·894 1 22·893 63·873 0·0041
f 22 1·934 1 1·934 5·395 0·1028
Residual 1·075 3 0·358
Corr. total 463·05 8

Table 4. ANOVA for impact strength model.

Response 3 Impact strength

ANOVA for response surface quadratic model
Analysis of variance table (partial sum of squares − type III)
p-value
Sum of Mean
Source squares df square F value Prob > F Remarks

Model 6180·040 5 1236·008 619·179 0·0001 Significant

f1 5133·375 1 5133·375 2571·576 <0·0001
f2 532·042 1 532·042 266·527 0·0005
f1 × f2 189·063 1 189·063 94·711 0·0023
f 12 321·734 1 321·734 161·172 0·0011
f 22 3·827 1 3·827 1·9173 0·2602
Residual 5·989 3 1·996
Corr. total 6186·03 8
580 S Velumani et al

>0·1000 indicate that the model terms are not significant. powerful tool for optimizing functions with numerous appli-
Hence, f 22 is not significant. cations. It requires only an objective function to guide their
search.
Genetic algorithms require design space to be converted
6.3 ANOVA for impact strength test
into genetic space. So, genetic algorithms work with a coding
The list of degrees of freedom, sum of squares and mean of variables. The data processed by GA consisted of a set of
square for impact strength model are given in table 4. The strings (or chromosomes) with a finite length in which each
fit was exact and R 2 value is 0·998. The model F-value of bit is called an allele (or a gene).
619·179 implies that the model is significant. There is only A selected number of strings are called as population and
0·01% chance that “model F-value” of this largeness could the population at a given time is a generation. Generation
occur due to noise. Values of “probability > F” less than of the initial population of strings is done randomly. New
0·0500 indicate that the model terms are significant. In this chromosomes can be generated using genetic operators. The
case, f 1 , f 2 , f 1 × f 2 , f 12 are significant model terms. Values genetic operators operate on the genes to replace their place
>0·1000 indicate that the model terms are not significant. So, within the chromosome. Since, the binary alphabet proposes
f 22 is not significant. the maximum number of schemata per bit of information of
any coding (Goldberg 1989), a binary encoding scheme is
traditionally used to represent the chromosomes using either
7. Genetic algorithm technique zeros or ones. Other encodings such as octal encoding, hexa-
decimal encoding, permutation encoding, value encoding
GAs were introduced by Prof John Holland (1975) of the Uni- and tree encoding are also possible.
versity of Michigan. Genetic algorithms are computerized Reproduction is the first operator applied on population.
search and optimization algorithms based on the mecha- Chromosomes are selected from the population to be par-
nics of natural selection (Rajasekaran and Vijayalakshmi Pai ents to crossover and produce offspring. The best one should
2005) and it can search the entire search space to find the survive and create new offspring according to Darwin’s
global optimum. Therefore, this technique has been used as a evolution theory of survival of the fittest.

Figure 7. Optimum fibre parameter values obtained from GA method for tensile strength.
 Optimization of mechanical properties of sisal–vinyl ester composite 581

Figure 8. Optimum fibre parameter values obtained from GA method for flexural strength.

In Roulette wheel selection, each member of the pool is Mutation of a bit involves flipping it, i.e. changing 0–1 and
assigned space on a Roulette wheel proportional to its fitness. vice versa with mutation probability. The mutation operator
The members with the greatest fitness have the highest pro- searches in the entire space in the given time. Restores lost
bability of selection. This selection technique works only for information or adds information to the population. Mutation
GA which maximizes its objective function. operation was performed on a child after crossover function.
Crossover, inversion and mutation are the three main Performed infrequently (for example, 0·005 probability of
genetic operators used for global searches, although there are altering a gene in a chromosome).
other types of genetic operators that can yield good results. In
the crossover operator, new strings are created by exchanging Child-1 1 1 0 1 0 0 0 0 1 0 0 1 1
information among strings of the mating pool. After mutation 1 1 0 1 1 0 0 0 1 0 0 1 1
Simple crossover involves two parents and crossover In the Matlab software, the following options have been
points are selected randomly. If two parents to be used for selected: population: double vector, population size: 20, fit-
generating new chromosomes are {Parent A: 1 0 0 1 0} and ness scaling: rank, selection: stochastic uniform, reproduc-
{Parent B: 1 0 1 1 0} and a crossover point was chosen ran- tion: elite count-2, mutation function: constraint dependent,
domly as 2; this produces the following children: {Child A: cross over: scattered and migration: forward direction.
1 0 | 1 1 0} and {Child B: 1 0 | 0 1 0}. Stopping criteria: Generations: 100, time limit: infinitive,
Parent-A 1 0 0 1 0 1 0 1 1 0 Child A fitness limit: infinitive, stall generation: 50, stall time limit:
Parent-B 1 0 1 1 0
» 1 0 0 1 0 Child B infinitive, function tolerance: 1e-6 and nonlinear constraint
tolerance: 1e-6.
Inversion: A string from the population is selected and the Results are plotted in figures 7–9 for tensile, flexural and
bits between two random sites are inverted as shown below: impact strengths, respectively.

1 0 1 1 0 0 1
8. Optimization of parameters
1 0 0 0 1 1 1
Bits between Optimization of responses includes the work of finding
sites inverted
the optimum set of experimental factors that will produce
582 S Velumani et al

Figure 9. Optimum fibre parameter values obtained from GA method for impact strength.

Table 5. Optimal fibre parameters obtained from factorial design and GA.

Fibre length (mm) Fibre content (wt%) Tensile strength (MPa) Flexural strength (MPa) Impact strength (J/m)
Factorial Factorial Factorial Factorial Factorial
design GA design GA design GA design GA design GA

50 50 34·05 34·048 44·697 44·698 – – – –

50 50 33·78 30·07 – – 48·86 48·922 – –
50 47·641 35·33 45 – – – – 92·88 93·491
50 50 33·94 34 44·697 44·701 48·856 48·87 – –
50 50 34·05 34·071 – – 48·852 48·853 92·74 92·741
50 50 34·17 34·172 44·697 44·699 – – 92·74 92·75
50 50 34·05 34·051 44·6976 44·697 48·852 48·852 92·7406 93·491

maximum value of responses. In this study, optimal para- 9. Results and discussion
meters were found using factorial design technique and GA
method through the Design Expert and Matlab software. The Sisal–vinyl ester composite shows the average values for
best possible combination of input parameters for maximum tensile, flexural and impact strengths of 44·698 MPa,
value of tensile, flexural and impact strengths are shown in 48·852 MPa and 92·74 J/m, respectively. In this work,
table 5. The optimization module in the software package mechanical properties of two main factors f 1 and f 2 and
searches for various combinations of input parameters and interaction effects of f 1 × f 2 and f 12 × f 22 were studied. All
satisfies the objective function as well as maintains the limit the terms were included in the regression model of second-
of factor values within the limit. For in between values of order response for getting minimum standard error. From
factors, the responses can be predicted using point prediction ANOVA result, the following suggestions were made: (i) in
mode of the software. the tensile strength analysis, the most significant variable
 Optimization of mechanical properties of sisal–vinyl ester composite 583

Table 6. Validation of mathematical model.

Experimental values Predicted values % of error

f 1 (mm) f 2 (wt% ) ts (Mpa) fs (Mpa) is (J/m) ts (Mpa) f s (Mpa) is (J/m) ts (Mpa) fs (Mpa) is (J/m)

25 40 38 44 72 40·3373 44·4543 74·2925 6·151 1·033 3·184

30 35 40·5 44·2 79 41·5185 45·4741 78·5741 2·515 2·883 0·539
35 30 40·8 44 80 42·3785 46·33 82·1087 3·869 5·295 2·636
40 25 43 46 85 42·9171 47·0269 84·8963 0·193 2·232 0·122
50 15 42·5 48 86 43·0306 47·9306 88·2306 1·248 0·145 2·594
Average absolute error 2·795 2·318 1·815

was f 1 . The interaction between f 1 and f 1 contributed maxi- (III) From the result of optimization, the best combination
mum. The interactions between the following variables were of fibre parameters ( f 1 = 50 mm and f 2 = 34·05 (wt%) are
less significant; f 2 × f 2 and f 1 × f 2 , (ii) in the flexu- determined to attain maximum value of tensile, flexural and
ral strength analysis, the most significant variable was f 1 . impact strengths.
The interaction between f 1 × f 2 contributed maximum. The (IV) From the results obtained (table 5), genetic algorithm
interactions between the following variables were less sig- proved as an efficient technique than factorial design method
nificant; f 1 × f 1 and f 2 × f 2 , (iii) in the impact strength in achieving the objective function.
analysis, the most significant variable was fibre length. The
interaction between f 1 × f 1 was maximum. The interac-
tions between the following variables were less significant; Acknowledgements
f 2 × f 2 and f 1 × f 2 , (iv) this paper reports the optimiza-
tion of fibre parameters in table 5 in order to achieve the The authors acknowledge the support of Mr G Tharanidharan
optimal responses. The optimal values were obtained from and S Gnanakumar, Department of Mechanical Engineering,
factorial design and GA are noted in table 5, (v) the opti- Kongu Engineering College, Erode, to carry out mechanical
mum parameter settings are f 1 = 50 mm, f 2 = 34·05% and testing.
in these settings, all responses are gaining maximum values,
ts = 44·698 (MPa), fs = 48·852 (MPa) and is = 92·74
(J/m) and (vi) confirmation experiments were conducted for References
five sets of conditions. The experimental values and the pre-
Cruz-Ramos and Carlos A 1986 Mechanical properties of
dicted values obtained from regression model were com-
reinforced thermoplastics (Springer) p. 65
pared. The percentage of error was calculated using the
Goldberg D E 1989 Genetic algorithms in search, optimization, and
following formula for the validation of regression model machine learning (Addison-Wesley)
(table 6): Jayabal S and Natarajan U 2011 Int. J. Adv. Manuf. Technol. 54 639
Jayabal S, Natarajan U and Sekar U 2011 Int. J. Adv. Manuf.
% of error =(experimental value − predicted value) Technol. 55 263
 Joly C, Gauthier R and Escouben M 1996 J. Appl. Polym. Sci. 61 57
experimental value × 100.
Joseph P V, Kuruvilla Joseph and Sabu Thomas 1999 Compos. Sci.
From table 6, the average absolute error for tensile Technol. 59 1625
strength = 2·795%, flexural strength = 2·318% and impact Noorunnisa Khanam P, Abdul Khalil H P S, Jawaid M,
strength = 1·815% was observed and better accuracy was Ramachandra Reddy G, Surya Narayana C and Venkata Naidu S
2010 J. Polym. Environ. 18 727
obtained using the developed regression models.
Pavithran C, Mukherjee P S, Brahmakumar M and Damodaran A D
1987 J. Mater. Sci. Lett. 6 882
10. Conclusions Rajasekaran S and Vijayalakshmi Pai G A 2005 Neural networks,
fuzzy logic and genetic algorithms (New Delhi: Prentice-Hall of
India) p. 228
The following conclusions were drawn:
Sotton M and Ferrari M 1989 L’industrie Textile 1197 58
(I) The fibre length plays a prominent role in deciding the Woodhams R T, Thomas G and Rodgers D K 1984 Polym. Eng. Sci.
mechanical properties of sisal composite compared with fibre 24 1166
content. Yadav Ranjana, Devi Archana, Tripathi Garima and Srivastava
(II) In the interactions between fibre length and fiber con- Deepak 2007 Euro. Polym. J. 43 3531
tent, fibre length in second-order polynomial is more signifi- Zhang S and Horrocks A R 2003 Prog. Polym. Sci. 28 1517
cant than fibre content in the tensile and impact tests. Zhong J B, Lv J and Wei C 2007 Express Polym. Lett. 10 681