Professional Documents
Culture Documents
1. Introduction
Plant-based natural bers consist mainly of cellulose, which has a high elastic modulus
of 138 GPa in its crystal.[1] It is expected that natural bers will be widely substituted
for conventional glass bers. From the viewpoint of effective utilization for the biomass, the application of composites that consist of the natural bers and biomass-based
biodegradable resin, designated as green composites, has also been studied.[2,3] These
natural bers mainly sustain the loads in green composites. Therefore, it is important
that basic mechanical properties such as the Youngs modulus and tensile strength of
natural bers are estimated precisely as well as the synthetic glass bers. To evaluate
the mechanical properties of natural bers, many researchers have been inclined to
assume uniform shape and size of bers, as they do with synthetic bers.[412]
Additionally, the material reliability of natural bers and synthetic bers have been
evaluated using Weibull statistics.[5,10,1216]
Plant-based natural bers consist of aggregations of many single ber cells, similar
to the kenaf bers portrayed in Figure 1.[17] A single ber cell is also an aggregation
of cellulose micro-brils (CMF), hemi-cellulose, and lignin. The tensile strength and
Youngs modulus of this single ber along the ber axis depend on the CMF angle.
Figure 1.
J. Noda et al.
The specic gravity, strength, and Youngs modulus of these bers also depend on the
lumen, placed in the center of bers.[18,19] Figure 1 shows that the cross-sectional
shape of natural bers is complicated, differing greatly from a circle shape. For textile
engineering, the cross-sectional area is estimated indirectly by the neness as in a
denier.[20] Measurement of the direct cross-sectional area of bers is nevertheless necessary to estimate the stress of bers exactly. Xu et al. precisely estimated the crosssectional area of sisal bers using image analysis based on many cross-sectional area
photographs obtained using optical microscopy.[21] They reported that the conventional
estimation method of the cross-sectional area as a circle shape underestimated the area.
They also reported that the strength was overestimated by approximately two times.
Although their method was able to estimate the cross-sectional area precisely, the specimen used for tensile tests was not directly applicable because cross-sectional photographs were required.
Suzuki et al. estimated the cross-sectional area of kenaf bers using image analysis
with the specimen after tensile tests.[17] They prepared specimens by embedding them
in resin. The cross-sectional shape of the specimens was then observed at 0.2 mm intervals along the ber axis. They reported that the cross-sectional area varied among individual bers and that the cross-sectional area along the ber axis also varied within
bers. The ber strength evaluated using their method was precise because the crosssectional area was estimated exactly. Silva et al. reported that the precise cross-sectional
area of sisal bers was measured using micrographs obtained from scanning electron
microscopy.[15] The cross-sectional shape of natural bers was found to be polygonal,
with 57 corners.[7]
We have proposed a new method of evaluating the ber cross-sectional area precisely, called DB (date base)-based approximation in an earlier paper.[22] This DBbased approximation method is based on the database for a distribution of actual ber
cross-sectional area obtained from cross-sectional pictorial images of ber-reinforced
composites. In this study, the DB-based approximation method was extended to various
polygon shapes in addition to ellipses to elucidate the appropriate shape as the crosssectional area of natural bers and to verify the validity of DB-based approximation.
Additionally, the effects of the cross-sectional area variation on the tensile strength and
Youngs modulus of natural bers were investigated. Finally, a novel Weibull distribution was proposed for the tensile strength of natural bers. The distribution accommodates the cross-sectional area variations that exist between bers and within bers. The
effects of the variations are discussed.
Figure 2.
J. Noda et al.
(d) Hexagon
2
(14874 m )
Figure 3.
(e) Octagon
2
(15701 m )
(f) Dodecagon
2
(17120 m )
(g) Icositetragon
(15658 m2)
(d) Octagon with four diagonal lines equalized to four projective widths along 0,
45, 90, and 135 directions
(e) Dodecagon with six diagonal lines equalized to six projective widths along 0,
30, 60, 90, 120, and 150 directions
(f) Icositetragon with 12 diagonal lines equalized to 12 projective widths
For each image, the cross-sectional areas of the shapes above were measured. The
actual cross-sectional area painted in black was estimated using image analysis. Consequently, the approximate cross-sectional areas were obtained as shown in Figure 3(b)(g)
for the case of the actual cross-sectional area of 12,615 m2. Results showed that these
approximate areas were larger than the actual area.
The relations between the actual cross-sectional area and the circle assumption and
polygon assumption of the icositetragonal shape for kenaf bers are shown in
Figure 4(a) and 4(b). The correlation coefcients between the actual and assumed areas
were, respectively, 0.800 and 0.909: the cross-sectional area using the polygon assumption of icositetragonal shape was estimated more precisely. The other coefcients for
the case of assumed shapes described above are shown in Table 1. These results show
that the coefcients for the circle and ellipse assumption were high values of
0.800.84. The coefcients for the polygon assumption increased to about 0.9 with
increasing projective width. However, results revealed that the increase of the projective
width was not necessary to obtain high coefcients because the increase in coefcients
was saturated in the case of the hexagon shape. This table shows the average crosssectional areas of measurements at the various assumed shapes. The approximate
expression using the least-squares method between the actual areas and areas using the
icositetragon assumption was y = 0.606x + 852. In this study, the cross-sectional areas
measured using the polygon assumption were transformed to their actual areas using
this equation. This method was denominated as a DB-based approximation because the
Figure 4a.
Figure 4b.
Table 1. Coefcient of correlation between real and assumption cross-sectional areas and average areas of measurements.
Fiber
Real
Circle
Ellipse
Hexagon
Octagon
Dodecagon
Icositet.
Coefcient of correlation
Cross-sectional area (m2)
5010
0.800
7356
0.840
7045
0.900
5783
0.880
6338
0.904
6672
0.909
6865
actual areas were estimated based on the database (DB) accumulated from the numerous measurements that were taken.
2.2. Tensile properties of bers
Kenaf single-ber specimens having 10 mm gage length were prepared as shown in
Figure 5. For each ber of the 36 specimens, the projective width along the ber direction was measured using the laser scan micrometer (LSM-500S; Mitutoyo Corp., Japan)
during the shift to the direction of arrows. The measurement length was 8 mm, except
for the 1 mm from the end of the gage length to avoid interference from erroneous
measurements with the paper board of specimens. In all, 81 measurements were taken
J. Noda et al.
Fiber
5
10
10
10
10
[mm]
Figure 5.
per ber along the ber direction at intervals of 0.1 mm. After the rst measurements,
the micrometer was rotated 15, as shown in Figure 6(b). The projective width along
the ber direction was then measured again. Such measurements were repeated from
12 directions of 0 to 165. A typical distribution of the projective width from 0, 60,
and 120 directions is shown in Figure 7. Results claried that multi-angle measurements along the ber direction are necessary because the projective widths are changed
at the direction and position of measurements. Using the obtained projective widths,
Fiber specimen
Laser
(a) First measurement position (0)
Figure 6.
Figure 7.
the cross-sectional areas at each position based on the shape dened by chapter 2.1,
Figure 3(b)(g) were calculated. Then the average cross-sectional area of each ber
was obtained. Then, tensile tests at 0.8 mm/min were conducted using these specimens.
Table 2 shows the average cross-sectional area, Youngs modulus, and tensile strength
at each assumed shape. In this table, the approximation of cross-sectional area using
the icositetragonal shape was the transformed area of icositetragon shape based on the
DB-based approximation. The approximation of Youngs modulus and tensile strength
were the estimated modulus and strength using the transformed area. Consequently,
results suggest that these Youngs modulus and tensile strength were estimated accurately because the calculated area using the DB-based approximation was close to the
actual area shown in Table 1.
3. Effect of cross-sectional area variation on tensile properties
3.1. Youngs modulus
Figure 8 shows the relation between the average cross-sectional area and Youngs modulus for 36 kenaf bers. Figure 9 shows the relation between the coefcient of variation
for the variation in within-ber cross-sectional area and Youngs modulus. From
Figure 8, the Youngs modulus of natural bers decreases concomitantly with the
Table 2.
Cross-sectional
area (m2)
Youngs
modulus
(GPa)
Tensile strength
(MPa)
Figure 8.
Appro.
(Icositet.)
6810
6365
6192
6367
6533
6572
4835
11.6
12.2
12.6
12.2
11.8
11.8
14.8
235
247
256
247
241
240
310
Figure 9.
J. Noda et al.
increase in the cross-sectional area. This cross-sectional area dependence of the Youngs
modulus was also reported by Baley [7] and by Charlet et al. [23]. They investigated
the relation between the diameter of natural bers and Youngs modulus. Aramid bers
reportedly present a similar tendency.[24] Results presented in Figure 9 showed that
the Youngs modulus also decreases with the increase in the coefcient of variation as
well as the average cross-sectional area. The reason why the Youngs modulus
decreases according to the variation of cross-sectional area is discussed in the following
chapter in conjunction with the effects of that variation on tensile strength.
3.2. Tensile strength
Figure 10 shows the relation between the average cross-sectional area and tensile
strength. Figure 11 shows the relation between the coefcient of variation for the variation in within-ber cross-sectional area and tensile strength. Doan et al. reported that
the relation between the tensile strength and the cross-sectional area of bers obtained
from the convenient area measurements was investigated.[25] No dependence of the
Figure 10.
Figure 11.
cross-sectional area on the tensile strength was found. However, because the tensile
strength depends strongly on the area, as shown in Figure 10, the size effect on
strength explained by the increase in the defect probability was expressed in this study.
Figure 11 results showed that the tensile strength also decreases along with the increase
in the coefcient of variation as well as the Youngs modulus.
Figure 1 shows that kenaf bers have a multi-cell structure grouped by some single
cells. The effect of cross-sectional area variation on tensile properties for kenaf bers
arises from the micro-structure of each single cell. The change in the micro-bril angle
of cellulose and the lumen size depends on the growth of natural bers.[18,19] Especially, in the case of bers that have large variation of the within-ber cross-sectional
area, because a part with large local deformation attributable to the small area exists,
the obtained Youngs modulus and tensile strength are expected to decrease.
4. Tensile strength distribution of natural bers
In this chapter, the strength distribution based on the Weibull statistics was proposed
considering the variation of cross-sectional area within bers and between bers. The
effect of these variations on tensile properties was investigated.
4.1. True Weibull parameter estimation except for the effect of cross-sectional area
variation
4.1.1. Conventional Weibull model
A Weibull model used for general bers to show the strength distribution is designated
as a straight bar model that considers only the variation of cross-sectional area between
bers. The cumulative probability of ber failure F is given by the following equation.
m
m
V r
A L r
F 1 exp
1 exp
V 0 r0
A0 L0 r0
(1)
Therein, m and 0, respectively, stand for the Weibull shape and scale parameters,
whereas the gauge volume, area, and length are denoted, respectively, as V0, A0, and
is the average cross-sectional area of a specimen. r is the average failure stress,
L0. A
10
J. Noda et al.
1
ln ln
1 Fi
L
A
ln
ln
m ln ri m ln r0
L0
A0
(2)
4.1.2. Weibull model considering the cross-sectional area variation within a ber
Because natural bers have the variation of cross-sectional area within bers along the
ber direction, a stepped bar model has been proposed in an earlier paper.[22] This
model consists of n cylindrical elements that have random average area and constant
length of x. It was assumed that the weakest ber element (the lowest stress element)
is independently broken on the size of cross-sectional area for neighborhood bers and
that the whole failure results from this element. Based on these assumptions, a Weibull
model considering the variation of the area within bers was proposed as presented
below.
The cumulative probability Fj of the ber failure stress for the jth element in a ber
is given as the following equation.
Vj rj m
Aj Dx rj m
Fj 1 exp
1 exp
A0 L0 r0
V0 r0
(3)
In that equation, Vj = Ajx and VjAj represent the volume and average area for
the jth element. Based on the weakest link model, the ber failure probability is given
as follows.
F 1
n
Y
j1
"
#
n m
Aj
rj
Dx X
1 Fj 1 exp
L0 j1 A0
r0
(4)
Pi
A
r
Aj Aj
(5)
where Pi denotes the failure load for the ith ber. In the case of x 0 considering
the continuum cross-sectional area variation, Equation (4) is expressed as presented
below.
m
ZL m1
r
1
A L
A
F 1 exp
R ; R
dx
L
A 0 L0 r0
A x
(6)
Therein, A(x) denotes a stochastic process that expresses the randomness of the
within-ber cross-sectional area along its axial direction. In the case of a uniform
Equation (6) coincides with Equation (1). However,
cross-sectional area, i.e. Ax A,
11
the ber failure probability is rewritten as Equation (7) when the change in cross-sectional area is measured discretely along the ber axial direction.
n m1
1X
A
n j1 Aj
(7)
1
1 Fi
ln
L
A
ln
ln Ri m ln ri m ln r0
L0
A0
(8)
Results show that the effect of the variation within bers was excluded by the term
ln Ri in Equation (8), and Equation (1) was corrected by the parameter R. The average
strength rF was derived according to the equation presented below.
1
1
A L m 1
rF r0 C 1
R m
m
A0 L0
(9)
1
rj
l
Zl
Pi
rdx
l
Zl
0
dx
2Pi
Ax Aj Aj1
(10)
L
A1
A2
Aj
An
P
l
Figure 12.
12
J. Noda et al.
m
n
m1
r
1 X
2A
A L
F 1 exp
R ; R
n j1 Aj Aj1
A0 L0 r0
(11)
Figure 13.
Table 3.
m
0
(MPa)
Equation
(2)
Equation
(8)
Equation
(18)
Equation
(19)
Equation
(20)
5.61
336
6.90
324
7.45
334
6.68
336
6.86
327
5.88
343
13
(12)
accounted for a stochastic variable between bers and when its probability
When A
the strength distribution Ft is given as shown
density function was dened as gA,
below.
Z
r gA
d A
F A;
m
Z 1
r
A L
d A
1
exp
g A
L
r
A
0
0
0
0
Ft
(13)
It was established that the following formula was derived from Equation (13) [26]
as
" k #
r m
A
Ft 1 exp
r0
A0
(14)
where is a correction index of the size scale. For short expressions, it was assumed
that L = L0 in Equation (14). As described above, some discussions using Equation
(14) were developed for the strength distribution for various bers.[2628] Here, the
Weibull distribution considering the variation of cross-sectional area within bers as
well as between bers was proposed as shown below.
r
; R
F F A;
(15)
The difference between Equations (12) and (15) is the parameter R, which signies
AR
is newly dened here, with A
larger
the variation of areas within bers. Also, A
when R > 1. Equation (15) can be rewritten as presented below.
than A
14
J. Noda et al.
m
A L
r
; r
1 exp
Fj A
A 0 L0 r 0
(16)
From this equation, it was qualied that this strength distribution is a Weibull distri without within-ber variation.
bution with uniform average cross-sectional area A
is dened similarly as g*(A
), Ft* takes
When the probability density function of A
the value presented below.
Z
; r g A
d A
F A
0 Z
m
1
r
A L
d A
1
exp
g A
L
r
A
0
0
0
0
Ft
(17)
Figure 14.
L r m
A
exp
A0 L0 r00
"
#
l1 2
1
A
p2 exp
dA
2S12
2pS1
(18)
15
j and S1 is its
where 1 is the mean of the average cross-sectional area distribution A
standard deviation. The scale and shape parameters, as estimated from Equation (2), are
used as r00 and m0 . For a stepped bar model considering the within-ber variation area,
Ft is given correspondingly as
Z
Ft 1
0
L r m
A
exp
A0 L0 r000
00
"
#
l2 2
1
A
p2 exp
dA
2S22
2pS2
(19)
, and S2 is its
where 2 is the mean of the average cross-sectional area distribution A
j
standard deviation. The scale and shape parameters estimated from Equation (8) are
used as r000 and m00 . For this study, Equations (18) and (19) were solved mathematically.
4.2.3. Probability density function using conditional distributions
The distributions of the average cross-sectional area for kenaf bers were claried as
according with a normal distribution. Results show that the average tensile strength also
depends on the average cross-sectional area, as shown in Figure 10. Because
r
, Equation (13) was reconsidered using the following equation.
Aj A
Z
r g A
j r
d A
F A;
0 Z
m
1
r
A L
j r
d A
1
exp
g A
A 0 L0 r0
0
Ft
(20)
When the two-dimensional normal distribution was dened using stochastic variables and Aj as X and Y, respectively, the conditional distribution of probability density function was given as
2
!2 3
j lY qXY SY =SX r
l
A
1
1
X
j r
5
p
p p exp4
g A
2
SY 1 q2XY
2pSY 1 q2XY
(21)
where X and Y, respectively, denote the means of and Aj. Also, SX and SY, respectively, represent the standard deviations of and Aj. Additionally, XY is the correlation
coefcient. The mean and standard deviation were calculated from experimentally
obtained results used as X and SX. Equation (20) was also solved mathematically.
4.2.4. Weibull parameter comparison
The Weibull parameters using the Weibull models considering the normal distribution
and the conditional two-dimensional normal distribution were estimated from Equations
(18), (19), and (20) for kenaf bers. Figure 15 portrays the obtained tensile strength
distributions as Weibull plots. The obtained scale and shape parameters are shown in
Table 3. For Equations (18) and (19) using the normal distribution, results showed that
the difference of the parameter R, which means the within-ber variation area, was not
large, the estimated shape parameter was larger and the estimated scale parameter was
smaller. However, in the case of Equation (20) using the conditional distribution considering the dependence of cross-sectional area on the tensile strength, both parameters
16
Figure 15.
J. Noda et al.
showed agreement with experimentally obtained results. Results show that the tensile
strength distribution of natural bers considering the variation of the cross-sectional
area within a ber and between bers can be formulated using Equation (20).
5. Conclusions
This study of kenaf bers was the rst to reveal correlation between the actual crosssectional areas obtained from many area optical microscopy measurements and the
assumed areas of circles, ellipses, and polygons having from six corners to 24 corners.
Then, precise shape of the cross-sectional area of natural bers was investigated to
improve cross-sectional area estimation methods. Results claried that the correlation
between the actual cross-sectional area and the assumed polygonal shape area was
higher than that between the actual area and the conventionally assumed circle shape.
Using the estimated cross-sectional area based on the DB-based approximation of
the polygon shape, the effect of cross-sectional area variation on Youngs modulus and
the natural ber tensile strength were investigated. Results showed that the Youngs
modulus and tensile strength decrease along with the increase of the coefcient of variation for cross-sectional area as they do also for the cross-sectional area.
Finally, the parameter R was newly proposed to assess the within-ber cross-sectional
area variation. The true Weibull parameters, excluding the effects of area variation, were
estimated experimentally. Then, a novel Weibull model of natural ber tensile strength
was formulated. It considers both within-ber and between-ber variations. The Weibull
parameters considering the area variation were estimated from this model mathematically.
Consequently, according to the consideration of the dependence of area variation on the
strength, the tensile strength distribution including the area variation was predicted
precisely.
References
[1] Fujii T, Nishino T, Goda K, Okamoto T. Developments and applications of environmentally
friendly composites. Tokyo: CMC Publishing Co., Ltd; 2005. Japanese.
[2] Inao T. Industrial products of plant origin material-effective use of plant origin plastics for
recycling society. Trans. Jap. Soc. Mech. Eng. 2006;109:5152. Japanese.
17
[3] Serizawa S, Inoue K, Iji M. Kenaf-ber-reinforced poly (lactic acid) used for electronic
products. J. Appl. Pol. Sci. 2003;100:618624.
[4] Hornsby PR, Hinrichsen E, Tarverdi K. Preparation and properties of polypropylene composites reinforced with wheat and ax straw bres. J. Mater. Sci. 1997;32:443449.
[5] Lodha P, Netravali AN. Characterization of interfacial and mechanical properties of green
composites with soy protein isolate and ramie ber. J. Mater. Sci. 2002;37:36573665.
[6] Okubo K, Fujii T, Yamamoto Y. Development of bamboo-based polymer composites and
their mechanical properties. Composites Part A. 2002;35:377383.
[7] Baley C. Analysis of the ax bres tensile behaviour and analysis of the tensile stiffness
increase. Composites Part A. 2002;33:939948.
[8] Joffe R, Andersons JA, Wallstrm L. Strength and adhesion characteristics of elementary
ax bres with different surface treatments. Composites Part A. 2003;34:603612.
[9] Gomes A, Goda K, Ohgi J. Effects of alkali treatment to reinforcement on tensile properties
of curaua ber green composites. JSME Inter. J., Series A. 2004;47:541546.
[10] Goda K, Sreekala MS, Gomes A, Kaji T, Ohgi J. Improvement of plant based natural bers
for toughening green composites-effect of load application during mercerization of ramie
bers. Composites Part A. 2006;37:22132220.
[11] Pickering KL, Beckermann GW, Alam SN, Foreman NJ. Optimising industrial hemp bre
for composites. Composites Part A. 2007;38:461468.
[12] Defoirdt N, Biswas S, Vriese LD, Tran LQN, Acker JV, Ahsan Q, Gorbatikh L, Vuure AV,
Verpoest I. Assessment of the tensile properties of coir, bamboo and jute bre. Composites
Part A. 2010;41:588595.
[13] Kompella MK, Lambros J. Micromechanical characterization of cellulose bers. Polym.
Test. 2002;21:523530.
[14] Zafeiropoulos NE, Baillie CA. A study of the effect of surface treatments on the tensile
strength of ax bres: Part II. Application of weibull statistics. Composites Part A.
2007;38:629638.
[15] Silva FA, Chawla N, Filho RDT. Tensile behaviour of high performance natural (sisal)
bers. Compos. Sci. Tech. 2008;68:34383443.
[16] Virk AS, Hall W, Summerscales J. Multiple Data Set (MDS) weak-link scaling analysis of
jute bres. Composites Part A. 2009;40:17641771.
[17] Suzuki K, Kimpara I, Saito H, Funami K. Cross-sectional area measurement and monolament strength test of kenaf bast bers. J. Jpn. Soc. Mat. Sci. 2005;54:887894. Japanese.
[18] Gassan J, Bledzki AK. Modication methods on nature bers and their inuence on the
properties of the composites. J. Eng. Appl. Sci. 1996;2:25522557.
[19] Madsen B, Thygesen A, Lilholt H. Plant bre composites porosity and volumetric interaction. Compos. Sci. Tech. 2007;67:15841600.
[20] The Society of Fiber Science and Technology. Fiber physical science. Tokyo: Maruzen Co.,
Ltd; 1962.
[21] Xu XW, Jayaraman K. An image-processing system for the measurement of the dimensions
of natural bre cross-section. J. Compu. Appl. Tech. 2009;34:115121.
[22] Tanabe K, Matsuo T, Gomes A, Goda K, Ohgi J. Strength evaluation of curaua bers with
variation in cross-sectional area. J. Jpn. Soc. Mat. Sci. 2008;57:454460. Japanese.
[23] Charlet K, Eve S, Jernot JP, Gomina M, Breard J. Tensile deformation of a ax ber.
Procedia Eng. 2009;1:233236.
[24] Steenbakkers LW, Wagner HD. Elasticity and mechanical breakdown of kevlar 149 aramid
bres by a probabilistic approach. J. Mater. Sci. Lett. 1988;7:12091212.
[25] Doan TTL, Gao SL, Mder E. Jute/polypropylene composites I. Effect of matrix modication. Compos. Sci. Tech. 2006;66:952963.
[26] Watson AS, Smith RL. An examination of statistical theories for brous materials in the
light of experimental data. J. Mater. Sci. 1985;20:32603270.
[27] Zhang Y, Wang X, Pan N, Postle R. Weibull analysis of the tensile behaviour of bers with
geometrical irregularities. J. Mater. Sci. 2002;37:14011406.
[28] Curtin WA. Tensile strength of ber-reinforced composites: III. Beyond the traditional
weibull model for ber strength. J. Compos. Mat. 2000;34:13011332.
[29] Andersons J, Sparnins E, Porike E. Strength and damage of elementary ax bers extracted
from tow and long line ax. J. Compos. Mat. 2009;43:26532664.