You are on page 1of 35

C. Zweben,' W. S. Smith,' andM. W.

Wardle'
Test Methods for Fiber Tensile
Strength, Composite Flexural
IVIodulus, and Properties of
Fabric-Reinforced Lanninates
REFERENCE: Zweben, C, Smith, W. S., and Wardle, M. W., "Test Methods for
Fiber Tensile Strength, Composite Flexural Modulus, and Properties of Fabric-Rein-
forced Laminates," Composite Materials: Testing and Design (Fifth Conference),
ASTM STP 674, S. W. Tsai, Ed., American Society for Testing and Materials, 1979,
pp.228-262.
ABSTRACT: This paper examines several aspects of fiber and composite testing,
including: fiber tension test methods, the three-point loading flexural test, and special
problems related to test methods and properties of fabric-reinforced composites. Some
major cause of property variability in hand lay-up laminates are studied. Tensile
strength properties of the important reinforcing fibersglass, graphite, aramid, and
boronare reviewed. Filament, yarn, and resin-impregnated strand test methods are
discussed. Tensile strength theories for untwisted yarns (fiber bundles) and strands are
summarized. A new theory for twisted fibrous structures, such as yarns and ropes, is
presented. Predictions of the various theories are compared with test data for Kevlar^
49 aramid. Standard flexural test procedures presented in ASTM Tests for Flexural
Properties of Plastics and Electrical Insulating Materials (D 790) are reviewed, and it is
shown analytically and experimentally that use of the shorter recommended span-to-
depth ratios can give artificially low modulus values. The importance of considering
characteristic dimensions of material internal structure in selection of test coupon
geometry, a topic ignored in most test methods, is discussed. This criterion is particu-
larly important for fabric composites because yarn bundle size, an important charac-
teristic dimension, can be large. The effects on mechanical properties of two types of
common nonuniformities found in hand lay-up fabric laminatesresin surface layers
and unequal layer thicknessare examined and found to be very significant. Finally,
it is shown that there is an apparent effect of number of plies on flexural modulus
which can be significantly reduced by modifying test data to account for resin surface
layer.
KEY WORDS: composite materials, test methods, laminates, filaments, fibers, yarns,
' Staff engineer, General Electric Space Center, Phila., Pa. 19101.
^Senior research chemist and research physicist, respectively, E. I. du Pont de Nemours &
Co., Inc., Textile Fibers Department, Experimental Station, Wilmington, Del. 19898.
' DU Pont registered trademark.
228
Copyright 1979 by ASTM International www. astm. org

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 229
twisted yarns, strands, fabric, woven roving, glass fibers, fiberglass, aramid fibers,
graphite fibers, boron fibers, tensile strength, flexure, bending, flexural strength,
flexural modulus, shear deformation
Tensile Strength
Introduction
The tensile strength of reinforcing fibers is one of their most important
and least understood properties. Graphite, aramid, and glass fibers are
produced in the form of multiple-filament yarns. The tensile strength of
these materials is measured by testing fibers in several forms: single fila-
ments, untwisted and twisted yarns, and resin-impregnated yarns, which
we shall call strands. Typically, each test method results in a different
strength value. In this section we consider the reasons for this. We begin
by a discussion of single filament strength characteristics and test methods.
We then discuss the strength of untwisted yarns, and how it relates to fila-
ment properties. Following this, we present a new theory for the strength of
twisted yarns, and consider how resin impregnation affects the failure
process and resultant strength of strands.
Filament Properties and Test Methods
The important reinforcing fibersglass, graphite, aramid, and boron
do not have unique tensile strengths. These materials are flaw sensitive,
and the breaking stress of a given filament is a statistical variable [1-6] ^
This flaw sensitivity has two important effects on strength that must be
considered in test design and interpretation of data. First, because the
probability of finding severe flaws increases as length increases, mean
strength decreases with increasing gage length. Second, there is
considerable scatter at a fixed length. Coefficients of variation {v stan-
dard deviation/mean) typically range between 0.1 and 0.3, depending on
the type of fiber. As a rule, coefficient of variation also varies with gage
length. Figures 1 through 4, adapted from the indicated references, show
how mean tensile strength varies with gage length for graphite, boron,
aramid, and glass filaments. Clearly, fiber strength cannot be represented
adequately by a single number. For example, the mean strength of the
graphite fibers of Fig. 1 decreases from about 2410 MPa (350 ksi) to about
1790 MPa (260 ksi) as gage length increases from 12.7 to 25.4 mm (0.5 to
1.0in.).5
There would appear to be two basic requirements for a good filament
test: good alignment and an acceptable number of clamp breaks. The first
""The italic numbers in brackets refer to the list of references appended to this paper.
^Original experimental data were measured in U. S. customary units.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
230 COMPOSITE MATERIALS (FIFTH CONFERENCE)
600
100
HIGH-MODULUS GRAPHITE
0 5 1.0
GAGE LENGTH (IN)
1.5
FIG. 1Variation with gage length of the tensile strength of high modulus graphite fila-
ments [4]. To convert from: psi to Pa, multiply by 6.89 X 10 ; in. to m, multiply by
2.54 X 10-^.
I
500
400
300
200
100
n
1 I I 1 1
BORON
^
(AFTER HERRING)
I I I I I
-
-
-
-
-
5 10
GAGE LENGTH (IN.)
20 50
FIG. 2Variation with gage length of the tensile strength of boron filaments [3]. To con-
vert from: psi to Pa, multiply by 6.89 X 10^: in. to m, multiply by 2.54 X 10~^.
item is obvious; the second requires some discussion. Fibers are loaded by
frictional stresses on their surface in the clamp region. The resultant stress
is uniform in the gage section, decreasing to zero in the clamps. The tensile
stress in the clamp region is not zero, and therefore some breaks here are
expected. However, it is important that the gripping mechanism not
damage fibers and cause excessive breaks outside the gage section. See Ref
7 for a more complete discussion of this problem.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 231
900
I 800
5 700
z 600
500
' ^ 400
: 0-
; K) 300
UJ
I
i
200
100
~\ 1 1 V I T ' I - , -
KEVLAR49 ARAMID
\II1r
_ i I I I 1 I
10
GAGE LENGTH (IN.)
_] I I L_l_ll
i <
5 (t
I-
4 tf)
lu
3
- 2
I
100
FIG. 3Variation with gage length of the mean tensile strength and ultimate strain of
Kevlar 49 aramid filaments [5]. To convert from: psi to Pa, multiply by 6.89 X /O''; in. to m,
multiply by 2.54 X 10'^.
^-
z
Id
900
800
700
600
500
(o - 400
CO
I - Q.
5 m 300
<
UJ
2 0 0 -
100
-
-

1 1 1 1 1 1 1 1 1 1
(AFTER METCALFE 8 SCHMITZ)
1 1 1 1 1
S-GLASS
E-GLASS
1 1 1 1 1
1 1
_
1 1
10
GAGE LENGTH (CM)
100
FIG. 4 Variation with gage length of the mean tensile strengths of S-glass and E-glass
filaments [I]. To convert from psi to Pa, multiply by 6.89 X 10^.
There are two ASTM test methods for filaments: ASTM Test for Tensile
Properties of Single Man-Made Textile Fibers Taken from Yarns and Tows
(D 2101) and ASTM Test for Tensile Strength and Young's Modulus for
High-Modulus Single-Filament Materials (D 3379). Because inorganic
fibers such as glass, graphite, and boron are relatively brittle, they are
easily damaged by pressure clamps, and ASTM Test D 3379 specifies that
they be bonded to cardboard tabs with resin or wax. Aramid fibers are not
as easily damaged, and can be tested with clamps lined with rubber or
another elastomeric material.
Because of the statistical nature of filament strength, the failure of struc-
tures made from groups of fibers, such as yarns or strands, are the end

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
232 COMPOSITE MATERIALS (FIFTH CONFERENCE)
result of complex processes. Therefore, it is not surprising that the
strengths of such structures differ from one another and from those
obtained in single-filament tests.
Untwisted Yarns
The simplest multiple-filament structure is the parallel-fiber bundle,
such as an untwisted yarn. Bundle strength depends on several filament
characteristics: cross-sectional area, stress-strain characteristics, strength,
and initial length (or slackness), all of which are statistical variables.
Reference 8 is the classic work on the subject, which is also studied in Refs
9 and 10, among others.
Mean bundle strength is almost always less than that of the mean
strength of the filaments of which it is composed. To explain this, we
present a very simplistic model. As a bundle is loaded, filaments break
randomly. When they do, the load they support is redistributed among the
unbroken filaments, subjecting them to an increase in stress. Once
filaments begin to break, the stress in the intact fibers is greater than
average fiber stress, defined as load divided by total cross-sectional area
(which includes broken and unbroken fibers). The bundle fails when the
unbroken fibers cannot sustain an increase in load. Because of the failure
process, in which scattered fiber breakage overloads the remaining fibers,
the average stress at failure is less than that of the mean strength of the
filaments tested individually.
Coleman [9] adapts Daniel's earlier analysis [8] and relates the strength
characteristics of bundles to those of the filaments of which they are com-
posed. He assumes that all fibers are linearly elastic to failure with the
same modulus (though this is not essential) and cross-sectional area. He
assumes that filament strength can be represented by a Weibull distribu-
tion of the form
F{a) = 1 - exp {-alan (1)
where a and /3 are parameters of the distribution and / is filament gage
length. The mean of this distribution is
a = a,(/)-' ^sr ( 1 + ^) (2)
where ai = a~^'^ and V is the gamma function.
The fiber strength coefficient of variation, c, is a function only of jS,
which is inversely dependent on scatter.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 233
""+
r . U + , -
- 1 (3)
Coleman shows that the asymptotic mean bundle strength is
at = am)-"^ exp ( - |
The bundle strength asymptotic coefficient of variation is
1
(4)
Cb =
(e-i/fi _ 1)1
(5)
where n is the number of fibers in the bundle.
To obtain the ratio of mean bundle strength-to-filament strength, we
divide Eq 4 by Eq 2
a
/ 3 - ^ e x p ( ^ ) r ( l + I (6)
This ratio is a function of |3 only which, as discussed above, depends solely
on filament strength coefficient of variation. Equation 6 is plotted in Fig.
5. It should be kept in mind that this expression applies when filaments
and bundles have the same gage length.
To determine the accuracy of Coleman's analysis, experimental data
were obtained for the strength of Kevlar 49 aramid filaments and un-
twisted 1420 denier yarn, which contains 1000 filaments. Table 1 presents
the results. Ten replicates for yarns and twenty for filaments were used.
Filament tests were performed using flat, rubber-surfaced grips
according to ASTM Test D 2101, with minor modifications. Yarns were
tested using Instron 4C textile clamps, following the procedures of ASTM
Testing Tire Cords, Tire Cord Fabrics, and Industrial Filament Yarns
Made from Man-Made Organic-Base Fibers (D 885). In many textile
clamps, such as those used here, yarn passes over a curved surface and
then through rubber-surfaced flat grips. ASTM yarn test methods such as
Specifications for and Testing Glass Yarns (D 578), Test D 885 and
Recommended Practice for Sampling Yarn for Testing (D 2258) specify
yarn length from nip to nip. (The nip is where the yarn "enters" the flat
grip.) This length can be much greater than the unsupported yarn length
between tangent points on the clamps. For Instron 4C textile clamps, the
unsupported length is about 64 mm (2.5 in.) smaller than the nip-to-nip

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
234 COMPOSITE MATERIALS (FIFTH CONFERENCE)
0 0.1 0.2 0.3 0.4 0.5
FILAMENT STRENGTH COEFFICIENT OF VARIATION, C
FIG. 5Ratio of mean bundle-to-filament strength as a function of filament strength coef-
ficient of variation [9].
length. Since the test methods specify that breaks occurring on the curved
surfaces or in the flat grips should be ignored, we believe that unsupported
length, rather than nip-to-nip length, should be reported, and that is the
procedure we follow in this paper.
Predicted bundle mean strengths for the two shorter gage lengths are in
good agreement with experimental data. The measured strength of the
longest bundles is greater than predicted. It is more difficult to separate
long filaments from a yarn without damaging them, and this may explain
the discrepancy. Also, some fiber entanglement is unavoidable. Figure 5
shows these three data points with filament strength coefficients of
variation as abscissa and experimental ratio of mean strengths as ordinate.
The variation in mean strength with gage length for filaments and
untwisted yarns (bundles) is shown in Fig. 6, which also presents data for
twisted yarns considered in the next section. We observe that the filament
strength curve is linear on the log-log axes, which is consistent with the
assumed WeibuU strength representation. We also note that the bundle
strength curve roughly parallels that for the filaments, as predicted by the
analysis.
Experimental bundle strength coefficients of variation are significantly
greater than predicted by Eq 5. This may result from factors related to
testing. However, it should be noted that, consistent with the analysis,
bundle coefficients of variation are significantly smaller than those for
single filaments.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 235
I
J
1
s-
1
1
u
l-l
CO
<
H
t
e
&
i/i
jS

g
J
to
a
a
e a
c
a
e
o c
05 u
u
2 g ) !
a
S2
' f i b I
rt lO O
fO v^ -^
o o o
0^ 0^ fN
O O O
O O O
in T^ 00
VO rt CT\
ro r^ <N
g
UI 00
o o o
00 r^
vO O O
^ * m
"3
E
E
o
s

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
236 COMPOSITE MATERIALS (FIFTH CONFERENCE)
tr 500
CO
0- 400
O
C 300
X
c5 200
100
SYMBOLS
O FILAMENTS
A UNTWISTED YARNS
D TWISTED YARNS
_ L _L_ - J _
__
_L-
4 5 7 10
GAGE LENGTH
20
(IN)
30 40 50 70 100
FIG. 6Variation with gage length of the mean tensile strength of Kevlar^ 49 aramid fila-
ments and untwisted and twisted yarn. To convert from: psi to Pa, multiply by 6.89 X 10^;
in. to m, multiply by 2.54 X I0~^.
Twisted Yarns
The introduction of twist in a yarn significantly changes the failure
process and resultant strength characteristics. In an idealized twisted yarn,
the filaments lie along helical paths. The angle between the tangent to a
helix and the yarn axis, designated 0, is a function of radius, r, starting at
zero at the center, and reaching a maximum value 0m at the outer surface.
Because of the twist, the effective modulus of a twisted yarn is lower than
that of an untwisted yarn by a factor cos^ 0 [//, 15],
Twist has another important effect, it causes radial stresses in a loaded
yarn. When a filament in a twisted yarn breaks, the separation of the
broken ends is resisted by interfiber frictional stresses (Fig. 7). The shear
stresses acting on the surface of the fiber "reload" it so that the effect of
the break is localized. At some distance from the break, the stress in a
broken fiber is essentially equal to that in an unbroken fiber. We can
-Rr=
S -
TWISTED YARN
SHEAR STRESSy ^TRANSVERSE STRESS
^ ; ^ ^ l * i ' ^ ' / / i l ^ ^ ' i - ^ ^ ^ BROKEN
FILAMENT
EFFECTIVE BUNDLE LENGTH,A
FIG. 7Model for the tensile strength analysis of twisted yam.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 237
consider the yarn to be composed of a chain of fiber bundles whose length
is a measure of the length of the stress perturbation near a fiber break. We
call this dimension A, the effective bundle length. This is now the mathe-
matical material strength model of Gilcer and Gurland [12] which was
used to model composite materials by Rosen [13] in a similar way.
Although the model proposed here includes a cos^ 9^ term to account
for the reduction in yarn effective modulus caused by twist, we neglect the
effect of radial variation of yarn stress on the statistical failure process.
That is, by using classical bundle theory to represent the strength of the
links, we implicitly assume that the strain in the fibers is uniform. This is a
significant limitation of the theory, as we shall see.
Giicer and Gurland assert that the mode ay* and standard deviation Sy
of the strength distribution of a material consisting of a chain of N links in
series, each link of which is a bundle of fibers, are given by
* = Uy
at - Sbfiim (7)
Sy = StftiN) (8)
where ai, is bundle mean strength, Sb is bundle strength standard devia-
tion, and
4\nN Inln AT In 47r
in which In is the natural, or Napierian logarithm. The number of links in
a twisted yarn is given by N = L/A.
Based on the above assumptions, the predicted mode at* and standard
deviation St of the distribution of a twisted yarn are given by
a,* = ay* cos^ 0 (11)
S, = Sy cos2 e (12)
To use Eq 11 to predict twisted yarn strength, we need to know effective
bundle length, the mean strength, and strength coefficient of variation for
bundles of length A and total yarn length L. We can define A by a trial-
and-error process, or by direct analysis. Bundle strength and coefficient of
variation are determined by direct measurement at length A or by extrap-
olating or interpolating from data obtained at several gage lengths.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
238 COMPOSITE MATERIALS (FIFTH CONFERENCE)
The analytical evaluation of effective bundle length A requires some dis-
cussion. Because the normal stress on a fiber in a twisted yarn is an
inverse function of radial distance from the yarn center, the axial distance
over which fiber stress is perturbed near a filament break, the perturbation
length, is also a function of radius. To define the perturbation length, we
assume that the interfiber shear stress on the surface of a broken yarn is
constant. Figure 8 shows the assumed stress distribution used in the
model. The perturbation length is now simply twice the transfer length, /,,
of Kelly and Davies [14], which is given by
/, =
Ofdf
(13)
where a/ is the stress in the fiber, df is fiber diameter, and T is the shear
stress on the fiber surface.
We assume that the shear stress T is given by
ftffn (14)
where a is the normal stress on the fiber and n, the coefficient of static
friction. A convenient estimate of the normal stress in a twisted yarn is
given in Ref 15. The maximum value, a*, which occurs at the center of
the yarn, is given by
a* = q(Qm) Of (15)
i I 1 1 ( i t 1 i 1 i i t t i t W W i 1 1 i
DC
t f t t t t t r t t t t f t t t t t t t t t f
SURFACE SHEAR STRESS
= * ^
I
I
FIBER TENSILE STRESS
FIG. 8Model for definition of transfer length, I t ' .

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 239
where
q(Q) = 3/4(1 - cos2 0) + 1/2 In (cos 9). (16)
We assume that fiber tensile stress, a/, is constant throughout the yarn.
The perturbation length at the yarn center is now
2liq(Q)
We define the effective bundle length, A, as the perturbation length at a
point where the normal stress is one half the maximum value. This is equal
to twice the perturbation length at the yarn center. That is, an analytical
estimate of effective bundle length is given by
(18)
2fji.q{Q)
Since yarn twist, T, is the physically imposed variable, rather than
surface helix angle 9^, it is convenient to have a relationship between
them. (9m can, of course, be measured directly, but this is tedious.) For
this purpose, we use a modified form of the expression given in Ref 15
tan Qm = 3.73 X IQ-^ ( J T (19)
where D is yarn denier, p is fiber specific gravity, <A is yarn packing factor
(equivalent to volume fraction), and T is twist in turns/metre (tpm). We
assume a packing factor of 0.8.
The model predicts that twisted yarn strength depends strongly on the
mean strength of bundles of fibers of length A. For fibers having a strong
length-strength dependence, A is an important characteristic length. As
twist increases, so does fiber normal stress and interfiber shear stress,
which causes A to decrease. Since bundle strength increases with decreasing
gage length, increasing twist increases one component affecting yarn
strength. However, increasing twist also decreases the cos^ 9 term, which
has the effect of reducing predicted yarn strength. The analysis predicts
that the strength of twisted yarns is governed by these two competing
factors.
Note that the effective bundle length depends on interfiber coefficient of
friction, so that applied finish, if any, can affect twisted yarn strength.
Figure 9 compares predicted and experimental data for the variation of
strength with twist of finish-free, Kevlar 49 aramid 1420 denier (nominal)

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
240 COMPOSITE MATERIALS (FIFTH CONFERENCE)
500
400
n 1 1 1 1 I r
100
1 2 3 4 5 6 7 8
TWIST (TURNS/INCH)
9 10
FIG. 9Variation with twist of the mean tensile strength of Kevlar 49 1420 denier yarn.
To convert from turns/in. to turns/m, multiply by 39.4.
yarn. The actual yarn denier is 1460 and the other parameters used in the
analysis are ii = 0.4 and p = 1.44. Ob was obtained from a plot of mean
bundle strength versus length, and we used an average coefficient of varia-
tion of 0.045 to find 5'*.
The experimental strength reaches a peak at about 79 tpm [2 turns/in.
(tpi)] and then drops off fairly sharply. The theoretical curve tracks the
initial strength rise reasonably well, but does not drop off as fast. We
believe that the poor agreement at higher twist results from neglecting the
variation in tensile strain caused by twist in the statistical strength portion
of the analysis. That is, although the theory accounts for the lower yarn
effective modulus by incorporating a cos^ 9^ term, it neglects the effect of
variation of tensile strain on the statistical failure process. This would seem
to be a fruitful area for future work.
Because the failure process of a chain of bundles is very different from
that of a simple fiber bundle, the analysis predicts a very different length-
strength dependence for these two types of structures. Figure 6 compares
the prediction of the theory developed here with experimental data for
Kevlar 49 aramid 1420 denier yarn twisted to 79 tpm (2 tpi). The two
curves are in reasonably good agreement. We note that, as predicted, the
slope of the twisted yarn curve is much flatter than that for untwisted yarn.
We note that the test methods for untwisted yarns discussed in the previ-
ous section also apply to twisted yarns.
Strands
Resin-impregnated yarnsstrandsare composite materials. The matrix

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 241
provides a mechanism for transmitting load from broken filaments to the
intact filaments in the cross section, as twist does in unimpregnated yarns.
The failure process is a complex one, and a number of possible failure
modes have been identified [5,13,16-21]. We present a brief, simplified
summary of the current understanding of the subject.
As a composite is loaded, scattered fiber breaks occur throughout the
material. The fracture of the first fiber can initiate a propagating crack
that causes catastrophic failure. An expression for the stress level at which
this occurs is given in Ref 16. This mode of failure appears to occur in
some boron/aluminum systems [22]. In most systems, however, first fiber
fracture does not appear to cause composite failure. The explanation for
the difference may be that boron fibers are much thicker than other
reinforcing fibers and release considerably greater amounts of energy when
they break. Further, the rough surface of boron fibers probably provides
better bonding to the matrix, permitting greater energy transfer to it, and
the relatively high strength of the aluminum matrix allows it to transmit
the energy to other fibers in the cross section without fracturing, as a resin
might.
In systems where scattered breaks accumulate, the failure process is
complex. Each break causes a local perturbation in stress. The stress in a
broken fiber is reduced in the vicinity of the break over a distance called
the ineffective length [13]. At the same time, localized tensile stress con-
centrations occur in surrounding fibers which increases the probability that
they will break [16]. A fiber break also results in localized shear stress con-
centrations which can lead to interfacial failure or matrix yielding or
failure. Any of these will reduce the stress concentration in surrounding
fibers and reduce the tendency for fiber breaks to propagate [23].
In Ref 19, it is proposed that the stress at which the first overstressed
fiber is expected to break is a lower bound on the strength of composites
that fail by fiber break propagation. A simple, approximate formula for
this stress is derived in Ref 5. Assuming that the fibers can be represented
by a WeibuU distribution, as in Eq 1, the lower bound stress is
aa = ai[4NL8(ki^ - l)]-'^^^ (20)
where
N = number of filaments,
L = strand length,
8 = ineffective length, and
ki = fiber tensile stress concentration factor for a square array of fibers
= 1.146 [24].
It is also suggested in Ref 19, without proof, that the formula for
composite tensile strength proposed in Ref 13, because it neglects stress

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
242 COMPOSITE MATERIALS (FIFTH CONFERENCE)
concentrations, should be an upper bound on composite strength. The
upper bound stress is given by
ff = a, (^6e)-' ^ (21)
where e is the Napierian base number. It is convenient to use the expression
for ineffective length proposed by Friedman [25]
8 = df
2 \G,
1 - v A
(22)
where
df = fiber diameter,
Ef = fiber extensional modulus,
Gm = matrix shear modulus, and
Vf = fiber volume fraction.
To compare the strength of strands with those of yarns, we impregnated
the Kevlar 49 aramid yarn described in the previous sections with a resin
made from a base resin mixture of (by weight) 100 parts 826 epoxy and 25
parts RD-2 reactive modifier to which was added 60-40 curing agent. The
ratio of resin mixture-to-curing agent was 100/24 by weight. The strands
were passed through a resin bath, wound on a frame, and cured for 1.5 h
at 75C (167F) and 1 h at 150C (302F).
Ten specimens were bonded to cardboard tabs and tested according to
procedures described in ASTM Test for Tensile Properties of Glass Fiber
Strands, Yarns, and Rovings used in Reinforced Plastics (D 2343). (It
should be noted that this method does not specify the impregnating resin
to be used, although this can significantly affect strand breaking strength.)
The average fiber stress at failure was 3410 MPa (495 ksi), and the coef-
ficient of variation was 0.053. The load carried by the resin is negligible,
and is neglected. For comparison. Fig. 6 shows that at a gage length of 254
mm (10 in.), mean filament and twisted yarn strengths are about 2690
MPa (390 ksi), and bundle strength is 2140 MPa (310 ksi). The reason for
the high composite strength will be discussed later.
Obviously, the "Rule of Mixtures," based on 254 mm (10 in.) filament
or yarn strength, does not predict strand strength very well. The extrapo-
lated filament strength at a 25.4 mm (1 in.) gage length, a length frequently
used for testing filaments, is about 3510 MPa (510 ksi), which is reasonably
close to that obtained in the composite. However, we see no logical reason
for selecting filament strength at a gage length of 25.4 mm (1 in.), or any
other arbitrary length, for comparison with composite strength. This sort
of correlation frequently appears in the literature. We believe that, in
general, any agreement that may be found is fortuitous.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 243
We should point out that although there may be an inverse relationship
between composite volume and strength, it appears to be much less severe
than that for filaments, bundles, and even twisted yarns [2,16].
To compare the analytical bounds of Eqs 20 and 21 with experimental
data, we need material parameters of the composite system and geometric
properties of the particular specimen. The material properties are filament
strength Weibull parameters a and /S, and ineffective length, 8, which
depends on fiber diameter, modulus and volume fraction, and resin shear
modulus. The geometric parameters are number of filaments, N, and
length, L, which are 1000 and 254 mm (10 in.), respectively.
The easiest way to estimate the Weibull parameters is to use Eq 2 and
the filament length-mean strength curve. If the strengths at two arbitrary
lengths, / and /*, are Oa and Ob, respectively, then
In
^ = / " ; (23)
> "
Instead of evaluating a, it is more convenient to determine ai which can be
done by using Eq 2 at / or /*. Following this approach, we find that /3 =
9.0 and ai = 3650 MPa (530 ksi).
To find 8, we use > = 124 GPa (18 X 10* psi), G = 1.31 GPa (0.19
X 10* psi) and an estimated volume fraction, v/ = 0.7. This provides an
ineffective length of 0.036 mm (0.0014 in.).
Using these data, we find that the lower and upper bounds are 2780
(403) and 5300 MPa (769 ksi), respectively, which do, in fact, bound the
ten experimental strength values, which range between 3080 (447) and
3620 MPa (526 ksi), with a mean of 3410 MPa (495 ksi). The lower bound,
which is 90 percent of the minimum experimental strength value and 81
percent of the mean, is in better agreement with the data than the upper
bound which is 146 percent of the maximum strength measured and 155
percent of the mean. Similar results were reported for other fiber-matrix
systems [19].
We noted earlier that strand strength is much higher than that of twisted
yarns or single filaments of the same length. We are now in a position to
understand why this is. As we have emphasized throughout this paper,
there is a strong length-strength dependence for most reinforcing fibers.
Therefore, the characteristic lengths associated with fibrous structures,
such as strands and twisted yarns, are key parameters. The characteristic
length for filaments and bundles is gage length. However, for twisted
yarns, the effective bundle length is also an important dimension, as is
ineffective length for composites. We find that for a twist level of 79 tpm

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
244 COMPOSITE MATERIALS (FIFTH CONFERENCE)
(2 tpi), which provides the maximum strength for the yarn tested, A is
about 5.1 mm (0.2 in.). The ineffective length for the strands, however, is
0.036 mm (0.0014 in.), more than two orders of magnitude smaller. Using
Eq 2 with the values of a and (3 presented above, mean filament strength
corresponding to 5 is 7170 MPa (1040 ksi), which is much greater than the
value of 4130 MPa (600 ksi) associated with A = 5.1 mm (0.2 in.). There-
fore, although the failure processes are different, it is not surprising that
strands are stronger than twisted yarns. (We note that although the
accuracy of data obtained by extrapolation over several orders of
magnitude may be questionable, the data provide useful qualitative infor-
mation for the purpose of illustrating the importance of reference length.)
Summary and Conclusions
In this section we have examined briefly the tensile strength characteris-
tics of some of the major reinforcing fibers and how these properties affect
the failure processes and resultant strengths in filament, yarn, and strand
tests. We have proposed a new theory for the strength of twisted yarn
which explains some of the phenomena observed in these structures.
Quantitative agreement with data for length-strength dependence is fairly
good. Predicted strength increase for small twist angles is reasonably
accurate, but at larger angles strength drops off much faster than
predicted. We believe that the discrepancy at large twist angles results
from neglecting the variation in fiber strain caused by twist in the statistical
strength portion of the theory. The failure process in resin-impregnated
yarns (strands) was reviewed and data compared with analytical bounds for
failure associated with the accumulation of scattered fiber breaks. The
experimental data fell within the bounds, but the upper bound was signifi-
cantly above the maximum strength measured.
Based on the results of this study, we conclude that to get an accurate
picture of fiber strength properties, it is necessary to test individual fila-
ments, preferably at more than one gage length. Testing fibers in struc-
tures such as untwisted yarns, twisted yarns, or strands masks many of the
basic fiber strength characteristics because they fail as a result of complex
failure processes. Yarn or strand tests may, however, be useful for material
screening or quality control.
Flexural Modulus Test
The three-point flexure test, used to determine flexural strength and
modulus, is one of the most common and also one of the most misunder-
stood tests for composite materials. Because of its simplicity and economy,
it is widely used to screen materials, and, frequently, to define material
properties. Figure 10 shows a schematic representation of the test, which is

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 245
COMPRESSION (-)
=Z] D I D
^P/ 2 H
max - 2 ^QZ
' ^ * " 4 bD
FIG. 10Three-point loading flexure test.
covered by ASTM Test D 790. This test specifies procedures that provide
reliable strength data, but, as we shall see, moduli values can be
misleading.
It is fair to say that the test measures flexural properties of laminates
rather than material properties. A good test for a material mechanical
property should have a single, uniform stress component in the gage
section, particularly when strength values are desired. The flexure test
satisfies neither criteria. According to simple Bernoulli-Euler elastic beam
theory, the state of stress is as shown in Fig. 10. The extensional stress
varies linearly from compression on the top face to tension on the bottom.
In addition, there is a parabolic distribution of shear stress across the
beam. The actual state of stress, particularly in the region of the supports
and central load, is far more complex [26]. We note that simple beam
theory only applies for small deflections. When deflections are large, as is
frequently the case for long beams, the simple formulas for stress must be
corrected, as discussed in ASTM Test D 790.
From a simplistic standpoint, the beam can fail in tension on the bottom
face, compression on the top face, by shear, or by some interaction of these
stresses. The mode of failure is strongly dependent on span-to-depth ratio
(sometimes called aspect ratio), L/D. Short beams usually fail in shear and
long ones by tension or compression [28]. Generally, the flexural stress at
failure is greater than either the tensile or compressive strengths of the
material from which it is composed [27]. This is explained by the nonuni-
formity of stress in the flexure test. The maximum tensile and compressive
stresses exist only in the outer fibers of the beam. Because of the scatter in
strength, the probability of failure in a small volume of material at a given
stress level is lower than that in a large volume. Therefore, for equal prob-

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
246 COMPOSITE MATERIALS (FIFTH CONFERENCE)
ability of failure, the outer-fiber stress in a flexure test must be higher than
in the same beam loaded in pure tension or compression.
As we stated earlier, ASTM Test D 790 procedures provide reliable
strength values. Particular attention should be paid to the large deflection
corrections specified for long beams. Figure 11 shows corrected and
uncorrected flexural stress-strain curves for two laminates, one made from
five plies of 0.82 kg/m^ (24 oz/yd^) E-glass woven roving (WR) and the
other five plies of Kevlar 49 aramid Style 1033 WR. Both used a polyester
resin.
We observe that the uncorrected E-glass stress-strain curve for L/D =
50 departs from that for L/D = 20 and displays significant nonlinearity.
However, the corrected curve falls right on top of the one for the smaller
L/D ratio, indicating that the greater nonlinearity in the curve for L/D =
50 results from specimen geometry, and not material properties.
On the other hand, Kevlar 49 aramid "yields" when loaded in compres-
sion, and the flexural stress-strain curve does in fact display a strong non-
linearity. Because the horizontal thrust correction in ASTM Test D 790 is
80
60
40
20
1
KEVLAR49, j
E- GL ASSx A/ /
///
/^^^^^--^
w
' 1
1 1
E-GLASS
, xL/ D-20
.X
/ L / [ ) = 5 0 CORRECTED _
/ L / [ , = 5 O UNCORRECTED
/ / ^
KEVLAR *49ARAMID
S-1033
%= 2 0
^ L/ Q =50 CORRECTED _
L/ Q =50 UNCORRECTED
-
1 1
1.0 2.0
STRAIN (%)
3.0
FIG. 11Effect of horizontal thrust correction on stress-strain curves for beams in three-
point loading. To convert from psi to Pa, multiply by 6.89 X 10-^.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 247
based on the deflection of an elastic beam, it does not strictly apply to
beams with material nonlinearity. Therefore, the corrected curve for L/D
= 50 does not coincide with that for L/D = 20. It is recommended that
strength tests be run at a span-to-depth ratio of 16/1 to eliminate effects of
geometric nonlinearity, particularly when the material itself displays some
nonlinearity, as does Kevlar 49 aramid in compression.
There is an important point about flexural modulus that requires some
consideration. Consider first a unidirectional composite in which the fibers
are uniformly distributed throughout the cross section, or are distributed
randomly so that the distribution is homogeneous in a statistical sense. If
the axial tensile and compressive moduli of such a material are equal, then
the flexural modulus should equal the extensional modulus.
Next, consider a laminate made from many layers of fabric, all of which
are oriented in the same direction. If the fabric is spaced uniformly
through the thickness, and compressive and tensile moduli are equal, then
the flexural modulus should equal extensional modulus. The case of a
laminate made from a few layers of fabric is discussed later. In the rest of
this section, we consider materials that meet the criteria discussed above.
It is common to find in the open literature flexural moduli that are sig-
nificantly lower than extensional moduli for the types of materials just
described. The most probable reasons for such discrepancies would appear
to be nonuniformity of reinforcement spacing or improper test procedures.
In a later section, we consider some types of nonuniformities that occur in
the hand lay-up process for fabric laminates.
As indicated earlier, the recommended procedures in ASTM Test D 790
are not always reliable for measuring flexural modulus. This is because
there is no consideration of the contribution of shear deflection to total
beam deflection. The significance of shear deflection in composite
materials loaded in three-point flexure has been considered theoretically
[28,29], but does not appear to be widely recognized throughout the indus-
try. We briefly review the theory and present some experimental data to
support the analysis.
As discussed earlier, both shear and flexural stresses exist in a beam
loaded at three points. Both types of stress contribute to beam deforma-
tion, as illustrated in Fig. 12. According to the BernouUi-Euler theory, the
deflection arising from flexural stress in a rectangular beam is
where P is load applied at center, L is span length, b is beam width, D is
beam depth, and E is beam extensional modulus.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
248 COMPOSITE MATERIALS (FIFTH CONFERENCE)
SHEAR
^= 4 GAs
BENDING
AK - -L t ?
^b " 4 EbD'
FIG. 12Beam deflection components under three-point loading.
Shear deflection in a centrally-loaded beam is given by (see Ref 30, for
example)
A, =
PL
AGA,
(25)
where G is the shear modulus in a plane defined by the beam axis and the
normal to the beam mid-plane, and As is effective cross-sectional area for
shear deformation. For a rectangular beam, As = bD/\.l, so that
A. =
0.3 PI
GbD
(26)
Because the ratio of shear modulus-to-extensional modulus is much lower
for composites, shear deformation is more important for composites than
for isotropic materials like metals.
The formula for computing flexural modulus presented in ASTM Test D
790 is based on Eq 24, which implicitly assumes that all deformation
results from bending, and ignores shear deflection. In a flexure test we
measure total deflection at the center. A,, which is the sum of bending and
shear components, that is. A, = A/ + Aj. The apparent flexural modulus
is given by
Ea =
PV
AbD^MX -f A,/A/)
(27)
To assure that the apparent flexural modulus is equal to the true
flexural modulus, shear deflection must be small compared to flexural
deflection. This requires that
t=-(f)(fr'
(28)
This term depends on the ratio of composite extensional modulus-to-shear

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 249
modulus, which is significantly greater for composites made from high-
modulus fibers, such as graphite and aramid, than for glass fibers. This
means that span-to-depth ratios must be greater for advanced-fiber com-
posites. However, even for glass-reinforced plastics (grp), prescribed L/D
values can be too low to give valid flexural modulus data, as we shall see.
Figure 13 shows how apparent flexural modulus varies with L/D for a
pultruded composite consisting of unidirectional Kevlar 49 aramid fibers
in a polyester matrix. As L/D increases, the apparent flexural modulus
asymptotically approaches the tensile modulus, as predicted by theory.
ASTM Test D 790 suggests L/D ratios of as low as 16/1. For this L/D, the
apparent flexural modulus is only about 65 percent of the true value.
Figure 14 shows how apparent flexural modulus varies with L/D for five-
ply laminates of E-glass 0.82 kg/m^ (24 oz/yd^) woven roving (WR) in
polyester resin. Figure 15 shows the same curve for five-ply laminates of
Kevlar 49 aramid Style 1350 WR in the same resin. In both cases, L/D
should be significantly greater than the values of 16/1 and 32/1 suggested
in ASTM Test D 790.
The results of this study suggest that to obtain valid data, flexural
modulus values should be determined using at least a 60/1 span-to-depth
ratio. To obtain good strength values, the same specimens can then be
o
o
16/1
I 1 1
^TENSILE MODULUS
^APPARENT FLEXURAL
MODULUS
- 40/ 1 -60/ 1
32/ 1
_L-
I
20 40 60 80 100
SPAN-TO-DEPTH RATIO, L/ D
120
FIG. 13Apparent flexural modulus as a function of span-to-depth ratio for a composite
with unidirectional Kevlar^ 49 aramid fibers in a polyester resin matrix. To convert from psi
to Pa, multiply by 6.89 X 10^.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
250 COMPOSITE MATERIALS (FIFTH CONFERENCE)
20 30 40 50
SPAN-TO-DEPTH RATIO, L/ D
FIG. 14Apparent flexural modulus as a function of span-to-depth ratio for 5-ply lami-
nates of E-glass woven roving in a polyester resin matrix. To convert from psi to Pa, multiply
by 6.89 X 10^.
-16/ 1 32/1 -40/1
I
-60/ 1
10 20 30 40 50 60
SPAN-TO-DEPTH RATIO, L/ D
70
FIG. 15Apparent flexural modulus as a function of span-to-depth ratio for 5-ply lami-
nates ofKevlar 49 aramid style 1350 woven roving in a polyester resin matrix. To convert from
psi to Pa, multiply by 6.89 X 10^.
broken at L/D 16/1. If this procedure is followed, care should be taken
not to damage specimens while determining modulus. Naturally, different
specimens can be used for modulus and strength, if desired. Alternatively,
if the shear modulus is known, Eq 27, which includes a shear correction
term, can be used to determine flexural modulus.
In summary, the procedures described in ASTM Test D 790 are adequate
for determining flexural strength, but, because the influence of shear
deformation is not considered, they can give misleading flexural moduli if

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 251
short spans are used. Correct values can be obtained by using large span-
to-depth ratios or a shear deformation correction.
Hand Lay-Up Fabric Laminates
Introduction
Glass fabric reinforced plastics have been used for many years for a
variety of applications ranging from boats to semistructural aircraft com-
ponents such as fairings and radomes. Although the reinforcement ef-
ficiency of fabrics is less than that of straight aligned fibers, such as are
found in prepreg tapes, they are much easier and cheaper to work with.
Because of this, recently there has been a strong trend toward the use of
graphite, aramid, and hybrid fabrics [31,32].
Normally, preimpregnated fabrics are used to make aircraft parts, but
most nonaerospace glass fabric structures are made by a wet lay-up
process, in which resin is combined with dry fabric at the time of fabrica-
tion. When the resin is applied manually, the procedure is called hand lay-
up. The resin content in laminates made from prepregs can be controlled
much more accurately than for those made by hand lay-up.
Aerospace components generally are vacuum bagged and cured in an
autoclave under pressure and temperature, which produces a uniform,
void-free material. Hand lay-up structures, on the other hand, are
generally made on a single convex or concave mold, and cured at room
temperature and atmospheric pressure, a procedure called contact
molding.
Because of the differences in fabrication methods, the physical and
mechanical properties of vacuum bagged, autoclave cured laminates made
from prepregs are significantly different from those that are laid up by
hand and contact molded. Prepreg fabric laminates have a relatively high
fiber volume fraction (0.5 to 0.7), low void content, and uniform internal
structure. The fiber volume fraction for hand lay-up laminates is generally
much lower (0.25 to 0.5), void content is higher, and the internal structure
can be extremely nonuniform.
Because of the factors discussed above, the mechanical properties of
hand lay-up composites are extremely variable. In this section we examine
some factors affecting laminate properties, including: thickness of resin
surface layer, eccentricity of reinforcement, and local variability of resin
content. We also examine cursorily how the number of plies affects flexural
modulus. Before we begin consideration of these items, some discussion of
test methods is required.
Composites are heterogeneous materials consisting of two very distinct
phases, fiber and matrix. For testing and design purposes, they are usually
treated as homogeneous anisotropic materials. For this assumption to be

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
252 COMPOSITE MATERIALS (FIFTH CONFERENCE)
valid, the dimensions of the structural component or specimen should be
large compared to the characteristic dimensions of the material internal
structure. For example, for a unidirectional composite, specimen width
and thickness should be large with respect to fiber diameter and fiber
spacing. When specimen dimensions are of the same order of magnitude as
structure dimensions, use of analytical formulas developed for
homogeneous materials, such as that for beam flexural stress, is in
question.
For fabrics, yarn width and fabric thickness are important characteristic
dimensions. Therefore, fabric laminate specimens should have many plies,
and width should be large compared to ply thickness and yarn width.
Another reason for using wide specimens is to minimize possible effects of
machining damage. Further, if there is any fabric skewing, some yarns
intersect specimen edges, which are stress free. The portions of these yarns
near the free edge will not be loaded to the same stress level as a
continuous yarn, and this can result in a lower apparent modulus. This
effect is less severe in a wide specimen with many yarns (note that skewing
itself reduces the real axial modulus, independent of specimen width).
For woven rovings, coarse fabrics commonly used to make boats and
other large structures, yarn width can be as large as 6.4 mm (0.25 in.) [Ply
thickness is much smaller, about 0.8 mm (0.3 in.).] Therefore, laminate
specimens for these materials should be at least 50 mm (2 in.) wide. It
should be noted that ASTM Test D 790 does not consider material struc-
ture dimensions in specifying specimen dimensions.
Effect of Resin Surface Layer on
Laminate Extensional and Flexural Moduli
As discussed earlier, the extensional modulus, Et, and flexural modulus,
/, of a laminate are equal when it is composed of many identical layers.
When there is some nonuniformity in the laminate, Ef and Et will generally
be different. In particular, a layer of resin on either or both of the surfaces
can significantly affect flexural modulus, and we consider this phenomenon
here.
Because fiber extensional modulus is much greater than resin modulus,
the direct contribution of resin to the ability of a fabric-reinforced laminate
to resist extensional and flexural deformation is usually small, and can be
neglected as a first approximation. Therefore, layers of resin on the outer
surfaces of a laminate have little effect on the deflection of a laterally
loaded beam or plate. However, surface resin layers do contribute to the
laminate thickness used in computing extensional and flexural moduli, and
can have a significant effect on the computed flexural modulus, as we shall
see.
Consider a laminate of total thickness t which has layers of resin on

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 253
either or both surfaces of combined thickness At, so that the thickness of
the portion of the laminate containing reinforcing fibers is
ta = t - At (29)
As discussed above, the contribution of the resin layers to resisting deflec-
tion can be ignored, to a first approximation, unless they are extremely
thick. That is, the lateral deformation of laminate of thickness t having a
total resin thickness A^ on the surfaces is almost the same as that of a
laminate of thickness to having no surface resin layers under the same load.
Similarly, the extensional deformations are the same.
We denote the flexural and extensional moduli of the laminate of thick-
ness t by Ef' and Et'. The corresponding properties of the laminate without
resin layers are Ef and Ei. Equating the laminate deflections under the
same load, P, we obtain the relationship from Eq 24 (with D replaced by t
and to).
E/=Ef(^jj (30)
Equating deformations under the same extensional load, we find
E,'=EJJ\ (31)
Using Eq 29 we can rewrite Eqs 30 and 31 in terms of total thickness and
resin layer thickness
AtV _ / . 3At^
E/=^E,^l--j - Er i ^ l - - ) (32)
E,' =E,(l-j\ (33)
We see that the resin layer has a much more severe effect on the effective
flexural modulus than on effective extensional modulus. We emphasize
again that the resistance to flexural and extensional deformation of the two
laminates is the same, but because of the extra resin, the effective moduli
of the thicker one are lower. To put it another way, as Eq 24 shows,
deflection under lateral load is governed by the beam flexural rigidities
Ef't^ and E/to^, which are equal, although the flexural moduli Ef' and Ef
differ. Similarly, the extensional stiffnesses E, 't and Etto are equal even
though extensional moduli diff'er.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
254 COMPOSITE MATERIALS (FIFTH CONFERENCE)
Figure 16 shows how the ratios of flexural and extensional moduli,
Ef'/Ef and Et'/Et, vary with the total surface resin layer thickness, A^,
normalized with respect to total laminate thickness t. We see that At has a
significant effect, particularly on the ratio of flexural moduli, because of
the cubic dependence on At/t.
The figure also shows how the ratio of flexural-to-extensional thickness
Ef'/Et' varies with At/t. We see that when there are layers of resin on
either or both surfaces, the effective flexural modulus of a laminate can be
significantly lower than the extensional modulus. For example, consider a
2.5 mm (0.1 in.) thick laminate with a total surface resin layer At of only
0.13 mm (0.005 in.), about the thickness of a piece of bond paper. The
analysis predicts that the flexural modulus will be about 10 percent lower
than the extensional modulus, a significant difference.
It is possible to measure surface layer thickness directly, but this may
require polishing specimen edges, which is time-consuming. An alternative
approximate approach is to make laminates with different numbers of plies
and observe how thickness varies with this parameter. Figure 17 shows how
thickness varies with number of plies for hand lay-up laminates of 0.82
kg/m^ (24 oz/yd^) E-glass woven roving in polyester resin. A polyester film
was used on top and bottom to provide smooth surfaces. We observe that
the points fall on a straight line, but the intercept on the vertical axis is
about 0.25 mm (0.010 in.). This suggests a fairly uniform ply thickness of
about 0.67 mm (0.0265 in.) and a total surface resin layer At of 0.25 mm
(0.010 in.). We will use this information on ply and resin layer thickness
later on.
We have found that both ply thickness and At vary significantly from
one laminator to another, and this is another source of variability. The
variability throughout the industry reported in Ref 33 is even more
0. 04 0. 08 0,12
NORMALIZED SURFACE THICKNESS, A l
t
FIG. 16Influence of resin surface layer on flexural and extensional moduli.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 255
280
240
200
160
120
80
40
n
1 1 1 1
-
-
/
/
- cf
1 1 1 1
I I I
/ -
/
/
-
-
-
-
2 4 6 8 10
NUMBER OF PLI ES
FIG. 17Variation of thickness with number of plies for E-glass woven roving/polyester
and lay-up laminates. To convert from in. to m, multiply by 2.54 X 10~'.
substantial. For example, tensile modulus was found to vary by a factor of
about 2.5. Obviously, this has significant implications for design and
reliability.
The results of this analysis indicate that flexural properties are extremely
sensitive to Ai, which is highly variable in a hand lay-up composite. This
suggests that caution should be exercised in interpreting test results. We
see that tensile modulus is much less sensitive to A^ and is probably a
better test to use when comparing materials.
Effect of Nonuniform Ply Thickness
In this section we consider another type of laminate nonuniformity that
can have a significant effect on material properties. When fabric layers are
not evenly spaced, a situation that is quite common, especially when resin
content is high, the laminate neutral axis is not at the midplane. This
results in extensional-bending coupling. When a tensile coupon is loaded in
a testing machine, bending stresses, as well as tensile stresses, are induced
in the specimen. As a result, strain distribution is not uniform throughout
the specimen. The strain on one surface is higher than the mean strain,
and it is lower on the other surface. Therefore, if extensional modulus is
measured by a single strain gage bonded to one surface, the computed
modulus will be either greater or less than the actual one. To overcome this
source of error, gages should be placed on both sides of the coupon.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
256 COMPOSITE MATERIALS (FIFTH CONFERENCE)
To assess the magnitude of the eccentricity effect, we use a simple rect-
angular beam model. We emphasize that the load is not eccentric. It acts
along the beam midplane. However, because of the nonuniformity of rein-
forcement spacing, the neutral axis, which is essentially the centroidal axis
of fiber cross-sectional area, is offset from the beam midplane.
As a first approximation, we neglect the effect of elastic deformations on
stress. This is a "conservative" assumption, which overestimates the
magnitude of bending stresses arising from the eccentricity. We consider a
beam of unit width and depth t. Let the distance between the central plane
and neutral axis be e. Under a tensile load P applied along the central
plane, the stresses at the outer surfaces are
t \ t
(34)
The average tensile stress in the beam, a = P/t, is unaffected by the bend-
ing stress.
To assess the magnitude of the stresses caused by eccentricity, we plot
Om/a as a function of e/t in Fig. 18. We see that an eccentricity of 2 per-
cent produces a bending stress that is 12 percent of the mean stress. As
discussed above, this would result in a 12 percent error in extensional
-NEUTRAL AXIS
<
a:
0.02 0.04 0.06 0.08
ECCENTRICITY, e/ t
0.10
FIG. 18Effect of reinforcement eccentricity on the ratio of bending-to-extensional stress.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 257
stiffness if a single strain gage were used. Based on examination of
numerous hand lay-up laminates, we believe that eccentricities of 2 percent
or greater are not uncommon.
Our discussion has emphasized the effect of eccentricity, and resultant
extensional-bending coupling, on test procedures. However, it is important
to recognize that coupling can reduce buckling loads significantly (see Ref
34, for example), and this should be considered in designing with materials
produced by hand lay-up.
Preliminary data suggest that local variation in ply thickness may also
affect strength properties. A five-ply laminate made from an experimental
aramid woven roving was observed to have a concentration of fabric layers
near one surface, resulting in a localized region of high fiber volume frac-
tion. The plies on the other surface were spread out, consequently, local
fiber content was low. Six specimens were loaded to failure in three-point
bending at an aspect ratio of 16/1. Three had the resin-rich region on the
top (compression) face and three on the bottom (tensile) face. All
specimens failed in compression on the top face, the typical failure mode
observed in flexure tests of laminate reinforced with woven roving made
from glass or aramid fiber. The specimens with the resin-rich regions in
compression failed at an average load of 1446 N (325 Ibf), compared to
only 1032 N (232 Ibf) for the other case.
These rather sketchy results suggest that local variations in resin content
can affect significantly the compression strength of woven roving
laminates. This is not surprising since in a composite the matrix is
required to support the fibers and prevent them from buckling. This is
particularly important in fabric-reinforced composites where fibers have
significant curvature.
Effect of Number of Layers on Extensional and Flexural Moduli
There appears to be relatively little detailed information in the open
literature about the properties of fabric-reinforced composites. One
important question that needs to be answered concerns the effective
flexural moduli of fabric-reinforced composites made from a small number
of plies. We take a preliminary look at this problem here.
As discussed earlier, a basic requirement for treating a structural ele-
ment or specimen as an effectively homogeneous continuum is that its
dimensions be large with respect to material characteristic dimensions.
This requirement is not satisfied for laminates made from a few layers of
fabric since thickness is of the same order of magnitude as such character-
istic dimensions as yarn bundle width. Despite this consideration, it is
convenient to treat composites made from one or more fabric plies as
effectively homogeneous materials.
As discussed earlier, when reinforcement is uniformly distributed

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
258 COMPOSITE MATERIALS (FIFTH CONFERENCE)
throughout a material, its flexural and extensional moduli are equal. A
single-ply fabric laminate does not satisfy this criterion. Fibers are
combined in yarns which are woven together, creating a complex internal
structure. Therefore, there is no reason to expect, a priori, that the exten-
sional modulus, Eto, and flexural modulus, E/n, of a single-ply laminate
are equal. In principle, JS'/o can be greater or less than Eto-
Consider next the properties of an n-ply laminate made from identical
layers with extensional and flexural moduli, ',0 and Eja- It can be shown
that the extensional modulus of a beam cut from this laminate is Et =
Eto, and its effective flexural modulus is
Ef = (l- ^) E.0 + (4-) ^^ (35)
We see that as the number of plies increases, the beam effective flexural
modulus approaches the extensional modulus, ^(O. When E/o and Eto are
equal, the beam flexural modulus is equal to its extensional modulus for
all values of n.
Now, let us look at some experimental data for flexural modulus of hand
lay-up fabric laminates as a function of the number of plies. Since resin
content can vary significantly, at the minimum some correction for fiber
volume fraction must be made to compare data from different laminates.
The most accurate way to do this is to measure volume fraction, v/, and
normalize data to some average value. It should be noted that the variation
of modulus with fiber volume fraction may not be linear for fabric com-
posites. Therefore, linear normalization ma^ not be accurate if the measured
Vf is significantly different from the reference value. For simplicity, we
assume that a linear correction is acceptable. Further, we assume that the
fabric is uniform, so that the fiber volume per unit area for each ply is
constant for all laminates. Based on this assumption, selecting a reference
ply thickness is equivalent to selecting a reference volume fraction.
The material system studied is the E-glass woven roving/polyester for
which Fig. 17 presents the variation in thickness with number of plies. It
will be recalled that the best fit line through the data points corresponds to
a ply thickness, to, of 0.67 mm (0.0265 in.) with a total surface resin layer
thickness of 0.25 mm (0.010 in.). In a previous section it was shown that
the present resin surface layer has a significant effect on measured flexural
modulus. To account for this, we have normalized data to a reference
laminate thickness by multiplying measured data by the ratio it/nt(,y,
where t is measured laminate thickness. For comparison, we have made a
"straight" volume fraction correction by multiplying by t/nto. Table 2
presents measured tensile and flexural moduli for one-, two-, three-, five-
and ten-ply laminates, along with measured, reference and computed resin

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 259
5
a
IS
s
I
s
pa
<
o
Z
..s
vC lO -tt ^ 00
fO ro ^ l^_ '-<
fO f ) ro <^ ''*'
--H (N m f^ r o
rO r*^ r'l f^ r*!
i-H <N ro ol f^
fS r' j ro r' l f i
i/> O lO i/> O
f s o ^' --H od
I/) O lO I/)
5'"
=1=
ve fO CJ^ fNI lO
(N lO I^ r*^ ' ^
i-< f S
.a "
^ So
ON ^ ^ ^ r*!
fo 1/) 00 -^ r-'
TH <N CD l O O

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
260 COMPOSITE MATERIALS (FIFTH CONFERENCE)
layer thicknesses. Tensile moduli normalized with respect to t/nto and
flexural moduli with t/nto and {t/ntoY corrections are also given.
Flexural moduli with a t/nto correction are shown by circles in Fig. 19.
We see that there is a strong variation of this quantity with number of
plies. However, use of a {t/ntoY correction (triangles) significantly reduces
the variation of flexural moduli for small values of n. Tensile moduli,
normalized with respect to t/nto, are indicated by squares. We note that
tensile modulus is relatively constant over the range studied, but even the
(t/ntoY normalization does not entirely remove the dependence of flexural
modulus on number of plies.
Based on the limited data in this study, we conclude that the apparent
strong variation of flexural modulus with number of plies may well be a
surface resin layer effect. However, we note that the {t/ntoY correction
does not completely eliminate this dependence. Additional study is
required to determine whether the effect is real.
Summary and Conclusions
In this section we have examined some aspects of hand lay-up fabric lami-
nates that affect tensile and flexural properties. We found that the presence
of a resin surface layer can reduce significantly the effective flexural
modulus, and that nonuniform spacing causes extensional-bending coupling
that can lead to incorrect tensile moduli measurements if only one strain gage
is used. We noted that this coupling can reduce structural buckling loads.
Finally, we briefly considered the influence of number of plies on tensile and
flexural moduli and found that a surface layer correction significantly
5.0
4.0
ID
O 3.0
in
2 2.0
o
s
1.0
&
0
-
1 1 1
0
o
SYMBOLS:
TENSILE
1 1 1 1 1 1
A
0 ^
A
n n
0
-(| ) CORRECTION
0 FLEXURAL - ( ^ ) CORRECTION
A FLEXURAL - (7^CORRECTION
1 1 1 1 1 1 1 1 1
3 4 5 6 7 8 9
NUMBER OF PLIES, n
10
FIG. 19Tensile and flexural moduli of hand lay-up, E-glass woven roving laminates as a
function of number of plies. To convert from psi to Pa, multiply by 6.89 X 10 .

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
ZWEBEN ET AL ON TEST METHODS 261
reduces dependence of effective flexural modulus on n. The properties of
fabric laminates is a topic requiring further study.
Summary of Recommendation for Testing Fabric-Reinforced Composites
In this paper we have considered several aspects of test methods for
fabric-reinforced composites. Here, we summarize our conclusions and
recommendations.
1. Specimen width generally should be large compared to yarn bundle
width. However, for carefully made specimens, narrower widths can give
reliable results.
2. For the three-point-loading flexural modulus test, the span-to-depth
ratio should be great enough to make the effect of shear deformation
negligible. The required aspect ratio depends on the ratio of extensional
modulus-to-shear modulus for the material. As a rule, L/D should be at
least 60/ 1.
3. To measure flexural strength, an aspect ratio of about 16/1 should be
used.
4. Because of material and load eccentricity, strain gages should be used
on both sides of a tensile coupon to eliminate effect of bending stresses.
5. The effect of surface resin layers on flexural modulus can be very
significant.
6. For laminates made from many uniform layers, tensile and flexural
moduli should be equal.
7. The effect of number of plies on tensile and flexural moduli requires
further study.
References
[/] Metcalfe, A. G. and Schmitz, G. K., Proceedings, American Society for Testing aijd
Materials, Vol. 64, 1964, p. 1075.
[2] Kies, J. A., "The Strength of Glass," NRL Report 5098 Naval Research Laboratory,
Washington, D.C., April 1958.
[3] Herring, H. W., "Selected Mechanical and Physical Properties of Boron Filaments,"
NASA TN D-3202, National Aeronautics and Space Administration, Langley, Va., 1966.
[4] Diefendorf, R. J. and Tokarsky, E., Polymer Engineering and Science, Vol. 15, No. 3,
1975, pp. 150-159.
[5] Zweben, C, Journal of Materials Science, Vol. 12, No. 7, 1977, pp. 1325-1337.
[6] Street, K. N. and Ferte, J. P., Proceedings, 1975 International Conference on Com-
posite Materials, Metallurgical Society of American Institute of Mining, Metallurgical,
and Petroleum Engineers, New York, 1976.
[7] Phoenix, S. L. and Sexsmith, R. G., Journal of Composite Materials, Vol. 6, 1972, pp.
322-337.
18] Daniels, H. E., Proceedings, The Royal Society, Vol. 183A, 1945, p. 405.
[9] Coleman, B. H., Journal of the Mechanics and Physics of Solids, Vol. 7, 1958, pp. 60-70.
[10] Phoenix, S. h., Fibre Science and Technology, Vol. 7, 1974, pp. 15-31.
[//] Gegauff, G., Bulletin, Societe Industrielle de Mulhouse, Vol. 77, 1907, pp. 153-213.
[12] Gucer, D. E. and Gurland, J., Journal of the Mechanics and Physics of Solids, Vol.
10, 1962, pp. 365-373.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.
262 COMPOSITE MATERIALS {FIFTH CONFERENCE)
[13] Rosen, B. W., "Mechanics of Composite Strengthening," in Fiber Composite Materials,
American Society for Metals, Metals Park, Ohio, 1965.
[14] Kelly, A. and Davies, G. J., Metallurgical Reviews. Vol. 10, 1965, pp. 1-77.
[15] Hearle, J. W. S., Grosberg, P., and Backer, S., Structural Mechanics of Fibers, Yams,
and Fabrics, Vol. 1, Wiley-Interscience, New York, 1969.
[16] Zweben, C, Journal, American Institute of Aeronautics and Astronautics, Vol. 6, No. 12,
Dec. 1968, pp. 2325-2331.
[17] Zweben, C. and Rosen, B. W., Journal of the Mechanics and Physics of Solids, Vol. 18,
1970, pp. 189-206.
[18] Zweben, C. in Composite Materials: Testing and Design, ASTM STP 460, American
Society for Testing and Materials, 1970, pp. 528-539.
[19] Zweben, C, Engineering Fracture Mechanics, Vol. 4, 1972, pp. 1-8.
[20] Rosen, B. W. and Zweben, C, "Tensile Failure Criteria for Fiber Composite Materials,"
NASA CR-2057, National Aeronautics and Space Administration, Washington, D.C.,
Aug. 1972.
[21] Scop, P. M. and Argon, A. S., Journal of Composite Materials, Vol. 1, 1967, pp. 92-99.
[22] Herring, Harvey, "Fundamental Mechanisms of Tensile Fracture in Aluminum Sheet
Unidirectionally Reinforced with Boron Filaments," NASA TR R-383, National Aero-
nautics and Space Administration, Washington, D.C., 1972.
[23] Zweben, C, Engineering Fracture Mechanics, Vol. 6, 1974, pp. 1-10.
[24] Hedgepeth, J. M. and Van Dyke, P., Journal of Composite Materials, Vol. 1, 1967, pp.
294-309.
[25] Friedman, E., "A Tensile Failure Mechanism for Whisker Reinforced Plastics Division,"
Society of the Plastics Industry, Washington, D.C., Feb. 1967.
[26] Berg, C. A., Tirosh, J. and Israeli, M. in Composite Materials: Testing and Design,
(Second Conference) ASTM STP 497, American Society for Testing and Materials,
1972, pp. 206-218.
[27] Bullock, R. E., Journal of Composite Materials, Vol. 8, 1974, pp. 200-206.
[28] MuUin, J. V. and Knoell, A. C, Materials Research and Standards. Vol. 10, 1970, pp.
16-33.
[29] Mallick, P. K. and Broutman, L. I., "Impact Behavior of Hybrid Composites," 30th
Anniversary Technical Conference, Reinforced Plastics/Composites Institute, Society of
the Plastics Industry, Inc., 1975.
[30] Handbook of Engineering Mechanics, W. Flugge, Ed., 1st ed., McGraw-Hill, New York,
1962.
[31] Stone, R. H., "Flight Service Evaluation of Kevlar 49/Epoxy Composite Panels in
Wide-Bodied Commercial Transport Aircraft," National Aeronautics and Space Admin-
istration, Report NASA CR-145141, NTIS Document No. N77-24199, March 1977.
[32] Zweben, C. and Norman, J. C, SAMPE Quarterly. Vol. 7, No. 4, July 1976, pp. 1-10.
[33] Scott, R. J., Fiberglass Boat Design and Construction. John de Graff, Inc., Tuckahoe,
New York, 1973.
[34] Whitney, I. M., "A Study of the Effects of Coupling Between Bending and Stretching
on the Mechanical Behavior of Layered Anisotropic Composite Materials," AFML-TR-
68-330, Air Force Materials Laboratory.

Copyright by ASTM Int'l (all rights reserved); Mon Feb 10 12:21:35 EST 2014
Downloaded/printed by
Universidad De Antioquia pursuant to License Agreement. No further reproductions authorized.

You might also like