Static and fatigue tests of a

woven glass fabric composite

under biaxial tension-torsion
loading
S. AMIJIMA*, T. FUJII* and M. HAMAGUCHI t

(* Doshisha University/t Dai-Nihon Ink Chemical Industries, Japan)

A preliminary study of the static and fatigue behaviour of a plain woven glass
fabric-reinforced laminate under uniaxial and biaxial loading is presented,
using newly developed thin-walled tubular specimens. Problems associated
with fabrication of the tubular specimen were investigated. Both the Tsai-Wu
and the Tsai-Hill strength criteria seem to be applicable to the static strength
data of the glass fabric composite under combined tension/compression-
torsion loading. Some biaxial fatigue test results are also given.

Key words: composite materials; static testing; fatigue testing; biaxial

loading; tubular specimens; plain woven glass cloth; polyester matrix

The fatigue failure process of fibre-reinforced compos- posites 3"4. A biaxiai stress state combining axial and
ites must be more complicated under biaxial loading shear stresses can be attained when the tubular
than under uniaxial loading. Fibre composites inher- specimen is simultaneously twisted along its longitudi-
ently have a heterogeneous microstructure. It can be nal axis while it is subject to axial loading 5.
inferred that fatigue degradation would be accelerated Today, fabrics such as plain woven glass cloth are also
due to the stress coupling effect on material damage widely used as reinforcement for composite materials in
when biaxial stresses, for instance axial and shear addition to unidirectional fibres. Thus knowledge of
stresses along fibres, are applied to composites. There- the fatigue properties of fibre composites using such
fore, material evaluation for fibre composites under reinforcement is required. Filament winding is a typical
biaxial stress states is required not only for static method used to fabricate thin-walled tubular speci-
loading but also for cyclic loading in many applications. mens. Alternatively, tubular specimens may also be
Various test methods have been employed to obtain fabricated by winding unidirectional prepreg tape on a
directly both static and fatigue properties of composite mandrel 6. A mould is used to control the dimensions of
materials under biaxial loading. The uniaxial test using the specimen, or the mould can be replaced by heat-
off-axis specimens is one of the simplest and cheapest shrinkage film. No fabrics were used in such cases.
methods although only a limited range of axial and
shear stress combinations can be used. The method Owen and Griffiths 7 successfully developed the wet
using fiat cruciform specimens in conjunction with a winding technique for resin-impregnated glass fabric to
special testing machine is a more advanced technique. fabricate thin-walled tubular specimens in order to
The use of flat plates enables specimens to be subjected conduct static and fatigue tests. Discontinuity at both
to arbitrary combinations of three stress components, ends of the fabric must exist when fabric is used as
but the disadvantage of this technique is that stress reinforcement for tubular specimens. There are other
concentrations at the corners of the specimens are problems accompanied with fabrication; for example,
unavoidable. Therefore, the test method employing how to eliminate voids and how to control the glass
thin tubular specimens has been widely used, where an content. The effects of these problems on fatigue test
internal pressure and/or a torsional moment along the results have not been revealed. In addition, the
longitudinal axis of the specimen is simultaneously problem of how to grip the specimen in the testing
applied in addition to the axial loading ~'2. However, a machine has still to be resolved.
wide range of stress combinations cannot be covered if
only the internal pressure is superimposed on the axial This paper deals with problems associated with the
loading. Among multi-stress states, the existence of fabrication of tubular specimens of a plain woven glass
shear stress parallel to the fibres is important since the fabric-reinforced composite, which are used for static
shear stress strongly affects the strength of fibre corn- and fatigue tests. Some results are reported for static
0010-4361/91/040281-09 (~) 1991 Butterworth-Heinemann Ltd
COMPOSITES. VOLUME 22. NUMBER 4. JULY 1991 281
and fatigue tests conducted on the tubular specimens was rotated at 5 rpm for 24 h at room temperature to
under biaxial tension/compression and torsion loading. prevent resin dropping. After solidification, the tube,
about 4 mm thick, was pulled off the mandrel and then
cured at 353 K for 4 h in an oven.
EXPERIMENTAL DETAILS
Fabrication Biaxial testing
Table 1 gives details of the plain woven glass cloth Tubular specimens fabricated as described above were
reinforcement and the unsaturated polyester resin mechanically ground to the configuration shown in Fig.
matrix used in the present study. Tubular specimens 2. Initially, relatively short specimens (type A, L --
were fabricated using the wet winding machine shown 230 mm) having a gauge length of 100 mm were used,
in Fig. 1, according to the following procedure. but this was later changed so that the specimen length
Fluorocarbon release agent was sprayed onto the was L -- 300 mm (type B) with the gauge length kept
chromium-plated mandrel (50 mm in diameter), and a the same.
layer of resin applied to the mandrel. Glass cloth, The specimens used by Krempi and Niu had tapered
350 mm wide, was then placed in contact with the tabs in both ends of the tube 5. Another specimen
mandrel and wound onto it by slow rotation of the requires metal flanges to be bonded to both ends of the
mandrel as resin was applied with a brush. Initial tube for fixing the specimen to the testing machine 8.
tension was given to the warp of the cloth by a dead These precautions, necessary for avoiding stress
weight. Care was taken to prevent the introduction of concentrations and to support both ends of the
voids between the glass cloth layers because it was specimen, and fixing the specimen to the machine rods
impossible to remove them later. After 12 layers of are time-consuming processes. Therefore, a chuck was
glass cloth were wound onto the mandrel, the mandrel developed for clamping the specimens in this study.
The tubular specimen was held tightly by tapered
Table 1. Composition of specimen collets as bolts were turned (see Fig. 3).
A servo-controlled fatigue testing machine
Matrix Polyester resin m Polylite
(SHIMAZU EHF-EB 10/TQ 1, of dynamic loading
FG-284 from Dai-Nihon Ink
capacities + 10 tonf (axial force) and + 100 kgf.m
Chemical Industries
(twisting moment) ) was used (see Fig. 4). This
Reinforcement Plain woven glass cloth - -
machine, which can apply axial and torsion loads
MG-252 from Asahi Glass
simultaneously, was operated by a mini-computer using
Works
a programme developed by the authors. A computer-
Catalyst Methyl ethyl ketone peroxide
controlled data acquisition system was also employed.
Accelerator Cobalt naphthenate 0.6%
solution Preliminary experiments were conducted under several
conditions in order to confirm the test procedure
Brush

l/i
Plain woven glass cloth

o__/MandreGea
Belt

, ~r_ . . . . . . . ~ Moor /

Fig. 1 Wet winding machine for tubular specimen

282 COMPOSITES. JULY 1991

 L = 230 (A type)
jt L = 300 (B type)

L
Fig. 2 Dimensions of tubular specimen (in mm)

Fig. 4 Computer-controlled servo-hydraulictension-torsion

biaxial test system

RESULTS AND DISCUSS~ON

Effect of specimen length on observed tensile
strength and improvement of the chuck
~sureplate The effect of specimen length on observed tensile
strength was examined in order to determine the
appropriate dimensions of the specimen for later tests.
Firstly, type A specimens, having overall length (L) of
230 mm and chuck length (l) of 65 mm, were used. No
initial tension (which will be discussed later) was
applied in fabrication. Fracture occurred at the
boundary between the gauge section and the gripped
portion, indicated by ~ in Fig. 3, for all (five) speci-
mens. It was inferred that such unexpected fracture was
Fig. 3 Assembly drawing of clamping fixture 1(~ denotes the due to a non-uniform stress distribution at the vicinity
fracture position for the shorter specimen, L =230 ram) of point @. Finite element calculations using three-
appropriate under biaxial loading. Later, general static dimensional isoparametric elements confirmed that the
and fatigue tests were conducted. Additional detailed axial stress concentration at the boundary between the
gauge section and the gripped portions is not significant
information on experimental procedures will be given
in the 'Results and discussion' section for convenience. as long as the tensile load is properly applied at the end
Static failure strength under monotonically increasing of the tube.
biaxial stresses were observed at several stress ratios 0~ It was inferred that the observed boundary fracture of
(= axial stress/shear stress). For fatigue tests, number the specimen was due to insufficient number of co|lets
of cycles to failure for axial tensile, pure shear and and the short gripped length of the specimen. The first
combined (biaxial) stress states were determined. In chuck designed had four collets. Thus the number of
both static and fatigue tests, proportional loading was collets was increased from four to six and the specimen
applied to the specimen. For fatigue tests, a sinusoidal was also modified from type A to type B by increasing
stress wave was used. All tests were carried out at room the chuck length (l) to 100 mm. Specimens then failed
temperature. within the gauge section, away from the boundaries.

COMPOSITES. JULY 1991 283

 Type B specimens were used thereafter for all other
tests.

Effect of initial tension applied to glass cloth on

observed tensile strength
Static tensile strength and Young's modulus were
226.2 MPa and 19.0 GPa respectively for specimens
which were fabricated by applying an initial tension of
100 N to the glass cloth in order to regulate alignment
of fibres in the longitudinal direction. The mean glass
volume content, Vf, was 38.3%. The value of Young's
modulus is reasonable, but the strength value is about
20% lower than that obtained from a dumb-bell type
specimen (Vf = 37.9%) fabricated by hand lay-up. The
reason for this discrepancy is considered to be as
follows.
The warp fibres in the glass cloth were stretched in the
circular direction of the tubular specimens by the initial
tension while the weft fibres in the axial direction
became not only highly crimped but contacted one
another, possibly causing damage. This strong waviness
in the weft fibres leads to a reduction in the axial
stresses supported by the tube and results in a decrease
in the observed strength of the tubular specimen.
Fig. 5(a) presents a photo-micrograph which shows the
cross-section parallel to the longitudinal direction of
the tube fabricated with initial tension. Fibre bundles of
the warp are not flat and maintain their round shapes.
Fig. 5(b) shows a similar cross-section of another
specimen fabricated without initial tension, where a
flattening of the fibre bundles can be seen.
The specimen fabricated without initial tension showed Fig. 6 Shearbuckling under static pure torsion
a tensile strength of 233.4 MPa and Young's modulus
of 17.1 GPa for Vf = 32.9%. This strength buckling apparently took place. Under this condition
(233.4 MPa) is higher than that for the specimen the pure shear strength is not obtained. In order to
fabricated with initial tension despite the lower glass prevent this shear buckling, a steel bar having a slightly
content. These observed values coincide with ones smaller (0.05 mm) diameter than the inside diameter of
obtained by testing a dumb-bell type specimen having the tube was inserted inside the specimen. The surface
almost the same glass content. Thus specimens for the of the bar was lubricated with a Fluorocarbon lubricant
following tests were fabricated with no initial tension. to reduce friction. Using the steel bar the shear
buckling, distinguishable by the naked eye, did not
Pure torsion and uniaxial compression tests occur during the torsion test. On the other hand, this
bar was not necessary under biaxial loading (tension-
Fig. 6 shows the final failure mode of the tubular torsion) since the shear buckling mode was probably
specimen under static torsion, revealing that shear avoided by tensile stress in the longitudinal direction.
A new anti-buckling fixture (see Fig. 7) was devised for
compression testing which seemed to be effective for
the tubular specimens used here. The essential function
of this fixture is similar to that used for compression
testing of thin composite plates (ASTM D695). The
outer and inner fixtures, intended to prevent buckling
of the thin tubular specimen, are machine-grooved to
minimize friction between the specimen walls and the
fixture. The observed compression strength, a reason-
able value for the glass cloth composite studied, was
close to the static tensile strength. No indication of
macroscopic compression buckling was found from the
appearance of the failed specimen.

Fig. 5 Magnified cross-sections of tubular specimens parallel Static strength under biaxial loading
to the longitudinal axis: (a) fabricated with initial tension; (b)
fabricated without initial tension Static and subsequent fatigue tests were conducted

284 COMPOSITES. JULY 1991

 Detail of B
Detail of A

54
,i

a,

I
t

114
Q1 84

ilil
4x~13
;kll
rll

,llll 60 10~
I 100
:1111
I : ill i

ii
/72 "

a b
Fig. 7 Anti-buckling fixture for compression testing of tubular specimens: (a) inner piece; (b) outer piece. Dimensions in m m
using specimens having lower nominal glass contents fully in an electric furnace. Fig. 9 shows glass cloth
(Vf about 27_+3%) than former ones. Fig. 8 shows the obtained from the gauge section after burning the
strength under uniaxial and biaxial loading. In this specimen. It is found that some pieces of glass cloth
figure the symbols denote the experimental results were removed irregularly near the end. Therefore, the
while the dotted lines denote the loading path (straight glass content in the gauge section is not expected to be
lines due to proportional loading). It can be seen from uniform. In order to check this, the gauge length of a
Fig. 8 that there is considerable scatter in the strength specimen was cut into pieces and their glass contents
values observed. measured. Large scatter (about 24 to 30%) was
Fabrication of the test specimen requires grinding the observed, with low glass content corresponding to the
outer surface of the specimen, resulting in damage to damage zone. Usually the mean glass content is used as
the glass fibres in the outer lamina. To examine the the nominal glass content and the data are mutually
degree of fibre damage, a specimen was burned care- compared. It is proposed that the large scatter of glass

COMPOSITES. JULY 1991 285

 120

100

_Tsai-Wu theo~
~ 50
ii//
0 Tsai-Hill th7
O9
¢.-
//i /
/t.
Transverse surface
I I I I I I
-300 -200 -100 0 100 200 300
Normal stress,o 1(MPa)
Fig. 8 Ultimate static strength under uniaxial and biaxial J
loading (open circles denote experimental results) a 1

•~ 6

'ongitodina,su,ace I

Fig. 9 Glass cloth of a tubular specimen damaged by grinding

content and partial damage in the glass fibres are
responsible for the strength scatter of the tubular
specimen.
The static strengths of the tubular specimen can be
compared with the predictions of both the Tsai-Wu
and the Tsai-Hill failure criteria. The Tsai-Wu failure
criterion is given by9:
Y
Fig. 10 Definition of stress components for tubular specimens

G + H = 1/X2t = 36.9 x 10-.6 1/(MPa) 2

Fio i + Fijo2= 1 (i=1,2 ..... 6) (1)
N = 1/S 2 = 195.6 x 10-6 1/(MPa) 2 (5)
where Fi and F~j are fracture tensors of the second and
fourth rank, respectively. The contracted stress The solid lines in Fig. 8 show the failure surfaces
notation is used except for the shear stresses 04=023, calculated by both criteria. Relatively good agreement
05=~31, 06=oiz. Applying this equation to the tubular between the calculated and the experimental results is
specimen yields: found.
FIo I + FllOl2 + F66062= 1 (2) Since the difference between the compressive and the
tensile strengths is small, the two theories give almost
where 01 and o6 represent the normal stress in the the same results except near the uniaxial compressive
longitudinal direction and the shear stress, respectively strength. Thus the static strength of the glass cloth
(see Fig. 10). F1,
F 11 and F66 have the following values composite subjected to biaxial stress can be estimated
for the glass cloth composite, where the ultimate tensile using both the Tsai-Wu and the Tsai-Hill criteria.
strength Xt = 164.4 MPa, the ultimate compressive
strength Xc= 160.1 MPa, and the ultimate shear
strength S=71.5 MPa: Effect of surface finishing of tubular specimen on
fatigue life
F1 = l / X t - 1/Xc = -163.3 x 10-6 1/MPa
It is well known that different surface finishing tech-
FH = 1 / ( X t X c ) =-37.9 × 10-6 1/(MPa) 2 (3) niques greatly affect the fatigue life of metal and
F66 = 1/S 2 = 195.6 x 10-6 1/(MPa) 2 polymer test specimens. As above mentioned, the
tubular specimens have a relatively large amount of
The Tsai-Hill failure criterion is given by: glass fibre damage and a rough surface due to machin-
(G + H)o~l + No~6= 1 (4) ing even at the virgin stage. Wang et al. reported that
surface finishing by grinding accelerated stiffness
where G, H and N are given by degradation under cyclic loading, but had no effect on

286 COMPOSITES. JULY 1991

 309 --

Failure
R

305 -

t~

170 E
I1)
}-- 301
Fig. 11 D i m e n s i o n s of d u m b - b e l l t y p e s p e c i m e n (in m m )

180
150 297 I I I I
•
0 1500 3000 4500 5100
ro 100 ~- Time (s)
Q. • Fig. 13 Temperature increase inside the tubular specimen
during zero-torsion fatigue test at a speed of 10 Hz

u~ 50 any greater fatigue damage at the lap regions of the

E glass cloth compared with that at other locations in the
E specimen.
x
Consequently it may be concluded that there is no fatal
difference between the fatigue behaviours of the
tubular and the dumb-bell type specimens in our
10 I I I I
102 103 104 105 106 07
laboratory. It is adequate to use the tubular specimen
developed in the present work to study the fatigue
Number of cycles to f a i l u r e ( N f ) characteristics of a woven glass fabric composite.
Fig. 12 Fatigue lives (S/Ndiagram) for tubular and dumb-bell However, the tubular specimen sometimes became
type specimens under uniaxial zero-tension loading slightly twisted during fatigue testing even when only
uniaxial (but high) load was applied. This distortion
fatigue life m. It has not been clear whether surface tends to decrease the fatigue life. It is believed that this
finishing affects the fatigue life of our tubular derives from the grinding damage, and that damage of
specimen. Thus, in order to examine this effect, fatigue the outermost fibres has to be kept minimum and
test results obtained from the tubular specimen were uniform. Misalignment of longitudinal fibres against
compared with those obtained using the dumb-bell type the loading axis also could be a reason for this distor-
specimen shown in Fig. 11. All fatigue tests were tion.
carried out in fluctuating tension (zero-tension condi-
tion; i.e., stress ratio R=0) at a speed of 10 Hz. In Temperature increase of specimen during torsion
order to achieve comparable surface finishing to that of fatigue
the tubular specimen, one side surface of the dumb-bell
specimen was also plain ground using a grinding stone As a preliminary test, a zero-torsion fatigue test was
of the same grain size as used for the tubular specimen. conducted at 10 Hz cyclic speed. A temperature
increase was experienced inside the specimen as shown
Fig. 12 shows the S/N curves obtained from both in Fig. 13. At specimen failure the temperature was
specimens, the lines being linearly fitted to the data by 33.9°C and the number of cycles to failure, Nf, was
means of the least-squares method. It can be seen that, 45 990. This temperature increase can be explained by
in addition to the static data, the fatigue data for the shear stress/shear strain hysteresis loss, which generates
tubular specimen also exhibit scatter. heat. In order to avoid this unexpected temperature
During specimen fabrication two ends of the glass cloth increase of the specimen, the cyclic loading speed was
strip are embedded in the matrix. This means that the limited to below 2 Hz. The air inside the tubular
tubular specimen has a discontinuity of glass fibres in specimen always has to be exhausted.
the laminae• The discontinuity may become a stress
concentration source, leading to easy debonding Some typical fatigue test results under biaxial
between laminae and reducing the fatigue life. tension-torsion loading
However, little difference was found between the S/N Fig• 14 presents the fatigue test results obtained under
diagrams of tubular and dumb-bell specimens. Owen uniaxial and biaxial (tension-torsion) loading. Every
and Griffiths 7 conducted zero-tension fatigue tests for cyclic load was pulsating (the stress ratio, R, was
both plain woven glass fabric composite specimens and approximately zero). For the biaxial fatigue test,
lap joint specimens• They concluded that, below 106 in-phase cyclic tensile and torsion loads were applied to
cycles, the lap had an effect on the fatigue strength but the specimen. In studying the multi-axial fatigue
above 106 cycles there was no effect• In our tests, behaviour of isotropic metals it is convenient to repre-
discontinuity of the reinforcement had no apparent sent the complex stress state under cyclic loading by an
effect on the fatigue life. Microscopic examination of •
equivalent or effective •
cychc stress I I . However, the
the tubular specimen after fatigue testing did not reveal proper values have not been established for represent-

C O M P O S I T E S . J U L Y 1991 287
 1.0 -- If the shear or axial stress has a linear effect on the
o Biaxial(ix=4:9) fatigue failure process under biaxial stress, then the
0.8 • Biaxial (¢t=1:1)
[] Tension higher degradation in the case of 0¢= 1:1 (in which the
.9 0.6 • Torsion specimen is subjected to a higher shear stress compo-
nent) would be expected. Therefore, the failure surface
09 in the multi-axial stress state at a given cycle is not
09
0.4- proportional to the static one. It is apparent that the
fatigue life of woven glass fabric composites is affected
0 by many factors.

I I I I I CONCLUSIONS
0.2
102 10 a 10 4 10 s 10 6 10r Problems associated with fabrication of tubular speci-
Number of cycles to failure, Nf mens of a plain woven glass cloth composite for biaxial
Fig. 14 N o r m a l i z e d f at i g u e lives u n d e r u n i a x i a l and biaxial static and fatigue tests have been discussed. Problems
loading with static and fatigue tests have also been examined,
and limited fatigue test results under biaxial loading
t~6 given. The following conclusions may be drawn from
this study.

Tsai-Wu crite~lien ~ Staticstrength 1) For static and fatigue tests under biaxial loading, a
new type of chuck having six tapered collets was
developed. A n appropriate tube configuration and
dimensions for biaxial loading tests on a glass fabric
composite were determined.
2) Initial tension applied to the glass cloth during
I/ / specimen fabrication reduces the axial tensile
strength of tubular specimens.
01 .max G1 3) A fixture designed to prevent compression buckling
of the tubular specimen was found to be effective
and useful for compression testing. A steel bar
inserter should be used to prevent shear buckling
when the tubular specimen is subjected to torsion
Fig. 15 Definition of stress ratio (stress state under biaxial moments.
loading)
4) Both Tsai-Wu and Tsai-Hill failure criteria can be
applied to predict the static strength of a plain
ing the multi-axial stress state in anisotropic fibre
woven glass cloth composite under biaxial loading.
composites, although Wang et al. t2 used the equivalent
Scatter of strength data is due to non-uniform and
stress for evaluating fatigue strength of their carbon
local damage produced by surface finishing of the
fibre-reinforced plastic under biaxial cyclic loading.
tube.
Here, every fatigue strength was normalized by the
mean static strength and plotted against the log-scaled
5) Little difference was found in the fatigue life
diagrams under .~ero-tension cyclic loading of
number of cycles to failure, i.e., a stress ratio was
tubular and dumb-bell type specimens. There is no
introduced for representing the normalized fatigue life
apparent effect of discontinuity and lap of the
under biaxial loading. The y-axis in Fig. 14 is the stress
reinforcement on the fatigue life of tubular speci-
ratio, defined as the ratio of the distances O B : O A on
mens.
the ot - o6 stress map (see Fig. 15). Point A on Fig. 15
represents the static strength (stress combination at
6) A temperature increase was experienced inside
tubular specimens at high cyclic speed under fatigue
failure); point B represents the maximum stress state
torsion loading. This temperature increase restricts
under cyclic biaxial loading.
the cyclic loading speed to below 2 Hz.
Although fatigue strength under uniaxial torsion 7) Fatigue life strongly depends on the biaxial stress
loading is always lower than that under uniaxial tensile ratio, and fatigue life degradation does not follow a
loading, fatigue life degradation under torsion loading simple relation with the biaxial stress ratio.
is less than that under tensile loading (here, the fatigue
life degradation is indicated by the slope of the S / N
REFERENCES
line). Two life degradations can be linearly extended to
the static strengths on Fig. 14. On the other hand, life 1 Owen,M.J. and Rice, D.J. 'Biaxialstrength behaviourof glass
degradations under biaxial loading cannot be extended. fabric-reinforcedpolyesterresins' Composites 12 No 1 (1981)
It is predicted that below 103 cycles rapid life degrada- p13
2 Francis, P.H., Walrath, D.E. and Weed, D.N. 'First ply failureof
tion occurs under biaxial stress. An interesting phe- G/E laminatesunder biaxial Ioadings'Fibre Sci and Techno112
nomenon found for the biaxiai fatigue lives is that the (1979) p 97
life degradation is lower in the case where biaxial stress 3 Puck,A. and Sehurm~mn,H. 'Die Zug/Druck-Torsionsprufung
ratio ix=4:9 (max. shear stress:max, normal stress) than an rohrformigenProbekorpern' Kunststoffe 72 (1982) p 9
4 Smith,E.W. and PsLseoe,KJ. "Therole of shear deformationin
that for 0c= 1:1. The cyclic shear stress component for the fatiguefailureof a glass fibre-reinforcedcomposite'Compos-
the former case (0t=4:9) is less than that for the latter. ites 8 No 4 (1977) p 237

288 COMPOSITES. JULY 1991

5 Krempl, E. and Niu, T.-M. 'Graphite/epoxy [+45]s tubes. Their 11 Kikukawa, M. et al. 'A comparison of the axial and reversed-
static axial and shear properties and their fatigue behavior under torsional strain cycling low-cyclefatigue strength of several
completely reversed load controlled loading' J Composite Mater structural materials' Bull JSME 15 No 86 (1972) p 889
16 (1982) p 172 12 Wang, S.S., China, E.S.-M. and Soeie, D.F. 'Biaxial fatigue of
6 Weed,D.N. and Francis, P.H. 'Process development for the fiber-reinforced composites at cryogenic temperatures. Part I:
fabrication of high-quality composite tubes' Fibre Sci and fatigue fracture life and mechanisms' A S M E J Engng Mater and
Techno110 (1977) p 90 Techno1104 (1982) p 129
7 Owen, M.J. and Griffiths, J.R. 'Evaluation of biaxial stress
failure surfaces for a glass fabric reinforced polyester resin under
static and fatigue loading' J Mater Sci 13 (1978) p 1521
8 Swanson, S.R. and Nelson, M. 'Failure properties of carbon/ A U THORS
epoxy laminates under tension-compression biaxial stress' Proc
3rd Japan-US Conf on Composite Materials (1986) T. F u j i i (to w h o m c o r r e s p o n d e n c e should be
9 Tsai, S.W. and Wu, E.M. "A general theory of strength for a d d r e s s e d ) a n d S. A m i j i m a are with the D e p a r t m e n t of
anisotropic materials' J Composite Mater I (1971) p 58 Mechanical Engineering, Doshisha University, Kyoto,
10 Wang, S.S., Chim, E.S.-M. and Socie, D.F. 'Tensile and tor-
sional fatigue of fiber-reinforced composites at cryogenic tem- J a p a n . M. H a m a g u c h i is with D a i - N i h o n I n k C h e m i c a l
peratures' A S M E J Engng Mater and Techno1104 (1982) p 121 Industries, Amagasaki, Osaka, Japan.

COMPOSITES, JULY 1991 289