Composites Part A 29A (1998) 869–886

1359-835X/98/$ - see front matter

q 1998 Published by Elsevier Science Ltd.
PII: S1359-835X(97)00090-0 All rights reserved.

A review of the biaxial strength of fibre-

reinforced plastics

Holger Thom
Bundesanstalt für Materialforschung und -prüfung, 12200 Berlin, Germany
(Received 10 February 1997; revised 5 August 1997; accepted 26 August 1997)

Several strength criteria for laminae and laminates are presented and compared numerically. It is concluded that
the von Mises hypothesis is not suitable for composites, instead strength criteria should be derived from Mohr’s
hypothesis. In industry mostly laminates are used. It would be desirable if the strength properties could be
determined for any arbitrary laminate from those of the laminae included. Since this is spoiled by interaction
between the layers, some of these interaction effects are described. In order to become established, a strength
criterion has to be verified experimentally. Thus, different techniques for biaxial testing with their characteristic
problems are shown. q 1998 Published by Elsevier Science Ltd. All rights reserved.

(Keywords: biaxial strength; D. mechanical testing; strength criteria)

Nomenclature INTRODUCTION
a,A scalar
a,A vector, tensor Initially, research in the field of anisotropic materials was
d diameter carried out on materials such as timber or crystals. In the
E elastic or Young’s modulus mid 1960s, research started to focus on fibre-reinforced
« strain
e ultimate strain plastics (FRPs). In contrast to isotropic materials, no failure
F factor of a strength tensor criterion has been established so far for these materials. This
G shear modulus lack of a criterion is due to the following additional
n Poisson’s ratio
P pressure problems.
J fibre volume fraction
S ultimate stress (absolute value) • The criterion cannot be based on the principal stresses due
j normal stress to anisotropy (like the von Mises hypothesis), but has to
t wall thickness be formulated with respect to the axes of anisotropy.
t shear stress
• Fibres ending at free edges of laminates induce inter-
Subscripts laminar stresses 1.
• FRPs are more sensitive to local stress concentrations,
1 fibre longitudinal direction (also principal axis of orthotropy)
2,3 fibre transverse directions therefore specimens have to be designed more accurately 2.
n,t normal, transversal to plane • Owing to the high modulus ratio in the two axes of ortho-
x,y,z arbitrary directions tropy, the mode of failure has to be checked (buckling) 2.
Superscripts • The transformation of the strength properties of different
plies to the properties of a laminate is not at all obvious;
þ property for tension Gerharz and Schütz3 even said that this was impossible.
¹ property for compression
• The term ‘failure’ is a question of definition, as with
ductile materials. The first change in behaviour is due
Calculations are done with the following material data for carbon/ commonly not to yielding but to initial fibre fractures or
epoxy-HT: microcracks in the resin of the off-axis plies.
S1þ ¼ 2150 MPa, S1¹ ¼ 1500 MPa, The lack of a reliable failure criterion is disadvantageous,
S2þ ¼ 35 MPa, S2¹ ¼ 200 MPa,
S12 ¼ 80 MPa. especially for composite materials. FRPs are mostly used in
lightweight constructions. If the point of failure is only
vaguely known, high safety factors may cause unnecessary
weight. Nevertheless industry uses very simple and
imprecise failure criteria like the maximum stress criterion.

869
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

This is due to today’s methods of production which still

cause high scattering, and because fatigue or impact are
usually the design restrictions of the component. However,
with the goal of composite design being theoretically
substantiated and not only experimentally proven, a reliable
criterion describing static strength will be one of the basic
constituents.
Reading the literature, one observes that initially criteria
were formulated theoretically but not verified experimen-
tally. Only later, when testing methods were developed and
materials were cheaper, can trustworthy results be found in
the literature. Still, there are a lot of unanswered questions,
especially since most researchers concentrated on laminates
without searching for the effects leading to the difference
between laminae and laminates.
A good failure criterion should take the following rules
into consideration;
• it should describe as many different types of materials and
loadings as possible,
• a minimum of experimental work should be necessary to Figure 1 Compressive strength of GFRP
determine the criterion’s coefficients.
The establishment of a failure criterion involves three failure results from microbuckling or kinking of the fibres
steps: the formulation of the criterion, the experimental within the resin. On SEM pictures, tensile, compressive and
verification, and the convincing of future users. This survey neutral zones on the fibres can often be discovered.
presents and compares some of the common criteria. Microbuckling is more likely to initiate at free edges or
Afterwards, different geometries of specimen and their close to inner voids, where lateral support is reduced. It can
problems are shown. The successful path from the be distinguished from kinking bands by the multiple
scientist’s ivory tower to the design department cannot be fractures of the fibres. Dow and Rosen6 tried to find the
found in this study. First we present a survey, although very actual mode of compressive failure. They distinguished an
brief, concerning the different modes of fracture and effects extension and a shear mode (Figure 1). In the extension
not known from isotropic monoliths. mode, the adjacent fibres will buckle with 1808 phase
shift, thus extending the matrix mainly in a direction
perpendicular to the fibres. Shear strains are the main
FAILURE MECHANISMS deformation in the shear mode, where the fibres buckle in
phase. The result of these analytical models is that a
Tensile fibre failure lamina of commonly used fibre volume fraction will fail in
the shear mode. The affiliated compressive strength is given
Typical for tensile fibre failure is pull-out on the failure by
surface. This also happens under very cold and dry, i.e.
brittle, conditions 4. A rise in temperature increases the pull- Gm
S1¹ ¼ (1)
out length. These observations show that local fibre 1¹J
fractures occur earlier than the final fracture, owing to Since the linear model caused overestimation of the com-
misalignment and scatter during the manufacture of the pressive strength at high fibre volume fractions, they also
fibres. The stresses of broken fibres are transferred to intact considered non-linear elastic behaviour. The compressive
fibres by the resin. A further increase in the stress results in strength of the non-linear modeal was still higher than the
debonding of the fibre due to stress peaks at its ends. While experimentally determined value. Wisnom7 looked at the
Miller and Wingert4 state that ‘‘specimen fracture is the influence of fibre misalignment on the compressive strength.
cumulative total of a series of nearly independent micro- Only 0.258 misalignment reduced the predicted strength
structural fracture events’’, Purslow 5 reported that he was from 2720 MPa to 1850 MPa for a carbon/epoxy lamina.
able to determine the direction of crack propagation on At 38 misalignment, a strength of only 700 MPa was left.
small surfaces. Since the crack process on a large area is A periodic deflection of the fibres was examined by Hahn
chaotic and propagates on several surfaces, the origin of the and Williams8. They found the compressive strength to be
crack cannot be ascertained. g12
S1¹ ¼ JG12 (2)
g12 þ pfh =l
Compressive fibre failure
G12 is the secant composite shear modulus, f0 the amplitude
This mode of failure is to a lesser degree determined not of the fibre deflection, and l is half of its wavelength. This
by the strength of the fibres but rather by their stability. The is in contrast to Rosen’s theory, since the equation

870
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

above becomes
S1¹ ¼ JG12 (3)
if no initial deflection is assumed. The difference is due to
the selection of the free body. In the model of Hahn and
Williams the forces are applied only to the fibre.
No matter which theory is used, the important fact is that
the compressive strength of the composite is somehow
proportional to the shear modulus of the matrix. This has
also been verified experimentally. In practice, the theo-
retical compressive strength can never be reached owing to
fibre misalignment and fibre curvature.
Figure 2 Delamination due to microcracks in off-axis plies

Matrix failure
interface of two plies this stress is much higher for the 6308
The mechanism of matrix failure is the most complicated.
specimen than for the 6 458 specimen.
It is presented here only very briefly. Matrix failure in a ply
The free-edge effect is also responsible for the difference
is called ‘inter-fibre’ or ‘intralaminar’ fracture, the separa-
in strength of differently stacked, but otherwise similar,
tion of two plies ‘delamination’ or ‘interlaminar’ fracture.
laminates. Schulte and Stinchcomb16 reported that the
The fracture mechanics differs according to the fracture
strength of a (08/908) laminate depended on the orientation
mode of normal or shear stress (mode 1, 2). The
of the outer ply. Also Whitney and Browning17 observed on
morphologies of both modes can be distinguished clearly 9.
a ( 6 458,908) s laminate that only one stacking sequence led
It is believed that inter-fibre fracture starts at the interface
to delamination. In the case of delamination jz was greater
of fibre and resin10. The crack then propagates through the
than zero, in the other stacking sequence smaller than zero.
interface into the resin. This process produces river
Pagano and Pipes18 examined this correlation in detail and
markings and textured microflow in the resin when loaded
concluded that the difference in strength can be put down to
in mode 1, and stacked lamellae (also called hackle marks)
the interlaminar normal stress. A correlation with inter-
when loaded in mode 2.
laminar shear stresses is less significant.
Delamination is caused by interlaminar stresses (see free-
edge effect). Interlaminar stresses often result out of
microcracks in off-axis plies. Thus, delamination is not
Cook–Gordon effect19
dependent primarily on the homogenous stress state11.
A unidirectional lamina failure can be attributed to any of This effect also leads to differences in strength between
the reasons mentioned above, whereas a laminate fails unidirectional laminae and multidirectional laminates.
ultimately usually owing to fibre failure12 (an exception is Swanson20 observed that a lamina had a higher strength
delamination or oblique, compressive inter-fibre cracking13). than a laminate, independent of the free-edge effect. This
became obvious when a tough resin was applied. Whereas a
normal, brittle resin decreases the ultimate strain by 5%, the
Free-edge effect
tough resin reduced it by 30%. An explanation is delivered
A speciality of compounds is the free-edge effect. by the Cook–Gordon effect. In off-axis plies of laminates,
Because of this effect, geometry can influence the macro- microcracks occur long before failure of the longitudinal
scopic strength. fibres (Figure 2). These microcracks lead to stress peaks in
Laminate theory does not include the constraints of a the adjacent longitudinal fibres. Those stress peaks are
compound. According to this theory it follows that multi- reduced by a debonding of the fibres. In case of a strong
directional compounds would have normal and shear interface, like the interface of the tough resin, this
stresses on edges with ending fibres. Pipes and Pagano14 mechanism no longer works as well.
calculated the stresses near a free edge with finite
differences. They found a strong increase in the shear
Geometric influences on strength
stress txz and to a lesser degree in jz near the free edge. This
three-dimensional stress state reaches approximately as far The strength of compounds also depends on further
into the compound as it is thick. Pipes et al. showed15 that geometric parameters. There is a definite correlation
failure due to the free-edge effect also depends on the angle between the thickness and the strength of a compound and
of two adjacent plies. The strength of a 6308 specimen is of a laminate’s ply16,21. Increasing thickness decreases the
greatly reduced by the free-edge effect. The ultimate strain strength. O’Brian and Salpekar21 found with laminates
is equivalent to that of the 308 off-axis specimen. However, consisting of 6, 8, 16, 32, 64 plies, a maximum strength at a
the ultimate strain of a 6 458 specimen is higher than that of thickness of 8 plies. For thin 908 off-axis plies the strain for
the 458 off-axis specimen. The free-edge effect does not the first microcrack increases progressively with decreasing
seem to influence the 6 458 specimen. An explanation for thickness (see Cook–Gordon effect)22. The strain, before
this difference is the value of txz . Near the free edge and the the first microcracking occurs, does not depend on S 2 but on

871
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

Table 1 Survey of the presented strength criteria

Criterion Distinction between Interaction term Application to UD-lamina,
failure modes MD-laminate, not specified
Analytic Experimental
Maximum stress/strain þ ¹ ¹ UD
Tsai–Hill ¹ þ ¹ n.s.
Hoffman ¹ þ ¹ n.s.
Franklin–Marin ¹ ¹ þ n.s.
Tsai–Wu ¹ ¹ þ n.s.
Puppo–Evensen ¹ þ ¹ n.s.
Hashin–Rotem þ ¹ ¹ UD
Wu–Scheublein ¹ ¹ þ MD
Tennyson–MacDonald–Nanyaro ¹ ¹ þ n.s.
Chang Chang þ ¹ ¹ UD
Truncated maximum strain þ ¹ ¹ MD
Cuntze þ ¹ þ UD
Puck þ ¹ þ UD

the energy release rate. Under certain thicknesses, micro- Since the strength of anisotropic materials cannot be
cracks are even suppressed completely. The strain necessary described in a principal stress state, he added shear stresses.
for microcracking also depends on the location of the ply in It is formulated for stresses parallel to the axes of
a laminate. Microcracks in an outer ply occur at lower stress orthotropy. Interaction terms are not taken into considera-
levels than in inner plies. tion. The different modes of failure are distinguished. For
Some of these measured differences can also occur as a the plane stress state it is of the following form:
result of the manufacturing process of the specimens.
j1 ¼ S1þ (j1 . 0) j1 ¼ ¹ S1¹ (1 , 0) (4)
Clements and Lee23 increased the strength of 08 specimens
by 15%–25% after polishing the edges. They concluded that j2 ¼ S2þ (j2 . 0) j2 ¼ ¹ S2¹ (j2 , 0)
the strength was decreased by micro damage to fibres and by
the different ratio of surface to volume of previously cut jt12 j ¼ S12
fibres. 38
When testing two 6 458 off-axis specimens, one narrow Pettit and Waddoups supplemented the maximum strain
and the other wide, one will observe that the narrow criterion in order to consider non-linear elastic responses.
specimen will fail owing to matrix failure, whereas the wide
specimen fails owing to fibre failure. The wide specimen has Tsai–Hill (or modified Hill) (1965) criterion39
long continuous fibres, the narrow one has only short fibres
with much less overlap24. This can also change Young’s Hill derived the criterion from the von Mises hypothesis
modulus25. for anisotropic materials. Later this was altered by Tsai for
composites. The criterion has interaction terms, but it does
not distinguish between tensile and compressive properties.
STRENGTH CRITERIA Fan 40 suggested the use of the tensile, or compressive,
strength as a function of the actual stress sign.
Over the years, many strength criteria have been developed. j j j j t
( 1 )2 ¹ 1 2 2 þ ( 2 )2 þ ( 12 )2 ¼ 1 (5)
This section presents some of the most important ones and S1 S1 S2 S12
explains their formulations. The criteria presented are all
macroscopic. Input data consist of the strength of a single
ply or a whole laminate as well as the stresses. From the Hoffman (1967) criterion41
macroscopic criteria, microscopic criteria have to be
The Hoffman criterion is a supplement to the Tsai–Hill
delimited, which are mostly used in combination with
criterion. Linear terms were added to consider the difference
finite elements. They are based more on the physical
of tensile and compressive strength:
processes of crack propagation, mainly on the energy
release rate. Because of their need for immense CPU time 1 1 1 1 j21 j j
( ¹ )j þ ( ¹ )j þ ¹ 1 2
and lavish modelling, their application in industry for large S1þ S1¹ 1 S2þ S2¹ 2 S1þ S1¹ S1þ S1¹
models is doubtful in the future. Further surveys of strength
criteria can be found in references26–36. j22 t12 2
þ þ ¹ þ( ) ¼1 ð6Þ
To start with, Table 1 shows the presented criteria, their S2 S2 S12
structure, and their application.
Franklin–Marin (or modified Marin) (1968) criterion27
37
Maximum stress/strain criterion (1920)
Franklin modified the criterion of Marin proposed in
This is the most simple criterion. Jenkins derived it from 1956. Marin had added terms to the von Mises hypothesis
the principal stress criterion for brittle, isotropic materials. to consider the different properties for tension and

872
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

compression. He did not add a term for the shear stress

(although it is formulated in the system of orthotropy), and
also did not consider the difference between compression
and tension for the transverse strength. Franklin removed
these weak points and added a further term for biaxial
strength, which has to be determined experimentally. This
term is supposed to increase the accuracy of the criterion.
The resulting equation is
S1¹ ¹ S1þ S2¹ ¹ S2þ j21 K12 j1 j2
þ ¹ j1 þ þ ¹ j2 þ þ ¹ ¹
S1 S1 S2 S2 S1 S1 S1þ S1¹
j22 t212
þ þ ¹ þ ¼1 ð7Þ
S 2 S2 S212
K12 is determined by an experiment with biaxial tension. Figure 3 Sensitivity of jbiax on F12
With the stresses j1 ¼ j2 ¼ j, K12 is determined by
 þ ¹
S1þ S1¹ 1 ¹ þ S1 S1
K12 ¼ 1 þ þ ¹ þ (S1 ¹ S1 ) þ (S2 ¹ S2 ) þ ¹
þ ¹
The remaining coefficients F12 , F13 and F23 have to be
S 2 S2 j S2 S2
determined experimentally. For plane stress and transverse
S1þ S1¹ isotropy, the criterion can be written as
¹ ð8Þ
j2
1 1 1 1 j21
The Hoffman criterian coincides when K12 ¼ 1. ( þ ¹ ¹ )j1 þ ( þ ¹ ¹ )j 2 þ þ ¹
S1 S1 S2 S2 S1 S1
j22 t12
Tsai–Wu (1971) criterion42 þ F12 j1 j2 þ þ ¹ þ ¼1 ð12Þ
S 2 S2 S12
The Tsai–Wu criterion is a tensor polynomial. The The term F12 has to be determined by a biaxial experiment.
formulation has the following form in Voigt’s tensor notation:
This determination is very complicated though, since F12
Fi ji þ Fij ji jj þ … ¼ 1, i ¼ 1…6 (9) has very little sensitivity to the biaxial stresses. Tsai and Wu
mentioned the following experiments for the determination:
Tsai and Wu neglected all summation terms beginning with
biaxial tension, a 458 off-axis specimen (recommended by
the cubic. The cubic term not only increases the number of
many Russian scientists), and a 458 shear test (recom-
unknowns significantly, but would also be responsible for an
mended by Tsai and Wu). Since the best experiment is
open-ended failure envelope.
also dependent on the material, a sensitivity study, as
The same formulation was already used by Zacharov p
shown in Figure 3, should be done. The parameter F12 on
(1963), but is not found in Western literature29. The
the abscissa is the standardized value of F12 :
formulation by Gol’denblat and Kopnov (like the poly-
nomial of Tsai–Wu, but square root of the second p F12
F12 ¼ p
 (13)
summation term) is not accorded more generality by Tsai F11 F22
and Wu, since the equation can be transformed to
Since Tsai and Wu also introduced a stability criterion to
p
2Fi ji þ Fij ji jj ¹ (Fi ji ) ¼ 1
2
(10) guarantee a closed failure envelope, F12
p
has to be in the
range of ¹ 1 # F12 # 1. The stability criterion has the fol-
With regard to curve fitting, the formulation is no better than lowing definition:
that of Tsai and Wu, though more complicated.
Fii Fjj ¹ Fij2 $ 0 no sum over i and j (14)
For transversly isotropic materials, the tensor of the
second order consists only of coefficients on the main Owing to problems with the experimental determination of
diagonal, owing to symmetry. Since strength cannot be F12 , several estimates can be found in the literature. Tsai and
p
dependent on the sign of shear stresses, coefficients with a Hahn43 proposed the value F12 ¼ ¹ 0:5, in order to fulfil the
single 4, 5 or 6 in the index must vanish. For transverse von Mises hypothesis in case of the isotropy. Narayanas-
isotropy the indices are equivalent. Under these assump- wani and Adelman44 even proposed to remove the interac-
p
tions, the equations for the coefficients are the following: tion term, since Pipes and Cole45 measured values of F12
contradicting the stability criterion. Pipes and Cole con-
1 1 1 1 1 1
F1 ¼ ¹ ¹ ¹ F2 ¼ þ ¹ ¹ F3 ¼ þ ¹ ¹
cluded, though, that their experiment was unsuitable for the
S1 S1 S2 S2 S3 S3 determination of the interaction factor. In Figure 4 it can be
p
1 1 1 seen that F12 is not negligible for jx ¼ 6 jy ¼ 6 j, txy ¼ 0.
F11 ¼ F22 ¼ F33 ¼ The stability criterion mentioned above was criticized by
S1þ S1¹ S2þ S2¹ S3þ S3¹
Sendeckyj28, since strength under hydrostatic pressure is
1 1 1
F44 ¼ F55 ¼ F66 ¼ ð11Þ unlimited for most materials. This criticism is unimportant
S223 S231 S212 for the plane stress state, though.

873
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

2 3
b=S2x ¹ 1=2S2z ¹ b=2S2x
6 7
6 7
A3 ¼ 6 a=S2y ¹ a=2S2y 7
4 5
1=S2z
2 3
1=S2yz 0 0
6 7
6 7
B¼6 1=S2zx 0 7 ð18Þ
4 5
1=S2xy
Depending on the actual stress sign, the tensile or the com-
pressive strength is used for Sx , Sy , and Sz .
Figure 4 Influence of F12 on the strength Further properties of this criterion are unlimited strength
for hydrostatic pressure, and an equivalence to the von
Mises hypothesis for isotropic materials. The factor g can
Puppo–Evensen (1972) criterion46 be further adjusted to certain materials by introducing an
Puppo and Evensen criticized the fact that some strength exponent. So far, nobody seems to have made use of this
criteria were not invariant against coordinate transforma- potential.
tion. They described the ‘paradox of the Hill-type criterion’. For plane stress the criterion can be rewritten as*
A balanced (equal number of identical plies in each j1 2 S j j j t
( ) ¹ g( 1 )( 1 )( 2 ) þ g( 2 )2 þ ( 12 )2 ¼ 1
direction) (08,908) s laminate has two sets of axes of S1 S2 S1 S2 S2 S2
orthotropy (in 98/908 and 6458). The failure envelope
j1 2 S j j j t
calculated with a non-invariant criterion is different for each g( ) ¹ g( 2 )( 1 )( 2 ) þ ( 2 )2 þ ( 12 )2 ¼ 1 (g , 1)
S1 S1 S1 S2 S2 S12
set. Consequently the question arises, which set of axes has
to be chosen, when the calculation is not done plywise. ð19Þ
Therefore, an interaction factor g is defined:
The axes of principal strength can differ from those of ortho-
S2xy tropy, when g is larger than unity. In this case they do not
g¼3 (15)
Sx Sy need to be orthogonal too. When g . 1 the following
equation has to be used, which in turn is derived from the
For isotropic materials this factor is equal to unity, for non- system rotated by 458:
interacting materials zero, and for only shear reinforced,
j j j j j j t
otherwise isotropic materials, larger than unity. With this ( 1 )2 ¹ 2h(g)( 1 )( 2 ) þ ( 2 )2 ¹ 2g(g)[( 1 ) 6 ( 2 )]( 12 )
S1 S1 S2 S2 S1 S2 S12
factor the axes of principal strength can now be defined as
t
those axes for which g is minimal. In the three-dimensional þ ( 12 )2 ¼ 1 (g . 1)
case there are three interaction factors: S12

S2yz S2zx S2xy 1

a¼3 b¼3 g¼3 (16) h(g) ¼ 1 ¹ f (g)
Sy Sz Sz Sx Sx Sy 2
p
The strength criterion is like that by Hill of a quadratic g(g) ¼ g=12[3f (g)=(f (g) ¹ 4) þ 1]f (g)
nature. In Voigt’s tensor notation it has the form
" # q
T
Ai 0 f (g) ¼ 0, 1[ (3=g þ 4)2 þ 240=g ¹ (3=g þ 4)] (20)
j Ri j ¼ 1 Ri ¼ i ¼ 1, 2, 3 (17)
0 B In the case that this criterion is used in the system of ortho-
tropy, the axes of principal strength do not have to be cal-
Thus it consists of three intersecting failure envelopes. The
culated. A check whether g is larger or smaller than unity is
submatrices Ai and B are defined by the following
sufficient. In the case g ¼ 1 the difference between eqn (19)
equations:
and eqn (20) vanishes.
2 2 3
1=Sx ¹ g=2S2y ¹ b=2S2z
6 7
6 7 Hashin–Rotem (1973/1980) criterion 47,48
A1 ¼ 6 g=S2y ¹ 1=2S2x 7
4 5
Apart from the maximum stress/strain criterion, the
b=S2z criteria presented so far do not distinguish between the
2 3 different failure modes, as they are based solely on curve
g=S2x g=2S2x ¹ 1=2S2y
6 7 fitting. It is not possible to determine whether it is the fibre
6 7 or the resin that fails. The calculated strength in the case of
A2 ¼ 6 1=S2y ¹ a=2S2z 7
4 5
a=S2z * The reference has a mistake in this equation.

874
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

biaxial tension also depends on the compressive strength in considered

the Tsai–Wu and the Hoffman criteria. To take remedial 1 1 2 1
action, Hashin supplemented a criterion, which was (j2 þ j3 )2 þ (j23 ¹ j2 j3 ) þ 2 (j212 þ j213 )
S2þ 2 2
S23 S12
developed by Hashin and Rotem, to describe fatigue of
composites. It takes distinct fibre and resin failure modes ¼ 1 (j1 þ j2 . 0) ð27Þ
into consideration.
In the case of matrix failure due to compression, an
The criterion starts out with transverse isotropy (S2 ¼ S3 ,
additional condition can be taken into account. Analogous
S23 ¼ S31 ). In order to guarantee invariance against rotation
to hydrostatic pressure, it is assumed that the strength is
around the fibre axis, the invariants of the stress tensor are
significantly higher in the case of equal stresses in both
used to derive the criterion. These invariants are
transverse directions than in the case of uniaxial stress. By
I1 ¼ j11 adding the linear terms to the previous equation, one arrives
at
I2 ¼ j22 þ j33 1 S2¹ 2 1
[( ) ¹ 1](j2 þ j3 ) þ 2 (j2 þ j3 )2
S2¹ 2S23 4S23
I3 ¼ j223 ¹ j22 j33
1 2 1
I4 ¼ j212 þ j213 þ (j23 ¹ j2 j3 ) þ 2 (j212 þ j213 ) ¼ 1 (j2 þ j3 , 0)
S223 S12
I5 ¼ 2j12 j23 j13 ¹ j22 j213 ¹ j33 j212 (21)
(28)
I5 cannot be used, owing to the linear shear stresses. Using a The distinction, if j2 þ j3 is larger or smaller than zero,
quadratic formulation, it follows that stands for the sign of the normal stress on the fracture
plane. For the derivation, please refer to the reference. How-
A1 I1 þ B1 I12 þ A2 I2 þ B2 I22 þ C12 I1 I2 þ A3 I3 þ A4 I4 ¼ 1
ever, from this derivation it can be concluded that matrix
(22) failure occurs in quadratic approximation always in a plane
From pure transverse and respectively longitudinal shear of maximal, transversal shear stress.
stress A3 and A4 are For the plane stress state, the equations can be rewritten as
j t
1 1 ( þ1 )2 þ ( 12 )2 ¼ 1 (j1 . 0)
A3 ¼ A4 ¼ (23) S1 S12
S223 S212
j1 ¼ ¹ S1¹ (j1 , 0)
Only stresses acting on the fracture plane are taken into
account for the different failure modes. It is assumed that j2 2 t12 2
tensile and shear stresses are mutually weakening, therefore ( þ) þ( ) ¼ 1 (j2 . 0)
S2 S12
causing fibre failure
j2 2 S¹ j t
j 1 ( ) þ [( 2 )2 ¹ 1] ¹2 þ ( 12 )2 ¼ 1 (j2 , 0) (29)
( þ1 )2 þ 2 (j212 þ j213 ) ¼ 1 (j1 . 0) (24) 2S23 2S23 S2 S12
S1 S12
It should be emphasized that it is very hard to measure S23 .
The linear term of j1 is neglected, since a uniaxial tensile However, Puck11 concluded that S23 ¼ S2þ since the plane of
test delivers only one value. An even more simple approxi- fracture is always rotated 458 to the pure shear stress state.
mation is achieved in the use of the maximum stress
criterion:
Wu–Scheublein (1974) criterion50
j1 ¼ S1þ (j1 . 0) (25)
The authors of this criterion saw the problem of quadratic
It is also used for the compressive strength. Rosen 49 approximation was that only orthotropic laminates can be
reported that microbuckling takes place in a shear mode, handled without plywise analysis. Since a multidirectional
but it was not known whether additional shear stress is laminate is often anisotropic, they added the cubic tensor in
weakening or strengthening. Therefore the tensor polynomial used by Tsai and Wu. This criterion is
not to be applied to unidirectional plies. Because of the
j1 ¼ ¹ S1¹ (j1 , 0) (26) number of coefficients, Wu and Scheublein restricted the
The description of the matrix strength is more complicated, derivation to the plane stress state. Analogous to Tsai and
since the plane of fracture is not known, a priori. There- Wu, they figured
fore the plane would have to be identified by a rotation, F1 j1 þ F2 j2 þ F11 j21 þ F22 j22 þ F66 j26 þ 2F12 j1 j2
similar to Mohr’s circle, for which the strength criterion
has a maximum. This is not only complicated, but also the þ 3F112 j21 j2 þ 3F221 j22 j1 þ 3F166 j1 j26
criterion would no longer be quadratic. The normal
þ 3F266 j2 j26 ¼ 1 ð30Þ
stress in the fibre direction is neglected. Again, the
linear term of the transversal normal stress cannot be The diagonal terms for the linear and quadratic tensor

875
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

remain the same: Chang Chang (1984) criterion54–56

1 1 1 1 1 The Chang Chang criterion considers also non-linear
F1 ¼ ¹ F2 ¼ ¹ F11 ¼ elastic behaviour of the matrix. It distinguishes between
S1þ S1¹ S2þ S2¹ S1þ S1¹
matrix cracking and fibre breakage or fibre–matrix shearing.
1 1
F22 ¼ F66 ¼ ð31Þ Since non-linearities between shear stresses and shear
S2þ S2¹ S212 strains are more important than those between normal
stresses and strains for composites, only the shear stress–
An experimental determination of the interaction terms
strain relation is considered to be non-linear. To incorporate
would end in an enormous number of tests. The best suited
the strain history in the stresses Chang, Scott and Springer
ratio of stresses for the determination depends also on the
combined the Yamada–Sun criterion57 with Sandhu’s strain
interaction term, to be measured. Thus, a number of itera-
energy criterion58,59. The resulting criterion is for both
tions would be necessary to give reliable results. Needing n
failure modes of the form
iterations, 5 þ 5n experiments would be necessary to deter-
Zg12
mine the ten coefficients. Since one normally averages each
j j12 dg12
value out of several tests, the number of tests would further ( 1=2 )2 þ Z0e12 ¼1 (35)
increase. Consequently, Wu and Scheublein used a hybrid S1=2
j12 dg12
method to reduce the number of experiments. In the first 0
iteration, the strengths are calculated by using a plywise The non-linear shear stress–strain relationship is expressed
analysis. The calculated points of the failure envelope can according to Hahn and Tsai60 as
be expressed piece by piece by a polynomial of the third
j
order. The calculated failure envelope delivers the biaxial g12 ¼ 12 þ aj312 (36)
strength of the laminate for further iterations. Finally the G12
optimal ratio of stresses will be known. The number of G12 is the initial ply shear modulus. The constant a can be
experiments has been reduced from 5 þ 5n to ten tests. calculated from the stress–strain curve of an off-axis test
The experimental interaction term has to be compared under an arbitrary angle with
with the calculated value, and in the case of significant «y ¹ «yelast
deviation, an experimental iteration has to follow. a¼ (37)
j3x
Using this non-linear relationship, the integration above can
Tennyson–MacDonald–Nanyaro (1980) criterion51–53 be performed, thus:
Following the same reasoning already mentioned by Wu j212 3
and Scheublein, Tennyson et al. also thought of using a þ aj4
j1 2 2G12 4 12
cubic tensor polynomial. The derivation is similar up to the ( ) þ 2 ¼1
S1 S12cr 3 4
point of the determination of the cubic interaction terms. þ aS
They too used the hybrid method, but needed only one 2G12 4 12cr
experiment to determine F12 (plus five tests for uniaxial
strengths). The cubic interaction terms are estimated with j212 3
þ aj4
this quadratic interaction term by setting the discriminant of j2 2 2G12 4 12
( ) þ 2 ¼1 (38)
the cubic tensor to zero. The stresses can be replaced by a S2 S12cr 3
loading factor l: þ aS412cr
2G12 4
j 1 ¼ k1 l j 2 ¼ k2 l j 6 ¼ k6 l (32) According to Yamada and Sun, the shear strength of a cross-
eqn (30) can be rewritten as ply laminate should be used instead of that of a lamina, since
the shear strengths of laminates are usually higher than the
al3 þ bl2 þ cl þ d ¼ 0 strenghths of laminae.

a ¼ 3(F112 k12 k2 þ F221 k22 k1 þ F166 k1 k62 þ F266 k2 k62 )

Truncated maximum strain (1990) criterion61,62
b ¼ F11 k12 þ F22 k22 þ F66 k62 þ 2F12 k1 k2
Hart-Smith created a very simple criterion based on the
c ¼ F1 k 1 þ F2 k 2 Tresca yielding condition. Its application is limited to thin
laminates (plane stress state) where ultimate failure is fibre
d¼ ¹1 (33) dominated. These laminates have to be made of strong stiff
fibres in a soft matrix, i.e. composites with n21 < 0. For such
Setting the discriminant to zero, it follows that
a laminate, only e1þ , e1¹ and n12 are needed. For better
27a2 þ a(4c3 þ 18bc) ¹ 4b3 ¹ b2 c2 ¼ 0 (34) visualization, the construction of the failure envelope is
shown instead of the descriptive formulae (Figure 5). The
The cubic interaction terms can now be calculated by insert- parameter gcrit is calculated by
ing the stresses responding to the three uniaxial and the one
biaxial strength. gcrit ¼ (1 þ n12 )max(e1þ ; je1¹ j) (39)

876
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

þ
The factor p12 indicates the absolute value of the failure
envelope’s slope when intersecting the t12 axis. It has to
be determined experimentally and is supposed to increase
the accuracy. The section of the failure envelope charac-
terized by mode B is expressed by a parabolic equation,
thus:
q
1
( S212 þ (p12 ¹ j )2 þ p ¹ j ) ¼ 1
2 12 2
S12

j2 R23
(j2 , 0, 0 # j j # þ p
¹
) (42)
j12 S12 1 þ 2p23
The inclination parameter again indicates the slope at the t12
axis. In this case it can be interpreted as a measure for
Coulomb’s friction, which increases the shear strength in
the case of pressure on the fracture plane. The factor R23
will soon be explained. The fracture angle is calculated by
Figure 5 Tresca’s and truncated maximum strength criterion’s failure
envelopes
searching for the global maximum of the effort (ratio of
current loading to possible strength):
r
j j ¹ jn )2 þ p ¹ j n
E(v) ¼ ( nt )2 þ ( nl )2 þ (p12 12 (43)
Puck (1996) criterion11 R23 S12 S12 S12

Like Hashin and Rotem, Puck also starts from Mohr’s One will observe that jn will remain constant at the value
hypothesis, which states that in the case of brittle materials ¹ R23 when calculating the fracture plane for mode C. Thus,
only stresses acting on the crack plane are responsible for the fracture plane can simply be calculated by
fracture. Because of the increased potential of computers, he s
no longer neglected the determination of the fracture plane R23
v ¼ arccos (44)
for matrix failure. The derivation of the criterion, presented ¹ j2
in this survey, is restricted to the plain stress state, since
interesting simplifications occur in this stress state. The The resistance of the material to transverse shear stress on
three-dimensional derivation is described in the reference. the fracture plane is called R23 by Puck. Since a specimen
Fibre failure is assumed to be independent of the stresses loaded with pure transverse shear stress fails at an angle of
other than that in the direction of the fibres. However, owing 458, the reason for failure is tensile and not shear stress on
to the distinct Poisson ratios of fibre and resin, normal fibre the fracture plane. Therefore R23 is not equivalent to the
stresses are induced by transverse stresses. The ultimate strength S23 . Instead, it can be measured by uniaxial com-
stress is somewhere between the uniaxial ultimate stress and pression, where R23 is distorted by Coulomb’s friction,
the stress calculated from ultimate strain. Since the though. However, Puck derived R23 from Mohr’s circle
difference is only very small, the strength is approximated based on transverse compression:
q
by S¹ S¹
R23 ¼ 2 cotanv ¼ 2 ( 1 þ p23 ¹v
¹ p23 ) (45)
1 j 1 «1 2 2
j þ 6j ¼ 1 (40)
2 S6
1 e1 ¹
The inclination parameter p23 ¹
is assumed to be tied to p12 by
¹
Puck mentions, though, that the hypothesis of maximum p23 p¹
¼ 12 (46)
fibre stress is not acceptable to others, and therefore needs S23 S12
further investigation. Now, the failure envelope for mode C can be expressed by
The inter-fibre fracture is described by three equations the following elliptical equation:
referring to different failure modes. Failure due to tension
t12 j S¹
and shear is called mode A, failure due to shear and lesser [( )2 þ ( ¹2 )2 ] 2 ¼ 1
compression mode B, and shear with stronger compression 2(1 þ p23 )S12
¹
S2 ¹ j2
mode C. While the fracture plane is perpendicular to the p
¹
transverse stress for modes A and B, it is oblique for mode j12 S12 1 þ 2p23
(j2 , 0, 0 # j j # ) (47)
C. In mode C the fracture occurs in a plane of maximal shear j2 R23
stress. The failure envelope is described for mode A by an The equations used for the derivation of the elliptical
elliptical equation: equation for mode C started out by neglecting the longitu-
s
þ dinal stress. Instead, Puck assumes that high longitudinal
t þ S2 2 j 2 2 þ j2
( 12 )2 þ (1 ¹ p12 ) ( ) þ p12 ¼ 1 (j2 . 0) stresses (0.7S1 ) reduce the transverse strength due to local
S12 S12 S2þ S12
fibre failure and fibre debonding. This reduction is
(41) expressed by an ellipse, which reduces the strength to

877
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

50% progressively, beginning at 0.7S1 : I3 ¼ t231 þ t212

1 j
fw2 þ (j 61 j ¹ 0:7)2 ¼ 1 (48) I4 ¼ (j2 ¹ j3 )2 þ 4t223
0:12 S1
Unity on the right side of the equation sign in eqns (41) and I5 ¼ (j2 ¹ j3 )(t231 ¹ t212 ) þ 4t12 t23 t31 (50)
(42) and eqn (47) can then be replaced by fw . The determi-
nation of the fracture angle offers the additional benefit that The first invariant is not based on the ‘smudged’ longitudi-
the danger of delamination due to an oblique crack (mode C) nal laminate stress but on the fibre stress times the fibre
can be noticed. volume ratio. This is done because longitudinal strength is
In the case of mixed modes, the three-dimensional assumed to depend only on fibre strength. It first appears to
determination of fracture angles becomes numerically be formalism but in case of thermal stresses it might just
unstable. This can also be deduced from the physical lead to a difference. One will observe that the strength
view, since the orientation of the voids also controls the depends on the sign of the shear stresses due to I5 . This
fracture angle when the effort is comparatively indifferent mistake was made on purpose, in order to take into con-
to the angle. Therefore, corners can be rounded with a sideration the distinct interactions of j2 (j3 respectively)
probabilistic formulation. The area of significant effort with t12 and t31 . In the plane stress state the linear shear
(more than 50%) over the variable fracture angle is an stress terms vanish. For the plane stress state the equations
appropriate measure for the reduction of strength. are
Jj1
F1j : aj1 ¼1
Cuntze (1995) criterion 63 S1þ

The strength criterion of Cuntze is chronologically Jj1 j t

F1t : at1 ¹ þ bt1 ¹2 ( þ ct1 ( 12 )2 ) ¼ 1
misplaced. However, it was influenced by the development S1 S1 R12
of the Puck criterion, which goes back to the year 1970. In
2
contrast to Puck, Cuntze did not determine the fracture j2 j2 2 j j 2 t12 Jj
þ þ b12 ( þ ) ( þ c12 þ ej12 ( þ1 )2 ) ¼ 1
j
F12 : aj12 j
angle. Instead, he used those invariants of the elasticity S2 S2 S123 S1
tensor that included the stresses responsible for the distinct
failure modes corresponding to Puck’s analysis. Also t12 2 j t2 Jj
isotropic materials or fabrics can be dealt with by using a F12 : a12 ( ) þ b12 2 3 12 ( þ c12 ( þ1 )2 ) ¼ 1
S12 S12 S1
different set of invariants. For a transverse isotropic material
the equations for the different failure modes are the j2 t j2 2 t t12 2 Jj
following†:
F2t : at2 ¹ þ b2 ( ¹ ) ( þ c 2 ( ) þ d2t ( ¹1 )2 ) ¼ 1 (51)
S2 S2 S12 S1
aj1 I1
F1j : ¼1
S1þ
Application of the strength criteria without distinct failure
at1 I1 bt1 I2 ct1 I3 modes
F1t : þ (þ )¼1
S1¹ S1¹ S2 There are different ways to apply criteria without distinct
failure modes to multidirectional laminates:
aj12 I2 bj12 I4 cj12 I22 ej12 I2 I3 f12
j
I5 gj12 I12
j
F12 : þ þ ( þ þ þ þ2 ) ¼ 1 • a plywise analysis,
S2þ S2þ
2
S2þ
2
S312 S312 S1 • application to the laminate as a whole,
• determination of the strengths in the direction of the axes
a12 I3 b12 I2 I3 c12 I5 d12 I12 of orthotropy with a plywise analysis, which are then
F12 : þ þ ( þ 2 )¼1 inserted in the criterion for the laminate as a whole.
S212 S312 S312 S1þ
The first method allows determination of the laminate’s
at2 I2 bt2 I4 ct2 I3 d2t I12
F2t : þ ( þ þ ¹2) ¼ 1 (49) strength with the data of the unidirectional plies. These data
S2¹ S2¹
2
S212 S2 can be found in tables. Experiments could sometimes be
The summation terms in brackets describe mixed mode fail- spared. However, the stresses in the plies would have to be
ure. A probabilistic calculation is replaced by these correc- calculated anew for every stress state. These calculations
tion terms. In the case of a porous matrix, also a ‘crumbly’ can be avoided by using the second method. Instead, the
fracture mode can be considered, which leads to failure strengths have to be determined experimentally for every
under hydrostatic pressure. The invariants are given by distinct laminate structure. The hybrid method tries to
combine the advantages of the previous two methods. The
I1 ¼ Jj1q
strengths, which were determined experimentally in the
second method, are now calculated with the first method.
I2 ¼ j2 þ j3
Further plywise analysis is not necessary. However, it is not
possible to determine the failing ply. This can only be done
† Formula of ref.63 changed according to64. by using the first method.

878
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

Figure 6 Comparison of the strength criteria in the j1 ¹ j2 graph

Figure 7 Comparison of the strength criterion in the j1 ¹ j12 and j2 ¹ j12 graph

Comparison of the strength criteria The difference in the diagrams in which j12 over j1 or j2
is drawn is no longer clear. The criteria similar to the Tsai–
The following shows graphically some of the presented
Wu criterion, only distinct in the interaction term, form the
strength criteria. The calculations are based on a carbon/
same failure envelope (Figure 7).
epoxy-HT lamina. With the exception of the Tsai–Wu and
From these graphs one would conclude that a biaxial test
the Puck criteria, the criteria needing experimental data are
with the two normal stresses would be best suited to
left out. For the Tsai–Wu criterion the value suggested by
determine the most accurate criterion. However, the Puck
Tsai and Hahn was used for the interaction term
p criterion seems to be very plausible and therefore a test in
(F12 ¼ ¹ 0:5). According to Puck the strength S23 was
the j2 ¹ t12 stress state seems to be more useful.
replaced by S2þ for the Hashin–Rotem criterion. The
However, it has not yet been ‘proven’ that the longitudinal
inclination parameters in the Puck criterion are taken from
¹ þ strength is independent of the transverse stresses. This still
the reference and are p12 ¼ 0:2=p12 ¼ 0:3.
has to be clarified. Hashin and Rotem, Puck, as well as
The most significant difference can be seen in the j1 ¹ j2
Cuntze derived their criteria from this assumption. Later this
graph (Figure 6). The Tsai–Hill criterion delivers the most
year the results of a project funded by the German Ministry
conservative values (the suitable strength is used for each
for Education and Science will be published in which the
quadrant). The increase in biaxial strength is the highest for
three-dimensional criterion of Puck was experimentally
the Tsai–Wu criterion. The maximum stress and the
examined 65.
Hashin–Rotem criterion have exactly the same failure
While tensor polynomials seem to be the most accurate
envelope and almost the same envelope as the Puck criterion
for laminae (Puck and Cuntze criterion are excluded), one
in this graph. The Puppo–Evensen criterion seems to work
observes with laminates that the maximum stress/strain
only with laminates with less orthotropy, since its failure
criterion describes failure better than the others. The reason
envelope is not closed.

879
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

Figure 8 Determination of strength with maximum stress and Tsai-Wu criterion for a [08, 6 608] s laminate

is the definition of the term failure. Experiments normally can be seen in Figure 9, the different specimens also have
record ultimate fracture as failure. Ultimate fracture is in different properties. While the differences in strength and
most cases caused by fibre failure. Consequently, the test modulus are small for tensile specimens, they become more
data are close to the lines of fibre failure in the maximum significant for compressive or in-plane shear tests. Much
stress/strain criterion (see Figure 8‡). Tensor polynomials higher compressive strengths can be achieved with the
and other criteria, that do not distinguish between failure specimen developed by Matthews and Haeberle67 as well as
modes, indicate failure at plainly lower stress levels, when with that developed by Curtis et al.68. The latter can be used
used in plywise analysis. This failure is mainly character- with the Celanese rig. Scatter in tensile tests can be reduced
ized by matrix fracture. Since matrix failure causes non- by cross-plies in the 908 direction69.
linear behaviour and reduces Young’s modulus, an accurate
determination of the beginning point of this fracture mode is
also of interest. A compromise solution could be the use of
the tensor polynomial for the determination of matrix failure
and the maximum stress criterion or eqn (24) for fibre
failure. This aspect might just be of historic value owing to
the expected superiority of the Puck and Cuntze criteria.

SPECIMEN TYPES

Just as there are macroscopic and microscopic failure

criteria, there are also two principal types of specimen.
Specimens of the first type are designed for a homogeneous
stress state in the test section, whereas those of the second
type have local stress peaks due to holes or notches. To
verify macroscopic failure criteria, specimens with a
homogenous stress state are much better suited since they
satisfy the demand of a homogeneous stress state in a
statistically significant test volume. Further requirements on
such a specimen are that failure has to start in the test section
and that it must in no case be caused by buckling.
Unidirectional (UD) laminae can be tested as well as
multidirectional laminates. Testing of UD laminae allows
the construction of a database for the different materials.
Interaction of stacked laminae can be examined by testing
laminates.
The uniaxial strengths needed for the failure criteria are
normally determined with special specimens which are
described by special standards, such as ASTM or CRAG. As

‡ Data points are not measured but only used for illustration. Reduction of Figure 9 Comparison of different compression tests (average and
strength is not considered. standard deviation)66

880
 A review of the biaxial strength of fibre-reinforced

Figure 12 Crossbeam by Boehler and Demmerle

Figure 10 Comparison of the failure criteria with the off-axis test

Figure 13 Flat plate

Contraction due to Poisson’s ratio is not supported by the

rigid tabs. Special attention has to be paid to small angles
ensuring that no single fibre is clamped by both grips.

Crossbeam
The crossbeam is a sandwich construction in which the
Figure 11 Crossbeam laminate to be tested is on one of the outer coats (Figure 11).
The stress state is induced by bending of the beams. Shear
can be engendered by changing the orientation of the
Off-axis specimen fibres. Bert et al.71 proposed an elliptical reduction of
thickness for better homogeneity of stresses. The ellipse
The off-axis specimen has the most simple geometry. It is is described by
a flat plate with a distinct angle between uniaxial loading r
and fibre orientation. The stress state is biaxial when b j2
¼ (53)
transformed parallel to the axes of orthotropy. The three a j1
plane stresses cannot be varied independently since they are It also guarantees failure in the gauge section. The stress
tied to each other by the transformation tensor. The state, however, in the test section is not determined by loads
difference between the distinct criteria is not significant as applied, and not by the geometry72. The only advantage to
can be seen in Figure 10. Pipes and Cole reported in ref.45 be found is in the quadrant compression/compression,
that the off-axis sample is not suitable for determination of whereby the sandwich construction increases stability.
the interaction term F12 . Matrix failure may be caused by Nowadays, this specimen is becoming popular again.
the free-edge effect which can result in ultimate failure Boehler and Demmerle73 optimized the test section with
before the actual strength is reached. The specimen has regard to the stress state and also with regard to the certainty
either to be designed with a high aspect ratio or clamped of the stresses with finite element calculations. As param-
with a pivoted grip to prevent moments induced by eters they used the sizes shown in Figure 12 (1/8 of the
orthotropy. Another method is the use of oblique rigid specimen) and in addition the ratio of the cross-sectional
end-tabs70. The angle to the longitudinal direction of the area of limbs and slots, the distribution of the slot width
specimen is given by and the limb width and the number of slots. The halfwidth
t and the number of slots were fixed for the numerical opti-
C̄ 16 exy gxy
cotv ¼ ¹ C̄ 11 ¼ C̄16 ¼ (52) mization. The resulting geometry depends on the material.
C̄ 11 jxx jxx Loads are applied by in-plane tension.

881
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

elements for glass/epoxy in the range of ¹ 158 # a # 208.

The analysis leads to the following conclusions: at 08 load
angle the stress state is almost homogenous; the higher the
absolute value of the load angle, the less homogenous the
stress state; determination of the shear modulus is accurate
while the measured shear strength is less, compared to other
types of specimen.

Thin-walled tube
The tube is the most versatile specimen. All three plane
stresses can be applied independently from each other. The
free-edge effect is not present. Axial stresses are applied by
axial tension or compression, hoop stresses by inner or outer
pressure, and shear stress by torsion. Three different kinds
of fibre orientation can be distinguished: unidirectional
laminae in axial or hoop orientation, helically wound
laminae, and multidirectional laminates. It is argued that
helically wound tubes are more general since the transfor-
mation of elasticity is also checked77. The main concern
Figure 14 Circular specimen might be though to find a compromise between high axial
loads for 08 and high pressures needed for 908 wound tubes.
However, owing to orthotropy, additional stresses are
Flat plate induced which may lead to other difficulties.
The term ‘thin’ has to be defined in a more conservative
The flat plate is pulled or pushed at all four edges manner than for isotropic materials, owing to orthotropy.
(Figure 13). Shear stresses are controlled by the fibre Rizzo and Vicario77 defined it as t=d # 0:02. It has to be
orientation. Whiffle-tree linkage grips are used to allow mentioned that their analysis was for a helical angle of 308
contraction. Nevertheless the stress state is not homogenous74, which is very unfavourable for a homogenous stress state.
since the grips are not infinitely small. An improvement can The most disadvantageous angles of helically wound
be achieved with an elliptical reduction of the thickness. laminae are 308 for axial loads and 608 for torsion and
This geometry is mainly accepted for notched specimens. pressure78. There are several analytic solutions in the
literature for the stress state in a tubular specimen, some of
Circular specimen them taking into consideration the clamping constraints78–82.
In the age of less and less expensive CPU time, finite element
Arcan, Hashin and Voloshin75 introduced a new methods should have preference, since they require less
geometry for biaxial tests in 1978 (Figure 14). The biaxial simplification of the problem.
stress state is controlled by the load angle a (a ¼ 08 Choo and Hull83 remark that inner pressure induces a
corresponding to pure shear): three-dimensional stress state on the inside of a tube. The
jx ¼ ja sina radial stress is calculated according to Timoshenko84 with
Pi ri r
jy ¼ ja sina jr ¼ [1 ¹ ( o )2 ] (55)
ro2¹ ri 2 r
txy ¼ ja cosa For a thin tube this can be approximated to a linear decrease
in radial stresses. The radial stress decreases for example
Pa
ja ¼ (54) the shear stress in the 458 plane, when axially compressed.
Atest section This shear stress has to be zero on the outside owing to
The cross-section is measured at the narrowest spot of the equilibrium. Consequently this influence cannot be
gauge section. neglected in general, as has been the case. Instead the
Marloff76 examined this specimen geometry with finite radial stress should be inserted in a three-dimensional
criterion with 0–0.5Pi .
The major source of stress inhomogeneities are the
collets, especially when loaded with pressure. Also in the
case of torsion or when helically wound tubes are tested,
attention has to be paid to the length of the specimen. When
tubes without reinforced end sections are used, failure
normally occurs near the grips. The most common design
for the reduction of stress concentrations is thickened end
Figure 15 Tube design by Swanson sections. A very successful design has been made by

882
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

Figure 16 Test facility by Cole and Pipes

Figure 18 Typical fabrication set-up

has to be examined. Wang and Socie1 reported that a tube

failed under outer pressure by delamination while a flat
specimen failed at higher stresses due to kinking.
Krempl and Niu89 observed that a 6 458 wound tube had
a different torsional strength for both turning directions.
They concluded that strength depended on the orientation
Figure 17 Collet by Lindholm et al. of the outer ply, and thus buckling was the actual failure
mode.
The explosion-like fracture impedes the determination of
Swanson (Figure 15)85,86,2. Much more complicated is the the failure mode, hence accompanying calculations and
attempt to match the radial displacement of the end sections SEM investigations sometimes have to be used. For the
with that of the gauge section. Such radially adjustable calculation of the critical buckling stresses, solutions can be
collets have been developed by Cole and Pipes72 as well as found in the references90–94. Since Ley et al.94 described
Lindholm et al.87,88. The more expensive test facility is buckling of ring-stiffened tubes, clamping constraints also
justified by cheaper, because shorter, specimens. can be taken into consideration.
Cole and Pipes used an arrangement as shown in The manufacturing process has to generate a laminate
Figure 16. There are no conventional collets but tabs structure comparable with industrially produced laminates
sealed with O-rings. Axial and hoop stresses are loaded only on the one hand, and to reduce scatter due to production
by hydraulic pressures. The pressures Pi and Po control the tolerances on the other. Consequently, the fibre volume ratio
hoop stresses. Both pressures together respectively Pax are J has to be at least 0.6. Tubes can be produced by filament
used to apply axial loads. Radial displacement of the stiffer winding (winding with very thin rovings), and winding with
tab section is forced by P1 and P2 . Torsion is loaded by a larger rovings, wide tapes, or fabrics. The use of tapes (at
mechanical link (not shown in the figure). A total least for 908 winding angles) and fabrics for tubes is
elimination of the constraints, especially when the tube is complicated, since the generation of continuous plies is
loaded simultaneously axially and torsionally, is not impossible. Either stiffness or strength is discontinuous,
achieved. Stress concentrations are induced mainly, depending on the chosen joint33. Cole and Pipes72 though,
although less significant when compared to conventional could not find a correlation between the location of
collets, by local bending of the tabs. Owing to this concept discontinuity and of transverse failure. They paid attention
of applying loads, a new problem appears: the three- to the angle of the lap joint, so that it did not coincide with
dimensional stress state, when loaded in axial tension, can any direction of minimal strength. Thick rovings often lead
no longer be neglected. This is especially true for to an inhomogenous fibre density95. At the top edges of
unidirectional lamina owing to the high Poisson ratio n32 . these rovings, regions with low fibre content are formed.
Lindholm et al. designed a radially adjustable collet The highest quality can be achieved by filament winding. To
without changing the loading procedure. As can be seen in achieve high fibre volume contents, surplus resin has to be
Figure 17, the collet consists of a grip separated in 2 3 12 pressed out of the laminate. This is done by winding the
segments. The clamping pressure is summed up by P1 and roving on a hollow mandrel, which can be internally
P2 . For radial displacement one of the pressures is increased pressurized. Typical cavity tools can be seen in Figures 18
while the other is decreased. To reduce the influences of the and 19. Before the winding of rovings, a vent cloth to draw
clamping constraints, one would be tempted intuitively to off volatile gases and a bleeder for surplus resin are wound
increase the length of the specimen. According to a finite around the mandrel. Separation between them and the
element analysis by Rizzo and Vicario77, a sample with high laminate is achieved by perforated Teflon films for
aspect ratio (l/d) leads to higher stress gradients in the cross- example81. Byung and Lehnhoff 96 even vacuumed surplus
section. resin and gases through porous spacers at the ends of the
Since strength is controlled by the failure mode, the mode tube. Vent cloth and bleeder are therefore not necessary to

883
 A review of the biaxial strength of fibre-reinforced

CONCLUSIONS

During the last 30 years many strength criteria have been

formulated. Most of them are based on the von Mises
hypothesis. This hypothesis though was formulated to
describe yielding of ductile, isotropic materials. Its
application to brittle, anisotropic, fibre-reinforced plastics
is doubtful. For example lower transverse tensile strength
would lead to higher biaxial compressive strength. Instead,
it is Mohr’s hypothesis that seems to describe the
micromechanical relationships more realistically. This
hypothesis takes only stresses acting on the failure plane
into account. Mohr’s hypothesis was further derived for
composite materials by Hashin and Rotem, Puck and
Cuntze. A good failure criterion should distinguish the
different failure modes. This is not only a step towards
Figure 19 Cavity tool by Whitney failure mechanics but also facilitates a much better
description of a laminate’s strength. Since distinct failure
modes produce cornered failure envelopes, probabilistic
aspects should be considered to round off these corners.
receive a laminate with high fibre volume ratio and little This is physically plausible and will increase further the
voids. Filamentary wound tubes do not necessarily need accuracy of the predicted strength.
removal of surplus resin since the winding process already Strength criteria from the 1960s described unidirectional
presses the resin out of the laminate. Determination of the laminae. Later it was thought that the strength of laminates
fibre volume ratio is difficult then, because some surplus could only be derived inaccurately from the properties of
resin remains on the outer surface. laminae. Thus, criteria for whole laminates were formu-
Inner layers can be weakened by winding tubes several lated. Nowadays the initial concept is becoming more
layers thick. The final strength of a layer is almost popular again. A distinction between failure modes is
independent of the applied tension of the winding process therefore necessary. Should the strength be described
as long as the layer tension is greater than zero. For no accurately by the new criteria, the next step would be to
tension, or worse, compression, strength is reduced by understand the interactions of the laminae in a laminate.
inaccurate fibre orientation. An applied tension of 5–10 N Also the application to woven fabrics is of interest, since
ensures parallel, well orientated fibres. However, when strength is reduced by the curvature of the fibres and
several layers are wound, each new layer compresses the changing thickness of the fabric.
mandrel and the previously wound layers. Knight97 For biaxial experiments the tubular specimen is widely
evaluated with calculations and experiments that a kevlar/ accepted. Besides the complicated fabrication processes, the
epoxy tube with 64 plies and 5.9 N applied tension on a introduction of loads causes problems. The influence of
polymethylenethacrylate mandrel (d ¼ 150 mm, t ¼ 3:3 mm) radial stresses due to internal or external pressure has not yet
with more than half of the plies cured in a state of been evaluated. Also testing of fabrics and prepeg tapes is
compression. With a tube consisting of 36 plies, 8.9 N not easy with this design. However, no other sample
applied tension, and an aluminium mandrel (d ¼ 150 mm, geometry causes less problems. When one is familiar with
t ¼ 6:2 mm) the layer tension fluctuated between 125 MPa the problems, designs, manufactures and when one
and 180 MPa. Therefore the mandrel could already be measures accurately, the tubular specimen is still best
inflated during the winding process. When choosing the suited for the determination of biaxial strength. When the
applied tension, it also should be considered that lower interaction of laminae in laminates is being investigated, the
applied tensions do not lead to coincident tension in the absence of the free-edge effect might be undesirable.
laminate, owing to rapid resin flow and cross-section Whether the off-axis test will provide the alternative is
flattening97. Knight states this correlation between layer doubtful. Probably a combination of different specimens
and applied tension for kevlar/epoxy laminates will be needed, particularly so that differences in strength
q due to geometry can be detected and measured. As fabrics
2
Flayer ¼ Fapplied ¹ 4, 52 (56) gain more and more popularity a new testing machine for
flat specimens would be desirable. The crossbeam might
Mistakes can also be made when measuring the loads. Fric- just be the starting point.
tion in the test set-up can reduce the applied loads or prevent
contraction. Internal combined with external pressures
induce axial stresses in the test set-up but not in the sample. ACKNOWLEDGEMENTS
When high hoop strains occur, the actual diameter instead of
the nominal diameter should be used to calculate the hoop This work was supported by DFG through Graduierten-
stress. Kollegg Polymerwerkstoffe at Technische Universität Berlin.

884
 A review of the biaxial strength of fibre-reinforced plastics: H. Thom

REFERENCES composite materials. In Analysis of the Test Methods for High Mod-
ulus Fibres and Composites, ASTM STP 521, American Society for
Testing and Materials, Philadelphia, PA, 1973, pp. 167–180.
1. Wang, J.Z. and Socie, D.F., Biaxial testing and failure mechanisms 25. Dickerson, E.O. and DiMartino, B., Off-axis strength and testing of
in tubular G-10 composite laminates. In Composite Materials: filamentary materials for aircraft application. Advanced Fibrous
Testing and Design (11th Conf.), ASTM STP 1206, American Reinforced Composites, 10th Natl. Symp.. Society of Aerospace
Society for Testing and Materials, Philadelphia, PA, 1992, 136–149. Materials and Process Engineers, 1966, 10, H23–H50.
2. Swanson, S.R., Christoforou, A.P. and Colvin, G.E. Jr., Biaxial 26. Griffith, J.E. and Baldwin, W.M., Failure theories for generally
testing of fiber composites using tubular specimens. Experimental orthotropic materials. Developments in Theoretical and Applied
Mechanics, 1988, 28, 238–243. Mechanics, 1962, 1, 410–420.
3. Gerharz, J.J. and Schütz, D., Schrifttumsrecherche zum Festigkeits- 27. Franklin, H.G., Classic theories of failure of anisotropic materials.
verhalten von Faserverbundwerkstoffen—Analyse zum Stand der Fibre Science and Technology, 1968, 1, 137–150.
Technik, TB-145, 1979. 28. Sendeckyj, G.P., A brief survey of empirical multiaxial strength
4. Miller, A.G. and Wingert, A.L., Fracture surface characterization of criteria for composites. In Composite Materials: Testing and
commercial graphite/epoxy systems. In Nondestructive Evaluation Design, ASTM STP 497, American Society for Testing and
and Flaw Criticality for Composite Materials, ASTM STP 696, Materials, Philadelphia, PA, 1971, pp. 41–51.
ed. R.B. Pipes, American Society for Testing and Materials, 29. Schneider, W., Versagenskriterien für Kunststoffe. Zeitschrift für
Philadelphia, PA, 1979. Werkstofftechnik, 1975, 6, 269–280, 339–348.
5. Purslow, D., Some fundamental aspects of composite fractography. 30. Och, F., Schädigungsgrenze bei Faserverbundstrukturen. DGLR
Composites, 1981, 12, 241–247. Symp., 1976, pp. 133–178.
6. Rosen, B.W. and Dow, N.F., Mechanics of failure of fibrous com- 31. Owen, M.J. and Griffiths, J.R., Evaluation of biaxial stress failure
posites. In Fracture, An Advanced Treatise, Vol. 6, ed. H. Liebowitz, surfaces for a glass fabric reinforced polyester resin under static and
Academic Press, New York, 1972, pp. 612–674. fatigue loading. J. Mater. Sci. , 1978, 13, 1521–1537.
7. Wisnom, M.R., The effect of fibre misalignment on the compressive 32. Owen, M.J. and Rice, D.J., Biaxial strength behavior of glass-
strength of unidirectional carbon fibre/epoxy. Composites, 1990, 21, reinforced polyester resins. In Composite Materials: Testing and
403–407. Design, ASTM STP 787, American Society for Testing and
8. Hahn, H.T. and Williams, J.G., Compression failure mechanisms in Materials, Philadelphia, PA, 1981, pp. 124–144.
unidirectional composites. In Composite Materials: Testing and 33. Owen, M.J. and Rice, D.J., Biaxial strength behavior of glass fabric-
Design (7th Conf.), ASTM STP 893, American Society for Testing reinforced polyester resins. Composites, 1981, 12, 13–26.
and Materials, Philadelphia, PA, 1986, pp. 115–139. 34. Tsai, S.W., A survey of macroscopic failure criteria for composite
9. Franz, H.E., Microfractography of fibre reinforced composite mate- materials. Journal of Reinforced Plastics and Composites, 1984, 3,
rials. Practical Metallography, 1991, 28, 404–419. 40–62.
10. Bascom, A.C. and Gweon, S.Y., Fractography and failure mechan- 35. Rowlands, R.E., Strength theories and their experimental correc-
isms of carbon fiber-reinforced composite materials. In Fracto- tion. In Failure Mechanics of Composites, ed. G.C. Sih and A.M.
graphy and Failure Mechanisms of Polymers and Composites, Skudra, North-Holland, Amsterdam, 1985, pp. 71–125.
ed. A.C. Roulin-Moloney, Elsevier, London, 1988, pp. 351–386. 36. Cui, W.C., Wisnom, M.R. and Jones, M., A comparison of failure
11. Puck, A., Festigkeitsanalyse von faser-matrix-laminaten: modelle criteria to predict delamination of unidirectional glass/epoxy speci-
für die Praxis, Hanser, München, 1996. mens waisted through the thickness. Composites, 1992, 23, 158–
12. Guess, T.R., Biaxial testing of composite cylinders: experimental– 166.
theoretical comparison. Composites, 1980, 11, 139–148. 37. Jenkins, C.F., Materials of construction used in aircraft and aircraft
13. Puck, A., Ein Bruchkriterium gibt die Richtung an. Kunststoffe, engines. Report to the Great Britain Aeronautical Research Com-
1992, 82, 607–610. mittee, 1920.
14. Pipes, R.B. and Pagano, N.J., Interlaminar stresses in composite 38. Petit, P.H. and Waddoups, M.E., A method to predict the nonlinear
laminates under uniform axial extension. J. Compos. Mater., behavior of laminated composites. J. Compos. Mater., 1969, 3, 2–
1970, 4, 538–548. 19.
15. Pipes, R.B., Kaminski, B.E. and Pagano, N.J., Influence of the free 39. Azzi, V.D. and Tsai, S.W., Anisotropic strength of composites.
edge upon the strength of angle-ply laminates. In Analysis of the Experimental mechanics, 1965, 000, 283–288.
Test Methods for High Modulus Fibres and Composites, ASTM 40. Fan, W.X., On phenomenological anisotropic failure critera. Com-
STP 521, American Society for Testing and Materials, Philadelphia, posites Science and Technology, 1987, 28, 267–278.
PA, 1973, pp. 218–228. 41. Hoffman, O., The brittle strength for anisotropic materials.
16. Schulte, K. and Stichcomb, W.W., Damage mechanisms— J. Compos. Mater., 1967, 1, 200–206.
including edge effects—in carbon fibre-reinforced composite 42. Tsai, S.W. and Wu, E.M., A general theory of strength for aniso-
materials. In Application of Fracture Mechanics to Composite tropic materials. J. Compos. Mater., 1971, 5, 58–80.
Materials, ed. K. Friedrich, Elsevier, Amsterdam, 1989, pp. 273– 43. Tsai, S.W. and Hahn, H.T., Introduction to Composite Materials,
325. Technomic, Westport, 1980.
17. Whitney, J.M. and Browning, C.E., Free-edge delamination of ten- 44. Narayanaswam, R. and Adelman, H.M., Evaluation of the tensor
sile coupons. J. Compos. Mater., 1972, 6, 300–303. polynomial and Hoffman strength theories for composite materials.
18. Pagano, N.J. and Pipes, R.B., The influence of stacking sequence on J. Compos. Mater., 1977, 11, 366–377.
laminate strength. J. Compos. Mater., 1971, 5, 50–57. 45. Pipes, B.R. and Cole, B.W., On the off-axis strength test for aniso-
19. Cook, J. and Gordon, J.E., A mechanism for the control of crack tropic materials. J. Compos. Mater., 1973, 7, 246–256.
propagation in all-brittle systems. Proc. R. Soc. London, Ser. A, 46. Puppo, A.H. and Evensen, H.A., Strength of anisotropic materials
1964, 282, 508–520. under combined stresses. AIAA J., 1972, 10, 468–474.
20. Swanson, S.R., Biaxial failure criteria for toughened resin carbon/ 47. Hashin, Z. and Rotem, A., A fatigue failure criterion for fiber
epoxy laminates. Experimental Mechanics, 1988, 28, 1075–1083. reinforced materials. J. Compos. Mater., 1973, 7, 448–464.
21. O’Brien, T.K. and Salpekar, S.A., Scale effects on the transverse 48. Hashin, Z., Failure criteria for unidirectional fiber composites.
tensile strength of graphite/epoxy composites. In Composite Journal of Applied Mechanics, 1980, 47, 329–334.
Materials: Testing and Design (11th Conf.), ASTM STP 1206, 49. Rosen, B.W., Mechanics of composite strengthening in fiber com-
American Society for Testing and Materials, Philadelphia, PA, posite materials. American Society for Metals, 1965, 000, 37–76.
1992, pp. 23–52. 50. Wu, E.M. and Scheublein, J.K., Laminate strength—a direct char-
22. Nairn, J.A. and Hu, S., Matrix microcracking. In Damage acterization procedure. In Composite Materials: Testing and Design
Mechanics of Composite Materials, ed. R. Talreja, Elsevier, (3rd Conf.), ASTM STP 546, American Society for Testing and
Amsterdam, 1994, pp. 187–243. Materials, Philadelphia, PA, 1973, pp. 188–206.
23. Clements, L.L. and Lee, P.R., Influence of quality control variables 51. Tennyson, R.C., MacDonald, D. and Nanyaro, A.P., Evaluation of
on failure of graphite/epoxy under extreme moisture conditions. the cubic polynomial strength criterion on the failure analysis for
In Composites for Extreme Environments, ASTM STP 768, composite materials. J. Compos. Mater., 1978, 12, 63–75.
ed. N.R. Adsit, American Society for Testing and Materials, 52. Tennyson, R.C., Nanyaro, A.P. and Wharram, G.E., Application of
Philadelphia, PA, 1982, pp. 161–171. the cubic polynomial strength criterion on the failure analysis for
24. Whitney, J.M., Free edge effects in the characterization of composite materials. J. Compos. Mater. Suppl. , 1980, 14, 28–41.

885
A review of the biaxial strength of fibre-reinforced plastics: H. Thom

53. Tennyson, R.C., Wharram, G.E. and Elliot, G., Application of the 75. Arcan, M., Hashin, Z. and Voloshin, A., A method to produce uni-
cubic strength criterion to the failure analysis of composite lami- form plane-stress states with applications to fiber-reinforced
nates. In Fracture of Composite Materials, ed. G.C. Sih and V.P. materials. In Composite Materials: Testing and Design (6th Conf.),
Tamuzs, Martinus Nijhoff, The Hague, 1982, pp. 53–66. ASTM STP 787, American Society for Testing and Materials,
54. Chang, F.K., Scott, R.A. and Springer, G.S., Failure of composite Philadelphia, PA, 1981, pp. 34–49.
laminates containing pin loaded holes—method of solution. 76. Marloff, R.H., Finite element analysis of biaxial stress test specimen
J. Compos. Mater., 1984, 18, 255–278. for graphite/epoxy and glass fabric/epoxy composites. In Composite
55. Chang, F.K., Scott, R.A. and Springer, G.S., Failure strength of Materials: Testing and Design (6th Conf.), ASTM STP 787,
nonlinearly elastic composite laminates containing a pin loaded American Society for Testing and Materials, Philadelphia, PA,
hole. J. Compos. Mater., 1984, 18, 464–477. 1981, pp. 34-49.
56. Chang, F.K. and Chang, K.Y., A progressive damage model for 77. Rizzo, R.R. and Vicario, A.A., A finite element analysis of lami-
laminated composites containing a stress concentrations. nated anisotropic tubes. J. Compos. Mater., 1970, 4, 344–359.
J. Compos. Mater., 1987, 21, 834–855. 78. Pagano, N.J. and Whitney, J.M., Geometric design of composite
57. Yamada, S.E. and Sun, C.T., Analysis of laminate strength and its cylindrical characterization specimens. J. Compos. Mater., 1970,
distribution. J. Compos. Mater., 1978, 12, 275–284. 4, 360–378.
58. Sandhu, R.S., Nonlinear response of unidirectional and angle-ply 79. Sherrer, R.E., Filament-wound cylinders with axial-symmetric
laminates. AIAA Paper no. 74-380, presented at 15th AIAA-ASME loads. J. Compos. Mater., 1967, 1, 344–355.
Structural Dynamics and Materials Conf., Las Vegas, NV, 1974. 80. Vicario, A.A. and Rizzo, R.R., Effect of length on laminated thin
59. Sandhu, R.S., Ultimate strength analysis of symmetric laminates, tubes under combined loading. J. Compos. Mater., 1970, 4, 273–277.
AFFDL-TR-73-137, 1974. 81. Whitney, J.M., Pagano, N.J. and Pipes, R.B., Design and fabrication
60. Hahn, H.T. and Tsai, S.W., Nonlinear elastic behavior of unidirec- of tubular specimens for composite characterization. In Composite
tional composite laminae. J. Compos. Mater., 1973, 7, 102–118. Materials: Testing and Design (2nd Conf.), ASTM STP 497,
61. Hart-Smith, L.J., A strain-based maximum-strain-stress failure cri- American Society for Testing and Materials, Philadelphia, PA,
terion for fibrous composites. Douglas Paper 8376, 31st AIAA/ 1971, pp. 52–67.
ASME/ASCE/AHS/ASC Structures, Structural Dynamics and 82. Widera, O.E. and Chung, S.W., A theory for non-homogeneous,
Materials Conf., Long Beach, CA, USA, 2–4 April 1990. In Proc. anisotropic cylindrical shells. J. Compos. Mater., 1972, 6, 14–30.
CP 902 Part 2, pp. 714–722. 83. Choo, V.K.S. and Hull, D., Influence of radial compressive stress
62. Hart-Smith, L.J., The truncated maximum stran composite failure owing to pressure on the failure modes of composite tube speci-
model. Composites, 1993, 24, 587–591. mens. J. Compos. Mater., 1983, 17, 344–356.
63. Cuntze, R.G., Bruchtyp-Festigkeitskriterien, formuliert mit 84. Timoshenko, S.P., Strength of Materials, Van Nostrand, Princeton,
Invarianten, die die Werkstoffsymmetrie des jeweiligen iso-/aniso- NJ, 1956.
tropen Werkstoffs beinhalten. Technischer Bericht, 04.12.95, 85. Toombes, G.R., Swanson, S.R. and Cairns, D.S., Biaxial testing of
MAN-Technologie Informationszentrum Augsburg, in Workshop composite tubes. Experimental Mechanics, 1985, 25, 186–192.
Comp. Forsch. in der Mech., 1995. 86. Swanson, S.R. and Christoforou, A.P., Response of quasi-isotropic
64. Cuntze, R.G., Personal note to author, 1996. carbon/epoxy laminates to biaxial stress. J. Compos. Mater., 1986,
65. Zum verbesserten Festigkeitsnachweis von Bauteilen aus Faser- 20, 457–471.
Kunststoff-Verbunden, BMBF Förderkennzeichen 03N8002. 87. Nagy, A. and Lindholm, U.S., Hydraulic grip systems for composite
66. Matthews, F.L., Mechanical testing and the relevance of standards. tube specimens, AFML-TR-73-239, 1973.
In Proc. I Mech E Conf.: Design of Composite Materials, 1989, 88. Lindholm, U.S., Nagy, A., Yeakley, L.M. and Ko, W.L., Design of a
pp. 13–22. test machine for biaxial testing of composite-laminate cylinders,
67. Matthews, F.L. and Haeberle, J.G., The compressive mechanical AFFDL-TR-75-83, 1975.
properties of fibre-reinforced plastics. Imperial College Report 89. Krempl, E. and Niu, T.-M., Graphite/epoxy [ 6 45] s tubes: their
and RAE Agreement 2037/235 XR/MAT, 1990 (Imperial College, static axial and shear properties and their fatigue behavior under
London). completely reversed load controlled loading. J. Compos. Mater.,
68. Curtis, P.T., Gates, J. and Molyneux, C.G., An improved engineer- 1982, 16, 172–187.
ing test method for measurement of the compressive strength of 90. Hess, T.E., Stability of cylindrical shells under compressive load-
unidirectional carbon fibre composites. Composites, 1991, 22, ing. American Rocket Society Journal, 1961, 31, 237.
363–368. 91. Cheng, S. and Ho, B.P.C., Stability of heterogeneous cylindrical
69. Hart-Smith, L.J., Backing out equivalent unidirectional lamina shells under combined loading. AIAA J., 1963, 1, 892–898.
strengths from tests on cross-plied laminates. Douglas Paper 92. Ho, B.P.C. and Cheng, S., Some problems in stability of hetero-
MDC 91K0078, presented at 37th Int. SAMPE Symp. and geneous aelotropic cylindrical shells under combined loading. AIAA
Exhibition, Anaheim, CA, 1992. J., 1963, 1, 1603–1607.
70. Sun, C.T. and Chung, I., An oblique end-tab design for testing off- 93. Whitney, J.M. and Sun, C.T., Buckling of composite cylindrical
axis composite specimens. Composites, 1993, 24, 619–623. characterization specimens. J. Compos. Mater., 1975, 9, 138–148.
71. Bert, C.W., Mayberry, B.L. and Ray, J.D., Behavior of fiber- 94. Ley, R.P., Gürdal, Z. and Johnson, E.R., Buckling of imperfect,
reinforced plastic laminates under biaxial loading. In Composite anisotropic, ring-stiffened cylinders under combined loads. AIAA
Materials: Testing and Design (1st Conf.), ASTM STP 460, J., 1994, 32, 1302–1309.
American Society for Testing and Materials, Philadelphia, PA, 95. Wempner, G.A. and Clodfelter, G.A., Mechanical behavior of
1969, pp. 362–380. filament-wound composite tubes. J. Compos. Mater., 1980, 14,
72. Cole, B.W. and Pipes, R.B., Filamentary composite laminates sub- 260–268.
jected to biaxial stress fields, AFFDL-TR-73-115, 1974. 96. Byung, S.M. and Lehnhoff, T.F., Polymeric composite tube fabrica-
73. Boehler, J.P., Demmerle, S. and Koss, S., A new direct biaxial tion. Journal of Engineering Materials and Technology, 1995, 117,
testing machine for anisotropic materials. Experimental Mechanics, 235–237.
1994, 34, 1–9. 97. Knight, C.E., Residual stress and strength loss in filament-wound
74. Daniel, I.M., Methods of testing composite materials. In Failure composites. In Composite Materials: Testing and Design (8th
Mechanics of Composites, ed. G.C. Sih and A.M. Skudra, North- Conf.), ASTM STP 972, American Society for Testing and
Holland, Amsterdam, 1985, pp. 277–373. Materials, Philadelphia, PA, 1986, pp. 413–422.

886