Composites: Part B 43 (2012) 280–289

Contents lists available at ScienceDirect

Composites: Part B
journal homepage: www.elsevier.com/locate/compositesb

A numerical assessment on two-dimensional failure criteria for composite

layered structures
P. Nali ⇑, E. Carrera
Department of Aeronautics and Space Engineering, Politecnico di Torino, Turin, Italy

a r t i c l e i n f o a b s t r a c t

Article history: This paper is devoted to the most popular failure criteria which are employed in the design and analysis
Received 15 March 2011 of anisotropic materials/layered plates. In-plane stress (or strain) states are considered and both classical
Received in revised form 16 June 2011 and advanced criteria are illustrated and compared. The advantages and disadvantages of different
Accepted 21 June 2011
approaches are discussed through FEM numerical results, by referring to composite material data cur-
Available online 25 June 2011
rently employed. Layer-wise plate models are adopted to provide quasi three-dimensional stress field
and to limit the error of modeling. Failure envelopes emphasizing the in-plane shear influence in the case
Keywords:
of biaxial traction/compression are proposed. The maximum strain, maximum stress, Tsai–Wu, Tsai–Hill,
A. Layered structures
C. Damage mechanics
Hashin and LaRC03 failure criteria are described and compared for various benchmark problems.
C. Laminate mechanics Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction valid today: material properties are usually considered at the mac-
roscale level through material homogenization, while microscale
It is an established fact that further advances in the use of lam- investigations require modeling fibers/matrix interfaces, leading
inated composites are subordinate to a better understanding of to extremely heavy computations. It follows that the calculation
their failure mechanisms. On the other hand, the analysis and sim- of the failure initiation in laminated composite structures is a very
ulation of the failure of composite laminated structures are quite challenging task and it is far from being totally solved.
cumbersome tasks. As anticipated by Hoffman [1], who accounted Extensive use of experiments is made in most advanced com-
for differences in tensile and compressive strength, the failure posite constructions. Quite expensive test campaigns have been
mechanisms are very different from those of traditional metallic conducted for composite made parts of aircrafts such as the A350
structures. The combination of various interfaces (fibers, matrix, or B787. Due to intrinsic tailoring, anisotropy and lay-out, the
layers) at a macro scale level requires a local dedicated analysis number of necessary experiments has increased remarkably with
to establish the initiation of failure mechanisms of a fiber, a crack respect to traditional metallic structures. Uncertainties could play
in the matrix or a delamination between two different layers. a predominant role in both experiments and analyses. In fact, many
As written by Tsai and Hahn [2], the ‘‘analysis of properties of additional sources of uncertainties arise, such as angle of fiber ori-
composite laminates is in general performed at three different levels entation, ply thickness, and mechanical/thermal/electric properties
of sophistication. At the first level, which is commonly called the of the lamina (fiber and matrix). It is also confirmed that the full
micromechanics approach, lamina properties are predicted in terms utilization of composites in the transportation industry is limited
of constituent properties. At the intermediate level, the lamina proper- mostly by matrix and/or interface failure rather than by fiber fail-
ties are determined experimentally and then used to predict the ure [2].
laminate properties in conjunction with an appropriate macromechanics Due to the above mentioned complicating effects, various Reg-
analysis such as the laminated plate theory. The last level is reserved ulation Authorities have introduced strict safety rules and regula-
for direct characterization of experimentally-determined laminate tions for the design and use of composite materials. Although
properties. For most current structural applications, a combination thousands of contributions are already available in literature, it ap-
of the last two levels is employed; that is design starts at the interme- pears mandatory that years to come should be devoted to a better
diate level. . .’’. In engineering practice, the above concept is still understanding of the above mentioned topics.
In the works by Hashin and Rotem [3], for the first time the fail-
⇑ Corresponding author. Address: Department of Aeronautics and Space Engi- ure of laminated composites has been attributed to different phys-
neering, Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Torino, Italy. ical phenomena: fiber-dominated and matrix-dominated failure
Tel.: +39 011 546 6869; fax: +39 011 564 6899. modes [3,4]. This approach has been further extended by Puck
E-mail address: [email protected] (P. Nali). and Schuermann [5,6] by addressing matrix compression failure

1359-8368/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesb.2011.06.018
 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289 281

with a model based on the Mohr–Coulomb criterion. The World stresses (or strain) components. Maximum strain and maximum
Wide Failure Exercise (WWFE), conceived and conducted by Soden stress criteria [21] belong to this group. In the remaining criteria,
and co-workers [7–10] and Puck and Schuermann [5,6] provided the failure in one direction may be sensitive to loads along other
an exhaustive assessment of the theoretical methods for predicting directions (including shear). This last group can be divided into
material initial failure in Fiber Reinforced Polymer composites the following two subgroups.
(FRP). The WWFE indicated the Puck failure criterion as one of
the most effective, the predicted failure envelopes being in good  Criteria proposing one single inequality to define the failure
correlation with the test results [6]. In any case, it should be under- envelope.
lined that, even when analyzing simple laminates that have been  Criteria proposing a combination of interactive and non-inter-
extensively studied and tested, the predictions of most theories active conditions.
differ significantly from the experimental observations [9]. After
WWFE, NASA Langley Research Center revisited existing failure The Hoffman [1], Tsai–Wu [22], Liu–Tsai [23] and Tsai–Hill
theories in order to identify the most accurate models and, when [24,25] are quadratic criteria and they belong to the first group,
possible, introduce some enhancements [11,12]. The result of this while the Hashin and Rotem [3], Hashin [4], Puck and Schuermann
activity is the criterion named LaRC03 [11] which can be employed [5,6] and LaRC03 [11] criteria pertain to the second. The latter are
in order to predict the initial failure of FRP laminated panels with considered phenomenological approaches since they aim to distin-
in-plane stress states. LaRC03 is inspired by Puck’s fundamental guish between different failure mechanisms. For example, accord-
assumption of a fragile fracture for the matrix failure in compres- ing to the Puck approach, the failure stress in the fiber direction is
sion and consequently implements the action plane concept [5,6], independent of loads along other directions, while shear and trans-
according to the Mohr–Coulomb theory. Concerning tensile matrix verse failure stresses take into account interactions [11].
cracking, LaRC03 is associated with Dvorak’s fracture mechanics For a fixed case-study, a Failure Index (FI) corresponds to each
approach [13], making use of the energy release rates associated failure criteria. An FI exceeding the unitary value means that fail-
with intralaminar crack propagation. Several assessments con- ure occurs, according to the applied criterion. In the following sec-
firmed the good agreement between the LaRC03 and Puck results tions, the analytical definition of the most popular two-
[11]. Subsequently, the LaRC04 criterion [14–18] was defined by dimensional failure criteria is proposed.
extending the approach to three-dimensional stress states. In pa-
per [19] the reader can find an exhaustive and detailed critical re- 2.1. Maximum stress and maximum strain criteria
view of the state of the art in composite failure theories, together
with a consistent number of references on this subject. Maximum stress criterion does not consider any interaction be-
Three-dimensional failure criteria are not included in the tween different stress components. The failure occurs when the
assessment proposed hereafter. This is due to the fact that, dealing stress in any direction exceeds the stress associated to the failure.
with layered composite panels, significant result enhancements If the cartesian material reference system is considered and fibers
are rarely obtained trough the inclusion of the three-dimensional are taken parallel to axis 1, the conditions for failure can be stated
stress state in the model. Moreover, for the sake of consistency, as:
the authors aim to propose assessments and comparisons among
failure criteria belonging to the same class. r11 P X T or r11 6 X C ; ð1Þ
In the next section, a brief overview of most popular two- r22 P Y T or r22 6 Y C ; ð2Þ
dimensional failure criteria for anisotropic materials, including
s12 P SL ; ð3Þ
the LaRC03, is given. The accuracy of various approaches is quanti-
fied by referring to experimental data. Additional comparisons are where r11, r22 and s12 are the two in-plane normal stresses and the
proposed in order to identify some load cases and stacking se- in-plane shear stress, in the same order. XT and XC are respectively
quences for which different failure criteria give rather distinct pre- the material strength in the fiber direction under tension and com-
dictions. The consequent scatter of results is commented on by pression (longitudinal tensile and compressive strengths). YT and YC
referring to the assumptions made in each criterion and suggesting are respectively the material strength normal to the fiber direction
the most reasonable approach. under tension and compression (transverse tensile and compressive
strengths). SL is the longitudinal shear strength.
The FI can be obtained as:
2. Brief overview of failure criteria  
r11 r22 s12
FI ¼ MAX ; ; L
; ð4Þ
The homogeneous material strengths parameters are typically
X Y S
established by test prediction in principal material directions. being X = XT or X = XC if r11 P 0 or r11 < 0, respectively and Y = YT
Plates and panels are easily subjected to in-plane multiaxial stress or Y = YC if r22 P 0 or r22 < 0, in the same order.
states during service and, depending on the stacking sequence, the The maximum strain criterion is obtained following the same
in-plane shear stress may play a significant role in the failure. A approach as for the maximum stress criterion, but considering cor-
number of failure criteria have been formulated in order to predict responding strains in the conditions for failure:
failure loads under arbitrary stress states. According to the classi-
fication proposed in Ref. [20], they can be grouped into two main 11 P T11 or 11 6 C11 ; ð5Þ
groups. 22 P T22 or 22 6 C22 ; ð6Þ
L
c12 P c 12 ; ð7Þ
 Failure criteria neglecting interactions between different stress
components. where 11, 22 and c12 are the two in-plane normal strains and the
 Failure criteria considering interactions between different in-plane shear strain, in the same order. T11 and C11 are respectively
stress components. the longitudinal tensile and compressive failure strain. T22 and C22
are respectively the transverse tensile and compressive failure
Criteria belonging to the first group are the simplest and they strain. cL12 is the in-plane shear failure strain.
usually propose one inequality for each one of the three in-plane In this case the FI is the following:
282 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289

 
11 22 c12
FI ¼ MAX ; ; ; ð8Þ
11 22 cL12

being  ¼ T11 or  ¼ C11 if 11 P 0 or 11 < 0, respectively and

 ¼ T22 or  ¼ C22 if 22 P 0 or 22 < 0, in the same order.
If the load acting on the structure is indicated as L, according to
maximum stress or maximum strain criterion, the corresponding
failure load LF can be obtained as:

LF ¼ jL; ð9Þ

where j = 1/FI.
Fig. 1 gives an example of failure envelope of maximum stress
and maximum strain criteria. Shear stress/strain interactions are
neglected in both approaches [21]. It can be noticed that the enve-
lope related to the maximum stress criterion can be obtained from
the plot related to the maximum strain criterion through the linear Fig. 2. Tsai–Hill envelopes.
transformation expressed by Hooke’s matrix. Nevertheless, since
failure stress and strain parameters are both obtained by testing,
this mathematical relation could not be rigorously satisfied in 2.3. Tsai–Wu criterion
practice.
The Tsai–Wu criterion was not derived from a physical basis,
but it was formulated in order to fit experimental results [22,23].
2.2. Tsai–Hill criterion It is an interactive approach considering in-plane shear stress ef-
fects. The failure condition is expressed by the following
The Tsai–Hill criterion was formulated by referring to distor- inequality:
tional energy [24,25] and is thus an interactive criterion that takes
into account the effect of the in-plane shear stress. A11 r211 þ 2A12 r11 r22 þ A22 r222 þ A66 s212 þ B1 r11 þ B2 r22 P 1; ð12Þ
The condition for failure is given by the following inequality:
where
r 2 r 2  2
11 r11 r22 22 s12 1
 þ þ P 1; ð10Þ A11 ¼ ;
X X 2 Y S L XT XC
1
A12 ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ;
where the failure parameters X and Y depend on the considered 2 XT XC Y T Y C
quadrant of the coordinate plane, being X = XT or X = XC if r11 P 0 1
or r11 < 0, respectively and Y = YT or Y = YC if r22 P 0 or r22 < 0, in A22 ¼ T C ;
Y Y
the same order. Tsai–Hill FI is given by the expression in Eq. (10). 1
The failure load corresponding to the Tsai–Hill criterion can be A66 ¼ L L ;
SS
computed as:
1 1
B1 ¼ T  C ;
LF ¼ jL; ð11Þ X X
1 1
pffiffiffiffi B2 ¼ T  C :
where j ¼ 1= FI. Y Y
An example of corresponding failure envelopes with varying s12 The expression given for A12 can be found in the work of Reddy [21].
is given in Fig. 2, where it can be noticed that the semimajor and Nevertheless, the A12 coefficient is often obtained/corrected by
semiminor axes of the four ellipses composing the plots are re- biaxial tests of laminae [25,23]. The Tsai–Wu FI is given by the
duced by the in-plane shear stress. expression in Eq. (12).
The failure load corresponding to the Tsai–Wu criterion is:

LF ¼ jL; ð13Þ

with
0 1
B 2 C
j ¼ MINB
@     ffiC
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A:
2 2
B1 r11  B2 r22 4A11 þ B1 r11 þ 2ð4A12 þ B1 B2 Þr11 r22 þ 4A22 þ B2 r22 þ 4A66 s12
2 2 2

An example of resulting failure envelope with varying s12 is given in

Fig. 3, where it is noticeable that the semimajor and semiminor axes
of the corresponding ellipse are reduced by the in-plane shear stress
effect.

2.4. Hashin criterion

The Hashin criterion proposes a combination of four interactive

and non-interactive conditions in order to distinguish between
Fig. 1. Envelopes of various failure criteria without in-plane shear. matrix or fiber failure caused by tension or compression. It was
 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289 283

Fig. 4. Angle of the fracture plane of a unidirectional lamina subjected to transverse

compression and in-plane shear.

Two examples of Hashin failure envelopes with varying s12 are

given in Figs. 5 and 6, where it can be noted that the fiber failure for
compression is not conditioned by the in-plane shear stress. Enve-
lopes in Fig. 5 are all obtained with a = 0° while a is set at 53° in
Fig. 6. According to Puck [11], the angle of the fracture plane a in-
Fig. 3. Tsai–Wu envelopes. creases as the in-plane shear decreases with respect to transverse
compression. That is, a more realistic collection of envelopes with
varying s could be obtained by plotting in one single chart the
first defined in work [3] and subsequently revised in work [4]. A envelopes pertaining to different a(r22, s12), with a taking a value
further version of this criterion was proposed by Sun et al. [26], between 0° and 53°. The function a(r22, s12) should be defined by
according to some empirical modifications. The Hashin criterion testing. Figs. 9 and 10 give a qualitative idea of the Hashin and
is interactive, the conditions for failure being given by the follow- LaRC03 envelopes with varying s12 and a: the two parameters
ing inequalities [4]. are assumed to vary with an hypothetical linear law.

Matrix failure for compression ðr22 P 0Þ : 2.5. LaRC03 criterion

 2  2
r22 s12
þ P 1: ð14Þ
YT SL The LaRC03 failure criterion was developed after the WWFE at
Langley Research Center [11,12] and suggests a combination of five
Matrix failure for tension ðr22 < 0Þ : interactive conditions in order to indicate a crack initiation in the
2 !2 3 matrix or a failure due to fiber compression. One non-interactive
 2 C  2
r22 4 Y 5 r22 þ s12 P 1: inequality is suggested for tensional fiber failure. A summary of
þ  1
2ST 2ST YC SL the LaRC03 criterion is proposed in the following.

 2  2
r11 s12 Matrix cracking for tension ðr22 P 0Þ :
Fiber failure for tension ðr11 P 0Þ : þ P 1: ð16Þ
XT SL ! !2 !2
r22 r22 s12
ð1  gÞ þg þ P 1; ð20Þ
r11 Y Tis Y Tis SLis
Fiber failure for compression ðr11 < 0Þ :  C
P 1: ð17Þ
X
T where g ¼ GGIIcIc ðLÞ is the toughness ratio, and (L) stands for the Longi-
S is the transverse shear strength which is very difficult to measure ðLÞ
tudinal direction. GIc and GIIc are the components of the fracture
experimentally. Work [11] suggests the following analytical
toughness G pertaining to the two modes of energy release during
relation:
  the crack propagation. Although GIc and GIIc can be measured from
cosðaÞ standard fracture mechanics tests, a formulation for calculating
ST ¼ Y C cosðaÞ sinðaÞ þ ; ð18Þ
tanð2aÞ these coefficients was proposed by Dvorak and Laws [13]. A more
simple formula can be used for thin embedded plies [11]:
where a is the angle of the fracture plane introduced by Puck and
Schuermann [5,6] and subsequently adopted in the LaRC03 criterion
[11]. As written in work [11], ‘‘Puck determined that the matrix
failures dominated by in-plane shear occur in a plane that is normal
to the ply and parallel to the fibers (a = 0°). For increasing the amounts
of transverse compression, the angle of the fracture plane a changes to
about 40°, and increases with compression to 53° ± 2° for pure trans-
verse compression’’. The angle of the fracture plane a is illustrated
in Fig. 4. Hashin’s FI is given by the highest value among the expres-
sions in Eqs. (14)–(37).
The failure load corresponding to the Hashin’s criterion is given
by:

LF ¼ jL; ð19Þ

where, in the case of linear analysis, j can be found by scaling the

results computed with load L by a varying factor jtry. After a num-
ber of iterations, if FI ’ 1 then j = jtry. This procedure is indepen-
dent of the failure criterion employed and is also used for the
LaRC03 criterion of Section 2.5 in the present study. Fig. 5. Hashin envelopes with a = 0°.
284 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289

The angle of the fracture plane a stays between 0° and 90°. smL
eff
and smT
eff can be obtained through Mohr’s circle [11]:

smT m T
eff ¼ hr22 cosðaÞ½sinðaÞ  g cosðaÞi; ð30Þ

m 
seff ¼ cosðaÞ s12 þ g r22 cosðaÞ ;
mL L m
ð31Þ

with hxi = x if x P 0, while otherwise hxi = 0. Internal material fric-

tion coefficients gT and gL are obtained by testing or calculated
through the following formulae:
1
gT ¼ ; ð32Þ
tanð2a0 Þ
SLis cosða0 Þ
gL ¼  : ð33Þ
Y C cos2 ða0 Þ
Usually, a0 = 53° from test data [5,6,11]. As suggested in work
[11], the fracture angle a can be obtained by searching for the max-
Fig. 6. Hashin envelopes with a = 53°. imum value of the expression in Eq. (34), within a loop over the
range of possible fracture angles between 0° and a0.
!2 Matrix cracking for compression (r22 < 0 and r11 P YC):
K0 Y T 2 !2 !2
g ¼ 1:12 22 ; ð21Þ
K044 SL sTeff sTeff
  þ P 1; ð34Þ
m2 ST SLis
with K022 ¼ 2 E12  E211 and K044 ¼ G112 .
T L
Y is and Sis of Eq. (20) are the in situ strengths [11] and for thin where the effective shear stresses sTeff and sLeff are:
embedded plies they can be calculated as:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sTeff ¼ hr22 cosðaÞ½sinðaÞ  gT cosðaÞi; ð35Þ
8GIc ðLÞ sLeff ¼ hcosðaÞ½js12 j þ gL r22 cosðaÞi: ð36Þ
Y Tis ¼ ; ð22Þ
ptK022
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Failure for fiber tension (r11 P 0):
8GIIc ðLÞ
SLis ¼ ; ð23Þ 11
ptK044 T11
P 1: ð37Þ

where t is the ply thickness. Failure for fiber compression (r22 < 0):
In the case of thick plates, the following formulae can be used: * +
pffiffiffi sm þ gL rm
12 22
Y Tis ¼ 1:12 2Y T ; ð24Þ P 1: ð38Þ
pffiffiffi SLis
SLis ¼ 2SL : ð25Þ
Failure for fiber compression (r22 P 0):
C
Matrix cracking for compression (r22 < 0 and r11 < Y ): ! !2 !2
!2 !2 rm22 rm22 sm12
s mT
s mT ð1  gÞ þg þ P 1: ð39Þ
eff
þ
eff
P 1: ð26Þ Y Tis Y Tis SLis
T
S SLis
The LaRC03 FI is given by the highest value among the expressions
s mL
and smT
eff eff are the effective shear stresses for a generic angle of in Eqs. (20), (26), (34), (37)–(39).
fracture plane a, in the misalignment coordinate frame m. The coor- The failure load corresponding to the LaRC03 criterion can be
dinate frame m is obtained by rotating the material reference sys- obtained by following the procedure described in Section 2.4.
tem of the angle u. The latter describes the fiber kinking Two examples of LaRC03 failure envelopes with varying s12 are
phenomenon, that is the imperfections in fiber alignment under given in Figs. 7 and 8. Envelopes in Fig. 7 are all obtained with
load. The in-plane stresses in the misalignment coordinate frame a = 0° while a is set at 53° in Fig. 8.
can be calculated as:

rm11 ¼ cos2 ðuÞr11 þ sin2 ðuÞr22 þ sinð2uÞs12 ; 3. Numerical results

rm22 ¼ sin2 ðuÞr11 þ cos2 ðuÞr22  sinð2uÞs12 ; ð27Þ As suggested by Tsai and Hahn [2], judicious linear analysis
sm12 ¼ cosðuÞsinðuÞr11 þ cosðuÞsinðuÞr22 þ cosð2uÞs12 ; seems preferable to complicated nonlinear analysis when predict-
ing the initial failure of layered composite panels. In accordance
with
with this, linear static analyses are addressed in this work and this
js12 j þ ðG12  X C ÞuC allow us to deal with a minor number of varying parameters,
u¼ ; ð28Þ focusing our attention on the peculiarities of the failure criteria
G12 þ r11  r22
considered. Problems are solved in MUL2 academic code through
and where the total misalignment angle for the case of axial com- the Finite Element Method (FEM). The penalty technique is applied
pression loading uC is: in order to impose the boundary conditions to the FEM model.
0 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
L  L 1
S S
B1  1  4 XisC þ gL XisC C 3.1. Considerations about failure envelopes
uC ¼ tan1 B
@ L
S
 C:
A ð29Þ
2 XisC þ gL The failure envelopes presented in this paper are obtained by
seeking the combinations of in-plane stresses leading to unitary
gL is introduced below through Eq. (33). FI. For this purpose, a dedicated routine able to process the
 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289 285

Table 1
AS4/3502 material properties.

Longitudinal Young’s modulus, E1 (GPa) 127.6

Transverse Young’s moduli, E2,3 (GPa) 11.3
Shear modulus, G12 (GPa) 6.0
Shear moduli, G13,23 (GPa) 6.0, 1.8
Poisson’s ratios, m12, m13, m23 (–) 0.3, 0.3, 0.36

Table 2
AS4/3502 failure coefficients.

XT (MPa) 1730
XC (MPa) 1045
YT (MPa) 66.5
YC (MPa) 255
SL (MPa) 95.1
T11 (–) 0.0138
Fig. 7. LaRC03 envelopes with a = 0°. C11 (–) 0.01175
T22 (–) 0.00436
C22 (–) 0.002
cL12 (–) 0.002

lopes may be strongly sensitive to the approximations made when

evaluating some physical parameters.
Figs. 2, 3, 5, 6, 7, 8, 9, 10 give the variation of the interacting cri-
teria envelopes for various levels of s12. Figs. 2 and 3 respectively
show that both for the Tsai–Hill and Tsai–Wu criteria, the in-plane
shear stress resizes the envelope with respect to the reference sys-
tem’s origin. Concerning the Hashing criterion, Figs. 5 and 6 show
that resizing occurs with respect to a point located along the enve-
lope’s edge pertaining to the fiber compression. Figs. 7 and 8 indi-
cate that, according to the LaRC03, the center of resizing is a point
along the envelope’s edge related to fiber failure due to traction.
Fig. 11 provides some comparisons among envelopes obtained
with various criteria and some WWFE experimental results. The
Fig. 8. LaRC03 envelopes with a = 53°. material is the composite unidirectional E-Glass/LY556 from Soden
et al. [7,8,14]: E1 = 53.48, E2 = 17.7, G12 = 5.83, all in GPa and
SL = 66.5, YT = 37.5, YC = 130.3, all in MPa. a0 corresponds to 53°. A
rather bad agreement can be observed among the criteria. In addi-
material failure coefficients according to the criteria described in
tion, experimental results are significantly scattered, which is quite
Section 2 was implemented in MUL2 code. The envelopes are plot-
common when dealing with composites. The LaRC03 envelope
ted point by point without implying any FEM computation: s12, r11
seems to match some experimental points properly, while the
and r22 are taken as independent variables. For a fixed value of s12,
maximum stress criterion looks inappropriate if the in-plane shear
an adequate number of (r11, r22) points are considered in order to
is low. In any case, test results should be interpreted from a statis-
cover the in-plane normal stress domain. For each of these points,
tical view point. A higher number of experimental points would
the FI corresponding to a selected criterion is computed. A dot is
lead to the formulation of more resolute comments about this
drawn in the chart only if FI ’ 1. In this way, after running the rou-
figure.
tine, the relevant failure envelope is obtained. It is also possible to
superimpose multiple plots, by varying the values of s12 and/or
selecting different failure criteria.
A graphite-epoxy fiber-reinforced ply is considered in the fol-
lowing. The material is AS4/3502 and the elastic and failure prop-
erties are listed in Tables 1 and 2, respectively. Failure envelopes in
Fig. 1 show that a failure analysis can lead to very different results,
depending on the criterion applied. For example, in the case of
biaxial compression, a large region of load cases not leading to fail-
ure, according to Tsai–Wu, would imply a FI P 1 by referring to
other failure criteria. In the case of s12 = 0, the Tsai–Hill criterion
can be seen as a variation of Maximum Stress or Hashin’s criterion,
with smoothed edges of the envelope. All the edges of the LaRC03
envelope, apart from the one corresponding to fiber failure for ten-
sion, depend on the in situ strengths of the embedded plies. In
other words, the ply strengths variation due to the presence of
layer interfaces is considered in LaRC03. Being aware that uncer-
tainties are unavoidable in the engineering practice, some of
curve-slope discontinuities which are present in the LaRC03 enve- Fig. 9. Hashin envelopes with varying s, a.
286 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289

Fig. 10. LaRC03 envelopes with varying s, a.

Fig. 11. Comparison among failure envelopes and WWFE test data for unidirec-
3.2. Convergence study tional composite E-Glass/LY556.

A [(+45°/45°)]s AS4/3502 laminated plate whose dimensions

are a = 445 mm, b = 381 mm and h = 0.13 mm is considered in the Reissner–Mindlin plate theory [27,28], that is, the in-plane dis-
following convergence study. The system is illustrated in placements respect a first-order equivalent-single-layer (ESL) kine-
Fig. 12a. A x, y, z cartesian reference system whose origin is at matical description (classical modeling). According to Layer-Wise
the plate central point is considered for the laminate. The x, y-axes (LW) kinematical descriptions, displacements of each layer are
are parallel to the a, b plate-edges, respectively. The lamination an- considered independently from the others and the continuity is
gle h is null for fiber parallel to the x-axis and is positive for coun- imposed at the layer-interfaces [29]. LWn indicates that a polyno-
terclockwise rotations. A uniform compressive loading of 1 MPa mial expansion of order n is considered for displacements of each
parallel to the x-axis is applied to the plate, which is clamped at layer. It is understood that LW models imply a higher number of
the opposite edge. Fig. 12b shows that the stress distribution is Degrees Of Freedoms (DOFs) compared to FSDT. FEs corresponding
not uniform on the surface of the plate: the r11 stress-map high- to various theories are implemented in the framework of the Car-
lights the lamination angle corresponding to the observed ply. rera Unified Formulation (CUF) [30]. The CUF permits the building
  of FE matrices according to a convenient hierarchical approach,
FEM results for the in-plane stress rxx 0; 2b ; 0 obtained by refer-
ring to various kinematical descriptions and different n  n regular favoring consistent comparisons among different theories. Table 3
meshes are listed confirms that the discretization error is significantly reduced even
 in the following convergence study. The stress with poor meshes. By comparing the rows of Table 3, it is notice-
state at point 0; 2b ; 0 is three-dimensional due to free-edge ef-
fects. As a consequence, this is one of significant positions to inves- able that the results are sensitive to the choice made for the plate
tigate in a first-ply failure analysis. Classical two-dimensional kinematical description. The results for the point investigated in
kinematical descriptions could lead to unrealistic results nearby Table 3 are illustrated in Fig. 13, where the whole thickness of
the edges of the plate. The present assessment emphasize the accu- the plate is considered and the 16  16 regular mesh is used.
racy of advanced layer-wise kinematical description with respect Fig. 13 shows that FSDT theory provides a constant through-the-
to classical modeling, when a free-edge point is considered. The thickness description of rxx and thus could easily lead to inaccurate
aim of this section is to propose an appropriate approach leading results. LW models provide a better description for rxx, although
to accurate failure results over all the volume of the plate, includ- some numerical oscillations can be detected for orders of expan-
ing boundary conditions. Each row of the Table 3 provides a con- sion greater than one. LW1 FEs are employed in the next computa-
vergence study with four-edge quadrangular FEs (Q4). The FEs tions since they limit the numerical oscillations and involve a
are formulated in accordance with the following theories. FSDT lower number of DOFs in comparison with other LWn descriptions.
(First Order Shear Deformation Theory) corresponds to the
3.3. Assessments of composite laminate failure results

In the following assessments, the plate already described in Sec-

tion 3.2 is considered with varying lamination angle h and different
combinations of boundary conditions and loading: monoaxial
compressive, monoaxial tensile, biaxial compressive, biaxial ten-
sile, biaxial combined compressive/tensile cases-study.
Since the point of failure is strongly dependent on boundary
conditions, loading, h and, in small part, on the criterion which is
applied, in the FEM computations the FI is evaluated over all the
volume of the plate by using post-processing routines. The corre-
sponding failure load is obtained by iterations, according to Eq.
(19).
Although in the literature classical criteria (like maximum
stress/strain, Tsai–Hill and Tsai–Wu) are rarely compared with
phenomenological criteria (like Hashin and LaRC03), in the follow-
Fig. 12. System’s description and static result. ing, results from both classical and phenomenological criteria are
 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289 287

Table 3
  The remaining criteria are quite acceptable, even if some differ-
Convergence study: rxx (MPa) at the plate point x ¼ 0; y ¼ 2b ; z ¼ 0 – computations ences can be detected for small lamination angles and around
performed with different kinematic descriptions and with n  n uniform meshes.
h = 65°.
FEs 44 88 12  12 16  16 Fig. 14b gives the same curves as Fig. 14a, with compressive
Q4-FSDT 0.88936 0.77756 0.77276 0.77416 loading along the y-direction (uniform pressure loading applied
Q4-LW1 0.88153 0.76775 0.76240 0.76238 only on edge 4 and clamp condition imposed only on edge 2).
Q4-LW2 0.88418 0.77027 0.77005 0.77654 Although the plate is slightly rectangular, it is certainly worth
Q4-LW3 0.88341 0.76901 0.76616 0.76930
Q4-LW4 0.88440 0.76860 0.76496 0.76818
pointing out the symmetrical trend of Fig. 14b and a.
Fig. 15a gives the failure results for the case of tensile loading
along the y-direction (uniform pressure loading applied only on
edge 4 and clamp condition imposed only on edge 2). Also in this
case, apart from the maximum strain criterion, a general accor-
dance is found among various failure criteria.
Fig. 15b provides the results in the case of tensile loading along
the x-direction (uniform pressure loading applied only on edge 3
and clamp condition imposed only on edge 1), confirming a strong
symmetry with Fig. 15a.

3.3.2. Biaxial loading

In the case-studies discussed in this section, a tensile or com-
pressive uniform pressure loading is applied on edges 3 and 4,
while edge 1 and 2 are clamped.
Failure results for the case of tensile equal biaxial loading along
the x and y-directions are provided in Fig. 16a. The curves of failure
are quite scattered: LaRC03 is the criterion leading to failure for the
highest load value. On the other hand, the maximum strain crite-
  rion is the approach which proposes the failure for the lowest load.
Fig. 13. rxx distribution along the thickness of the plate, plate point 0; 2b  16  16
regular mesh. All the curves are mostly symmetrical with respect to the axis
h = 45°.
Fig. 16b provides the results for the case of compressive equal
compared at the same level in order to propose an objective critical biaxial loading along the x and y-directions. Also in this case the
discussion of the different approaches. curves are significantly scattered. The Tsai–Wu criterion implies
The 20  20 regular mesh of LW1 FEs is used in the following the failure for the highest loading. The failure load for LaRC03 is
analyses. the lowest and this is due to the fact that the criterion takes into
account the fiber kinking phenomenon, that is, the fiber misalign-
3.3.1. Monoaxial loading ment under loading is considered, leading to a decrease in corre-
A compressive loading along the x-direction is considered in sponding failure load [11]. As for Fig. 16a, the chart is mostly
Fig. 14a, where the set of experimental failure results obtained symmetrical with respect to the axis h = 45°.
by Shuart [31] are plotted. According to Fig. 12a, a uniform pres- The case of biaxial equal load, with compression along the x-
sure loading is applied on edge 3, while edge 1 is clamped. Each an- direction and traction along the y-direction is illustrated in
gle of lamination involves an independent FEM analysis and the Fig. 17a. Some differences among the curves can be noticed. For
points composing the curves of failure are obtained by processing most of the lamination angles, the failure load computed by
the FEM results consistently with different failure criteria (as de- LaRC03 is higher than the others.
scribed in Eq. (19)). It can be noticed that in this case the maximum Fig. 17b shows the results for the biaxial equal load, with trac-
strain criterion is not able to approach the experimental results. tion along the x-direction and compression along the y-direction.

Fig. 14. Failure results for monoaxial compressive stress, width varying ply orientations.
288 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289

Fig. 15. Failure results for monoaxial tensile stress, width varying ply orientations.

Fig. 16. Failure results for rxx = ryy, width varying ply orientations.

Fig. 17. Failure results for jrxxj = jryyj, width varying ply orientations.

Also in this case there is no strong agreement among the curves. In 4. Conclusions
particular, when h ranges from 20° to 50°, the plot of the maximum
strain criterion appears strongly anomalous compared with the A brief review of the most popular two-dimensional failure cri-
others. It can be noticed that the trend of Fig. 17a and b is rather teria for composite layered structures has been proposed in this
symmetrical. work. The envelopes of maximum stress, maximum strain, Tsai–Hill,
 P. Nali, E. Carrera / Composites: Part B 43 (2012) 280–289 289

Tsai–Wu, Hashin and LaRC03 criteria have been obtained and [9] Hinton MJ, Kaddour AS, Soden PD. A comparison of the predictive capabilities
of current failure theories for composite laminates, judged against
described. For each approach, the in-plane shear effect has also
experimental evidence. Compos Sci Technol 2002;62(12–13):1725–97.
been illustrated. Different case-studies have been introduced in [10] Hinton MJ, Soden PD. Predicting failure in composite laminates: the
order to compare the failure results in the case of monoaxial and background to the exercise. Compos Sci Technol 1998;58(7):1001–10.
biaxial loading. Layered plates have been modeled through layer- [11] Dávila CG, Camanho PP. Failure criteria for frp laminates in plane stress. Tech.
Rep. NASA/TM-2003-212663; 2003.
wise kinematical description in order to reduce the error of model- [12] Dávila CG, Jaunky N, Goswami S. Failure criteria for frp laminates in plane
ing. Numerical solutions have been obtained by FEM in MUL2 stress. In: Proceedings of 44th AIAA/ASME/ASCE/AHS/ASC structures,
academic code. It was found that the LaRC03 criterion gives results structural dynamics, and materials conference; 2003.
[13] Dvorak GJ, Laws N. Analysis of progressive matrix cracking in composite
in good accordance with the test experiences referred. The laminates. II. First ply failure. J Compos Mater 1987;21:309–29.
maximum strain criterion appears the most inappropriate in all the [14] Pinho ST, Dávila CG, Camanho PP, Iannucci L, Robinson P. Failure models and
case-studies considered. A common agreement among different criteria for frp under in-plane or three-dimensional stress states including
shear non-linearity. Tech. Rep. NASA/TM-2005-213530; 2005.
criteria is not obtained in the case of biaxial compressive load, [15] Pinho ST, Iannucci L, Robinson P. Physically-based failure models and criteria
where the LaRC03 criterion is the most conservative, being the only for laminated fibre-reinforced composites with emphasis on fibre kinking. Part
one which takes into account the fiber kinking phenomenon. The I: Development. Composites Part A: Appl Sci Manuf 2006;37(1):63–73.
[16] Pinho ST, Iannucci L, Robinson P. Physically based failure models and criteria
experimental results currently available in literature are rather for laminated fibre-reinforced composites with emphasis on fibre kinking. Part
scattered and most often consist of a low number of experimental II: Fe implementation. Composites Part A: Appl Sci Manuf 2006;37(5):766–77.
points. Additional test campaigns are needed in order to better [17] Maimí P, Camanho PP, Mayugo JA, Dávila CG. A continuum damage model for
composite laminates. Part I – Constitutive model. Mech Mater 2007;39(10):
quantify the accuracy of the failure criteria in the case of biaxial
897–908.
and combined shear/axial loading. [18] Maimí P, Camanho PP, Mayugo JA, Dávila CG. A continuum damage model for
composite laminates. Part II – Computational implementation and validation.
Acknowledgement Mech Mater 2007;39(10):909–19.
[19] Orifici AC, Herszberg I, Thomson RS. Review of methodologies for composite
material modelling incorporating failure. Compos Struct 2008;86:194–210.
This work was partially carried out in the framework of Pied- [20] Material Solutions Polymer Composites. <http://www.admc.esrtechnology.
mont regional STEPS contract. com/background/failure/>.
[21] Reddy JN. Mechanics of laminated composite plates and shells, theory and
analysis. 1st ed. Boca Raton (FL): CRC Press; 1997.
References [22] Tsai SW, Wu EM. A general theory of strength for anisotropic materials. J
Compos Mater 1971;5(1):58–80.
[1] Hoffman O. The brittle strength of orthotropic materials. J Compos Mater [23] Liu KS, Tsai SW. A progressive quadratic failure criterion for a laminate.
1967;1(2):200–6. Compos Sci Technol 1998;58(7):1023–32.
[2] Tsai SW, Hahn HT. Failure analysis of composite materials. In: Winter annual [24] Hill R. Theory of mechanical properties of fiber-strengthened materials. III.
meeting of ASME; 1975. p. 73–96. Self-consistent model. J Mech Phys Solids 1965;13:189–98.
[3] Hashin Z, Rotem A. A fatigue criterion for fiber-reinforced materials. J Compos [25] Tsai SW. Strength theories of filamentary structures fundamental aspects of
Mater 1973;7:448–64. fibre reinforced plastic composites. New York: WileyInterscience; 1968.
[4] Hashin Z. Failure criteria for unidirectional fiber composites. J Appl Mech [26] Sun CT, Quinn BJ, Oplinger DW. Comparative evaluation of failure analysis
1980;47:329–34. methods for composite laminates. Tech. Rep. DOT/FAA/AR-95/109, 20591.
[5] Puck A, Schuermann H. Failure analysis of frp laminates by means of physically Washington (DC): Office of Aviation Research; 1996.
based phenomenological models. Compos Sci Technol 1998;58(7):1045–67. [27] Reissner E. The effect of transverse shear deformation on the bending of elastic
[6] Puck A, Schuermann H. Failure analysis of frp laminates by means of physically plates. J Appl Mech 1945;12:69–76.
based phenomenological models. Compos Sci Technol 2002;62(12–13): [28] Mindlin RD. Influence of rotatory inertia and shear in flexural motions of
1633–62. isotropic elastic plates. J Appl Mech 1951;18:1031–6.
[7] Soden PD, Hinton MJ, Kaddour AS. A comparison of the predictive capabilities [29] Reddy JN. A generalization of two-dimensional theories of laminated
of current failure theories for composite laminates. Compos Sci Technol composite plates. Commun Appl Numer Methods 1987;3:173–80.
1998;58(7):1225–54. [30] Carrera E. Theories and finite elements for multilayered plates and shells: a
[8] Soden PD, Hinton MJ, Kaddour AS. Biaxial test results for strength and unified compact formulation with numerical assessment and benchmarking.
deformation of a range of e-glass and carbon fibre reinforced composite Arch Comput Method Eng 2003;10(3):215–96.
laminates: failure exercise benchmark data. Compos Sci Technol 2002;62(12–13): [31] Shuart MJ. Failure of compression-loaded multidirectional composite
1489–514. laminates. AIAA J 1987;27(9):1274–9.