You are on page 1of 9

COMPOSITES

SCIENCE AND
TECHNOLOGY
Composites Science and Technology 66 (2006) 2054–2062
www.elsevier.com/locate/compscitech

Accounting for residual stresses in FEM analyses of laminated


structures using the Puck criterion for three-axial stress states
C. Schuecker, D.H. Pahr, H.E. Pettermann *

Austrian Aeronautics Research (AAR)/Network for Materials and Engineering at the Institute of Lightweight Design and Structural Biomechanics,
Vienna University of Technology, Vienna, Austria

Received 14 February 2005; received in revised form 22 December 2005; accepted 26 December 2005
Available online 23 February 2006

Abstract

A computer program for handling non-proportional loading in First Ply Failure analyses of layered fiber reinforced composites is
developed. To this end, stresses are split up into a constant and a variable part, such that for the latter the assumption of proportional
increase of stress components is valid. The risk of failure is assessed in terms of a stress based failure envelope and four risk parameters,
which are defined as linear combinations of the constant and variable stress contributions. Amongst others, the Puck failure criteria for
plane and three-axial stress states are implemented for stress assessment. By hooking the program up onto a Finite Element package as a
post-processing tool, it can be utilized in structural failure analysis. The approach is used to study the effect of residual ply stresses on the
failure behavior of a typical laminated structure under complex loading.
 2006 Elsevier Ltd. All rights reserved.

Keywords: A. Polymer–matrix composites; C. Residual stresses; Laminates; Failure criterion; Computational simulation

1. Introduction (maximum stress/strain criteria [1]) and quadratic failure


criteria like the Tsai-Hill, Tsai-Wu, or Hoffman criterion
In the aerospace and many other industries, structural [1,2]. More advanced approaches are mechanism based
components made from fiber reinforced polymer (FRP) failure criteria initiated by the works of Hashin [3], such
laminates are becoming increasingly important due to their as the Puck failure criterion [4–7] and variations thereof
great potential for weight savings. In the design process, the [8–10]. In addition to the risk of failure, these mechanism
availability of reliable modeling tools to predict the strength based criteria yield information on failure modes and frac-
of a component is crucial. One of the most commonly used ture plane angles. Recently, in a World Wide Failure Exer-
approaches for failure analysis of laminated structures is the cise [11–13], an extensive study was conducted comparing
First Ply Failure (FPF) concept [1,2]. It states that failure the predictive capabilities of a number of failure criteria
occurs, if the stress state in one of the laminate’s layers for uni-directional (UD) laminates. In this study it is
reaches a certain stress limit. For multi-axial stress states, found, that out of the failure criteria tested, the ones devel-
failure criteria are used to define stress limits by way of a oped by Puck [5] and Cuntze [8] give the best results.
failure envelope. Several failure criteria for orthotropic Another observation made in this study is that there can
materials have been presented in the past decades. be significant variations in predicting laminate failure
Among the most commonly used criteria are those depending on whether or not residual curing stresses are
assuming no interaction between ply-stress components taken into account [11]. The study emphasizes the impor-
*
Corresponding author. Tel.: +43 1 5880131726; fax: +43 1
tance of using physically meaningful failure criteria and
5880131799. accounting for production related stresses in order to fur-
E-mail address: pettermann@ilsb.tuwien.ac.at (H.E. Pettermann). ther improve laminate failure predictions.

0266-3538/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compscitech.2005.12.024
C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062 2055

Notation

General notation w direction of shear component of traction vector


1, 2, 3 ply coordinates in rnt  rnl plane
l, n, t fracture plane coordinates RA
23 ; sc coordinates of Mode B/Mode C transition point
( )t, ( )c value ( ) for tension, compression, respectively in r22  r12 plane
k risk parameter hfp fracture plane angle
kcmb risk parameter of combined stress state
kcon risk parameter of constant stress state Variables in example problem
kvar risk parameter of variable stress state c, c0 winding angle, winding angle in cylindrical part,
ktot risk parameter of constant + variable stress respectively
state r, r0 distance from axis of rotation, cylinder radius,
Rkl first ply failure strength values respectively
rij stress tensor t, t0 ply thickness, ply thickness in cylindrical part,
rij stress state at ply failure respectively
rcon
ij constant stress state
rvar
ij variable stress state Abbreviations
FF fiber failure
Parameters in Puck’s action plane failure criterion FPF first ply failure
fw1 weakening factor for high longitudinal stresses FRP fiber reinforced polymer
p12 slope parameter at R12 (pt12 for r22 P 0, pc12 for IFF inter fiber failure
r22 < 0) UD uni-directional
p23 slope parameter at R23 (pt23 for r22 P 0, pc23 for
r22 < 0)

In conventional FPF analyses residual stresses cannot be failure, rij (with respect to some FPF criterion), by multi-
included, since proportional increase of all stress components plication with a given stress tensor, rij,
is assumed. In the present paper, the ‘superposition method’
F ðrij ; Rt;c t;c
kl Þ ¼ F ðkrij ; Rkl Þ  1. ð1Þ
[5,14,15] for studying combined stress states within the FPF
concept is adopted. Combining the approach with the Finite The function F ðrij ; Rt;c
kl Þ refers to any suitable failure cri-
Element Method (FEM), a post-processing tool for struc- terion, with FPF strength values Rkl, and superscripts t and
tural analysis is developed. The software concept is open to c denoting tension and compression, respectively (Fig. 1).
several failure criteria, though the present study focuses on Eq. (1) implies proportional increase of all stress compo-
Puck’s failure hypothesis. Through the implementation of nents with load, which is not the case, if stress contribu-
the Puck failure criteria for plane and three-axial stress tions are caused by independent load mechanisms, e.g. a
states, failure modes and fracture plane angles can be pre- combination of temperature and mechanical loads. In such
dicted in addition to the risk of failure. The superposition a case, one of the load mechanisms needs to be identified as
method is applicable to any combination of independent the load of primary interest, and the corresponding stresses
load mechanisms, but in the present study it is used to look
into the effects of residual stresses. An example problem dem-
onstrating the features of this structural analysis tool is pre-
sented. The case study also shows how residual stresses affect
the failure behavior of a laminated structure.

2. Superposition method for combined stress states a b

2.1. FPF and combined stress states

The concept of FPF operates on the ply level, and


failure envelopes are defined with reference to a local ply-
coordinate system with axes 1 in fiber direction, 2 transverse c d
to fibers in-plane, and 3 transverse out-of-plane. Assuming
Fig. 1. Schematic illustration of risk parameters: (a) FPF under propor-
linear elastic, orthotropic material behavior, the risk of tional loading, k, vs. FPF of combined stress states, kcmb; (b) constant risk
failure is characterized by a scalar risk parameter k. It is parameter, kcon; (c) variable risk parameter, kvar and (d) total risk
defined as the factor that yields the stress state at ply parameter, ktot.
2056 C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062

are defined as variable stress tensor, rvar


ij . The sum of stress ular structure in order to provide for easy extension to
states due to all other load mechanisms is defined as con- additional failure criteria. Plane as well as three axial stress
stant stress tensor rcon
ij . For linear elastic material, the cur- states can be handled by the routine. The four risk param-
rent stress tensor is the sum of constant and variable eters presented above are available for output, addition-
stresses, and the risk parameter for combined stress states, ally, failure mode and fracture plane angle can be
kcmb, is defined as predicted depending on the failure criterion applied.
t;c The program can be used as a stand alone tool, e.g. for
F ðkcmb rvar con
ij þ rij ; Rkl Þ  1 ð2Þ
evaluating combined stress states and constructing failure
(cf. Fig. 1a). This risk parameter, kcmb, is a proper measure envelopes of laminates that were analysed by lamination
for the risk of failure under combined stresses, but may not theory. In order to enable failure analysis of complex com-
give a complete picture of the situation. Therefore, addi- posite structures, it is hooked up onto the FEM Program
tional risk parameters are introduced. A constant risk ABAQUS, Inc., Pawtucket, RI as a post-processing subrou-
parameter, kcon, is defined based on the constant stress ten- tine. Any shell or continuum type FEM model suitable for
sor (cf. Fig. 1b) studying laminated structures can be used. A two step
t;c FEM-stress analysis needs to be performed, in which the
F ðkcon rcon
ij ; Rkl Þ  1; ð3Þ
constant and variable load cases are applied consecutively.
in order to ensure that the constant load can be sustained if In the post-processing routine, the layer stresses computed
the variable load is zero (i.e. kcon > 1 is required). Knowl- by FEM are used to determine the failure relevant vari-
edge of the constant risk parameter is also important for ables. These are stored by ABAQUS as intrinsic variables,
assessing the reliability of results, since the influence of so they can be accessed by any compatible post-processor,
loading path may not be negligible for kcon ! 1 (see discus- e.g. for visualization as contour plots.
sion at end of Section 2.1). It should be noted, that the use of FPF criteria in com-
The variable risk parameter, kvar, (Fig. 1c) shows if the bination with three-axial stress states is somewhat ambigu-
variable load can also be carried when there is no constant ous. Criteria for FPF are designed for use at the ply level
load and gives a comparison to the case where constant assuming a homogeneous stress state over ply thickness.
stresses are neglected, With increasing layer thickness and more triaxial stress
F ðkvar rvar t;c states, stress gradients in thickness direction are not neces-
ij ; Rkl Þ  1. ð4Þ
sarily negligible, thus violating requirements of the FPF
Finally, a total risk parameter, ktot, is defined to identify approach. The question arises, whether the use of averaged
‘direction sensitive’ load cases, ply stresses or local (material point) stresses give more real-
F ðktot ðrvar con t;c istic predictions of the risk of FPF in 3D models. In the
ij þ rij Þ; Rkl Þ  1. ð5Þ
present work, the latter option is adopted, hence, the pre-
tot
Comparing k to the other three risk parameters can diction of risk of failure is a conservative one. Additionally,
indicate if the variable stresses are ‘parallel’ to the failure high stress gradients can be recognized as variations of the
surface, in which case a small variation of one stress com- risk parameter over ply thickness, which would be an indi-
ponent of the variable or constant stress tensors can lead to cation that the FPF approach reaches its limit of applica-
a significant change of kcmb (Fig. 1d). bility at the respective location.
As mentioned previously, the superposition method can
only be used if material behavior is approximated as linear 2.3. Failure criteria
elastic. Otherwise, laminate failure depends on the loading
path, and the superposition of stress states is not valid. Given the results of the World Wide Failure Exercise
Non-linear material behavior is most pronounced in shear [13] and the capability of Puck’s failure criterion to addi-
loading conditions and becomes relevant when ply shear tionally predict failure modes and fracture plane orienta-
stresses reach approximately 50% of ultimate shear tions, it seems to be one of the most potent failure
strength. The load path dependency of laminate failure is criteria at the time. Therefore, the Puck failure criterion
discussed in [8,16,17], where it is indicated that there is is focused on in the present work, even though it is not
no influence of loading path as long as no damage occurs the only failure criterion implemented in the CNV software.
or if the damage mechanism does not change during load- Since the criterion is not very well known and there have
ing. These observations need to be kept in mind when look- been several variations proposed in the literature [4–7,19],
ing at failure predictions using the superposition method. the version of the Puck failure criterion used in the present
Particularly, when the constant stress contributions are study will be reviewed in the following.
high (kcon 6 2) caution is advised. Unlike most of the commonly used failure criteria (e.g.
Tsai-Hill, Tsai-Wu, Maximum Stress Criteria), the Puck
2.2. Implementation failure criterion is designed specifically for continuous fiber
reinforced composites and is, therefore, restricted to
A software CNV is developed based on the superposition transversally isotropic materials. It uses separate
method and the considerations in Section 2.1, using a mod- definitions for different parts of the failure envelope
C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062 2057

depending on the failure mode. Fiber failure (FF) and inter


fiber failure (IFF) are distinguished as general failure nt
p c23
modes. Assuming that only the stress component in fiber A
R 23
direction, r11, is relevant for FF, the maximum stress crite-
t
rion is used to predict FF p 23
  p c
t
R22
 r11  12
  6 1. ð6Þ
Rt;c  t
R12 p 12
11
nl nn
IFF is viewed as brittle failure leading to fracture in a
Fig. 3. Puck fracture plane failure surface for three-axial stress states
plane that is parallel to the ply’s fiber direction and defined
(stresses with respect to fracture plane coordinate system).
by the fracture plane angle, hfp, as shown in Fig. 2. The
physical basis for Puck’s IFF criterion is given by Mohr’s is open for rnn ! 1, signifying that pressure on a plane
fracture hypothesis for brittle materials. It states, that fail- does not lead to failure of this plane, but rather to shear
ure occurs in that material plane, which offers the lowest failure of an inclined plane.
failure resistance towards the stresses acting on the plane. According to Puck, the actual fracture plane is the plane
Consequently, the failure surface is defined with respect that has the highest risk of failure. In the general case of
to a fracture plane coordinate system l–n–t (cf. Fig. 2) three-axial stress states (referred to as ‘Puck 3D’ in the fol-
and is a function of the fracture plane stresses (rnn, rnt, rnl), lowing), the fracture plane angle is determined by a mini-
only. Note, that for the present model the direction of l and mum search for the plane of lowest risk parameter
1 always coincide. (k(h) ! min). A prediction of hfp = ±90 is interpreted as
Two cases of IFF are distinguished, depending on the delamination. Typically, the function k(h) has several local
fracture plane’s normal stress component, rnn, being tensile minima, which need to be determined and compared in
or compressive. The corresponding failure surface is given order to find the global minimum. In cases where the values
in [4] of k at two (or more) local minima are similar, a small var-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
iation of the stress state can cause the global minimum to
u  2  2
u 1 pt2w rnt rnl pt2w jump from one local minimum to another, leading to an
t  r 2 þ þ þ rnn ¼ 1
nn abrupt change of predicted fracture plane angle.
Rt22 RA 2w RA 23
R12 RA2w
In plane stress, hfp can be determined analytically if the
for rnn P 0 relation pc2w =RA 2w ¼ const is assumed. Plugging this relation
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi into Eq. (7), the ‘Puck 2D’ formulation can be derived.
u c !2  2  2
u p c
t 2w r2 þ rnt þ rnl þ p2w r ¼ 1 Transformation of the fracture plane stresses, rnn, rnt,
nn nn
RA 2w RA 23
R12 RA
2w
rnl, to ply stresses, r22, r12, leads to the more commonly
known set of equations distinguishing three IFF modes
for rnn < 0 (Fig. 4, left) [4]
Rc22 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2  2ffi
with RA
23 ¼ 1 pt12 r12 pt
2ð1 þ pc23 Þ 2
þ 12 r22 ¼ 1;
t  r22 þ
pt;c R22 R12 R12 R12
2w pt;c pt;c rnl
and ¼ 23
cos 2
w þ 12
sin2 w with w ¼ arctan . hfp ¼ 0 for r22 P 0 (Mode A)
RA RA R rnt sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
2w 23 12
2
ð7Þ pc12 r12 pc
r222 þ þ 12 r22 ¼ 1;
R12 R12 R12
Some estimates for physically realistic values of slope  
parameters pt12 , pc12 , pt23 , and pc23 are given in [18]. A sche- r22  RA
hfp ¼ 0 for r22 < 0 and 0 6   6 23 (Mode B)
matic representation of the fracture plane failure envelope r12 sc
" 2  2 # c
and its parameters is shown in Fig. 3. Note that the surface r12 r22 R22
þ ¼ 1;
2ð1 þ pc23 ÞR12 Rc22 r22
sffiffiffiffiffiffiffiffiffiffi
RA 23
cos hfp ¼ for r22 < 0
θfp r22
 
3 t n r12  sc
and 0 6   6 A (Mode C)
22 r 23 R
l
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pc pc
2
with sc ¼ R12 1 þ 2pc23 ; 12 ¼ 23
R12 RA
1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 23 !
R12 Rc22
Fig. 2. Definition of fracture plane coordinate system and fracture plane and RA23 ¼ 1 þ 2p c
12 1 . ð8Þ
2pc12 R12
angle, hfp.
2058 C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062

τc σ12
σ12
A R12 Inter fiber
R 23 R12
failure
σ22 0.5R12 fiber
c t
R
22
t
R22 R22 failure
t
0.7R11
t
mode A σ22 R11
R12
mode C
mode B σ 11
Fig. 4. Puck failure surface for plane stress states (stresses with respect to ply coordinate system).

In Eq. (8) the definition of hfp for Mode C has two pos- z
sible solutions in the domain p/2 6 h 6 + p/2, implying
that the planes +hfp and hfp are equally likely to fail. Ø50
Since the fracture planes are always parallel to the fiber
direction, there is no influence of longitudinal stresses on
IFF in Eqs. (7) and (8). However, Puck suggests to use a
weakening factor, fw1, to scale the failure envelopes at high
longitudinal stresses according to
distributed
F Puck ðrij ; Rt;c t;c
kl ; p kl Þ ¼ fw1 ð9Þ load

with the weakening factor being defined as 0


= 19.5˚
8
> 1 for jr11 j < 0:7Rt;c
11 ; = -75K
< sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fw1 ¼     2 pi = 4MPa
> 1  1  r11t;c   0:7
: for jr11 j P 0:7Rt;c
0:12 R 11 .
11
Ø150
ð10Þ
Its effect on the Puck 2D failure envelope in the three-
dimensional stress space of ply-stress components is shown
in Fig. 4, right. A somewhat modified formulation of this Fig. 5. Pressure vessel geometry and applied loads.
weakening factor can be found in [19], where longitudinal
weakening is suggested for jr11 j P 0:5Rt;c
11 if specific mate-
rial data is not available. a lid (not shown) mounted to the inside of the rim. The
laminate lay-up of the vessel is summarized in Table 1. In
3. Example problem the cylindrical part, the basic angle ply laminate is rein-
forced by additional 90-layers on the inside and outside.
A typical engineering structure is chosen to demonstrate The thickness of each 90-layer is 0.2 mm, that of each
the features of the developed analysis tool. Due to its angle-ply 0.6 mm. Due to the filament winding process
geometry and complex loading, the structure is difficult there is a gradual change in fiber angle, c, and thickness,
to analyse analytically. In some regions of the structure t, of each layer along the dome contour. The winding con-
three-axial stress states occur, which calls for 3D modeling. dition for geodesic winding is given in [20]
These stresses can be evaluated by the Puck 3D criterion, r sin c ¼ const; ð11Þ
while Puck 2D can be used for assessing plane stress states
where r is the distance from the axis of revolution. For a
computed via layered shell elements. In both cases, the dis-
circular dome contour Eq. (11) yields the fiber angle as a
tribution of risk parameters, failure modes and fracture
function of axial coordinate, z,
plane angles are predicted.

3.1. Problem definition


Table 1
The case study is concerned with a filament wound pres-
Laminate design of filament winding pressure vessel (0 = axial direction)
sure vessel, typical for propellant storage in space craft.
Lay-up Ply thickness [mm]
The geometry of the vessel consists of a cylindrical center
part with a spherical dome on each end (Fig. 5). At the Cylinder [90/ + 19.5/  19.5]s t90 = 0.2; t19.5 = 0.6
Dome [ + c/  c]s (c = 19.5 . . . 90) t = 0.6 . . . 1.91
top of the dome there is an opening which is covered by
C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062 2059
" #
r0 sin c0 needs to be modeled. Note that axisymmetric modelling
c ¼ arcsin pffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ; ð12Þ is not possible in the problem at hand. Due to the laminate
r20  z2
lay-up, the specified loads lead to displacements in hoop
with initial winding angle, c0 = 19.5, and cylinder radius, direction, which violate assumptions of standard axisym-
r0 = 75 mm (z = 0 at the dome/cylinder conjunction). metric FEM-modeling.
Due to the changing fiber angle and shell diameter, the Near the dome openings, the stress state is more com-
layer thickness varies according to plex due to lid forces and the higher ratio of shell thickness
r0 cos c0 versus diameter. A computationally efficient way to exam-
t ¼ t0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; ð13Þ ine the three-axial stress state near the rim more closely is
cos c r20  z2
the submodeling technique, which is realized by employing
with t0 = 0.6 mm denoting the thickness of each angle ply a 3D continuum submodel. In the submodel, the three rows
at z = 0. Eqs. (12) and (13) lead to a maximum ply thick- of shell elements closest to the dome opening are replaced
ness and fiber angle at the dome openings of 1.91 mm by continuum elements (cf. Fig. 6, right). Each ply is mod-
and 90, respectively. The vessel is made of the intermedi- eled by four elements over its thickness using 20-noded
ate carbon fiber/epoxy material AS4/3501-6. Transversely hexahedral elements. For comparison, a coarsely meshed
isotropic material data is taken from [21] and summarized submodel with two elements per layer is analysed as well,
in Table 2. Interlaminar strength data for the Puck 3D cri- showing good agreement with the refined model. Free edge
terion is derived from in-plane strength data given in Table effects [14,22] are not considered in the present study.
2 by a 10% reduction, as suggested in [4,5]. Therefore, the results of the first two rows of elements
The vessel is operated at room temperature. According are not relevant and are not displayed in the figures of
to material specifications [21], the stress free temperature the results section. The same strategy is applied to the ele-
corresponds to 177 C, however, a reduction of thermally ment rows near the shell/submodel transition zone where
induced stresses due to relaxation can be expected and is results are unreliable due to the locally distorted deforma-
assumed to amount to 50%. This yields an effective temper- tion field.
ature load of DT = 75 C, which is applied to the struc-
ture as constant load. An internal pressure of 4 MPa 3.2. Results
causes variable loads. In addition to uniform pressure,
there is a radially cosine-distributed load at the rim of In this section, results of the structural failure analysis
the vessel due to the pressure acting on the lid (cf. Fig. 5). are presented. For the material in use, the influence of
For most of the structure, shell thickness is very small the ratio p2w =RA
2w varying with w (cf. Eq. (7)) is small. Con-
compared to the vessel diameter and the gradients of exter- sequently, the Puck 2D and Puck 3D formulations basi-
nal loads are small. Therefore, a layered shell model is cally yield the same results for plane stress states. Since
assumed to be sufficiently accurate for predicting the over- Puck 2D is much more computationally efficient, risk
all behavior of the vessel (Fig. 6, left). The shell model con- parameters of the shell model are computed using Puck
sists of quadrilateral 8-noded shell elements employing first 2D, while stress states of the continuum submodel are
order lamination theory. Because of globally symmetric assessed by Puck 3D. Note, that in the following stresses
geometry, loading, and response, only half of the vessel are referred to the respective ply’s coordinate system, which
is defined through rotation about the shell normal by an
angle ±c. Therefore, the 1-directions of the coordinate sys-
tems of the middle and outer layers point in opposite direc-
Table 2
Material data of carbon fiber/epoxy UD-layer, AS4/3501–6 [21] tions at the dome opening where c = 90.
Elastic constants
E1 [GPa] 126 3.2.1. Constant temperature load
E2 = E3 [GPa] 11 Subjecting the vessel to a temperature change causes
G12 = G13 [GPa] 6.6 stresses by two effects. The first one is related to the
m12 = m13 0.28 angle-ply lay-up of the laminate. Since the ply thermal
m23 0.4
expansion in fiber direction is smaller than the one in trans-
verse direction, a temperature reduction leads to tensile ply
Tension Compression
stresses transverse to the fibers, r22 > 0. These stresses
Strength data depend upon the laminate’s ply angles and have their max-
R11 [MPa] 1950 1480
R22 [MPa] 48 200
imum for a [±45] lay-up. As a result of the low tensile ply-
R12 [MPa] 79 79 strength in transverse direction, these stresses have a high
p12 0.35a 0.3a potential to cause failure. The second effect is due to the
p23 0.27a 0.27a doubly curved geometry in the dome section. Because of
Coefficients of thermal expansion the differing thermal expansion properties of the laminate
a11 [1/K] 1 · 106 in meridional and circumferential directions, laminate
a22 = a33 [1/K] 26 · 106
strain components in these directions are not equal, which
a
Following Puck’s guidelines for carbon fiber laminates [18].
2060 C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062

Fig. 6. FEM-mesh; global shell model (left), and cut open view of 3D continuum submodel (right).

results in a bending moment. The combination of these two 3.2.2. Variable pressure load
effects leads to the distribution of constant risk parameter According to analytical solutions the highest section
as shown in Fig. 7. forces in a vessel subjected to internal pressure will occur
In the shell model view (Fig. 7, left) the outer angle ply is in circumferential direction and are proportional to the
displayed, which is the most critical one. There, two critical vessel diameter. The cylindrical part is reinforced with
regions of constant risk parameter can be discerned. They 90 layers which carry a major part of these circumferential
are located where the difference between fiber angles of laminate stresses. At the beginning of the dome, however,
adjacent layers is high, i.e. in the cylindrical section fiber directions are in ±19.5 only, without further rein-
between the 90 and the 19.5 layer, and near the rim of forcement, while section forces are still the same as in the
the dome where fiber angles are approximately ±45. cylinder. Consequently, the relatively low transverse
Because of the coarse shell discretization near the dome strength is of relevance and the lowest variable risk param-
opening, it is useful to obtain more accurate values for eter can be expected in this area. The distribution of kcmb is
the second location by considering the continuum model, shown for the outer angle ply in Fig. 8, with a critical risk
a slice of which is shown in Fig. 7, right. While kcon does parameter of kcmb = 1.05 near the conjunction of the cylin-
not vary over laminate thickness in the cylindrical part, der and dome sections (Fig. 8, left). With the reducing shell
in the dome section there is a gradient in thickness direc- diameter of the dome, fiber angles increase, so that the
tion. Corresponding to the bending moment due to the higher strength fibers are able to carry a growing portion
doubly curved geometry the highest risk of failure occurs of the load. Additionally, the shell thickness increases as
in the outermost layer with a constant risk parameter of well, hence, no failure due to the internal pressure is to
kcon = 2.4. be expected there. A look at the submodel (Fig. 8, right)
confirms this general assessment up to the point where
the fiber orientation is ±45. For c > 45 the risk of tensile
matrix failure increases again due to the unfavorable fiber

Fig. 7. Distribution of constant risk parameter, kcon; outermost angle ply Fig. 8. Distribution of combined risk parameter, kcmb; outermost angle
of global shell model – Puck 2D (left), and slice of 3D continuum ply of global shell model – Puck 2D (left), and slice of 3D continuum
submodel – Puck 3D (right). submodel – Puck 3D (right).
C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062 2061

orientation with regard to meridional laminate stresses


(minimum kcmb = 1.8 in region without compressive r33
stresses).
Whether or not residual stresses are taken into account
has no influence on the spatial distribution of the risk of
failure. However, the magnitude of the combined risk
parameter, kcmb, is significantly lower than that of the var-
iable risk parameter with a critical value of kvar = 1.3 at the
dome/cylinder conjunction. The lowest total risk parame-
ter is also found at that location with ktot = 1.04.

3.2.3. Failure mode


Both temperature and pressure load mainly cause tensile
layer stresses in transverse direction as long as plane stress
is assumed. Therefore, tensile matrix failure is predicted Fig. 10. Distribution of fracture plane angle, hfp, in the continuum
throughout the shell model. The distribution of failure submodel predicted by Puck 3D failure criterion.
modes in the continuum model is shown in Fig. 9, illustrat-
ing that tensile matrix failure predominates here as well. Fig. 11). According to Puck’s hypothesis, purely compres-
Only the region near the application of the distributed lid sive stresses do not lead to failure directly, rather, they
forces exhibits compressive matrix failure. However, risk cause shear failure at an inclined angle. For example, pure
parameters are very high there, and FPF of this mode out-of-plane compressive stresses would lead to a fracture
unlikely. plane angle slightly below 40. Here, in-plane stresses are
present at the same time and, in combination with the com-
3.2.4. Fracture plane angle pressive out-of-plane stresses, lead to the observed fracture
In plane stress conditions, tensile matrix failure leads to angle. It is noted that in the region where the failure mode
a fracture plane that is perpendicular to the laminate plane, switches from tension to compression (cf. Figs. 9 and 10),
i.e. hfp = 0 (cf. Eq. (8), Mode A). This is the case every- the prediction of hfp is difficult, since the global minimum
where in the shell model. The distribution of fracture plane of k(h) switches between several local minima.
angles in the continuum model is shown in Fig. 10. Let us
first consider the area where tensile failure is predicted (cf. 3.2.5. Validity of the approach
Fig. 9). There, the fracture plane angle deviates up to 18 As pointed out in Section 2.1, the presented approach is
from the zero-degree orientation with different sign valid only as long as failure is independent of the loading
depending on the sign of fiber orientation in the layer. path, and it is necessary to ascertain that this assumption is
The deviation is attributed primarily to out-of-plane shear not violated. In the example problem, constant stresses are
stresses. Due to the parabolic distribution of macro-shear well below the expected failure stress, since kcon P 2.4. Shear
stresses, the absolute values of hfp are higher at the two stresses and compressive transverse stresses, which are most
middle layers (18 at maximum) than at the outer layers likely to induce non-linear material behavior, are negligible
(jhfpj 6 11). The sign of the predicted fracture plane angle in the constant load case. During loading the failure mode
is related to the definition of the respective ply coordinate does not change, except for a small area near the introduction
system. of lid forces. At that location, however, constant stresses are
In the area of compressive matrix failure, high fracture very low, and stress components increase almost proportion-
plane angles are predicted with a maximum of hfp  50
(Fig. 10). Since the only compressive stresses caused by var-
iable loads appear in the out-of-plane direction (r33 < 0)
they must be responsible for this failure mode (cf.

Fig. 9. Distribution of failure mode in the continuum submodel predicted Fig. 11. Distribution of out-of-plane normal stress, r33, (ply-coordinate
by Puck 3D failure criterion. system), in the continuum submodel.
2062 C. Schuecker et al. / Composites Science and Technology 66 (2006) 2054–2062

ally with load. As a result it can be assumed, that failure is composite structures, vol. 348 of CISM courses and lectures. New
independent of the loading path and the application of the York, Austria: Springer Verlag Wien; 1994. p. 73–110.
[3] Hashin Z. Failure criteria for unidirectional fiber composites. J Appl
superposition method is legitimate. Mech 1980;47:329–34.
[4] Puck A. Festigkeitsanalyse von Faser-Matrix-Laminaten. München
4. Conclusions Wien, Germany: Carl Hanser Verlag; 1996.
[5] Puck A, Schürmann H. Failure analysis of FRP laminates by means
In the current study an approach for evaluating com- of physically based phenomenological models. Comp Sci Tech
1998;58:1045–67.
bined stress states, showing constant and variable charac- [6] Puck A, Schürmann H. Failure analysis of FRP laminates by means
ter, is presented. Superposition of stress states is used in of physically based phenomenological models. Comp Sci Tech
conjunction with the concept of FPF to predict the onset 2002;62:1633–62.
of failure in laminated composite structures. The risk of [7] Juhasz J, Rolfes R, Rohwer K. A new strength model for application
failure and its sensitivity to various load combinations is of a physically based failure criterion to orthogonal 3D fiber
reinforced plastics. Comp Sci Tech 2001;61:1821–32.
characterized by four risk parameters. A modular com- [8] Cuntze RG, Freund A. The predictive capability of failure mode
puter code is developed, which is easily adaptable for concept-based strength criteria for multidirectional laminates. Comp
additional failure criteria and can handle plane as well Sci Tech 2004;64:343–77.
as three-axial stress states. The Puck fracture plane criteria [9] Cuntze RG. The predictive capability of failure mode concept-based
for plane and three-axial stress states are employed. By strength criteria for multidirectional laminates–part B. Comp Sci
Tech 2004;64:487–516.
hooking the code up onto a FEM program as post-pro- [10] Dávila CG, Camanho PP. Failure criteria for FRP laminates in plane
cessing tool, stress analysis of complex structures can be stress. Tech Rep TM-2003–212663, NASA, 2003.
performed, while additionally assessing the distribution [11] Soden PD, Hinton MJ, Kaddour AS. A comparison of the predictive
of the risk of failure, failure mode, and fracture plane capabilities of current failure theories for composite laminates. Comp
angle. Sci Tech 1998;58:1011–22.
[12] Hinton MJ, Kaddour AS, Soden PD. A comparison of the predictive
The features of the developed software tool are dem- capabilities of current failure theories for composite laminates,
onstrated for a filament winding pressure vessel with judged against experimental evidence. Comp Sci Tech 2002;62:
residual curing stresses. The spatial distribution of the 1725–97.
risk of failure is predicted, showing that the most critical [13] Soden PD, Kaddour AS, Hinton MJ. Recommendations for design-
location is at the intersection of dome and cylinder. ers and researchers resulting from the world-wide failure exercise.
Comp Sci Tech 2004;64:589–604.
Despite additional lid forces acting near the dome open- [14] Pahr DH, Schuecker C, Rammerstorfer FG, Pettermann HE.
ing, failure in this region is unlikely. Comparison of com- Numerical investigations of perforated laminates in the presence of
bined and variable risk parameters emphasizes the residual ply stresses. J Comp Mat 2004;38:1977–91.
importance of taking residual stresses into account. Fur- [15] Palanterä M. Theoretical background of ESAComp analyses, version
thermore, failure modes and fracture plane angles are 1.5. Helsinki University of Technology, Finland, 1999.
[16] Choo VKS. Effect of loading path on the failure of fibre reinforced
predicted by the Puck failure criterion. Finally, the valid- composite tubes. J Comp Mat 1985;19:525–32.
ity of using the superposition method in the presented [17] Lee CS, Hwang W, Park HC, Han KS. Failure of carbon/epoxy
example is discussed. composite tubes under combined axial and torsional loading 2.
Fracture morphology and failure mechanism. Comp Sci Tech
Acknowledgement 1999;59:1789–804.
[18] Puck A, Kopp J, Knops M. Guidelines for the determination of the
parameters in Puck’s action plane strength criterion. Comp Sci Tech
The funding of the Austrian Aerospace Research 2002;62(3):371–8 (9) 1275.
(AAR)/Network for Materials and Engineering by the [19] Schürmann H. Konstruieren mit Faser-Kunststoff-Verbunden. Berlin
Austrian Federal Ministry of Economics and Labor is Heidelberg New York, Germany: Springer Verlag; 2005.
gratefully acknowledged. [20] Liang C-C, Chen H-W, Wang C-H. Optimum design of dome
contour for filament-wound composite pressure vessels based on a
shape factor. Compos Struct 2002;58:469–82.
References [21] Soden PD, Hinton MJ, Kaddour AS. Lamina properties, lay-up
configurations and loading conditions for a range of
[1] Jones RM. Mechanics of composite materials. second ed. Philadel- fibre-reinforced composite laminates. Comp Sci Tech 1998;58:
phia, PA, USA: Taylor & Francis Inc.; 1999. 1011–22.
[2] Rammerstorfer FG, Starlinger A. Lamination theory and failure [22] Pahr DH, Rammerstorfer FG. A fast multiscale analysing tool for
mechanisms in composite shells. In: Hult J, Rammerstorfer FG, the investigation of perforated laminates. Compos Struct 2004;82:
editors. Engineering mechanics of fiber reinforced polymers and 227–39.

You might also like