-
Failure Analysis of a Quasi-isotropic Laminated Composite Plate
with a
Hole in Compression
Dr. O.H. Griffin, Jr.
by
Nirmal Iyengar
'
Thesis submitted to the Faculty of the
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Master of Science
in
Engineering Mechanics
Dr. Z. Gurdal, Chairman
June 1992
Blacksburg, Virginia
Dr. S.L. Hendricks
-
FAILURE ANALYSIS OF A QUASI-ISOTROPIC
LAMINATED COMPOSITE PLATE WITH A
HOLE IN COMPRESSION
ABSTRACT
The ability to predict failure of laminated composites in
compression has been doggedly pursued
by researchers for many years. Most have, to a limited extent,
been able to predict failure for a
narrow range of laminates. No means, as yet, exist for
predicting the strength of generic laminates
under various load conditions. Of primary concern has been the
need to establish the mode at
failure in compression. Even this has been known to vary for
fiber and matrix dominated laminates.
This study has been carried out to analyze the failure of
specimens with a hole made of laminates
with various quasi-isotropic stacking sequences. Different
stacking sequences are achieved by
rotating a [±45/90/0]9 stacking sequence laminate as a whole
with respect to the loading axis of the
specimens. Two- and three-dimensional finite element models,
using commercial packages, were
ge~erated to evaluate the stresses in the region of the hole.
Two different compressive failure
prediction techniques based on distinctly different failure
modes have been used. The validity of
these techniques was measured against experimental data of
quasi-isotropic specimens tested.
To investigate the applicability of the failure criteria for
different laminated composite plates,
analyses were repeated for specimens with different stacking
sequences resulting from the rotation
of the laminate.
The study shows the need for the use of three-dimensional
analysis of the stress state in the
vicinity of the hole in order to be able to accurately predict
failure. It also shows that no one mode
of failure is responsible for limiting the strength for all
laminate orientations but rather the mode
changes with change in stacking sequence. The failure of the
laminate with a hole was seen to be
very sensitive to the stacking sequence. Experimental data
presented also shows that the peak
strength obtainable from the laminate analyzed, [±45/90/0]9 , is
going to be in the off-axis
configuration rather than on-axis placement of the stacking
sequence with respect to the loading
direction.
-
! ~
I I I I
I I I I
ACKNOWLEDGEMENTS
The author would like to thank the following people for their,
direct or indirect, contribution to this
work:
Dr.Z. Gurdal, committee chairman, for taking on the author as
his student and for
his advice and patience during the course of this work.
Ors. O.H. Griffin, Jr. and S.L. Hendricks for serving on the
authors graduate
committee.
Many colleagues of the author who helped him through the
learning curve on
PATRAN and ABACUS, notably and
and who were innocent victims of the authors
frustrations of trying to understand compressive failure, and
who in spite of it all
provided valuable insights to the problem.
and for their hospitality during the defense.
The •permanent• occupants of ESM Computer Lab namely,
and who helped maintain a sense of humour when all was not
going right.
, the author's in-laws who provided invaluable family
support
when it was most needed.
and who provided the author and his family with valuable
professional and domestic assistance whenever needed.
, of NKF Engineering, who showed the author that curiosity is
a
trademark of a good engineer, and that the need to learn should
never stop.
for their yoghurt, after all one has to eat, and healthy
too.
Finally the last but, most definitely, not the least, to the
authors wife , who
provided the moral and financial support through it all, his
daughter , who
is still trying to figure out why she sees her father only every
three weeks, and
their dogs, . and , who provided the family security and
companionship
during the past few years.
iii
-
TABLE OF CONTENTS
1
INTRODUCTION...................................................................................................................
1
1.1 Compression Failure of Laminated
Composites.............................................................
2 1.2 Failure of Panel with Stress
Raisers...............................................................................
3 1.3 Review of
Literature.........................................................................................................
3
1.3.1 Failure Response of Laminated Composites in
Compression.................................. 4 1.3.2 Failure
Criteria for Laminated Composites
with Stress Concentrations in Compression
................................................................ 10
1.3.3 Summary of Literature Review
...................................................................................
13
1.4 Compression Strength Test
Data....................................................................................
13 1.5 Scope of Investigation
......................................................................................................
19
2 :'FINITE ELEMENT MODELS
................................................................................................
21
2.1 Introduction
.......................................................................................................................
21 2.2 Model Description
............................................................................................................
23
2.2.1 Two-Dimensional Model.
............................................................................................
23 2.2.2 Three-dimensional Model.
..........................................................................................
25
2.3 Boundary Conditions
........................................................................................................
27 2.3.1 Symmetry
....................................................................................................................
27 2.3.2 Traction Free Boundary
Conditions...........................................................................
29
2.4 Comparison Between 2-D and 3-D Models
....................................................................
32
3 FAILURE CRITERIA
.............................................................................................................
40 3.1 Introduction
.......................................................................................................................
40 3.2 Mechanical Properties
......................................................................................................
41
3.2.1 Characteristic Distance
...............................................................................................
42 3.2.2 Buckling Wavelength
..................................................................................................
43
3.3 Failure Prediction
.............................................................................................................
45
iv
-
TABLE OF CONTENTS (contd)
3.3.1 Kinking Model.
............................................................................................................
45
3.3.2 Delamination Criterion ..................................
;················-'············································
50
4 STRESS ANALYSIS
.............................................................................................................
52 4.1 Introduction
.......................................................................................................................
52
4.2 Two-Dimensional Model Stresses
...................................................................................
57
4.3 Three-Dimensional Model
Stresses................................................................................
63
4.3.1 Three-Dimensional Model Stress Correction
.............................................................
75
5 FAILURE ANALYSIS
...........................................................................................................
80
5.1 Introduction
.......................................................................................................................
80
5.2 Initiation of Kinking
...........................................................................................................
82
5.2.1 Two-dimensional Model.
.............................................................................................
82
5.2.2 Three-dimensional Model..
.........................................................................................
83
5.3' Inception of Delamination
................................................................................................
90
6 CONCLUSIONS AND RECOMMENDATIONS
....................................................................
93
7 REFERENCES
......................................................................................................................
95
VITA.....................................................................................................................................
99
v
-
LIST OF FIGURES
Figure 1-1: Kink Band Formation
..................................................................................................
7
Figure 1-2: Laminate Configuration
...............................................................................................
14
Figure 1-3: Compressive Strength of [45/-45/90/0bs Laminate
.................................................... 16
Figure 2-1: Plate Specimen Dimensions and Loading
.................................................................
22
Figure 2-2: Finite Element Model.
.................................................................................................
24
Figure 2-3: Comparison of cr33 and -r23 in Half and Quarter
Symmetric models .......................... 28
Figure 2-4: 3-D finite element model - undefromed state
............................................................ 30
Figure 2-5: 3-D finite element model - deformed state
................................................................
31
Figure 2-6: Through the thickness strain distribution for 2-D
and 3-D models ........................... 33
Figure 2-7a: Comparison of strains between 2-D and 3-D models,
0=0 and =O ....................... 34 Figure 2-7b: Comparison of
strains between 2-D and 3-D models, 0=90 and =O
..................... 34 Figure 2-7c: Comparison of strains between
2-D and 3-D models, 0=-45 and =O .................... 35 Figu~.e
2-8: Comparison of .stresses obtained from 2-D and 3-D Models at q,
= 0.... ...... ......... 36 Figure 2-9: Comparison of stresses
obtained from 2-D and 3-0 Models at q, = 45............ ...... 36
Figure 2-10: Comparison of stresses obtained from 2-D and 3-D
Models at q, = 90 .................. 37 Figure 2-11: Comparison of
stresses obtained from 2-D and 3-D Models at q, = -45
................. 37 Figure 3-t: Model of a fiber under combined
axial and shear forces ..........................................
44
Figure 3-2: Forces on fiber adjacent to hole
................................................................................
49
Figure 4-1a: Distribution of cr11 around hole at mid laminae at
= 0 .......................................... 53
Figure 4-1b: Distribution of -r12 around hole at mid laminae at
= 0 .......................................... 53
Figure 4-2a: Distribution of cr11 around hole at mid laminae at
= 90 ......................................... 54
Figure 4-2b: Distribution of -r12 around hole at mid laminae at
= 90 ......................................... 54
Figure 4-3a: Distribution of cr11 around hole at mid laminae at
q, = -45 ........................................ 55 Figure 4-3b:
Distribution of -r12 around hole at mid laminae at = -45
........................................ 55
Figure 4-4a: Distribution of cr11 around hole at mid laminae at=
45 ......................................... 56
Figure 4-4b: Distribution of -r12 around hole at mid laminae at
= 45 ......................................... 56
Figure 4-5a: Distribution of cr11 around hole at mid laminae at
= 15 ......................................... 58
Figure 4-5b: Distribution of -r12 around hole at mid laminae at
= 15 ......................................... 58
vi
-
LIST OF FIGURES (contd)
Figure 4-6a: Distribution of a11 around hole at mid laminae at
cl>= 0 .......................................... 59 Figure
4-6b: Distribution of t 12 around hole at mid laminae at cl> =
0................................... ....... 59 Figure 4-6c:
Distribution of t 13 around hole at mid laminae at cl> =
0.......................................... 60 Figure 4-7a:
Distribution of 0 11 around hole at mid laminae at cl>= 45
......................................... 61 Figure 4-7b:
Distribution of t 12 around hole at mid laminae at cl>= 45
......................................... 61 Figure 4-7c:
Distribution of t 13 around hole at mid laminae at cl>= 45
......................................... 62 Figure 4-8: Variation
in axial stress, a 11 with laminate orientation (cl>= Oto 45)
......................... 64 Figure 4-9: Variation in axial stress,
a11 with laminate orientation (cl>= Oto -45)
........................ 64 Figure 4-10: Variation in shear stress,
t 12 with laminate orientation (cl>= Oto 45)
...................... 65 Figure 4-11: Variation in shear stress, t
12 with laminate orientation (cl>= Oto -45)
..................... 65 Figure 4-12: Variation in shear stress, t
13 with laminate orientation (cl>= Oto 45)
...................... 66 Figure 4-13: Variation in shear stress, t
13 with laminate orientation (cl>= Oto -45) ....................
66 Figure 4-14a: Distribution of a11 around hole at mid interfaces
at cl>= 0 ...................................... 67 Figure
4-14b: Distribution of t 12 around hole at mid interfaces at
cl>= 0 ...................................... 67
z Figure 4-14c: Distribution of t 13 around hole at mid
interfaces at cl>= 0 ...................................... 68
Figure 4-15a: Distribution of a11 around hole at mid interfaces at
q, = 0 ...................................... 69 Figure 4-15b:
Distribution of t 12 around hole at mid interfaces at cl> = 0
...................................... 69 Figure 4-15c:
Distribution of t 13 around hole at mid interfaces at cl> = 0
...................................... 70 Figure 4-16: Variation in
axial stress, a 11 , at midplane, with laminate orientation
(cl> = O to 45)
.................................................................................................................
72 Figure 4-17: Variation in axial stress, a11 , at midplane, with
laminate orientation
(cl>= 0 to -45)
................................................................................................................
72 Figure 4-18: Variation in shear stress, t 12, at midplane, with
laminate orientation
(cl> = Oto 45)
.................................................................................................................
73 Figure 4-19: Variation in shear stress, t 12, at midplane, with
laminate orientation
(cl>= Oto -45)
................................................................................................................
73 Figure 4-20: Variation in shear stress, t 13, at midplane, with
laminate orientation
(cl>= Oto 45)
.................................................................................................................
74 Figure 4-21: Variation in shear stress, t 13, at midplane, with
laminate orientation
(cl>= 0 to -45)
................................................................................................................
74
vii
-
LIST OF FIGURES (contd)
Figure 4-22a: Comparison of 't°12 and t 12 for 9=0 and ip:o
.......................................................... 77
Figure 4-22b: Comparison of 't°12 and t 12 for 9=0 and' fP=15
......................................................... 77 Figure
4-22c: Comparison of t 0 12 and t 12 for 9=0 and fP=15
......................................................... 78
Figure 5-1: Compressive failure distribution based on fiber
kinking (2-D FEM Model) .............. 81
Figure 5-2: Compressive failure distribution based on fiber
kinking (3-D FEM Model).............. 84
Figure 5-3: Compressive failure distribution based on fiber
kinking using corrected t 12 • •• • ••••••• 85
Figure 5-4: Initiation of matrix failure at mid laminae (3-0 FEM
Model) ..................................... 87
Figure 5-5: Compressive failure distribution based on fiber
kinking at the interfaces ................ 88
Figure 5-6: Compressive failure distribution based on fiber
kinking at the interfaces
using corrected t 12 ••• •••••••••••••••••••• •• ••••••••••
••••••••••• •••••••• ••••••• ••••••• ••••• •• ••••• ••••• •• • ••••
•• • •••• ••• 89
Figure 5-7: Inception of delamination
...........................................................................................
91
LIST OF TABLES
Table 1-1 : Experimental Results of [±45/90/0]38 Laminate
Specimens ....................................... 15
Table 2-1: Material Properties of AS1/T3201
..............................................................................
23
Table 3-1: AS1/T3201 Lamina Strengths
....................................................................................
42
viii
-
1 INTRODUCTION
With the increased use of laminated composite materials for
primary structural components, there
exists a need to fully understand their capabilities. What makes
laminated composites attractive
as structural materials is their ability to be tailored to
achieve specific material properties and
strength for a given laminate configuration.
For a material to be successfully utilized for structural
purposes it must satisfy certain fundamental
requirements; one being the ability to maintain structural
integrity in the presence of discontinuities.
Discontinuities, in the form of holes, are necessary in
structural panels to allow for joining, access
openings for maintenance, and piping and electrical
penetrations. The existence of these
discontinuities produce areas of local stress concentration, and
it is necessary to be able to predict
the response of these panels under various loading conditions if
they are to be used for structural
purposes.
Experimental research [1-4] has shown the detrimental effect of
discontinuities such as notches,
holes etc., on the load carrying capacity of laminated
composites. The effect of holes on the load
carrying capacity is greater in laminated composites than in
metals. Being made of brittle fibers,
the ability of high performance laminated composites to undergo
local plastic deformation in an
area of high stress concentration is limited [1]. It has also
been shown [3-10] that the failure
mechanism is more complex in compression loading than in
tension. The modes of failure in
compression are particularly sensitive to the presence of holes,
notches, and other similar stress
raisers. Once the mechanism of failure propagation can be
identified and the load carrying
capability predicted, the material can then be tailored to
accommodate designed discontinuities.
Also the effect of discontinuities, introduced subsequent to
initial design, can readily be assessed.
Introduction 1
-
1.1 Compression Failure of Laminated Composites
Prediction of strength of a laminated composite depends on two
factors, which are material strength
and stacking sequence. The material strength of a laminate is
dependent not only on the properties
of its constituents, but also on a number of other faqors
governed by the manufacturing process.
Flaws incorporated during manufacturing that can affect laminate
strength are non-uniform fiber
distribution, fiber waviness, and matrix voids. Another factor
whose effect on strength is not known
but at present is under investigation is fiber-matrix
interaction. Due to the number of ingredients
influencing compressive strength there is no completely
satisfactory account, as yet, of
compressive failure [6].
In order to assess the compressive strength of a laminate, it is
necessary that the mode of failure
be understood. Studies of failure on plates in compression [1,2,
11-26], for a variety of laminates,
accompanied by experimental verification, have been presented
over the years. None of the
predictions, however, seem general enough to be able to predict
the failure over a range of
laminate configurations. There does seem to be some consensus on
the nature of failure being
localized and was initially attributed to a loss of fiber
stability [11-21 ].
Unlike their isotropic counterparts where stability, whether
global or local, is based on geometric
considerations, stability of laminated composites is to be
evaluated at global, macro and micro
levels [3]. Due to the makeup of fiber reinforced composites,
inhibiting the loss of stability at the
global (orthotropic failure) and macro (lamina failure) levels,
will precipitate it at the micro
(fiber/matrix failure) level. At the micro level the loss of
stability is, as mentioned earlier, only to a
limited extent dependent on the geometry of the fiber. Research
[2,22-24] has shown evidence that
fiber failure may not be due to this •1oss of stability• at the
micro level, but rather a phenomenon
called fiber •kinking• based on observations of kink bands in
the failed specimens.
The theory of compressive strength of a laminate being dependent
on kinking has been studied
by many researchers [2,4,9,10,26]. Experimental evidence has
shown [15,17,20,21] that the
kinkbands originate at a free edge and propagate towards the
center of the plate eventually
resulting in catastrophic failure. Kinking is also said to
initiate from a defect, notch, or stress raiser,
and grows in a manner analogous to a dislocation. Experimental
evidence also seems to indicate
that the initiation of kinking and subsequent catastrophic
failure, are often not very far apart, thus
making it difficult to chart the progress of this phenomenon [21
]. Recent studies, however, indicate
Introduction 2
-
that failure in compression cannot be attributed to one mode for
all laminates, as suspected earlier,
but is more likely to be progressive in nature [2,27]. These
researchers (2,27] have recorded
delamination along with kinking in specimens in which the
failure process was partly captured by
interrupting loading prior to catastrophic failure. The exact
order and effect of one failure mode on
another has never been evaluated.
1.2 Failure of Panels with Stress Raisers
As mentioned earlier the presence of stress raisers is
detrimental to the load carrying capacity of
a laminated composite panel in compression. The failure of these
panels has, in general, been
assessed at the macro and micro levels. The studies on the macro
level use a fracture mechanics
approach to obtain Stress Correlation Factors (SCF) or Stress
Intensity Factors (SIF) for a variety
of laminates. Experimental studies [28] Were also carried out to
evaluate the effect of finite width
and hole size. These studies were confined to the lo5s in load
carrying capacity and did not
address the mode of failure. To provide a generalized assessment
of the failure, researchers have
focused on the micromechanics in order to evaluate the initial
mode of failure [4,9, 1 O].
~-
Most discussions on failure have been restricted to
unidirectional loading. Many practical loading
situations are unlikely to be unidirectional and the existence
of combined compressive and shear
loads is a reality. Even under unidirectional loading shear
stresses generated due to complex
stresses existing at edges of stress raisers [9, 1 O] brings up
the need to understand the effect of
combined loading on such panels with geometric discontinuities.
The response of a laminated plate
to combined loading can be simulated by analyzing a plate with a
hole. The hole or other
geometric discontinuities provide the complex stress state to be
analyzed, and the failure process
can be scrutinized better by focussing on the region around the
hole.
1.3 Review of Literature
It is apparent that the failure process in compression is very
complex. Lack of knowledge of the
micromechanical behavior of a laminated composite plate under
compression inhibits proper
understanding of the failure process. This, however, has not
deterred researchers in attempting to
analyze the problem. Excellent surveys of existing literature
have been presented by Gurdal (4],
Greszczuk (20], and Shuart (29] to name a few. However, in order
to bring the task on hand into
Introduction 3
-
perspective the process is repeated. The literature survey is
discussed in two parts, namely, the
failure response and the failure criteria of laminated
composites in compression. In the former, the
modes of failure in compression with experimental evidence are
discussed. In the latter, analytical
models and criteria used to predict failure of laminated
composites with geometric discontinuities
have been discussed
1.3.1 Failure Response of Laminated Composites in
Compression
For the past two decades, since it was first recognized, the
failure of a laminated composite in
compression was attributed to short wavelength buckling
(microbuckling) and had been the focus
of many researchers. Most analytical work done so far is based
on the models proposed by Rosen
[11 ]. He proposed two modes under which microbuckling could
take place, the extension mode and
the shear mode. Rosen's model were based on the stability of a
beam (representing the fiber) on
an elastic foundation (representing the matrix). Analytically he
presented the two failure modes as
follows:
~ension Mode: (1.01)
Shear Mode: (1.02)
where:
a,cr - Compressive strength of composite Ei. Em - Elastic
modulus of fiber and matrix respectively V1 - Volume fraction of
fiber
Gm - Shear modulus of matrix
h - Width of fiber
L - Buckling wavelength of fiber
m - Fiber buckle mode number
Rosen [11] neglected the second term of the latter equation on
the basis of its magnitude; the
argument being that the buckling wave length to fiber diameter
ratio was very large. Thus reducing
Introduction 4
-
the shear mode equation to :
(1.03)
Rosen [11] however recognized that the predictions obtained by
his models overestimated the
strength of the laminates. He attributed this difference to the
possible loss of shear modulus
resulting from plasticity of the matrix [6]. Schuerch [12] also
obtained similar equations as Rosen,
but he extended his model to include inelastic microbuckling as
he noted that the strain levels
exceeded the yield strain of his matrix.
Hayashi [13] introduced the concept of "shear instability"
associated with compressive loading. He
proposed the idea that in the case of structural members with
high flexural rigidity compared to
shear rigidity, buckling along with shear deformation will take
place. The buckling strength is
therefore attributed largely to the shear modulus of the
laminate. Foye [14] was also soon to come
to the conclusion that, for unidirectional composites, the
ultimate strength in longitudinal
compression is limited by. their shear modulus. He obtained a
modified stress strain law for an
element under shear, indicating that in compression the
effective shear modulus would decrease
With the increase in load.
(1.04)
At failure the element would experience a complete loss of shear
stiffness resulting in shear
instability or crippling failure. Noting that his predicted
strengths were higher than experimental
values, he included in his analysis model the effect of voids
and matrix fillers.
In order to fully understand the behavior of a fiber in an
elastic matrix Hermann and Mason [15]
and Sadowsky et al. [16] experimented with single fibers
isolated in a matrix to study short
wavelength buckling of the fiber. Hermann and Mason [17] found
that predicted failure of their
model related well to Rosen's [11] extension mode. They also
showed that the initial waviness of
the beam (fiber) had a significant impact on failure. Sadowsky
et al. [16] took into consideration
residual thermal stresses, sensing that these stresses could
cause waviness and fiber buckling and
came to the conclusion that shear deformation had negligible
effect on the compressive strength.
Introduction 5
-
They also measured the strain during buckling and found that the
predicted strains were more than
an order of magnitude greater than that measured experimentally.
Crawford [17) was the first to
incorporate initial waviness of the lamina into the expression
of stiffness. This resulted in
interlaminar stresses being generated. Based on this, he
concluded that the laminate failure was
a consequence of these interlaminar stresses, rather than short
wavelength buckling, though the
exact mode of such a failure was not suggested.
To further study the effect of the matrix in fiber buckling,
Lager and June [18) tested Boron fibers
in two epoxy systems. They based their model on that of Rosen's
[11 ], but replaced the Elongation
and Shear Modulii in Equations 1.01 and 1.03 by their respective
Tangent Modulii. Based on
experimental data they then replaced the tangent modulii by an
"effective" modulus which is
defined as a product of an "influence coefficient" and the
material modulus. The value of the
influence coefficient was determined experimentally to be 0.63.
They also established that the
strength of a composite in compression is largely dependent on
matrix strength. The study also
showed that for fiber volume fractions less than 10%, failure
correlated well with the extension
model while for large volume fractions the shear mode was
observed. Since most of the
experimental work was still over predicting the failure of a
laminated composite in compression De
F~rran and Harris [19] conducted experiments on polyester
reinforced steel wire. They questioned
the validity of Rosen's premise, that planar buckling was
occurring. They concurred with what Lager
and June [19) had alluded to, in that the buckling had an out of
plane component and so
considered a three-dimensional helical buckling approach. The
hypothesis, of microbuckling being
affected by interlaminar stresses, having been presented it was
not until much later that Guynn and
Bradley [27] and Waas et al. [2] would provide experimental
evidence that in- and out-of-plane
microbuckling does indeed occur. This type of microbuckling
could be a result of the interlaminar
stresses.
To investigate the effect of matrix properties on the strength
of a laminated composite in
compression Greszczuk [20,21,30) performed many experiments by
varying the matrices and
volume fractions and considering different failure modes to come
to the following conclusions. For
composites with a low-modulus resin, failure in compression is
generally governed by
microbuckling, however, as the resin modulus is increased the
transformation to a non
microbuckling failure takes place. His data showed the effect of
fiber matrix interface strength on compressive strength indicating
an interaction type failure. Nonlinear material and geometric
effects
of the fiber and matrix were not considered.
Introduction 6
-
T' T' -····~
_______ ...
. 111 " 1111
T'
Figure 1-1: Kink Band Formation (After Ref. 22)
7 Introduction
-
At this point, with still no means for accurately accounting for
the compressive strength Berg and
Salama [22) through their studies of fatigue of graphite epoxy
in compression, introduced the
concept of kinking, to laminated composites. They showed the
existence of kink bands due to
microbuckling, in the specimens failed as a result of
compressive fatigue. They concluded that axial
matrix cracking was essential to the initiation of microbuckling
instability. This longitudinal cracking
permits kink bands on conjugate planes to link up leading to
catastrophic failure. They postulated
a kink process by which microbuckled fibers undergo shear
deformation to the point that they break
off into bits (Figure 1-1) due to tensile failure of the
fibers.
Based on work by Berg and Salama [22), and Weaver and Williams
[23), Evans and Adler [24)
produced an exhaustive study on kinking. They minimized the
plastic work done to obtain the kink
inclination and minimized the elastic strain energy to determine
the kink boundary. They also
utilized a model for statistical fiber fracture and matrix
enhancement to determine parameters and
an expression for critical kink formation stress. This
expression confirmed the need for high matrix
strength to suppress kink formation. Chaplin [1] attributed the
low compressive strength to the
presence of a defect from which local failure can propagate. He
suggested use of fracture
mechanics approach observing that the failure occurs at one
region and then rapidly propagates
across the specimen. Defects which act as stress concentrators
are responsible for the initial ·:;;·
failure. Chaplin [1] compared the failure to shear instability
that occurs in un-reinforced resin and
so felt the term microbuckling to be inappropriate.
Utilizing the above information Parry and Wronski [25)
investigated the compressive strength of a
plate with a notch. They supported the Berg and Salama [22)
claim that the principal mode of
failure at the notch was due to the formation of a kink band.
Later, experiments carried out by
Rhodes et al. [26), and Waas et al. [2] also support this
assertion. Rhodes [26) described the failure
sequence beginning with matrix failure at 85% load which was
interfiber in nature. Following matrix
failure, upon further loading shear crippling in plies of the
same orientation were initiated resulting
in specimen failure. Waas et al. [2] also observed failure at
the hole edge in the load carrying ply
(0 degree) and at the location of the maximum compressive
stress. They also found the
microbuckling to originate at the hole surface and persist well
into the interior of the specimen
leading to catastrophic failure.
Departing from the traditional view of fiber failure, Shuart and
Williams [31] conducted experiments
on angle ply laminates, with ±45 degree and ±45 degree dominated
plies, with holes to conclude that the failure of the former was
due to matrix shear and the latter was a combination of matrix
Introduction 8
-
failure and delamination. On close inspection of the damaged
specimens they observed kink band
formation and transverse matrix cracking. Continuing work in
compression failure of laminated
composites, Shuart [29] developed a model with the fibers in the
lamina as a plate and the matrix
acting as an elastic foundation. He used linear analysis to
determine stress, strain and mode shape
using short wave buckling criteria. He extended this study to
include the effect of volume fraction '
on the buckling stress and then incorporated the effect of fiber
imperfections in a non-linear
analysis. He considered,the failure of the lamina based on the
buckling of the outer lamina, in-
plane matrix shearing and interlaminar shear strains from the
fiber imperfections. He also provided
a method to evaluate the load carrying capacity of a O degree
dominated lamina in compression.
Shuart [32] extended his previous non-linear model to include
the effects of out-of-plane ply
waviness, in-plane fiber waviness and fiber scissoring. He used
this model to predict the failure
of [±9/•9ls. composite laminates in compression. Using
experimental data, he showed his model
to provided excellent agreement ,for laminates with fiber angles
greater than 45 degrees, 9 > 45, while that for those with fiber
angles less than 45 degrees, 9 < 45, was not as good. Again he
found the dominant failure modes to be interlaminar shearing,
in-plane matrix shearing and matrix
compression failure.
So far the failure of a laminated composite in compression had
been attributed to one failure mode.
The possible interaction of two modes of failure were first
recorded by Guynn and Bradley [27].
Guynn and Bradley [27] provided evidence, based on experiments
on compressive failure of
AS4/PEEK laminate specimens with open holes, of the existence of
two failure modes but
discounted their interaction. Based on their observations of
specimens that were inspected on
interruption of the tests, prior to catastrophic failure, that
the principal mode of failure was shear
crippling. They also found evidence of in- and out-of-plane
microbuckling prior to the shear
crippling, which was also accompanied by local delamination.
They concluded that this local
de lamination was a result of the shear deformation of the
fibers, caused by microbuckling, however,
they did not believe that the delamination propagated to become
macroscopic delamination, until
final compressive failure when large scale brooming was
observed. Waas et al. [2] on the other
hand observed, depending on the specimens tested, the formation
of delamination buckling prior
to actual final failure. The damage initiated by microbuckling
led to delamination cracks and
suQ&equently to catastrophic failure. The growth of the
delamination was in large part dependent
on the percentage of 0 degree plies. In laminates with a small
percentage of 0 degree plies
delamination size grew till the area buckled leading to failure.
The larger percentage of O degree
plies led to brittle failure while a small percentage of 0
degree plies led to a more ductile failure.
Beyond failure initiation by microbuckling, the delamination and
buckling growth stage was
Introduction 9
-
influenced by the overall stiffness. Waas et al. [2] also
presented initial 0 degree strains which
seemed to initiate failure.
1.3.2 Failure Criteria for Laminated Composites with Stress
Concentrations in Compression
The failure criteria of plates with cutouts, holes, and notches
have been studied for years, however,
most of them relate to tensile loading. Many of the criteria
proposed have been based on failure
at the laminate level without any emphasis on the failure modes.
Whitney and Nuismer [3]
continued work by Waddoups et al. [5] to obtain average and
point stress criteria for the failure of
laminated plates with holes. They presented characteristic
distances for both criteria, based on
work by Bowie [33] on isotropic material. They, however,
postulated that the characteristic
distances were constant for various laminates and
discontinuities. Nuismer and Labor [7] extended
the concepts established above to laminated plates under
compressive loads. However, the
characteristic distance was found to be much greater in
compression than was observed in tension.
They attributed this difference to the greater redistribution of
stress prior to failure in compression.
Garbo and Rogonowski [8] conducted experiments and provided
strength analysis for composite ·:~
laminates under compressive and tensile loading. They postulated
that the failure of a composite
laminate can be correlated with analytical predictions of point
stress at a characteristic distance
away from the edge of the stress concentration. Garbo and
Rogonowski [8] also extended the
above theory to predict strength of a laminate, by utilizing the
point stress criteria to evaluate the
strength of each of the laminae. They recommended that, since,
in a laminated composite, the
peak stress does not occur at the same point at the hole edge,
it is necessary to evaluate the
stresses adjacent to the hole to obtain the combination of
stresses that will lead to failure.
Gurdal [4] expanded the work on failure criteria of laminated
plates with cutouts, holes and notches
by combining micromechanics and the laminate point stress
criterion. Using information provided
by Rhodes et al. [26], on the occurrence of kinking at the
boundary of a stress raiser and the work
by Greszczuk [21,30] on point stress criteria for plates with
holes, he recommended that the two
could be coupled to evaluate the failure strength of plates with
notches. He also noticed that most
of the prior work had restricted itself to the effect of the
local compressive stresses only. Supported
by Greszczuk's [21] observations that microbuckling occurs at a
higher load than experimentally
obtained, Gurdal [4] suggested that the discrepancy may be
caused by shear stresses induced by
the presence of the discontinuity. He then used an energy
formulation to analyze a beam on an
Introduction 10
-
elastic foundation, subject to in-plane and shear loading and
obtained equations for stress in the
fiber due to bending. Details of the formulation are presented
in Chapter 3. The model assumed
failure due to bending rather than microbuckling instability, as
had been the norm till then. The
shear loading in the model helped account for the shear stresses
generated due to the presence
of the notch. To prove the validity of the model he used the
same variational formulation to do a
stability analysis to obtain a buckling expression similar to
that of Rosen's [ 11 J microbuckling model in the shear mode.
Whereas Rosen [11 J neglected the second term Gurdal [4J retained
it. He
argued that experimental evidence provided by Evans and Adler
[24J had shown that the buckling
wave length to fiber diameter ratio did not justify ignoring the
second term in Equation 1.02. He
used two-dimensional finite element models to obtain the stress
distribution in notched plates and
using experimental data he evaluated the characteristic distance
for the quasi-isotropic laminate
to be 1.5 mm for plates with notches. Similarly for plates with
holes the distance was shown to be
2.5 mm. Gurdal [4J extended his model of plates with cracks to
represent those with holes as more
data was available for that purpose. He supported his decision
by correlating the analysis with
experiments conducted by Haftka and Starnes [9J for laminates
with varying degrees of orthotropy.
The results, however, showed the predicted values to be 40%
greater than the experimental
values. Gurdal and Haftka [1 OJ extended the earlier work to
predict failure of a plate with a hole
}lnder combined loading. They modeled the plate so as to obtain
predictions for various laminate
orientations. This was done by rotating the principal axis of
orthotropy of the laminate with respect
to the load direction. The stresses around the hole were
evaluated using techniques proposed by
Savin [34J and corrected for finite width from Hong and Crews
[28J. The study clearly showed the
anisotropy of strength of a quasi-isotropic plate under
compression. Gurdal and Haftka [1 OJ also
concluded that for a plate under combined in-plane and shear
loading, rotating the laminate axis
of orthotropy in a direction opposite to the applied shear load,
can increase its load carrying
capacity, thus suggesting a possible use of orthotropy for
strength improvements of plates with
holes.
Some other attempts to predict the compression strength of
laminated composites plates with
holes, using more deterministic approaches are described below.
Bums et al. [35J analyzed the
compressive failure of notched angle ply laminated composites
utilizing a three-dimensional finite
element model and compared the results with experimental data. A
point stress method and
classical quadratic tensor and maximum stress failure criteria
were used to evaluate the failure.
Burns et al. [35J showed that failure analysis at the hole edge
gave extremely conservative
strengths. They concluded that the lack of consistent results
was due to the point stress criteria
being used and that failure in a laminate was not dependent on
point failure. They found that the
Introduction 11
-
in-plane shear stress, 't12, was dominant in the midplane with
the out-of-plane shear stress, 't13,
dominant at the interface. Observing the points at which failure
was initiated, they concluded, that
a combination of the two shear stresses were required to
initiate failure. The characteristic distance
theory had limited success in predicting failure.
Chang and Lessard [36] considered the progressive failure of a
laminated composite containing
a hole subject to compressive loading. The stress analysis was
performed using a finite element
model considering the material and geometric non-linearities
based on finite deformation theory.
Failure criteria were also formulated for first ply failure,
based on a semi statistical method and a
damage progression model. Property degradation models for
various failure have been provided.
Good correlation was found when compared with experimental data
for the laminates in question.
The model considered ply thicknesses, clustering and finite
widths, however the possibility of
delamination was not considered.
In a very different approach, Soutis et al. [37] use a modified
Linear Elastic Fracture Mechanics
technique to predict failure of a Carbon Fiber-Epoxy laminate,
[(±45/0J3]., with a hole. Based on
experimental observations, they inferred that failure is
initiated by microbuckling at the hole edge
)Yhich they equated with a "crush zone". This crush zone was
considered to grow stably under
increasing load till a critical length was reached, when
unstable growth results in catastrophic
failure. Using a non-linear approach to crack growth they
predicted good results for the
compressive strength of the laminate, though their prediction of
length of the microbuckled region
(crush zone) was greater than observed. This technique did not
in any way address progressive
failure and, being based on a fiber failure mode only, may have
its limitations when utilized for
laminates without fiber dominant layers.
Considering the hypothesis that interlaminar stresses may play a
part in compression failure of
laminated composites, and that the magnitude of these stresses
[38] in the vicinity of a hole cannot
be ignored, Lagace and Saeger [34] proposed a failure criteria
based on the inception of
delamination at a hole edge. They used a methodology based on an
eigen function solution
technique to evaluate the interlaminar stresses near a hole for
a plate in tension. They then used
Hashim-Rotem type quadratic failure criteria proposed earlier
(Lagace and Brewer [39]) to
determine the load at the onset of delamination. They plotted
the load at onset of delamination
versus the laminate orientation, ~. to show the anisotropic and
asymmetric distribution of strength of the quasi-isotropic
laminates. The failure criterion was based on a single point stress
and no
consideration was given to progressive failure initiated by
either microbuckling or matrix cracking.
Introduction 12
-
1.3.3 Summary of Literature Review
In summary, the literature review shows the extensive research
done on the failure prediction of
laminated composites in compression. Most of the research,
however, has been restricted to the
compressive strength of unidirectional laminates a~ has been
based on the assumption of short
wavelength fiber buckling (microbuckling) in the plane of the
lamina. Attempts at correlating
analytical with experimental data by incorporating manufacturing
defects such as voids, fiber
waviness and scissoring and debonding have been studied.
Research has also covered the failure
of the fiber as a result of kinking.
Based on evidence that the mode of failure in compression is
related to the type of laminate (fiber
or matrix dominated), research on matrix initiated failures have
also been done. Having accepted
the fact that compression failure of a laminate is initiated at
a notch or similar stress raiser
researchers have developed models to predict compression
strength of laminates with holes with
reasonable success. The influence of interlaminar stresses on
compression strength has also to
a limited extent been covered, though it has long been suspected
to play a part in the failure
mechanism.
Most of the above studies were based on planar failure models
and those that studied interlaminar
stresses did not include in-plane failure modes. As a result, in
spite of the extensive work being
done there is no general means of predicting failure of a
laminated composite with a hole in
compression.
1.4 Compressive Strength Test Data
As discussed in the previous section, existing techniques for
the prediction of strength of laminated
composite plates with holes were applicable to only specific
laminate systems, with none being
general enough to be applicable for a range of laminate
configurations. The above prediction
techniques are very sensitive to specific laminate
configurations, whether unidirectional or angle
plies. In order to assess the capability of predictive models
for laminates in compression, it was necessary to obtain
experimental data with which the analytical predictions of failure
could be
compared. Also in order that the generality of the analysis
technique could be demonstrated, it was
decided that the use of a quasi-isotropic laminate was
necessary. Earlier models had been used
for predicting the strength of unidirectional and angle ply
laminates separately; the response of a
Introduction 13
-
Figure 1-2: Laminate Configurations
Introduction 14
.I
-
Table 1-1: Experimental Results of [±45/9010],. Laminate
Specimens
Orientation Number of Off-axis Failure Loads Average Failure
Angle Specimens of Specimens Loads
,
-80 2 6728.8 7105.3 7481.7
-67.5 4 6405.9 6640.2 6700.2 6875.0 6875.0
-45 5 6446.8 6941.2 6703.2 6647.0 6813.96 6667.2
-22.5 3 . 8296.5 7923.3 8078.1 8014.5
-15 5 8126.5 7800.9 . 7919.9 7941.9 7803.0 7926.9
-10 5 8155.2 8416.7 8256.8 8227.5 8218.9 8265.4
-5. 5 7561.6 n38.o 7597.0 7492.0 7857.0 7336.1
0 5 7200.0 7080.0 7199.8 7266.2 7122.8 7330.0
10 5 7042.0 6n9.8 6940.7 6872.6 7094.1 6914.9
22.5 5 6756.9 7078.8 6970.6 6693.4 7378.8 6945.2
45 5 6929.6 7189.6 6740.5 6861.3 6325.4 6396.8
Introduction 15
-
--T·--r·---]-1--r-
l l
i l
i :
: :
: :
l 1
l l
~ ....................... i ........................ !
....................... j··· ··················1·······--......
o-······f···············-···
l l
0 co
0 C")
0 0 C") I
0 co I
0 O> I
G)
- ca c ·-E ca ..J en C") ...... 0 - 0 0) - It) oi::t' I - It)
-&
-oi::t' .....
c:: 0 -..::::; - 0 as .... 0
.c -
.... c::n
- 0 Q) c G) ..
15> c::
-"' <
( G
) >
·c;; en G) .. c. E 0 CJ Cf) I .
I T""
G)
I .. ~
I c::n i!
16
-
quasi-isotropic laminate is a combination of the above two
configurations. Further the generality
of the technique could be corroborated by applying it to various
laminate orientations that will be
described later.
Results of a test program carried out earlier [40], for
quasi-isotropic laminates of [±45/90/0b. and
[±45/90/0k. stacking sequences are presented here (See Table 1-1
). Specimens in Reference [40]
were cut from a large panel with a [±45/90/0J. layup so that
laminates with varying angles of
laminate orientation, cl> (as defined in Figure 1-2), were
obtained. That is, a specimen cut at cl>= 15 degrees would have
a stacking sequence of [60,-30,-75,15]. with respect to the
principal axes
of the specimen. This layup is still quasi-isotropic, however it
is suspected that the strength of the
specimen would be different than a specimen cut at cl> = O
degrees because of the change in the stacking sequence which would
affect the local stresses around the hole. The loading direction
was
kept constant and the laminate orientation cl> changed; this
was analogous to keeping the laminate orientation constant and
varying loading angle thus simulating the effect of combined shear
and
axial loading on the laminate. The counter clockwise direction
was considered to be a positive
rotation and is referred to as the laminate orientation cj>.
The specimens were 1.5 inches long, 1.0 inch wide and 0.133 inch
thick (24 ply) with a circular hole of 0.20 inch diameter.
'The compressive strength obtained at off-axis laminate
orientations (cl>* 0) were normalized by the strength at cl>
= O and shown in Figure 1-3. On plotting the normalized strengths,
P/P 0 versus laminate orientation, cj>, the curves obtained for
the specimens with two laminate configurations tested were
practically coincidental with one another. The only difference
between the two laminate
configurations being the total number of plies, and therefore
the total laminate thickness. For this
reason the following discussion has been based on the
experimental data points obtained from the
testing of [±45/90/0b. laminate specimens only. The results of
the experimental data shown in
Figure 1-3 are described below:
• The single roost distinctive feature of the failure envelope,
at laminate orientations -90 <
cl> < 90, is that its is anisotropic and asymmetrical
about cl> = 0. By anisotropic, it is meant that the strength
varies as the laminate is rotated about the z-axis.
• The failure loads at cl> = 45 and cl> = -45 are
approximately the same, within experimental scatter.
• The trend at the extremes indicates that the load at cl> =
90 would be close to or equal to that at cl>= 0.
• The failure load (strength) peaks at about cl> = -15
degrees and is approximately 12% greater than that at cl> = O
and about 19% greater than the value at cl> = 15 degrees. This
is rather unusual because based on CL T, the axial stiffness of the
laminate does not
change if it is rotated in the +15 or -15 degree direction. At
the lamina level the axial
Introduction 17
-
" I ~ If
1:
I I I
•
stiffness coefficients, A1• are the same for a lamina with a
fiber angle of +0 as that with -a. The difference, however, in the
two laminate orientations,~= 15 and -15 is the stacking sequences.
In these orientations the former becomes a [60,30,-75,15}. laminate
and the
latter [30,-60,75,-151. laminate respectively. As can be seen in
these two laminate
configurations, for each 0 in the former there exists a -0 in
the latter and vice versa. So the lamina stiffness coefficients of
the two are the same. The relative through-the-thickness
locations of the two laminates are different. The stacking
sequence seems be the only
feature that probably accounts for the difference in compressive
strengths of the two
laminate orientations. It can be conjectured that the strength
of the laminate in
compression is not only a function of the total laminate
stiffness but also one of the relative
location, i.e. stacking sequence, of the various axial stiffness
components. It is for this
reason that it is essential to determine the three-dimensional
stress state for this laminate
in various orientations. A two-dimensional model is not able to
capture these subtle
differences.
Another manner in which the stacking..sequence can be considered
to have significant
impact on compressive strength is in the through-the-thickness
location of the primary load
bearing laminae. This can be seen by comparing the orientations
at angles in steps of 45
degrees, that is at ~ = 0, 45, 90 and -45 degrees. It seems to
indicate that when the primary load carrying laminae, i.e. one in
which a = ~. is at the top and bottom of the laminate during
rotation, there is a great reduction in the strength. For example,
when ~ = 0 the laminate configuration is [±45/90/0J. while in the ~
= ±45 orientations they are
[0/90/±45]. and [90/0/•451. respectively. In the former
condition (~ = 0) the primary load
bearing plies are at the midplane of the laminate while in the
latter (~ = ±45) they are on or near the surface. This could result
in out-of-plane buckling due to the lack of transverse
support for the surface layers. This hypothesis is in part
supported by experimental
observations of Guynn et al. [27), who suggest that the
out-of-plane buckling does occur
and may be due to the influence of transverse stresses on the
laminate.
The asymmetry of compression strength has tremendous
significance to real life applications. In
order to optimize the design of a quasi-isotropic material for a
structure that may experience
tension and compression loading it may be necessary to utilize
the material in an off-axis
configuration. Sun and Zhou [41] have also presented research
showing the anisotropic behavior
of tensile strength in off-axis configurations of
quasi-isotropic laminates. The angle of laminate
orientation, ~. exhibiting maximum strength in tension, was
different from that shown in Figure 1-3 for compression.
Introduction 18
-
1. I
1.5 Scope of Investigation
There seems to be adequate evidence supporting the hypothesis
that compression failure of a
plate with a hole is initiated by fiber kinking at the hole
surface [2,4,9, 10,27,37]. This kinking leads
to fiber breakage causing kink bands and eventually resulting in
catastrophic failure. Research
[4,9, 1 O] shows the significant contribution of shear to this
mode of failure. Evidence of this is more
apparent when considering laminates that have no O degree plies
in the loading direction, i.e. all
plies are angle plies. Even under a uniaxial compressive load
combined compressive and shear
stresses are developed due to the stiffness coupling in the
laminae and the presence of geometric
discontinuities. Analyses to date have had reasonable success in
predicting the failure strength of
specific laminated composites under a compressive load, though a
more general approach is
desired. The objective of the present work is to evaluate two
existing models of compression failure
and attempt to predict the strength of a laminated composite
plate with a hole under varying
laminate orientations, ~·
As discussed earlier, to date most work on microbuckling and
kinking has utilized a two-
dimensional approach based on the assumption that the response
is planar. Based on inference
·from limited experimental data, as discussed in Section 1 .4
above, that the variation in normalized
strength, P/P 0 , was the same for the two laminate
configurations tested regardless of their
thickness, and for the sake of simplicity, this study is based
on the analysis of a [±45/90/0].
laminate. Also for the reasons discussed in Section 1.4 above,
in this study a three-dimensional
linear finite element model of the above laminate is used to
analyze the stress distribution around
a hole. Comparison between the stress distribution in this model
and a similar two-dimensional
model is presented. The stresses from the finite element models,
along with the kinking model
proposed by Gurdal [4], are used to predict failure of a
quasi-isotropic laminate under compression.
The kinking model failure criteria is applied to the stresses at
the edge of the hole to evaluate
failure of the laminate, at varying angles of rotation of the
quasi-isotropic laminate. The analysis
is then compared with experimental data available as described
in Section 1.3. As it was suspected
that transverse stresses may contribute to compressive failure,
the kinking model was modified to
incorporate them. Also in order to fully evaluate the effect of
the inter1aminar stresses, the onset
of delamination was evaluated based on a criteria proposed by
Lagace and Saeger [34]. Even
though it is suspected [2,39] that the failure is progressive in
nature, the present study ignores the
progressive nature of failure and does not use ply discount
technique. Also interaction between the
two modes of failure will not be considered.
Introduction 19
-
In Chapter 2 the two finite element models, two-dimensional and
three-dimensional, are described,
while Chapter 3 discusses the kinking and delamination models
along with the failure criteria that
have been used. In Chapter 4 the stresses obtained from the two
finite element models have been
discussed while Chapter 5 discusses the result of the failure
analysis using criteria and models
presented in Chapters 2 and 3. Conclusions and recommendations
have been covered in Chapter
6.
Introduction 20
-
2 FINITE ELEMENT MODELS
2.1 Introduction
In order to compare the effect of different analysis models on
the failure prediction of a plate with
a hole loaded in compression, two finite element models were
generated. The first is a two-
dimensional model which could account for in-plane stresses
only. The second model is a three-
dimensional model capable of providing all six stresses at a
point. The primary objective of creating
the two-dimensional model was to provide a basis by
corroborating data obtained by Gurdal [4],
as to the effect of in-plane compressive and shear forces on the
failure strength of the laminated
plate in compression. Due to discrepancies between the
analytical and experimental values
observed by Gurdal and Haftka [1 O], the natural progression
seemed to be the study of the effects
of out-of-plane stresses on the failure of the plate. It has
long been suspected [18] that these
stresses may play a significant role in the mechanism
contributing to failure of the abovementioned
plate due to the out-of-plane nature of the kink formation.
Recently experimental evidence by
Guynn et al. [27] have substantiated this claim. In addition to
the out-of-plane microbuckling/kink
band formation, researchers [2,27,37] have also detected the
existence of delamination during the
failure process. The prediction of the inception of delamination
also requires evaluation of the
transverse stresses which makes a three-dimensional analysis
necessary.
Many techniques have been put forth for the evaluation of the
complex stress state along the edge
of a hole in a plate. These vary from being based on an
elasticity approach [42], to using
Lekhnitskii complex function based solution [43], and finally
the three-dimensional finite element
approach. Of the above, the three-dimensional finite element
approach was chosen in order to
compare and contrast with similar techniques (two-dimensional)
utilized earlier. It will be shown
later that the use of this model not only provides the
out-of-plane stresses but also provides in-
plane stresses that are quite different those obtained from a
similar two-dimensional model.
Finite Element Models 21
-
y
c .£ x 0
~ .-
------- 1 .50 in --------
z
0.048 in
Figure 2-1: Plate Specimen Dimensions and Loading
Finite Element Models 22
-
Both finite element models were generated using a commercially
available software, PATRAN [44)
(version 2.5), for pre- and post-processing while the finite
element analysis was carried out using
ABAQUS [45) (version 4.8). The dimensions of the plate are as
shown in Figure 2-1. The
distribution of stresses around the hole are described by angle
a., with a. = 0 being at x = 0.1 O and y = 0.00. The laminate
orientation, as described in Figure 1-2, is denoted by cj>. The
properties of the laminae are assumed to be hornogenous and are
based on those of a commercially available
graphite/epoxy (AS1/T3201) provided in Table 2-1. The layup of
the plate is [±45/90/0). which
exhibits quasi-isotropy at the laminate level.
Table 2-1: Material Properties of AS1/T320f
E11 E22 E33 \>12 \>13 \>23 G12 G13 G23
20.0 1.30 1.25 0.30 0.30 0.49 1.03 0.90 0.90
* All rnodulii in msi
The above properties are not the exact properties of the
specimens described in Chapter 1, but
rather reflects those of similar commercially available.
material.
2.2 Model Description
2.2.1 Two-dimensional Model
The two-dimensional model consists of a 0.048 in thick plate in
which the properties of the various
laminae were smeared to provide the global properties of the
laminate. By virtue of the layup it is
quasi-isotropic in the x-y plane. The analysis of the
two-dimensional model is based on Classical
Lamination Theory (CL T) and a plane stress state is considered.
Since no through-the-thickness
variation of strains are allowed, a single element was used to
model the laminate in the z direction. Eight node quadratic
elements were used in this model, except at places where due to
geometric
necessity (i.e. at the fine-coarse mesh interface), triangular
elements were utilized. These elements
consist of quadratic lagrangian interpolation functions, so are
capable of providing better
Finite Element Models 23
I 'i
'' '
-
r
Finite Element Models
-G> "C 0 E ... c G> E G> -G> G> ~ c ii: .. N
I N
~ ::J C> ii:
24
-
compliance to the plate in the vicinity of the hole, than linear
lagrangian function elements. Since
interlaminar stresses are ignored in Classical Lamination Theory
(CL T), considerable freedom can
be exercised in generating this model due to the numerous planes
of symmetry which will be
elaborated on later. However to facilitate comparison, the nodal
distribution of the two models in
the x-y plane has been kept the same. The nodal points were
located so as to limit computation
time and also prevent distortion of the elements. The maximum
aspect ratio (Vt) of the elements
was 41.667, which is well within the limits for these quadratic
elements, as established by
Vidussoni [46].
The distribution of the nodes at the point of interest,
surrounding the hole, has been kept fine and
then gradually expanded away from it. The nodes were distributed
radially along the hole to a point
0.15 inches, beyond which the distribution was on a cartesian
basis as shown in Figure 2-2. The
nodal distribution was chosen so as to be sensitive to the
radial variation of stresses along the hole
edge.
2.2.2 Three-dimensional Model
The primary objective of this model was to study the magnitude
and effect of the three-dimensional
stress state on the compression failure of a plate with a hole.
The modeling procedure is thus
critical for proper evaluation of stresses. Three-dimensional
finite element models are often not
given consideration as analytical tools as, at the present state
of the art of computers, they are not
cost effective. Many techniques have been presented over the
years, with an emphasis on
improved cost-effectiveness, to provide the three-dimensional
information within the area of interest,
yet utilize a two-dimensional model away from it. A study by
Dong [47] provides Qood overview on alternatives to complete
three-dimensional modeling that can be utilized. Of these the
Local-Global
technique has been the widely accepted [46,48] procedure. Bums
et al. [35] recommends that for
Local-Global analysis of a notched composite laminate, the
local-global interface should be at least
five times the laminate thickness away from the hole boundary.
This interface for our model would
have to be 0.240 inches from the hole boundary or 0.340 inches
from the center of the hole. Given
the dimensions of our specimen, this technique, would not have
provided any substantial reduction
in size of the finite element model. Since the objective of this
study, however, is principally to
evaluate the three-dimensional effect on compressive strength,
the cost effectiveness of this
technique will not be belabored.
Finite Element Models 25
-
Limitations on the size of the model are imposed by the
capability of the computing system to
process the data. Most of the limitations experienced were
hardware based, primarily inadequate
swap diskspace, resulting in the use of a barely acceptable
model. The number of elements
chosen, particularly in the region of the hole, was such so as
to provide accurate representation
of the stress distribution, and satisfaction of boundary
conditions. The greater the number of nodes
in the x-y plane the better the representation of the transverse
stresses. Similarly the greater the
number of nodes through the thickness the more accurate
satisfaction of the boundary conditions
at the hole edge. It can be see that as the degree of accuracy
of the stresses obtained from the
finite element model increases the size of the model grows
exponentially. Since both of the above
requirements cannot be satisfied due to hardware limitations the
model has to be compromised.
Pagano [38) has shown that the effect of interlaminar stresses
are confined to a distance of
approximately one lamina thickness away from the free edge,
which in our case is the hole
boundary where the in-plane stresses are the greatest. As the
stresses in this region are known
to be singular, a large number of nodes is needed in this region
to properly represent the
distribution of transverse stresses. Too few nodes would cause
the model to be too stiff and so
over/under estimate the magnitude of these stresses, while too
many nodes would rapidly increase
the size of the model. -·'
In order to ensure the proper modeling of the stress gradients,
and to satisfy the traction-free
boundary conditions as closely as possible, it is desirable to
use two elements per layer through
the thickness. Hardware limitations permitted only one element
per layer to be used to represent
the thickness of each lamina. Vidussoni [46] has shown that
reasonably adequate representation
can be provided with one 20 node element per layer for cross ply
laminates, though, two such
elements per layer would definitely provide more accurate
results. This has been corroborated by
Thompson [48) though she used more elements. Depending on the
local distribution, the difference
between using one or two elements per layer may not be great due
to the fact that, in the event
of using two elements, the aspect ratio of the element for the
same planar dimensions increases
as compared to the one element case. Thus the gain in
sensitivity may not have the desired result,
as the change in aspect ratio effects the stiffness
characteristic of the element [46]. Also, in order
to maintain adequate sensitivity in the x-y plane at least two
elements have been accommodated
within one laminate thickness from the hole edge.
In the x-y plane, the distribution of the nodes is the same as
that in the two-dimensional model and
is shown in Figure 2-2. At least two rings of elements are
located within a region of one laminate
Finite Element Models 26
-
thickness from the hole edge. The density of radial distribution
of the nodes was made to vary with
the distance from the hole edge. Away from the hole, where the
stresses are assumed to
gradually reduce to far field values, larger aspect ratios (Vt)
have been used. Twenty node brick
elements have been used all over except at the location where
the distribution changes from radial
to cartesian where, appropriately, 15 node triangular wedge
elements have been used.
2.3 Boundary Conditions
2.3.1 Symmetry
The use of boundary conditions to represent planes of symmetry
on the three-dimensional model
is far more critical than that of the two-dimensional model. In
a two-dimensional model of a quasi-
isotropic plate, two planes of symmetry can be exploited. As
such a quarter plate model can be
used for the two-dimensional model. By way of definition a model
with a midplane of symmetry
(symmetric laminate) is called a half symmetric model. The
increase in the number of planes of
symmetry reduces the order of symmetry of the plate by a factor
of two, e.g. a two-dimensional
symmetric laminate plate will be an eighth symmetric model due
to assuming symmetry about the
planes through the three orthogonal axes. In a three-dimensional
model the number of elements
and the total degrees of freedom being large, symmetry provides
a necessary means of reducing
the size of the model. For all symmetric laminates the midplane
automatically presents a plane of
symmetry. Burns [35] has shown that no other plane of true
symmetry exists for a three-
dimensional models of quasi-isotropic laminates.
Despite the lack of true symmetry, in order to reduce the
computational effort other symmetry
conditions were investigated. One such condition which proved to
yield accurate stress distribution
was introduction of the symmetry about the x-z plane. For the 0
and 90 degree laminae the
coefficients of mutual influence 1111.i• which are defined as
[35]
Finite Element Models
(2511-2513-551 ) cos38sin8- (2522-2512 -SH) cos8sin38 S11cos'8+
(2S12+S11 ) sin28cos26+S22sin'8
(2.01)
27
-
'tx z
0.01.-----.----...-....... ---....-...-.......
-..----..-----.-----.
0
-0.011
.. .f~ .. l : ! l i ~ !
~
I ......... -............ 1 ................... .
; l
··········f···-·····--··+······-·····-···············+·····················-··-···+····-·······················
~'. I 11~--i I --& - Ctr. Sym. ~ -0.1
----------------------
0 10
Angle a 1311
Figure 2-3a: Comparison of a In half and zz
quarter symmetric models
110 ~
0.1 ..-........ -.----...-........ -.--....-..-........
-..---.-..-................ ---.---..
0.011 . .
···················-----···-·--··-···········-··········· ..
······· ··········-·· ............................... ..
0
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
411 10
Angle a 1311
Figure 2-3b: Comparison of 't In half and xz quarter symmetric
models
110 ~
Finite Element Models 28
-
where:
'Y1i - Shear strains
£1 - Normal strains
S11 - Compliance coefficients
9 - Fiber angle are zero about the plane perpendicular to the
loading axis. However, this is not the case with angle
ply laminae. Thus, an in-plane strain will result in shear
strains in the off-axis plies. The effect of
imposing symmetry on the x-z plane causes erratic behavior of
the through-the-thickness stresses,
az and 'txz• in the immediate vicinity of the hole along the
x-axis.
Figures 2-3 show the distribution of these stresses along the
periphery of the hole in three-
dimensional models, with and without assumed symmetry about the
x-z plane. The stresses are
normalized by the average far-field stress, att• which is
obtained by dividing the force reactions of the nodes at the
supports, i.e. X=-0.75, by the laminate area at that location. The
location of the
support is at a distance such that the stresses there are
unaffected by the existence of the hole.
The stress distribution of both, the half and quarter symmetric
models coincide over most of the
hole boundary except when approaching the plane of assumed
symmetry (x-axis) due to the
imposed boundary constraint, Ui = 0.00. Similar behavior has
also been shown by Burns [35], which -~
he describes to be of localized nature. This imposition of
symmetry is therefore not recommended
where absolute accurate representation of the stress state is
necessary throughout the edge of the
hole. Thus a quarter symmetry model of a quasi-isotropic
laminate would be considered adequate
for a failure analysis of an axially loaded condition as the
point of interest is along the y axis, i.e.
away from the axis of symmetry. The use of a half symmetry model
would result in significantly
more computation time with no particular gain in accuracy.
2.3.2 Traction-free Boundary Conditions
In finite element modeling, isoparametric displacement elements
satisfy equilibrium by means of
static geometric equations between the internal forces and the
external loads. The finite elements
are based on a displacement based variational principle and as
such continuity of displacements
is Satisfied. Displacements are thus generated at each point of
integration. The strains at these
points are then generated, on an elemental basis, as derivatives
of the displacements. The
stresses at each strain are then obtained by multiplying the
strains by the stiffness coefficient
matrix. It is thus seen that the stresses are discontinuous at
the interfaces due to the difference
Finite Element Models 29
-
,.. I
Finite Element M
odels
CD ... cu ... 0 "C CD E ... 0 """" CD "C c :s I -CD "C 0 E -cu c
0 ·-0 c CD E ·-"C I CD CD ... .c I-.. ~ I C\I CD ... :s en ·-LL
30
-
s1ap
ow 1
uawa
13 a
11u1
:1
"T1 -· cc c ~ CD N I c.n •• -f ::r ~ CD CD I c. -·
. ~": 3 CD :J tn -· 0 :J m - 3 0 c. CD -I c. a 0 ~ 3 CD c. en ..
m .. CD
-
in material properties in the two adjacent layers and due to the
lack of continuity of the derivatives
of the displacements across the element boundaries. This is
acceptable for in-plane stresses and
strains, however, equilibrium conditions require interlaminar
stresses to be continuous at the
interfaces. These have been taken care of by using an averaging
interpolation technique. By using
these displacement elements it is not possible to satisfy the
boundary conditions at the traction free
edges. They will not be identically zero [48]. It is for'this
reason that, in order to attain convergence,
the number of elements used to model the thickness be increased
in order to attain traction free
edges. It has been shown that in spite of system limitations and
the small number of elements per
layer used in the models, satisfactory results can be obtained
to provide valuable information
[35,48].
2.4 Comparison Between 2-D and 3-D Models
So far most of the work done [3,4, 1 O], on the ~nalysis of
failure of a plate with a hole has been
restricted to the use of two-dimensional finite element models,
assuming planar failure. Due to the
disparity between experimental and analytical values of strength
obtained, it has become necessary
to evaluate the three-dimensional effect on the response of a
laminate to in-plane loading. A two-
Climensional model was used to correlate the results obtained
earlier [1 O] and compare with a
three-dimensional model that is used to build upon it. Prior to
performing a failure analysis it is
necessary to understand the response of the respective models
under the uniaxial load. Figures
2-4 and 2-5 show the "undeformed" and "deformed" shapes of the
three-dimensional finite element
model. In Figure 2-5 an "exagerration factor" (maximum model
dimension/maximum displacement)
of 1 O was used.
In order to generalize a failure theory for quasi-isotropic
laminates, it is necessary to validate it for
all laminate orientations,~- To ensure this Gurdal & Haftka
[10] recommended that the model of a quasi-isotropic laminate
[±45/90/0]. be subject to a load that was rotated through 180
degrees.
This is analogous to keeping the load constant and varying the
fiber angles of the laminae by the
same amount resulting in a variety of quasi-isotropic laminates,
as described in Section 1.3. In this
case, they showed, that it was necessary to perform only one
two-dimensional finite element
analysis, subsequent orientations of the laminate could be
accomplished by transformation of the
loading axis. This process is efficient and economical and was
considered for the three-dimensional
model as well. However, as will be shown later this procedure is
not conducive for use with three-
dimensional analysis.
Finite Element Models 32
-
r
z
3h/8
··································f····································1····:············~················l.-..........................
.
i + ! I ;
·································-'····································•·····
·····A·················::'=,,,::····································
i I ~ ~
h/4
h/8
··································f····································l·················4r··············f···········
i i I i
..--~~~-'~ : I
1~~=1 * 0 .__.__.__.__.__,&,__..__..__..__..__"'-___
~,..._...__,_..._~ ............ ~
-0.00045 -0.0004 -0.00035
Strain £ xx
-0.0003 -0.00025 lg24.pm
Figure 2-6: Through the thickness strain distribution for '" 2-D
and 3-D models at x=0.50 and y=-0.33
(symmetric about z-axis)
Finite Element Models 33
-
O.OOO&r-....-....... --...--.,.--.--...-.--....-.......
--...--..--.--...-.--....--.
Eu -0.0005
-0.001
., I i
" i I i I i
" i ••..•. ,. ........... r······························ I
i
_:_j ____ _ --.- 3-0 Model (0)
- -ll - 2·0Model(O) -0.002 .__ ____ ..__ __ _..._.__....__._ __
..__...a..._..._.__....__,
0 46 10 135
Angle a
Figure 2-7a: Comparison of strains between 2-D and 3-D models,
9=0 and '=0
-0.0005 ___ ....... __....__ ............ __..._...._ __ .......
__. ___ ............ __._...._ __ ..... 0 45 10 135 180
Angle a
Figure 2· 7b: Comparison of strains between 2-D and 3-D models,
9::90 and t = O
Finite Element Models 34
-
r
0.001
.--...--.---.--...-------.-----------.-----.---....-..---e-- 3-D
Modi! (..CS) - -II - 2-0 Modi! (..CS)
0.0005
. ' : ········-r-···················:a:·········r······· . . . .
i ~ i . . . . . . i \ i i ..
-0.0005
-0.001 L-.......__;,..__,.....,.&-......................
..._...L-_.___.__,..._....L.. .............. ,__...._....,1 0 45 10
135 180
Angle a
Figure 2-7c: Comparison of strains between 2-D and 3-D models,
9=-45 and '=0
Finite Element Models 35
-
r
au
4r;:::c:::c:::z:==::::::i====~,....----...--."""."""r-----_,..-,
_..__ 3-0 Model (0)
- -b - 3-0 Modi! (45)- •• -45
: Jrq::~:::=::··~+~··:::~::~::~:-= : ! !
1 ······························!··················
············!············ .................. L.
.......................... .
I I : !
0 ~ .... ~ ...... .e-tS::·-··-1··-··--··-·1 ~...._ ........
_..__.,__.._...._ ........ ......i.. ....... -....__._....a...
........ __ ..._....__~
0 45 80 135 180
. Angle ex if29.pll Figure 2·8: Comparison of stresses obtained
by
rotation and transformation of 3·0 model· at • = O degrees
__._ 3-0 Model (45)- •· 0
- -b - 3-0 Model to>- t•45 -+- ~D Model (0)- + • Onmd. 45
. ~~:=~--~~:::~··-~:~~::= : : : I I . i oa... ...............
4=:tP-'. .. :'.'.':--r-.
:
i 0 80 135 180
Angle Cl
Figure 2·9: Comparison of stresses obtained by rotation and
transformation of 3-0 model at • = 45 degrees
Finite Element Models 36
-
a.,
0.1------.-. ....... --------------------------------. . : ., .
__ ;;,_;;_~;~~--J_L.fL.: ...