Geometrically Non-Linear Analysis of Layered Composite Plates and Shells

IRENEUSZ KREJA
GEOMETRICALLY NON-LINEAR
ANALYSIS OF LAYERED
COMPOSITE PLATES AND
SHELLS
Gdańsk 2007
PREFACE
This monograph includes the results of investigations carried out by the author since
1992. The whole period was divided into several research intervals of a different intensity.
It all started during the author participation in the DFG-Research Project "Theory and
Nonlinear FEM-Analysis of Elastic and Elasto-Plastic Anisotropic Structures Including
Material Damage" performed under a supervision of Prof. Dieter Weichert and Dr-Ing.
Rüdiger Schmidt at the University of Wuppertal, Germany in the period of 1991-19931.
Concurrently to, and to some extend separately from the main subject of the DFG project a
scientific cooperation with Dr-Ing. Rüdiger Schmidt was initiated that was devoted to the
FEM implementation of the Moderate Rotation Theory of anisotropic shells proposed by
Schmidt & Reddy [426]. First versions of computer programs for the geometrically non-
linear analysis of layered shells were prepared by the author already in Wuppertal. The
results of this period of research were presented in two conference reports (Bödefeld et al.
[70], Kreja & Schmidt [263]) and one journal paper (Kreja et al. [269]). After his return to
the Gdansk University of Technology in 1993 the author continued the research searching
for possible improvements in the accuracy of the results by including additional terms in
strain-displacement relations, improving procedures of the rotation update (Kreja &
Schmidt [265]) and exploring assumed natural strain approach (Kreja & Schmidt [264]).
After several months of break related to the intensive involvement in other research projects
the author returned to the investigations on large deformations of anisotropic shells resulted
in a computer implementation of the large rotation theory of anisotropic shells proposed by
Librescu [290]. The results of that research activity were presented at several international
conferences (Ferro et al. [165], Kreja [260], Kreja & Schmidt [266, 267]) and in one journal
paper Kreja & Schmidt [268].
The present report starts with an extensive literature survey on the main concepts of
theoretical models for multi-layered plates and shells. Then a systematic construction of a
computational model is presented for a large rotation analysis of elastic laminated shells
including a Finite Element Method implementation of the proposed algorithm. An essential
part of the present report is devoted to the examination on the relevance of various
approximation decisions in the large deformation analysis of plate and shell problems. A
number of sample problems of non-linear, large rotation response of composite laminated
structures are discussed. The report ends with a short summary of main conclusions and
some recommendations for the potential areas of future research. The supplemented list of
references contains 512 items cited in the text.
1
Kreja, I., Schmidt, R., Teyeb, O. & Weichert, D.: Geometrically nonlinear analysis of inelastic shell structures
including ductile damage, Cahiers de Mécanique, Université des Sciences et Technologies de Lille 1-2, 1994, 1-
106.
ACKNOWLEDGEMENTS
I would like to take this opportunity to acknowledge my colleagues and co-workers for
stimulating ongoing discussions and their encouragement to work on the subject of this
report. I look with a special appreciation for the many years of research collaboration with
Prof. Rüdiger Schmidt, actually at the RWTH Aachen. I would like also to express my
gratitude towards Dr Czesław Branicki and Prof. Jacek Chróścielewski, who many years
ago introduced me to the world of the computer analysis of structures.
I would like to express particular thanks to the peer reviewers, Prof. Wojciech
Pietraszkiewicz and Prof. Eligiusz Mieloszyk for their extremely helpful and constructive
comments, which have been incorporated to improve the final version of the manuscript.
A special recognition is extended to Mrs. Agnieszka Sabik for her thorough proof-
reading of the manuscript.
I am also grateful to my school, Gdańsk University of Technology for supporting my
research work financially and organizationally during all those years.
My special tribute goes to Maestro Tomasz Stańko for a very stimulating atmosphere
for the writing of this monograph he has provided with his magic trumpet.
Finally, my deepest gratitude goes to my family that never ceased to give me support,
encouragement, and understanding, especially my beloved wife Anna Mariola.
DEDICATION
Almost at the end of the writing of this monograph I was terrified with the news of the
violent death of the distinguished scientist and world-wide recognized authority in the field
of layered shell structures, Prof. Liviu Librescu, who was killed in the shooting at the
Virginia Tech University on April 16, 2007. Prof. Librescu died as a hero; before the
gunman killed him Prof. Librescu saved lives of his students by blocking the door with his
own body.
I have never met him personally, but we had a chance to exchange some e-mails.
Prof. Librescu was also a reviewer of one of my earlier papers.
I would like to dedicate this work to His memory.
Ireneusz Kreja
Gdańsk, May 2007
CONTENTS
LIST OF SYMBOL AND ABBREVIATIONS.................................................................................. 9
1. INTRODUCTION ........................................................................................................................ 13
1.1. Plates and shells – from Nature to space industry ......................................................... 13
1.2. Laminated composites and sandwich panels................................................................... 13
1.3. Scope and objectives of this report .................................................................................. 16
2. LITERATURE REVIEW AND MODELING CONSIDERATIONS ....................................... 18
2.1. Development of theoretical models for plates and shells................................................ 18
2.2. Geometrically non-linear analysis of plates and shells .................................................. 22
2.3. Theoretical models for layered thin-walled structures .................................................. 25
2.4. Numerical implementation of plate and shell theories .................................................. 35
2.5. Modeling considerations resulted from the presented review ...................................... 51
3. INCREMENTAL FORMULATION OF NONLINEAR SHELL ANALYSIS ........................ 52
3.1. Incremental formulation .................................................................................................. 52
3.2. Shell geometry .................................................................................................................. 53
3.3. Shell deformation ............................................................................................................. 57
3.4. Strain-displacement relations .......................................................................................... 59
3.5. Virtual work principle ..................................................................................................... 72
3.6. Constitutive relations ....................................................................................................... 77
4. FINITE ELEMENT METHOD IMPLEMENTATION............................................................. 82
4.1. Finite element discretization of the problem ................................................................... 82
4.2. Incremental equilibrium equations of FE model ............................................................ 84
4.3. Solving of incremental equilibrium equations ............................................................... 86
4.4. Finite elements used in this study .................................................................................... 90
4.5. Computer implementation of the proposed FEA algorithm .......................................... 92

8 Contents
4.6. Assessment of selected elements in linear analysis ......................................................... 93
5. LARGE DEFORMATION ANALYSIS EXAMPLES ............................................................ 103
5.1. Instability of clamped-hinged circular arches............................................................... 103
5.2. Clamped laminated shallow arch under point load...................................................... 108
5.3. Hinged laminated cylindrical panels under point load ................................................ 110
5.4. Glass-epoxy cylinder under internal pressure ............................................................. 116
5.5. Asymmetric cross-ply simply supported plate strip .................................................... 117
5.6. Clamped laminated cylindrical panels under point load ............................................. 118
5.7. Stretching of an open cylinder ....................................................................................... 125
5.8. Pinched hemispherical shell with 18° hole .................................................................... 129
5.9. Axial compression of composite cylindrical panel ........................................................ 133
5.10. Buckling of composite cylindrical panels with square cut-outs ................................. 139
6. CONCLUSIONS AND FUTURE PERSPECTIVES ................................................................ 148
6.1. Concluding Remarks....................................................................................................... 148
6.2. Original Contribution ..................................................................................................... 150
6.3. Recommendations and Future Perspectives.................................................................. 150
7. REFERENCES ............................................................................................................................ 151
SUMMARY IN ENGLISH ............................................................................................................. 177
SUMMARY IN POLISH ................................................................................................................ 178

LIST OF SYMBOLS AND ABBREVIATIONS
Symbols
0 0
aαβ − covariant components of the surface metric tensor in the middle surface Ω
0 αβ 0
a − contravariant components of the surface metric tensor in the middle surface Ω
0 0
aα − covariant base vector of the middle surface Ω
0
bαβ − components of the surface metric tensor of the second order
0
cαβ − components of the surface metric tensor of the third order
0
C – initial configuration, at time 0
1
C – actual configuration, at time t
2
C – searched configuration, at time t+∆t
[C] − constitutive matrix in the 3D constitutive relation
0
gαβ − components of the metric tensor in the shell space
m m m m
d – director, local position vector d = g3 = a3
0
dV − volume element in the initial configuration 0C
0
dΩ − midsurface area element in the initial configuration 0C
m
0E – Green-Lagrange strain tensor
m m
Eαβ − components of the Green-Lagrange strain tensor in the configuration C
E – Young modulus for the isotropic material
Ea – Young modulus in the direction of the material a-axis for the orthotropic material
m
0F – deformation gradient
1
0 FS – components of the balanced forces vector
2 i
f − components of external body forces (acting per unit volume element)
0
gi − covariant base vectors in the space of the shell
0 0
G − metric tensor at any arbitrary point P in the shell space
Gab – “in-plane” shear modulus for the orthotropic material
Gbc, Gac – transverse shear moduli for the orthotropic material
h − shell thickness
[H] − constitutive matrix in the 2D constitutive relation
1 (U )
0K ST − components of the first part of the incremental stiffness matrix
1 (G )
0 K ST − components of the geometrical stiffness matrix
2 mn m
0 L − components of the effective stress resultants in the configuration C
0 0
n − base vector normal to the initial middle surface Ω
Nk − isoparametric shape function associated with node k
NNE − number of nodes of the isoparametric element
2 i
p − components of external surface forces
m m m
r – position of an arbitrary point P of the middle surface in the configuration C
m k m
r – position vector of the node k at the shell mid-surface in the configuration C
m
R – position vector of an arbitrary point P at the configuration mC
[R] − rotation matrix
r, s – natural coordinates for the isoparametric element
2 mn m
0 S − components of the second Piola-Kirchhoff stress tensor in the configuration C
10 List of symbols and abbreviations
[T] − transformation matrix between the material axes (a, b, c) and the coordinate
1 2 3
system (θ , θ , θ )
m m
V - displacement vector in configuration C
m
Vi - components of the displacement vector referred to the undeformed shell space
m
υi - components of the displacement vector referred to the undeformed shell mid-surface
αk - ply orientation angle
ε − prescribed tolerance in the convergence criterion
m
Г i − Christofel symbol of the second kind (components of the base vector gi derivative)
δαβ − Kronecker delta
δ2eij − variation of components of the infinitesimal strain tensor in the configuration 2C
δui − covariant components of the virtual displacement vector
2
δWe − external virtual work in the configuration 2C
2
δWi − internal virtual work in the configuration 2C
{∆q} − vector of displacement increments
m
ϕα − rotation angle, (α = 1, 2)
m
ϕαβ - components of the displacement gradient
κ − transverse shear correction coefficient
0 0
−shifter tensor in the initial configuration C
0
0 β
µα − components of the shifter tensor
0
µ − determinant of the shifter tensor
λ – load parameter
ν – Poisson coefficient
νab – “in-plane” Poisson coefficient for the orthotropic material
νbc, νac – transverse Poisson coefficients for the orthotropic material
σ
2 ij
− components of the Cauchy stress tensor in the configuration 2C
α
θ − general convected coordinates in the middle surface, (α = 1, 2)
3
θ − thickness coordinate taking values from the interval (-h/2, h/2)
m
« − Rodrigues rotation vector
m m
Ω − middle surface of the shell in the configuration C
m m
[ ℜ( « & − rotation tensor
Abbreviations
2D – two dimensional
3D – three dimensional
ANS – Assumed Natural Strain (... approach)
CLT – Classical Lamination Theory
DKT – Discrete Kirchhoff Theory
DL – Discrete-Layer (... theory)
dof(s) – degree(s) of freedom
EAS – Enhanced Assumed Strain (... method)
ESL – Equivalent Single Layer (... model)
FE – Finite Element
FEA – Finite Element Analysis
FEM – Finite Element Method
FOSD – First Order Shear Deformation (... theory)
FRC – Fiber Reinforced Composite
HOSD – Higher Order Shear Deformation (... theory)
List of symbols and abbreviations 11
LRT – Large Rotation Theory

LRT5 – LRT with 5 displacement parameters (inextensible director)
LRT56 – LRT with 5 aggregate and 6 incremental displacement parameters (inextensible
director)
LRT6 – LRT with 6 displacement parameters (extensible director)
MITC – Mixed Interpolation of Tensorial Components
MRT – Moderate Rotation Theory
MRT5 – MRT with 5 displacement parameters (inextensible director)
RT5 – TOSD model of shells with 5 displacement parameters (Başar et al. [37])
RT7 – TOSD model of shells with 7 displacement parameters (Başar et al. [37])
RVK – Refined von Kármán (... theory)
RVK5 – RVK with 5 displacement parameters (inextensible director)
SLR – Simplified Large Rotation (by Dennis & Palazotto [138, 139])
SOSD – Second Order Shear Deformation (... theory)
SRI – Selective Reduced Integration
TL – Total Lagrangian (... formulation)
TOSD – Third Order Shear Deformation (... theory)
UL – Updated Lagrangian (... formulation)
URI – Uniformly Reduced Integration
12 List of symbols and abbreviations
Chapter 1
INTRODUCTION
1.1. Plates and shells – from Nature to space industry

Plates and shells are often used as structural members of engineering constructions -
just to mention walls and roofs of buildings, decks of bridges, dam walls, various tanks and
containers, hulls of ships, bodies of automobile vehicles, wings and fuselages of aircrafts,
bowls of telescopes or spaceship bodies. A structural element with one dimension, the
thickness, being much smaller than the two others one can categorize as a thin-walled
surface structure or a shell, then a plate is just a special case of a shell characterized by a
flat surface. Some authors would rather prefer to keep the term shell only for a thin-walled
structure with a curved surface. One can believe that the form of a (curved) shell was
invented by the Nature (a hard outer covering of some animals or bird’s eggs) and it was
just adopted by ancient builders who early recognized advantages of that phenomenal
structure which, being compact and light itself, can carry considerable loads. It can be
supposed that first man-made shells appeared as parts of housing constructions, from
primitive huts to vaults in buildings. For long time architects and builder masters depended
just on their own experience and intuition2 although many of ancient shell structures lasted
until our days, despite wars and other disasters.
1.2. Laminated composites and sandwich panels

Composites are made of two or more materials, combined together to obtain a new
matter with properties that are superior to those of individual components. A typical
example of a composite is a sandwich beam with cross-section that is very close to an ideal
profile for a pure bending loading, Gere & Timoshenko [172]. Two thin outer faces of a
sandwich beam carry almost the whole load whereas a light core between the two faces has
only to keep them apart at a constant distance. Such a sandwich beam can be regarded as a
composite structure. One can guess that distinguishing between a composite material and a
composite structure is not always straightforward. Reinforced concrete is a quite standard
building material in civil engineering. A concrete beam reinforced with discrete steel bars is
just a composite structure. On the other hand, a similar beam but made of fiber-reinforced
concrete can be regarded as a structure build of a composite material.
Generally among the main constituents of a composite material one can recognize:
reinforcement components, matrix and fillers (see e.g. Jones [231], Vasiliev & Morozov
[475], Vinson & Chou [476], also Website of ACMA [487]). The reinforcement mostly
takes the form of fibers or whiskers that are used to provide the required strength and
stiffness. The matrix provides a compliant support for the reinforcement; it contributes to
load transfer and also gives environmental protection. Fillers are other substances that are
2
An impressing description of a building process of big cathedrals in Middle Ages can be found in a famous novel
The Pillars of the Earth by Ken Follett (William Morrow, New York 1989).
14 Chapter 1. Introduction
added to the matrix to reduce the cost but sometimes also to improve some mechanical
properties of the composite. It is quite obvious that mechanical properties of composites
depend mainly on the choice of material components used for the composite but they are
also considerably influenced by the applied fabrication technique.
Probably the most suitable for structural applications among all composite materials
there are fiber reinforced composites (FRC), with the reinforcement taking the form of
either continuous (long) fibers (Halle & Kelly [188]) or whiskers (short fibers) (Chou &
Kelly [117]). Composites reinforced with continuous fibers frequently appear as fiber
reinforced composite laminates, see Fig. 1.1.
Fig. 1.1: An example of a FRC three-layer laminate
A typical fiber reinforced composite laminate is made of a number of unidirectional

fiber reinforced composite layers (Jones [231], Vasiliev & Morozov [475], and Vinson &
Chou [476]). A composite layer with a parallel system of reinforcement fibers represents an
orthotropic medium with three mutually orthogonal planes of symmetry. Usually a
composite layer consists of high modulus fibers (typically they are glass, boron or graphite
fibers) embedded in a matrix (epoxy or polyamide). A resulting fiber-reinforced material
(composite material) combines a high strength with a light weight; see Fig. 1.2 for a
simplified comparison of specific tensile strength (tensile strength/specific weight) of some
selected composite laminates and standard construction materials.
Aluminum
Structural steel
High-strength steel
Aluminum alloy Composite laminates composed with 60% fibre and

40% epoxy resin per volume
Glass-epoxy composite
Aramid-epoxy
composite
Carbon-epoxy
composite
0 10 20 30 40 50 60 70 80 90
3
Specific tensile strength [10 m]
Fig. 1.2: Simplified comparison of specific tensile strength of some selected materials
1.2. Laminated composites and sandwich panels 15
Considering its light weight, a lamina of fiber reinforced composite is remarkably

strong along the fiber direction. However, the same lamina is considerably weaker in all
off-fiber directions. To address this issue and withstand loadings from multiple angles, one
would use a laminate constructed by a number of laminas oriented at different directions. A
laminated composite panel can be considered as an optimal structure with effective
utilization of composite material directional properties (Vasiliev & Morozov [475]);
however, fiber reinforcement can be applied also in a three-dimensional layout (compare
Tong et al. [468]). Composites can contain more than one type of fiber material; then they
are known as the hybrid composites (Chou & Kelly [117], Vasiliev & Morozov [475]).
A sandwich shell and a thin composite laminate take a form of a layered shell that is
built of several laminas bonded together with some adhesives. A typical sandwich shell
consists of a light core and two thin outer faces, which can take a form of either isotropic or
laminated composite panels. The light core, similarly as in the case of the aforementioned
sandwich beam, does not transfer significant forces; the most important for the core is its
low weight. When the core is made as a composite then the requirement for reinforcement
is minimal and the contribution of light fillers significantly increases. Very often the fillers
are simple empty voids, and then one can talk about porous cores or foams. Another option
is a core build as a 3D structure, e.g. corrugated panels and spatial lattices of beams or
plates. Probably the most popular type of a sandwich core is based on two-dimensional
cellular geometries with large-scale cells (see e.g. Librescu & Hause [291], Hohe & Becker
[199], Hohe & Librescu [200]); here the flagship example is a hexagonal sheet structure
that visually resembles a product of the apiarian industry, therefore is commonly known as
a honeycomb (Fig. 1.3). Other examples of multilayered anisotropic structures are laminates
made of different isotropic layers e.g. employed for better thermal insulation or noise
suppression. Considering the application of layered structures, one should not forget also
smart thin-walled sandwich structures with piezoelectric layers embedded for passive or
active vibration or sound control (see e.g. Reddy [397] or Lentzen & Schmidt [282])
upper face
adhesive
bond
honeycomb
core lower face
Fig. 1.3: Sandwich plate with a honeycomb core
It is easy to find a prototype of all modern composite laminates in a traditional layered

wood material - plywood, which is made from thin sheets of wood veneer, stacked together
with strong adhesives under heat and pressure, each layer glued at right angles to the
adjacent layers, Piskunov [369]. Plywood is well known for its resistance to cracking and
shrinkage, with high strength-to-weight ratio. For over one hundred years this material has
been used for building boats, and it has also contributed to the development of aviation
since skins of many early aircrafts were constructed of plywood including even military
16 Chapter 1. Introduction
aircrafts like the famous British bomber Mosquito used in the II World War. In the 1940s
some parts of military equipment used by the U.S. Air Force and Navy were made from a
glass-epoxy composite commonly know as the fiberglass. Many structural parts of
contemporary jet airplanes are made of more advanced composite materials, from boron-
epoxy stabilizers of the 1970’s fighter F-15 to a large-scale use of composites in the
fuselages and wings of the new passenger airliner Boeing 787 and the mega-liner Airbus
A380 (see e.g. Ford [168], Hachenberg [187]). Similarly as for other high-tech products,
the earliest applications of modern composites were limited mostly to the aerospace
industry and military engineering. Gradually with the development of the production
technologies, when composites became cheaper and more available, they found also
applications in the manufacturing of automobiles, pressure vessels, and other weight-
efficient mechanical structures. Quite early, composites were used for production of
advanced sport equipment, such as sailing boats, skis or tennis racquets.
Since production costs of advanced composite materials remained at relatively high
level as compared with other building materials, for a long time their application in civil
engineering was rather limited; on the other hand, the community of civil engineers was
accustomed to the usage of traditional composites like reinforced concrete, laminated wood
or steel-concrete girders. The first applications of modern composites in civil constructions
took place in the Middle East at the turn of the 1960’s and 1970’s: they were the dome
structure erected by British engineers in Benghazi (Libya) in 1968 and the roof at Dubai
Airport (the United Arab Emirates) built in 1972, Hollaway [201]. In the USA in the late
1960’s fiberglass was used in construction of some fuel storage tanks, Stockton [447]. In
1976 the first FRC pedestrian bridge was build in Israel, Telang et al. [463]. The Miyun
Bridge built in China in 1982 is commonly recognized as the first vehicular composite
bridge, Lopez-Anido & Karbhari [295]. More remarkable examples of the fiber reinforced
composites applications in civil engineering structures can be found in recent publications
by Hollaway [201], Lopez-Anido & Karbhari [295], Van Den Einde et al. [473], and at the
Website of Composites IQ [488]. Lopez-Anido & Karbhari [295] and Stockton [447]
discussed exhaustively the advantages of fiber reinforced composites in the context of civil
infrastructure applications. Besides the already mentioned high strength and the low weight
of the FRC materials, one should add here also the corrosion resistance, enhanced fatigue
life, and especially the ability to tailor directional properties of the final product to precise
geometrical specifications. Those potential benefits from FRC application should be
balanced against the higher initial cost and lower stiffness of composites, when compared
to traditional building materials. Noteworthy are also such features of composites like
controllable thermal properties, reduced electrical conductivity and tailored magnetic
properties, Stockton [447].
1.3. Scope and objectives of this report

This report explores a topic of a mathematical modeling of a mechanical behavior of
laminated plates and shells. An appropriate computational representation of laminated
composites and sandwich panels is a subject of current interest of specialists in various
fields of structural engineering, from aerospace and automobile industry to civil
engineering. Analysis of laminated composite panels is usually based on various refined
plate and shell theories, however, some computational models are derived directly from 3D
continuum mechanics.
1.3. Scope and objectives of this report 17
The main goal of this research is to develop a numerical model for geometrically
nonlinear analysis of moderate thick laminated composites and sandwich panels based on a
general large rotation layered shell theory. An essential part of the present report is devoted
to a practical finite element implementation of the proposed computational model. A
special emphasis is focused on the examination on the consequences of various
approximation decisions concerning the description of large rotations.
The following fundamental assumptions have been adopted in this work:

a) elastic material behavior;
b) moderate thickness of the shells allowing for application of the Equivalent Single
Layer models3 within the First Order Shear Deformation Theory;
c) large deformations with finite rotations but small strains (inextensible thickness);
d) smooth geometry of the middle surface without kinks and branches.
The content is organized as follows. Chapter 2 provides a state-of-the-art in the

modeling of laminated composite and sandwich panels; the main concepts in constructing
theoretical models for plates and shells as well as the fundamental hypotheses concerning
the computational treatment of multi-layered panels are reviewed. Chapter 3 presents a
systematic construction of a mathematical model of a geometrically nonlinear mechanics of
elastic laminated shells assuming an Equivalent Single Layer approach. Next the finite
element implementation of the proposed computational strategy is outlined in Chapter 4.
Chapter 5 is devoted to analysis of selected numerical examples. Finally, conclusions and
recommendations are given in Chapter 6.
3
A detailed explanation of terminology related to the numerical modeling of composite plates and shells can be
found in the next chapter.
Chapter 2
LITERATURE REVIEW AND MODELING CONSIDERATIONS
2.1. Development of theoretical models for plates and shells

The systematic analysis of thin-walled surface structures became possible after the
progress in mathematical sciences during the Renaissance. However the foundations of
modern plate and shell theories were laid only in the 19th century by G. Kirchhoff and
A. E. H. Love, whose names are associated with the classical (Kirchhoff-Love) theory of
thin shells4. For a long time theoretical models of plates and shells were constructed
following the idea of a dimensional reduction – such a model assumes that the behavior of a
three-dimensional shell-like body can be described with the two-dimensional surface
model. A big advancement in the elaboration of plate and shell theories accompanied the
technological boom of the 20th century. The refined plate and shell theories accounting for
the effects of transverse shear deformations formulated at the turns of the 1940s and the
1950s by H. Hencky, L. Bollé, E. Reissner [404] and R. D. Mindlin [314] are commonly
named after the latter two as Mindlin-Reissner type theories. A natural appetite for a more
efficient use of building materials by thinning the walls of the shell, on one hand, and
increasing the loads, on the other hand, caused the need for expansion of the range of
analysis beyond the limits of linear theory – the problems of large deformations, buckling
and material non-linearity have been considered for plates and shells by many researchers.
The non-linear models of thin (Kirchhoff-Love type) shells are usually associated with the
name of W. T. Koiter [251]5, whereas the large deformation analysis of shear deformable
(Mindlin-Reissner type) shells is frequently linked to P. M. Naghdi [321, 322].
Although the theoretical models of plates and shells can be categorized according to
different criteria, the main question remains the same: how accurately the considered model
can represent behavior of a real shell-like three-dimensional construction?6
In the classical methodology called the “derived approach”, the shell theory is
constructed starting from a three-dimensional continuum mechanics model. The governing
equations of a three-dimensional continuum can be reduced into a two-dimensional shell
theory by adopting appropriate a priori assumptions for the displacement field (see e.g.
4
The list of References attached to this report is limited only to texts which were available to the author; therefore
it does not include many famous historical papers often cited in the literature. Detailed reviews on historical
developments in the theory of plates and shells can be found e. g. in comprehensive studies presented by Jemielita
[227, 228]. An extensive bibliography on plates and shells is available also in the works of Bischoff et al. [63],
Chapelle & Bathe [105], Chróścielewski et al. [119], Dauge et al. [136], Naghdi [321], Noor [325, 326],
Pietraszkiewicz [362], Saleeb et al. [412], Valid [471] or Wempner [489].
5
It is quite symptomatic that over ten years before the publication of the paper by Koiter [251] in 1966, the non-
linear theory of thin shells was presented in a series of articles by Russian researches: Mushtari, Galimov and
Alumäe. However, their results were published mainly in Russian language and therefore they were hardly
recognized outside the East Europe (compare Pietraszkiewicz [362]).
6
We can recall here after Dauge et al. [136] a very significant remark of W.T. Koiter and J.G. Simmonds: “Shell
theory attempts the impossible: to provide a two-dimensional representation of an intrinsically three-dimensional
phenomenon” (Koiter, W.T. & Simmonds, J.G.: Foundations of shell theory, Theoretical and Applied Mechanics,
Springer: Berlin, 1973, 150–176).
2.1. Development of theoretical models for plates and shells 19
Habip [186], Koiter [250], Librescu [290] or Naghdi [320])7. Within the First Order Shear
Deformation (FOSD) theory a linear variation of a displacement field over the thickness of
the thin three-dimensional body is assumed, resulting in a standard Mindlin-Reissner type
shell model8 (see Fig. 2.1).
Fig. 2.1: Simplified representation of the FOSD model
Due to a constant shear distribution across the shell thickness, the FOSD models
require an appropriate transverse shear correction typically defined by energy
considerations.9 In its conventional version, with the constraint of a constant shell
thickness, a FOSD type theory represents a “5-parameter shell model”, where the
kinematics of the shell can be described sufficiently with 3 translational and 2 rotational
displacement parameters of the shell midsurface. Extended versions of the FOSD type
theory can include thickness changes of the shell: a “6-parameter shell model”10 accounts
7
In the mathematical literature one can find rigorous validations of some classical shell models performed for
selected sets of boundary and load conditions, Dauge et al. [136]. A mathematically precise justification of the
Koiter model for Kirchhoff-Love shells can be obtained by means of the asymptotic expansion method (see e.g.
Ciarlet & Lods [127], Gratie [177] also Niordson & Niordson [324]). A similar validation for the general,
asymptotically correct Naghdi model of Mindlin-Reissner plates and shells can be obtained only within variational
reduction methods (see e.g. Arnold [18], Sutyrin & Hodges [453]). For more reading in this subject we can refer to
the texts of Arnold et al. [22], Chapelle & Bathe [105], Ciarlet & Lods [125], Ciarlet et al. [126], Reissner [403] or
Pitkäranta [371]. (Two other frequently cited resources on asymptotic analysis of shells were not available to the
author, therefore they can be just recited here, e.g. after Chróścielewski et al. [119]: Ciarlet, P.G.: Mathematical
Elasticity, Vol. III: Theory of Shells, North-Holland, Amsterdam 2000; Lewiński, T. & Telega, J. J.: Plates,
Laminates and Shells. Asymptotic Analysis and Homogenization, World Scientific Publishing Co., Singapore
2000.)
8
It is quite common in the present-day literature to identify the FOSD model with the “Reissner-Mindlin
plate/shell theory”. However, in the present report a more accurate phrase “Reissner-Mindlin kinematics” has been
adopted instead, following the suggestion of E. Ramm who reminded in [382] that E. Reissner and R. D. Mindlin
proposed in fact two different plate theories. Quite recently, practical differences in the two plate theories have
been examined by Wang et al. [482]. Ramm [382] pointed also out, that the contribution of Mindlin can be
considered as an extension of earlier works of H. Hencky and L. Bollé. More details on a historical context in the
development of theory of plates and shells can be found in comprehensive studies presented by Jemielita [227,
228]. Jemielita not only backed up Ramm’s opinion that credits for the development of the plate theory commonly
known as the “Mindlin plate theory” should also go to H. Hencky and L. Bollé, but he recommended even to use
the name “Hencky-Bollé plate theory” as being more correct one from a historical point of view.
9
The most frequently a shear correction factor for isotropic plates and shells is taken as k=5/6 after Reissner or
k=π2/12 after Mindlin. However, in a more advanced considerations the shear correction factor can depend on the
Poisson ratio (see e.g. Altenbach [6], Jemielita [227, 228], Mindlin [314]). Wittrick [497] indicated that for general
orthotropic plates the effective shear stiffness can be determined only for assumed displacement modes.
10
In the “6-parameter shell model” considered here, the 6th parameter represents only the deformation
corresponding to the thickness stretch. In the literature one can find also other “6-parameter shell models”, where
a three-parameter description of displacements is accompanied by a three-parameter representation of rotations
without any additional parameter related to the thickness stretch. Such 6-parameter shell formulations account for
drill rotations though the transverse normal strains are neglected or recovered from the plane stress or the
20 Chapter 2. Literature Review and Modeling Considerations
for a linear through-the-thickness distribution of the transverse normal displacement

(Pietraszkiewicz [360], Simo et al. [440], Braun et al. [81], Bischoff & Ramm [60],
Bischoff et al. [63]), whereas a “7-parameter shell model” (Bischoff & Ramm [59, 60],
El-Abbasi & Meguid [154], Ramm [382], Brank et al. [78], Brank [74]) corresponds to
quadratic distribution of transverse displacement11. When the effect of transverse shear
strains as well as the thickness change can be neglected (e.g. for thin isotropic panels), the
rotations are not independent parameters any more, and the resulting “3-parameter shell
model” corresponds to the classical Kirchhoff-Love theory of thin shells (Koiter [251],
Simmonds & Danielson [434], Pietraszkiewicz [359, 362, 363], Pietraszkiewicz &
Szwabowicz [365]). The orthogonality condition of the shell director (compare Fig. 2.2) is
probably the most frequently recognized attribute of the Kirchhoff-Love model; see e.g.
Green & Zerna [178], Koiter [250], Naghdi [320, 321], Reissner [403], Pietraszkiewicz
[359, 360, 362], Libai & Simmonds [289], Başar & Krätzig [43], Jemielita [228], Arnold et
al. [22], Bischoff et al. [63], Chróścielewski et al. [119].
Fig. 2.2: Simplified representation of the Kirchhoff-Love model
A further simplification of the model can lead to the membrane shell theory, where all
stress and strain components can be considered as being constant through the shell
thickness (see e.g. classical handbooks of Flügge [166], or of Girkmann [175]).
When advancing in the opposite direction, i.e. increasing the order of a complication
of the model, one can arrive at the Higher Order Shear Deformation (HOSD) theory;
Bercha & Glockner [56], Cho et al. [114], Hildebrand et al. [195], Phan & Reddy [355],
Reddy [393], Reddy & Liu [401] in the range of small displacements and Reddy [396],12
Librescu [290], Başar [35], Balah & Al-Ghamedy [31] for geometrically non-linear
problems13. It is quite obvious that the HOSD models are capable to represent the warping
of the deformed cross-section (see Fig. 2.3), whereas in the FOSD model that state of
deformation was approximated in the average sense by a constant transverse shear strain, γ.
incompressibility conditions (see e.g. Chróścielewski et al. [120], Sansour & Bufler [417], Sansour & Wagner
[419], Wiśniewski [496]). For an extensive discussion of the drill rotation formulations we can refer to the book of
Chróścielewski et al. [119].
11
It was indicated by Bischoff and Ramm [60], that linearly varying transverse normal strains resulted from
quadratic distribution of transverse displacement in the 7-parameter shell model allows for an exact reproduction
of the three-dimensional solution. See also Ramm & Wall [385, 386].
12
Reddy’s paper [396] contains a review of various third-order plate theories accounting for the transverse shear
stress conditions on bounding surfaces.
13
Simmonds [433] questioned the plausibility of higher order shell theories pointing out that “except in
exceptional circumstances, one cannot refine the classical, first-approximation theory of shells without
simultaneously evoking the 3-dimensional theory of elasticity to refine the boundary conditions”. (Similar critical
remarks can be found in Libai & Simmonds [289])
2.1. Development of theoretical models for plates and shells 21
An alternative way to construct a shell theory is known as the “direct approach”. Here
the deformation measures and balance equations are postulated directly for a shell modeled
as a one- or more-director Cosserat surface, or a surface with three independent rigidly
rotating versors (or the rotation tensor), without any direct references to the three-
dimensional continuum mechanics (see e.g. Eriksen & Truesdell [156], Green & Zerna
[178], Naghdi [321]). The geometrically exact14 stress-resultant 5-parameter shell
formulation proposed by Simo et al. [435, 436, 437]15 can also be linked to this category.
Similar models but accounting additionally for drill rotations where formulated by
Ibrahimbegović [217], Ibrahimbegović & Frey [222], Sansour & Boćko [415], Sansour &
Bednarczyk [414], Sansour & Wagner [419].
Fig. 2.3: Warping and average shear deformation of the cross-section
Bercha & Glockner [55] confronted a linear shell model based on the “direct
approach” with a corresponding formulation obtained by means of the “derived approach”
and concluded that “Cosserat surface theory may be considered to be equivalent to a first
order shear theory”. Bischoff et al. [63], who performed a similar but much detailed
assessment accounting also for non-linear problems, indicated that the material law
postulated in a “direct approach” neglects an initial curvature of the shell what results in a
“restricting applicability of the formulation to thin shells”.
Another concept for developing of a 2-dimensional shell theory was initiated by
Reissner in 197416 and developed by Libai & Simmonds [289]. Their approach can be
placed between the other two discussed above, therefore it is known as the “mixed
approach”. The main idea is “to use those equations of three-dimensional continuum
mechanics that are independent of material properties to derive corresponding rod or shell
equations, but to postulate the form of those rod or shell equations that depend unavoidably
on material properties” [289]. Working along similar lines, Chróścielewski et al. [120]
formulated the governing equations of the two-dimensional shell model by exact through-
the-thickness integration of the fundamental laws of continuum mechanics: balance of
linear and angular momentum. Appropriate shell strain and bending measures were
constructed exactly on the two-dimensional level as work-conjugate counterparts of the
stress resultants and couples. In such a shell model the simplifying assumptions are
introduced only in the two-dimensional constitutive relations (see also Chróścielewski et al.
[119], Chróścielewski et al. [121]). Slightly different methodology although also classified
14
The term “geometrically exact” can be fairly misleading: in Simo et al. [437] on page 22 on can find the
following explanation: “Accepting the kinematic assumption which defines the class of admissible motions, the
geometry of the shell, as well as the balance equations, are treated exactly”.
15
An extension of this model including through-the-thickness deformations can be found in Simo et al. [440].
16
Reissner, E.: Linear and Nonlinear Theory of Shells, Thin Shell Structures, Y. C. Fung and E. E. Sechler, eds.,
Prentice-Hall, Englewood Cliffs, NJ, 1974, 29–44 (recited here after Chróścielewski et al. [119]).
as mixed approach was proposed by Valid, who presented “a simple surface theory” [471,
472]. His intrinsic formulation “directly starts from a surface endowed with mechanical
properties” and utilizes the two-dimensional variational principles “directly inspired by the
three-dimensional theory”. Valid did not follow the classical model of the Cosserat surface
with directors although he assumed direct definitions of surface stresses. A similar idea can
be recognized in the formulation of linear plate theory presented by Sutyrin & Hodges
[453] who employed the variational-asymptotical method to split the original three-
dimensional problem into a one-dimensional through-the-thickness analysis and two-
dimensional shell analysis equivalent to the classical Reissner–Mindlin model. Later that
concept was extended for the geometrically non-linear shell theory by Yu & Hodges [505]
(see also Yu et al. [506]). An interesting idea of a special three-dimensional outer-surface-
related shell theory particularly advantageous in the contact analysis was proposed by
Schlebusch et al. [422, 423].
Any survey of theoretical models for plates and shells should not omit the degenerated
solid approach (see e.g. Bischoff et al. [63]), where the behavior of a shell is described
without introduction of any surface theory. A computational model of a shell in this
conception, Ahmad [2], originates in a straight line from the continuum mechanics but it is
also strictly related to the Finite Element Method and therefore a more detailed description
of that methodology is located in Section 2.4 of this report.
A visible tendency in the current research is directed to the development of a general
shell theory as the response to the great demand of the engineering community for a
universal, all-purpose shell model being applicable in the analysis of different thin-walled
structures. However, until now none of those versatile models is the most advantageous
solution for an every problem. Therefore, there is still a need for the research in the field of
specific shell theories17 dedicated to membrane shells, thin shells, shallow shells, axial
shells, layered shells, etc. (see e. g. texts of such well established experts in that field like
Ciarlet et al. [126], Libai & Simmonds [289], Axelrad [30], Pietraszkiewicz [363] and
Chapelle & Bathe [105]).
2.2. Geometrically non-linear analysis of plates and shells

In the previous section the theoretical models of thin-walled surface structures were
classified according to different methodologies applied to obtain the particular formulation.
Another essential distinction between various plate and shell theories is based on the degree
of non-linearity considered in the individual model. Shells are among those structures
where the state of deformation during “regular” exploitation very often goes outside the
limits of the linear response, Ramm & Wall [385, 386].
Considering the possible sources of non-linearity, one usually distinguishes between
material non-linearity and geometrical non-linearity; see e.g. Bathe [46], Crisfield [129],
Başar & Weichert [44], and Zienkiewicz & Taylor [510].
The material (or physical) non-linearity is related to non-linear stress-strain relations.
In view of the structural safety, one should rather avoid overloading the shell structure
beyond the threshold of elastic behavior. However, there are quite enough situations when
the material non-linearity cannot be neglected; one can mention here, for example, the limit
state examination of steel-structures, where the application of the material non-linear,
elastic-plastic analysis is crucial to the correctness of accomplished results.
17
Axelrad [30] used instead the phrase “specialized branches of shell theory”.
2.2. Geometrically non-linear of analysis of plates and shells 23
When the effect of large deformations is so significant that the linear approximation of
the strain-displacement relations cannot be accepted, one has to implement a geometrical
non-linear analysis. The geometrical non-linearity is also essential in the stability analysis
of structures. An additional line of separation in the scale of large deformations goes
between small strains and large strains. A large strain analysis for thin-walled structures
usually means that the thickness change during deformations is taken into account; see e. g.
Hughes & Carnoy [210], Simo et al. [440], Schieck et al. [421], Betsch et al. [57], Bischoff
& Ramm [59], Brank et al. [78], Başar & Kintzel [41], Başar & Grytz [38], Ibrahimbegović
[221].
Various approximate geometrically non-linear shell theories can be found in the
literature with a gradation established usually with respect to the scale of displacements
and/or rotations. The first systematic classification of large deformation shell theories
according to the magnitude of rotations was proposed by Pietraszkiewicz [360, 362], who
introduced four categories of shell theory: theory of finite, large, moderate and small
rotations, where the latter is in fact equivalent to a linear shell theory. In the literature, the
terms finite rotations and large rotations are very often treated as synonyms for each other
and also for unrestricted rotations. Then, the phrase moderate rotations becomes a kind of
an umbrella term characterizing the whole area between the linear shell theory and the
non-linear analysis of large (=finite=unrestricted) rotations. However, some authors set
apart one additional category which is located in that classification between the linear
theory and the moderate rotation theory (MRT); this simplified non-linear theory of shells
is based on von Kármán assumptions, what means that non-linear terms in the strain-
displacement relations contains only the derivatives of the transverse deflection18. Since the
original von Kármán-type non-linearity was assumed for thin (Kirchhoff) plates, its
extension for Reissner–Mindlin shells can be indicated as the Refined von Kármán (RVK)
theory, see e. g. Reddy [392, 396], Reddy & Chandrashekhara [400], Schmidt & Reddy
[426], Palmerio et al. [342, 343], Sing et al. [442], He [193], Fares et al. [158], Fares &
Youssif [157], Woo et al. [498]. Then, due to more non-linear terms retained in the strain-
displacement relations, each MRT can be considered as a more advanced approximation of
the non-linear model than the RVK theory19. Various MRT shell models have been
considered e. g. by Koiter [251], Pietraszkiewicz [359, 363], Librescu & Schmidt [292],
Schmidt & Reddy [426], and Schmidt & Weichert [427], Palmerio et al. [342], Altman &
Palmerio [9], Kreja & Schmidt [263, 264, and 265], Kreja et al. [269], Libai & Simmonds
[289], Carrera & Parisch [97]. Shell theories accounting for small strains and large rotations
were presented e. g. by Habip [186], Simmonds & Danielson [434], Pietraszkiewicz [359,
360, 362, 363], Reissner [405], Ramm & Matzenmiller [383], Başar [34, 35], Librescu
[290], Simo et al. [435, 437], Stander et al. [446], Wriggers & Gruttmann [500, 501], Başar
& Krätzig [42], Dennis & Palazotto [139], Saleeb et al. [412], Holzapfel [202], Parisch
[348], Büchter & Ramm [83, 82, 84], Chróścielewski et al. [120, 121] (see also [118, 119]),
Sansour & Bufler [417] (see also [414, 415, 419]), Başar et al. [37]), Damjanić et al. [134],
18
Such an assumption seems to be justified mainly by the observation that a plate is stiffer in-plane directions than
in the transverse direction. This approach was initiated by von Kármán in his non-linear theory of plates (1910);
however, quite often the name of Marguerre is also linked to that model due to his non-linear theory of shallow
shells published in 1938 (see e.g. Ciarlet & Gratie [124], Ferreira & Barbosa [162], Teng & Hong [464], Vassilev
[474]). A similar idea can be recognized also in the Donnell-Mushtari-Vlasov linear theory of shallow-shells;
Axelrad [30], Schmidt & Reddy [426], Teng & Hong [464], Vassilev [474], Woźniak [499].
19
Classical shell theories accounting for moderate rotations are frequently linked in the literature with the names
of Mushtari, Galimov, Sanders, Koiter and Novozhilov; see e.g. Axelrad [30], Pietraszkiewicz [362, 363], Teng &
Hong [464] or Toorani & Lakis [469].
Ibrahimbegović [217, 218], Jiang & Chernuka [230], Brank et al. [79, 80], Park et al.
[351], Valid [471, 472], Libai & Simmonds [289], Meek & Wang [313], Wiśniewski [496],
Zahrouni et al. [507], Masud et al. [309], Wang & Thierauf [483], Massin & Al Mikdad
[307], Campello et al. [86], Kulikov & Plotnikova [273] and Eremeyev [155]. Large
rotations with large strains were considered e. g. by Simo et al. [440], Schieck et al. [421].
Bischoff & Ramm [59], Libai & Simmonds [289], DiCarlo et al. [145], Brank et al. [78],
Başar & Kintzel [41], Başar et al. [39], Pimenta et al. [367], Brank [74], and
Ibrahimbegović [221].
Regardless the magnitude of strains considered in the particular shell model, a proper
description of finite rotations remains a considerable challenge itself. Mathematically, finite
rotations are represented by the rotation matrix, Crisfield [129], Korn & Korn [253]; or
rather by the proper orthogonal tensor belonging to the rotation group SO(3), Bauchau &
Trainelli [50], Betsch et al. [58], Chróścielewski et al. [120]. It is obvious that nine
elements of rotation tensor (or rotation matrix) can be expressed by smaller number of
independent parameters. It is well known that in the flight dynamics three parameters
named pitch, roll and yaw, Korn & Korn [253], are used to control an orientation of an
object in the 3-dimensional space. In such a case, the general rotation may be expressed as
a (non-commuting) sequence of rotations about three follower axes, characterized by
Cardan (Tait-Bryant) angles, which are very often identified with Euler angles, Büchter &
Ramm [83], Korn & Korn [253]. Another option is to consider rotations about 3 axes fixed
in space. All approaches based on the use of three Euler-like angles rotated in a certain
sequence are not free from singularities (often called gimbal locks), Crisfield [129]. The
issue of appropriate parameterization of rotations were examined in details e.g. by Bauchau
& Trainelli [50], Betsch et al. [58], Büchter & Ramm [83], Cheng & Gupta [110], Crisfield
[129], Pietraszkiewicz & Badur [364], Wang & Thierauf [483], Wiśniewski [496] and
Ibrahimbegović [219]. A singular-free parameterization of rotations can be obtained for
instance by the application of Euler–Rodrigues parameters (which can be identified with
the components of a quaternion) but at the expense of one additional parameter and an extra
constraint, Bauchau & Trainelli [50]. Appropriate modifications of that methodology can
lead to singular-free and also non-redundant vectorial parameterizations given by
parameters of Cayley, Gibbs, or Rodrigues or by the exponential map, Crisfield [129]. A
very detailed and mathematically formal discussion of the finite rotations in 3-dimensional
space can be found in the classical book of Altmann [10]20.
It is quite obvious that the application of three rotation parameters (in vectorial
parameterizations of finite rotations) results in 6-dof shell models (see e.g. Parisch [348],
Chróścielewski et al. [120, 121], Sansour & Bufler [417], Ibrahimbegović & Frey [222],
Sansour & Boćko [415], Ibrahimbegović [218], Sansour & Bednarczyk [414],
Chróścielewski [118], Wiśniewski [496], Campello et al. [86], Sansour & Wagner [419],
Chróścielewski et al. [119]. On the other hand, one can notice that the most of the
formulations for smooth shells neglect the drilling rotation of the director; therefore just
two rotation parameters are enough to describe the motion of the shell director. Brank &
Ibrahimbegović [77] wrote in that case about ”constrained finite rotations” which were
”unrestricted in size, and on the other hand constrained in the 3-d space in the direction of
the shell director”. The resulting shell model possesses five independent external
parameters at each point of the middle surface (5 degrees of freedom). One of the earliest
20
A rich collection of different formulas useful in the numerical treatment of finite rotations in 3-D space can be
found also in the Appendix of a quite recent paper by Felippa & Haugen [159]. A historical review of the
developments in that field an interested reader can find in e.g. Cheng & Gupta [110], Dai [132], Altmann [11].
2.2. Geometrically non-linear of analysis of plates and shells 25
practical applications of the 5-dof shell model accounting for finite rotations was presented
in 1976 by Ramm [378, 377] who employed Euler-like angles. Ramm formulation was
improved in Ramm & Matzenmiller [383]; comparable methodologies were applied also in
Hsiao & Chen [206]21, Stander et al. [446], Wriggers & Gruttmann [500, 501],
Rammerstorfer et al. [389], Başar [35], Başar et al. [37], Damjanić et al. [134], Brank et al.
[79], Kreja & Schmidt [266, 267, 268]. In 1990 Simo et al. [437] introduced a finite
rotation 5-dof shell formulation based on exponential mapping. Similar formulations can be
found also in Saleeb et al. [412], Parisch [348], Büchter & Ramm [84], Brank et al. [80],
Betsch et al. [58], Brank & Ibrahimbegović [77]. It is worth to notice that Balah and
Al-Ghamedy [31] extended that methodology also for the TOSD shell model.
Lee & Lee [280] combined 5-dof and 6-dof shell formulations in one computational
model, i.e. the smooth regions of the shell were analyzed with the 5-dof model, whereas a
6-dof formulation was applied for folded and boundary regions.
Review of theoretical models for a large rotation analysis of shells should not ignore
the continuum based shell models where the behavior of the shell-like construction is
represented without any reference to rotational degrees of freedom. This approach is closely
associated with the numerical implementation of theoretical models; therefore a more
detailed discussion of that concept will be presented in Section 2.4.
Quite often mathematical models of shells are categorized in the literature according to
the numerical strategy chosen for the solution of large displacement problem - one should
mention here especially a big group of shell models based on so called co-rotational
formulation. However, since this topic is strictly related to computational aspects, its
description was also located in Section 2.4.
As a final remark of the present section, it is worth to refer to a rather justified opinion
of Bischoff et al. [63] who stated that the “distinction between moderate and large
rotations is of mostly historical value”; however, as one can notice, quite many authors still
perform research in the field of simplified non-linear formulations, see e.g. Carrera &
Parisch [97], Teng & Hong [464], Jaunky & Knight [226], Kim & Voyiadjis [242], Axelrad
[30], Shen [429], Amabili & Paidoussis [12].
2.3. Theoretical models for layered thin-walled structures

The increasing use of laminated composites and sandwich panels demands a better
understanding of the behavior of multilayered thin-walled structures (see e.g. Hachenberg
[187], Vasiliev & Morozov [475]).
Generally, analysis of such complex structure as a fiber reinforced composite laminate
can be performed either from the micro-mechanical or macro-mechanical point of view,
Jones [231], Guz [183]. It is quite obvious that a precise study of interaction between the
fibers and the matrix can be examined in detail only in the micro-mechanical scale.
However, costs of micro-mechanical scale calculations of any real structure are still far too
high for practical applications; a much more realistic option but still quite costly is multi-
scale modeling (see e.g. Pagano & Yuan [336], Takano et al. [458], Schmauder [424], Pahr
21
Hsiao & Chen [206] tested six different strategies for finite rotation description in finite element analysis of
shells based on application of degenerated isoparametric elements. Their studies included three different forms of
Euler-like angles as well as the finite rotation vector method; however, as a matter of fact, no one of the shell
problems analyzed by Hsiao & Chen experienced finite rotations (compare Kreja [260]).
& Rammerstorfer [337], and a recent review paper by Ladevéze [275]). If one is interested
in an overall performance of a thin-walled structure made of fiber reinforced composite
laminates, then the macro-mechanical modeling can be applied, where all micro-scale
effects are smeared in a phenomenological material model.
The macro-mechanical models of laminated plates and shells are usually constructed
according to an appropriate lamination theory, where it is assumed that the laminated panel
is made up by a certain number of layers, which are supposed to be perfectly bonded
together. In such a model a single layer is considered as an elementary and homogenous
part of the structure. Therefore, even in a case of fiber reinforced composite laminates each
lamina can be considered as a complete physical entity instead of a collection of isolated
components. The effective properties of a layer made of any heterogeneous material can be
obtained by the homogenization that may be understood as “finding a homogeneous
comparison material that is energetically equivalent to a given microstructured material”,
Böhm [71] (see also Pagano & Yuan [336], Schmauder [424] or Zhang & Evans [508]).
Basically, talking about 2D computational models for multilayered shells one can
distinguish between two primary categories of lamination theories: the Equivalent Single
Layer (ESL) model22 or Discrete-Layer (DL) theory23, Noor & Burton [329], Kulikov
[270], Carrera [92], Reddy [394], Rohwer et al. [407].
Equivalent Single Layer models

In the ESL model the entire laminate is represented by a single-layer panel with
macro-mechanical properties estimated as a weighed average of the mechanical properties
of each lamina.24 The Equivalent Single Layer model used in conjunction with the classical
Kirchhoff-Love theory of thin shells/plates is commonly known as the Classical Lamination
Theory (CLT) (see e.g. Ambartsumyan [14], Jones [231], Rohwer et al. [407], Vinson &
Chou [476]). Sun & Chin [450] presented the von Kármán-type CLT model for a
geometrically non-linear analysis of laminated composite plates; a similar large
displacement formulation for thin composite shells was given by Saigal et al. [411].
Composite multilayered shells are typically characterized by a large ratio of Young's
modulus to shear modulus, even if they are relatively thin (see e.g. Chou & Kelly [117],
Jones [231] or Vinson & Chou [476]). Therefore, the CLT formulation, which does not
account for the transverse-shear deformation, is rather inadequate for accurate prediction of
the elastic behavior of composites. In order to overcome that limitation a refined lamination
theory is required that accounts for transverse shear deformation. An extension of the
FOSD model for laminated plates was described by Whitney & Pagano [493], whereas
Dong & Tso [146] presented FOSD model for composite shells (see also Jones [231],
Reddy & Arciniega [399], and Rohwer et al. [407], Vinson & Chou [476], Ghugal &
Shimpi [173]). Reissner [402] developed a FOSD model designated for sandwich shells.
Various variants of CLT and FOSD shell theories were examined in a linear static analysis
of cylindrical laminated shells by Chandrashekhara & Pavan Kumar [102, 101]. An
interesting asymptotic formulation of the FOSD multilayered plate theory based on the
22
Some authors (see e.g. Kulikov [271], Noor & Burton [329], Noor et al. [330] and Shu [431]) used to categorize
this group as “global approximation theories”.
23
Discrete-layer models appear also very often in the literature as layer-wise theories, see e.g. Başar & Ding [36],
Carrera [91, 92], Carrera & Demasi [95, 96].
24
A single layer model replacement for the heterogeneous panel can also be constructed with a help of the
homogenization technique, see e.g. Chen et al. [109], Lewiński [284], Rabczuk et al. [376] or Tanov & Tabiei
[459].
2.3. Theoretical models for layered thin-walled structures 27
mixed Hellinger-Reissner variational principle was presented by Tarn & Wang [461]. Large
deformation formulations based on the FOSD model was presented e.g. by Reddy &
Chandrashekhara [400], Schmidt & Reddy [426], Palmerio et al. [342], Kreja et al. [269].
It was already stated in Section 2.1 of the present report that FOSD models require an
appropriate transverse shear correction due to a constant shear distribution across the shell
thickness resulting from the linear interpolation of the displacement field in that direction.
One should realize, however, that the estimation of a shear correction for laminated
composites is much more complicated than for homogeneous panels. In the literature one
can find innumerable proposals of different formulas for appropriate shear correction
factors, depending on material properties and also on such geometrical characteristics of the
laminate as stacking sequence of the layers and their ply angles. Most of those formulas
have been determined by matching the transverse shear strain energy predicted by the
FOSD plate model with that obtained from the three-dimensional elasticity theory (see e.g.
Dong & Tso [146], Jemielita [229], Vlachoutsis [477], Whitney [492], Wittrick [497]).
Noor & Peters [331] calculated the transverse shear correction factors for multilayered
cylindrical panels by means of the predictor-corrector approach (see also Sze et al. [454]).
A similar methodology was applied by Auricchio & Sacco [26, 27], who determined the
shear correction in an iterative manner. Both those tactics based on the comparison between
the shear energy computed for the transverse shear stress obtained from constitutive
relations and the shear energy calculated for transverse shear stress recovered from the
three-dimensional equilibrium. Another approach was proposed by Pai in [338] where not
only the transverse shear strain energy but also the shear stress resultants estimated with the
FOSD model were balanced with those calculated by the layer-wise higher-order shear
theory.
An interesting alternative to the employment of the shear correction factors was
proposed by Rolfes & Rohwer [408] who introduced an ”improved” transverse shear
stiffness in the FOSD model as calculated with the assumption of a cylindrical bending
mode and by utilizing the differential relation between the resulting transverse shear forces
and bending moments.25 Altenbach [7, 6] estimated the shear stiffness of layered plates by
comparing the forces and moments calculated from two-dimensional and three-dimensional
models. Tanov & Tabiei [460] presented a Corrected FOSD model where they enforced a
parabolic shear strain distribution across the shell thickness, what not only improved a
profile of the transverse shear stress but also eliminated the need for using a shear
correcting factor. Although the authors of [460] classified themselves their model as the
“displacement-based formulation” it should be rather classified as the mixed formulation
based on the Reissner partial-mixed variational principle, Reissner [404]. Similar strategy
was adopted by Fares & Youssif [157], Fares et al. [158], and Auricchio & Sacco [28, 26,
27] who presented a collection of different Refined FOSD models, all based on independent
approximations of the transverse stress fields introduced in the (partial-)mixed variational
principle, but varied in the final number of unknowns. A slight different version of Refined
FOSD theory, although corresponding to some extent with the model of Tanov & Tabiei
[460], was proposed by Qi & Knight [374] and was labeled as a Consistent FOSD in the
subsequent paper of those authors, Knight & Qi [249]. In their approach the effective
transverse shear strains of the FOSD was treated as the stress-weighted average of the
through-the-thickness transverse shear strains based on the equivalent shear strain energy.
Evaluation of the effective shear stiffness in that formulation is comparable to the
application of the transverse shear correction factor.
25
Similar procedure was incorporated also in the commercial FEA system MSC-Nastran [318].
For thick panels a significant improvement of results can be obtained by applying the
HOSD models, where the conventional displacement equations of FOSD are supplemented
with various higher-order terms. In the most popular methodology of the “derived
approach” a 2-D plate/shell model is constructed by applying a power series expansion of
the displacements and strains with respect to the thickness coordinate. Starting with that
formulation one can obtain a shear deformation model of an arbitrary order depending on
the selected level of truncation. In 1984 Reddy [393] developed a Third Order Transverse
Shear Deformation (TOSD) theory for laminated plates assuming a cubic representation of
the displacement field with respect to the height coordinate (see also Khdeir et al. [236],
Phan & Reddy [355], Reddy [394, 395] and Reddy & Arciniega [399]). An extension of
that model for laminated shells was presented by Reddy & Liu [401]. By using the
condition of vanishing transverse shear stresses on top and bottom surfaces they reduced a
set of unknowns to exactly the same five displacement components as in the FOSD model.
Three various TOSD models of multilayered plates were examined in the linear analysis by
Bose & Reddy [67]. Reddy extended his TOSD model for the von Kármán-type non-linear
plate theory in [396]. Dennis [137] and Simitses [432] used a similar approach to obtain
analytical TOSD solutions for circular laminated cylindrical panels modeled within the
range of the von Kármán geometric non-linearity by means of the modified Galerkin
method. A corresponding formulation for large rotation shell theory was presented by Başar
[35] and by Başar et al. [37], who considered also a 7-dof model where two extra
displacement variables were included due to enriched approximation of the displacement
field across the thickness of the panel26. A similar TOSD model with 7 dofs was proposed
for geometrically non-linear shells by Balah & Al-Ghamedy [31], who additionally applied
singularity-free description of finite rotations based on exponential mapping after Simo et
al. [437]. Another implementation of the Reddy concept of TOSD [393, 401] can be found
in the Simplified Large Rotation (SLR) formulation proposed for laminated shells by
Palazotto and co-workers, Dennis & Palazotto [138, 139], Naboulsi & Palazotto [319], and
Tsai et al. [470]. The second-order shear theory (SOSD) for laminated shells in the range of
moderate rotations was examined by Sacco & Reddy [410] who found that inclusion of
second order terms did not significantly improve the linear solution over the FOSD model.
Moita et al. [315] applied the HOSD model to investigate the buckling behavior of
laminated panels. Piskunov [369] described an iterative analytical theory of composites,
where starting from the CLT formulation, one can obtain a HOSD-equivalent model by
successive approximations. Quite a general geometrically non-linear HOSD laminated shell
theory was presented by Librescu [290].
The 3-D elasticity solutions of laminated plates (see e.g. Pagano & Hatfield [335])
exhibited rapid changes of the displacement profile at the interfaces between two
contiguous layers. This phenomenon is commonly classified as the zig-zag effect. To
account for that feature the kinematical model of the layered shell should be enhanced by
adding some warping functions that are capable to represent the deformed profile with a
different slope in each layer.27 The ability of different shear deformation theories to
represent the deformation profiles of a layered panel is illustrated in Fig. 2.4.
26
More comments on the Başar model will be presented in Chapter 3.
27
A zig-zag effect can be recognized also in displacement fields obtained in some stress based formulations (see
e.g. Ambartsumyan [14]).
Fig. 2.4: Deformation profiles of a layered panel represented by different shear deformation models
By assuming a piecewise linear approximation for the warping function one in fact
adopts the FOSD hypothesis for each individual layer of a multilayered shell (see e.g.
Brank [73], Brank & Carrera [75, 76], Carrera [94], Di Sciuva [141], Toledano &
Murakami [466]). However, in the pioneering zig-zag model for multilayered plates28
presented by Ambartsumyan [14] the resulting through-the-thickness distribution of in-
plane displacement field is cubic in each layer; similar piecewise TOSD zig-zag models
were considered by Di Sciuva [142], Di & Rothert [144, 143], Lee et al. [279], Toledano &
Murakami [467]. The warping function can be given explicitly, for example as a zig-zag
function connected with two additional unknowns for the whole cross section (Toledano &
Murakami [466, 467], Di & Rothert [144, 143], Carrera [87, 94]). Savithri & Varadan [420]
presented a TOSD formulation for composite plates, where the zig-zag effect was included
by application of Heaviside step function in the description of the displacements
distribution across the plate thickness. Another way is to construct the warping function by
invoking interlayer shear stress continuity conditions and zero shear traction boundary
conditions on the upper and lower bounding surfaces. Then the resulting model contains
exactly the same number of unknowns as the standard FOSD formulation (Di Sciuva [141,
142], Lee et al. [279], Librescu & Schmidt [293], Schmidt & Librescu [425], He [193], Shu
[431]), but requires C1 type continuity in the FEM implementation. Lee & Cao [278]
proposed a predictor-corrector method of laminated shells analysis based on zig-zag models
of Di Sciuva [141] and Lee et al. [279]. An interesting variant of that approach was
presented by Soldatos & Watson [444] who proposed to enhance a standard 5-dof model
for small displacement analysis of laminated plates by an introduction of special shape
functions, which could be determined a posteriori from the stress equilibrium equations.
Working along similar lines, Cho et al. [114] proposed the Efficient Higher-Order Shell
Theory based on an overall cubic distribution of in-plane displacements combined with a
piecewise linear profile. Similar model but with cubic function replaced with the sinus
function was proposed by Idlbi et al. [224] and by Fernandes [161]. Arya et al. [23]
proposed zig-zag model enhanced by the application of several trigonometric functions in
the displacement field. Hassis [191] in his higher order plate theory introduced a warping
function constructed on the basis of deformation modes of the normal fiber treated as a
geometrical beam. Such an approach resembles the earlier idea of Sutyrin & Hodges [453],
who applied the variational-asymptotical method to split the 3-D analysis of plate
deformation into two separate reduced-dimensional problems: a 2-D Reissner–type plate
theory analysis and a 1-D through-the-thickness analysis (an extension of that idea for non-
28
According to an extensive historical review on zig-zag models developments published recently by Carrera [93],
the first model accounting for the zig-zag effect was proposed for multi-layered beams by Lekhnitskii in 1935.
linear shell theory was presented by Yu & Hodges [505] and Yu et al. [506]). Quite
recently, Kim & Cho [238] presented an Enhanced FOSD theory for laminated plates
constructed as weighed least-square approximation of a 3-D theory. The warping function
incorporated in that formulation was obtained with the HOSD plate theory, Cho et al. [114],
and the resulting effective transfer shear stiffness could be considered as being analogous to
that used in the Consistent FOSD of Knight & Qi [249].
Discrete-layer theories
Despite the fact that the performance of ESL models can be significantly improved by
inclusion of various warping functions29, it is almost impossible to construct a universal
ESL model which would be equally efficient for symmetrically and asymmetrically
laminated panels. Therefore, the next step on the way to increase the accuracy of the
multilayered shell models has to go beyond the limits of a single layer model, i.e. it is
necessary to consider each layer separately within discrete-layer (DL) theories named also
the layer-wise formulations30. The laminate in the DL theory is treated as a stack of laminas
bounded together by appropriate conditions at ply interfaces. Since each lamina is treated
individually, the number of unknowns in DL theories depends on the number of layers, N.
Kulikov [270] (see also Piskunov [369]) claimed that the DL theory originated from the
layer-wise description of sandwich panels introduced by Grigoluk in the 1950s.31 A layer-
wise theory presented for laminated plates by Mau in 1973 [310] used 4N+1 displacement
unknowns accompanied by 2(N-1) additional unknown variables representing the
interlamina shear stresses. In 1978 Pagano [334] presented an approximate theory for stress
analysis in composite laminates, where he assumed in-plane stresses represented within
each layer by linear functions of the thickness coordinate. The stress equilibrium equations
were expressed in force and moment resultants and formulated separately for each layer,
and the set of equations were supplemented with appropriate interface conditions. A final
number of unknowns in the Pagano model [334] was equal to 13N. One can notice that the
order of computational complexity of both just mentioned formulations is relatively high,
especially that Mau [310] as well as Pagano [334] suggested to model each physical layer
with two or three computational sub-layers to provide a satisfactory accuracy of the results.
Much more economical DL theories were based on an independent shear deformation of the
director associated with each individual layer and involving just 3+2N displacement
unknowns (three global displacements for the whole laminate and two local rotations for
each layer). An example of such an approach was a layer-wise laminate plate theory
proposed by Mawenya & Davies [311] and described also by Reddy [395, 394]. A
corresponding laminated shell model was presented by Chaudhuri & Seide [108] as the
29
See for instance Carrera [94], who showed that FOSD models with zig-zag functions were more effective in a
laminated plate analysis than HOSD models without zig-zag functions.
30
It is worth to notice that some authors used to extend the term “layer-wise formulation” also to include those
equivalent single layer models which were enhanced by addition of some warping functions (see e.g. Rohwer
[407]). To some extent such an approach can be justified by the common in the both models abandonment of the
Cz1 requirements, what means that functions describing the displacement distribution in thickness direction can
exhibit rapid changes of slopes at the interfaces between two contiguous layers. However, the main difference
between the ESL model with the warping function and the DL formulation is that the number of unknowns in the
ESL model does not depend on the number of layers (usually after taking advantage of some compatibility
conditions to eliminate local unknowns).
31
On the other hand, any DL model was mentioned neither in a review article on layered shells theories by
Ambartsumyan [13] from 1962, nor in a survey of developments in the analysis of sandwich structures published
three years later by Habip [185].
“layerwise constant shear-angle theory” (see also Chaudhuri [107]), and by Noor & Burton
[330, 329] as the “discrete layer shell theory”. A similar formulation but with 3(N+1)
unknowns due to accounting for 3-D effects was presented for laminated shells by
Huttelmaier & Epstein [216] as well as by Masud & Panahandeh [308]. A “multi-director
formulation”32 of Pinsky & Kim [368] accounted for visco-elastic material behavior and
large deformations including the thickness stretching; therefore the number of unknowns in
that model was extended to 3+4N. Cho & Averill [116] combined the zig-zag model of Di
Sciuva [141] with a layer-wise formulation obtaining a “First order zig-zag sublaminate
plate theory” with 5(N+1) unknowns. The generalized laminated plate theory of Reddy
[289] (see also Barbero et al. [32]) offers a quite universal description of the layer-wise
model with an arbitrary order of displacement interpolation within each layer assuming
kinematical variables located at the interfaces. A similar concept was considered by
Gaudenzi et al. [171] who, however, imposed the continuity of interlaminar stresses only at
selected interfaces in order to limit the total number of unknowns. Carrera [89] presented a
mixed variational formulation of a layer-wise multilayered plate theory with variable fields
of displacements and transverse stresses interpolated by Legendre polynomials of a chosen
order (see also Carrera [91], Carrera & Demasi [95, 96]). A displacement formulation of
that model was considered in Carrera [89] as a special reduced variant of the layer-wise
plate theory with limited number of unknowns but also without continuity of transverse
shear and normal stresses. Başar [35] and Başar et al. [37] introduced DL models with
inextensible multi-director (3+2N unknowns) for finite rotation analysis of composite
shells. A corresponding model based on the geometrically exact shell formulation of Simo
et al. [435, 436, 437] was presented by Vu-Quoc et al. [478] with the main assumption that
“the transverse fiber across the whole multilayer shell deforms as a chain of rigid links that
are connected to each other by universal joints”. Başar et al. [40] and Braun et al. [81]
applied the 7-parameter FOSD shell theory33 for each single layer, what resulted in DL
formulations with 3+4N unknowns. Başar & Ding [36] considered the transverse normal
strains in their DL models with 3+3N, 3+4N, and 3+6N unknowns. Gruttmann & Wagner
[182] presented a DL multilayered shell model based on the HOSD theory with 3+9N
unknowns. Williams & Addessio [494] constructed a DL model for analysis of plate
delamination problem; in their model the layer displacement variables were supplemented
with interfacial traction terms and appropriate evolution laws describing the damage
growth.
It is quite interesting that traces of DL formulations can also be found in some ESL
zig-zag models. As a typical example one can consider the ESL zig-zag theory of Di Sciuva
[141], who started his derivations assuming a multi-director description of the displacement
field with independent rotational parameters in each layer. In the next step of the
formulation those parameters were eliminated by invoking the shear stress constraints. As it
was mentioned earlier in this chapter, the final number of unknowns in the ESL model of
32
It is crucial to distinguish between two quite different concepts of shell kinematics that are quite often identified
in the literature with the common label of the “multi-director model”. On the one hand, Pinsky & Kim [368],
Braun et al. [81] and Wagner & Gruttmann [481] used this term to describe DL models with independent director
vectors in each layer, on the other hand, Krätzig [254] presented a multi-director single-layer shell model that can
be considered as being analogous to the general HOSD shell theory of Librescu [290], but in contrast to the latter,
belonging to the category of the “direct approach” methodologies (see also Başar et al. [40], Krätzig & Jun [256,
257]). A prototype of that formulation was given earlier by Naghdi (see e.g. equation (2.22) in Naghdi [322]) but
without introducing the name: “multi-director model”. To avoid any possible confusion, in the present report the
expression “multi-director model” stands only for the DL formulations being equivalent to that presented by
Pinsky & Kim [368].
33
As described earlier in this chapter (see Section 2.1)
Di Sciuva [141] (and also in other models of that kind by Librescu & Schmidt [293],
Schmidt & Librescu [425], He [193], Shu [431]) was equal to five. A quite similar concept
can be recognized in the laminate theories proposed by Li & Liu [286], where the
displacement field was expressed with global components of TOSD theory and local
components of DL model combined within so called global-local superposition technique.
After taking advantage of continuity conditions the model constructed by Li & Liu [286]
used 13 layer independent variables, what makes six more unknowns than in a standard
TOSD model.
Three-dimensional and combined models

It should be emphasized that either the ESL or DL formulations described above are in
fact 2-D models; one can quite easily imagine that multilayered shells can be analyzed also
with the use of 3-D models based on elastic three-dimensional continuum. However,
practical application of such models is very limited due to the dramatic increase of the
number of unknowns. Some of analytical 3-D solutions presented by Pagano in 1969 and
1970 for selected examples of simply supported laminated composite plates under
distributed transverse loads34 serve as classical benchmark problems. They are still used by
many researchers for the purpose of validating their diverse theories of laminated plates
(see also Pagano & Hatfield [335], Pagano [333, 334]). Bogdanovich & Yushanov [65]
presented a 3-D displacement-assumed variational analysis of laminated composite plates.
Williams & Addessio [494] extended Pagano 3-D model to include the problem of
delamination. Desai et al. [140] and Feng & Hoa [160] modeled layered composites as a
stack of 3-D solid sub-elements along the thickness of composite laminates.
It seems quite obvious that any attempt to use a detailed 3-D model for the whole
analyzed panel very soon would finish in exceeding reasonable limits of practical
application. Therefore a concept of global-local analysis (Feng & Hoa [160], Cho &
Averill [116], Kong & Cheung [252])35 appears to represent a rational compromise between
the requirements of precise analysis and realistic costs of calculations. The primary step is
to select those regions of the analyzed structure, named local zones, where a more detailed
analysis (3-D or at least DL model) is required due to vicinity of holes, constrained
boundaries or other discontinuities. One can assume that the remaining parts of the
structure, the global zones are more regular and therefore they can be treated with accepted
accuracy as regions of a single layer panel with equivalent laminate properties (ESL
model). Depending on specifics of the applied approach a special treatment of transition
zones may be necessary, see Feng & Hoa [160], Kong & Cheung [252]. By combining
local and global zones in one computational model one can significantly improve the
accuracy of the analysis staying within a moderate range of computational costs.
Recovery of transverse stress

Using a FOSD model with a proper estimation of a transverse shear stiffness one can
attain a quite satisfying accuracy of a global response of moderately thick laminated panels.
34
The most frequently cited papers of Pagano were unavailable to the author of the present report. One can just
recite them e.g. after Pagano & Hatfield [335]: Pagano, N.J.: Exact Solutions for Composite Laminates in
Cylindrical Bending, Journal of Composite Materials, 3, 1969, 389-411; Pagano, N.J.: Exact Solutions for
Rectangular Bidirectional Composites and Sandwich Plates, Journal of Composite Materials, 4, 1970, 20-34.
35
Yu et al. [504] used a similar methodology, but the separation into local (DL) and global (ESL) regions was
performed within the thickness of the laminate.
However, whereas the FOSD results for deflections and rotations or even for in-plane
stresses are acceptable, a direct application of constitutive relations must result in a wrong
profile of the transverse shear stresses (being constant across the thickness of each layer). A
simple correction of the FOSD model resulting in a much more realistic distribution of the
transverse shear stress can be obtained when the transverse stresses are calculated from the
stress equilibrium condition of the in-plane stress components (see e.g. Cen et al. [99],
Hossain et al. [205], Rolfes & Rohwer [408], Sze et al. [454], Alfano et al. [4]). Carrera
[90], Das et al [135] and Rohwer et al. [407] showed that using the stress equilibrium for
the estimation of transverse stresses can also be very effective for HOSD or DL models36.
A slightly different post-processing method was constructed by Cho & Kim [115], who
applied a displacement approximation of the HOSD theory to reinterpret the results from
the FOSD analysis by matching rotational variables of both kinematical models (similar
approach can be found also in Cho & Kim [113]). The improved displacement field
predicted by that procedure provided a satisfactory accuracy of transverse shear stress
calculated directly from the constitutive relations. It is also worthy to notice that in mixed
formulations transverse stresses can be calculated directly as primary variables (see e.g.
Auricchio & Sacco [28]) or at least the shear stress profiles can be substantially improved
by direct calculation of stress resultants being primary variables, Auricchio & Sacco [26,
27]. Post-processing method of stress recovery can also be applied in geometrically non-
linear analysis of laminated shells as presented e.g. by Lee & Lee [277], Park et al. [350].
Lee & Lee [277] developed an equilibrium-based stress recovery method that utilizes the
in-plane stresses and shear forces obtained by a shell element analysis and one-dimensional
FEM approximation introduced along the thickness in the post-processing phase. Park et al.
[350] calculated transverse stress in finite rotation analysis by piecewise integration of the
three-dimensional stress equilibrium equations in the thickness direction. A review on a
priori and a posteriori methods of transverse stress evaluation in multilayered plates was
presented by Carrera [90] and Kant & Swaminathan [234].
Sandwich panels as a special case of multilayered structures

It is worth to notice that sandwich panels can be in general considered as multilayered
structures and as such they can be analyzed with most of the theoretical models surveyed
above. However, there is also quite a broad category of computational models that are
specially adjusted to deal with sandwich panels by a compliance with the specific features
of those structures. A typical sandwich panel is constructed from two thin face-sheets
separated by a thick but usually lightweight core. As a rule, the core is made of a low
strength material: the most popular are foam or honeycomb cores but one can also find a
core constructed as a 3-D truss-structure. All those materials are characterized by low in-
plane and flexural rigidities, for that reason usually only shear stiffness is considered for the
core, whereas the thin plate/shell model is applied for the faces. Such assumptions were
introduced in one of the earliest sandwich plate models proposed by Hoff [198] in 1950,
similar approach was also adopted in one of the first non-linear finite element formulations
for sandwich plates presented by Schmit Jr. & Monforton [428] in 1970. On the other hand,
the thickness of outer face-sheets is significantly smaller than the height of the core; hence
already in 1947 Reissner [402] suggested that the bending stiffness of facings about their
own midsurface can be neglected. In 1991 Lewiński [283] proposed a more general
formulation of a sandwich plate model, which by introducing some simplifying
36
This strategy was applied also in the CLT models; see e.g. Ambartsumyan [14] or Jones [231].
assumptions can be reduced to that of Hoff [198] or Reissner [402]. Glockner & Malcolm
[176] and Malcolm & Glockner [301] constructed computational model based on the
Cosserat surface theory, where they treated the face-sheets as membranes of negligible
thickness. Holzapfel & Wimmer [203] investigated influence of shear deformation in the
geometrically non-linear Finite Difference Method analysis of sandwich plates.
Marcinowski [306] performed the geometrically non-linear analysis where sandwich shells
were modeled in a manner analogous to that proposed by Reissner [402]. Das et al. [135]
proposed a HOSD sandwich shell model with seven weighted-average displacement
variables. Vu-Quoc et al. [478] presented a DL formulation for sandwich shells based on
geometrically exact shell formulation of Simo et al. [435, 436, 437]. Borsellino et al. [66]
performed experimental tests and numerical simulations of sandwich structures with
composite facings.
Developments in the analysis and modeling of sandwich structures were reviewed in
1965 by Habip [185], and more recently by Burton & Noor [85] (1995), Librescu & Hause
[291] (2000) and by Hohe & Librescu [200] (2004).
Material modeling in analysis of multilayered shells

The most popular, linear elastic material models of multilayered shells are very well
established in the technical literature. Probably the most frequently cited resources in that
context are handbooks of Jones [231] and of Vinson & Chou [476], both published in 1975;
more recent texts were prepared by Nettles [323], Stockton [447], and Vasiliev & Morozov
[475]. Phenomenological observations of most high-strength fiber reinforced composites
seem to justify assumption of their linearly elastic response; however, one can easily
indicate also many cases where material non-linearity should be taken into consideration.
Nevertheless, the label of “composite materials” covers so large range of materials that it is
just impossible to get a general approach of the mechanical behavior; the following short
survey is limited to some selected examples only.
A constitutive model applied by Hu [207] in buckling analysis of fiber-composite
plates accounted for elastic material non-linearity restricted only to in-plane shear terms37.
Abu-Farsakh et al. [1] examined inelastic static response of laminated composite beams
using the secant modulus model. Woo et al. [498] analyzed laminated orthotropic plates
using anisotropic elastic-plastic material model, based on Prandtl–Reuss flow rule with
37
This issue corresponds to the important question of an effective experimental determination of the in-plane shear
stiffness of a composite plate. It seems quite obvious that the non-linearity of the relation between in-plane shear
stress and in-plane shear strain was treated by Hu [207] as a material non-linearity. However, another option was
chosen by Pai & Palazotto [340], who handled this problem exclusively within geometrical non-linearity. Pai &
Palazotto [340] presented a geometrically non-linear co-rotational formulation for laminated shells accounting for
large strains and a change of fiber directions during deformation of a laminate. They used the right-stretch (Biot)
strain tensor and the work-conjugate Jaumann-Biot stress tensor, declaring that only those strain and stress
measures allow for using “experimentally obtained material constants in the constitutive equations” [340]. As far
as we know, their model has not been confronted with any popular benchmark problem for large deformation of
laminated shells. However, the results of a corresponding geometrically-exact curved beam model by Pai &
Palazotto [341] were very promising and the team decided to tackle with the extremely challenging task of a
computer simulation of inflation of a tire, Greer Jr. & Palazotto [179], though the material properties were
assumed as linear elastic. The effect of fiber rotation was also investigated by Wisnom [495], who examined how
much the change of fiber directions during deformation of a glass-epoxy laminate in ±45° tension tests can
influence the measured in-plane shear stiffness. Wisnom [495] concluded that the effect of fiber rotation is small at
shear strains below 7%, what means that this factor can be neglected in small strain analysis. Thus, the proposition
of Pai & Palazotto [340] to account for that effect in large strain analysis seemed to be quite reasonable; however,
according to the recent paper of Felippa & Haugen [159] application of a co-rotational formulation is restricted to
the range of small strains.
strain hardening and Huber–Mises yield criterion modified by introducing the parameters
of anisotropy. Kłosowski & Woźnica [248] (see also Bouhafs et al. [69]) analyzed a
rheological behavior of laminated plates and shells assuming constitutive relations given in
visco-elastic models of Perzyna, Chaboche and Bodner-Partom. Wagner & Gruttmann
[481] employed visco-plastic material model in an examination of delamination problems
in layered panels. Başar et al. [40] applied a hyper-elastic Mooney-Rivlin type constitutive
model in a large strain analysis of a sandwich shell with a rubber core.
An interesting example of micro-macro modeling of composite materials based on the
homogenization theory of periodic media was described recently by Takano et al. [458]. In
this formulation the composite is treated as the assembly of periodic microscopic structures.
Assuming that microscopic periodicity remains in the local region also under large
deformation, the local region is replaced by the homogenized model. However, in general,
due to large deformations the change of microstructures in one local region is different
from that in other region; therefore during deformation the microstructures have to be
updated for local regions. An application of that procedure in an analysis of knitted fabric
composite materials seems to be especially promising. More on a multi-scale modeling of
composite materials can be found in a recent book by Böhm [71] (see also Pagano & Yuan
[336] and Ladevéze [275]).
A review of recent developments in theoretical modeling of composite materials
including inelastic behavior and damage was given by Dvorak [151], who among other
things indicated also growing “ability to design physical properties of composite material
systems and structures for different specific purposes”, [151]. Hohe & Becker [199]
presented a survey on material representation for cellular sandwich cores, including plastic
and non-linearly elastic models.
More on multilayered plates and shells modeling

This chapter does not pretend to present a complete survey of all theoretical models
proposed for layered thin-walled structures, but rather to show the most relevant ideas in
that field. Extensive reviews on analysis of multilayered plates and shells were presented
e. g. by Reddy [394, 395, 392], Noor & Burton [329], Noor et al. [330], Simitses [432],
Carvelli & Savoia [98], Bose & Reddy [67], Jaunky & Knight [226], Kant & Swaminathan
[234], Carrera [91, 92, 93], Ferreira & Fernandes [163], Toorani & Lakis [469], Altenbach
& Altenbach [8], Carrera & Demasi [95, 96], Ghugal & Shimpi [173], Piskunov &
Rasskazov [370], Qatu [373], Piskunov [369], Reddy & Arciniega [399], Rohwer et al.
[407]. See also collection of papers edited by Guz [183] and Hult & Rammerstorfer [215].
2.4. Numerical implementation of plate and shell theories

Analytical vs. numerical solutions
Since the most natural form of governing equations for classical shell theories were
partial differential equations of stress/force equilibrium associated with appropriate
boundary conditions, it seems quite obvious that the possibilities of obtaining the analytical
closed-form solutions were very limited. For a long period a practical implementation of
various plate and shell theories was restricted only to calculations performed with
approximated procedures for idealized computational models, concerning the geometry,
loading and boundary conditions (see for example popular engineering handbooks by
Girkmann [175] or Flügge [166]). Already in the 19th century the bending problem of
simply supported rectangular plates was solvable by expanding the loading and the
searched deflection into double trigonometric series as proposed by Navier in 1820. At the
turn of the 19th and 20th centuries, Levy introduced an application of single trigonometric
series to solve the bending problems of rectangular plates having two opposite simply-
supported edges. Both those strategies could be extended to the cylindrical panels and
shallow shells with double curvature possessing rectangular plan-forms, nevertheless large
portion of problems still reminded outside the range of attainable exact solutions. In 1908
Walter Ritz presented an approximate solution to the problem of rectangular clamped plate
by assuming a series of admissible trial functions that meet boundary conditions (see
Taylor [462]). In the Ritz method (or Rayleigh-Ritz method) the unknown solution is
approximated by a linear combination of N known basis functions parameterized by N
unknown coefficients that can be calculated based on the principle of stationary potential
energy. An alternative approximate method of solving the bending problem of rectangular
plates was proposed by Boris Galerkin in 1915. Similarly as in the Ritz method, the
unknown function was replaced by a series approximation, but Galerkin multiplied the
approximated differential equation by each function in the series and the whole result
integrated over area of the plate. By setting that integral to zero he obtained the system of
algebraic equations that can be solved to obtain the discrete parameters. Today we use to
classify the Galerkin method as a member of the Weighted Residual Methods, Korn & Korn
[253]. Since the trail function used in the Weighted Residual Method is only an
approximation, the differential equation is not satisfied and non-zero residual remains. If
we multiply the residual by a weighting function and equate the weighted integral to zero,
we can understand such an operation as enforcing the satisfaction of differential equation
over the whole region but in a weak (integral) sense. An extensive description of Ritz
method and Weighted Residual Methods, also in the context of the plate and shell analysis,
can be found in the recent book of Reddy [398]; see also Zienkiewicz & Taylor [510].
Computerized analysis of thin-walled structures

One can imagine that the practical usability of any method of solving a partial
differential equation by converting it into a system of algebraic equations is determined by
the limits of an available solver. Therefore, it is quite obvious that the appearance of analog
computers changed a lot in that field. Nowadays the majority of “analytical solutions” for
plates and shells are obtained numerically. One of the most popular approach is to convert
the differential equilibrium equations into a system of algebraic equations e.g. by
expressing the loads and displacements in terms of either single trigonometric series (Levy
type solutions, see e.g. Vinson & Chou [476], Phan & Reddy [355], He [193], Cho & Kim
[115], Idlbi et al. [224], Williams & Addessio [494], Chandrashekhara & Pavan Kumar
[102], Hassis [191]) or double trigonometric series (Navier type solutions, see e.g. Reissner
[402], Mau [310], Vinson & Chou [476], Reddy & Liu [401], Wittrick [497], Reddy [394,
397], Barbero et al. [32], Chandrashekhara & Pavan Kumar [101], Di & Rothert [144, 143],
Simitses [432], Shu [431], Tarn & Wang [461], Fares et al. [158], Kim & Cho [238]).
Pagano [334] presented a simplified analytical DL model for multi-layered plates with
governing equations based on layer equilibrium and interlaminar continuity conditions
expressed in terms of force and moment resultants and interlaminar tractions. Sun & Chin
[450] presented closed-form solutions of differential equations describing the cylindrical
bending for laminated composite plates assuming a CLT model with the von Kármán-type
geometrical non-linearity.
2.4. Numerical implementation of plate and shell theories 37
A real progress in practical analysis of thin-walled surface structures was strictly

related to the development of more advanced computerized methods, mainly the Finite
Element Method (FEM), but also the Finite Difference Method (FDM) (see e.g. Holzapfel
[202], Holzapfel & Wimmer [203]). Both those methods can be considered as numerical
schemes for approximate solution of partial differential equations. However, finite
differences can be considered as an approximation to the differential operator38, whereas
FEM offers an approximate solution of differential equations by converting the problem to
its weak, variational form. In that sense FEM can be treated as a form of Galerkin
variational method39, although the test functions of Galerkin method are assumed for the
whole analyzed domain, while FEM is characterized by a mesh discretization of a
continuous domain into a set of discrete sub-domains (finite elements) with test functions
(shape functions) defined locally within a single finite element. Due to that feature FEM is
much more versatile than Ritz or Galerkin methods in the treatment of arbitrary shapes,
different boundary conditions and various loading. Without any question, today FEM
predominates in the computational analysis of structures40. Comprehensive presentations of
theoretical basis and practical aspects of the Finite Element Analysis (FEA) can be found in
many handbooks, with the most recognized classical books written by Zienkiewicz &
Taylor [510] and also by Bathe [46], Belytschko et al. [52] or Crisfield [129].
Computational aspects of non-linear analysis

The non-linear analysis of plates and shells requires a proper treatment of the non-
linearity of the governing equations. The dominant numerical methodology of solving non-
linear equations consists in a combination of an incremental analysis and equilibrium
iterations; see e.g. Bathe [46], Crisfield [129], Kleiber [244], Waszczyszyn [485] or
Zienkiewicz & Taylor [510]. Probably, in the most natural form the solution of any non-
linear equilibrium equation can be presented as an equilibrium path in the load-
displacement space. In the incremental analysis (known also as the continuation method)
the total load is applied gradually in small increments that allowed for an acceptance of
linearization assumptions. After solving a sequence of linear incremental equations, one
linear equation for each increment, the final solution is composed as a collection of
equilibrium points in the load-displacement space, forming a discrete version of the
equilibrium path.
The incremental approach was used already in 1960s (see Clough & Wilson [128] or
Crisfield [129]). A pure incremental strategy is effective only for smooth non-linearities but
in case of severe non-linearities the error of approximation can accumulate after several
increments and the obtained solution can depart away from the true response (see e.g.
Crisfield [129]). Since the solution of the linearized incremental equation can be considered
as the first step in a general iterative Newton-Raphson method of solving equations (Korn
& Korn [253]), one can imagine that by preceding with the next steps of those iterations the
accuracy of incremental solution can be significantly improved. The (standard) Newton-
38
Zienkiewicz & Taylor [510] classified the FDM as one of the prospective forms of the Generalized FEM. On the
other hand, they also consider mesh-free methods as a generalization of the FDM.
39
Very interesting observations on historical roots of FEM appeared in a recent anniversary essay by Taylor [462],
according to whom one should look for the creators of FEM among pioneers of variational methods like W. Ritz,
B. Galerkin, I. G. Bubnov or R. Courant (see also Zienkiewicz & Taylor [510]).
40
It is quite symptomatic that FEM is placed as number one in the list of Top Ten Computational Methods of the
20th Century published by Dan Givoli in Expressions, the periodic of International Association of Computational
Mechanics (no. 11, September 2001, pages 5-9).
Raphson method requires creating a new tangent stiffness matrix in every equilibrium
iteration. During the pioneering period of non-linear FEA, at the turn of the 1960s and
1970s, a very limited power of computers persuaded the analysts to search for time saving
strategies. Very popular was a modified Newton-Raphson method that kept a constant
tangent stiffness matrix for all equilibrium iterations in one increment. Another option was
a flexible strategy where the iterations were turned on only at the specified level of
unbalanced forces. The beginnings of the non-linear FEA of shell structures were portrayed
in a state-of-art survey of Dupuis et al. [150]; see also Crisfield [129], Clough & Wilson
[128] and Zienkiewicz & Taylor [510]. In practical implementation, especially for
extremely non-linear problems, the modified Newton-Raphson method very often can be
found as the less efficient one due to a slow convergence or even a lack of the convergence;
Bathe [46]. It is worth to mention here also a group of quasi-Newton or secant iteration
methods that can be an alternative to the Newton-Raphson iterations. The tangent stiffness
matrix used in the Newton-Raphson iterations is replaced in quasi-Newton methods with a
secant stiffness matrix what can accelerate the convergence in particular applications. The
most popular quasi-Newton methods are the BFGS and Davidon method; Bathe [46],
Crisfield [129], Waszczyszyn [485] or Zienkiewicz & Taylor [510]. What is an additional
essential advantage of those methods, the stiffness matrix is created and inversed only once
for a single increment and during iterations only the inverse of the coefficient matrix is
continuously modified.
The incremental solution of non-linear equilibrium equations in the FEA needs an
appropriate steering parameter. The load control of incremental solution is probably the
most natural choice; however it allows tracing the equilibrium path only until the tangent
stiffness matrix remains positive defined. In a vicinity of any critical point (limit load or
bifurcation point) the tangent stiffness matrix becomes ill-conditioned, and it is singular at
the point, so it is impossible to obtain a convergent solution unless the control parameter is
changed, e.g. by enforcing a prescribed increments of a selected displacement component.
Generally, a displacement control is not a universal answer; it fails for complicated
equilibrium paths with a snap-back behavior and in the case of a general problem with
many dofs the choice of a suitable displacement component as a control parameter is not
trivial. One possible remedy in such a case is a special procedure, which every time, when
the convergence becomes slower, can automatic change the selected displacement
component for the one with the biggest relative increment in the previous increment as
suggested by Marcinowski [305]. Another option was presented by Chróścielewski &
Schmidt [122], who introduced a control window technique allowing for automatically
switching the incremental control between load and displacement parameters according to
the current route of the equilibrium path examined in the assumed control window. A
similar strategy was used already in 1968 by Leicester [281], who presented quite complex
equilibrium paths obtained for shallow shells applying Navier method for solving
governing differential equations in the incremental manner. A combined load-displacement
control was applied also in a geometrically non-linear FEA by Sabir & Lock [409] in 1972.
Practically, the most universal type of incremental control can be obtained by a proper
combination of the load and displacement parameters; one can imagine that by introducing
any load-displacement constraint a generalized arc-length parameter can be defined. The
arc-length control method, commonly known as the Riks-Wempner method, allows one for
tracing any complicated equilibrium path, including also bifurcation problems. There are
many existing variants of this method in the literature: see e.g. Crisfield [129], Ramm [379,
381] and Waszczyszyn [485].
Description of large deformations in the non-linear continuum mechanics is related to

the choice of the reference frame; generally the spatial (Eulerian) coordinates or material
(Lagrangian) coordinates can be used. It is commonly accepted that for solids the material
reference frame is recommended, Başar & Weichert [44]. In the incremental formulation of
a large deformation analysis of solids one can choose between the Total Lagrangian and
the Updated Lagrangian description; Bathe [46], Kleiber [244]. In the Total Lagrangian
(TL) description the governing equations are formulated with respect to the initial
configuration which stays unchanged during the deformation, whereas in the Updated
Lagrangian (UL) description the reference configuration is the one that has been obtained in
the previous increment. In the large displacement analysis of slender structures like beams
and shells, beside the TL or UL one can apply also the third type of a description, that is the
Co-rotational description (CR) based on the separation of the pure deformation and the
rigid rotation. In the CR description rotations of the body are referred to the immovable
base configuration, whereas all strains and stresses are calculated with respect to the co-
rotated configuration; an extensive report on co-rotational FE formulations was recently
presented by Felippa & Haugen [159], see also books by Belytschko et al. [52] and
Crisfield [129].
Finite element modeling of thin plates and shells

According to MacNeal [299] the first FEM plate analysis was published in 1961. The
majority of early FEM calculations for plates were performed with the classical Kirchhoff
theory, where the state of deformation is represented only with one variable, the transverse
deflection, w. In that pioneering period of FEM most implementations were based on the
displacement formulation. Every standard handbook of FEM listed the convergence
requirements that are necessary for assuring the monotonic convergence of the FEM
solution with the increasing density of the mesh (see e.g. Bathe [46] or Zienkiewicz &
Taylor [510]). Probably in their most condensed form, the criteria for monotonic
convergence of displacement based FEM formulation were expressed by Bathe [46]: “For
monotonic convergence, the elements must be complete and the elements and mesh must be
compatible”. In the same chapter Bathe wrote a commentary which is very crucial for FEA
of plates and shells: “the requirements of compatibility are difficult to satisfy in plate
bending analysis, and particularly in thin shell analysis if the rotations are derived from
the transverse displacements”. That remark explains why for a long time the development
of effective finite elements based on the Kirchhoff–Love constraints was a challenging task.
To ensure a continuous profile of the deformed plate, the shape functions should provide
not only the continuity of w but also continuity of derivatives of w, so called C1-continuity.
The C1-continuity requirements were satisfied e.g. by including derivatives of deflection, w
in the set of nodal parameters (see e.g. Bischoff et al. [63], Reddy [398], Wempner [489],
and Zienkiewicz & Taylor [510]). Such an approach introduced some difficulties related to
physical interpretations of those additional nodal degrees of freedom, mainly with respect
to description of boundary conditions and loading. Among the pioneers of the FEA was
Sir John H. Argyris who in 1966 published a paper [17] on Matrix displacement analysis41
of plates and shells using triangular and quadrilateral elements, where the transverse shear
deformations and non-linear effects were ignored. His quadrilateral elements satisfied the
41
At that time the term Finite Element Method was not yet established as the generally accepted label. However,
another famous pioneer of FEM, Ray W. Clough, introduced this term already in 1960 (see Clough & Wilson
[128]).
continuity of slopes only at nodal points, but the kinematic compatibility of triangular
elements were even poorer; nevertheless their performance was quite satisfactory. It was
few years later, when Bruce Irons introduced the patch test as a necessary condition for
convergence also for nonconforming elements42; compare Zienkiewicz & Taylor [510]. In
1972 Sabir & Lock [409] analyzed large displacements of thin cylindrical panels using
conforming cylindrical shell elements with 24 dofs; later those examples became probably
the most popular set of benchmark problems for a geometrically non-linear analysis of
shells. In 1968 Wempner et al. [490] presented an alternative formulation of thin shell
elements launching a new idea of Discrete Kirchhoff Theory (DKT) elements. They noticed
that shell elements based on Mindlin-Reissner kinematics could be “unduly stiff” when the
length-to-height ratio was big, that was quite natural for thin shell applications. To repair
that deficiency, at the midpoints of every elemental edge they imposed constraints
analogous to the Kirchhoff hypothesis, what resulted in more flexible element and much
faster convergence. Meek & Ristic [312] and Meek & Wang [313] considered a flat
triangular shell element constructed as a combination of the discrete Kirchhoff plate
element with Loof nodes (DKL) and the linear strain triangle (LST) membrane element; the
resulting element had 24 degrees of freedom. A slightly different formulation of a very
similar LST + DKL triangular shell element was presented by Poulsen & Damkilde [372].
Crisfield et al. [130] (see also Crisfield [129]) considered a very effective co-rotational flat
triangular shell element with 12 dofs.
Aforementioned difficulties related to the compatibility requirements of the
Kirchhoff–Love theory based plate/shell elements could also be avoided when the
displacement formulation on finite elements was replaced with the hybrid stress
formulation by T.H.H. Pian in the 1960s, see e.g. Pian [356], Morley [316, 317]. The
hybrid stress elements of Pian [356, 357] were based on the Hellinger-Reissner mixed
variational principle with the stress field assumed over the interior of the element and the
displacement field defined over element’s interface. Qing-Hua [375] proposed an efficient
finite element formulation for a geometrically non-linear analysis of thin shells by using
mixed variational approach.
More on constructing effective thin shell elements can be found e.g. in the classical
FEM handbook of Zienkiewicz & Taylor [510]; see also Reddy [398].
Plate and shell finite element models accounting for transverse shear
The above mentioned complications in the creation of plate and shell elements based
on Kirchhoff-Love theory created a strong impulse for searching for alternative trouble-free
formulations. One of the new ideas which appeared at that time was published in 1970 by
Ahmad et al. [2], who proposed to construct shell finite elements by a proper modification
of isoparametric solid finite elements. That modification included mainly an
implementation of the basic assumption of the Mindlin-Reissner kinematics, which
introduced a linear variation of displacements across the shell thickness. Some years later
Ramm [377] described that modification process as a “degeneration” of 3-D elements
(Fig. 2.5) into 2D elements (Fig. 2.6), and very soon the name of the “degenerated
isoparametric shell elements” was commonly accepted for those new elements43; Parisch
[345, 346, 347], Bathe & Bolourchi [47], Dvorkin & Bathe [152], Oliver & Oñate [332],
42
Bathe [46] suggested calling the path test “the completeness condition on an element assemblage”.
43
An alternative label “continuum based elements” also appears in the literature; see e.g. Betsch et al. [57],
Dvorkin & Bathe [152], Klinkel et al. [246, 247], Liao & Reddy [287, 288], and Parisch [349].
Huang & Hinton [209], Bathe & Dvorkin [48], Belytschko [51], Huang [208], Hsiao &
Chen [206], Cheung & Chen [111], Gilewski & Radwańska [174], Kreja & Cywiński
[262], Rammerstorfer et al. [389], Jiang & Chernuka [230], Ziyaeifar & Elwi [512]44, Yang
et al. [502], Lee & Lee [280], Zienkiewicz & Taylor [510], Kim et al. [237], Woo et al.
[498].
It is worthy to notice that alongside with the development of degenerated shell
elements an alternative concept of “solid-shell elements” also found many followers, who
especially appreciated a simple treatment of finite nodal rotations represented just by
relative displacements between the nodes at the top surface and the reference surface, see
e. g. Kanok-Nukulchai et al. [233], Hughes & Liu [212, 213], Kim & Lee [243], Laschet &
Jeusette [276], Parish [349], Park et al [351], Sansour [413], Hauptmann & Schweizerhof
[192], Klinkel et al. [246, 247], Masud et al. [309], Krätzig & Jun [256, 257], Sze et al.
[456], Sze & Zheng [457], Fontes Valente et al. [167], Kim et al. [237], Kulikov &
Plotnikova [273], Vu-Quoc & Tan [479].
Fig. 2.5: Solid-shell 3-D element (3 degrees of freedom at each node)
Fig. 2.6: Thick shell 2D element (5 degrees of freedom at each node)
Certainly, the Ahmad idea introduced a visible dissension into the community of shell
analysts. For a long time the “old school” did not want to accept the FE shell model that
44
Ziyaeifar & Elwi [512] presented degenerated shell elements with the parabolic and unsymmetrical distribution
of the transverse shear strain obtained by introducing new shape functions and extra degrees of freedom.
was built ignoring the heritage of shell theory. On the other hand, advocates of the new
concept pointed out the simplicity of the shell element formulation based on 3-D continuum
mechanics. Ramm [377] in 1976 and Kanok-Nukulchai et al. [233] in 1981 presented two
similar schemes, where the classical shell concept and the degeneration concept where
shown as two equivalent paths, both starting from a real shell-like 3-D structure and both
arriving at the FEM shell element. However, it was not until 1992 when Büchter & Ramm
[82, 84] presented a systematic comparison of shell theory and degeneration, and then the
shell traditionalists and the supporters of the degeneration started to tolerate the others’
point of view. A very detailed analysis of similarities and differences between those two
schools was recently presented by Bischoff et al. [63]; see also Malinen [302].
Today it is commonly accepted that Ahmad idea started an important period in the
chronicles of the FE shell analysis. For years the degenerated isoparametric shell elements
(and to some extend also finite elements based on the FOSD shell theory) became a
dominating strategy in the FEA of plates and shells45.
Locking - the disease and various medicines

Shortly after publication of Ahmad et al. [2] it was found that, similarly as in the case
of shell/plate elements based on FOSD, the convergence of degenerated shell elements was
very slow for thin plate/shell applications due to an over estimation of the element stiffness.
In computational mechanics such a phenomenon is known as the “locking”. Different types
of locking were named and described in the FEM literature (see e.g. Bathe [46] or
Zienkiewicz & Taylor [510]). Four of them are associated with FEA of plates and shells:
“shear locking”, “membrane locking”, “Poisson thickness locking” and “curvature
thickness locking”; Bischoff et al. [63]. Shear locking typically appears if very thin plates or
shells are analyzed with the low-order FEM formulation that accounts for shear
deformation; due to inadequate representation of deformation modes in the FE model, a
portion of energy which is associated with the transverse shear deformation is
overestimated for large values of the length-to-height ratio; Kreja & Cywiński [261].
Membrane locking that arises in FEA of curved panels is caused by the similar interaction
between bending and membrane energies; Stolarski & Belytschko [449]. The thickness
locking may appear in large deformation analysis accounting for a thickness stretching of
the shell, in those formulations which utilize the complete 3-D constitutive law. The
Poisson thickness locking occurs in the 6-parameter shell theory where a non-zero value of
the Poisson coefficient produces a linear through-the-thickness variation of the transverse
normal stress, what stays in a conflict with the unvarying through-the-thickness transverse
normal strain. The curvature thickness locking can be observed in some FEM shell models
based on the 7-parameter shell theory; a parasitic transverse strain can be developed in a
pure bending when low-order (straight) elements are used to model curved members
(Bischoff & Ramm [59], Bischoff et al. [63], Brank [74], El-Abbasi & Meguid [154]); an
analogous phenomenon known as trapezoidal locking appears also in some difference
vector formulations (Hauptmann & Schweizerhof [192], Krätzig & Jun [256, 257], Sze et
al. [456]). It was also observed (Heppler & Hansen [194], Ramm & Matzenmiller [383],
45
To support that statement we can quote here a very characteristic sentence from the review on plate and shell
finite elements presented by T. Belytschko at the 1986’ Winter Annual Meeting of ASME [51]: “Elements which
require C1 continuity because of a Kirchhoff-Love shell hypothesis seem to have faded from the scene, both as
topic of research and as a tool in both general purpose and special purpose finite element programs”. Another
characteristic passage comes from the 2000’ review paper by Yang et al. [502]: “Over the past two decades,
computational shell analysis has been, to a large extent, dominated by the so-called degenerated solid approach”.
Chinosi et al. [112] or Zienkiewicz & Taylor [510]) that higher order elements (e.g. 16-
node element with Lagrange interpolations) are much less sensitive to locking.
In the FEM literature of the 1970s and the 1980s one can find a huge number of
publications dealing with different aspects of the locking and a possible treatment of that
deficiency. In 1971 Pawsey & Clough [354] concurrently with Zienkiewicz et al. [511]
revealed an apparent paradox that behavior of degenerated elements can be significantly
improved by lowering the order of numerical integration used for evaluation of stiffness
matrices. The technique of a "reduced integration" became a very popular remedy for the
locking of degenerated elements as well as for FOSD plate/shell elements; see e.g. Hughes
& Liu [212, 213], Ramm & Stegmüller [384], Belytschko [51], Hughes & Hinton [211],
Laschet & Jeusette [276], Gilewski & Radwańska [174], Kreja & Cywiński [262],
Marcinowski [305], Zienkiewicz & Taylor [510], Bischoff et al. [63]. However, it was also
found that the reduced integration procedure lowered rank of the element stiffness matrices;
some spurious hourglass (zero-energy) mechanisms appeared which, depending on the
boundary conditions and an applied mesh of elements, could cause a singular solution.
Additionally to various variants of a selective reduced integration (Ramm [377], Parisch
[346], Bathe & Bolourchi [47], Arnold & Brezzi [20], Auricchio & Lovadina [24], Masud
& Panahandeh [308]), some stabilization methods (Belytschko et al. [53], Jacquotte &
Oden [225], Liu at al. [294]) were also considered in searching for possible improvements
of the reduced integration technique (see also Zienkiewicz & Taylor [510]).
A significant progress in the development of effective FOSD plate/shell elements is
related to the application of special “improved” interpolation schemes for selected strain
components. One of the first successful attempts in that field was reported by MacNeal
[298] in 1978 for element QUAD446 with a special treatment of transverse shear strains.
Quite a similar concept was presented three years later by Hughes & Tezduyar [214], who
assumed additionally that the transverse deflection should be interpolated with a
polynomial one order higher than that assumed for rotation. In 1983 Tessler & Hughes
[465] introduced the idea of continuous transverse shear edge constraints in a construction
of their “Mindlin-type four-node quadrilateral element”. A similar 4-node shell element
with independent interpolation of strains was proposed by Wempner et al. [491]. One of the
most recognized shell elements of that kind was developed at the MIT by the working team
of Klaus-Jürgen Bathe; in 1984 Dvorkin & Bathe [152] presented a four-node continuum
mechanics based (degenerated) shell element being a successor to earlier developments of
that team (Bathe & Bolourchi [47]) but modified by introduction of assumed interpolation
of transverse shear strains. The selected strain components at integration points were
re-interpolated on the base of their values at some special sampling points. In their next
publication, Bathe & Dvorkin [48] introduced the code name MITC4 for the 4-node shell
element based on Mixed Interpolation of Tensorial Components and MITC8 for its 8-node
counterpart. For the curved element MITC8 it was also necessary to provide a special
treatment of in-plane strains to avoid the membrane locking. Since then, the family of
MITCn plate/shell elements (with n standing for varying number of nodes) served as a
typical example of the Assumed Strain mixed variational approach application (see also
Bathe [45, 46], Bathe et al. [49], and Chapelle & Bathe [105]). In 1986 Park & Stanley
[353] presented the shear and membrane locking-free 9-node shell element (9-ANS)
formulated within the Assumed Natural Strain approach (see also Park et al. [352]). They
remarked that, if specialized to a rectangular four-node case, their element would result in a
46
Modified versions of that element are still available in the FEM system NASTRAN [318] (see also Hoff [196]
and Hoff et al. [197])
FE model equivalent to the earlier proposals of MacNeal [298], Hughes & Tezduyar [214]
or Dvorkin & Bathe [152]. Another concept of mixed formulation for Naghdi shell model
finite elements was given by Arnold & Brezzi [21] and simultaneously by Bramble & Tong
[72]. Various schemes of mixed formulation for triangular and quadrilateral plate elements
were examined by Arnold et al. [19]. Groenwold & Stander [180] developed a flat four
node shell element as a combination of a MITC4 plate element and a 4-node membrane
element with drilling degrees of freedom. A co-rotational implementation of the MITC type
4-node shell element was presented by Jiang & Chernuka [230]. One should notice that, in
spite of the widespread recognition of the plate/shell elements from the MITCn family, very
often in the literature the whole group of methods with special interpolations of selected
strain components is labeled after Park & Stanley [353] as the Assumed Natural Strain
(ANS) methods. Alternative formulations of 9-ANS and 8-ANS shell elements were
presented also by Chang et al. [103], Huang [208], Huang & Hinton [209], Stander et al.
[446] and by Stolarski [448]. Kim & Lee [243] examined four different ANS schemes for a
solid-shell element with 18 nodes.
The Stress Projection Method introduced by Belytschko et al. [54] can also be
classified as a kind of Assumed Strain formulation, similarly the Discrete Shear Gap
method of Bletzinger et al. [64]. It seems that also the linked interpolation approach47
proposed by Zienkiewicz & Taylor [510], can be associated with the group of the Assumed
Strain methods. Auricchio & Taylor [29] (see also Auricchio & Lovadina [24, 25], Bischoff
& Taylor [62]) introduced an improved interpolation of transverse displacement that was
kinematically linked to nodal rotations; resulting interpolation functions were one degree
higher for transverse deflection than for bending rotations48. A corresponding formulation
of plate elements but named a “field consistence approach” was presented by Luo &
Eriksson [297]. A field-consistent shell element was developed also by Somashekar et al.
[445]. Auricchio & Sacco [26, 27] used the linked interpolation approach to construct finite
elements for the analysis of laminated composite plates. Another anti-locking technique of
that kind was recently described by Wanji & Cheung [484] who applied the Timoshenko
beam function to define the rotation and deflection on the element boundary; their approach
closely resembles the idea used almost 20 years earlier by Hughes & Tezduyar [214] and
Tessler & Hughes [465]. In 1986 Simo & Hughes [438] revealed the variational
foundations of the Assumed Strain methods linking them to Hu-Washizu variational
principle. The main idea of the mixed variational formulation lies in the relaxation of some
constraints (e.g. those of vanishing shear strain for thin shells) which are taken into account
by means of Lagrange multipliers representing additional unknowns, other than
displacements. The equivalence between some mixed models and displacement models
with reduced integration was analyzed by Malkus & Hughes [303], Noor & Andersen [327]
and Shimodaira [430].
Aside from ANS, another popular technique of avoiding locking problems in FEA is
the Enhanced Assumed Strain (EAS) method; Simo & Rifai [439], Andelfinger & Ramm
[16]. While the ANS method lowered the polynomial order of interpolation for some
47
It is also known as the “anisotropic interpolation”, Hughes & Tezduyar [214], or "interdependent variable
interpolation", Tessler & Hughes [465]. According to Zienkiewicz & Taylor [510], the linked interpolation was
used for the first time for beams by Fraeijs de Veubeke [169].
48
In his recent book Reddy [398] used analogous interpolation for his “consistent finite element model of the
Timoshenko beam theory“, showing next its equivalence with the under-integrated isoparametric beam element.
Similar considerations for shear locking of beam elements were presented also by Kreja & Cywiński [261] in
1988. Ten years later Luo [296] presented more general examination accounting also for membrane locking
analysis within his newly introduced field consistence approach.
selected strain components, in the EAS method the conventional strain fields resulting from
differentiation of displacements are augmented with additional independent strain field
which is incompatible. When those additional strain fields are orthogonal with the
corresponding stress fields, they do not contribute to the element energy. Simo & Rifai
[439] started their derivation of the EAS method from the three field variational principle of
Hu–Washizu; however, Bischoff & Ramm [59] declared that “the EAS formulation is not a
mixed method; but it is rather a displacement model with a limited softening effect”. One
can trace back the origin of the EAS method to the first experiments with incompatible
displacement fields from the early 1970s (see e.g. Clough & Wilson [128], Zienkiewicz &
Taylor [510]). Andelfinger & Ramm [15, 16], Yeo & Lee [503], Bischoff et al. [61]
discussed the equivalence of the EAS method with the assumed stress hybrid elements
based on the Hellinger-Reissner principle; Pian [356]. On the other hand, few years earlier
the similarity between the incompatible displacement model and the assumed stress hybrid
model was indicated by Pian & Tong [358]. Locking-free degenerated shell elements based
on the EAS formulation were presented e.g. by César de Sá et al. [100] with a further
extension to the large rotation analysis in Fontes Valente et al. [167]. Recently,
Chróścielewski & Witkowski [123] showed that the EAS method can also be used for the
shell model with unsymmetrical strain measures. Campello et al. [86] presented an
interesting proposition of a 6-node triangular shell element with a compatible quadratic
interpolation scheme for the displacements (based on discrete parameters at six nodes) and
a non-conforming linear interpolation of the rotations (with discrete parameters only at 3
mid-side nodes)49. Although, the authors themselves classified their element as a “pure
displacement-based” that needed “no numerical tricks such as ANS, EAS or reduced
integration with hourglass control (...) to improve its performance”, one can also recognize
there some aspects of the EAS idea.
Fundamentals of the hybrid formulation of plate elements can be found e.g. in the
handbook of Zienkiewicz & Taylor [510]. Horrigmoe & Bergan [204] presented flat
triangular and quadrilateral shell elements based on a hybrid stress model and co-rotational
formulation. Cheung & Chen [111] examined hybrid degenerated elements for linear
analysis of plates and shells. Hybrid shell elements for large rotation analysis were
presented e.g. by Saleeb et al. [412], Sansour [413], Sansour & Boćko [415, 416], Sansour
& Bufler [417] and Sansour & Kollmann [418]. Some large rotation results obtained with
hybrid shell elements were also presented by Duan [149]; however, without any details on
the strategy adopted for the finite rotation treatment.
Quite often, the EAS method is combined with the ANS approach in the same element
formulation, for example, the ANS technique is employed to avoid the transverse shear
locking and the curvature thickness locking, whereas the EAS scheme is used to circumvent
the membrane locking and the Poisson thickness locking, see e.g. Andelfinger & Ramm
[16], Betsch et al. [57], Bischoff & Ramm [59], Brank et al. [78], Braun et al. [81],
Hauptmann & Schweizerhof [192], Klinkel et al. [247], Krätzig & Jun [256, 257], Vu-Quoc
& Tan [479], or Wagner & Gruttmann [481].
Equivalent Single Layer FE models of laminated composites and sandwich panels

It seems that the most straightforward construction of a 2D FE model for a laminated
composite or a sandwich panel can be obtained within the ESL approach. One can simply
49
This element was extended later by Pimenta et al. [367] for the inclusion of the thickness changes according to
the 7-parameter theory.
employ one of existing finite elements prepared for homogeneous plates or shells and all
necessary modifications in that case consist in the introduction of an anisotropic material
model with material parameters estimated according to selected lamination theory. Rao
[391] performed a linear FE analysis of shallow laminated shells using 48 dof finite
elements based on the CLT (Classical Lamination Theory); a similar element was used in
geometrically non-linear analysis by Saigal et al. [411].
For the reasons described earlier in this chapter, it seems quite obvious that shear
deformation theories are more suitable as the basis for a construction of finite elements to
model laminated composites. There is a big number of finite elements for laminated shells
that are formulated within the FOSD (First Order Shear Deformation) theory. A basic FE
formulation of the classical linear FOSD theory of layered shells can be found in the book
of Reddy [398]. Large displacement FOSD FE analysis of layered plates (in the range of
von Kármán non-linearity) was presented in a review paper of Reddy [392]; a
corresponding FE model for laminated composite shells was examined by Reddy &
Chandrashekhara [400]. Palmerio et al. [343] described a 9-node shell element for moderate
rotation FOSD analysis of laminated shells. Different aspects of the FE implementation of
the FOSD moderate rotation shell theory were examined by Kreja et al. [269] (see also
Kreja & Schmidt [263, 264, and 265]). Large rotation FEA of laminated shells within the
FOSD theory was considered by Kreja & Schmidt [266, 267, and 268]; see also Kreja
[260]. In 1979 Panda & Natarajan [344] constructed their FE FOSD laminated plate model
as a displacement based degenerated isoparametric element with quadratic in-plane
interpolation and reduced integration; the corresponding FE formulation for laminated
shells was presented by Chang & Sawamiphakdi [104]. A very similar element was used by
Jun & Hong [232] (see also Kweon & Hong [274]) in a non-linear UL analysis of
cylindrical composite panels performed with the arc-length control method. Wagner [480]
analyzed large deformations and buckling of cylindrical composite laminated shells using a
4-node finite element with reduced integration and hour-glass stabilization. Ferreira &
Barbosa [162] presented a 9-node element based on the Marguerre shallow shell theory (in
the range of von Kármán non-linearity) and ANS approach. Bödefeld et al. [70] analyzed
large deformations of axi-symmetric laminated shells using multi-layered version of the
degenerated 2-D shell element introduced for large rotation analysis by Kreja & Cywiński
[262]. Laschet & Jeusette [276] performed post-buckling analysis of laminated composites
applying an under-integrated solid-shell element possessing only translational degrees of
freedom. Rikards et al. [406] analyzed buckling and vibration of composite stiffened shells
using triangular FOSD shell elements with selective integration. The concept of FOSD
finite elements based on mixed interpolation of tensor components (MITC elements) was
extended for laminated plates by Alfano et al. [4], and for laminated shells by Haas & Lee
[184] and Hossain et al. [205]. Somashekar et al. [445] examined a 4-node field-consistent
shell element for a linear analysis of laminated composite panels. Groenwold & Stander
[181] developed a 4-node shells element for layered composites starting from their own flat
24 dof element for isotropic shells (Groenwold & Stander [180]). Dorninger [147] (see also
Dorninger & Rammerstorfer [148]) extended the non-linear formulation of the degenerated
shell element of Ramm [377, 378] to include the anisotropic layered material behavior of
laminated composites. Brank et al. [79] presented an ANS formulation of a 4-node FOSD
shell element for large-rotation analysis of laminated elastic shells; however, only one out
of eight numerical examples was devoted to multilayered shell problem, and the magnitude
of rotations in that particular example stayed within the range of moderate rotations
(compare Kreja [260]). Recently, Han et al. [189] performed a large deformation analysis
of laminated shells using an element-based 9-node stress-resultant ANS shell element with
54 dofs. Kim [239] (see also Kim & Voyiadjis [242]) developed under-integrated 8-node
non-linear composite FOSD shell element based on corotational formulation. A modified
version of that element based on the ANS formulation was examined by Kim & Park [241]
(see also Kim et al. [240]). A co-rotational formulation was applied also by Barut et al.
[33], who analyzed large displacements of shallow laminated shells applying triangular
finite elements. Quite recently, Pai [339] developed a 4-node laminated shell element
using the co-rotational formulation of Pai & Palazotto [340] and the energy-consistent
FOSD theory of Pai [338]; with 14 dofs per node including the derivatives of deflections
that element was declared by the author to be locking-free. Hashagen et al. [190] adopted
the solid-like shell element introduced by Parisch [349] for homogeneous structures to
perform materially and geometrically non-linear analysis of fiber reinforced metal
laminates. Kulikov & Plotnikova [272, 273] presented an extended mixed field formulation
of a multilayered shell element with fundamental unknowns consisted of six displacement
parameters, eleven strains and eleven stress resultants; however, the non-displacement
unknowns were eliminated on the element level resulting in the FE model with
displacement dofs only. Cen et al. [99] proposed a 4-node laminated FOSD plate element
based on utilization of Timoshenko beam theory (a mixed field formulation corresponding
to the ANS) combined with a hybrid stress approach for improving the accuracy of stress
recovery; a very similar approach was also used by Zhang & Kim [509] in a construction of
their 20 dof and 24 dof quadrilateral laminated plate elements. An eighteen-node hybrid-
stress solid-shell element for laminated structures was presented by Sze et al. [456] and by
Sze & Zheng [457].
Looking for a possible improvement of the FOSD results Tanov & Tabiei [460]
proposed a simply correction to a standard FE FOSD shell model by enforcing a parabolic
shear strain distribution across the shell thickness. Fares & Youssif [157], Fares et al. [158],
and Auricchio & Sacco [28, 26, and 27] presented a collection of different finite shell
elements based on the refined FOSD theory and mixed variational principle. In 1985 Phan
& Reddy [355] constructed a 4-node finite element based on the Reddy TOSD theory of
laminated plates [393] assuming Hermite interpolation of the transverse deflection and
Lagrange interpolation of the other displacement unknowns. An extension of that FE model
for inclusion of the von Kármán non-linearity was presented by Reddy [394]. Finite
elements constructed according to various TOSD plate theories were examined also by
Bose & Reddy [68]. High order interpolation shell elements based of FOSD and TOSD
small displacement theories were described recently by Reddy & Arciniega [399]. Dennis
& Palazotto [138, 139] (see also Chaplin & Palazotto [106], Tsai et al. [470] and Naboulsi
& Palazotto [319]50) developed finite shell elements based on their own Simplified Large
Rotation (SLR) TOSD theory of cylindrical laminated shells; a combination of Hermite and
Lagrange interpolation schemes was applied, similarly as used earlier by Phan & Reddy
[355], however, a quadratic shape functions were used for “in-plane” displacement
components, u and v. Das et al. [135] presented a rather complex formulation of a triangular
finite element based on HOSD model with seven weighted-average displacement variables;
a special procedure based on the hybrid energy functional was applied to satisfy the C1
50
Naboulsi & Palazotto [319] examined additionally two other FE models of cylindrical composite shells: one
described as FOSD model accounting for large rotations treated with Euler angles, and the other being a discrete
layer formulation following the co-rotational concept of Pai & Palazotto [340]. However, description of the finite
elements applied by Naboulsi & Palazotto [319] for those additional models was by far insufficient; e.g. there is no
a single word on a locking sensitivity of those elements in the text.
inter-element continuity requirements what resulted in the FE with 13 dofs per node. Moita
et al. [315] used 80 dof finite shell elements based on the HOSD theory in the buckling
analysis of laminated panels. Başar et al. [37] developed 4-node ANS shell elements based
on a large-rotation TOSD theory of laminated shells; variants with 7 and 5 dofs per node
were considered. Balah & Al-Ghamedy [31] presented a similar 4-node ANS shell element
for the TOSD formulation with seven degrees of freedom but they applied exponential
mapping of finite rotations instead of Euler angles used by Başar et al. [37].
A separate group among the FE implementations of the ESL models consists of finite
elements constructed according to the zig-zag deformation theory with interlaminar stress
continuity (compare a review article of Carrera [93]). Various FE realizations of the
theoretical zig-zag model of Toledano & Murakami [466, 467] were presented by Carrera
and co-workers; their FE formulations with seven displacement unknowns at each node
were characterized by the C0 type continuity. Carrera [88] described 4-, 8- and 9-node plate
elements following his own theoretical model, Carrera [94]; selectively and uniformly
reduced integration schemes were considered. A corresponding multilayer 4-node shell
element was presented by Carrera & Parisch [97], who started from the existing finite-
rotation assumed strain shell element proposed for homogeneous shells by Parisch [348].
Another FE implementation of the zig-zag model of Carrera [94] was prepared by Brank &
Carrera [76, 75]; their formulation based on the ANS shell element by Brank et al. [79]. As
it was mentioned earlier the zig-zag model proposed by Di Sciuva [141] required a C1 type
continuity in the FEM implementation, therefore the triangular fully conforming
multilayered plate element presented by Di Sciuva [142] had 10 dofs at each node, with
first and second derivatives of the transverse deflection in the list of displacement
unknowns.
Discrete Layer FE models of laminated composites

Mawenya & Davies [311] presented a linear bending analysis of laminated plates
employing a DL finite element with quadratic interpolation and 3+2N dofs at each node
(three global translations for the whole laminate with two local rotations for each layer). A
similar FE formulation of a DL model for laminated plates by accounting for the von
Kármán non-linearity was developed by Reddy [394]. An analogous DL shell element was
constructed within the degenerated formulation by Rammerstorfer et al. [389]. Chaudhuri
& Seide [108] described a corresponding triangular multilayered shell element based on
their “layerwise constant shear-angle theory” with quadratic in-plane interpolation; more
recently, Chaudhuri [107] applied that element in the analysis of angle-ply composite
plates. The DL shell elements of Pinsky & Kim [368] accounted for large deformations
including the thickness stretching with the number of nodal unknowns extended to 3+4N.
Similar discrete layer FE models of composite shells based on the 7-parameter FOSD large
rotation shell theory were developed by Braun et al. [81]. Başar & Ding [36] and Başar et
al. [40] examined 4-node ANS/EAS shell elements for various DL models accounting for
the thickness stretching using 3+3N, 3+4N and 3+6N dofs per node. A DL multilayered
shell model with 3+9N unknowns based on the HOSD theory was implemented into FEM
by Gruttmann & Wagner [182].
Vu-Quoc & Tan [479] presented a DL formulation based on a solid-shell element
without any rotational degrees of freedom what made it especially well suited for modeling
multilayer shells with geometrical thickness discontinuities like ply drop-offs or
(piezoelectric) patches. An analogous finite element formulation of DL model of laminated
panels was considered by Dakshina Moorthy & Reddy [133]; however, their analysis was
performed for a simplified 2D geometry of a vertical cross-section. They applied 6-node

plane element with quadratic interpolation for the in-plane approximation and a linear
function for the thickness approximation; the EAS formulation was employed to prevent
locking. A similar concept was applied by Krätzig & Jun [255, 256, 257], who considered a
quite general formulation of the discrete-layer models with two different layer-wise
refinement concepts, internal for improved modeling of complicated stress states and
external for better kinematic approximation properties. In the two examples presented by
Krätzig & Jun [256], an automatic 3-D refinement procedure based on the error estimation
was applied; the obtained final h-refined FE meshes directly corresponded to some extend
with the global-local solution concept. Desai et al. [140] applied 3-D finite elements in a
layer-wise discretization of layered composites (i.e. using one element per each layer across
the thickness). The set of nodal parameters in their mixed formulation based 3-D finite
elements contained displacements and transverse stress components therefore the through-
the-thickness continuity requirements of displacements and transverse stress fields were
satisfied automatically. A similar hybrid approach was used by Feng & Hoa [160] in their
multilayered 3-D model build as a stack of 8-node solid sub-elements along the thickness of
composite laminates. Two dimensional shell elements were combined with 3-D solid
elements in materially and geometrically non-linear analysis of composites by
Rammerstorfer et al. [390]. A slight different variant of a global-local FEA can be obtained
by a combination of DL and ESL composite shell elements as it was done by Kong &
Cheung [252] for linear analysis and by Gruttmann & Wagner [182] in non-linear
applications. Yu et al. [504] presented a detailed FEA of composites plates using a discrete
layer approach where a single layer is modeled with the mixed-field eight-node plate
element with the total number of 104 (stress-displacement) unknowns. Within a global-
local concept proposed by Yu et al. [504] some physical layers can be modeled as a single
global (ESL) region; nevertheless, the numerical size of the problem remained enormous.
Crisfield et al. [130] introduced special interface elements to analyze delamination problem
in composites.
Aitharaju & Averill [3] (see also Cho & Averill [116]) proposed an interesting tactic to
circumvent problems of the C1 continuity requirements related to the application of the zig-
zag model of Di Sciuva [141] in the FEM. Their shell element had a form of an 8-noded
cube with 5 dofs (three translations and two rotations) at each node; therefore they could
increase the refinement of the model by using more than one element in the thickness
direction.
Special FE models for sandwich structures

Generally, the most of FE models of multilayered plates and shells reviewed above in
the context of composite panels can be directly applied also in the analysis of sandwich
structures; however, there is also a group of finite elements exclusively dedicated to
sandwich structures. In 1970 Schmit Jr. & Monforton [428] presented one of the first non-
linear finite element models of sandwich plates, accounting only for membrane and bending
deflections of the outer faces and transverse shear deformations of the core. Marcinowski
[306] constructed his 8-node shell element for a geometrically non-linear analysis of
sandwich shells assuming after Reissner [402] that outer facings worked as thin membranes
without bending stiffness. Das et al. [135] used a hybrid-stress formulation to develop a
new 3-node triangular HOSD sandwich finite element with 39 degrees of freedom.
Vu-Quoc et al. [478] analyzed sandwich shells applying a geometrically exact 4-node shell
element with selective reduced integration based on a general DL multilayered shell
formulation. An interesting way of putting the idea of ESL modeling of sandwich panels
into practice was presented by Tanov & Tabiei [459] who suggested performing a FEA of
any sandwich shell with an existing FE FOSD model of homogeneous shells, simply
entering equivalent material parameters provided by their sandwich homogenization
procedure.
Analysis of composites and sandwich panels with commercial FEA codes

The growing contribution of laminated composite applications in engineering
structures initiated an increasing interest in appropriate analysis tools; as a result some
professionals reached for ready computational models offered by the commercial Finite
Element Analysis systems. Ali [5] applied the MSC Nastran [318] to perform a linear
analysis of a petrol engine oil sump pan made of fiber-glass composite. Rolfes & Rohwer
[408] analyzed laminated composite plates using the MSC Nastran with their self-written
preprocessor and postprocessor to implement the ”improved” transverse shear stiffness for
the FOSD model together with the special procedure to evaluate the transverse shear
stresses. Kreja [259] employed MSC Nastran in a stability analysis of cylindrical composite
shells. Sze et al. [455] performed geometrically non-linear analysis of selected benchmark
problems of laminated shells using SR4 element of the FEA system ABAQUS. Hu [207]
applied ABAQUS with user defined composite material model where non-linearity of
strain-stress relation was associated with in-plane shear. Eason & Ochoa [153] presented
the modeling progressive damage in composites with the ABAQUS. Manet [304]
investigated an application of ANSYS 5.2 in the analysis of sandwich structures behavior
examining various finite elements available in that FEA system. Quite recently, Borsellino
et al. [66] applied ANSYS 5.6 to perform a 2-D (plane) computer simulation of static
mechanical tests for sandwich panels.
More reading on different FE shell models

There is a huge literature on the finite element analysis of plate and shell structures,
therefore it was simply impossible to cover in this limited survey a full variety of diverse
ideas presented in the professional literature on this subject. A detailed discussion of
different aspects of the FE implementation of various shell models can be found in the
fundamental handbooks of Bathe [46], Belytschko et al. [52], Chapelle & Bathe [105],
Chróścielewski et al. [119], Crisfield [129], and Zienkiewicz & Taylor [510]. Interesting
collections of papers dedicated to this subject appeared in the volumes edited by Hughes &
Hinton [211], Ibrahimbegović & Krätzig [223], Krätzig & Oňate [258], Noor et al. [328],
Pietraszkiewicz & Szymczak [366], Ramm [380], Ramm & Wall [387], and by
Rammerstorfer [388]. Noteworthy review articles on FEM shell elements were published
by Belytschko [51], Bischoff et al. [63], Gilewski & Radwańska [174], Ibrahimbegović
[220], Wempner [489], and Yang et al. [502]. Problems of FEM modeling of multilayered
shells were studied e.g. by Ferreira & Fernandes [163], Noor & Burton [329], Qatu [373],
Toorani & Lakis [469]; FEA of sandwich structures was surveyed e.g. by Librescu &
Hause [291], Burton & Noor [85].
2.5. Modeling considerations resulted from the presented review

The main conclusions one can draw from the presented survey on theoretical and
numerical modeling of plate and shell structures with a special emphasis given to the
laminated composite and sandwich structures are as follows:
a) A single numerical model capable for a universal representation of all layered
composite and sandwich panels does not exist; depending on the particular problem
different formulations can be the most effective choice. Frequently, the best results can
be obtained with a combined global-local analysis where various numerical models are
used for separate parts of the structure.
b) A detailed through-the-thickness representation of deformation profiles and/or
distribution of stresses usually accompany small or moderate displacement
formulations, on the other hand, most of large rotation analyses are performed for a
simplified FOSD type models.
c) In contrast to the state of the art in the field of isotropic shells, there is lack of a
commonly accepted set of benchmark problems for laminated shells undergoing large
rotations; very often an advanced formulation accounting for finite rotations in
laminated shells is illustrated with numerical examples where rotations stay well within
the range of small rotations.
Chapter 3
INCREMENTAL FORMULATION OF NONLINEAR SHELL

ANALYSIS
We start this chapter introducing an idea of an incremental description of a body

motion, and then the basic nomenclature regarding the shell geometry is established,
followed by an analysis of large deformations of a layered shell within the First Order
Shear Deformation (FOSD) theory. Next the strain-displacement relations are established
for the large rotation shell theory including a detailed study of the implementation of the
director inextensibility together with a proper treatment of finite rotations. Afterward the
obtained formulae are confronted with relations proposed by other authors. Since the main
goal of the presented considerations is preparing a finite element formulation of the
problem, we will omit the derivation of the equilibrium differential equations and directly
proceed to the corresponding variational expressions in terms of the Principle of Virtual
Displacements and incremental Total Lagrangian (TL) formulation. Concurrently stress
resultants are introduced, that are energetically conjugate to the applied strain measures.
Appropriate constitutive relations are constructed assuming the Equivalent Single Layer
(ESL) model with global constitutive relations established with enhanced Lamination
Theory adequate for the FOSD theory.
3.1. Incremental formulation

The motion of the 3-D shell-like body in space is considered assuming the existence of
static effects only. In the incremental formulation the following three configurations of the
body (Fig. 3.1) are considered, Bathe [46]:
0
- the initial configuration C, at time 0;
1
- the actual configuration C, at time t;
2
- the searched configuration C, at time t+∆t.
1
C
1
0
d
C 0
n
1
V DV
2
2 2
C
V d
0 0
r
x3 R 1
R
2
R
x2
x1
Fig. 3.1: Shell body motion in space

3.1. Incremental formulation 53
The respective configurations are characterized by left superscripts 0, 1, and 2. Thus,

m m
the position vector of an arbitrary point P of the configuration C is denoted by R;
0 1 2
whereas the time instants t, t, and t are used for the time variable equal 0, t, and t+∆t,
respectively.
3.2. Shell geometry

A comprehensive discussion of geometry of a three dimensional shell-like body in
undeformed configuration can be found in classic texts of Green & Zerna [178], Naghdi
[320, 321] Pietraszkiewicz [359, 360, 362] or Başar & Krätzig [43]; therefore only
fundamental issues of the topic are described in the following, merely in the range required
as a basis for the further derivations.
It is assumed that the middle surface of a shell, Ω, stays smooth in every configuration;
m m m
hence, the position vector, r, of an arbitrary point P of the middle surface Ω in the
m α
configuration C can be represented as a function of general convected coordinates θ
(α = 1, 2) and the time variable t:
m
r = r (θ α , t = mt ) . (3.1)
0
Correspondingly, in the initial configuration (m = 0), C, the position vector of an arbitrary
0 0
point P of the middle surface Ω is described as
0
r = r (θ α , t = 0) . (3.2)
0
At every point of the middle surface Ω (see Fig. 3.2) one can introduce a covariant base
0 0 0
vector triad ( a1, a2, n)51, Green & Zerna [178], Başar & Krätzig [43]:
∂0r 0 0
a1 × 0 a 2
0
aα ≡ ≡ r ,α , α = 1, 2 and 0
n = 0a3 = 0a3 = . (3.3)
∂θ α 0
a1 × 0 a 2
1 2 3 α
The coordinate system (θ , θ , θ ) is defined in such a way that θ (α = 1, 2) denote
3
convected curvilinear surface coordinates of the shell mid-surface Ω, and θ is the thickness
coordinate taking values from the interval (-h/2, h/2) with h standing for the initial shell
3
thickness. In the undeformed configuration the coordinate θ is measured in the direction
0 m m
that is perpendicular to Ω. Consequently, the position vector R of an arbitrary point P*
m
in the shell space in the configuration C can be represented as a function of general
i
convected coordinates θ (i = 1, 2, 3):
m
R = R (θ 1 ,θ 2 ,θ 3 , m t ) . (3.4)
0 0
In the initial configuration C the position vector of an arbitrary point P* in the shell space
can be described as
0
R (θ 1 , θ 2 , θ 3 ) = 0 r (θ 1 , θ 2 ) + θ 3 0 n . (3.5)
51
Symbol × in (3.3) represents the vector product.
54 Chapter 3. Incremental formulation of nonlinear shell analysis
3
q
1
0 0
n= g 3 0
q
g1
0
a1
0
g2 0
P*
0
a2 0
P
2
q
0
R 0
r
x3
x2
x1
Fig. 3.2: Shell geometry in the initial configuration
The covariant base vectors in the space of the shell can be calculated as derivatives of
m
the position vector R; Green & Zerna [178], Başar & Krätzig [43], and Başar & Weichert
[44]:
∂mR m (3.6)
m
gi ≡ ≡ R ,i , i = 1, 2, 3.
∂θ i
Consequently, using (3.5) and (3.6) one can get
0
gα = 0 a α + θ 3 0 n,α , α = 1, 2,
(3.7)
0
g3 = 0 n = 0 a3 = 0a3 , ( )
3 0 0
what indicates that for θ = 0 it is gk = ak.
Assuming that the symbol “•” denotes the scalar product of two vectors, one can write
obvious properties of the base vectors:
0
aα ⋅ 0n = 0 , 0
n ⋅ 0 n = 1 and 0
n ⋅ 0 n,α = 0 . (3.8)
0 α 0
The contravariant base vectors , a , in the middle surface Ω can be constructed in such a
way that
0
a α ⋅ 0 a β = δ αβ , (3.9)
where δαβ stays for the Kronecker delta:

1 for i = j . (3.10)
δi j = 
0 for i ≠ j
3.3. Shell deformation 55
The covariant and contravariant components of the surface metric tensor in the middle
0
surface Ω can be determined as
0
aαβ = 0aβα = 0 a α ⋅ 0 a β and 0 aαβ = 0a βα = 0 a α ⋅ 0 a β . (3.11)
These components are interrelated according to the following equation:

0
aαβ 0a βγ = δ αγ . (3.12)
They can be used to rise or lower the indices of the base vectors:
0
a α = 0aαβ 0
a β or 0
a α = 0aαβ 0 a β . (3.13)
The first fundamental form of the middle surface (Başar & Krätzig [43], Başar &
Weichert [44], and Green & Zerna [178]) can be written in the following form:
( ds )
0 2
= 0dr ⋅ 0dr = 0 a α ⋅ 0 a β dθ α dθ β = 0aαβ dθ α dθ β . (3.14)
i
Assuming a usual notation of a vector w = w gi and a second order tensor
ij
T = T gi ⊗ gj (see e.g. Başar & Weichert [44])52 one can write the following rules of
differentiation:
∂w ∂ ∂w i ∂g i , (3.15)
w, k ≡ = ( w i
g i ) = g i + wi
∂θ k
∂θ k
∂θ k
∂θ k
∂T ∂ ∂T ij ∂g ∂g
T, k ≡ = (T ij
g ⊗ g ) = g i ⊗ g j + T ij ki ⊗ g j + T ij g i ⊗ kj . (3.16)
∂θ ∂θ ∂θ ∂θ ∂θ
k k i j k
m m
Introducing Christofel symbols of the second kind Г ij= Г ji as components of the base
vector gi derivative:
∂gi (3.17)
≡ gi ,k = Γikmg m
∂θ k
together with a short description of component derivatives
∂wi ∂T ij (3.18)
≡ wi , k , ≡ T ij , k ,
∂θ k ∂θ k
one can obtain the following formulae:
w,k = (wi ,k + wm Γkm
i
) gi = wi gi , (3.19)
k
T,k = (T ij ,k + Γkni T nj + Γkmj T mj ) gi ⊗ g j = T ij gi ⊗ g j , (3.20)

k
where the vertical line stands for the covariant derivatives of the vector and tensor
components.
The second fundamental form of the middle surface, Green & Zerna [178], Başar &
Krätzig [43], Başar & Weichert [44], can be presented as follows
52
Symbol ⊗ denotes the tensor product.
0
dr ⋅ 0 a 3 = − 0bαβ dθ α dθ β = − 0bαβ dθα dθ β = − 0bβα dθα dθ β (3.21)
with the components of the surface metric tensor of the second order defined as
0
bαβ = 0bβα = − 0 a α ⋅ 0 n, β = − 0 a β ⋅ 0 n,α . (3.22)
By differentiation of (3.8) on can obtain also another useful relation

0
bαβ = 0 a α , β ⋅ 0 n = 0 a β ,α ⋅ 0 n . (3.23)
The derivatives of the base vectors are given by the Gauss-Weingarten formulae,
Green & Zerna [178], Başar & Krätzig [43], and Başar & Weichert [44]:
δ 0 δ 0
0
a α , β = 0 Γαβ
m 0
a m = 0 Γαβ a δ + 0 Γαβ
3 0
a 3 = 0 Γαβ a δ + 0bαβ 0 n , (3.24)
0
a α , β = − 0 Γβαm 0 a m = − 0 Γβδ
α 0 δ 0 α 0 3 α 0 δ 0 α0
a − Γβ 3 a = − 0 Γβδ a + bβ n , (3.25)
0
n, β = 0 a 3 , β = Γ3mβ 0 a m = Γ3δβ 0 a δ + Γ33β 0 a 3 = − 0bβδ 0 a δ + 0 . (3.26)
Using (3.26) one can derive from the first equation of (3.7) the following relation
between the base vectors in the space of the shell 0gα and the base vectors of the middle
surface 0aδ (compare Başar & Krätzig [43]):
0
(
g α = 0 a α + θ 3 0 n,α = δ αδ − 0bαδ θ 3 ) 0
a δ = 0µαδ 0 a δ , (3.27)
with 0 µαβ standing for the components53 of the shifter tensor54 0 in the initial configuration
0
C:
0
= 0 g m ⊗ 0 a m = 0µεδ 0 a δ ⊗ 0 a ε + 0 a 3 ⊗ 0 a 3 ,
0
T = 0 a m ⊗ 0 g m = 0µεδ 0 a ε ⊗ 0 a δ + 0 a 3 ⊗0 a 3 , (3.28)
0
µαδ = δ αδ − 0bαδ θ 3 , µα3 = 0µ3δ = 0,
0 0
µ 33 = 1;
in the company of the inverse of the shifter tensor 0 assumed as:
0
= 0 a k ⊗ 0 g k = 0 ( µ −1 )δε 0 a δ ⊗ 0 a ε + 0 a 3 ⊗ 0 a 3 ,
0
T = 0 g k ⊗ 0 a k = 0 ( µ −1 )δε 0 a ε ⊗ 0 a δ + 0 a 3 ⊗ 0 a 3 , (3.29)
0
[
( µ −1 )αδ = ( 0µ ) −1 δ αδ + ( 0bαδ − 0bββ 0δ αδ ) θ 3 , ] 0
( µ −1 )α3 = 0 ( µ −1 )δ3 = 0, 0
( µ −1 ) 33 = 1;
where 0µ is the determinant of the shifter tensor55 in the initial configuration 0C:
0
µ = det( 0 = 0 µ βα . (3.30)
Using (3.28) and (3.29) one can also write56
53
Detailed investigations regarding properties of the shifter components 0µαβ and 0(µ−1)αβ were presented by
Naghdi [320]. The symbols 0µαβ were called “translators” by Pietraszkiewicz [360], who indicated that they can
also be considered as components of the metric tensor when expressed in the mixed basis: 0G = 0µik 0ai⊗0g k.
54
It is worth to notice, that also another definitions of the shifter tensor are possible, see e.g. Bischoff et al. [63],
where the shifter tensor was introduced as Z = 0gi⊗0ai; by comparing this expression with (3.29) one can easily
find that Z is equivalent to 0-T.
55
Green and Zerna [178] called this quantity “a surface invariant” and marked it with a symbol “h”.
0
gk = 0 0a k and 0
g m = 0 T 0 a m . (3.31)
0
The metric tensor at any arbitrary point P* in the shell space can be defined as
0
G = 0 g k ⊗0 g k = 0 g k ⊗0 g k = 0g ij 0 g i ⊗0 g j = 0g ij 0 g i ⊗0 g j , (3.32)
where
0
g ij = 0 g i ⋅ 0 g j , 0
g ij = 0 g i ⋅ 0 g j and 0
gi ⋅ 0g j = 0g j ⋅ 0gi = δij . (3.33)
Applying eq. (3.27) one obtains

0
(
gαβ = 0 gα ⋅ 0 g β = 0 a α ⋅ 0 a β + θ 3 0 n,α ⋅ 0 a β + 0 a α ⋅ 0 n, β + θ 3 ) ( ) 2 0
n,α ⋅ 0 n, β ,
0
(
gα 3 = 0 gα ⋅ 0 g 3 = 0 a α + θ 3 0 n,α )⋅ 0
n = 0µαδ 0 a δ ⋅ 0 n = 0, (3.34)
0
g 33 = 0 g 3 ⋅ 0 g 3 = 0 n ⋅ 0 n = 1.
Introducing the components of the surface metric tensors of the third order, Woźniak [499]
0
cαβ = 0 n,α ⋅0 n, β = 0bαλ 0bβλ (3.35)
and using (3.11) and (3.22), one can present the components of the space metric tensor as
0
gαβ = 0aαβ − 2 θ 3 0bαβ + θ 3 ( ) 2 0
cαβ . (3.36)
Following Green & Zerna [178] the volume element in the initial configuration 0C can be
introduced as
0
( )
dV = 0 g1 × 0 g 2 ⋅ 0 g 3 dθ 1dθ 2 dθ 3 = 0
g dθ 1 dθ 2 dθ 3 , (3.37)
0
where g stands for the determinant of a matrix containing covariant components of the
metric tensor 0G:
0
(
g = det[ 0 g ij ] = 0 g ij = 0 g1 × 0 g 2 ⋅ 0 g 3 . ) (3.38)
Analogously, the midsurface area element can be expressed (see Green & Zerna [178]) as
0
dΩ = 0 a1 × 0 a 2 dθ 1dθ 2 = 0
a dθ 1dθ 2 , (3.39)
with 0a standing for the determinant of a matrix containing covariant components of the
surface metric tensor 0a:
0
[ ]
a = det 0 aαβ = 0a11 0a22 − 0a12 0a12 . (3.40)
3.3. Shell deformation

It is worthy to notice that due to shear deformations allowed in the presented model,
m
the base vector a3 is perpendicular to the middle surface only for m=0. According to the
56
See also Başar et al. [40].
Reissner-Mindlin kinematics of the First Order Shear Deformation (FOSD) model straight
lines normal to the undeformed shell midsurface remain straight after deformation;
m
therefore the position vector R can be presented as
m
R (θ 1 ,θ 2 ,θ 3 ) = m r (θ 1 ,θ 2 ) + θ 3 m d , (3.41)
m m m
where d = g3 = a3 stands for the local position vector, called "director"57. Within the
FOSD model it is assumed that
∂md (3.42)
m
d = m d (θ 1 ,θ 2 ) and = 0.
∂θ 3
m
The deformation gradient at point P* in the shell space is
m
0 F = m g k ⊗ 0 g k and m
0 FT = 0 g k ⊗ m g k . (3.43)
With (3.32) and (3.43) one can write the Green-Lagrange strain tensor as
m
E ≡ m0 E ≡ 12 ( m
0 FT m0 F − 0 G = 12 ) ( m
gi ⋅ m g j − 0gi ⋅ 0g j ) 0
gi ⊗ 0g j . (3.44)
Substitution of (3.41) into (3.6), due to (3.42), yields

m
gα = m r,α + θ 3 m d,α = m aα + θ 3 m d,α . (3.45)
m
g 3 = 0 + θ 3 m d ,3 + m d = m d
m
With (3.45) components of the metric tensor at any arbitrary point P* in the shell space can
be calculated as follows58 (cf. Pietraszkiewicz [360], f.(3.1.13)):
m
gαβ = m g α ⋅m g β = m a α ⋅m a β + θ 3 ( m
) ( )
d,α ⋅m a β + m a α ⋅m d, β + θ 3
2 m
d,α ⋅m d, β ,
m
gα 3 = m gα ⋅m g 3 = m a α ⋅ m d + θ 3 m d,α ⋅ m d, (3.46)
m
g 33 = m g 3 ⋅m g 3 = m d ⋅ m d.
Consequently, components of the Green-Lagrange strain tensor can be arranged in the
following three groups:
• in-plane and bending terms:
2 m Eαβ = m g α ⋅ m g β − 0 g α ⋅ 0 g β = mgαβ − 0gαβ = m a α ⋅m a β − 0 a α ⋅0 a β +
+θ3 ( d, ⋅ a +
m
α
m
β
m
a α ⋅m d, β − 0 n,α ⋅ 0 a β − 0 a α ⋅ 0 n, β + ) (3.47a)
+ (θ ) ( d, ⋅ d,
3 2 m
α
m
β − 0 n,α ⋅ 0 n, β , )
57
In the classical theory of thin shells, constructed according to the direct approach (cf. Chapter 2), a shell is
regarded as an inextensible one-director Cosserat surface. This surface is defined by a middle surface of a shell
element and a unit vector at each point of the surface; see e.g. Eriksen & Truesdell [156], Green & Zerna [178],
and Naghdi [321].
58
In Woźniak [499] on page 339, one can find expression (1.21) describing components of the metric tensor at the
deformed configuration in a compact form:
g = ma − 2 θ 3 mb + (θ 3 ) mc .
m 2
αβ αβ αβ αβ
As one can notice, this formula is fully consistent with our equation (3.36) defining components of the metric
tensor at the initial configuration. However, if maαβ, mbαβ and mcαβ represent components of the surface metric
tensors at the deformed configuration, the considered equation is valid only when the Kirchhoff-Love constraints
of the theory of thin shells are applied additionally (see e.g. Pietraszkiewicz [360]).
• transverse shear terms:

2 mEα 3 = m g α ⋅ m g 3 − 0 g α ⋅ 0 g 3 = mgα 3 − 0gα 3 = (3.47b)
= m a α ⋅ m d + θ 3 m d,α ⋅ m d,
• transverse normal terms:
2 mE33 = m g 3 ⋅ m g 3 − 0 g 3 ⋅ 0 g 3 = mg33 − 0g33 = m d 2 − 0 n 2 = m
d 2 −1 . (3.47c)
As one can observe in (3.47), the assumptions (3.41) of the FOSD model resulted in
quadratic distributions of membrane and bending strains across the thickness of the shell
(3.47a), with linear distributions of transverse shear strains (3.47b). The transverse normal
strains, in view of (3.47c), are constant across the thickness. From (3.47c), it is also evident
that assuming inextensibility of the director one gets transverse normal strains equal to
zero.
3.4. Strain-displacement relations

Large Rotation Theory
m m
A displacement vector V in configuration C (m>0) (see Fig. 3.1) according to eq.
(3.41) can be written as
m (0) m (1)
m
V = m R − 0 R = ( m r − 0 r ) + θ 3 ( m d − 0 n) = V + θ 3 V , (3.48)
what indicates a linear variation of displacements across the thickness of the shell in the
FOSD model. The displacement vector can be expressed in terms of components referred to
the base vectors of the undeformed shell space
m
V = mVi 0 g i = mV i 0 g i , (3.49)
or in terms of the components referred to base vectors of the undeformed shell mid-surface
m
V = m υ α 0 a α + mυ 3 0 n = mυ α 0 a α + mυ 3 0 n . (3.50)
The following relationships are valid for the displacement vector components:
V α = ( 0 µ −1 )β mυ β ,
α
m
Vα = 0µαβ mυ β , m m
V 3 = mV3 = mυ 3 = mυ3 , (3.51)
where 0 µαβ and (µ )

0 −1 α
β
denote the components of the shifter tensor and its inverse (see
(3.28) and (3.29)).
m
According to (3.48), components of the displacement vector V can be presented as
m ( 0) m (1)
υi (θ 1 ,θ 2 ,θ 3 ) = υi (θ 1 ,θ 2 ) + θ 3 υi (θ 1 ,θ 2 ) .
m
(3.52)
Consequently, components of the Green strain tensor can be derived in the following form:
• in-plane and bending terms:
m ( 0) m (1) m ( 2)
Eαβ (θ 1 , θ 2 , θ 3 ) = Eαβ (θ 1 ,θ 2 ) + θ 3 Eαβ (θ 1 , θ 2 ) + (θ 3 )
2
m
Eαβ (θ 1 , θ 2 ) , (3.53)
where (cf. Pietraszkiewicz [360], f.(3.1.15)):

m (0) m (0) m (0) m (0) m m (0) m (0)
(0)
2 Eαβ = m a α ⋅m a β − 0 a α ⋅0 a β = ϕ βα + ϕαβ + ϕαλ ϕ λβ + ϕα 3 ϕ β 3 ,
m (1)
2 Eαβ = m d,α ⋅m a β + m a α ⋅m d, β − 0 n,α ⋅ 0 a β − 0 a α ⋅ 0 n, β =
m (1) m (1) m (0) m (0) m (0) m m (0) m
(1) (1)
δ δ λ (3.54)
= ϕαβ + ϕ βα − ϕδα b − ϕδβ b + ϕ 0
β
0
α α ϕ λβ + ϕ βλ ϕ λα
m (0) m (1) m (0) m (1)
+ ϕα 3 ϕ β 3 + ϕ β 3 ϕα 3 ,
m ( 2)
2 Eαβ = m d,α ⋅m d, β − 0 n,α ⋅ 0 n, β =
m (1) m (1) m (1) m m (1) m (1)
(1)
δ λ
= − ϕδα b − b 0
β
0
α ϕ λβ + ϕαλ ϕ λβ + ϕα 3 ϕ β 3 ,
with the following substitutions:
m (n) m (n) m (n)
ϕαβ = υα β − 0bαβ υ3 , (3.55a)
m (n) m (n) m (n)

ϕαλ = υ λ α − 0bαλ υ3 , (3.55b)
m (n) m (n) m (n)

ϕα 3 = υ3,α + 0bαλ υλ , (3.55c)
m (n) m (0) m (0)

υα β = υα , β − 0 Γαβλ υλ ; (3.55d)
• transverse shear terms:

m (0) m (1)
m
Eα 3 (θ 1 ,θ 2 ,θ 3 ) = Eα 3 (θ 1 ,θ 2 ) + θ 3 Eα 3 (θ 1 ,θ 2 ) , (3.56)
where
m (0) m (1) m (0) m (0) m m (0) m (1)
(1)
2 Eα 3 = m a α ⋅ m d = υα + ϕα 3 + ϕαλ υ λ + ϕα 3 υ3 , (3.57)
m (1) m (1) m (1) m (1) m m (1) m (1)
(1)
2 Eα 3 = m d ,α ⋅ m d = ϕα 3 − 0bαλ υ λ + ϕαλ υ λ + ϕα 3 υ3 ;
• transverse normal terms:
m (0) m (1) m (1) m
( d ⋅ m d − 0 n ⋅0 n ) = υ3 + 12 υ k υ k .
(1)
m
E33 (θ 1 , θ 2 ,θ 3 ) = E33 (θ 1 , θ 2 ) = 1
2
m
(3.58)
Up to this moment, no simplifying assumptions were introduced regarding the

magnitude of deformations; therefore the strain-displacement equations presented above
can be treated as a part of the FOSD Large Rotation Theory (LRT). Since all six
3.4. Strain-displacement relations 61
components of the displacement vector (3.52) are involved in the formulae above, we may
label this formulation for further references as LRT6 (the Large Rotation Theory with 6
parameters).
Inextensibility of director in 5 parameter theories

The straightforward interpretation of the director inextensibility assumed in the present
considerations results in the following equation59:
m
E33 = 0 . (3.59)
The simplest realization of the director inextensibility resulting in a pure 5-parameter

theory can be obtained just by assuming
m (1)
υ3 = 0 , (3.60a)
what leads to the replacement of the FOSD hypothesis (3.52) with the following form
m (0) m (1)
m
υα (θ 1 , θ 2 , θ 3 ) = υα (θ 1 , θ 2 ) + θ 3 υα (θ 1 , θ 2 ), α = 1, 2 (3.60b)
m ( 0)
m
υ3 (θ 1 , θ 2 , θ 3 ) = υ3 (θ 1 , θ 2 ).
m (1)
Bearing in mind that, according to (3.58), υ3 is equal to the linear (dominating) part of
the component mE33, one can consider (3.60a) as a simple (approximate) realization of the
constraint (3.59).
As a consequence of the assumption (3.60a) one gets
m (0) m (0) m (0) m (1) m (1) m (1) m (1)
ϕαβ = υα β − 0bαβ υ3 but ϕαβ = υα β − 0bαβ υ3 = υα β , (3.61a)
1
424 3
0
m (0) m (0) m (0) m (1) m (1) m (1) m (1)
ϕαλ = υ λ α − 0bαλ υ3 but ϕαλ = υ λ α − 0bαλ υ3 = υ λ α , (3.61b)

123
0
m (0) m (0) m (0) m (1) m (1) m (1) m (1)
ϕα 3 = υ3,α + 0bαλ υλ but ϕα 3 = υ3,α + 0bαλ υλ = 0bαλ υλ , (3.61c)
123
0
The above considerations result in the final form of strain-displacement relations (in
the configuration mC) for the five-parameter shell theory, as listed below:
m (0) m (0) m (0) m (0) m m (0) m (0)
(0)
2 Eαβ = ϕ βα + ϕαβ + ϕαλ ϕ λβ + ϕα 3 ϕ β 3 , (3.62a)
59
One should notice that the assumption (3.59) will be used only for the shell kinematics, whereas the constitutive
relationships to be introduced in Section 3.6 will be modified according to the condition of the plane stress state in
the shell.
m (1) m (1) m (1) m (0) m (0) m (0) m m (0) m

(1) (1)
δ δ λ
2 Eαβ ≅ υα β + υ β α − ϕδα b − ϕδβ b + ϕ 0
β
0
α α υλ β + ϕ βλ υλ α
(3.62b)
m (0) m (1) m (0) m (1)
+ 0bβδ ϕα 3 υδ + 0bαδ ϕ β 3 υδ ,
m
m ( 2) m (1) m (1) (1) m (1)
 m (1)
 m (1)

2 Eαβ ≅ − 0bβδ υδ α − 0bαλ υ λ β + υ λ α υ λ β +  0bαλ υ λ   0bβδ υδ  , (3.62c)
  
m (0) m (1) m (0) m (0) m

(1)
2 Eα 3 ≅ υα + ϕα 3 + ϕαλ υ λ , (3.62d)
m (1) m (1) m
(1)
2 Eα 3 ≅ υ λ υλ α . (3.62e)
m (1) m (1)
The parameters υ1 and υ 2 are often interpreted in the literature as the rotations ϕ1
and ϕ2 about the mid-surface base vectors60 0a2 and 0a1, respectively (see Fig. 3.3).
However, as it will be shown later in this report, this holds true only for small and moderate
rotations (e.g. linear, small deflection analysis or non-linear analysis based on the refined
von Kármán theory or moderate rotation theory).
0 3
a
0 1 0 2
a a
j2 j1
Fig. 3.3: Rotation angles in FEM models
The strain-displacement relations presented in (3.62) characterize a five-parameter

variant of the large rotation theory (LRT5) which was applied by many researches, e.g. in
Ferro [164], Ferro et al. [165], Bouhafs et al. [69], Kłosowski & Woźnica [248]. However,
as it was indicated by Başar & Kintzel [41], a large rotation theory based on the hypothesis
(3.60b) with only five kinematical parameters is not capable of treating finite rotations.
Therefore in the following an enhanced approach will be developed.
60
One should remember that according to (3.9) we have 0a1•0a2 = 0a1•0a2 =0
Enhanced interpretation of rotations

Looking for a more accurate formulation one can express the changes of the
inextensible director by using Euler angles (see e.g. Korn & Korn [253]). One can imagine
0
that the transformation of any vector from the initial configuration C to the configuration
m
C can be achieved by two subsequent rotations, first about the mid-surface base vector
0
a1, then about 0a2. With the two rotation angles from Fig. 3.3, the matrix operators for
such a conversion are as follows:
1 0 0   cos( mϕ1 ) 0 sin( mϕ1 ) 
[ m
]
R1 ( ϕ 2 ) = 0 cos( ϕ 2 ) sin( ϕ 2 ) ,
 m m
[ ] 
R2 ( mϕ1 ) =  0 1 0
 .(3.63)
,
0 − sin( m ϕ 2 ) cos( mϕ 2 ) − sin( mϕ1 ) 0 cos( mϕ1 )
 
and the rotation matrix for the whole transformation can be obtained as
 cos(mϕ1 ) − sin( mϕ1 ) sin( mϕ 2 ) sin( mϕ1 ) cos(mϕ 2 ) 

[ ][ ]
[R ] = R2 ( mϕ1 ) R1 ( mϕ2 ) =  0 cos(mϕ 2 ) sin( mϕ 2 )
 .(3.64)

− sin( mϕ1 ) − cos(mϕ1 ) sin( mϕ 2 ) cos(mϕ1 ) cos(mϕ 2 )
 
m
In particular, this transformation can be used to describe the director d introduced in
0
(3.41), as the image of the vector n obtained by performing subsequent rotations with the
m m 0 0
rotation angles ϕ2 and ϕ1 about the mid-surface base vectors a1 and a2, respectively:
m
d = sin( mϕ1 ) cos( mϕ 2 ) 0 a1 + sin( mϕ 2 ) 0 a 2 + cos( mϕ1 ) cos( mϕ 2 ) 0 n . (3.65)
m
Now, turning back to the relation (3.56) for the displacement vector V in the
m
configuration C, one obtains
m (1)
( )
V = m d − 0 n = sin( m ϕ1 ) cos( m ϕ 2 ) 0 a 1 + sin( m ϕ 2 ) 0 a 2 + cos( m ϕ1 ) cos( m ϕ 2 ) − 1 0 n . (3.66)
m m
It is obvious that for small and moderate values of the angles ϕ1 and ϕ2 one can assume
that
sin( m ϕα ) ≅ m ϕα and cos( m ϕα ) ≅ 1 . (3.67)
Thus for small and moderate rotations Eq. (3.66) yields

m (1)
V ≅ m ϕ1 0 a 1 + m ϕ 2 0 a 2 + 0 0 n . (3.68)
Hence, for the small or moderate rotation theory the kinematical hypothesis (3.60b) is
justified, however, as it was already pointed out, this approximation cannot be accepted for
large rotations. Therefore the exact relation (3.66) will be applied in the following
derivations for the numerical implementation of the large rotation theory based on the
enhanced interpretation of rotations.
For the purpose of the incremental description, it is necessary to construct a
corresponding relation for the displacement increment. Starting with the Taylor series
2
expansion of the displacement vector at the configuration C in the vicinity of the actual
1
configuration C, one obtains
(1) (1)
2 (1) 1 (1) 1 (1)
∂V (1)
∂V (3.69)
V= V+ ∆ V ≅ V+ ∆ϕ1 + ∆ϕ 2 ,
∂ϕ1 ∂ϕ 2
t t
where the higher-order terms have been neglected. As a consequence the linearized
incremental relation reads
 (1) 
 ∆ υ(11)   cos( ϕ1 ) cos( ϕ 2 ) − sin(1ϕ1 ) sin(1ϕ 2 ) 
1 1
∆ υ  =   ∆ϕ1  (3.70)
0 cos(1ϕ 2 )  ∆ϕ .
 (12)  
 ∆ υ  − sin( ϕ1 ) cos(1ϕ 2 )
1
− cos( ϕ1 ) sin( ϕ 2 )  2 
1 1
 
3
Also, it is quite obvious that the exact relation that can be obtained taking advantage of
formula (3.66) reads:
2 (1) 1 (1)
( )
(1)
∆ V = V − V = sin( 2ϕ1 ) cos( 2ϕ 2 ) − sin(1ϕ1 ) cos(1ϕ 2 ) 0 a1 + (3.71)
( ) 0 2
(
+ sin( ϕ 2 ) − sin( ϕ 2 ) a + cos( ϕ1 ) cos( ϕ 2 ) − cos( ϕ1 ) cos( ϕ 2 ) n.
2 1 2 2 1 1
) 0
Assuming now the usual incremental decomposition 2ϕα =1ϕα + ∆ϕα and taking
sin( ∆ϕα ) ≅ ∆ϕα and cos( ∆ϕα ) ≅ 1 (3.72)
one gets from (3.71) exactly the same relation as (3.70).

Consequently, we obtain a numerical formulation based on the increments of 5
parameters similar as in the LRT5 formulation. Nevertheless, it should be stressed that the
approximated relation (3.70) is used only during the increment and the order of such an
approximation is within the range of the usual linearization approximation applied in the
incremental Lagrangian formulation. It is important that after the increment is performed,
the new configuration must be updated according to the exact relation
 2 (1)   1 (1) 
 1   cos(∆ϕ ) − sin(∆ϕ ) sin(∆ϕ ) sin(∆ϕ ) cos(∆ϕ )   υ1  0
υ
 2 (1)   1 1 2 1 2
  (1)    .(3.73)
1
υ
 2  = 0 cos( ∆ϕ 2 ) sin( ∆ϕ 2 )   υ 2  − 0
 2 (1)  − sin(∆ϕ ) − cos(∆ϕ ) sin(∆ϕ ) cos(∆ϕ ) cos(∆ϕ )  1 (1)  1
 υ3   1 1 2 1 2   υ + 1
3
 
   
It is necessary to note that the transformation based on the Euler angles yields a unique
result only for angles from the interval (-π/2; π/2) (see e.g. Korn & Korn [253]), however
in the present formulation this restriction applies only to the increments of the rotations and
not to the total rotations (see also Betsch et al. [58]).
It is essential to understand also that the updating process as described by (3.73) is
applied in every equilibrium iteration before the balanced forces are established. Such a
strategy accompanied by a suitable incremental-iterative procedure with a flexible size of
load increment guarantees the desired accuracy of the analysis.
One should remember that all 6 parameters61 which are necessary to describe the
m
displacement vector V as given in Eq. (3.52) must be continuously stored and updated
during the analysis. It is quite obvious that such formulation is not a real “six- parameter
theory”, neither a “five-parameter theory”, therefore in the present paper it will be referred
to as the “LRT56” formulation62.
Remark 3.1: Starting from the assumption of the inextensible director

m
d 2
= md ⋅ md = 1 (3.74)
one can draw the following expression

m (1)
m
d,α ⋅ m d = 2 Eα 3 = 0 , (3.75)
what indicates a constant shear deformation throughout the thickness of the shell in the
LRT56 formulation,
m (1) m (0)
Eα 3 (θ 1 ,θ 2 ) = 0 ⇒ m
Eα 3 (θ 1 ,θ 2 ,θ 3 ) = Eα 3 (θ 1 ,θ 2 ) , (3.76)
on the contrary to the linear distribution of the shear strains along the thickness of the shell
obtained in the simplified LRT5 model (compare eq. (3.62e)).
Remark 3.2: The motion of the inextensible director can be also described using the
Rodrigues rotation vector m« as presented in the Fig. 3.4.
m
w m
d
m
w
z2
+z
2
0
n +z 1
z1 0 n
m d= z2
m
w z1
0
n
Fig. 3.4: Rotation of inextensible director defined by Rodrigues rotation vector
Assuming that |md| = |0n|=1 and by using very basic trigonometric calculus one can get
from Fig. 3.4 the following relations
sin( mω ) (3.77)
z1 = (cos( mω ) − 1) 0 n and z2 = m
« × 0n
m
ω
m (0) m (1)
61
Six scalar displacement components: υi and υ i , i =1, 2, 3
62
A similar treatment of large rotations was applied for isotropic shells by Ramm & Matzenmiller [383] and by
Wriggers & Gruttmann [500, 501], and for composite shells by Başar et al. [37] (see also Başar & Ding [36]) as
well as by Brank et al. [79].
with mω, standing for the magnitude of the rotation equal to the norm of the rotation vector:
ω=
m m
« = ( mω1 ) 2 + ( mω2 ) 2 +( mω3 ) 2 (3.78)
m
Finally the director in the configuration C, md can be expressed as
m
d = 0 n + z1 + z 2 = cos( mω ) 0 n +
sin( mω )
m
ω
m
[ ]
« × 0 n = m ℜ( m « ) 0 n (3.79)
with the rotation tensor [mℜ( m« & depending on the Rodrigues rotation vector m«
[ m
]
ℜ( m « ) = I +
sin( mω )
m
ω
( m « +
1 − cos( mω )
m
ω2
(m « , (3.80)
where the skew-symmetric tensor ( m« known as a spin tensor is introduced as
 0 − m ω3 ω2 
m
 m 
m
« × z = ( « z and ( «
m m
 ω3 0 − ω1  ei ⊗ e j .
m (3.81)
− mω 2 m
ω1 0 

Consequently, after appropriate transformations one can obtain an analogous relation to that
given by (3.70):
 1
ω cos(1ω ) − sin(1ω ) 1 1 sin(1ω ) ω cos(1ω ) − sin(1ω ) 1 2 
1
 (1)   ω1 ω2 + ( ω2 ) 
 ∆ υ(11)   (1ω )3 1
ω (1ω )3 
∆ υ  = − sin( ω ) − ω cos( ω ) − sin( ω ) (1ω ) 2
1 1 1 1 1
ω cos(1ω ) − sin(1ω ) 1 1 ∆ω 
 (1)  
− ω1 ω2   1 
2 1
ω ( ω)
1 3 1
( ω)
1 3
∆ω
∆υ     2
sin(1ω ) 1 sin(1ω ) 1
   
3
− 1 ω1 − 1 ω2
 ω ω 
(3.82)
The treatment of finite rotations as presented above provides an alternative strategy to
the application of Euler angles; however, one should notice that the components of the
rotation vector (3.78) are expressed here with respect to the general Cartesian reference
frame. A finite rotation formulation based on of Rodrigues rotation vector was
implemented for FEA of isotropic shells by Simo et al. [437]63 and later it was applied also
for laminated shells e.g. by Carrera & Parisch [97], Başar et al. [40], Vu-Quoc et al. [478],
Balah and Al-Ghamedy [31].
LRT strain-displacement relations proposed by other authors

After some standard transformations, it can be proved that the strain-displacements
relations presented for the LRT6 formulation in Section 3.4.1 of this report are equivalent
to those given by Habip [186], Pietraszkiewicz [360], Başar & Kintzel [41] as well as by
63
Comprehensive analyses of various constrained finite rotation formulations were presented by Betsch et al. [58]
and Brank & Ibrahimbegović [77]. Additional resources on that subject are listed in Section 2.2 of this report.
Büchter & Ramm [82, 84]64. The same expressions can be obtained also from the general
formulae given for large deformation analysis of laminated shells by Librescu [290], who
similarly as Habip [186] started with the Green strain tensor in the shell space expressed as
m
(
Eij = 12 mVi j + mV j i + mV k i
m
Vk j ), (3.83)
where (.) i stands for the covariant derivative with respect to the metric of the undeformed
shell space. Librescu [290] incorporated a power series expansion of the displacements and
3
strains with respect to the thickness coordinate θ :
 m (k )

υi (θ 1 ,θ 2 ,θ 3 ) = ∑ (θ 3 )
p
υi (θ 1 ,θ 2 ) , (3.84)
m k
k =0  
 m (k )

( )
q
Eij (θ 1 ,θ 2 , θ 3 ) = ∑  θ 3
k
m
Eij (θ 1 , θ 2 ) . (3.85)
k =0  
Using such a general approach one can develop a shear deformation model of an arbitrary
order, depicted by the number p. It is evident that putting p=1 in (3.84) one obtains the
relation (3.52) which exemplifies the FOSD model. The corresponding strain-
displacements relations can be calculated from (3.85) – after some routine operations the
resulted expressions agree entirely with those presented here in equations (3.53-54) and
(3.56-58).
Non-linear strain-displacement relations for large rotation shell theories with
inextensible director, corresponding to those obtained for the LRT56 model, were given
also by Başar [34], Başar & Krätzig [42], Brank et al. [79] and Parisch [348]65; however,
authors of those papers simplified their strain-displacements relations by neglecting also
m
quadratic terms in Eαβ :
m ( 2) m (0) m (1)
Eαβ (θ 1 ,θ 2 ) = 0 ⇒ m
Eαβ (θ 1 ,θ 2 ,θ 3 ) = Eαβ (θ 1 ,θ 2 ) + θ 3 Eαβ (θ 1 ,θ 2 ). (3.86)
Analogous simplifications were applied also by Başar and Krätzig in their linear theory of
shear deformable shells [43] (see also Hossain et al. [205]).
Neither a constant value of the transverse shear strains obtained in (3.76) nor a linear
distribution of the shear deformations throughout the thickness of the shell described in
(3.56), agrees with a quadratic dependency of the transverse shear deformation that can be
observed with respect to the thickness coordinate in the 3D elastic solutions for single layer
thick plates (see e.g. Ambartsumyan [14], Jemielita [227], Wittrick [497]). Among several
different ideas for a possible improvement in the reproduction of the transverse shear
strains within the range of the ESL models, one should list the refined theories proposed for
large rotation analysis of laminated shells by Başar [35] and Başar et al. [37]. Those
proposals can be considered as extensions of the Third Order Shear Deformation (TOSD)
theory proposed for composite plates by Reddy [393, 394]. It is possible to obtain a TOSD
nonlinear theory for laminated shells from the formulae given by Librescu [290]. However,
putting p=3 in (3.84) one obtains the higher level of truncation in the series expansion for
64
Originally, Büchter & Ramm [82, 84] presented the strain-displacement relations by means of the absolute
tensor notation.
65
An extension for multilayered shells was presented by Carrera & Parisch [97].
strains in (3.85) than in Başar’s formulation [35]. The two variants of the TOSD finite
rotation theory proposed in Başar et al. [37] differ by the number of independent
displacement variables: the model RT7 exploits 7 independent displacement parameters but
in the RT5 model the number of displacement variables was limited to 5 by means of the
following two constraints, Başar [35]:
3
• zero transverse shear strains on the outer surfaces (θ = ± h/2);
• zero length-change across the thickness ( m E33 (θ 1 ,θ 2 ,θ 3 ) = 0 ).
The resulting strain-displacements relations of the RT5 model are as follows
m (0) m (1) m ( 2) m ( 3)
m
Eαβ (θ 1 ,θ 2 ,θ 3 ) = Eαβ (θ 1 ,θ 2 ) + θ 3 Eαβ (θ 1 ,θ 2 ) + θ 3 ( )
2
Eαβ (θ 1 ,θ 2 ) + θ 3 ( ) 3
Eαβ (θ 1 ,θ 2 ) ,
(3.87)
m (0) m (0)
m
Eα 3 (θ 1 ,θ 2 ,θ 3 ) = Eα 3 (θ 1 ,θ 2 ) −
h2
( )
4 3
θ
2
Eα 3 (θ 1 ,θ 2 ) . (3.88)
The strain terms that are underlined in above equations were not included in the FOSD
m
LRT model. As compared to (3.53), the strain component Eαβ of the RT5 model contains
one additional term (underlined in (3.87)) being the third order function of the thickness
3
coordinate θ . The underlined term in (3.88) provides a quadratic distribution of the
m
transverse shear strain Eα3 across the thickness of the shell – it has replaced the part, which
3
was a linear function of the thickness coordinate θ in the FOSD LRT model, eq. (3.48).
From the constraints of the RT5 model it follows that the second term in (3.88) is explicitly
m
related to the first term of the strain component Eα3. Also the new term of the strain
m
component Eαβ can be related to the same set of displacement parameters as used in the
FOSD LRT. In a range of small displacements Başar’s RT5 formulation [35, 37] is
equivalent to the HST model proposed for linear analysis of laminated shells by Reddy &
Liu [401]. Since the RT5 model uses exactly the same five displacement parameters as the
FOSD, it can be categorized as a “Refined” FOSD (see also Auricchio & Sacco [28],
Khdeir et al. [236], Knight & Qi [249]) or as “Enhanced” FOSD, Kim & Cho [238]. A
“real” TOSD formulation corresponding to the RT7 model was proposed also by Balah &
Al-Ghamedy [31].
Another implementation of the Reddy concept of TOSD [393] (see also Reddy & Liu
[401]) can be found in a large rotation formulation for laminated shells of Dennis &
Palazotto [138, 139] (see also Chaplin & Palazotto [106], Tsai et al. [470]). However, their
model was simplified by neglecting all non-linear terms for the transverse shear strains.
Additionally, an approximate procedure has been applied for an updating of rotations, in
consequence, their model was not capable to deal with really large rotations – one can
guess that such an observation caused the authors to call their approach the Simplified
Large Rotation (SLR) formulation in their subsequent papers – see e.g. Naboulsi &
Palazotto [319].
One can expect that both variants of the TOSD finite rotation theory proposed by Başar
et al. [37] should provide a more accurate solution for transverse shear strains in a single
layer composite then that which can be achieved with the FOSD. However, for a multilayer
composites, especially those with asymmetrical stacking sequence, the distribution of the
transverse shear strains estimated with the TOSD is not drastically superior to the FOSD
solution (see e.g. Başar et al. [37] or Krätzig & Jun [256]). According to Reddy [395]
among different single-layer theories developed for the laminated composites “the FOSD
seems to provide the best compromise between accuracy and computational efficiency”.
As it was stated earlier, one of fundamental assumptions adopted for the large rotation
theory of anisotropic shells developed in this report was the limited magnitude of strains.
Customarily such an assumption justifies a disregard for the thickness change66; however,
the constraint given by (3.59) can introduce a significant error into the strain energy density
if it is not accompanied with an appropriate modification of the constitutive relations (cf.
Pietraszkiewicz [361]).
It is worthy noting that the LRT6 equations given in this section are free from the
constraint of the inextensible director. Nevertheless, Simo et al. [440] indicated that the
direct application of such a 6-parameter formulation suffers from the defect of locking due
to appearing of artificial transverse normal stresses, and suggested instead a multiplicative
decomposition of the director field into the inextensible director (a unit vector) and the
scalar thickness parameter (see also Braun et al. [81], Bischoff & Ramm [60], Bischoff et
al. [63]). A further improvement can be achieved when a 7-parameter formulation is
3
applied where transverse normal strain varies linearly along θ ; Bischoff & Ramm [59, 60],
El-Abbasi & Meguid [154], Ramm [382], Brank et al. [78], Brank [74]. It seems quite
obvious that strain-displacement relations utilized within 6- and 7-parameter formulations
are more general than those obtained for the LRT6 model in this report. Nevertheless, in the
limit case, i.e. assuming zero variation of thickness change parameters, they should be
reducible to the set of equations equivalent to those of the LRT6 theory.
Simplified strain-displacement relations for Moderate Rotation Theory

By assuming a moderate magnitude of rotations (see e.g. Pietraszkiewicz [360]) one
can replace the strain-displacement relations of the LRT5 shell theory (3.62) with the
following approximate relations of the MRT5 shell theory67:
m (0) m (0) m (0) m (0) m (0)
2 Eαβ ≅ ϕ βα + ϕαβ + ϕα 3 ϕ β 3 , (3.89a)
m (1) m (1) m (1) m (0) m (0)

2 Eαβ ≅ υα β + υ β α − ϕδα 0bβδ − ϕδβ 0bαδ +
(3.89b)
m (0) m (1) m (0) m (1)
+ 0bβδ ϕα 3 υδ + 0bαδ ϕ β 3 υδ ,
m ( 2) m (1) m (1)
 m (1)
 m (1)

2 Eαβ ≅ − 0bβδ υδ α − 0bαλ υλ β +  0bαλ υλ   0bβδ υδ  , (3.89c)
  
m (0) m (1) m (0) m (0) m

(1)
2 Eα 3 ≅ υα + ϕα 3 + ϕαλ υλ , (3.89d)
66
Assumption of the unchanging shell thickness during deformation is equivalent to the inextensibility of the
director.
67
Compare Schmidt & Reddy [426], Schmidt & Weichert [427], Palmerio et al. [342], Kreja & Schmidt [263, 264,
265], and Kreja et al. [269].
m (1) m (1) m
(1)
2 Eα 3 ≅ υ λ υλ α . (3.89e)
It is interesting to notice that the transverse shear strain components presented above
for the MRT5 formulation (3.89d-e) are exactly the same as used in the LRT5 theory
(3.62d-e).
Simplified strain-displacement relations for Refined von Kármán Theory

A further simplification of the strain-displacement relations can be obtained according
to the von Kármán type non-linearity68 by retaining only those nonlinear terms that contain
derivatives of the transverse deflection (RVK5 model):
m (0) m (0) m (0) m (0) m (0)
2 Eαβ ≅ ϕ βα + ϕαβ + υ3,α υ3, β , (3.90a)
m (1) m (1) m (1) m (0) m (0)

2 Eαβ ≅ υα β + υ β α − ϕδα 0bβδ − ϕδβ 0bαδ , (3.90b)
m ( 2) m (1) m (1)
2 Eαβ ≅ − 0bβδ υδ α − 0bαλ υ λ β , (3.90c)
m (0) m (1) m (0)

2 Eα 3 ≅ υα + ϕα 3 , (3.90d)
m (1)
2 Eα 3 ≅ 0 . (3.90e)
As one can notice, nonlinear terms in the RVK5 model appear only in the membrane
strain components. The RVK5 model is characterized also by a constant value of the
transverse shear strains similarly as in the LRT56 concept, but in contrast to the LRT5 and
MRT5 formulations.
Final inventory of strain-displacement relations for various non-linear FOSD models

To make easier a detailed comparison of various non-linear FOSD shell theories
considered above, the strain-displacement relations obtained for the LRT6, LRT56, LRT5,
MRT5, and RVK5 models are gathered together in Table 3.1 (membrane and bending strain
terms), Table 3.2 (transverse shear strain terms), and Table 3.3 (transverse normal strain
terms).
68
Compare e.g. Reddy [392], Reddy & Chandrashekhara [400], Schmidt & Reddy [426], and Palmerio et al. [342].
Table 3.1: Strain-displacement relations for various variants of FOSD shell theory:
membrane and bending terms
m (0) m (1) m ( 2)
Eαβ (θ 1 ,θ 2 ,θ 3 ) = Eαβ (θ 1 ,θ 2 ) + θ 3 Eαβ (θ 1 ,θ 2 ) + (θ 3 ) theory:

2
m
Eαβ (θ 1 ,θ 2 )
m (0) m (0) m (0) m (0) m (0) m (0)

LRT6
ϕ βα + ϕαβ + ϕαλ ϕ λβ + ϕα 3 ϕ β 3 LRT56
LRT5
m (0) m (0) m (0) m (0) m (0)
ϕ βα + ϕαβ + ϕα 3 ϕ β 3 MRT5
2 Eαβ =
m m (0) m
(0) (0) m (0)
ϕ βα + ϕ αβ + υ 3,α υ 3,β RVK5
m (0) m (0)
ϕ βα + ϕαβ linear
m (1) m (1) m (0) m (0) m (0) m

(1)
ϕαβ + ϕ βα − ϕδα b − ϕδβ b + ϕαλ ϕ λβ +
0 δ 0 δ LRT6
β α
m (0) m m (0) m (1) m (0) m (1)
(1)
+ ϕ βλ ϕ λα + ϕα 3 ϕ β 3 + ϕ β 3 ϕα 3 LRT56
m (1) m (1) m (0) m (0) m (0) m
(1)
υα β + υ β α − ϕδα b − ϕδβ b + ϕαλ υλ β + 0 δ
β
0
α
δ
m (0) m m (0) m (1) m (0) m (1)

LRT5
m (1) (1)
λ 0 δ 0 δ
2 Eαβ = + ϕ β υ λ α + bβ ϕα 3 υδ + bα ϕ β 3 υδ
m (1) m (1) m (0) m (0)

υα β + υ β α − ϕδα 0bβδ − ϕδβ 0bαδ +
MRT5
m (0) m (1) m (0) m (1)
+ 0bβδ ϕα 3 υδ + 0bαδ ϕ β 3 υδ
m (1) m (1) m (0) m (0)
RVK5
υα β + υ β α − ϕ δα 0 bβδ − ϕ δβ 0 bαδ
linear
m (1) m (1) m (1) m

(1) m (1) m (1) LRT6
δ λ
− ϕ δα b − b 0
β
0
α ϕ λβ + ϕαλ ϕ λβ + ϕα 3 ϕ β 3 LRT56
m (1)
m (1) m (1) m (1)
 m (1)
 m (1)

− 0 bβδ υ δ α − 0bαλ υ λ β + υλ α υλ β +  0 bαλ υ λ   0 bβδ υ δ  LRT5
m ( 2)   
2 Eαβ = m (1) m (1)
 m (1)
 m (1)

− 0 bβδ υδ α − 0bαλ υ λ β +  0 bαλ υ λ   0 bβδ υδ  MRT5
  
m (1) m (1)
RVK5
− 0 bβδ υ δ α − 0bαλ υλ β
linear
transverse shear terms
m ( 0) m (1)
m
Eα 3 (θ 1 ,θ 2 ,θ 3 ) = Eα 3 (θ 1 ,θ 2 ) + θ 3 Eα 3 (θ 1 ,θ 2 ) theory:
m (1) m (0) m (0) m

(1) m (0) m (1) LRT6
υα + ϕα 3 + ϕαλ υλ + ϕα 3 υ3 LRT56
m (0)
2 Eα 3 = m (1) m (0) m (0) m
(1) LRT5
υα + ϕα 3 + ϕαλ υ λ MRT5
m (1) m (0) RVK5
υα + ϕ α 3 linear
m (1) m (1) m (1) m m (1) m (1)

(1)
ϕα 3 − 0bαλ υλ + ϕαλ υλ + ϕα 3 υ3 LRT6
m (1) 0 LRT56
m (1)
2 Eα 3 = m (1)
LRT5
υ λ υλ α
MRT5
RVK5
0
linear
transverse normal terms
m (0)
m
E33 (θ 1 ,θ 2 ,θ 3 ) = E33 (θ 1 ,θ 2 ) theory:
m (1) m (1) m
(1)
υ3 + 1
υ k
υ LRT6
2 k
m (0) LRT56
E33 = LRT5
0 MRT5
RVK5
linear
3.5. Virtual work principle

An implementation of the proposed shell theory into the Finite Element Method can be
constructed basing on an appropriate variational principle; Kleiber [244, 245], Bathe [46],
Reddy [398], Zienkiewicz & Taylor [510], Chapelle & Bathe [105], Bischoff et al. [63],
and Chróścielewski et al. [119]. Depending on the assumed set of unknowns one can
choose between mixed functionals and single field functionals.69 The mixed variational
69
The single field variational principles are also called the irreducible variational principles; Zienkiewicz &
Taylor [510].
3.5. Virtual work principle 73
principle of Hu-Washizu70 exploits an independent interpolation of three fields:

displacements, stresses and strains. Among all possible formulations of a two-field
functional, probably the most popular is the variational principle of Hellinger-Reissner,
with independent interpolation of displacements and stresses.71 The principle of minimum
complementary energy and the principle of minimum potential energy are examples of
single field functionals. The principle of minimum complementary energy can be
considered as equivalent to the virtual forces work principle; in the resulting numerical
model stresses appear as the only free variables72. The principle of minimum
potential energy constitutes a proper weak form for a pure displacement formulation. For
conservative loading the principle of minimum potential energy is equivalent to the virtual
displacements work principle.
The virtual work principle states equilibrium between the internal virtual work, 2δWi,
and the external virtual work, 2δWe, in the configuration 2C:
δWi − 2δWe = 0 .
2
(3.91)
Assuming that all external interactions are represented by body and surface forces the
virtual displacements external work can be taken as
2
δWe = ∫ 2 f i δui 2 dV + ∫ 2 p i δuiA 2 dA , (3.92)
2 2
V A
where 2f i and 2pi are components of external forces, acting per unit volume element and unit
surface element, respectively, and δui are covariant components of virtual displacement
vector (δuiA stands for virtual displacement of the surface). Since the equation (3.91)
describes an equilibrium state in the configuration 2C, the admissible field of virtual
displacements δui can be treated as the variation of the displacements in this configuration:
δui = δ(2ui).
The internal virtual work in 2C, which is yet to be determined, can be expressed as
δWi = ∫ 2σ ij δ 2eij 2 dV ,
2
(3.93)
2
V
where 2σij denote the components of the Cauchy stress tensor in the configuration 2C and
δ2eij are the variations of components of the infinitesimal strain tensor in the same
configuration. Within the Total Lagrangian formulation all quantities in the unknown
configuration 2C are referred to the initial configuration 0C. This can be achieved by using
the following relation:
70
In the literature the three-field mixed variational principle is commonly associated with the names of H.-C. Hu
and K. Washizu who independently proposed suitable variational principles in 1955. However, quite recently
Felippa indicated that already in 1951 Fraeijs de Veubeke presented a more general four-field variational principle,
in which additionally an independently varied surface tractions were considered (see: Felippa C.A.: Fraeijs de
Veubeke: neglected discoverer of the “Hu-Washizu Functional”, IACM Expressions 12, 2002, 8-10).
71
The mixed variational principles of Hu-Washizu and of Hellinger-Reissner are very useful for the development
of extremely efficient Enhanced Assumed Strain (EAS) shell and plate elements, see e.g. Betsch et al. [57], Fontes
Valente et al. [167], Simo & Hughes [438], Yang et al. [502], Schlebusch et al. [422, 423], and Zienkiewicz &
Taylor [510].
72
Although a direct recovery of stresses in the numerical models based on the principle of minimum
complementary energy seems to be very attractive, however, in practical applications, pure stress formulations are
very seldom due to difficulties with a choice of approximation functions; Zienkiewicz & Taylor [510]. Much more
popular are Hybrid Stress Elements (Pian [356], Ramm [382], Zienkiewicz & Taylor [510]) based on the two-field
Hellinger-Reissner variational principle
∫ σ ij δ 2 eij 2 dV = ∫ 02 S mn δ 2 Emn 0 dV ,
2
(3.94)
2 0
V V
where 02Smn are components of the second Piola-Kirchhoff stress tensor and 2Emn are
components of the Green-Lagrange strain tensor. Both tensors are measures of the
corresponding quantities in the configuration 2C but referred to the initial configuration 0C,
Bathe [46], Kleiber [244].
The volume element 0dV is related to the midsurface area element 0dΩ as
0
dV = 0µ dθ 3 0dΩ , (3.95)
where 0µ is the determinant of the shifter tensor in the initial configuration 0C (compare
3.30).
With the assumption of zero transverse normal stresses the internal virtual work can be
expressed as:
2
δWi = ∫ 02 S αβ δ 2Eαβ 0 dV + ∫ 202S α 3 δ 2Eα 3 0 dV . (3.96)
0 0
V V
Following (3.53) and (3.56), the membrane, bending and transverse shear strains can be
expanded in terms of the thickness coordinate 3θ
   
δWi = ∫ 02 S αβ δ 2 Eαβ + θ 3 δ 2 Eαβ + (θ 3 ) δ 2 Eαβ  0 dV + ∫ 202S α 3 δ 2 Eα 3 + θ 3 δ 2 Eα 3  0 dV .
(0) (1) ( 2) (0) (1)
2 2
0
V   0
V  
(3.97)
With (3.95) the volume integration in (3.97) can be split into through-the-thickness
pre-integration followed by the area integration over the middle surface:
h
 (0) 
( )
2 (1) ( 2)
2
δWi = ∫∫ S αβ δ 2 Eαβ + θ 3 δ 2 Eαβ + θ 3 δ 2 Eαβ  0 µ dθ 3 0dΩ +
2 2
0
0
Ω − h2  
h
. (3.98)
2
 2 (0) (1)
0
∫ ∫2 S
α3
+ 2
0 δ E α3 + θ 3
δ 2
E α 3  µ dθ
3 0
dΩ
0
Ω − h2  
Introducing the n-th order effective stress resultants obtained by the through-the-
thickness pre-integration of the stress components
h
( )
(n) 2
n0
L = ∫ S ij θ 3 µ dθ 3 ,
2 ij 2
0 0 (3.99)
− 2h
the virtual internal work, 2δWi, can be expressed as the area integral over the middle surface
in the initial configuration 0C.
The incremental decomposition of the stress tensor in the configuration 2C can be
expressed for the stress resultants as
(n) (n) (n)
0L = 01 Lij + 0 Lij .
2 ij
(3.101)
Remark 3.3: The effective stress resultants defined by (3.99) are based on the second Piola–
Kirchhoff stress tensor referred to the initial configuration 0C, therefore they lack a clear
physical interpretation; they can be merely considered as force variables energetically
conjugate with the corresponding strain components. Much more realistic are stress
resultants constructed by the integration of the Cauchy stress in the current configuration
2
C:
h
σ ij (θ 3 ) 2 µ dθ 3 .
(n) 2
n
L = ∫
2 ij 2
2 (3.102)
− h2
Commonly the zeroth-order resultants (i.e. those, that are obtained by putting n=0 in
(3.102)) are interpreted as membrane73 and shear forces as portrayed in Fig. 3.5.
q
3
q
3
q 2
q 1
q
2 (0)
s
22
L22
s
11 (0)
11
L
q 1
q 3
q 3
q 2
q
1
(0)
q
2
s 21
L21
s
12
(0)
L12
q
1
q 3
q 3
q
2
q
1
q (0)
2
L23
s
23
(0)
s
13 L13
q 1
Fig. 3.5: Zeroth-order stress resultants
Consequently, the first order stress resultants obtained from (3.102) for n=1 can be treated
as effective membrane and shear moments (compare illustration in Fig. 3.6).
A physical meaning of the second-order membrane stress resultants (Fig. 3.7) is a little
more questionable. Some authors prefer to call them the bimoments (compare Bischoff &
Ramm [60], also Krätzig & Jun [255-257]); however, such interpretation would be more
73
Strictly speaking, such membrane forces correspond to the symmetric pseudo-stress resultant tensors; compare
e.g. Başar & Krätzig [43]; for a more detailed consideration on the physical interpretation of those force variables
see also Başar [34].
justified when quadratic varied stress terms are considered together with a portion of the
stress that is through-the-thickness constant as presented schematically in Fig. 3.8.
q
3
q
3
q 2
s
22
q 1
q 2
(1)
s
11
L22
(1 )
11
L
q 1
q
3
q
3
q 2
(1)
q
1
s 21 L21
q
2
s 12
(1)
L12
q
1
q
3
q
3
q
2
q 1
(1)
q 2 L23
s
23 (1 )
13
s L
13
q
1
Fig. 3.6: First-order stress resultants
q 3
q
3
q
2
s
22
q 1
(2)
q 2
L22
s 11 (2)
11
L
q
1
q 3
q
3
q 2
q 1
s 12
s 21 q 2
(2)
(2)
L12 L21
q
1
Fig. 3.7: Second-order membrane stress resultants

+ = = +
Fig. 3.8: Bi-moment interpretation of the second order stress resultants
3.6. Constitutive relations

A typical composite shell made of orthotropic fiber-reinforced material can be
analyzed as a layered structure, with the fibers of the reinforcement in each lamina placed
in the surfaces parallel to the shell mid-surface. An orthogonal system of the material axes
(a, b, c) in each lamina is introduced in such a way that the a-axis is aligned with the
direction of the reinforcement whereas the c-axis is normal to the shell mid-surface (Fig.
3.9)
c
a
Fig. 3.9: System of material axes for fiber reinforced composite
A linear elastic 3-D orthotropic material is characterized by 9 independent engineering

material constants: Ea, Eb, Ec, νab, νbc, νac, Gab, Gbc, and Gac (see e.g. Green & Zerna [178],
Jones [231], and Vinson & Chou [476]). Assuming zero value of the transverse normal
stress component74 (in the c-axis direction) the 3-D incremental constitutive relation for an
individual lamina in the material axes can be presented as follows:
 0 S aa   0 ε aa 
 bb   ε 
 0 S   0 bb  (3.103)
0  S ab
= [C ]
m 2 0 ε ab ,
 S bc  2 ε 
0   0 bc 
 0 S ca  2 0 ε ca 
with the constitutive matrix
74
Please notice, that such assumption formally disagrees with the director inextensibility, which was however
postulated only for the shell kinematics (cf. Section 3.4).
 Ea ν ab Ea 
1 −ν ν 0 0 0 
1 −ν abν ba
 ab ba

 ν Ea Eb
ab
0 0 0  . (3.104)
[C m ] = 1 −ν abν ba 1 −ν abν ba 
 0 0 Gab 0 0 
 
 0 0 0 κGbc 0 
 0 0 0 0 κGca 

The set of well known engineering constants (Ei, νij and Gij)75 in (3.104) is supplemented by
the shear correction factor κ, which can be taken as equal to 5/6 after Reissner (see also
Chapter 2).
The transformation of stress and strain vectors between the material axes (a, b, c) and
1 2 3
the coordinate system (θ , θ , θ ) can be illustrated by the relations
 0 ε aa   0 ε 11   0 S 11   0 S aa 
 ε   ε   22   bb 
 0 bb   0 22   0 S   0 S  (3.105)
2 0 ε ab  = [T ]  2 0 ε 12   0 S  = [T ]
12 T ab
and 0 S 
2 ε  2 ε   S 23   S bc 
 0 bc   0 23  0  0 
 2 0 ε ca  2 0 ε 31   0 S   0 S 
31 ca
with transformation matrix

 cos 2 (α k ) sin 2 (α k ) 12 sin( 2α k ) 0 0 
 
 sin (α k ) cos (α k ) − 2 sin( 2α k )
2 2 1
0 0 , (3.106)
[T ] = − sin(2α k ) sin(2α k ) cos(2α k ) 0 0 
 
 0 0 0 cos(α k ) − sin(α k ) 
 0 0 0 sin(α k ) cos(α k ) 

1
where the ply orientation angle, αk, is measured between the θ -axis and the a-axis as
shown in Fig. 3.10.76
q
3
c
b
reinforcement
fibers q
2
ak
q
1
a
Fig. 3.10: Ply orientation angle
75
Note that νab Eb= νba Ea (no summation over repeated indices).
76 1 2 3
A general curvilinear coordinate system (θ , θ , θ ) was used until now in a description of the shell geometry;
1 2
however, a coordinate system with perpendicular axes θ and θ better fits an orthogonal system of reinforcements.
3.6. Constitutive relations 79
After such a transformation the incremental constitutive relation for an individual lamina in
1 2 3
the coordinate system (θ , θ , θ ) reads:
 0 S 11   0 ε11 
 22   ε 
 0 S   0 22  (3.107)
0S 
12
= [C ]  2 0 ε12 ,
 S 23  2 ε 
0   0 23 
 0 S 31   2 0 ε 31 
where
c11 c12 c13 0 0
c c22 c23 0 0 
 12 (3.108)
[C] = [T ]T [C]m [T ] = c13 c23 c33 0 0 .
 
0 0 0 c44 c45 
 0 0 0 c45 c44 
Performing the pre-integration of the 3-D constitutive relations through the thickness of the
whole shell one obtains the 2-D constitutive relation
{ S} = [H ]{ E}.
0 0
(3.109)
Introducing matrix notation one can put the stress and strain components in the vector
form77
 {0 N }  {¤ 0

}
{ M}
 0 

 {¤ 0

}
{0 S} =  {0 B}  and {0 E} =  {¤ 0

} (3.110)
 { Q} 
 0 


{ 0

}
 {0 P}   { 0

}
with sub-vectors
 ( 011
)
  (1)11   ( 211
)

0 L 0L  0 L
 (0)   (1)   ( 2)  (3.111)
{0 N } =  0 L22 , {0 M} =  0 L22 , {0 B} =  0 L22 ,
 ( 012
)   (1)12   ( 212
) 
0L  0L  0L 

     
 ( 0)23   (1)23 
{0 Q} =  0 (L0) , {0 P} =  0 (L1)  . (3.112)
 0 L31   0 L31 
77
One should notice that the introduced constitutive relations are constructed in a quite general way to match the
different non-linear models considered in this report: from the LRT56, through LRT5, MRT5 and RVK5 to a pure
linear formulation.
 (k ) 
 0 ε 11   (i ) 
 (k ) 
{¤ }
0
k
=  0 ε 22 , k = 0, 1, 2; { }
0
i 2 ε 
=  0 (i )23 , i = 0, 1.
(3.113)
 (k )  2 0 ε 31 
2 0 ε 12 
 
With the substitutions (3.110-113) a more detailed form of the constitutive relation (3.109)
can be presented as
 {0 N }  [A]3 x 3 [B ]3 x3 [D ]3 x3 [0]3 x 2 [0]3 x 2  {0 ¤ }
{ M}  [B ]
 0   3 x 3
[D ]3 x3 [E ]3 x3 [0]3 x 2 [0]3 x 2  {0 ¤ }
 0  = [D ]3 x 3
{ B} [E ]3 x3 [F ]3 x3 [0]3 x 2 [0]3 x 2  {0 ¤ }, (3.114)
 { Q}   [0] [0]2 x3 [0]2 x3 [S A ]2 x 2 [S B ]2 x 2  {0 }
 0   2 x3
 {0 P}   [0]2 x 3 [0]2 x3 [0]2 x3 [S B ]2 x 2 [S D ]2 x 2  {0 }
where
 a11 a12 a13  h
a a45 
[A]3 x3 = a12 a23 , [S ]
2
a22 A 2x2
=  44  , aij = ∫ cij 0µ dθ 3 , . (3.115)
a13 a23 a33  a45 a55  −h 2
b11 b12 b13  h

b b45 
[B ]3 x3 = b12 b23 , [S ]
2
b22 B 2x2
=  44  , bij = ∫ cij θ 3 0µ dθ 3 , . (3.116)
b13 b23 b33  b45 b55  −h 2
 d11 d12 d13  h

d d 45 
[D]3 x3 = d12 d 23 , [ ] ( )
2
µ dθ 3 , . (3.117)
2 0
d 22 S D 2 x 2 =  44  , dij = ∫ cij θ 3
 d13 d 23 d 33  d 45 d 55  −h 2
e11 e12 e13  h
[E ]3 x3 = e12 e22 e23 , ( )

2
eij = ∫ cij θ 3
3 0
µ dθ 3 , (3.118)
e13 e23 e33  − h2
 f11 f12 f13  h
[F ]3 x3 =  f12 f 23 , ( )
2
f 22 f ij = ∫ cij θ 3
4 0
µ dθ 3 . (3.119)
 f13 f 23 f 33  − 2h
Remark 3.4: It is worth to notice that some authors prefer to introduce the shear correction
factors directly to the pre-integrated constitutive relation (3.109) or (3.114). Usually the
part of the constitutive matrix related to the transverse shear is reduced to the sub-matrix SA
only; then the correction proposed by Dong & Tso [146] (see also Whitney [492] and
Vlachoutsis [477]) looks as follows:
3.6. Constitutive relations 81
( k ) 2 a .
[S( ] A 2x2 =  23 44
a45
2  (3.120)
 a45 (k13 ) a55 
A little different transverse shear correction was proposed by Noor & Peters [331] (see also
Zhang & Kim [509])
 (k ) 2 a ( k 23 k13 ) a45  .
[S( ]
A 2x2 =  23 44
(k13 ) 2 a55 
 (3.121)
( k 23 k13 ) a45
Using (3.101) and (3.109) one can write a matrix form of the internal virtual work
(3.100)
2
0
(
δWi = ∫ {δ 20E}
T
{ S}+ {δ E} [H ]{ E}) dΩ .
1
0
2
0
T
0
0
(3.122)
Ω
Introducing the index notation an alternative form of the above equation can be presented
as
 2 (n)
 (n) 2
 ( n ,m ) (m)  1 (i )
 (i ) 1
 (i, j ) ( j)  
δWi = ∫ ∑ δ 2 Eαβ  01Lαβ + ∑  H αβχδ 0 E χδ  + 2∑ δ 2 Eς 3  01Lς 3 + 2∑  Ξςη 0 Eη 3  0 dΩ
2
0
Ω  n =0  m =0   i =0  j =0  
(3.123)
( n,m ) (i, j )
αβχδ ςη
with H and Ξ representing appropriate components of the constitutive matrix [H].
The virtual work principle established above (3.122 and 3.123) for the assumed ESL
material representation can be used to build the incremental equilibrium equations after
introducing the discretization of the Finite Element Method in the next chapter.
Chapter 4
FINITE ELEMENT METHOD IMPLEMENTATION
4.1. Finite element discretization of the problem

m
The shell geometry in the configuration C is defined in the FOSD model by the
m
position vector R (compare (3.41)). Assuming the isoparametric finite element
formulation the position vector can be interpolated inside each finite element as
( ) ∑ (N )
NNE NNE
m
R (r , s ) = ∑ N k (r , s) m r k + θ 3 k (r , s) m d k , (4.1)
k =1 k =1
m k
where r is the position vector of the node k at the shell mid-surface; r and s stand for the
element natural coordinates; Nk represents the shape function associated with the node k,
and NNE means the number of nodes of the element.
Probably the most popular quadrilateral surface finite elements are those based on the
Lagrange (Fig. 4.1) or Serendipity (Fig. 4.2) interpolation schemes. The detailed
prescription of interpolation functions for those elements can be found in most FEM
handbooks (see e.g. Bathe [46] or Zienkiewicz & Taylor [510]).
a) b) c)
4-node 9-node 16-node
Fig. 4.1: Lagrange family of shell elements: a) linear; b) quadratic; c) cubic interpolation
a) b) c)
4-node 8-node 12-node
Fig. 4.2: Serendipity family of shell elements: a) linear; b) quadratic; c) cubic interpolation
m m
The displacement vector V (3.48) in the configuration C can be interpolated in an
analogous way as the geometry:
NNE
 (0)
 NNE
 (1)

m
V (r , s ) = ∑  N k (r , s ) m V k  + θ 3 ∑  N k (r , s ) m V k  . (4.2)
k =1   k =1  
4.1. Finite element discretization of the problem 83
For the sake of the simplified notation in the following derivations a specific
substitution has been introduced
(0)
m
uM = mυi for M = i = 1, 2, 3, (4.3)
(1)
m
uN = υ jm
for N − 3 = j = 1, 2, 3.
One can apply (4.3) to construct a quite general form of the strain-displacement relations
given in (3.53), (3.56) and (3.62)
m
0 Eij = BijM mu M + 12 GikP G jkQ mu P muQ , (4.4)
where BijM and GkjP stand for two linear operators.

m
With (4.2) the displacement components uI for the LRT56 model based on the relation
(3.66) can be expressed by the nodal displacement parameters as
∑ {N }
NNE
m
u M (r , s) = k (r , s ) m q Mk for M = 1, 2, 3,
k =1
∑ {N }
NNE
m
u 4 (r , s ) = k (r , s) sin( m q 4k ) cos( m q5k ) , (4.5)
k =1
∑ {N }
NNE
m
u 5 (r , s ) = k ( r , s ) sin( m q5k ) ,
k =1
∑ {N }
NNE
m
u 6 (r , s) = k (r , s ) cos( m q 4k ) cos( m q5k ) − 1,
k =1
m k
where qI represents the corresponding displacement parameters at the node k. One should
note that due to the use of the exact formula (3.66) the six displacement parameters are
expressed in (4.5) as non-linear functions of five nodal parameters, while the use of the
simplified variant (3.68) would lead to linear relations only.
The derivatives of the strain tensor components with respect to the nodal displacement
1
parameters at the actual configuration, C, will be calculated next as
∂Eij ∂u M ∂u
= BijM + GikP G jkM 1u P M , (4.6)
∂qR ∂qR ∂qR
∂ 2 Eij
( ) ∂q∂ u∂q ∂u P ∂u M ,
2
= BijM + GikP G jkM 1u P M
+ GikP G jkM (4.7)
∂qR ∂qQ R Q ∂qQ ∂qR
At this stage one should notice that due to the non-linear (trigonometric) relation
introduced in (3.66), the second derivative of the displacement with respect to the nodal
displacement parameter does not vanish, in contrast to the simplified formulation LRT5
where the second derivative of the linear relation (3.68) is zero.
The derivatives given above are used in the linearized expression for the variation of the
strain tensor components
∂Eij ∂ 2 ( Eij )
δ ( 02Eij ) ≅ δqS + δqS ∆qT , (4.8)
∂qS ∂qS ∂qT
and for the strain increment
84 Chapter 4. Finite Element Method implementation
∂Eij
∆Eij ≅ ∆q S . (4.9)
∂qS
4.2. Incremental equilibrium equations of FE model

By introducing the relations (4.6)-(4.9) into (3.123), the internal virtual work
expression for the FEM model of the shell can be presented as
 
2 1 1 (U )
(
1 (G ) 1 ( II )
1442443
) 
δWi = δqS  0 FS + 0 K ST + 0 K ST ∆qT + 0 K STR ∆qT ∆qR  . (4.10)
 JS 
 
According to the commonly accepted FEM terminology the particular entities of the above
equation can be named as follows:
• 01FS stands for the components of the balanced forces vector, 01 F , { }
 ∂Eαβ ∂Eδ 3 1 δ 3  0
∫  ∂q
1 αβ
1
0 FS = L
0 +2 
0 L  dΩ =
0
Ω S ∂q S 
(4.11)
 2  (n) (n)  1 
(i )

 ∂ Eαβ 1 αβ   ∂ Eδ 3 1 (i )δ 3  0
= ∫ ∑  0 L  ∑
+ 2 0 L  dΩ,
Ω n =0 
∂q S  i = 0  ∂q S 
  
0
 
• 1
0K (U ) ST represents the first part of the incremental stiffness matrix containing the
constitutive stiffness term and the initial displacement stiffness term, 01 K u , [ ]
 2 2  ( n ) ( n ,m ) (m)
  (i ) ( i , j ) ( j ) 
  ∂ Eαβ αβχδ ∂ E χδ  1 1
 ∂ Eς 3 ςη ∂ Eη 3  0 (4.12)
1
0 K (U )
ST = ∫ ∑∑  H  + 4∑∑  Ξ  dΩ,
Ω n =0 m =0 
∂q S ∂qT  i =0 j =0  ∂q ∂q T 
 
0 S
  
• 1
0 K ( G ) ST denotes the components of the geometrical stiffness matrix, [ K ],
1
0 g
 ∂ 2 ( Eαβ ) 1 αβ ∂ 2 ( Eδ 3 ) 1 δ 3  0
K ( G ) ST = ∫0  ∂q S ∂qT 0 L + 2 ∂q S ∂qT 0 L  dΩ =
1
0
Ω  (4.13)
 2  2 (n)
)  1  2
(i )

  ∂ ( Eαβ ) 1 ( nαβ   ∂ ( Eδ 3 ) 1 (i )δ 3  0
= ∫ ∑  0L  + 2∑  0 L   dΩ,
Ω n =0 
∂q ∂q  i = 0  ∂q S ∂qT 
  S T 
0
 
• 1
0K ( II ) STR symbolizes the additional object, which according to Kleiber [244] one can
call “the second-order stiffness matrix”; the product ( 01 K ( II ) STR ∆qT ∆q R ) can be
understood as a component of the vector {J( 1 q, ∆q )} containing additional terms,
which are non-linear with respect to the displacement increments {∆q}
4.2. Incremental equilibrium equations of FE model 85
 2 2  2 ( n ) ( n ,m ) ( m)
 1  2
(i ) ( j)

  ∂ ( Eαβ ) αβχδ ∂ E χδ  1
 ∂ ( Eδ 3 ) ( i ,δηj ) ∂ Eη 3  0
1
K ( II )
= ∫ ∑∑  H + 4 ∑∑ Ξ dΩ.
∂qR   ∂qR 
STR
0  ∂qS ∂qT i = 0 j = 0  ∂qS ∂qT
Ω n =0 m =0 
 
0
  
(4.14)
An alternative, matrix form of (4.10) reads
δWi = {δq}T ({01 F}+ ([01 K u ]+ [01 K g ]) {∆q}+ {J( 1 q , ∆q )}).
2
(4.15)
By denoting as { R}
2
0
the vector of nodal forces resulted from the loads acting in the
2
configuration C, one can write a very general form of the external virtual work
δWe = {δq}T {20 R}.
2
(4.16)
With (4.15) and (4.16) one can obtain from (3.91) the incremental equilibrium equations as
([ K ]+ [ K ]) {∆q} = { R}− { F}− {J( q, ∆q )}.
1
0 u
1
0 g
2
0
1
0
1
(4.17)
The obtained relation represents a standard form of the incremental equation of the
quasi-static motion of the structure and most of the terms have been efficiently explained
above. However, the geometrical stiffness matrix (4.13) needs an additional comment. With
(4.7) the expression for the components 01K (G ) ST can be presented in a more detailed form
as
 ∂u P ∂u M 1 αβ ∂u P ∂u M 1 α 3  0
∫0  GαkP G βkM + 2 GαkP G3kM  dΩ +
1
0 K (G ) ST = 0L 0L
∂q S ∂qT ∂q S ∂qT
14444444444444444444444  4443
Ω
4244444444444444444444
1 ( G1)
0 K ST
 ∂ 2 u M 1 αβ ∂ 2 uM 1 α 3  0
+ (
∫  αβM αkP βkM P ∂q S ∂qT 0
 B + G G 1
u L + 2 B )
α 3M + GαkP G 3 kM
1
u P (
∂q ∂qT
0L
 dΩ , )
0
1 Ω
44444444444444444244444444444S 44 444  43
1 (G 2)
0 K ST
(4.18)
where the geometrical stiffness matrix is split into two parts. The first part of the
geometrical stiffness matrix, 01K ( G1) ST , is a “regular” geometrical stiffness matrix that is
present in every standard TL formulation. However, the additional part, denoted as
ST , results from a non-zero value of the second displacement derivative taken with
1 (G 2)
0K
respect to the nodal displacement parameters, which appears in the first term of (4.7). As it
was shown earlier, the non-zero value of that derivative is characteristic for the LRT56
formulation utilizing the trigonometric relation (3.66), while for the LRT5 formulation
based on the linearized relation (3.68) this derivative vanishes and, as a consequence, the
ST part does not appear in LRT5. Up to the author knowledge, the essential role of
1 (G 2)
0K
that additional part of the geometrical stiffness matrix was indicated for the first time in
1977 by Frey & Cescotto [170] for the non-linear analysis of beams.78
78
Surana [451] introduced such additional geometric stiffness matrix for axisymmetric shells, Ramm &
Matzenmiller [383] used it for the degenerated shell elements; a generalized formulation for finite elements with
rotational degrees of freedom was presented by Surana [452].
A standard aggregation of equilibrium equations (4.17) for all finite elements results in
the following incremental equilibrium equation for the whole FE model:
1
K ∆u = 2 R − 1 F − J ( 1 u , ∆u ) , (4.19)
1
where ∆u is the global vector of displacement increments, K denotes the global
2 2 1
incremental stiffness matrix, R represents the global load vector in configuration C, F
1
stands for the global vector of balanced forces in the actual configuration C.
4.3. Solving of incremental equilibrium equations

Load control incremental analysis
A pure incremental solution of the non-linear problem can be obtained by applying a
standard recurrence procedure based on the approximated version of (4.19):
( n −1)
K ∆u = n R − ( n −1) F , (4.20a)
n
u= ( n −1)
u + ∆u = ( n −1)
u+ ( ( n −1)
K ) ( R−
−1 n ( n −1)
F, ) (4.20b)
where ∆u represents the increment of displacements; (n-1)K and (n-1)F stand, respectively,
for the tangent stiffness matrix and the vector of balanced forces, both resulted from the
state of displacements estimated in the previous increment (n-1)u.
1
Due to omitting the non-linear term J( u, ∆u) in (4.20), the error of approximation in
(4.20a) can accumulate after several increments and the obtained solution (4.20b) can
considerably drift away from the true response. To improve the accuracy of the incremental
solution an iteration process of the standard Newton-Raphson method can be employed in
every single increment as:
K ( 2 u ( i −1) ) δu ( i ) = 2 R − F( 2 u ( i −1) ) , (4.21a)
∆u ( i ) = ∆u (i −1) + δu (i ) , (4.21b)
2
u (i ) = 2 u (i −1) + δu (i ) , (4.21c)
for i = 1, 2, ... where 2u(0) = 1u and ∆u(0) = 0. The equilibrium iterations (4.21) should be
performed until the desired accuracy of solution can be obtained. The convergence criterion
used in the presented algorithm can be described by the relation
δu (i ) < ε ∆u (i ) , (4.22)
where || . || stands for the Euclidean norm in the displacement space, and ε is the prescribed
tolerance.
As it was already mentioned in Section 2.4, the application of the load-control method is
practically limited only for a pre-buckling region. Therefore, to extend the field of application
for problems with a more complex equilibrium paths (i.e. those possessing critical points) one
should apply a more universal type of incremental control as discussed in Section 2.4 of the
present report.
4.3. Solving of incremental equilibrium equations 87
Riks-Wempner-Ramm arc-length control method

The arc-length control method, commonly known as the Riks-Wempner method,
allows for tracing quite complex equilibrium paths by applying a generalized arc-length
control parameter obtained as an appropriate combination of the load and displacement
incremental control parameters. Many variants of this method were described in the
literature: see e.g. Crisfield [129], Kleiber [244], Waszczyszyn [485], and Zienkiewicz &
Taylor [510]. In 1980 Ramm [379, 381] proposed his own variant of the arc-length control
method that later on became popular in the literature under the name of
Riks-Wempner-Ramm method.
Assuming a proportional loading option, one can write
2
R = 2λ R ref , (4.23)
where 2λ stands for the load parameter in configuration 2C, and Rref represents the global
reference loads vector.
With the use of (4.23), equation (4.21a) can be transformed into
K (1 u ) ∆u ( 0 ) = 2λ R ref − F(1 u ) , (4.24)
and
K ( 2 u ( i −1) ) δu ( i ) = 2λ R ref − F( 2 u ( i −1) ) =
(4.25)
= δλ( i ) R ref + J( 2 u ( i −1) ).
The first approximation of the load parameter increment ∆λ(0) is limited by the constraint
equation
(∆u ) (0) T
(
∆u ( 0 ) + ∆λ( 0 ) )2
= ds 2 , (4.26)
which can be considered as forcing the general arc-length on the equilibrium path to be
equal to the prescribed value ds. The point of the exact solution (∆u, ∆λ) on the equilibrium
path (see Fig. 4.3) is pointed out by the sum of the tangent vector
r
t ( 0 ) = [∆u ( 0 ) , ∆λ( 0 ) ] (4.27)
and the correction vector

v
δ∆ = [δu, δλ ] . (4.28)
Assuming that both those vectors are mutually perpendicular, their scalar product has to
fulfill the relation
r v
t ( 0 ) ⋅ δ∆ = 0, (4.29)
or
(∆u ) δu + ∆λ
(0) T ( 0)
δλ = 0 . (4.30)
Practically, the correction of displacement increment is calculated in the iterative process

v
with successive corrections δ∆( i ) = [δu (i ) , δλ(i ) ] ; therefore, the relation (4.30) is replaced
with
(∆u ) δu(0) T (i )
+ ∆λ( 0 ) δλ( i ) = 0 . (4.31)
In the original Riks-Wempner method (see e.g. Waszczyszyn [485]), the constraint
equation (4.31) is supplemented to the system of equations (4.25), what results in a non-
symmetric and non-banded structure of the extended global stiffness matrix. Hence the
special, non-standard, procedure for solving the system of equations should be employed.
The Ramm variant avoids these difficulties by utilizing the following two-steps procedure.
It is assumed, that the unknown correction of the displacement increment δu (i ) can be
expressed by
δu ( i ) = δλ( i )δu R ( i ) + δu J ( i ) (4.32)
with δu R (i ) and δu J (i ) obtainable from the equations
K ( 2 u ( i −1) ) δu R
(i )
= R ref , (4.33)
( i −1)
K(2 u J ) δu ( i ) = J( 2 u ( i −1) ). (4.34)
l
(0)
t
l
2 (0)
(1)
dD
t
2
l
(1)
dl
(0)
t
l Dl
2 (2) (0)
(2)
2
l
(3) t
l
2
(3)
t
l
1
D u d u
(0)
2 (0) 2 (1) 2 (2) 2 (3)

u u u u
1 2
u u u
Fig. 4.3: Arc-length control method with constant direction of searching
By substituting the relation (4.32) into the constraint equation (4.31) one can calculate the
correction of the load parameter increment as
δλ(i ) = −
(∆u ) δu ( 0) T
J
(i )
. (4.35)
(∆u ) δu + ∆λ
(0) T
R
(i ) (0)
4.3. Solving of incremental equilibrium equations 89
v
The vector δ∆( i ) = [δu (i ) , δλ(i ) ] indicates the point in which the hyper-plane normal to the
r
tangent vector t ( 0 ) is punctured by the new tangent line t(i) (Fig. 4.3). Such approach
requires the vector ∆u(0) to be stored in the computer memory during the whole iteration
process at one increment. The alternative solution implemented in this report is to use a
new, updated hyper-plane in each new iteration (see Fig. 4.4); then the constraint equation
takes the form of
(∆u ) δu( i −1) T (i )

+ ∆λ( i −1) δλ( i ) = 0 (4.36)
with
∆λ( i ) = ∆λ( i −1) + δλ( i ) , (4.37)
and the formula (4.35) for the correction of the load parameter increment is replaced with
δλ
(i )
=−
(∆u ) δu ( i −1) T
J
(i )
. (4.38)
(∆u ) δu + ∆λ( i −1) T
R
(i ) ( i −1)
The application on the updated hyper-planes is similar to the idea of searching for the
solution on the sphere as suggested by Crisfield [129].
(0)
t
l
2 (0)
(1)
t
l
2 (1)
l
2 (2)
l
2
t(2)
l
1
2 (0) 2 (1) 2 (2)

u u u
1 2
u u u
Fig. 4.4: Arc-length control method with updated hyper-planes
To improve the effectiveness of the algorithm the value of the control parameter can be
updated according to the number of iterations necessary to achieve the convergence in the
previous increment
NDIT
∆λ( 0 ) ( new) = ∆λ( 0 ) ( old ) , (4.39)
NITE
where NDIT is the desired number of iterations in one increment, and NITE stands for the
number of iterations in the last increment. When the convergence is of the oscillatory
nature, i.e. if
δλ(i −1) δλ(i ) < 0 , (4.40)
with
δλ(i ) < δλ(i −1) , (4.41)
then the algorithm can be improved by the application of the relaxation coefficient α = 0.5
used in the calculation of the load parameter and displacement increments as follows:
δλ(i ) ( new ) = α δλ(i ) ( old ) , (4.42)
(
δu (i ) ( new ) = α δλ(i )δu R (i ) + δu J (i ) . ) (4.43)
A very important aspect related to the Ramm version of the Riks-Wempner method is
to trace the sign of the stiffness matrix determinant, and to begin the unloading process
every time when the coefficient matrix is negative defined. Practically, it is enough to count
the number of negative entries on the diagonal of the decomposed tangent stiffness matrix,
what can be done during the solving the system of equations. The sign of the load
parameter increment should be changed for the odd number of negative entries. Naturally,
there is no need to implement such procedure when the Riks-Wempner method is used in
its classical version with the extended matrix of coefficients (compare Waszczyszyn [485]).
4.4. Finite elements used in this study
Isoparametric shell elements with an independent interpolation of translational and

rotational displacement components have many advantages, like relatively simple structure
and easy satisfaction of all continuity requirements. However, when the shell becomes thin,
their rate of convergence (with an increasing number of elements) is too slow for effective
engineering analysis, due to the commonly known locking phenomenon as discussed already
in Section 2.4 of the present study. One popular remedy for that difficulty is an application of
the reduced integration technique, where the numerical integration of the element stiffness
matrix is performed with the lower order of the quadrature rule. It was discovered quite early
that the 4-node element with bi-linear interpolation was not able to provide correct results
when used with a reduced integration, therefore elements with higher order of interpolation
have been selected for an application in the present report. There have been three types of
elements considered: the 8-node Serendipity element with a bi-quadratic interpolation and
Lagrange elements with bi-quadratic (9-node) and bi-cubic (16-node) interpolations. For both
elements with a bi-quadratic interpolation (8-node and 9-node), the 9-point (3×3)
Gauss-Legendre numerical integration scheme is used for the full (exact) integration, while
the use of 4 (2×2) points means the reduced integration. The 16-node element requires 16-
point (4×4) numerical integration scheme for the full integration and 9-point (3×3) for a
4.4. Finite element used in this study 91
reduced one (see e.g. Bathe [46] or Zienkiewicz & Taylor [510]). From the contents of the
previous chapter one can conclude that the separation of the transverse shear terms from the
bending and membrane terms can be easy performed in a computer implementation of the
proposed algorithm. The application of the reduced integration only to transverse shear terms
with full integration kept for remaining terms resulted in the selective integration scheme. A
combination of different integration and interpolation schemes resulted in nine types of shell
elements listed in Table 4.1.
Table 4.1: Shell elements with various integration schemes

Integration scheme
Element Interpolation bending &
transverse shear
membrane
s
8-FI 3× 3 3×3
8-SRI r 3×3 2×2
8-URI 2×2 2×2
s
9-FI 3×3 3×3
9-SRI r 3×3 2×2
9-URI 2×2 2×2
s
16-FI 4×4 4×4
16-SRI r 4×4 3×3
16-URI 3×3 3×3
It was already mentioned in Chapter 2 that the reduced integration technique is not a
universal tool - in many cases the spurious zero-energy mechanisms, due to the lowered rank
of the elements stiffness matrices, can manifest in a singular solution (depending on the
boundary conditions and an applied mesh of elements). According to many sources, a better
way to avoid the locking phenomenon seems to be the Assumed Natural Strain approach
(compare the survey presented in Section 2.4). The formulation of the shear and membrane
locking-free 9-node shell element based on the Assumed Natural Strain approach (9-ANS)
as proposed by Park & Stanley [353] (see also Park et al. [352]) was adopted in the present
report as an alternative choice. The essence of this technique lies in the "improved"
interpolation of natural-coordinate strain components; the assumed strain components at
integration points are re-interpolated on the base of their values at some special sampling
points. The sampling points for each strain component are located at so called Barlow
points, where considered component obtained from the standard interpolation scheme
approaches its best approximation. The locations of the sampling points used in this study
for the 9-ANS element are given in Table 4.2, where a = (3)-1/2. The essence of this
technique lies in the "improved" interpolation of natural-coordinate strain components based
on the assumed strain components at integration points. The full integration (3×3) scheme is
used for the 9-ANS element what provides a proper rank of the stiffness matrix.
Table 4.2: Interpolation schemes for various strain components in 9-ANS element
Assumed strain interpolation
Components εrr and εr3 Components εss and εs3 Component εrs is linear
linear in r and quadratic in s quadratic in r and linear in s both, in r and in s
s s s
a
a
r r r
a
a
a a a a
- element nodal points - sample points of strain interpolation
4.5. Computer implementation of the proposed FEA algorithm
A family of computer programs incorporating the proposed FEA formulation for the
linear and non-linear problems of multi-layered shells has been written in FORTRAN. The
organization of the developed computer programs follows the standards introduced for the
FEM algorithms by Bathe [46].79 In particular, the storage system of the stiffness matrix
basing on the sky-line scheme has been applied with appropriate procedures for the
aggregation and for solving of the system of equations. Quadrilateral isoparametric finite
elements with 4, 8, 9 and 16 nodes are available for the analysis of plates and shells. The
non-linear equilibrium paths can be traced within the incremental analysis based on the
arc-length control method (Ramm [379, 381]) with the Newton-Raphson iterations.
79
See also earlier publications of Bathe et al.: Bathe, K.-J., Wilson, E.L. and Iding, R.H., NONSAP - A Structural
Analysis Program for Static and Dynamic Response of Nonlinear Systems, SESM Report 74-3, Department of
Civil Engineering, University of California, Berkeley 1974; Bathe, K.-J. & Wilson, E. L., Numerical Methods in
Finite Element Analysis, Prentice-Hall, Englewood Cliffs, New Jersey, 1976; Bathe, K. -J., Finite Element
Procedures in Engineering Analysis, Prentice Hall, Englewood Cliffs, New Jersey, 1982.
4.6. Assessment of selected elements in linear analysis 93
4.6. Assessment of selected elements in linear analysis

Pinched cylinder with end diaphragms
A pinched cylinder with end diaphragms as presented in Fig. 4.5 is known in the
literature as one of the most demanding benchmark tests for isotropic cylindrical shells (see
e.g. Bathe & Bolourchi [47], Dvorkin & Bathe [152], Bathe & Dvorkin [48], Belytschko et
al. [54], Heppler & Hansen [194], Chang et al. [103], Simo et al. [436], Cheung & Chen
[111], Groenwold & Stander [180], Kreja et al. [269], Chinosi et al. [112], Hauptmann &
Schweizerhof [192], Masud & Panahandeh [308], Sze et al. [456], Reddy & Arciniega
[399]). Due to the symmetry of the problem, only one octant of the cylinder is modeled
with a regular mesh of shell elements. For the geometrical and material data presented in
Fig. 4.580, the analytical solutions for selected displacement components can be estimated
after Heppler & Hansen [194] as given in Table 4.3.
The first tests were performed for the 8- and 9-node elements presented in the previous
section.
Rigid A
diaphragm
B D
wB uD L
P C P
wC
L R = 300 in
R
L = 300 in
h = 3 in
6
E = 3•10 psi
Rigid n = 0.3
diaphragm P = 1 lb
h
Fig. 4.5: Pinched cylinder with end diaphragms
Table 4.3: Analytically estimated displacements of pinched cylinder with end diaphragms
Displacement component Analytical result [194]
radial deflections at point C wCref = 1.8248 × 10-5 in
radial deflections at point B wCref = 5.2222 × 10-8 in
axial displacement at point D uDref = 4.5711 × 10-7 in
80
To keep the compatibility with the reference solutions a non-metrical US system units have been used in this
example (1 in = 25.3995 mm; 1 lb = 4.44822 N; 1 psi = 6.895 kPa).
The results of the normalized radial deflection under the point load, wC/wCref, obtained
with an increasing number of elements are presented in Fig. 4.6.
1.2
Normailzed radial deflection wC
0.8
9-FI
0.4 9-SRI
9-URI
8-FI
8-SRI
8-URI
9-ANS
0.0
3 5 7 9 11 13 15 17 19 21
Number of nodes per edge
Fig. 4.6: Convergence study for radial deflection at the point C
Comparing the results for uniformly reduced (URI), selective reduced (SRI) and full
integrations (FI) one can notice a considerable influence of the shear locking in this
example. On the other hand, the influence of the membrane locking decreases with the use
of more refined meshes. It should be noted that the 9-URI element yields a rapid
convergence of the solution, which, however, slightly differs from the reference solution
[194] for the most refined meshes. The performance of the 9-ANS element is also very
good; whereas the convergence obtained for the 8-URI element is considerably slower and
only a little better than those of 9-SRI and 8-SRI elements. Results obtained with the full
integration (8-FI and 9-FI) manifest a significant locking.
The convergence study of axial displacement uD shown in Fig. 4.7 seems to confirm
the superiority of the 9-URI element, but the performance of the 9-ANS and 8-URI
elements is also very good here.
To verify above observations the convergence for the radial deflection of the point B,
at 90° to the point load, has been considered in Fig. 4.8. This particular convergence study
can be treated as the most severe one for the problem considered, and, paradoxically, it is
rarely noted in other presentations of this example in the literature (compare Heppler &
Hansen [194], Kreja et al. [269], Reddy & Arciniega [399]). From Fig. 4.8 one can see that
results of the 9-URI element are not acceptable, as they do not converge to the exact
solution, similarly as those calculated with the 9-SRI element.
A confirmation of this observation can be found in Fig. 4.9, where the radial deflection
distribution at the line B-A obtained with the 10×10 mesh of elements is shown. The
characteristic oscillations of the results obtained with 9-SRI and 9-URI elements indicate
the presence of hour-glass mechanisms in those two models.
1.2
1.0
Normalized axial displacement u D

0.8
0.6
9-FI
9-SRI
9-URI
0.4 8-FI
8-SRI
8-URI
9-ANS
0.2
0.0
3 5 7 9 11 13 15 17 19 21
Fig. 4.7: Convergence study for axial displacement at the point D
80.0
Normailzed radial deflection wB
40.0
0.0
9-FI
9-SRI
9-URI
-40.0 8-FI
8-SRI
8-URI
9-ANS
-80.0
3 5 7 9 11 13 15 17 19 21
Fig. 4.8: Convergence study for radial deflection at the point B
It is interesting to notice that no hour-glass mechanisms were found in the solutions

obtained with the under-integrated Serendipity 8-node elements (8-SRI and 8-URI). The
8-SRI element is not free from locking but the results based on the 8-URI element are
almost identical with those obtained with the 9-ANS element and quite close to the
analytical solution [194].
In the following the performance of the 16-node Lagrange element is examined starting
with the deformation profile for the line BA obtained for a regular mesh of 10x10 16-node
elements applying various schemes of integration (Fig. 4.10).
2.8
2.6
2.4 9-FI
9-SRI
2.2 9-URI
2.0 8-FI
Radial deflection (10 -7 in)
1.8 8-SRI
8-URI
1.6 9-ANS
1.4
1.2
1.0
0.8 wB - analytical [194]
0.6
0.4
0.2
0.0
-0.2
-0.4
0 100 200 300

(B) Distance from the center (in) (A)
Fig. 4.9: Deformation profile (radial deflections) along the line BA
16
16-FI
12 16-SRI
16-URI
Radial deflection (10 -8 in)
8-URI
8
-4
-8
0 100 200 300

(B) Distance from the center (in) (A)
Fig. 4.10: Deformation profile along the line BA calculated with 16-node elements
As one can observe in Fig. 4.10 the reduced integration generates the hour-glass
mechanisms also for 16-node Lagrange elements. The characteristic oscillations in the
deformation profile obtained with 16-SRI and 16-URI elements are very similar to those
recorded for 9-SRI and 9-URI elements. However, now the difference between the profiles
obtained with selective and uniformly reduced integration is much smaller. By keeping the
same number of nodal points at the edge as for the 10x10 mesh of 16-node elements one
can obtain a regular mesh of 16x16 8-URI elements which produces almost the same
deformation profile as that calculated with 16-FI elements.
The detailed comparison of the results calculated with the 16-FI, 8-FI and 8-URI
elements for the selected displacement components at points C, D and B is presented in
Table 4.4 for the increasing number of nodes at the edge. As one can observe from the
numbers displayed in Table 4.4, the 16-FI element is not free from locking, nevertheless its
convergence is significantly better than that of 8-FI. It is also interesting to notice that for
dense meshes the results of the 16-FI elements almost match with those of 8-URI; however,
they are quite different from the analytical solutions given by Heppler & Hansen [194]. One
should be aware that analytical solutions were obtained by Heppler & Hansen [194]
neglecting the shear deformation. Therefore the value of wC obtained by Reddy &
Arciniega [399] with the high-order interpolation shell elements based on the Third Order
Shear Deformation theory seems to be more appropriate as the reference solution (compare
the bottom line in Table 4.4).
Table 4.4: Selected displacements of pinched cylinder with end diaphragms

No. wC × 105 [in] uC × 107 [in] wB × 108 [in]
of
nodes 16-FI 8-FI 8-URI 16-FI 8-FI 8-URI 16-FI 8-FI 8-URI
7 0.4529 0.1953 1.4842 2.1189 1.1588 4.6130 101.015 39.4964 -62.825
13 1.3760 0.8307 1.7697 4.3008 3.4061 4.5926 20.4892 90.3204 3.4709
19 1.6860 1.3007 1.8054 4.5547 4.1744 4.5787 5.8429 35.5171 4.6255
25 1.7735 1.5288 1.8217 4.5744 4.4056 4.5785 4.9603 17.2825 4.8000
31 1.8076 1.6453 1.8308 4.5775 4.4886 4.5785 4.8924 11.0606 4.8675
37 1.8242 1.7111 1.8363 4.5782 4.5254 4.5785 4.8962 8.4255 4.8931
43 1.8335 1.7515 1.8398 4.5784 4.5444 4.5785 4.9039 7.1149 4.9043
49 1.8392 1.7778 1.8422 4.5784 4.5553 4.5785 4.9092 6.3847 4.9099
55 1.8429 1.7957 1.8439 4.5785 4.5621 4.5785 4.9124 5.9431 4.9129
61 1.8456 1.8084 1.8452 4.5785 4.5665 4.5785 4.9144 5.6597 4.9146
67 1.8475 1.8176 1.8463 4.5785 4.5695 4.5785 4.9156 5.4693 4.9156
73 1.8489 1.8245 1.8472 4.5785 4.5716 4.5785 4.9164 5.3368 4.9163
[194] 1.8249 4.5711 5.2222
[399] 1.8465 - -
Looking at the results of the 16-node and 8-node elements one should remember that
with the same number of nodes at the edge, the total number of degrees of freedom for the
16-node element model is over 30% higher than those of the 8-node element model. For
that reason the results given in Table 4.4 are illustrated in Fig. 4.11÷4.13 as new
convergence plots presented for the increasing total number of degrees of freedom.
2.0
Radial displacement wC * 10-5 [in] 1.6
1.2
16-FI
0.8 8-FI
8-URI
9-ANS
Reference solution [152, 194]
0.4
Reference solution [399]
0 4000 8000 12000 16000 20000

Total number of dofs
Fig. 4.11: Radial deflection at the point C for increasing total number of dofs
5
Axial displacement uD * 10-7 [in]
3 16-FI
8-FI
8-URI
9-ANS
2 Reference solution [194]
0 4000 8000 12000 16000 20000

Fig. 4.12: Axial displacement at the point D for increasing total number of dofs
Looking at the graphs given in Fig. 4.11÷4.13 one can notice that the convergence
obtained for the 16-FI elements is notably better than that of the 8-FI elements. However,
by comparison with the results of the 9-ANS or 8-URI elements it is obvious that, contrary
to the statement of Heppler & Hansen [194], the 16-node Lagrange element is not free from
locking. Additionally, when comparing the performance of the 16-FI and 8-URI elements
one should remember that the total number of integration points for the 16-FI element
model is equal to 177.78% of the total number of integration points for the 8-URI element
model.
12
8
Radial displacement wB * 10-7 [in]
-4 16-FI
8-FI
-8 8-URI
9-ANS
-12
-16
0 4000 8000 12000 16000 20000

Fig. 4.13: Radial deflection at the point B for increasing total number of dofs
The total number of degrees of freedom for the 9-ANS element model is equal to that
of the 16-FI element model. Looking at the graphs of the radial deflection at the point C
presented in Fig. 4.11, one can observe that the convergence of the 9-ANS element model
is slightly better than that of the 8-URI element model, even, when referenced to the total
number of dofs. However, two other convergence plots in Fig. 4.12 and Fig. 4.13 seem to
indicate that the performance of the 8-URI element is better for the axial displacement at
the point D and for the radial deflection at the point C. Additionally, due to the full
integration scheme the total number of integration points for the 9-ANS element model is
equal to 225% of the total number of integration points for the 8-URI element model. Since
the non-linear analysis requires repeated formulations of tangent stiffness matrices and
multiple solutions of the system of equations, the 8-URI element appears to be a more
rational choice in this example than the 9-ANS element.
Hemispherical shell with 18° hole

In the second example we examine the performance of the 8-URI, 8-SRI and 16-FI
elements in a linear analysis of a doubly curved shell. The hemispherical shell with 18° hole
(Fig. 4.14) has been chosen from the list of standard test problems proposed for shell elements
by MacNeal & Harder [300].
2P
2P
2P
Fig. 4.14: Hemispherical shell with 18° hole
The shell is loaded at the equator of the hemisphere by two pairs of mutually opposite
horizontal forces one pair of pinching forces and one pair of stretching forces acting along
two perpendicular radial directions. Here, similarly as in the previous example, we can test a
proper representation of inextensional bending, but additionally, because the shell response is
dominated by rigid body rotations about normals to the shell surface, this is a challenging test
for the representation of rigid body motions. The problem has been considered by many
authors; see e.g. Belytschko et al. [54], Chang et al. [103], Simo et al. [436], Stander et al.
[446], Cheung & Chen [111], Groenwold & Stander [180], Chróścielewski et al. [121],
Kreja et al. [269], Sze & Zheng [457], Sze et al. [456], Başar et al. [39], Kulikov &
Plotnikova [273], Vu-Quoc & Tan [479], Chróścielewski et al. [119], Reddy & Arciniega
[399], and Brank [74].The material properties used are: E = 6.825×107 psi and v = 0.3, the
radius R = 10 in, the thickness h = 0.04 in, and the loading force P = 1 lb. Due to the
symmetry of the problem, only one quarter of the shell has been modeled with regular meshes
of elements, as shown in Fig. 4.15.
The results for the inward radial displacement under the pinching force calculated with
the increasing number of elements are presented in Table 4.5.
The analytical solution of the inward radial displacement under the pinching force was
given by MacNeal & Harder [300] as win = 0.094 in. Later, Simo et al. [436] reported another
analytical solution based on asymptotic expansions, win = 0.093 in. Looking at the numbers
presented in Table 4.5 one can notice that the solutions obtained with the 8-URI elements
converge at the level win = 0.0937 in. Reddy & Arciniega [399] obtained very close results
using the high-order interpolation 81-node shell elements based on the TOSD theory and
selective reduced integration. Due to a significant difference between the results obtained
with the 8-URI and 8-SRI elements, one can conclude that the membrane locking is a decisive
factor in this example. It is also interesting to notice that the 16-FI element provides a better
convergence than that of the 8-SRI element.
z
e x-sym
n metry
la
plane
p
y
tr
e
m
m
y
-s
y
P
y
P
x
Fig. 4.15: FE model of hemispherical shell with 18° hole
Table 4.5: Hemispherical shell - radial inward displacement under the pinching force
No. of nodes win * 102 [in]
at the edge 16-FI 8-SRI 8-URI
7 2.42339 0.27627 3.94087
13 8.63613 2.15702 9.19061
19 9.21257 5.40948 9.34673
25 9.31337 7.58081 9.36580
31 9.34087 8.53985 9.37050
37 9.34980 8.95051 9.37222
43 9.35340 9.13985 9.37292
49 9.35533 9.23493 9.37324
73 9.35943 9.34583 9.37360
Reference [300] 9.40
The latter observation finds a confirmation also in convergence plots of the inward
radial displacement under the pinching force presented in Fig. 4.16 for the increasing total
number of degrees of freedom.
The superior performance of the 8-URI element in the linear benchmark examples
considered above shows a very promising potential for the application to large deformation
analysis in the next chapter.
0.1
0.08
Radial inward displacement, win [in]
0.06
16-FI
0.04
8-SRI
8-URI
0.02 Reference solution [436]
0 4000 8000 12000 16000 20000

Fig. 4.16: Radial inward displacement under the pinching force for increasing total number of dofs
Chapter 5
LARGE DEFORMATION ANALYSIS EXAMPLES
5.1. Instability of clamped-hinged circular arches

In the first example we will test how the incorporation of different terms in the strain-
displacement relations affects the response of a clamped-hinged isotropic circular arch
subjected to a point load at the crown. An analytical solution for the clamped-hinged
circular arches with the varying value of the angle , in the range of large rotations was
given by DaDeppo & Schmidt [131].
We start with the arch characterized by a subtending angle , = 100 degrees as presented
in Fig. 5.1.
Fig. 5.1: Clamped-hinged arch with , = 100 degrees
The thickness of the arch is taken as h=R/400. Twenty 8-URI elements have been used
in the discretization of the arch in the analyses performed within the following
formulations: RVK5, MRT5, LRT5 and LRT56. The graphs of the normalized vertical and
horizontal displacements at the crown, w/R and u/R, are presented in Fig. 5.2 and Fig. 5.2,
respectively, versus the dimensionless load PR2/EI (with R being the radius of the arch, P
the vertical force acting at the crown, E the Young modulus and I the moment of inertia of
the cross-sectional area). The results of the LRT56 model agree very well with the
analytical solution of DaDeppo & Schmidt [131]. To verify the accuracy of the LRT56
formulation beyond the limits of the available analytical solution [131] an additional
reference solution obtained with a degenerated beam type element Nash2D (Kreja &
Cywiński [262]) has been provided. An excellent agreement in the entire post-buckling
region can be observed in Fig. 5.2 between the results of the LRT56 and Nash2D. It is
noteworthy that the results obtained with the LRT5 formulation differ significantly from
the response calculated with the LRT56 or Nash2D. It is evident that without the proper
updating of rotations the LRT5 model is unable to predict a correct solution for the
analyzed problem. The range of the acceptable accuracy of the LRT5 solution is just a little
beyond the limits of the moderate rotation theory (see Fig. 5.2) and is comparable to the
accuracy obtained with the MRT5 formulation. On the other hand, the RVK5 model gives
the solution that is surprisingly close to the correct results – there are visible quantitative
104 Chapter 5. Numerical examples
differences, but the shape of the equilibrium path estimated with the RVK5 is almost the
same as obtained with the LRT56 formulation.
LRT56
LRT5
20.0 MRT5
RVK5
DaDeppo & Schmidt [131]
Nash2D
Dennis & Palazotto [139]
15.0
Load parameter PR2/EI
10.0
5.0
Range of
moderate rotations
0.0
0.00 0.20 0.40 0.60 0.80
Normalized displacement w/R
Fig. 5.2: Normalized downward displacement at the crown for 100-degree arch
LRT56
LRT5
20.0 MRT5
RVK5
Nash2D
15.0
10.0
5.0
0.0
-0.08 -0.04 0.00 0.04 0.08
Normalized displacement u/R
Fig. 5.3: Normalized horizontal displacement at the crown for 100-degree arch
5.1. Instability of clamped-hinged circular arches subjected to point load 105
The 100-degree arch had been analyzed also by Dennis & Palazotto [139]. Their model
gave much softer response than the analytical solution of DaDeppo & Schmidt [131], as it
is shown in Fig. 5.2. Dennis & Palazotto [139] explained that this difference could be
caused by the inclusion of the middle surface extensibility and transverse shear deformation
in their model, both of which are not taken into account by DaDeppo & Schmidt [131]. The
present results do not support this conclusion since our LRT56 model includes these effects
and nevertheless no differences with the results of DaDeppo & Schmidt [131] can be
observed. Looking in more detailed way we notice that there are just two main differences
between the LRT56 model and that used by Dennis & Palazotto [139]: firstly, the LRT56
bases on the FOSD theory whereas the parabolic transverse shear stress distribution of the
TOSD model was assumed by Denis & Palazotto [139]; secondly, the LRT56 incorporates
the exact non-linear strain-displacement relations while Dennis & Palazotto [139] used
non-linear terms only for the in-plane strains. In order to make possible a more detailed
examination of the results two modified models have been created based on the present
formulation just by dropping all non-liner terms for the transverse shear strains in the LRT5
and LRT56 formulations named m-LRT5 and m-LRT56, respectively.
The results of the m-LRT56 model (see Fig. 5.4) have slightly changed with respect to
LRT56, but they are still away from the solution given by Dennis & Palazotto [139].
However, the equilibrium path predicted with the m-LRT5 model are placed very close to
the curve of Denis & Palazotto [139], what can suggest that their approach neglects the
proper updating of the rotational degrees of freedom. One can also observe that the refined
representation of the transverse shear strains incorporated by Dennis & Palazotto [139] is
not relevant for the accuracy of the solution for a rather thin arch (R/h=400) considered
here.
20.0 Dennis & Palazotto [139]

LRT56
LRT5
m-LRT56
m-LRT5
15.0
10.0
5.0
0.0
0.00 0.20 0.40 0.60 0.80
Fig. 5.4: Additional comparison study for the normalized horizontal displacement
at the crown for 100-degree arch
In the next example of a clamped-hinged isotropic circular arch considered here the
subtending angle increased to 215 degrees. The arch is subjected to a vertical point load at
the crown as presented in Fig. 5.5. Similarly, as in the previous example, the analyzed
model of the arch consists of twenty 8-URI elements.
Fig. 5.5: Clamped-hinged arch, with , = 215 degrees
The graphs of the normalized vertical displacement, w/R and the normalized horizontal
displacement, u/R of the crown point versus the dimensionless load PR2/EI are given in
Fig. 5.6 and Fig. 5.7, respectively.
LRT56
LRT5
MRT5
RVK5
Nash2D
11.0 Li [59]
10.0 DaDeppo & Schmidt [131]
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
Range of
1.0 moderate rotations
0.0
-1.0
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25
Fig. 5.6: Normalized vertical displacement at the crown for 215-degree arch
5.1. Instability of clamped-hinged circular arches subjected to point load 107
The results obtained with the LRT56 agree very well with the analytical solution as far
as it was given by DaDeppo & Schmidt [131]. Again, the post-buckling path is positively
verified by a comparison with the results of the large rotation analysis performed with a
degenerated beam type element Nash2D (Kreja & Cywiński [262]).
LRT56
LRT5
MRT5
RVK5
Nash2D
11.0 Li [59]
10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
-1.0
-0.2 0.0 0.2 0.4 0.6 0.8
Normalized displacement u/R
Fig. 5.7: Normalized horizontal displacement at the crown for 215-degree arch
The 215-degree arch was analyzed also by Hsiao & Chen [206], Ibrahimbegović & Frey
[222], Li [285], Surana [452], and by Zienkiewicz & Taylor [510]. Li [285] showed just
eight points of the equilibrium path in the pre-buckling range and one separate point after
the snap-through took place. Those points correspond quite well with the LRT56 solution
(see Fig. 5.6 and Fig. 5.7), but in the vicinity of the limit point Li solution [285] shows
differences with the LRT56 and also with the analytical solution of DaDeppo & Schmidt
[131]. The MRT5 model provides excellent results in the range of validity of the theory as
marked on Fig. 5.6. In this example the importance of the proper updating procedure of the
rotations is manifested by the big discrepancy of the LRT5 solution, where not only
quantitative differences can be observed, but even a qualitatively different response of the
structure is predicted (see Fig. 5.6 and Fig. 5.7). The range of acceptable accuracy of the
LRT5 approach is only slightly bigger than that of the MRT5. The RVK5 model gives a
quite good solution in the range of moderate rotations; however, outside that range the path
of the RVK5 significantly differs from all other solutions.
Quite recently Kapania & Li [235] presented their solution obtained with geometrically
exact curved beam elements. As far as it is possible to recognize from the figure given in
[235], one can observe an excellent agreement of their solution with the LRT56 results (see
Table 5.1 for comparison of the limit load points).
It should be mentioned that the main purpose of the presented example was to test the
behavior of the numerical model. However, the author is aware that the response of a real
structure can considerably differ from the presented solution due to a possible contact
between the deformed beam and the support in the post-buckling region as considered by
Simo et al. [441].
Table 5.1: Comparison of limit load points for the clamped-hinged 215-degree arch
Normalized load PR2/EI
Model
Max. limit point Min. limit point
LRT56 8.9712 -0.7304
Nash2D (3-node degenerate beam element) 8.9718 -0.7306
Kapania & Li [235] (4-node curved beam element) 8.9727 -0.7360
DaDeppo & Schmidt [131] (analytical) 8.97 -
5.2. Clamped laminated shallow arch under point load

The symmetrical buckling of the shallow circular laminated arch was analyzed earlier
with the use of degenerated shell elements by Liao & Reddy [287, 288]. The geometry of
the arch is presented in Fig. 5.8 with R = 100 in, h = 2 in, β = 0.707 and the width of the
arch taken equal to 1 in. Due to symmetry of the problem one half of the arch has been
modeled with five 8-URI elements. The material properties used are Ea = 25·106 psi,
Eb = 1·106 psi, Gac = 5·105 psi, Gac = 5·105 psi, Gbc = 5·105 psi and νab = 0.25. The arch
consists of two layers, each layer has the same thickness 1 in and is made of the same
orthotropic material, yet a different ply orientation has been used in each layer. The ply lay-
up for the considered example is [0, 90], i.e. in the bottom layer the material a-axis
coincides with the θ1 axis of the shell coordinate system, whereas in the upper layer the
material axes are rotated by 90 degrees (compare Fig. 3.10).
Fig. 5.8: Clamped shallow laminated arch

5.2. Clamped laminated shallow arch under point load 109
The obtained results are presented in Fig. 5.9 as the equilibrium paths in the space of
the non-dimensional load parameter, P = 10 PR 2 β / π 2 EI , and the central vertical deflection
w.
LRT56
LRT5
12
MRT5
RVK5
10 Liao & Reddy [287]
Load parameter, (10 PR2q/p2EI)
NASHL
0
0 5 10 15 20 25 30 35 40 45
Central deflection, (in)
Fig. 5.9: Central vertical deflection for the clamped laminated arch
Looking at the evident differences in the responses obtained with the MRT5, LRT5 and
the LRT56 models one can conclude that the considered example deals with finite rotations.
The range of moderate rotations ends before the load parameter reaches the value 6.0 and,
consequently, the MRT5 solution does not predict the snap-through instability, which is
evident in both the large rotation formulations, LRT5 and LRT56. However, the lack of
proper treatment of the rotations in the LRT5 approach is manifested by overestimation of
the maximum load level and total disagreement with LRT56 in the post-buckling path. Here
again, similarly as before for the 100-degree arch, the RVK5 model gives the solution that
is surprisingly close to the results of the LRT56, much closer than that of the LRT5
formulation. The maximum load level predicted by the LRT56 model agrees quite well
with the reference solution of Liao & Reddy [287], but the further part of their post-
buckling path differs from the LRT56 response. To allow for the final verification of the
obtained results additional large rotation computations have been performed for the
analyzed arch with the own program NASHL81 based on the degenerated layered beam
element, Bödefeld et al. [70], and a very good agreement has been obtained in the entire
post-buckling range with the LRT56 solution.
81
The NASHL program was developed as the extension of the Nash2D program for large rotation analysis of
isotropic 2D shells, Kreja & Cywiński [262].
5.3. Hinged laminated cylindrical panels under point load

The hinged cylindrical panel under the point load presented in Fig. 5.10 was analyzed
in different variants by many researchers. In fact, it originates from the family of isotropic
cylindrical panels proposed by Sabir & Lock [409] and has been used afterwards as a very
popular benchmark test for the non-linear analysis of isotropic shells (see e.g. Horrigmoe
& Bergan [204], Bathe & Bolourchi [47], Ramm [379], Chang & Sawamiphakdi [104],
Hughes & Liu [212], Parisch [347], Noor & Andersen [327], Oliver & Oñate [332], Pinsky
& Kim [368], Huang [208], Liao & Reddy [287, 288], Hsiao & Chen [206], Palmerio et al.
[343], Saleeb et al. [412], Simo et al. [437], Chróścielewski et al. [120], Sansour & Bufler
[417], Ibrahimbegović & Frey [222], Jiang & Chernuka [230], Sansour & Bednarczyk
[414], Kreja et al. [269], Marcinowski [305], Meek & Ristic [312], Meek & Wang [313],
Ferreira & Barbosa [162], Massin & Al Mikdad [307], Sze & Zheng [457], Fontes Valente
et al. [167], Chróścielewski et al. [119]).
The layered orthotropic variant of the panel has been analyzed first by Saigal et al.
[411] with the formulation neglecting the effect of shear deformations. Laschet & Jeusette
[276] analyzed the same panels with the 3-D isoparametric multi-layer finite element;
Dorninger [147] applied the 16-node degenerated shell element. Kim & Voyiadjis [242]
calculated a large displacement response for laminated panels using the co-rotational
formulation. Additionally, these laminated panels were analyzed with the finite rotation
formulations by Wagner [480], Brank et al. [79] and Balah & Al-Ghamedy [31]. Quite
recently, Sze et al. [455] considered this example as a member of the group of popular
benchmark problems for geometric non-linear analysis of shells.
Fig. 5.10: Hinged laminated cylindrical panel under point load
The geometry and the coordinate system used in the analysis are presented in Fig. 5.10.
The following dimensions have been assumed: R = 2540 mm, L = 254 mm and β = 0.1. The
thickness of the shell changed from h = 12.6 mm for the panel A, through h = 6.3 mm for
the panel B, to h = 3.15 mm for the panel C. The parameters of the orthotropic material are
Ea = 3.3 kN/mm2, Eb = 1.1 kN/mm2, Gab = Gac = Gbc = 0.66 kN/mm2 and νab = 0.25. The ply
orientation is characterized by the angle α as indicated in Fig. 5.10.
First, the panel A with the cross ply lay-ups [0/90/0] and [90/0/90] is considered and
the obtained results are presented in Fig. 5.11 as the graphs of the central vertical deflection
versus the central force. Due to symmetry of the problem one-quarter of the shell is
modeled with 2x2 8-URI elements. One can conclude that the entire deformation of the
panel stays well within the range of small rotations because almost no difference can be
5.3. Hinged laminated cylindrical panels under point load 111
observed between the LRT56, MRT5 and RVK5 solutions. Own results agree also quite
well with the given reference solutions of Saigal et al. [411], Laschet & Jeusette [276],
Brank et al. [79] and Sze et al. [455]. Similar results were reported also by Dorninger [147],
Wagner [480] and Kim & Voyiadjis [242].
4.0
LRT56
3.5 MRT5 Ply lay-up [90/0/90]
RVK5
3.0 Saigal et. al. [411]
Laschet & Jeusette [276]
2.5 Brank et.al. [79]
Central force [kN]
Sze et al. [455]

2.0
1.5
1.0
0.5
0.0 Ply lay-up [0/90/0]
-0.5
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0
Central displacement [mm]
Fig. 5.11: Central deflection for the cross ply laminate panel A (h=12.6 mm)
Next, the same panel A has been analyzed for the angle ply lay-up [-45/+45]. Since this
problem is asymmetrical it was necessary to model the whole panel (mesh 4x4 8-URI
elements). The obtained results are presented in Fig. 5.12 together with the reference
solutions of Saigal et al. [411] and of Laschet & Jeusette [276]. Here again, there are almost
no differences between the results of the LRT56, MRT5 and the RVK5 analyses. All three
our models give solutions that are very close to the response predicted by Laschet &
Jeusette [276]; however, there is a visible disagreement with the reference solution given by
Saigal et al. [411]. One should notice that Saigal et al. [411] analyzed just one-quarter of
the shell, assuming biaxial symmetry, what was not correct for the case of the angle-ply
lamination82. Dorninger [147] and Kim & Voyiadjis [242] reported results that were very
similar to those of Saigal et al. [411], but their calculations were also performed for one
quarter of the shell. To allow for a more detailed examination of the angle ply laminate
,additional calculations were performed with the LRT56 model for the whole panel and for
one quarter of the shell assuming two different lay-ups [-45/+45] and [+45/-45]. The results
of those calculations are presented in Fig. 5.13 together with the reference solutions.
82
This problem was considered e.g. by J. N. Reddy (“A note on symmetry considerations in the transient response
of unsymmetrically laminated anisotropic plates,” International Journal for Numerical Methods in Engineering,
Vol. 20, pp. l75-l94, l984) who indicated that biaxial symmetry conditions were not valid for unsymmetrically
laminated composites (see also Reddy [394]).
2.5
LRT56
MRT5
2.0 RVK5
Saigal et. al. [411]
Central force [kN] Laschet & Jeusette [276]
1.5
1.0
0.5
Ply lay-up [-45/+45]
0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0
Fig. 5.12: Central deflection for the angle ply laminate panel A (h=12.6 mm)
2.5
LRT56 results
whole panel [+45,-45]
whole panel [-45,+45]
2.0
one quarter [+45,-45)
one quarter [-45,+45]
Central force [kN]
Saigal et. al. [411]

1.5
1.0
0.5
0.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0
Fig. 5.13: Additional comparison for the angle ply laminate panel A (h=12.6 mm)
As one can observe in Fig. 5.13, the same equilibrium paths were obtained for the both
lay-ups ([-45/+45] and [+45/-45]), when the whole panel was represented in the FE model.
On the other hand, the choice of the stacking sequence ([-45/+45] or [+45/-45]) was
relevant when the reduced model of a one quarter with biaxial symmetry boundary
conditions was used in calculations. Our results obtained for the whole panel agree quite
well with the reference solution of Laschet & Jeusette [276]. However, when the reduced
model was used for the stacking sequence [-45/+45] the obtained response is very close to
the solution of Saigal et al. [411].
Brank et al. [79] introduced a laminated panel with the cross ply lay-ups [0/90/0] and
[90/0/90] and with the thickness reduced by half (h = 6.3 mm) hopping probably to gain a
more pronounced snapping behavior. Our results for the panel B (h = 6.3 mm) with the
cross ply lay-ups [0/90/0] and [90/0/90] are presented in Fig. 5.14 and Fig. 5.15,
respectively.
1.0
LRT56
0.8 MRT5
RVK5
0.6 Brank et al. [79]
Sze et al. [455]
Force [kN]
0.4
0.2
0.0
-0.2
Ply lay-up [0/90/0]
-0.4
0 5 10 15 20 25 30 35
Fig. 5.14: Central deflection for the cross ply [0/90/0] laminate panel B (h=6.3 mm)
1.0
LRT56
MRT5
0.8
RVK5
Brank et al. [79]
0.6 Sze et al. [455]
Force [kN]
0.4
0.2
0.0
-0.2
Ply lay-up [90/0/90]
-0.4
0 5 10 15 20 25 30 35
Fig. 5.15: Central deflection for the cross ply [90/0/90] laminate panel B (h=6.3 mm)
One can observe in Fig. 5.14 and Fig. 5.15 that the equilibrium paths for the panel B
(h = 6.3 mm) with the cross ply lay-ups are represented by evidently more complex curves
than for the panel A (h = 12.6 mm). Nevertheless, here again the responses for LRT56,
MRT5 and RVK5 formulations seem to be the same. Our calculations for the panel B have
been performed for one quarter of the shell modeled with a 4x4 mesh of 8-URI elements.
The same problem has been considered also by Balah & Al-Ghamedy [31], who obtained
almost the same solutions as the one given by Brank et al. [79]. The results obtained for all
three our models match the reference solutions of Brank et al. [79] and also that given by
Sze et al. [455].
The panel B with the angle ply lay-up [-45/45] has been analyzed in our next example,
which according to our knowledge does not possess any reference solution yet. The
calculations performed for the whole panel with the 4x4 mesh of 8-URI elements have
shown that a bifurcation point exists on the equilibrium path below the maximum load
point. To force the asymmetric response an additional horizontal force 0.001 P has been
applied as the load imperfection. The symmetric response has been obtained when the
horizontal displacements of the mid-point were fixed with additional constraints. Both
curves are presented in Fig. 5.16. Due to a limited magnitude of rotations in the present
example almost no difference has been obtained between the LRT56 and MRT5 models.
0.75
LRT56 and MRT5 give the same solution
symmetrical response (add. constrains)
asymmetrical response (load imperfection)
0.50
Central force [kN]
0.25
0.00
Ply lay-out [+45/-45]
-0.25
0 5 10 15 20 25 30 35
Fig. 5.16: Central deflection for the angle ply laminate panel 6.3 mm thick
In the third case of the present analysis even a further reduction of the panel thickness
was presumed taking h = 3.15 mm (panel C). Two cross ply lay-ups, [0/90/0] and [90/0/90],
have been considered for the panel C with the corresponding results presented in Fig. 5.17
and Fig. 5.18, respectively. However, even for this very thin panel the results of the LRT56,
MRT5 and RVK5 are very similar. Nevertheless, after a closer look at the graphs in
Fig. 5.17 and Fig. 5.18 one can conclude that, paradoxically, the results of the RVK5
formulation seemed to be more accurate in that particular case than those of the MRT5
model.
One can remark that although the decrease of the thickness of the shell has reduced its
stiffness, nevertheless, the range of rotations has not exceeded the limits of small rotations.
Because of that the considered example of the hinged shallow laminated cylindrical panel
cannot serve as a proper test problem for the large rotation shell analysis.
0.2
LRT56
MRT5
RVK5
0.1
Central force [kN]
0.0
Ply lay-out [0/90/0]

-0.1
0 5 10 15 20 25 30 35
Fig. 5.17: Central deflection for the [0/90/0] laminate panel C (h=3.15 mm)
0.30
0.25 LRT56
MRT5
0.20 RVK5
0.15
Central force [kN]
0.10
0.05
0.00
-0.05
-0.10 Ply lay-up [90/0/90]
-0.15
0 5 10 15 20 25 30 35
Fig. 5.18: Central deflection for the [90/0/90] laminate panel C (h=3.15 mm)
5.4. Glass-epoxy cylinder under internal pressure

A short cylinder shown in Fig. 5.19 is clamped at both ends and subjected to a
uniformly distributed internal pressure. Geometry of the shell is described by the radius
R = 20 in, the length 2L = 20 in and the thickness h = 1 in. The panel is made of one layer
glass-epoxy composite. Properties of a glass-epoxy orthotropic material are characterized
by the following elastic constants: Ea = 7500 ksi83, Eb = 2000 ksi, νab = 0.25, Gab = 1250 ksi
and Gbc = Gac = 625 ksi.
clamped edge
h
b R
c L
b
a
w L
clamped edge
Fig. 5.19: Glass-epoxy cylinder under internal pressure
Taking advantage of the axial symmetry and the planar symmetry with respect to the
mid-height plane, one band of shell elements, with proper boundary conditions has been
used in the discretization. Two elements in a mesh were enough to obtain the converged
solutions presented in Fig. 5.20. It was assumed that the pressure load is displacement
independent.
The reference results of Chang & Sawamiphakdi [104] are available only for the
internal pressure up to 10 ksi, but the present analysis has been performed for extended load
level to examine differences between results obtained with different formulations. An
additional reference solution has been provided with the axi-symmetric analysis performed
with the program Nash2D (Kreja & Cywiński [262]). One can notice almost no difference
between the solutions obtained with the RVK5 and MRT5 models, both those plots are very
close to the reference results of Chang & Sawamiphakdi [104], whereas the LRT5, LRT56
and Nash2D predict another response. Especially interesting is the fact that there is no
difference between LRT5 and LRT56 solution, what indicates that large rotations are not
really present in the analyzed panel. Looking at the strain-displacements given in Table 3.1
one can notice that the MRT5 (similarly as the RVK5) formulation does not include all
membrane strain terms. One can expect that for the high internal pressure the expansions in
the analyzed example are relatively large and, in consequence, the omitted strain terms can
83
1 ksi = 1000 psi = 6895 kPa
5.4. Glass-epoxy cylinder under internal pressure 117
be quite essential for the results. This seems to explain the difference between results of the
LRT (LRT56 and LRT5) and MRT5 models for the case of moderate rotations.
60.0
RVK5
50.0 MRT5
LRT5
LRT56
40.0 Nash2D
Internal pressure (ksi)
Chang & Sawamiphakdi [104]
30.0
20.0
10.0
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
radial displacement (in)
Fig. 5.20: Radial deflection of the glass-epoxy cylinder under internal pressure
5.5. Asymmetric cross-ply simply supported plate strip

We consider a simply supported asymmetric laminate [0/90] as shown in Fig. 5.21
assuming the following data: a = 9.0 in, b = 1.5 in, h = 0.04 in, Ea = 2.0 x 107 lb/in2,
νab = 0.30, Eb = 1.4 x 106 lb/in2, Gab = Gbc = Gca = 0.7 x 106 lb/in2.
p
h 0°
3
x 90°
///////// /////////
a
simply supported
2
x
free edge
b
1
x
a
simply supported
free edge
Fig. 5.21: Cylindrical bending of the composite plate strip

This example was initially investigated by Sun & Chin [450], who presented large
deflection cylindrical bending results of a pinned composite plate obtained with the von
Kármán plate theory. Reddy [396], who analyzed this problem with his TOSD model for
von Kármán-type non-linear plate theory, showed that really large deflections could be
obtained for the simply supported plate. Başar et al. [37] repeated this analysis using the
fully non-linear TOSD formulation accounting for finite rotations.
Due to the dual symmetry of the analyzed structure, a quarter of the plate is modeled
with nine 8-URI elements. The results obtained with the MRT5, LRT5 and LRT56
formulations are shown in Fig. 5.22 together with the reference solutions of Reddy [396]
and Başar et al. [37].
Our LRT56 results agree very well with those of Başar et al. [37] what allows one for
the conclusion that in this particular example the refined representation of the transverse
shear strains does not improve the accuracy of the solution. Both solutions are quite
different from the results of Reddy [396], which is due to the fact that his model based on
the von Kármán-type plate theory is not capable of dealing with large rotations. Reddy
solution [396] gives an acceptable prediction of the strip deflection only for the range w/h <
10, whereas our MRT5 model is capable of dealing with bigger deflections (w/h < 17),
which are a little beyond the range of moderate rotations marked in Fig. 5.22. It is
interesting that using the full non-linear strain-displacement relations but without the proper
updating of the rotations (the LRT5 formulation) one cannot obtain better results than with
the MRT5.
6.0
5.0 MRT5
LRT5
4.0 LRT56
Pressure load [lb/(in²)]
Reddy [396]
Baºar et al. [37]
3.0
2.0
Range of
1.0 moderate rotations
0.0
0 20 40 60 80 100
Normalized transverse deflection (v/h)
Fig. 5.22: Normalized central deflection of the plate strip under pressure load
5.6. Clamped laminated cylindrical panels under point load

Tsai et al. [470] analyzed a series of deep cylindrical laminated panels as presented in
Fig. 5.23, assuming the following geometrical data: R = 12 in, L =5.5 in, β = 0.5, and
5.6. Clamped laminated cylindrical panels under point load 119
h = 0.04 in. Various stacking sequences of a laminate made of a glass-epoxy composite

were considered by Tsai et al. [470], with the orthotropic material of a single layer
characterized by Ea = 20.46·106 psi, Eb= Ea / n, Gab = Gac = 0.62·Eb, Gbc = 0.31·Eb, and
ν12 = 0.313, with n standing for a variable degree of orthotropy.
Fig. 5.23: Clamped cylindrical panel under point load
In the first case the 8-layer cross-ply [0/90/0/90]s laminate is considered with n=5. The
central deflection of the panel is shown in Fig. 5.24 versus the central load.
240
RVK5
200 MRT5
LRT5
LRT56
160 Tsai et al. [470]
Central force [lb]
120
80
40
0
0 0.5 1 1.5 2 2.5 3
Central deflection [in]
Fig. 5.24: Central deflection of the clamped laminated panel [0/90/0/90]s with E1=5E2
As one can observe in Fig. 5.24 the buckling load level predicted with the LRT56
formulation is in a quite good agreement with the solution given by Tsai et al. [470]. Very
similar results for the pre-buckling range have also been obtained with the LRT5, MRT5
and RVK5 models. However, big differences appeared in the post-buckling range of the
equilibrium path. Quite surprisingly, the RVK5 results are much closer to the LRT56
solution than those obtained with the LRT5 and MRT5 models. The reference solution
[470] is quite close to the results of the RVK5 and LRT56 but the difference is still
significant. According to Tsai et al. [470], their model differs from the LRT56 in two
aspects: 1) the transverse shear stress distribution is parabolic through the shell thickness;
2) linear strain-displacement relations are assumed for the transverse shear strains. To allow
for more adequate comparison of the results we have repeated the analysis with the
modified model m-LRT56 created by dropping all non-linear terms in the transverse shear
strains of LRT56 similarly as introduced in the example of the 100-degree circular arch (see
Section 5.1).
240
m-LRT5
200 m-LRT56
LRT5
LRT56
160 Tsai et al. [470]
Central force [lb]
120
80
40
0
0 0.5 1 1.5 2 2.5 3
Fig. 5.25: Additional comparative study for the clamped laminated panel [0/90/0/90]s, n=5
Almost no difference between the results of m-LRT56 and LRT56 can be observed in
Fig. 5.25, what can support the opinion of Tsai et al. [470] that non-linear terms in the
transverse shear strains are insignificant in the considered example. However, the m-LRT56
solution still differs from that given by Tsai et al. [470]. In the second attempt the m-LRT5
model based on the LRT5 approach (i.e. obtained by dropping all non-linear terms in the
transverse shear strains of LRT5) has been applied and this time the obtained results were
almost identical with those of Tsai et al. [470]. The conclusions are the following:
a) similarly as in the LRT5 model, the formulation applied by Tsai et al. [470] does
not perform a proper updating of rotations;
b) the proper updating of the rotational degrees of freedom for thin composite shells
can be more essential for the accuracy of the solution than the refined
representation of the transverse shear strains incorporated in the TOSD model of
Tsai et al. [470].
An additional support for the above conclusions can be found in Table 5.2 where the
snap-through loads calculated for the clamped cylindrical panel with different models are
displayed in the descending order, from 176.50 lb for the MRT5, through 174.57 lb for the
LRT5, and 172.32 lb for LRT56, to 167.11 lb for the RVK5. The value measured at the
graph in Tsai et al. [470] (169 lb) is slightly higher than predicted with the m-LRT5
formulation (167.99 lb), where all non-linear terms were omitted for the transverse shear
strains. It is also worth to notice that the difference between the snap-through load
estimated with the models m-LRT56 and LRT5 is insignificant.
Table 5.2: Comparison of snap-through loads for laminated panel [0/90/0/90]s, n=5
Model snap-through load [lb]

MRT5 176.50
LRT5 174.57
m-LRT56 172.53
LRT56 172.32
Tsai et al. [470] 169
m-LRT5 167.99
RVK5 167.11
In the next study we check how the change of the ply orientation can affect the LRT56
response of the 8-layer laminate. Additionally to the cross-ply laminate [0/90/0/90]s
considered above, three other stacking sequences of a laminate are considered for the case
n=5 as presented in Fig. 5.26. Analyses of symmetrical cross-ply schemes [0/90/0/90]s and
[0/0/90/90]s were performed with the 5×5 mesh of the 8-URI elements for one quarter of
the shell, whereas the 10×10 mesh was used to model the whole panel with the angle-ply
schemes [-45/45/-45/45]s and [0/-45/45/90]s.
300
E1 = 5 E2
250
[0/90/0/90]s
[-45/45/-45/45]s
200
[0/0/90/90]s
Central force [lb]
[0/-45/45/90]s
150
100
50
0
0 1 2 3
Fig. 5.26: Influence of the ply orientation for the laminate with n=5
As one can observe in Fig. 5.26, the maximum snapping load (172.32 lb) was
calculated for the cross-ply laminate [0/90/0/90]s. A very similar response was obtained for
the angle-ply scheme [-45/45/-45/45]s. However, a slightly lower level of snapping load
(154.27 lb) was recorded in that case. It is quite interesting that the equilibrium path for the
angle-ply scheme [0/-45/45/90]s (with the snapping load 103.75 lb) is almost identical with
the path obtained for the cross-ply laminate [0/0/90/90]s (the snapping load 103.32 lb).
When the degree of orthotropy is raised to n=15, one can observe a bigger diversity
among the results for various ply lay-ups, as shown in Figure 5.27. The decrease of the
snapping load is more rapid for the [0/0/90/90]s scheme (from 103.32 lb for n=5 to 62.46 lb
for n=15) than for the [0/90/0/90]s one (respectively, from 172.32 lb to 145.07 lb).
300
E1 = 15 E2
250
[0/90/0/90]s
[-45/45/-45/45]s
200
[0/0/90/90]s
Central force [lb]
[0/-45/45/90]s
150
100
50
0
0 1 2 3
Fig. 5.27: Influence of the ply orientation for the laminate with n=15.
In the following, a more detailed study on the influence of the degree of orthotropy on
the performance of the laminate is presented for selected stacking sequences of a cross-ply
laminate, assuming a variable degree of orthotropy n = 1, 2, 5, 10, 15 and 30. We start with
the [0/90/0/90]s laminate. Looking at the graphs presented in Fig. 5.28, one can notice that
the increase of the degree of orthotropy n for the [0/90/0/90]s laminate is followed by the
essential decrease of the critical snapping load only for values n ≤ 10. When Eb is already
small, what corresponds to n > 10, the further reduction of Eb does not affect the critical
snapping load dramatically. This is due to the fact that for the high degree of orthotropy the
bending characteristic of the shell is practically dominated by Ea, which is kept constant.
The results obtained for the [0/0/90/90]s laminate (see Fig. 5.29) seem to confirm the earlier
observation, that the diversity of the snapping load levels for different values of the degree
of orthotropy n is bigger in that case than for the [0/90/0/90]s laminate. It is interesting that
for n=1, what corresponds to a nearly isotropic composite, the critical snapping load for the
[0/90/0/90]s laminate is almost the same as for the [0/0/90/90]s scheme.
450
Ply lay-up [0/90/0/90]s
400
E1= 1 E2
350 E1= 2 E2
E1= 5 E2
Central force [lb] 300 E1=10 E2
E1=15 E2
250 E1=30 E2
200
150
100
50
0
0 1 2 3
Fig. 5.28: Influence of the degree of orthotropy for the laminate panel [0/90/0/90]s
450
Ply lay-up [0/0/90/90]s
400
E1= 1 E2
350 E1= 2 E2
E1= 5 E2
300 E1=10 E2
Central force [lb]
E1=15 E2
250 E1=30 E2
200
150
100
50
0
0 1 2 3
It is quite obvious that a higher bending stiffness could be anticipated for the
[0/90/0/90]s laminate than for the [0/0/90/90]s one, because in the former case the
reinforcement along the θ 2-direction is located closer to the outer surfaces. For the same
reason, one can expect to obtain a higher overall bending stiffness by changing the stacking
sequence from [0/90/0/90]s to [90/0/90/0]s and even a bigger increase of the stiffness can be
anticipated for the [90/90/0/0]s laminate. Indeed, the results presented in Fig. 5.30 and Fig.
5.31 for the [90/0/90/0]s and [90/90/0/0]s stacking sequences, respectively, seem to fully
confirm that supposition; however, the critical snapping load for n=1 is nearly the same for
all four cross-ply schemes: [90/0/90/0]s, [90/90/0/0]s, [0/90/0/90]s and [0/0/90/90]s.
450
Ply lay-up [90/0/90/0]s
400
E1= 1 E2
350 E1= 2 E2
E1= 5 E2
300 E1=10 E2
Central force [lb]
E1=15 E2
250 E1=30 E2
200
150
100
50
0
0 1 2 3
500
Ply lay-up [90/90/0/0]s
450
E1= 1 E2
400
E1= 2 E2
350 E1= 5 E2
E1=10 E2
Central force [lb]
300 E1=15 E2
E1=30 E2
250
200
150
100
50
0
0 1 2 3
With a little amazement one can observe in Fig. 5.30 that despite of reduction of the
overall panel stiffness accompanying the increase of the degree of orthotropy, the value of
the critical snapping load for the [90/0/90/0]s laminate varies in a very limited range as
compared with the previous considered stacking sequences. Even more surprising
observation can be made for the [90/90/0/0]s laminate, where the reduction of the overall
stiffness for the increased degree of orthotropy n is accompanied quite paradoxically by the
increase of the critical snapping load. In addition, an explicit relation between the degree of
orthotropy and the level of critical snapping load for all four cross-ply schemes considered
above is presented in Fig. 5.32.
500
[90/90/0/0]s
400
Critical snapping load [lb]
[90/0/90/0]s
300
200
[0/90/0/90]s
100
[0/0/90/90]s
0
0 5 10 15 20 25 30
Degree of orthotropy, n=E1/E2
Fig. 5.32: Interrelation between the degree of orthotropy and the level of critical snapping load
5.7. Stretching of an open cylinder

In the next example a stretching of a short laminated cylinder, as shown in Fig. 5.33, is
considered with R = 4.953 in, L =5.175 in, h = 0.094 in. The open cylinder loaded by two
opposite stretching forces in its middle section exemplifies a commonly accepted
benchmark test for the non-linear analysis of isotropic shells. This example became very
popular in the literature (see e.g. Wriggers & Gruttmann [500, 501], Sansour & Bufler
[417], Damjanić et al. [134], Jiang & Chernuka [230], Sansour & Boćko [415], Brank et al.
[79], Park et al. [351], Sansour [413], Sansour & Bednarczyk [414], Chróścielewski [118],
Campello et al. [86], Fontes Valente et al. [167], Chróścielewski et al. [119], Pimenta et al.
[367]). Therefore, it is quite understandable that it was also included by Sze et al. [455] in
their list of popular benchmark problems for geometric non-linear analysis of shells.
Recently also a composite variant of this example has been proposed by Masud et al. [309].
Fig. 5.33: Stretching of a short cylinder with free ends
The results obtained for the isotropic variant (E = 10500 ksi and ν = 0.3125) are
presented in Fig. 5.34 together with the reference solutions of Sansour & Bednarczyk [414],
Chroscielewski [118], Masud et al. [309], and Sze et al. [455].
LRT56 (mesh 5 x 8)
LRT56 (mesh 10 x 16)
LRT56 (mesh 15 x 24)
Sansour & Bednarczyk [414]
Chróœcielewski [118]
Masud et al. [309]
Sze et al. [455]
40
30
wA - wB
Load [106 lb]
20
10
0
0 1 2 3 4 5
Deflection [in]
Fig. 5.34: Radial deflections at the points A and B for the isotropic cylinder
5.7. Stretching of an open cylinder 127
Due to the symmetry of the problem, only one octant of the cylinder is modeled with
regular meshes of the 8-URI shell elements; the density of the FE discretization was
gradually refined, from 5×8 in the first mesh, through 10×16 in the second mesh to the final
mesh of 15×24 elements. The 10×16 mesh was dense enough to provide a convergent
LRT56 solution, which is in a quite good agreement with the results of Chróścielewski
[118], Sansour & Bednarczyk [414] and Sze et al. [455], while the solution given by Masud
et al. [309] remains noticeably different.
The next computations for the stretching of the free end cylinder are performed
following Masud et al. [309] for the laminated shell [0/90] with the following material
properties Ea = 30500 ksi, Eb = 10500 ksi, Gab = Gac = Gbc = 4000 ksi and νab = 0.3125.84
Geometry of the shell remains the same as given in Fig. 5.33. The radial displacement of
the loaded point A calculated with the LRT56 approach is compared in Fig. 5.35 with the
reference solution of Masud et al. [309] using different density of the mesh.
40 LRT56, 8-URI, mesh 5 x 8

LRT56, 8-URI, mesh 10 x 16
LRT56, 8-URI, mesh 15 x 24
Masud et al. [309], 256 elements
30 Masud et.al. [309], 384 elements
Load [106 lb]
20
10
0
0 1 2 3
Point A displacement [in]
Fig. 5.35: Radial deflections at the point A for the laminate cylinder [0/90]
One can notice that differences between the reference solutions for 256 and 384
elements appear at a quite low load level, whereas differences between 5x8 and 10x16
meshes of the LRT56 model are visible only in the vicinity of the snap-through region.
Increasing the density of the mesh in the LRT56 approach from 10x16 to 15x24 elements
resulted in almost no change of the response. In contrary to Masud et al. [309], using the
dense meshes in the LRT56 model a larger value of the snap-trough load is obtained than
for the coarse mesh (5x8).
Interesting observations can be made when the LRT56 solution for stretching of the
composite cylinder is compared with the results of RVK5, MRT5, LRT5 models (see
Fig. 5.36).
84
Please notice that the metric units used by Masud et al. [309] have been replaced here by USCS units to obtain
more realistic data.
RVK5, mesh 10x16

MRT5, mesh 10 x 16
LRT5, mesh 10 x 16
MRT56, mesh 10 x 16
LRT56, mesh 10 x 16
Masud et al. [309], 384 el.
40
30
Load [106 lb]
20
10
0
0 0.5 1 1.5 2 2.5 3
Displacement of point A [in]
Fig. 5.36: Displacement at the point A for different models of the laminate cylinder [0/90]
One can notice that the snap-through phenomenon is not manifested in the MRT5 or
LRT5 results, and both those models give stiffer responses than the LRT56 approach or the
reference solution of Masud et al. [309]. It seems to by quite accidentally that the MRT5
results are closer to the LRT56 solution than those of the LRT5 formulation. One can notice
a trace of the snap-through behavior in the equilibrium path obtained for the RVK5
formulation; nevertheless, this formulation is characterized by the stiffest response.
It is quite significant that by combining the MRT equations with the enhanced updating
of rotations (analogous to that used in the LRT56 approach) one can obtain a new model
marked here as MRT56, which provides a better estimation of the wA displacement than the
LRT5 approach. One can conclude that inclusion of additional terms in the LRT equations
is not as much relevant in the current example as the proper treatment of the rotational
degrees of freedom.
Another interesting observations can be made by comparing the results calculated for
the [0/90] ply lay-up with the solution obtained for the [90/0] laminate. As one can notice
by looking at the graphs presented in Fig. 5.37, the magnitude of the snap-through load for
stretching of the composite cylinder can be increased by 10% just by changing the stacking
sequence from [0/90] to [90/0].
5.7. Stretching of an open cylinder 129
40
wA - wC - wB
Laminate layout:
30 [0/90]
Load [106 lb] [90/0]
20
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Deflection [in]
Fig. 5.37: Influence of the lamina sequence on the response in the stretching test
5.8. Pinched hemispherical shell with 18° hole

The hemispherical shell with an 18 degree hole under two inward and two outward
forces as presented in Fig. 5.38 can be considered as probably the most popular benchmark
test for the finite rotation analysis of isotropic shells (see e.g. Stander et al. [446], Başar &
Krätzig [42], Saleeb et al. [412], Simo et al. [437], Büchter & Ramm [83, 84],
Chróścielewski et al. [120], Sansour & Bufler [417], Jiang & Chernuka [230], Sansour &
Boćko [415], Parisch [349], Park et al. [351], Sansour [413], Sansour & Bednarczyk [414],
Betsch et al. [57], Chróścielewski [118], Bischoff & Ramm [59], Betsch et al. [58],
Hauptmann & Schweizerhof [192], Zahrouni et al. [507], El-Abbasi & Meguid [154],
Masud et al. [309], Lee & Lee [280], Wang & Thierauf [483], Massin & Al Mikdad [307],
Fontes Valente et al. [167], Kulikov & Plotnikova [273], Vu-Quoc & Tan [479], Pimenta et
al. [367], Campello et al. [86], Chróścielewski et al. [119], Brank [74], Chróścielewski &
Witkowski [123]). Due to the high popularity of this example it seems obvious that it could
not be also omitted in the list of popular benchmark problems for geometric non-linear
analysis of shells given by Sze et al. [455]. We can also remind that the small deformation
analysis of this problem can be found in Section 4.6 of the present report.
The radius is R = 10 in and the thickness is h = 0.04 in. For the isotropic case the
material properties are E = 6825·104 psi and ν = 0.3. Due to the double symmetry of the
problem only one quarter of the shell is analyzed with appropriate boundary conditions (see
also Fig. 4.15).
The radial displacements of the isotropic shell computed within the LRT56 approach at
the points subjected to stretching and pinching forces, respectively, are presented in
Fig. 5.39 together with selected reference solutions.
Fig. 5.38: Large deformation of a hemispherical shell with 18 degree hole
Fig. 5.39: Radial deflections of the isotropic hemispherical shell with 18 degree hole
As one can observe in Fig. 5.39, the mesh of 15x15 8-URI elements (regularly
distributed in spherical coordinates) can provide an excellent agreement of the LRT56
results with the reference solutions of Chróścielewski et al. [120], Simo et al. [437], Stander
et al. [446], and Sze et al. [455]. The solution given by Saleeb et al. [412] seems to be just a
5.8. Pinched hemispherical shell with 18° hole 131
rough approximation to a convergent solution, similarly as the LRT56 results obtained with
a coarse mesh of 8x8 8-URI elements. On the other hand, the mesh of 12x12 8-URI
elements already offers an acceptable approximation to the correct solution.
Following the inspiration given in the previous example a composite variant of the
hemispherical shell with 18 degree hole is considered next. To keep the analyzed problem
as realistic as possible the material data have been adopted from the paper by Tsai et al.
[470] (compare Section 5.6): Ea = 20.46·106 psi, Eb = 4.092·106 psi, νab = 0.313,
Gab = Gac = 2.53704·106 psi, and Gbc = 1.26852·106 psi. To obtain a similar range of
deflections as for the isotropic case the shell thickness has been increased to h = 0.08 in.
Keeping in mind the hemispherical geometry of the analyzed shell it was assumed that the
fiber reinforcement (material a-axis of the composite) is running in the circumferential
direction (parallel to the equator). One should remember that the generating lines
(meridians) do not keep the interval (distance) – in the considered example the length of the
bottom edge of the shell is 3.24 times the length of the upper edge. The results of the
LRT56 computations performed with the regular mesh of 12x12 8-URI elements for the
composite hemisphere are presented in Fig. 5.40 and Fig. 5.41, respectively, for the inward
and outward radial deflections together with the solutions computed with other models.
Since no reference solutions for the analyzed problem are available in the literature, the
comparative calculations have been carried out in the MSC Nastran [318] computer code
using a 20x20 mesh of QUAD4 elements in a co-rotational formulation85.
200
RVK5
MRT5
LRT5
160 LRT56
MRT56
Loading force P [lb]
m-LRT5
120 MSC Nastran
80
40
0
0 1 2 3 4 5 6 7 8 9
Inward deflection [in]
Fig. 5.40: Inward radial deflection of the composite hemispherical shell
The inward deflection (Fig. 5.40) is almost twice as big as the outward displacement
(Fig. 5.41), and differences among curves obtained for various models are much more
visible than in the second graph (Fig. 5.41). The responses for the inward deflection
obtained with the MRT5 and LRT5 approaches are surprisingly similar; however their
approximation can be accepted only up to the load level of 20 lb. A better accuracy can be
85
See also Hoff [196], Hoff et al. [197] and MacNeal [299].
achieved for the RVK5 and the m-LRT5 models (up to 40 lb). The MRT56 approach allows
one to obtain a very good approximation up to 70 lb. A very good agreement between the
LRT56 model and the MSC/NASTRAN solution can be observed in Fig. 5.40.
200
RVK5
MRT5
LRT5
160 LRT56
MRT56
m-LRT5
Load force P [lb]
120 MSC Nastran
80
40
0
0 1 2 3 4 5
Outward deflection [in]
Fig. 5.41: Outward radial deflection of the composite hemispherical shell
Looking at the graphs of the outward radial deflections in Fig. 5.41 one can notice that
the interrelation between equilibrium paths for various models has significantly changed as
compared with the state in Fig. 5.40. Now the MRT5 model offers much better
approximation than the LRT5 formulation; however, the RVK5 results are still superior to
both those models and also to the MRT56 or m-LRT5. Here again the LRT56 solution
matches the results of the MSC Nastran.
The influence of the circular fibre reinforcement on the response of the composite
hemispherical shell is investigated in the parametric study illustrated in Fig. 5.42.
Fig. 5.42: Deflections for P = 60 lb as function of the orthotropy ratio

5.8. Pinched hemispherical shell with 18° hole 133
It was assumed that a constant set of material parameters Eb = 1.364·106 psi,

Gab = Gac = 0.84568·106 psi, Gbc = 0.42284·106 psi and νab = 0.313 is supplemented with
Ea = n · Eb, where n is the orthotropy ratio from the range between 5 and 30. Such an
assumption corresponds to an increasing contribution of the circular fibre reinforcement.
The values of the inward and outward deflections for the constant load level P = 60 lb are
presented in Fig. 5.42 as a function of the orthotropy ratio, n. A gradual increase of the
shell stiffness can be observed for the growing orthotropy ratio.
5.9. Axial compression of composite cylindrical panel

In the next numerical example an axial compression of a 16-layer composite
cylindrical panel is consider (see Fig. 5.43) assuming that the straight edges AB and CD are
simply supported with possibility to move along the generatrix, whereas the both curved
edges are clamped. The lamination scheme can be described as [45/-452/45/04]S. Each
lamina is made of carbon-epoxy composite XAS-914C with the following parameters:
Ea = 130·106 kPa, Eb = 10·106 kPa, νab = 0.3 and Gab = Gac = Gbc = 5·106 kPa. Geometry of
the panel is characterized by the thickness h = 16×0.125 = 2 mm, the radius R = 250 mm,
the length L = 540 mm and the opening angle β = 1.6848 rad.
Fig. 5.43: Composite cylindrical panel under axial compression
The origin of this example is referred to the experimental and numerical study by Snell
& Morley86 which was, however, not available to the author of the present report. The
buckling of this composite cylindrical panel was analyzed numerically also by Jun & Hong
[232], Laschet & Jeusette [276], Wagner [480] and Brank & Carrera [75]. Jun & Hong
[232] performed their non-linear buckling analysis using 8-node degenerated shell elements
within the Updated Lagrangian formulation. Laschet & Jeusette [276] presented results of
the linear and non-linear buckling analysis computed with three-dimensional degenerated
isoparametric multilayered 16-node finite elements (3 translational dofs per node). Wagner
[480] calculated the linear buckling load of the panel employing different meshes of 4-node
shell elements with reduced integration and hourglass control. Brank & Carrera [75]
86
Snell M., Morley N.: The compression buckling behaviour of highly curved panels of carbon fibre reinforced
plastic, Proceedings of 5th Int. Conf. on Composite Materials, (1985), 1327-1354.
applied 4-node mixed ANS shell elements based on the refined FOSD theory with finite
rotations. Results of the buckling analysis performed with MSC Nastran were presented by
Kreja [259].
It is quite symptomatic that descriptions of the analyzed panel given by the authors of
the four papers cited above are not quite consistent. There are some differences in the
interpretation of boundary conditions on the straight edges which are described as “simply
supported” – for instance Jun & Hong [232] and Wagner [480] constrained only radial and
circumferential translations at all nodes lying on the straight edges. However, due to the
isoparametric formulation of the applied finite elements this approach does not fix the
rotations about the normals to the edge. One can expect that the deformation of the panel
obtained in this model strongly depends on the number of nodes assumed along the straight
edges. Details of the boundary conditions applied by Laschet & Jeusette [276] are not clear
– just from the figure given in their paper one can expect that they applied an additional
row of shell elements on both sides of the panel. Brank & Carrera [75] admitted themselves
that they met some problems with a proper description of boundary conditions in that
example.
Another difficulty of this particular example is related to a proper representation of the
loading conditions. The graphs in Fig. 5.44 and Fig. 5.45 illustrate differences between the
results of two LRT56 models: in the first approach (model 1) the panel was compressed by
the axial load (pressure) uniformly distributed on the curved AD edge of the panel; in the
second version (model 2) a rigid movement of the whole edge AD was enforced.
160
120
Axial loading [kN]
80
40
model 1
LRT56, 20x16 8-URI
model 2
0
0 0.4 0.8 1.2 1.6
Axial displacement, u [mm]
Fig. 5.44: Axial deflection of the compressed cylindrical panel
As one can observe in Fig. 5.44 and Fig. 5.45, differences caused by those two
different interpretations of the loading conditions are visible only close to the buckling
stage and in the post-buckling paths. The maximal critical load obtained for model 2 (140.9
kN) is slightly larger than that calculated for model 1 (137.7 kN). The results of Laschet &
5.9. Axial compression of composite cylindrical panel 135
Jeusette [276] seem to suggest that those authors applied the pressure loading without any
additional enforcement of the rigid edge constraint.
160
stage III
stage II
120
Axial loading [kN]
stage IV
80
stage I
model 1
40 LRT56, 20x16 8-URI
model 2
0
-2 0 2 4 6 8
Central transverse deflection, w [mm]
Fig. 5.45: Central transverse deflection of the compressed cylindrical panel
Since the axial deflection presented in Fig. 5.44 for model 1 is in fact a deflection of
the central point of the edge AD, it is interesting to examine the distribution of the axial
deflection along that edge as presented in Fig. 5.46.
0.0
0.2
Deflection of the loaded edge [mm]
Prebuckling state P=137.8 kN

0.4
Postbuckling state P=97.5 kN
0.6
0.8
1.0
1.2
1.4
0.0 105.3 210.6 315.9 421.2
Distance along curved edge [mm]
Fig. 5.46: Axial displacements at the edge AD for model 1

By comparing the almost symmetrical shape of the edge AD at the pre-buckling state
with the asymmetrically deformed edge AD for the post-buckling state in Fig. 5.46, one can
conclude that the freedom of axial deformation of the edge AD in model 1 can stand behind
a more flexible response of model 1 in the post-buckling deformation phase with respect to
model 2 as manifested by the equilibrium paths in Fig. 5.44 and Fig. 5.45.
The buckling load calculated with the LRT56 formulation is in a quite good agreement
with reference solutions as it is shown in Table 5.3. The only exception is the solution of
Brank and Carrera [75] which noticeably differs from all the others. The difference with
respect to the experimental results is contained within the range of just several percents.
One can observe that the increase of a mesh density results in a decrease of the estimated
buckling load.
Table 5.3: Buckling load for the cylindrical panel with simply supported straight edges
Critical load [kN]

Model Mesh Linear Incremental
buckling analysis
model 1 10×8 - 138.9
LRT56 16×8 - 137.8
8-URI elements 20×16 - 137.7
model 2 10×8 - 143.5
LRT56 16×8 - 143.3
8-URI elements 20×16 - 140.9
8-node elements
8×10 - 143.2
Jun & Hong [232]
16-node elements 8×10 143.9 137.8

Laschet & Jeusette [276] 12×18 140.3 -
4×12 145.6 -
4×16 142.2 -
4-node elements
4×20 140.8 -
Wagner [480]
4×40 140.0 -
4×80 139.6 -
4-node elements
32×32 - 150
Brank & Carrera [75]
MSC Nastran QUAD4 20×20 144.56 144.35

elements 40×40 141.56 142.34
Kreja [259] 80×80 140.34 140.38
Snell & Morley Experiment 134
As the values of the critical load estimated in the linear buckling analysis are very
close to those obtained from the non-linear incremental analysis, one can conclude that the
pre-buckling deformations do not differ too much from the linear solution. That conclusion
can be also supported by the fact that negligible differences were obtained among values of
the critical load calculated with the formulations RVK5, MRT5, LRT5 and LRT56.
The deformation profiles of the central generatrix line are presented in Fig. 5.47 for the
selected deformation stages marked earlier in Fig. 5.45 with dots.
-1
1
Vertical deflection [mm]
4 Profiles:
initial
5 stage I
stage II
6 stage III
stage IV
7
0 90 180 270 360 450 540
Longitudinal coordinate [mm]
Fig. 5.47: Deformation profiles of the central generatrix line
The deformation profile of the stage III can be directly compared with the first
buckling mode of the panel; an excellent agreement can be observed between the
corrugated shape representing the stage III in Fig 5.47 and the buckling mode determined
within the MSC Nastran system for the mesh of 80×80 QUAD4 elements as presented in
Fig 5.48.
Fig. 5.48: The first buckling mode (MSC Nastran 80×80 QUAD4 elements, Pcrit = 140.34 kN)
A similarly good harmony can be observed between the deformation profiles for the
stages II and IV in Fig 5.47 and the deformation patterns obtained for the corresponding
deformation stages with the MSC Nastran system as shown in Fig. 5.49 and Fig. 5.50.
Fig. 5.49: Pre-buckling deformation at P=136.60 kN (MSC Nastran 80×80 QUAD4 elements)
Fig. 5.50: Post-buckling deformation at P=110.26 kN (MSC Nastran 80×80 QUAD4 elements)
One can easily notice an obvious similarity of the considered problem to the stability
analysis of an isotropic cylindrical panel under axial compression being the classical
illustration of the buckling problem with a non-symmetrical bifurcation point; compare e.g.
Waszczyszyn et al. [486], Chróścielewski et al. [119] and Ramm & Stegmüller [384].
Beside of different boundary conditions, the basic difference lies herein in a layered
structure of the panel. Due to a non-symmetrical lamination sequence applied in the
considered panel, the bending form of deformation appears since the very beginning of the
loading process, what is in contradiction with the isotropic case, where the transverse
deflection does not occur in the pre-buckling phase. Nevertheless, one can imagine that by
analogy to the classical procedure for the isotropic shells, a corresponding imperfection
sensitivity analysis can be performed also for the buckling of the layered panel.
A load imperfection was assumed in a form of an additional vertical force acting in the
centre of the panel, Pimp = ε P, with ε standing for the imperfection factor and P being the
total axial load. A positive value of the imperfection factor ε corresponds to the vertical
imperfection force Pimp directed downwards, whereas a negative sign of ε means the
imperfection force acting upwards. The calculations were performed for the imperfection
factor ε equal to -0.01, -0.005, -0.001, 0, 0.001, 0.005 and 0.01, with the results illustrated
in Fig. 5.51 as the graphs of the central radial deflection vs. the total axial load.
160
e =-0.001
e =-0.005
120 e =0.0 e =0.001
e =0.005
Axial loading [kN]
e =-0.01 e =0.010
80
40
0
-4 0 4 8 12
Central transverse deflection, w [mm]
Fig. 5.51: Imperfection sensitivity of the layered panel buckling
One can observe that the equilibrium paths presented in Fig. 5.51 for the positive
values of the imperfection factor are very similar to the analogous paths known in the
literature for the isotropic cylindrical panels (see e.g. Waszczyszyn et al. [486]). However,
in contrast to the isotropic case, introduction of the additional transverse load acting
upwards could not increase the limit load of the considered composite panel due to some
bifurcation points that appear in the equilibrium paths for the negative value of the
imperfection factor ε.
5.10. Buckling of composite cylindrical panels with square cut-outs

In the last example we analyze a similar 16-layer composite cylindrical panel as
considered above (see Fig. 5.43). The main differences lay in a centrally located square
hole and different boundary conditions at the straight edges AB and CD, which now remain
free of any support. The panel is axially compressed by enforcing a rigid movement of the
whole edge AD. Buckling of such panels made of graphite-epoxy composite AS4/3501-6
was examined by Chaplin & Palazotto [106] with the following material parameters:
Ea = 135.8·106 kPa, Eb = 10.9·106 kPa, Gab = Gac = 6.4·106 kPa, Gbc = 3.2·106 kPa and
νab = 0.276. Geometry of the panel is described by the radius R = 304.8 mm, the length
L = 508 mm and the opening angle β = 1 rad. The assumed layer stacking sequence is
[0/45/-45/90]2S with the total thickness of the panel h = 16×0.127 = 2.032 mm. The three
analyzed panels differ by sizes of the cut-outs, as presented in Fig. 5.52:
Case I - no cut-out;
Case II - central square cut-out 50.8 mm × 50.8 mm;
Case III - central square cut-out 101.6 mm × 101.6 mm.
D D D
A A A
u u u
C C C
B B B
Case I Case II Case III
Fig. 5.52: Analyzed exemplars of the axially loaded composite panels
The convergence study performed for the Case I has proved that the mesh of 12×20
8-URI elements provides almost identical response as the one obtained with the refined
discretization of 24×40 8-URI elements. However, since some singularities can be expected
in the two other cases due to the existence of cut-outs, the more dense mesh 24×40 has been
selected as the fundamental discretization pattern.
The equilibrium paths obtained with different formulations for the Case I are presented
in Fig. 5.53 together with the reference solution of Chaplin & Palazotto [106]. As one can
observe in Fig. 5.53 the plots obtained with the RVK5, MRT5, LRT5, m-LRT5 and LRT56
formulations are relatively close to each other. However, near the maximum load level it is
visible a discrepancy between the simplified solutions of the RVK5, MRT5 and m-LRT5
models and the results of the LRT56. On the other hand, the LRT5 model provides almost
the same solution as the LRT56. It is worth remaining that an analogous situation occurred
earlier in Section 5.4, and there it was explained by the influence of membrane strain terms
that were omitted in the MRT5 and RVK5 models. One should also notice that the m-LRT5
formulation includes the same membrane strain terms as the LRT56, but those two models
yield different responses in the current example due to the lack of the non-linear transverse
shear strain terms in the m-LRT5 formulation. It is quite interesting that the reference
solution given by Chaplin & Palazotto [106] agrees with the LRT56 results only up to the
load level of about 25 lb, and above that level the equilibrium path given by Chaplin &
Palazotto [106] is significantly different. The separation of the equilibrium paths of Chaplin
& Palazotto [106] and the LRT56 formulation has a character of the Y-junction, what can
suggest that the former experienced a jump from the fundamental equilibrium path to the
post-bifurcation equilibrium path. To verify this deduction a linear buckling problem has
also been solved in MSC/Nastran for the considered cylindrical shell. The five lowest
eigenvalues calculated for model A (24×40 QUAD4 elements) and model B (48×80
QUAD4 elements) are gathered in Table 5.4. The corresponding buckling modes obtained
for model A (24×40 QUAD4 elements) are presented in Fig. 5.54. The non-linear solution
obtained with the MSC/Nastran agrees quite well with the LRT56 as shown in Fig. 5.53.
5.10. Buckling of composite cylindrical panels with square cut-outs 141
60
50
Axial load [kN] 40
30
LRT56
20 LRT5
m-LRT5
MRT5
10 RVK5
Chaplin & Palazotto [106]
Nastran, 24x40 QUAD4
0
0 0.2 0.4 0.6 0.8 1 1.2
Axial displacement [mm]
Fig. 5.53: Axial deflection for Case I
Table 5.4: Linear buckling load for the cylindrical panel, Case I
Eigenvalues of the buckling load [kN]
N
Model A Model B
1 24.43 24.40
2 27.55 27.49
3 28.47 28.46
4 29.31 29.31
5 36.91 36.81
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5
Pcrit=24.43 kN Pcrit=27.55 kN Pcrit=28.47 kN Pcrit=29.31 kN Pcrit=36.91 kN
Fig. 5.54: Linear buckling modes obtained for Case I with 24×40 QUAD4 elements
Numbers presented in Table 5.4 show that in a case of a linear buckling analysis there
are very little differences between the results for the models A and B. It is also quite
characteristic that, on the contrary to the previous example, the lowest eigenvalue computed
in the linear buckling analysis for the panel with free straight edges (24.4 kN) is
significantly smaller than the critical load estimated in the incremental analysis (52.8 kN
for the LRT56 model). Looking again at the graphs in Fig. 5.53, one can observe that the
above-mentioned Y-junction is located at the load level near the lowest eigenvalue
determined in the linear buckling analysis (24.4 kN). This observation seems to support the
opinion that the graph given by Chaplin & Palazotto [106] does not represent the
(fundamental) equilibrium path for an ideal structure. Moreover, a closer look at the post-
buckling deformation obtained with the LRT56 model allows one to recognize a very close
similarity of that post-buckling deformation pattern and the fifth linear buckling mode, as
shown in Fig. 5.55.
Radial deflection [mm]
-5 0 5 10 Nastran:
0 buckling mode 5
D
E
A
127
Height coordinate [mm]
254
C
B F
381
LRT56 results:
edge AB
edge DC
central line EF
508
Fig. 5.55: Comparison of post-buckling deformation and the 5th linear buckling mode for Case I
To decisively verify the explanation suggested above of the discrepancy between the
solutions of Chaplin & Palazotto [106] and the LRT56, additional computations have been
performed where the axial load was supplemented with a very small load imperfection
taken as two radial forces acting outward in the middle of each straight edge of the panel
and being equal to 0.0001 fraction of the axial load. According to the images presented in
Fig. 5.54, such imperfection should correspond to the first buckling mode related with the
lowest eigenvalue of the buckling load.
As one can observe in Fig. 5.56, almost the same results were obtained with the LRT56
and MSC/Nastran for the load imperfection example.87 The fact that the LRT56 and
MSC/Nastran solutions for the load imperfection example almost match the graph of the
reference solution of Chaplin & Palazotto [106] entirely confirms, in author opinion, the
87
It is quite interesting that numerical round-off errors appearing in the MSC/Nastran computations performed in a
single precision mode can also act as a kind of imperfection directing the solution into the post-bifurcation path;
see e.g. Kreja [259].
hypothesis, that the solution of Chaplin & Palazotto [106] experienced a jump from the
fundamental equilibrium path to the post-bifurcation path related to the first buckling mode
of the panel. However, it is also important to remark that in a case of a structure that is as
strongly sensitive to imperfections as the analyzed panel, the results obtained for the ideal
structure should by no means be used to determine the load capacity.
60
50
40
Axial load [kN]
30
P1 = 24.4 kN
20
LRT56
10 Chaplin & Palazotto [106]
LRT56 with imperfection
Nastran with imperfection
0
0 0.2 0.4 0.6 0.8 1 1.2
Fig. 5.56: Influence of load imperfection on the axial displacement (Case I)
The five lowest eigenvalues obtained for the composite cylindrical panel with a central
square cut-out of 50.8 mm × 50.8 mm (Case II) in the linear buckling analysis performed
with the MSC/Nastran (regular mesh of 3776 QUAD4 elements88) are presented in Fig.
5.57 together with the corresponding buckling modes.
Fig. 5.57: Linear buckling modes obtained for Case II with 3776 QUAD4 elements
88
The mesh of 48x80 elements minus 8x8 elements for the opening.
By comparing the Figs 5.54 and 5.57 one can notice that the cut-off has not caused the
change of the first five buckling modes89.
The results of the non-linear analysis for the Case II are presented in Fig. 5.58 together
with the reference solution of Chaplin & Palazotto [106]. Here again, the pre-buckling
equilibrium paths obtained with different formulations are relatively close to each other;
however, the reference solution given by Chaplin & Palazotto [106] slightly overestimates
the buckling load (30 kN) as compared with the LRT56 solution (28.3 kN). In contradiction
to the previous example (Case I - the panel without a hole), a visible discrepancy can be
observed in the post-buckling region between the results of the LRT5 and LRT56 models.
It is quite unexpected, especially seeing that the simplified solution of the RVK5
formulation almost matches the path of the LRT56 model. On the other hand, the RVK5
solution is also in a very good agreement with the results of the MSC/Nastran non-linear
analysis performed with the regular mesh of 3776 QUAD4 elements.
35
30
25
Axial load [kN]
20
15 LRT56
LRT5
m-LRT5
10
MRT5
RVK5
5 Chaplin & Palazotto [106]
Nastran, 3776 QUAD4 el.
0
0 0.2 0.4 0.6 0.8 1
Fig. 5.58: Axial deflection for Case II
Contrary to the previously considered Case I, the post-buckling deformation estimated

for the Case II matches the first linear buckling mode as presented in Fig. 5.59. In such
circumstances, the panel with the cut-off should not be as much sensitive to imperfections
as in the Case I.
The solution of the linear buckling analysis for the composite cylindrical panel with a
central square cut-out of 101.6 mm × 101.6 mm (Case III) modeled in the MSC/Nastran
with the regular mesh of 3584 QUAD4 elements90 resulted in the first five buckling modes
presented in Fig. 5.60. By comparing the Figs 5.57 and 5.60 one can notice that the second
and the third modes interchanged after increasing dimensions of the cut-off.
89
The apparent dissimilarity of the 3rd and the 4th buckling modes for the Case I and the Case II resulted from the
opposite signs of the displacements.
90
The mesh of 48x80 elements minus 16x16 elements for the opening.

-5 0 5 10 15 Nastran:
0 buckling mode 1
D
E
A
127
254
C
B F
381
LRT56 results:
edge AB
edge DC
central line EF
508
Fig. 5.59: Comparison of post-buckling deformation and the 1st linear buckling mode for Case II
Fig. 5.60: Linear buckling modes obtained for Case III with 3584 QUAD4 elements
The non-linear equilibrium paths calculated for the Case III are illustrated in Fig. 5.61.
Now, the difference between the buckling load given by Chaplin & Palazotto [106] (23.16
kN) and the LRT56 solution (19.51 kN) is much bigger than in Case II. The formulations
LRT5 and LRT56 provided almost identical results as the MSC/Nastran model of 3584
QUAD4 elements. The results of the simplified RVK5 formulation are again closer to the
LRT56 solution than the results of the MRT5 model.
In the same way as in the Case II the post-buckling deformation calculated for the Case
III is very similar to the first linear buckling mode, as one can observe in Fig. 5.62. Here
again, a possible additional imperfection of the panel should not cause any qualitative
difference of the response.
25
20
Axial load [kN]

15
LRT56
10 LRT5
m-LRT5
MRT5
5 RVK5
Nastran, 3584 QUAD4 el.
0
0 0.2 0.4 0.6 0.8 1 1.2
Fig. 5.61: Axial displacement for Case III

-5 0 5 10 15 Nastran:
0 buckling mode 1
D
E
A
127
254
C
B F
381
LRT56 results:
edge AB
edge DC
central line EF
508
Fig. 5.62: Comparison of post-buckling deformation and the 1st linear buckling mode for Case III
In order to clearly show how the introduction of a central square cut-off can influence
the response to the axial compression of the considered composite cylindrical panel, the
equilibrium paths obtained with the LRT56 formulation for all three analyzed cases are
gathered in Fig. 5.63 together with the results of Chaplin & Palazotto [106]. Despite of
some quantitative differences, a visible qualitative similarity can be observed between the
results of Chaplin & Palazotto [106] and the LRT56 model. As one could anticipate, the
buckling load of the panel can be significantly reduced by the cut-off, and this reduction
progresses with the increase of the cut-off dimensions. However, looking at the graphs in
Fig. 5.63 one can also observe the change of the character of equilibrium paths for the
Cases I, II and III.
40
case I
30
case II
Axial load [kN]
20 case III
10
LRT56
0
0 0.2 0.4 0.6 0.8 1
Fig. 5.63: Axial displacement for all three cases

Chapter 6
CONCLUSIONS AND FUTURE PERSPECTIVES
6.1. Concluding Remarks

With the increasing application of laminated composites and sandwich panels in
various fields of structural engineering, there is a great concern about their appropriate
computational representation. The issue of a proper mathematical modeling of a mechanical
behavior of laminated plates and shells undergoing large rotations has constituted the main
topic of the present report.
A substantial part of the work has been devoted to a state-of-the-art review of
available methods for analysis of laminated plates and shells. The literature survey has been
focused on three major topics: theoretical models for multi-layered plates and shells,
geometrically non-linear analysis with a special emphasis on large rotation treatment, and
numerical implementation of various plate and shell theories. Almost five hundred texts
have been included in the survey with more than half of them devoted directly to the
problems of composite and/or sandwich structures. As the first conclusion from that review,
one could notice a lack of a single numerical model capable for a universal representation
of all layered composite and sandwich panels; depending on the particular problem
different formulations could be recognized as the most effective ones. Due to obvious
limitations of the computational potential, the degree of complexity of the through-the-
thickness representation of deformation profiles decreased as a rule with the increase of the
range of rotations included in the particular model. One could also observe that the
collection of benchmark problems available in the literature for laminated shells undergoing
large rotations is very limited.
In the next part of the report a systematic construction of a numerical model for a
geometrically non-linear analysis of elastic laminated shells undergoing large rotations has
been presented adopting the Equivalent Single Layer (ESL) approach within the scope of
the First Order Shear Deformation (FOSD) theory. Additionally, a smooth geometry of the
middle surface without kinks and branches has been assumed. The strain-displacement
relations established for the large rotation shell theory (LRT) have been consistently
tailored according to the assumption of the director inextensibility with an appropriate
treatment of finite rotations. Afterwards a set of various approximated variants of the strain-
displacement formulae has been considered and confronted with relationships proposed by
other authors. Consequently, the series of large deformation formulations for laminated
shells have been constructed with gradually increased and clearly distinguished levels of
geometrical non-linearity, from the Refined von Kármán Theory (RVK5), through the
Moderate Rotation Theory (MRT5) and the (Simplified) Large Rotation Theory with 5 dofs
(LRT5) to the Large Rotation Theory (LRT56) formulation with an enhanced treatment of
finite rotations based on the application of Euler angles.
Starting with the Total Lagrangian approach the incremental strategy of the large
deformations analysis has been established with the Newton-Raphson equilibrium iterations
and the Riks-Wempner-Ramm arc-length control method. Standard quadrilateral finite
elements based on the Lagrange and Serendipity interpolation schemes have been
considered for the possible application in the computer implementation of the proposed
6.1. Concluding Remarks 149
algorithms. After a set of numerical tests performed within the range of small
displacements for various integration and interpolation schemes (including the Assumed
Natural Strain approach), the 8-node Serendipity shell element with a uniformly reduced
integration has been selected as the most promising choice for applications in the large
deformation analysis.
The largest part of the present report contains a comparative finite element analysis of
various sample problems of a non-linear, large rotation response of composite laminated
plate and shell structures. Everyone from the ten examples selected from the literature has
been analyzed with the LRT56 formulation and in all cases this model was positively
validated by confronting the obtained results with the available reference solutions.
Additionally, the results calculated with the simplified non-linear models, like the RVK5,
MRT5 and LRT5 have been included in the comparison. Furthermore, some selected
examples have been analyzed also with the modified formulations, as the m-LRT5 and m-
LRT5691 or MRT56.92.
A detailed examination of the obtained results leads to the following conclusions:
a) the proper updating procedure of the rotations is of extreme importance as soon as
the range of moderate rotations is exceeded; as shown in example 5.7, the proper
treatment of the rotational degrees of freedom can be much more relevant to the
accuracy of the results than the inclusion of all terms in the LRT equations;
b) the results obtained with the simplified large rotation formulation (LRT5) are
acceptable only for small and moderate rotation problems; a more accurate
representation of the strains in the LRT5 formulation as compared to the MRT5
model very seldom resulted in a better accuracy of results; compare e.g. examples
5.1, 5.5 and 5.7;
c) the refined representation of the transverse shear strains incorporated in the TOSD
model is less essential for the accuracy of the large rotation solution for moderately
thick composite shells than the proper updating of the rotational degrees of freedom;
see e.g. examples 5.5 and 5.6;
d) modified approaches, that use approximate strain-displacement relations by
neglecting all non-linear terms in the transverse shear strain-displacement relations
as the m-LRT5 and m-LRT56 models, cannot be positively validated; see examples
5.1 and 5.6;
e) paradoxically, the most simplified non-linear model of the RVK5 formulation
surprisingly often provided better solutions than the more elaborated models of the
MRT5 and the LRT5; check examples 5.1, 5.2, 5.6, 5.8 and 5.10;
f) by confronting the results obtained with the gradually increased and clearly
distinguished levels of geometrical non-linearity, it has been possible to indicate the
examples that have qualified to be benchmark problems for laminated shells
undergoing large rotations; see examples 5.1, 5.2, 5.5, 5.7 and 5.8;
g) a quite unexpected behavior of the clamped cylindrical composite panel can be
observed for some lamination schemes, where the reduction of the overall stiffness
can be accompanied quite paradoxically by the increase of the critical snapping load
(see Example 5.6).
91
The two modified models m-LRT5 and m-LRT56 have been created just by dropping all non-liner terms for the
transverse shear strains in the LRT5 and LRT56 formulations, respectively; compare the descriptions given in
Section 5.1 and Section 5.6.
92
The MRT56 model is based on the MRT equations supplemented with the enhanced updating of rotations
(analogous to that used in the LRT56 approach); see Section 5.7 for a more detailed explanation.
150 Chapter 6. Conclusions and future perspectives
6.2. Original Contribution

As far as the author is concerned, the following aspects can be considered as original
contribution of this report:
a) the systematic construction of a computational model for a large rotation analysis
of elastic laminated shells with a special emphasis to the practical finite element
implementation in a family of the self-developed computer programs;
b) the methodical examination on the consequences of various approximation
decisions concerning the description of large rotations;
c) the selection of examples that meet the requirements of benchmark problems for
laminated shells undergoing large rotations;
d) the comprehensive state-of-the-art literature survey on available methods of
analysis for laminated plates and shells.
6.3. Recommendations and Future Perspectives

The present report has demonstrated that the proposed LRT56 formulation can be
considered as a valuable tool for a geometrically nonlinear analysis of moderately thick,
laminated elastic shells undergoing large rotations; however, there is still much work that
can be pursued in this area. The following is a suggestion of research directions that may be
identified as interesting and potentially profitable:
a) taking into consideration geometrical imperfections and thermal effects, especially
in the context of the buckling analysis;
b) construction of special transition finite-elements that allow to connect the 2-D and
3-D models in a global-local analysis;
c) application of modern low-order finite elements based on the ANS concept;
d) extension of the analysis to account for the materially non-linear problems;
e) developing procedures for predicting composite damage like matrix cracking, fiber
failure and delamination;
f) investigation of composites built as functionally graded materials (FGM);
g) analysis of electro-mechanical coupling in smart laminated structures with
piezoelectric layers.
Chapter 7
REFERENCES
1 Abu-Farsakh, G.A., Barakat, S.A. & Al-Zoubi, N.R.: Effect of material nonlinearity in
unidirectional composites on the behavior of beam structures, Int. Journal of Solids &
Structures 37, 2000, 2673-2694.
2 Ahmad, S., Irons, B. M. & Zienkiewicz, O. C.: Analysis of thick and thin shell structures by
curved finite elements, Int. Journal for Numerical Methods in Engineering 2, 1970, 419-451.
3 Aitharaju, V. R. & Averill, R. C.: C0 zigzag kinematic displacement models for the analysis of
laminated composites, Mechanics of Composite Materials & Structures 6, 1999, 31-56.
4 Alfano, G., Auricchio, F., Rosati, L. & Sacco, E.: MITC finite elements for laminated
composite plates, Int. Journal for Numerical Methods in Engineering 50, 2001, 707-738.
5 Ali, R.: Use of finite element technique for the analysis of composite structures, Computers &
Structures 58, 1996, 1015-1023.
6 Altenbach, H.: On the determination of transverse shear stiffnesses of orthotropic plates,
Zeitschrift für angewandte Mathematik und Physik (Journal of Applied Mathematics and
Physics) - ZAMP 51, 2000, 629-649.
7 Altenbach, H.: An alternative determination of transverse shear stiffnesses for sandwich and
laminated plates, Int. Journal of Solids & Structures 37, 2000, 3503-3520.
8 Altenbach, J. & Altenbach, H.: Trends in engineering plate theories, Eksploatacja i
Niezawodność (Maintenance & Reliability, Polish Academy of Sciences Quarterly) 4/2001,
2001, 21-30.
9 Altman, W. & Palmerio, A. F.: A refined theory for anisotropic shells in the presence of small
strains and moderate rotations, Int. Journal of Non-Linear Mechanics 26, 1991, 609-618.
10 Altmann, S. L.: Rotations, Quaternions and Double Groups, Clarendon Press, Oxford, 1986.
11 Altmann, S. L.: Clifford algebra, symmetries, and vectors, Int. Journal of Quantum Chemistry
60, 1996, 359-372.
12 Amabili, M. & Paidoussis, M. P.: Review of studies on geometrically nonlinear vibrations and
dynamics of circular cylindrical shells and panels, with and without fluid-structure interaction,
Applied Mechanics Reviews 56, 2003, 349-381.
13 Ambartsumyan, S. A.: Contributions to the theory of anisotropic layered shells, Applied
Mechanics Reviews 15, 1962, 245-249.
14 Ambartsumyan, S. A.: Theory of Anisotropic Plates, Technomic Publishing Co., Inc.,
Stamford, 1970.
15 Andelfinger, U. & Ramm, E.: An assessment of hybrid-mixed four-node shell elements, in
[388], 1992, 31-45.
16 Andelfinger, U. & Ramm, E.: EAS-elements for two-dimensional, three-dimensional, plate and
shell structures and their equivalence to HR elements, Int. Journal for Numerical Methods in
Engineering 36, 1993, 1311-1337.
17 Argyris, J. H.: Matrix displacement analysis of plates and shells, Prolegomena to a General
Theory, Part I, Ingenieur-Archiv 35, 1966, 102-142.
18 Arnold, D. N.: Questions on Shell Theory, Workshop on "Elastic Shells: Modeling, Analysis,
and Computation", Mathematical Sciences Research Institute, Berkeley, CA, April 21, 2000
(Available at http://www.ima.umn.edu/~arnold/talks/shells.pdf - last visited on December 15,
2006)
152 Chapter 7. References
19 Arnold, D. N., Boffi, D. & Falk, R.S.: Remarks on Quadrilateral Reissner-Mindlin Plate
Elements, WCCM V: Proceedings of the Fifth World Congress on Computational Mechanics,
July 7-12, 2002, Vienna, Austria, H.A. Mang, F.G. Rammerstorfer & J. Eberhardsteiner (eds.),
CD-ROM, Vienna 2002.
20 Arnold, D. N. & Brezzi, F.: The partial selective reduced integration method and applications
to shell problems, Computers & Structures 64, 1997, 879-880.
21 Arnold, D. N. & Brezzi, F.: Locking-free finite element methods for shells, Mathematics of
Computation 66, 1997, 1-14.
22 Arnold, D. N., Madureira, A. & Zhang, S.: On the range of applicability of the Reissner–
Mindlin and Kirchhoff–Love plate bending models, Journal of Elasticity 67, 2002, 171-185.
23 Arya, H., Shimpi, R. P. & Naik, N. K.: A zigzag model for laminated composite beams,
Composite Structures 56, 2002, 21-24.
24 Auricchio, F. & Lovadina, C.: Partial selective reduced integration schemes and kinematic
linked interpolations for plate bending problems, Mathematical Models and Methods in
Applied Sciences 9, 1999, 693-722.
25 Auricchio, F. & Lovadina, C.: Analysis of kinematic linked interpolation methods for
Reissner-Mindlin plate problems, Computer Methods in Applied Mechanics & Engineering
190, 2001, 2465-2482.
26 Auricchio, F. & Sacco, E.: Partial-mixed formulation and refined models for the analysis of
composite laminates within an FSDT, Composite Structures 46, 1999, 103-113.
27 Auricchio, F. & Sacco, E.: A mixed-enhanced finite-element for the analysis of laminated
composite plates, Int. Journal for Numerical Methods in Engineering 44, 1999, 1481-1504.
28 Auricchio, F. & Sacco, E.: Refined First-Order Shear Deformation Theory models for
composite laminates, Journal of Applied Mechanics, Trans. ASME 70, 2003, 381-390.
29 Auricchio, F. & Taylor, R. L.: A shear-deformable plate element with an exact thin limit,
Computer Methods in Applied Mechanics & Engineering 118, 1994, 393-412.
30 Axelrad, E. L.: Shell theory and its specialized branches, Int. Journal of Solids & Structures
37, 2000, 1425-1451.
31 Balah, M. & Al-Ghamedy, H. N.: Finite element formulation of a third order laminated finite
rotation shell element, Computers & Structures 80, 2002, 1975-1990.
32 Barbero, E. J., Reddy, J. N. & Teply, J.: An accurate determination of stresses in thick
laminates using a generalized plate theory, Int. Journal for Numerical Methods in Engineering
29, 1990, 1-14.
33 Barut, A., Madenci, E. & Tessler, A.: Nonlinear analysis of laminates through a Mindlin-type
shear deformable shallow shell element, Computer Methods in Applied Mechanics &
Engineering 143, 1997, 155-173.
34 Başar, Y.: A consistent theory of geometrically non-linear shells with an independent rotation
vector, Int. Journal of Solids & Structures 23, 1987, 1401-1413.
35 Başar, Y.: Finite-rotation theories for composite laminates, Acta Mechanica 98, 1993, 159-176.
36 Başar, Y. & Ding, Y.: Interlaminar stress analysis of composites: layer-wise shell finite
elements including transverse strains, Composites Engineering 5, 1995, 485-499.
37 Başar, Y., Ding, Y. & Schultz, R.: Refined shear-deformation models for composite laminates
with finite rotations, Int. Journal of Solids & Structures 30, 1993, 2611-2638.
38 Başar, Y. & Grytz, R.: Incompressibility at large strains and finite-element implementation,
Acta Mechanica 168, 2004, 75-101.
39 Başar, Y., Hanskötter, U. & Schwab, Ch.: A general high-order finite element formulation for
shells at large strains and finite rotations, Int. Journal for Numerical Methods in Engineering
57, 2003, 2147-2175.
7. References 153
40 Başar, Y., Itskov, M. & Eckstein, A.: Composite laminates: nonlinear interlaminar stress
analysis by multi-layer shell elements, Computer Methods in Applied Mechanics &
Engineering 185, 2000, 367-397.
41 Başar, Y. & Kintzel, O.: Finite rotations and large strains in Finite Element shell analysis,
Computer Modeling in Engineering and Sciences 4, 2003, 217-230.
42 Başar, Y. & Krätzig, W. B.: Introduction into finite-rotation shell theories and their operator
formulation, Computational Mechanics of Nonlinear Response of Shells, W. B. Krätzig and E.
Oñate (eds.), ICES'88 Conference, Atlanta 1988, Springer Verlag, Berlin, 1990, 3-30.
43 Başar, Y. & Krätzig, W. B.: Theory of Shell Structures, VDI Verlag GmbH, Düsseldorf, 2000.
44 Başar, Y. & Weichert, D.: Nonlinear Continuum Mechanics of Solids, Springer Verlag, Berlin,
2000.
45 Bathe, K. -J.: Some advances in finite element procedures for nonlinear structural and thermal
problems, Computational Mechanics - Advances and Trends, Winter Annual Meeting of ASME,
Anaheim 1986, A. K. Noor (ed.), AMD - Vol.75, 1986, 183-216.
46 Bathe, K. -J.: Finite Element Procedures, Prentice Hall, Englewood Cliffs, New Jersey, 1996.
47 Bathe, K. -J. & Bolourchi, S.: A geometric and material nonlinear plate and shell element,
Computers & Structures 11, 1980, 23-48.
48 Bathe, K. -J. & Dvorkin, E.: A formulation of general shell elements - the use of mixed
interpolation of tensorial components, Int. Journal for Numerical Methods in Engineering 22,
1986, 697-722.
49 Bathe, K. -J., Iosilevich, A. & Chapelle, D.: An evaluation of the MITC shell elements,
50 Bauchau, O. A. & Trainelli, L.: The vectorial parameterization of rotation, Nonlinear
Dynamics 32, 2003, 71-92.
51 Belytschko, T.: A review of recent developments in plate and shell elements, Computational
Mechanics - Advances and Trends, Winter Annual Meeting of ASME, Anaheim 1986,
A. K. Noor (ed.), AMD - Vol.75, 1986, 217-230.
52 Belytschko, T., Liu, W. K. and Moran, B.: Nonlinear Finite Elements for Continua and
Structures, John Wiley & Sons, Chichester 2000.
53 Belytschko, T., Ong, J. S. -J., Liu, W. K. & Kennedy, J. M.: Hourglass control in linear and
nonlinear problems, Computer Methods in Applied Mechanics & Engineering 44, 1984, 251-
276.
54 Belytschko, T., Stolarski, H., Liu, W. K., Carpenter, N. & Ong, J. S. -J.: Stress projection for
membrane and shear locking in shell finite elements, Computer Methods in Applied Mechanics
& Engineering 51, 1985, 221-258.
55 Bercha, F. G. & Glockner, P. G.: Thick shell and oriented surface theories, Journal of
Engineering Mechanics ASCE 98, 1972, 823-833.
56 Bercha, F. G. & Glockner, P. G.: Thick shell stress resultants, Journal of Engineering
Mechanics ASCE 98, 1972, 757-762.
57 Betsch, P., Gruttmann, F. & Stein, E.: A 4-node finite element for the implementation of
general hyperelastic 3D-elasticity at finite strains, continuum based three-dimensional shell
element for laminated structures, Computer Methods in Applied Mechanics & Engineering
130, 1996, 57-79.
58 Betsch, P., Menzel, A. & Stein, E.: On the parametrization of finite rotations in Computational
Mechanics - A classification of concepts with application to smooth shells, Computer Methods
in Applied Mechanics & Engineering 155, 1998, 273-305.
59 Bischoff, M. & Ramm, E.: Shear deformable shell elements for large strain and rotations, Int.
Journal for Numerical Methods in Engineering 40, 1997, 4427-4449.
60 Bischoff, M. & Ramm, E.: On the physical significance of higher order kinematic and static
variables in a three-dimensional shell formulation, Int. Journal of Solids & Structures 37,
2000, 6933-6960.
61 Bischoff, M., Ramm, E. & Braess, D.: A class of equivalent enhanced assumed strain and
hybrid stress finite elements, Computational Mechanics 22, 1999, 443-449.
62 Bischoff, M. & Taylor, R. L.: A three-dimensional shell element with an exact thin limit,
ECCM-2001, European Conference on Computational Mechanics, June 26-29, 2001, Cracow,
Poland.
63 Bischoff, M., Wall, W. A., Bletzinger, K.-U. & Ramm, E.: Models and Finite Elements for
Thin-walled structures, Encyclopedia of Computational Mechanics, Vol. 2: Solids and
Structures, E. Stein, R. de Borst & T. J. R. Hughes (eds.), John Wiley & Sons, Ltd., London
2004, 59-137.
64 Bletzinger, K.-U., Bischoff, M. & Ramm, E.: A unified approach for shear-locking-free
triangular and rectangular shell finite elements, Computers & Structures 75, 2000, 321-334.
65 Bogdanovich, A.E. & Yushanov, S.P.: Three-dimensional variational analysis of Pagano's
problems for laminated composite plates, Composites Science & Technology 60, 2000, 2407-
2425.
66 Borsellino, C., Calabrese, L. & Valenza, A.: Experimental and numerical evaluation of
sandwich composite structures, Composites Science & Technology 64, 2004, 1709-1715.
67 Bose, P. & Reddy, J. N.: Analysis of composite plates using various plate theories Part 1:
Formulation and Analytical Solutions, Structural Engineering & Mechanics 6, 1998, 583–612.
68 Bose, P. & Reddy, J. N.: Analysis of composite plates using various plate theories Part 2:
Finite element model and numerical results, Structural Engineering & Mechanics 6, 1998,
727–746.
69 Bouhafs, B., Woźnica, K. & Kłosowski, P.: The large rotations theory of elasto-viscoplastic
shells subjected to the dynamic and thermal loads, Engineering Computations 20, 2003, 366-
389.
70 Bödefeld, J., Kreja, I. & Schmidt, R.: Large deformation static analysis of axisymmetric
orthotropic plates and shells, Shell Structures: Theory and Applications, Proc. the 5th Conf.
SSTA, Janowice, 15-18 October 1992, R. Nagórski (ed), Warsaw Univ. Technol. Press,
Warsaw 1992, 55-56.
71 Böhm, H. J. (Ed.): Mechanics of Microstructured Materials, CISM Courses and Lectures No.
464, Springer-Verlag, Vienna 2004.
72 Bramble, J.H. & Tong, S.: A locking-free finite element method for Nagdhi shells, Journal of
Computational and Applied Mathematics 89, 1998, 119-133.
73 Brank, B.: On composite shell models with piecewise linear warping function, Composite
Structures 59, 2003, 163-171.
74 Brank, B.: Nonlinear shell models with seven kinematic parameters, Computer Methods in
Applied Mechanics & Engineering 194, 2005, 2336-2362. (see [387])
75 Brank, B. & Carrera, E.: A family of shear-deformable shell finite elements for composite
structures, Computers & Structures 76, 2000, 287-297.
76 Brank, B. & Carrera, E.: Multilayered shell finite element with interlaminar continuous
stresses: a refinement of the Reissner-Mindlin formulation, Int. Journal for Numerical Methods
in Engineering 48, 2000, 843-874.
77 Brank, B. & Ibrahimbegović, A.: On the relation between different parametrizations of finite
rotations for shells, Engineering Computations 18, 2001, 950-973.
78 Brank, B., Korelc, J. & Ibrahimbegović, A.: Nonlinear shell problem formulation accounting
for through-the-thickness stretching and its finite element implementation, Computers &
7. References 155
Structures 80, 2002, 699-717.

79 Brank, B., Perić, D. & Damjanić, F. B.: On implementation of a nonlinear four node shell finite
element for thin multilayered elastic shells, Computational Mechanics 16, 1995, 341-359.
80 Brank, B., Perić, D. & Damjanić, F. B.: On large deformations of thin elasto-plastic shells:
implementation of a finite rotation model for quadrilateral shell element, Int. Journal for
Numerical Methods in Engineering 40, 1997, 689-726.
81 Braun, M., Bischoff, M. & Ramm, E.: Nonlinear 3-dimensional analysis of composite and
laminate plate and shell structures, Recent Developments in Finite Element Analysis, T. J. R.
Hughes, E. Oñate & O. C. Zienkiewicz (ed.), CIMNE, Barcelona, 1994, 215-224.
82 Büchter, N. & Ramm, E.: Comparison of shell theory and degeneration, in [388], 1992, 15-30.
83 Büchter, N. & Ramm, E.: Large rotations in structural mechanics - overview, in [388], 1992, 1-
13.
84 Büchter, N. & Ramm, E.: Shell theory versus degeneration – A comparison in large rotation
finite element analysis, Int. Journal for Numerical Methods in Engineering 34, 1992, 39-59.
85 Burton, W. S. & Noor, A. K.: Assessment of computational models for sandwich panels and
shells, Computer Methods in Applied Mechanics & Engineering 124, 1995, 125-151.
86 Campello, E. M. B., Pimenta, P. M. & Wriggers, P.: A triangular finite shell element based on
a fully nonlinear shell formulation, Computational Mechanics 31, 2003, 505-518.
87 Carrera, E.: C0 Reissner-Mindlin multilayered plate elements including zig-zag and
interlaminar stress continuity, Int. Journal for Numerical Methods in Engineering 39, 1996,
1797-1820.
88 Carrera, E.: Cz0 requirements—models for the two dimensional analysis of multilayered
structures, Composite Structures 37, 1997, 373-383.
89 Carrera, E.: Mixed layer-wise models for multilayered plates analysis, Composite Structures
43, 1998, 57-70.
90 Carrera, E.: A priori vs. a posteriori evaluation of transverse stresses in multilayered
orthotropic plates, Composite Structures 48, 2000, 245-260.
91 Carrera, E.: An assessment of mixed and classical theories on global and local response of
multilayered orthotropic plates, Composite Structures 50, 2000, 183-198.
92 Carrera, E.: Developments, ideas, and evaluations based upon Reissner’s Mixed Variational
Theorem in the modeling of multilayered plates and shells, Applied Mechanics Reviews 54,
2001, 301-329.
93 Carrera, E.: Historical review of zig-zag theories for multilayered plates and shells, Applied
Mechanics Reviews 56, 2003, 287-308.
94 Carrera, E.: On the use of the Murakami’s zig-zag function in the modeling of layered plates
and shells, Computers & Structures 82, 2004, 541-554.
95 Carrera, E. & Demasi, L.: Classical and advanced multilayered plate elements based upon
PVD and RMVT. Part 1: Derivation of finite element matrices, Int. Journal for Numerical
Methods in Engineering 55, 2002, 191-231.
96 Carrera, E. & Demasi, L.: Classical and advanced multilayered plate elements based upon
PVD and RMVT. Part 2: Numerical implementations, Int. Journal for Numerical Methods in
Engineering 55, 2002, 253-291.
97 Carrera, E. & Parisch, H.: An evaluation of geometrical nonlinear effects of thin and
moderately thick multilayered composite shells, Composite Structures 40, 1998, 11-24.
98 Carvelli, V. & Savoia, M.: Assessment of plate theories for multilayered angle-ply plates,
99 Cen, S., Yuqiu Long, Y. & Yao, Z.: A new hybrid-enhanced displacement-based element for
the analysis of laminated composite plates, Computers & Structures 80, 2002, 819-833.
100 César de Sá, J. M. A., Natal Jorge, R. M., Fontes Valente, R. A. & Almeida Areias, P. M.:
Development of shear locking-free shell elements using an enhanced assumed strain
formulation, Int. Journal for Numerical Methods in Engineering 53, 2002, 1721–1750.
101 Chandrashekhara, K. & Pavan Kumar, D. V. T. G.: Assessment of shell theories for the static
analysis of cross-ply laminated circular cylindrical shells, Thin-Walled Structures 22, 1995,
291-318.
102 Chandrashekhara, K. & Pavan Kumar, D. V. T. G.: Static response of composite circular
cylindrical shells studied by different theories, Meccanica 33, 1998, 11-27.
103 Chang, T. Y., Saleeb, A. F. & Graf, W.: On the mixed formulation of a 9-node Lagrange shell
element, Computer Methods in Applied Mechanics & Engineering 73, 1989, 259-281.
104 Chang, T. Y. & Sawamiphakdi, K.: Large deformation analysis of laminated shells by finite
element method, Computers & Structures 13, 1981, 331-340.
105 Chapelle, D. & Bathe, K.J.: The Finite Element Analysis of Shells - Fundamentals, Springer
Verlag, Berlin 2003.
106 Chaplin, C. P. & Palazotto, A. N.: The collapse of composite cylindrical panels with various
thickness using Finite Element Analysis, Computers & Structures 60, 1996, 797-815.
107 Chaudhuri, R. A.: Analysis of laminated shear-flexible angle-ply plates, Composite Structures
67, 2005, 71-84.
108 Chaudhuri, R. A. & Seide, P.: An approximate method for prediction of transverse shear
stresses in a laminated shell, Int. Journal of Solids & Structures 23, 1987, 1145-1161.
109 Chen, C.-M., Kikuchi, N. & Farzad R.-A.: An enhanced asymptotic homogenization method of
the static and dynamics of elastic composite laminates, Computers & Structures 82, 2004, 373-
382.
110 Cheng, H. & Gupta, K. C.: An historical note on Finite Rotations, Journal of Applied
Mechanics, Trans. ASME 56, 1989, 139-145.
111 Cheung, Y. K. & Chen, W.: Generalized hybrid degenerated elements for plates and shells,
112 Chinosi, C., Della Croce, L. & Scapolla, T.: Hierarchic finite elements for thin Naghdi shell
model, Int. Journal of Solids & Structures 35, 1998, 1863-1880.
113 Cho, M. & Kim, J.-S.: A postprocess method for laminated shells with a doubly curved nine-
noded finite element, Composites, Part B 31, 2000, 65-74.
114 Cho, M., Kim, K.-O. & Kim, M.-H.: Efficient higher-order shell theory for laminated
composites, Composite Structures 34, 1996, 197-212.
115 Cho, M. & Kim, M.-H.: A postprocess method using a displacement field of higher-order shell
theory, Composite Structures 34, 1996, 185-196.
116 Cho, Y.B. & Averill, R.C.: First-order zig-zag sublaminate plate theory and finite element
model for laminated composite and sandwich panels, Composite Structures 50, 2000, 1-15.
117 Chou, T.-W. & Kelly, A.: Mechanical properties of composites, Annual Review of Materials
Science 10, 1980, 22–59.
118 Chróścielewski, J.: Rodzina elementów skończonych klasy C0 w nieliniowej
sześcioparametrowej teorii powłok (Family of C0 finite elements in six-parameter non-linear
theory of shells), (in Polish), Zeszyty Naukowe Politechniki Gdańskiej (Scientific Bulletin of
GUT) 540, 1996, 1-291.
119 Chróścielewski, J., Makowski, J. & Pietraszkiewicz, W.: Statyka i dynamika powłok
wielopłatowych: Nieliniowa teoria i metoda elementów skończonych (Statics and dynamics of
multi-shell structures: Non-linear theory and FEM), (in Polish), Institute of Fundamental
Technological Research, PAS Warszawa, 2004.
120 Chróścielewski, J., Makowski, J. & Stumpf, H.: Genuinely resultant shell finite elements
7. References 157
accounting for geometric and material non-linearity, Int. Journal for Numerical Methods in
Engineering 35, 1992, 63-94.
121 Chróścielewski, J., Makowski, J. & Stumpf, H.: Finite element analysis of smooth, folded and
multi-shell structures, Computer Methods in Applied Mechanics & Engineering 141, 1997, 1-
46.
122 Chróścielewski, J. & Schmidt, R.: A solution control method for nonlinear finite element post-
bucking analysis of structures, Proc. EUROMECH-Colloquium 200, J. Szabo (ed.),
Matrafüred, Hungary 1985, 19-33.
123 Chróścielewski, J. & Witkowski, W.: Four-node semi-EAS element in six-field nonlinear
theory of shells, Int. Journal for Numerical Methods in Engineering 68, 2006, 1137–1179.
124 Ciarlet, P.G. & Gratie, L.: On the existence of solutions to the generalized Marguerre–von
Kármán equations, Mathematics and Mechanics of Solids 11, 2006, 83 –100.
125 Ciarlet, P.G. & Lods, V.: Asymptotic analysis of linearly elastic shells. I. Justification of
membrane shell equations, Archive for Rational Mechanics & Analysis 136, 1996, 119-161.
126 Ciarlet, P.G., Lods, V. & Miara, B.: Asymptotic analysis of linearly elastic shells. II.
Justification of flexural shell equations, Archive for Rational Mechanics & Analysis 136, 1996,
163-190.
127 Ciarlet, P.G. & Lods, V.: Asymptotic analysis of linearly elastic shells. III. Justification of
Koiter's shell equations, Archive for Rational Mechanics & Analysis 136, 1996, 191-200.
128 Clough, R. W. & Wilson E. L.: Early Finite Element research at Berkeley, Presentation at the
Fifth U.S. National Conference on Computational Mechanics, Aug. 4-6, 1999 (available at
http://www.edwilson.org/History/fe-history.pdf - last visited on December 15, 2006)
129 Crisfield, M. A.: Non-linear Finite Element Analysis of Solids and Structures, John Wiley &
Sons, Chichester 1997.
130 Crisfield, M.A., Jelenic, G., Mi, Y., Zhong, H.-G. & Fan, Z.: Some aspects of the non-linear
finite element method, Finite Elements in Analysis & Design 27, 1997, 19-40.
131 DaDeppo, D. A. & Schmidt, R.: Instability of clamped-hinged circular arches subjected to a
point load, Journal of Applied Mechanics, Trans. ASME 42, 1975, 894-896.
132 Dai, J. S.: An historical review of the theoretical development of rigid body displacements
from Rodrigues parameters to the finite twist, Mechanism and machine Theory 41, 2006, 41-
52.
133 Dakshina Moorthy, C. M. & Reddy, J. N.: Modelling of laminates using a layerwise element
with enhanced strains, Int. Journal for Numerical Methods in Engineering 43, 1998, 755-779.
134 Damjanić, F. B., Brank, B. & Perić, D.: A non-linear stress resultant shell finite element, Int.
Conf. Of Numerical Methods in Continuum Mechanics, Stara Leśna, 19-22 Sept. 1994, 1994,
100-107.
135 Das, M., Barut, A., Madenci, E. & Ambur, D.R.: Complete stress field in sandwich panels with
a new triangular finite element of single-layer theory, Computer Methods in Applied
Mechanics & Engineering 194, 2005, 2969-3005.
136 Dauge, M., Faou, E. & Yosibash, Z.: Plates and Shells: Asymptotic Expansions and hierarchic
Models, Encyclopedia of Computational Mechanics, Vol. 1: Fundamentals, E. Stein, R. de
Borst & T. J. R. Hughes (eds.), John Wiley & Sons, Ltd. 2004, 199-236.
137 Dennis, S. T.: A Galerkin solution to geometrically nonlinear laminated shallow shell
equations, Computers & Structures 63, 1997, 859-874.
138 Dennis, S. T. & Palazotto, A. N.: Transverse shear deformation in orthotropic cylindrical
pressure vessels using a higher-order shear theory, AIAA Journal 27, 1989, 1441-1447.
139 Dennis, S. T. & Palazotto, A. N.: Large displacement and rotational formulation for laminated
shells including parabolic transverse shear, Int. Journal of Non-Linear Mechanics 25, 1990,
67-85.
140 Desai, Y. M., Ramtekkar, G. S. & Shah, A. H.: A novel 3D mixed finite-element model for
statics of angle-ply laminates, Int. Journal for Numerical Methods in Engineering 57, 2003,
1695-1716.
141 Di Sciuva, M.: An improved shear deformation theory of moderately thick multilayered
anisotropic shells and plates, Journal of Applied Mechanics, Trans. ASME 54, 1987, 589-596.
142 Di Sciuva, M.: A third-order triangular multilayered plate finite element with continuous
interlaminar stresses, Int. Journal for Numerical Methods in Engineering 38, 1995, 1-26.
143 Di, S. & Rothert, H.: A solution of laminated cylindrical shell using an unconstrained third-
order theory, Composite Structures 32, 1995, 667-680.
144 Di, S. & Rothert, H.: Solution of laminated cylindrical shell using an unconstrained third-order
theory, Computers & Structures 69, 1998, 291-303.
145 DiCarlo, A., Podio-Guidugli, P. & Williams, W. O.: Shells with thickness distension, Int.
Journal of Solids & Structures 38, 2001, 1201-1225.
146 Dong, S. B. & Tso, F. K. W.: On a laminated orthotropic shell theory including transverse
shear deformation, Journal of Applied Mechanics, Trans. ASME 39, 1972, 1091-1096.
147 Dorninger, K.: Computational analysis of composite shells at large deformations, Computer
Aided Design in Composite Material Technology, Proceedings of the Int. Conf. Southampton,
1988, C. A. Brebbia, W. P. de Wilde & W. R. Blain (eds.), Springer-Verlag, Berlin 1988, 7/66-
7/75.
148 Dorninger, K. & Rammerstorfer, F. G.: A layered composite shell element for elastic and
thermoelastic stress and stability analysis at large deformations; Int. Journal for Numerical
149 Duan, M.: An efficient hybrid / mixed element for geometrically nonlinear analysis of plate
and shell structures, Computational Mechanics 35, 2004, 72-84.
150 Dupuis, G. A., Hibbitt, H. D., McNamara, S. F. & Marcal, P. V.: Nonlinear material and
geometric behavior of shell structures, Computers & Structures 1, 1971, 223-239.
151 Dvorak, G. J.: Composite materials: Inelastic behavior, damage, fatigue and fracture, Int.
152 Dvorkin, E. N. & Bathe, K. -J.: A continuum mechanics based four-node shell element for
general nonlinear analysis, Engineering Computations 1, 1984, 77-88.
153 Eason, T. G. & Ochoa, O. O.: Modeling progressive damage in composites: a shear deformable
element for ABAQUS, Composite Structures 34, 1996, 119-128.
154 El-Abbasi, N. & Meguid, S.A.: A new shell element accounting for through-thickness
deformation, Computer Methods in Applied Mechanics & Engineering 189, 2000, 841-862.
155 Eremeyev, V. A.: Nonlinear micropolar shells: theory and applications, in [366], 2005, 11-18.
156 Ericksen, J. L. & Truesdell, C.: Exact theory of stress and strain in rods and shells, Archive for
Rational Mechanics & Analysis 1, 1958, 295-323.
157 Fares, M. E. & Youssif, Y. G.: A refined equivalent single-layer model of geometrically non-
linear doubly curved layered shells using mixed variational approach, Int. Journal Non-Linear
Mechanics 36, 2001, 117-124.
158 Fares, M. E., Zenkour, A. M. & El-Marghany, M. Kh.: Non-linear thermal effects on the
bending response of cross-ply laminated plates using refined first-order theory, Composite
Structures 49, 2000, 257-267.
159 Felippa, C.A. & Haugen, B.: A unified formulation of small-strain corotational finite elements:
I. Theory, Computer Methods in Applied Mechanics & Engineering 194, 2005, 2285-2335.
160 Feng, W. & Hoa, S. V.: Partial hybrid finite elements for composite laminates, Finite Elements
in Analysis and Design 30, 1998, 365-382.
7. References 159
161 Fernandes, A.: A mixed formulation for elastic multilayer plates, Computes Rendus Mecanique
331, 2003, 337-342.
162 Ferreira, A. J. M. & Barbosa, J. T.: Buckling behaviour of composite shells, Composite
Structures 50, 2000, 93-98.
163 Ferreira, A. J. M. & Fernandes, A. A.: A review of numerical methods for the analysis of
composite and sandwich structures, Multimaterials Technologies – Solutions and
Opportunities, DOGMA Conference in Utrecht, 24-25 October 2000, 71-76.
164 Ferro, V.: Analyse géométriquement non linéaire des plaques en grandes rotations,
(Geometrically nonlinear analysis for laminated plates and shells with large rotations) (in
French) Ph.D. thesis, EUDIL, Univ. des Sciences et Tech de Lille, Villeneuve D’Ascq, France,
1998.
165 Ferro, V., Kreja, I. & Weichert, D.: Geometrically non-linear analysis of laminated shells, Shell
Structures: Theory and Applications, Proc. the 6th Conf. SSTA 98, Gdańsk – Jurata, 11-13
October 1998, J. Chróścielewski & W. Pietraszkiewicz (eds.), TUG Press, Gdańsk, 1998, 113-
114.
166 Flügge, W.: Stresses in shells, Springer-Verlag, New York 1967.
167 Fontes Valente, R. A., Natal Jorge, R. M., Cardoso, R. P. R., César de Sá, J. M. A.
& Grácio, J. J. A.: On the use of an enhanced transverse shear strain shell element for problems
involving large rotations, Computational Mechanics 30, 2003, 286-296.
168 Ford, T.: Aerospace composites, Aircraft Engineering and Aerospace Technology 69, 1997,
334–342.
169 Fraeijs de Veubeke, B. M.: Displacement and equilibrium models, in Stress Analysis, O. C.
Zienkiewicz & G. Hollister (eds.), Wiley, London, 1965, 145-197; reprinted in Int. Journal for
170 Frey, F. & Cescotto, S.: Some new aspects of the incremental Total Lagrangian description in
nonlinear analysis, Proc. Int. Conf. on Finite Elements in Nonlinear Solid and Structural
Mechanics, Geilo, Norway, 1977, ed. Bergan, P. G. et al., Tapir Press, Norwegian Institute of
Technology, Trondheim, 1977, Vol.1, 323-343.
171 Gaudenzi, P., Barboni, R. & Mannini, A.: A finite element evaluation of single-layer and
multi-layer theories for the analysis of laminated plates, Composite Structures 30, 1995, 427-
440.
172 Gere, J. M. & Timoshenko, S. P.: Mechanics of Materials, 3rd ed., PWS Publishing Company,
Boston, 1990.
173 Ghugal, Y. M. & Shimpi, R. P.: A review of refined shear deformation theories of isotropic
and anisotropic laminated plates, Journal of Reinforced Plastics and Composites 21, 2002,
775-813.
174 Gilewski, W. & Radwańska, M.: A survey of finite element models for the analysis of
moderately thick shells, Finite Elements in Analysis and Design 9, 1991, 1-21.
175 Girkmann, K.: Flachentragwerke: Einführung in die Elastostatik der Scheiben, Platten,
Schalen und Faltwerke, Springer-Verlag, Wien, 1956.
176 Glockner, P. G. & Malcolm, D. J.: Lagrangian formulation of sandwich shell theory, Journal
of Engineering Mechanics ASCE 99, 1973, 445-456.
177 Gratie, L.: Asymptotic analysis of nonlinearly elastic shells with variable thickness, Journal of
Theoretical and Applied Mechanics (Polish Society of Theoretical and Applied Mechanics) 41,
2003, 487-508.
178 Green, A. E. & Zerna, W.: Theoretical Elasticity, Oxford University Press, London, 1960.
179 Greer Jr., J. M. & Palazotto, A. N.: Application of a total Lagrangian Corotational FE scheme
to inflation of tires, Int. Journal of Solids & Structures 34, 1997, 3541-3570.
180 Groenwold, A. & Stander, N.: An efficient 4-node 24 dof thick shell finite element with 5-
point quadrature, Engineering Computations 12, 1995, 723-747.
181 Groenwold, A. & Stander, N.: A 24 d.o.f. four-node flat shell finite element for general
unsymmetric orthotropic layered composites, Engineering Computations 15, 1998, 518-543.
182 Gruttmann, F. & Wagner, W.: Coupling of 2D- and 3D-composite shell elements in linear and
nonlinear applications, Computer Methods in Applied Mechanics & Engineering 129, 1996,
271-287.
183 Guz, A. N. (ed.): Micromechanics of composite materials: Focus on Ukrainian Research,
Applied Mechanics Review 45, 1992, 13-101.
184 Haas, D. J. & Lee, W.: A nine-node assumed-strain finite element for composite plates and
shells, Computers & Structures 26, 1987, 445-452.
185 Habip, L. M.: A survey of modern developments in the analysis of sandwich structures,
Applied Mechanics Reviews 18, 1965, 93-98.
186 Habip, L. M.: Theory of elastic shells in the reference state, Ingenieur-Archiv 34, 1965, 228-
237.
187 Hachenberg, D.: The Role of Advanced Numerical Methods in the Design and Certification of
Future Composite Aircraft Structures, WCCM V: Proceedings of the Fifth World Congress on
Computational Mechanics, July 7-12, 2002, Vienna, Austria, H.A. Mang, F.G. Rammerstorfer
& J. Eberhardsteiner (eds.), CD-ROM, Vienna 2002.
188 Halle, D. K. & Kelly, A.: Strength of fibrous composite materials, Annual Review of Materials
Science 2, 1972, 405–462.
189 Han, S. C., Kim, K. D. & Kanok-Nukulchai, W.: An element-based 9-node resultant shell
element for large deformation analysis of laminated composite plates and shells, Structural
Engineering and Mechanics 18, 2004, 807-829.
190 Hashagen, E., Schellekens, J. C. J., de Borst, R. & Parisch, H.: Finite element procedure for
modelling fibre metal laminates, Composite Structures 32, 1995, 255-264.
191 Hassis, H.: A high-order theory for static-dynamic analysis of laminated plates using a special
warping model, European Journal of Mechanics A / Solids 21, 2002, 323-332.
192 Hauptmann, R., & Schweizerhof, K.: A systematic development of `solid-shell' element
formulations for linear and non-linear analyses employing only displacement degrees of
freedom, Int. Journal for Numerical Methods in Engineering 42, 1998, 49-69.
193 He, L.-H.: Non-linear theory of laminated shells accounting for continuity conditions of
displacements and tractions at layer interfaces, Int. Journal of Mechanical Sciences 37, 1995,
161-173.
194 Heppler, G. R. & Hansen, J. S.: A Mindlin element for thick and deep shells, Computer
Methods in Applied Mechanics & Engineering 54, 1986, 21-47.
195 Hildebrand, F. B., Reissner, E. & Thomas, G. B.: Notes on the foundations of the theory of
small displacements of orthotropic shells, finite symmetrical deflections of thin shells of
revolution, NACA TN No. 1833, Washington 1949.
196 Hoff, C. C.: Improvements in Linear Buckling and Geometric Nonlinear Analysis of
MSC/NASTRAN’s Lower Order Elements, MSC 1993 World Users’ Conference, Arlington,
VA, May 24-28, 1993.
197 Hoff, C. C., Harder, R. L., Campbell, G., MacNeal, R. H. & Wilson, C. T.: Analysis of shell
structures using MSC/NASTRAN’s shell elements with surface normals, Proc. 1995 MSC
World Users’ Conf. Universal City, CA, 8-12 May, 1995.
198 Hoff, N. J.: Bending and Buckling of Rectangular Sandwich Plates, NACA TN No. 2225,
Washington 1950.
199 Hohe, J. & Becker, W.: Effective stress-strain relations for two-dimensional cellular sandwich
7. References 161
cores: Homogenization, material models, and properties, Applied Mechanics Review 55, 2002,
61-87.
200 Hohe, J. & Librescu, L.: Advances in the Structural Modeling of Elastic Sandwich Panels,
Mechanics of Advanced Materials & Structures 11, 2004, 395-424.
201 Hollaway, L.C.: The evolution of and the way forward for advanced polymer composites in the
civil infrastructure, Construction & Building Materials 17, 2003, 365-378.
202 Holzapfel, G. A.: Application of the Hermitian-Method to thin shell structures undergoing
finite rotations, Computational Mechanics ’91, Theory and Applications. Proceedings of
ICCES’91. S.N. Atluri, D.E. Beskos, R. Jones and G. Yagawa (Eds), ICES Publications, W.G.
Wolfe Associates, Alpharetta, Georgia/USA, 1991, 372-375.
203 Holzapfel, G. A. & Wimmer, H.: Geometrisch nichtlineare Sandwichplatten unter
Berücksichtigung von Schubdeformationen, (Geometrically non-linear analysis of sandwich
plates including shear deformations) (in German), Zeitschrift für angewandte Mathematik und
Mechanik (Journal of Applied Mathematics and Mechanics) - ZAMM 74, 1994, 235-241.
204 Horrigmoe, G. & Bergan, P. G.: Nonlinear analysis of free-form shells by flat finite elements,
205 Hossain, S. J., Sinha, P. K. & Sheikh, A. H.: A finite element formulation for the analysis of
laminated composite shells, Computers & Structures 82, 2004, 1623-1638.
206 Hsiao, K. -M. & Chen, Y. -R.: Nonlinear analysis of shell structures by degenerated
isoparametric shell element, Computers & Structures 31, 1989, 427-438.
207 Hu, H.-T.: Buckling analyses of fiber-composite laminate plates with material nonlinearity,
Finite Elements in Analysis & Design 19, 1995, 169-179.
208 Huang, H. C.: Implementation of assumed strain degenerated shell elements, Computers &
Structures 25, 1987, 147-155.
209 Huang, H. C. & Hinton, E.: A new nine node degenerated shell element with enhanced
membrane and shear interpolation, Int. Journal for Numerical Methods in Engineering 22,
1986, 73-92.
210 Hughes, T. J. R. & Carnoy, E.: Nonlinear finite element shell formulation accounting for large
membrane strains, Computer Methods in Applied Mechanics & Engineering 39, 1983, 69-82.
211 Hughes, T. J. R. & Hinton, E. (eds.): Finite Element Methods for Plate and Shell Structures,
Pineridge Press Ltd, Swansea 1986.
212 Hughes, T. J. R. & Liu, W. K.: Nonlinear finite element analysis of shells. Part I. Three-
dimensional shells, Computer Methods in Applied Mechanics & Engineering 26, 1981, 331-
362.
213 Hughes, T. J. R. & Liu, W. K.: Nonlinear finite element analysis of shells. Part II. Two-
dimensional shells, Computer Methods in Applied Mechanics & Engineering 27, 1981, 167-
181.
214 Hughes, T. J. R. & Tezduyar, T. E.: Finite elements based upon Mindlin plate theory with
particular reference to the four-node bilinear isoparametric element, Journal of Applied
215 Hult, J. & Rammerstorfer, F. G. (eds.): Engineering Mechanics of Fibre Reinforced Polymers
and Composite Structures, CISM Courses and Lectures No. 348, Springer-Verlag, Vienna -
New York 1994.
216 Huttelmaier, H. P. & Epstein, M.: A finite element formulation for multilayered and thick
shells, Computers & Structures 21, 1985, 1181-1185.
217 Ibrahimbegović, A.: Stress resultant geometrically nonlinear shell theory with drilling rotations
– Part. I. A consistent formulation, Computer Methods in Applied Mechanics & Engineering
118, 1994, 265-284.
218 Ibrahimbegović, A.: On assumed shear strain in finite rotation shell analysis, Engineering
Computations 12, 1995, 425-438.
219 Ibrahimbegović, A.: On the choice of finite rotation parameters, Computer Methods in Applied
220 Ibrahimbegović, A.: Recent developments in nonlinear analysis of shell problem and its finite
element solution, First MIT Conference on Computational Fluid and Solid Mechanics, K.J.
Bathe (ed.), Elsevier Science Ltd., 2001, 251-254.
221 Ibrahimbegović, A.: Non-linear shell theory with finite rotations and finite strains: recent
achievements, in [366], 2005, 19-31.
222 Ibrahimbegović, A. & Frey, F.: Stress resultant geometrically nonlinear shell theory with
drilling rotations – Part. II. Computational aspects, Computer Methods in Applied Mechanics
& Engineering 118, 1994, 285-308.
223 Ibrahimbegović, A. & Krätzig, W. B. (eds): Shells: Theoretical formulation, mathematical
analysis and finite element implementation, (Special Issue of) Computers & Structures 80,
Issues 9-10, 2002, 697-906.
224 Idibi, A, Karama, M, & Touratier, M.: Comparison of various laminated plate theories,
225 Jacquotte, O.-P. & Oden, J. T.: Analysis of hourglass instabilities and control in
underintegrated finite element methods, Computer Methods in Applied Mechanics &
Engineering 44, 1984, 339-363.
226 Jaunky, N. & Knight Jr., N.F.: An assessment of shell theories for buckling of circular
cylindrical laminated composite panels loaded in axial compression, Int. Journal of Solids and
Structures 36, 1999, 3799-3820.
227 Jemielita, G.: On the winding paths of the theory of plates, Journal of Theoretical and Applied
Mechanics (Polish Society of Theoretical and Applied Mechanics) 31, 1993, 317-327.
228 Jemielita, G.: Meandry teorii płyt i powłok (On the winding paths of the theory of plates and
shells), (in Polish) [499], 2001, 45-146.
229 Jemielita, G.: Coefficients of shear correction in transversely nonhomogeneous moderately
thick plates, Journal of Theoretical and Applied Mechanics (Polish Society of Theoretical and
Applied Mechanics) 40, 2002, 73-84.
230 Jiang, L. & Chernuka, M. W.: A simple four-noded corotational shell element for arbitrarily
large rotations, Computers & Structures 53, 1994, 1123-1132.
231 Jones, R. M.: Mechanics of composite materials, McGraw-Hill, Tokyo, 1975.
232 Jun, S. M. & Hong, C. S.: Buckling behavior of laminated composite cylindrical panels under
axial compression, Computers & Structures 29, 1988, 479-490.
233 Kanok-Nukulchai, W., Taylor, R. L. & Hughes, T. J. R.: A large deformation formulation for
shell analysis by the FEM, Computers & Structures 13, 1981, 19-27.
234 Kant, T. & Swaminathan, K.: Estimation of transverse/interlaminar stresses in laminated
composites – a selective review and survey of current developments, Composite Structures 49,
2000, 65-75.
235 Kapania, R. K. & Li, J.: A formulation and implementation of geometrically exact curved
beam elements incorporating finite strains and finite rotations, Computational Mechanics 30,
2003, 444-459.
236 Khdeir, A. A., Reddy, J. N. & Librescu L.: Analytical solution of a refined shear deformation
theory for rectangular composite plates, Int. Journal of Solids & Structures 23, 1987, 1447-
1463.
237 Kim, C.H., Sze, K. Y. and Kim, Y. H.: Curved quadratic triangular degenerated- and solid-
shell elements for geometric non-linear analysis, Int. Journal for Numerical Methods in
7. References 163
Engineering 57, 2003, 2077-2097.

238 Kim, J.-S. & Cho, M.: Enhanced First-Order Shear Deformation Theory for laminated and
sandwich plates, Journal of Applied Mechanics, Trans. ASME 72, 2005, 809-817.
239 Kim, K. D.: Buckling behaviour of composite panels using the finite element method,
240 Kim, K. D., Lomboy, G. R. & Han, S. C.: A co-rotational 8-node assumed strain shell element
for postbuckling analysis of laminated composite plates and shells, Computational Mechanics
30, 2003, 330-342.
241 Kim, K. D. & Park, T. H.: An 8-node assumed strain element with explicit integration for
isotropic and laminated composite shells, Structural Engineering & Mechanics 13, 2002, 387-
410.
242 Kim, K. D. & Voyiadjis, G.Z.: Non-linear finite element analysis of composite panels,
Composites, Part B 30, 1999, 365-381.
243 Kim, Y. H. & Lee, S. W.: An assumed strain solid element model for geometrically nonlinear
shell analysis, in [328], 1989, 241-259.
244 Kleiber, M.: Incremental Finite Element Modeling in Non-Linear Solid Mechanics, J. Wiley &
Sons, New York, 1989.
245 Kleiber, M. (ed.): Handbook of Computational Solid Mechanics: Survey and Comparison of
Contemporary Methods, Springer Verlag, Berlin, 1998.
246 Klinkel, S., Gruttmann, F. & Wagner, W.: A continuum based three-dimensional shell element
for laminated structures, Computers & Structures 71, 1999, 43-62.
247 Klinkel, S., Gruttmann, F. & Wagner, W.: A robust non-linear solid shell element based on a
mixed variational formulation, Computer Methods in Applied Mechanics & Engineering 195,
2006, 179-201.
248 Kłosowski, P. & Woźnica, K.: Numerical treatment of elasto viscoplastic shells in the range of
moderate and large rotations, Computational Mechanics 34, 2004, 194-212.
249 Knight Jr., N. F. & Qi, Y.: On a consistent first-order shear deformation theory for laminated
plates, Composites, Part B 28B, 1997, 397-405.
250 Koiter, W.T.: A consistent first approximation in the general theory of thin elastic shells, Proc.
I.U.T.A.M. Symposium on the theory of thin elastic shells (Delft, August 1959), North Holland
Publishing Company, Amsterdam, 1960, 12-33.
251 Koiter, W.T.: On the nonlinear theory of thin elastic shells. Proceedings of the Koninklijke
Nederlandse Akademie van Wetenschappen B69, 1966, 1-54.
252 Kong, J. & Cheung, Y.K.: Three-dimensional finite element analysis of thick laminated plates,
253 Korn, G. A. & Korn, T. M.: Mathematical Handbook for Scientists and Engineers, Dover
Publications, Inc. Minaola, New York, 2000.
254 Krätzig, W. B.: ‘Best’ transverse shearing and stretching shell theory for nonlinear finite
element simulations, Computer Methods in Applied Mechanics & Engineering 103, 1993, 135-
160.
255 Krätzig, W. B. & Jun, D.: Layered higher order concepts for D-adaptivity in shell theory, First
MIT Conference on Computational Fluid and Solid Mechanics, K.J. Bathe (ed.), Elsevier
Science Ltd., 2001, 297-301.
256 Krätzig, W. B. & Jun, D.: Multi-layer multi-director concepts for D-adaptivity in shell theory,
257 Krätzig, W. B. & Jun, D.: On ‘best’ shell models – From classical shells, degenerated and
multi-layered concepts to 3D, Archive of Applied Mechanics 73, 2003, 1-25.
258 Krätzig, W. B. & Oñate, E. (eds.): Computational Mechanics of Nonlinear Response of Shells,
ICES’88 Conference, Atlanta 1988, Springer Verlag, Berlin, 1990.

259 Kreja, I.: Stability analysis of cylindrical composite shells in MSC/Nastran, Archives of Civil
and Mechanical Engineering 5, 2005, 31-41.
260 Kreja, I.: Critical examination of benchmark problems for large rotation analysis of laminated
shells, Shell Structures: Theory and Applications, Proc. the 8th Conf. SSTA, Gdańsk – Jurata,
12-14 October 2005, W. Pietraszkiewicz & C. Szymczak (eds), Taylor & Francis Group,
London 2006, 481-485.
261 Kreja, I. & Cywiński, Z.: Is reduced integration just a numerical trick, Computers & Structures
29, 1988, 491-496.
262 Kreja, I. & Cywiński, Z.: Degenerated isoparametric finite elements in nonlinear analysis of
2D-problems, Computers & Structures 41 1991, 1029-1040.
263 Kreja, I. & Schmidt, R.: Moderate rotation theory FEM analysis of laminated anisotropic
composite plates and shells, Shell Structures: Theory and Applications, Proc. the 5th Conf.
SSTA, Janowice, 15-18 October 1992, R. Nagórski (ed), Warsaw Univ. Technol. Press,
Warsaw 1992, 55-56.
264 Kreja, I. & Schmidt, R.: Moderate Rotation Shell Theory in FEM application, Proceedings of
Gdansk University of Technology Civil Eng series LI.1995, 229-249.
265 Kreja, I. & Schmidt, R.: On the FEM implementation of the Moderate Rotation Shell Theory
for elastic anisotropic shells, Computer methods in mechanics. Proceedings of the XII Polish
Conference on Computer Methods in Mechanics. Warsaw-Zegrze, Poland, 9-13 May 1995,
Military University of Technology, Faculty of Mechanics, Warsaw 1995, 173-174.
266 Kreja, I. & Schmidt, R.: On finite rotation theory and FE-analysis of composite shells. 1999
ASME Mechanics & Materials Conference. June 27 - 30, 1999, Blacksburg, Virginia. Book of
abstracts. R.C. Batra & E.G. Henneke (eds.), Virginia Tech, Blacksburg 1999, 317.
267 Kreja, I. & Schmidt, R.: On the FEM implementation of the large rotation shell theory for
elastic anisotropic shells. Shell Structures: Theory and Applications, Proc. the 7th Conf. SSTA,
Gdańsk – Jurata, 9-11 October 2002, P. Kłosowski & W. Pietraszkiewicz (eds), GUT Press,
Gdańsk 2002, 139-140.
268 Kreja, I. & Schmidt, R.: Large rotations in First Order Shear Deformation FE analysis of
laminated shells, Int. Journal Non-Linear Mechanics 41, 2006, 101-123.
269 Kreja, I., Schmidt, R. & Reddy, J. N.: Finite elements based on a first-order shear deformation
moderate rotation shell theory with applications to the analysis of composite structures, Int.
Journal Non-Linear Mechanics 32, 1997, 1123-1142.
270 Kulikov, G. M.: Non-linear analysis of multilayered shells under initial stress, Int. Journal of
Non-Linear Mechanics 36, 2001, 323-334.
271 Kulikov, G. M.: Analysis of initially stressed multilayered shells, Int. Journal of Solids &
Structures 38, 2001, 4535-4555.
272 Kulikov, G. M. & Plotnikova, S. V.: Simple and effective elements based upon Timoshenko–
Mindlin shell theory, Computer Methods in Applied Mechanics & Engineering 191, 2002,
1173-1187.
273 Kulikov, G. M. & Plotnikova, S. V.: Non-linear strain-displacement equations exactly
representing large rigid-body motions. Part I: Timoshenko-Mindlin shell theory, Computer
274 Kweon, J. H. & Hong, C. S.: An improved arc-length method for postbuckling analysis of
composite cylindrical panels, Computers & Structures 53, 1994, 541-549.
275 Ladevéze, P.: Multiscale Modeling and Computational Strategies for Composites, WCCM V:
Proceedings of the Fifth World Congress on Computational Mechanics, July 7-12, 2002,
Vienna, Austria, H.A. Mang, F.G. Rammerstorfer & J. Eberhardsteiner (eds.), CD-ROM,
Vienna 2002.
7. References 165
276 Laschet, G. & Jeusette, J.-P.: Postbuckling finite element analysis of composite panels,
277 Lee, K. & Lee, S. W.: A post-processing approach to determine transverse stresses in
geometrically nonlinear composite and sandwich structures, Journal of Composite Materials
37, 2003, 2207-2224.
278 Lee, K. H & Cao, L.: A predictor-corrector zig-zag model for the bending of laminated
composite plates, Int. Journal of Solids & Structures 33, 1996, 879-897.
279 Lee, K. H., Senthilnathan, N. R., Lim, S. P. & Chow, T.: An improved zig-zag model for the
bending of laminated composite plates, Composite Structures 15, 1990, 137-148.
280 Lee, W. J. & Lee, B. C.: An effective finite rotation formulation for geometrical non-linear
shell structures, Computational Mechanics 27, 2001, 360-368.
281 Leicester, R. H.: Finite deformations of shallow shells, Journal of Engineering Mechanics
ASCE 94, 1968, 1409-1423.
282 Lentzen, S. & Schmidt, R.: Geometrically nonlinear composite shells with integrated
piezoelectric layers, Proceedings in Applied Mathematics and Mechanics 4, 2004, 63-66.
283 Lewiński, T.: On displacement-based theories of sandwich plates with soft core, Journal of
Engineering Mathematics 25, 1991, 223-241.
284 Lewiński, T.: On recent developments in the homogenization theory of elastic plates and their
application to optimal design: Part I, Structural Optimization 6, 1993, 59-64.
285 Li, M.: The finite deformation theory for beam, plate and shell. Part I. The two-dimensional
beam theory, Computer Methods in Applied Mechanics & Engineering 146, 1997, 53-63.
286 Li, X. & Liu, D.: Generalized laminate theories based on double superposition hypothesis, Int.
287 Liao, C. -L. & Reddy, J. N.: An Incremental Total Lagrangian Formulation for General
Anisotropic Shell-Type Structures, Report No. CCMS-87-16. Virginia Polytechnic Institute and
State University, Blacksburg, Virginia, 1987.
288 Liao, C. -L. & Reddy, J. N.: A continuum-based stiffened composite shell element for
geometrically nonlinear analysis, AIAA Journal 27, 1989, 95-101.
289 Libai, A. & Simmonds, J.G.: The Non-Linear Theory of Elastic Shells, 2nd ed., Cambridge
University Press, New York, 1998.
290 Librescu, L.: Refined geometrically non-linear theories of anisotropic laminated shells,
Quarterly of Applied Mathematics 45, 1987, 1-22.
291 Librescu, L. & Hause, T.: Recent developments in the modeling and behavior of advanced
sandwich constructions: a survey, Composite Structures 48, 2000, 1-17.
292 Librescu, L. & Schmidt, R.: Refined theories of elastic anisotropic shells accounting for small
strains and moderate rotations, Int. Journal of Non-Linear Mechanics 23, 1988, 217-229.
293 Librescu, L. & Schmidt, R.: Substantiation of a shear-deformable theory of anisotropic
composite laminated shells accounting for the interlaminar continuity conditions, Int. Journal
of Engineering Science 29, 1991, 669-683.
294 Liu, W. K., Ong, J. S. -J. & Uras, R.S.: Finite element stabilization matrices - A unification
approach, Computer Methods in Applied Mechanics & Engineering 53, 1985, 13-46.
295 Lopez-Anido, R. & Karbhari, V. M.: Fiber Reinforced Composites in Civil Infrastructure,
Emerging Materials for Civil Infrastructure – State of the Art, R. A. Lopez-Anido & T. R.
Naik (eds), ASCE, Reston 2000, 41-78.
296 Luo, Y.-H.: Explanation and elimination of shear locking and membrane locking with field
consistence approach, Computer Methods in Applied Mechanics & Engineering 162, 1998,
249-269.
297 Luo, Y.-H. & Eriksson, A.: An alternative assumed strain method, Computer Methods in
Applied Mechanics & Engineering 178, 1999, 23-37.
298 MacNeal, R.: A simple quadrilateral shell element, Computers & Structures 8, 1978, 175-183.
299 MacNeal, R. H.: Perspective on finite elements for shell analysis, Finite Elements in Analysis
& Design 30, 1998, 175-186.
300 MacNeal, R. H. & Harder, R. L.: A proposed standard set of problems to test finite element
accuracy, Finite Elements in Analysis & Design 1, 1985, 3-20.
301 Malcolm, D. J. & Glockner, P. G.: Nonlinear sandwich shell and Cosserat surface theory,
Journal of Engineering Mechanics ASCE 98, 1972, 1183-1203.
302 Malinen, M.: On the classical shell model underlying bilinear degenerated shell finite
elements: general shell geometry, Int. Journal for Numerical Methods in Engineering 55, 2002,
629-652.
303 Malkus, P. S. & Hughes, T. J. R.: Mixed finite element methods - Reduced and selective
integration techniques: A unification of concepts, Computer Methods in Applied Mechanics &
Engineering 15, 1978, 63-81.
304 Manet, V.: The use of ANSYS to calculate the behaviour of sandwich structures, Composites
Science & Technology 58, 1998, 1899-1905.
305 Marcinowski, J.: Large deflections of shells subjected to an external loads and temperature
changes, Int. Journal of Solids & Structures 34, 1997, 755-768.
306 Marcinowski, J.: Geometrically nonlinear static analysis of sandwich plates and shells, Journal
of Theoretical and Applied Mechanics (Polish Society of Theoretical and Applied Mechanics)
41, 2003, 561-574.
307 Massin, P. & Al Mikdad, M.: Nine node and seven node thick shell elements with large
displacements and rotations, Computers & Structures 80, 2002, 835-847.
308 Masud, A. & Panahandeh, M.: Finite-Element Formulation for Analysis of Laminated
Composites, Journal of Engineering Mechanics ASCE 125, 1999, 1115-1124.
309 Masud, A., Tham, C. L. & Liu, W. K.: A stabilized 3-D co-rotational formulation for
geometrically nonlinear analysis of multi-layered shells, Computational Mechanics 26, 2000,
1-12.
310 Mau, S. T.: A refined laminated plate theory, Journal of Applied Mechanics, Trans. ASME 40,
1973, 606-607.
311 Mawenya, A. S. & Davies, J. D.: Finite element bending analysis of multilayer plates,
Int. Journal for Numerical Methods in Engineering 8, 1974, 215-225.
312 Meek, J. L. & Ristic, S.: Large displacement analysis of thin plates and shells using a flat facet
finite element formulation, Computer Methods in Applied Mechanics & Engineering 145,
1997, 285-299.
313 Meek, J. L. & Wang, Y.: Nonlinear static and dynamic analysis of shell structures with finite
rotation, Computer Methods in Applied Mechanics & Engineering 162, 1998, 301-315.
314 Mindlin, R. D.: Influence of rotatory inertia and shear in flexural motions of isotropic elastic
plates, Journal of Applied Mechanics, Trans. ASME 18, 1951, 31-38.
315 Moita, J. S., Mota Soares, C. M. & Mota Soares, C. A.: Buckling behaviour of laminated
composite structures using a discrete higher-order displacement model, Composite Structures
35, 1996, 75-92.
316 Morley, L. S. D.: An assumed stress hybrid curvilinear triangular finite element for plate
bending, Int. Journal for Numerical Methods in Engineering 20, 1984, 529-548.
317 Morley, L. S. D.: Hellinger-Reissner principles for plate and shell finite elements, Int. Journal
for Numerical Methods in Engineering 20, 1984, 773-777.
7. References 167
318 MSC/NASTRAN Encyclopedia for Version 69, Mac Neal-Schwendler Corp., 1996.
319 Naboulsi, S. K. & Palazotto, A. N.: Non-linear static-dynamic finite element formulation for
composite shells, Int. Journal of Non-Linear Mechanics 38, 2003, 87-110.
320 Naghdi, P. M.: Foundations of elastic shell theory, Progress in Solid Mechanics, Vol. 4, I. N.
Sneddon & R. Hill (eds.), North-Holland, Amsterdam, 1963, 1-90.
321 Naghdi, P. M.: The theory of shells and plates, Handbuch der Physik, Vol. VIa/2, S.Flügge &
C. Truesdell (eds.), Springer-Verlag, New York-Heidelberg-Berlin, 1972, 425-640.
322 Naghdi, P. M.: Finite deformations of elastic rods and shells, Finite Elasticity, Proceedings of
the IUTAM Symposium, Lehigh University, Bethlehem, PA, USA, August 10-15, 1980, D.E.
Carlson & R. T. Shield (eds), Martinus Nijhoff Publishers, The Hague/Boston/London, 1982,
47-103.
323 Nettles, A.T.: Basic Mechanics of Laminated Composite Plates, NASA RP-1351, MSFC,
Alabama 1994.
324 Niordson, F. I. & Niordson, C. F.: On the accuracy of the asymptotic theory for cylindrical
shells, Archive of Applied Mechanics 69, 1999, 677-689.
325 Noor, A. K.: Bibliography of monographs and surveys on shells, Applied Mechanics Reviews
43, 1990, 223-234.
326 Noor, A. K.: List of Books, Monographs, Conference Proceedings and Survey Papers on
Shells, in [328], 1989, vii-xxxiii.
327 Noor, A. K. & Andersen, C. M.: Mixed models and reduced/selective integration displacement
models for nonlinear shell analysis, Int. Journal for Numerical Methods in Engineering 18,
1982, 1429-1454.
328 Noor, A. K., Belytschko, T. & Simo, J. C. (eds.): Analytical and Computational Models of
Shells, Winter Annual Meeting of ASME, San Francisco, 1989, ASME, CED-Vol. 3, New York
1989.
329 Noor, A. K. & Burton, W. S.: Assessment of computational models for multilayered composite
shells, Applied Mechanics Reviews 43, 1990, 67-97.
330 Noor, A. K., Burton, W. S. & Peters, J. M.: Assessment of computational models for
multilayered composite cylinders, Int. Journal of Solids & Structures 27, 1991, 1269-1286.
331 Noor, A. K. & Peters, J. M.: A posteriori estimates for shear correction factors in multilayered
composite cylinders, Journal of Engineering Mechanics ASCE 115, 1988, 1225-1244.
332 Oliver, J. & Oñate, E.: A Total Lagrangian formulation for the geometrically nonlinear
analysis of structures using finite elements. Part I. Two-dimensional problems: Shell and plate
structures, Int. Journal for Numerical Methods in Engineering 20, 1984, 2253-2281.
333 Pagano, N. J.: Free edge stress fields in composite laminates, Int. Journal of Solids &
Structures 14, 1978, 401-406.
334 Pagano, N. J.: Stress fields in composite laminates, Int. Journal of Solids & Structures 14,
1978, 385-400.
335 Pagano, N. J. & Hatfield, S.J.: Elastic behavior of multilayered bidirectional composites, AIAA
Journal 10, 1972, 931-933.
336 Pagano, N. J. & Yuan, F. G.: The significance of effective modulus theory (homogenization) in
composite laminate mechanics, Composites Science and Technology 60, 2000, 2471-2488.
337 Pahr, D. H. & Rammerstorfer, F. G.: A fast multi-scale analyzing tool for the investigation of
perforated laminates, Computers & Structures 82, 2004, 227-239.
338 Pai, P. F.: A new look at shear correction factors and warping functions of anisotropic
laminates, Int. Journal of Solids & Structures 32, 1995, 2295-2313.
339 Pai, P. F.: Total-Lagrangian formulation and finite-element analysis of highly flexible plates
and shells, Mathematics and Mechanics of Solids, OnlineFirst 2005, published online at the
http://mms.sagepub.com/cgi/content/abstract/1081286505055474v1 on September 21, 2005
(last visited on December 15, 2006)
340 Pai, P. F. & Palazotto, A. N.: Nonlinear displacement based finite element analysis of
composite shells - A new total Lagrangian formulation, Int. Journal of Solids & Structures 32,
1995, 3047-3073.
341 Pai, P. F. & Palazotto, A. N.: Large-deformation analysis of flexible beams, Int. Journal of
Solids & Structures 33, 1996, 1335-1353.
342 Palmerio, A. F., Reddy, J. N. & Schmidt, R.: On a moderate rotation theory of elastic
anisotropic shells - Part 1. Theory, Int. Journal of Non-Linear Mechanics 25, 1990,
687-700.
343 Palmerio, A. F., Reddy, J. N. & Schmidt, R.: On a moderate rotation theory of elastic
anisotropic shells - Part 2. FE analysis, Int. Journal of Non-Linear Mechanics 25, 1990, 701-
714.
344 Panda, S. C. & Natarajan, R.: Finite element analysis of laminated composite plates, Int.
Journal of Non-Linear Mechanics 14, 1979, 69-79.
345 Parisch, H.: Geometric nonlinear analysis of shells, Computer Methods in Applied Mechanics
& Engineering 14, 1978, 159-178.
346 Parisch, H.: A critical survey of the 9-node degenerated shell element with special emphasis on
thin shell application an reduced integration, Computer Methods in Applied Mechanics &
Engineering 20, 1979, 323-350.
347 Parisch, H.: Large displacements of shells including material nonlinearities, Computer
348 Parisch, H.: An investigation of a finite rotation four node assumed strain shell element, Int.
349 Parisch, H.: A continuum-based shell theory for non-linear applications, Int. Journal for
350 Park, B. C., Park, J. W. & Kim, Y. H.: Stress recovery in laminated composite and sandwich
panels undergoing finite rotation, Composite Structures 59, 2003, 227-235.
351 Park, H. C., Cho, C. & Lee, S. W.: An efficient assumed strain element model with six dof per
node for geometrically non-linear shells, Int. Journal for Numerical Methods in Engineering
38, 1995, 4101-4122.
352 Park, K. C., Pramono, E., Stanley, G. M. & Cabiness, H. A.: The ANS shell elements: Earlier
developments and recent improvements, in [328] CED - Vol.3, 1989, 217-239.
353 Park, K. C. & Stanley, G. M.: A curved C0 shell element based on assumed natural-coordinate
strains, Journal of Applied Mechanics, Trans. ASME 53, 1986, 278-290.
354 Pawsey, S. F. & Clough, R. W.: Improved numerical integration of thick shell finite elements,
Int. Journal for Numerical Methods in Engineering 3, 1971, 575-586.
355 Phan, N. D. & Reddy, J. N.: Analysis of laminated composite plates using a higher-order shear
deformation theory, Int. Journal for Numerical Methods in Engineering 21, 1985, 2201-2219.
356 Pian, T. H. H.: State-of-the-art development of hybrid/mixed finite element method, Finite
Elements in Analysis & Design 21, 1995, 5-20.
357 Pian, T. H. H.: Some notes on the early history of hybrid stress finite element method, Int.
358 Pian, T. H. H. & Tong, P.: Relations between incompatible displacement model and hybrid
stress model, Int. Journal for Numerical Methods in Engineering 22, 1986, 173-181.
359 Pietraszkiewicz, W.: Introduction to the Non-linear Theory of Shells, Mitt. Inst. Mech., No. 10,
Ruhr-Universität Bochum, 1977.
7. References 169
360 Pietraszkiewicz, W.: Finite rotations and Lagrangian Description in the Non-linear Theory of
Shells. Polish Scientific Publishers PWN Warszawa, 1979.
361 Pietraszkiewicz, W.: Consistent second approximation to the elastic strain energy of a shell,
Zeitschrift für angewandte Mathematik und Mechanik (Journal of Applied Mathematics and
Mechanics) - ZAMM 59, 1979, T206-T209.
362 Pietraszkiewicz, W.: Geometrically nonlinear theories of thin elastic shells, Advances in
Mechanics 12, 1989, 51-130.
363 Pietraszkiewicz, W.: Teorie nieliniowe powłok (Non-linear theories of shells), (in Polish) in
[499], 2001, 424-497.
364 Pietraszkiewicz, W. & Badur, J.: Finite rotations in the description of continuum deformation,
Int. Journal of Engineering Sciences 21, 1983, 1097-1115.
365 Pietraszkiewicz, W. & Szwabowicz, M. L.: Entirely Lagrangian nonlinear theory of thin shells,
Archives of Mechanics 33, 1981, 273-288.
366 Pietraszkiewicz, W. & Szymczak, C. (eds): Shell Structures: Theory and Applications, Proc.
the 8th Conf. SSTA, Gdańsk – Jurata, 12-14 October 2005, Taylor & Francis/ Balkema, London
2005, 481-485.
367 Pimenta, P. M., Campello, E. M. B., Wriggers, P.: A fully nonlinear multi-parameter shell
model with thickness variation and a triangular shell finite element, Computational Mechanics
34, 2004, 181-193.
368 Pinsky, P. M. & Kim, K. O.: A multi-director formulation for nonlinear elastic-viscoelastic
layered shells, Computers & Structures 24, 1986, 901-913.
369 Piskunov, V. G.: An iterative analytical theory in the mechanics of layered composite systems,
Mechanics of Composite Materials 39, 2003, 1-16.
370 Piskunov, V. G. & Rasskazov, A. O.: Evolution of the theory of laminated plates and shells,
Int. Applied Mechanics (Plenum Publ. Corp.) 38, 2002, 135-166.
371 Pitkäranta, J.: Shells and finite elements: from classicism to modernism, The Advanced School
and Workshop on ‘Modelling and Numerical Simulation in Continuum Mechanics’, Coimbra,
July 14-18, 2003, I. N. de Figueiredo, L. F. Menezes & J. H. Videman (eds.), Centro
International de Matemática 23, 2003, 82-103.
372 Poulsen, P. N. & Damkilde, L.: A flat triangular shell element with loof nodes, Int. Journal for
373 Qatu M. S.: Recent research advances in the dynamic behavior of shells: 1989-2000, Part 1:
Laminated composite shells, Applied Mechanics Reviews 55, 2002, 325-350.
374 Qi, Y., Knight Jr., N. F.: A refined first-order shear deformation theory and its justification by
plane-strain bending problem of laminated plates, Int. Journal of Solids & Structures 33, 1996,
49-64.
375 Qing-Hua, Q.: Geometrically nonlinear analysis of shells by the variational approach and an
efficient finite element formulation, Computers & Structures 55, 1995, 727-733.
376 Rabczuk, T., Kim, J. Y., Samaniego, E. & Belytschko, T.: Homogenization of sandwich
structures, Int. Journal for Numerical Methods in Engineering 61, 2004, 1009-1027.
377 Ramm, E.: A plate/shell element for large deflections and rotations, Proceedings of US-
Germany Symp., MIT Boston, August 1976, 264-293.
378 Ramm, E.: Geometrisch nichtlineare Elastostatik und Finite Elemente (Geometrically
nonlinear elastostatics and Finite Elements) (in German), Bericht Nr.76-2, Institut für
Baustatik der Universität, Stuttgart 1976.
379 Ramm, E.: Strategies for tracing the nonlinear response near limit points, Proc. Europe-US
Workshop on Nonlinear Finite Element Analysis in Structural Mechanics, Bochum 1980, W.
Wunderlich et al. (eds.), Springer-Verlag 1981, 63-89.
380 Ramm, E. (ed.): Buckling of Shells: Proceedings of a State-of-the-art Colloquium, Universität

Stuttgart, Germany, May 6-7, 1982, Springer-Verlag, Berlin 1982.
381 Ramm, E.: The Riks/Wempner approach - an extension of the displacement control method in
nonlinear analyses, Recent Advances in Non-linear Computational Mechanics, Pineridge Press
Ltd, Swansea 1982, 62-86.
382 Ramm, E.: From Reissner Plate Theory to Three Dimensions in Large Deformation Shell
Analysis, Zeitschrift für angewandte Mathematik und Mechanik (Journal of Applied
Mathematics and Mechanics) - ZAMM 80, 2000, 61-68.
383 Ramm, E. & Matzenmiller, A.: Large deformation shell analyses based on the degeneration
concept, Finite Element Methods for Plate and Shell Structures, Vol.1: Element Technology,
365-393, eds. T.J.R. Hughes, E. Hinton, Pineridge Press Ltd., Swansea, 1986.
384 Ramm, E. & Stegmüller, H.: The displacement finite element method in nonlinear buckling
analysis of shells, in [380], 1982, 201-235.
385 Ramm, E. & Wall., W. A.: Shells in Advanced Computational Environment, WCCM V:
Proceedings of the Fifth World Congress on Computational Mechanics, July 7-12, 2002,
Vienna 2002.
386 Ramm, E. & Wall., W. A.: Shell structures - a sensitive interrelation between physics and
numerics, Int. Journal for Numerical Methods in Engineering 60, 2004, 381-427.
387 Ramm, E. & Wall., W. A. (eds.): Computational Methods for Shells, (Special Issue of)
388 Rammerstorfer, F. G. (ed.): Nonlinear Analysis of Shells by Finite Elements, CISM Courses
and Lectures No. 328, Springer-Verlag, Vienna - New York 1992.
389 Rammerstorfer, F. G., Dorninger, K. & Starlinger, A.: Composite and sandwich shells, in
[388], 1992, 131-194.
390 Rammerstorfer, F. G., Dorninger, K., Starlinger, A., Skrna-Jakl, I. C.: Computational methods
in composite analysis and design, in [215], 1994, 209-231.
391 Rao, K. P.: A rectangular laminated anisotropic shallow thin shell finite element, Computer
392 Reddy, J. N.: Analysis of layered composite plates accounting for large deflections and
transverse shear strains, Recent Advances in Nonlinear Computational Mechanics, E. Hinton,
D. R. J. Owen & C. Taylor (eds.), Pineridge Press Ltd, Swansea 1982, 155-202.
393 Reddy, J. N.: A simple higher-order theory for laminated composite plates, Journal of Applied
394 Reddy, J. N.: On refined computational models of composite laminates, Int. Journal for
395 Reddy, J. N.: On the generalization of displacement-based laminate theories, Applied
Mechanics Reviews 42, 1989, S213-S222.
396 Reddy, J. N.: A general non-linear third-order theory of plates with moderate thickness, Int.
Journal of Non-Linear Mechanics 25, 1990, 677-686.
397 Reddy, J.N.: On laminated composite plates with integrated sensors and actuators, Engineering
Structures 21, 1999, 568-593.
398 Reddy, J. N.: Energy Principles and Variational Methods in Applied Mechanics, John Wiley &
Sons, Ltd., New York 2002.
399 Reddy, J. N. & Arciniega, R. A.: Shear deformation plate and shell theories: From Stavsky to
Present, Mechanics of Advanced Materials and Structures 11, 2004, 535-582.
400 Reddy, J. N. & Chandrashekhara, K.: Nonlinear analysis of laminated shells including
transverse shear strains, AIAA Journal 23, 1985, 440-441.
7. References 171
401 Reddy, J. N. & Liu, C. F.: A higher-order shear deformation theory of laminated elastic shells,
Int. Journal of Engineering Sciences 23, 1985, 669-683.
402 Reissner, E.: Small bending and stretching of sandwich-type shells, NACA Report No. 975,
1950, 483-508.
403 Reissner, E.: On consistent first approximations in the general linear theory of thin elastic
shells, Ingenieur-Archiv 40, 1971, 402-419.
404 Reissner, E.: A consistent treatment of transverse shear deformations in laminated anisotropic
plates, AIAA Journal 10, 1972, 716-718.
405 Reissner, E.: A note on two-dimensional finite-deformation theories of shells, Int. Journal
Non-Linear Mechanics 17, 1982, 217-221.
406 Rikards, R., Chate, A. & Ozolinsh, O.: Analysis for buckling and vibrations of composite
stiffened shells and plates, Composite Structures 51, 2001, 361-370.
407 Rohwer, K., Friedrichs, S. & Wehmeyer, C.: Analyzing laminated structures from fibre-
reinforced composite material – an assessment, Technische Mechanik 25, 2005, 59-79.
408 Rolfes, R. & Rohwer, K.: Improved transverse shear stresses in composite finite elements
based on First Order Shear Deformation Theory, Int. Journal for Numerical Methods in
Engineering 40, 1997, 51-60.
409 Sabir, A. B. & Lock, A. C.: The application of finite elements to the large deflection
geometrically non-linear behaviour of cylindrical shells, Variational Methods in Engineering
2. C. A. Brebbia & H. Tottenham (eds.). Southampton University Press, 1972, 7/66-7/75.
410 Sacco, E. & Reddy, J.N.: On first- and second-order moderate rotation theories of laminated
plates, Int. Journal for Numerical Methods in Engineering 33, 1992, 1-17.
411 Saigal, S., Kapania, R. K. & Yang, T. Y.: Geometrically nonlinear finite element analysis of
imperfect laminated shells, Journal Composite Materials 20, 1986, 197-214.
412 Saleeb, A. F., Chang, T. Y., Graf, W. & Yingyeunyong, S.: A hybrid/mixed model or non-
linear shell analysis and its applications to large-rotation problems, Int. Journal for Numerical
413 Sansour, C.: A Theory and finite element formulation of shells at finite deformations involving
thickness change: Circumventing the use of a rotation tensor, Arch. Appl. Mech. 65, 1995, 194-
216.
414 Sansour, C. & Bednarczyk, H.: The Cosserat surface as a shell model, theory and finite-
element formulation, Computer Methods in Applied Mechanics & Engineering 120, 1995, 1-
32.
415 Sansour, C. & Boćko, J.: Hybrid stress versus hybrid strain finite elements for a geometrically
exact theory of shells with drilling degrees of freedom, Int. Conf. of Numerical Methods in
Continuum Mechanics, Stara Leśna, 19-22 Sept. 1994, 1994, 159-166.
416 Sansour, C. & Boćko, J.: On hybrid stress, hybrid strain and enhanced strain finite element
formulations for a geometrically exact shell theory with drilling degrees of freedom, Int.
417 Sansour, C. & Bufler, H.: An exact finite rotation shell theory, its mixed variational
formulation and its finite element implementation, Int. Journal for Numerical Methods in
Engineering 34, 1992, 73-115.
418 Sansour, C. & Kollmann, F. G.: Families of 4-node and 9-node finite elements for a finite
deformation shell theory. An assessment of hybrid stress, hybrid strain and enhanced strain
elements, Computational Mechanics 24, 2000, 435-447.
419 Sansour, C. & Wagner, W.: Multiplicative updating of the rotation tensor in the finite element
analysis of rods and shells – a path independent approach, Computational Mechanics 31, 2003,
153-162.
420 Savithri, S. & Varadan, T. K.: Large deflection analysis of laminated composite plates, Int.
Journal Non-Linear Mechanics 28, 1993, 1-12.
421 Schieck, B., Pietraszkiewicz, W. & Stumpf, H.: Theory and numerical analysis of shells
undergoing large elastic strains, Int. Journal of Solids & Structures 29, 1992, 689-709.
422 Schlebusch, R., Matheas, J. & Zastrau, B. W.: A three-dimensional surface-related shell theory
for the treatment of contact problems, WCCM V: Proceedings of the Fifth World Congress on
Computational Mechanics, July 7-12, 2002, Vienna, Austria, H.A. Mang, F.G. Rammerstorfer
& J. Eberhardsteiner (eds.), CD-ROM, Vienna 2002.
423 Schlebusch, R., Matheas, J. & Zastrau, B. W.: On surface-related shell theories for the
numerical simulation of contact problems, Journal of Theoretical and Applied Mechanics
(Polish Society of Theoretical and Applied Mechanics) 41, 2003, 623-642.
424 Schmauder, S.: Computational Mechanics, Annual Review of Materials Research 32, 2002,
437–465.
425 Schmidt, R. & Librescu, L.: Further results concerning the refined theory of anisotropic
laminated composite plates, Journal of Engineering Mathematics 28, 1994, 407-425.
426 Schmidt, R. & Reddy, J. N.: A refined small strain and moderate rotation theory of elastic
anisotropic shells, Journal of Applied Mechanics, Trans. ASME 55, 1988, 611-617.
427 Schmidt, R. & Weichert, D.: A refined theory of elastic-plastic shells at moderate rotations,
Zeitschrift für angewandte Mathematik und Mechanik (Journal of Applied Mathematics and
Mechanics) - ZAMM 69, 1989, 11-21.
428 Schmit Jr., L.A. & Monforton, G.R.: Finite deflection Discrete Element Analysis of sandwich
plates and cylindrical shells with laminated faces, AIAA Journal 8, 1970, 1454-1461.
429 Shen, H.-S.: Postbuckling of Shear Deformable Laminated Cylindrical Shells, ASCE Journal
of Engineering Mechanics 128, 2002, 296-307.
430 Shimodaira, H.: Equivalence between mixed models and displacement models using reduced
integration, Int. Journal for Numerical Methods in Engineering 21, 1985, 89-104.
431 Shu, X.-P.: A refined theory of laminated shells with higher-order transverse shear
deformation, Int. Journal of Solids & Structures 34, 1997, 673-683.
432 Simitses, G. J.: Buckling of moderately thick laminated cylindrical shells: A review,
Composites Part B 27B, 1996, 581-587.
433 Simmonds, J.G.: Some comments on the status of shell theory at the end of the 20th century -
Complaints and correctives, AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics,
and Materials Conference and Exhibit, 38th, and AIAA/ASME/AHS Adaptive Structures
Forum, Kissimmee, FL, Apr. 7-10, 1997, Collection of Technical Papers, AIAA-1997-1074,
1997, 1912-1921.
434 Simmonds, J.G. & Danielson, D. A.: Nonlinear shell theory with finite rotation and stress-
function vectors, Journal of Applied Mechanics, Trans. ASME 39, 1972, 1085-1090.
435 Simo, J. C., Fox, D. D. & Rifai, M. S.: On a stress resultant geometrically exact shell model.
Part I: Formulation and optimal parametrization, Computer Methods in Applied Mechanics &
Engineering 72, 1989, 267-304.
Part II: The linear theory; Computational aspects, Computer Methods in Applied Mechanics &
Engineering 73, 1989, 53-92.
Part III: Computational aspects of the nonlinear theory, Computer Methods in Applied
438 Simo, J. C. & Hughes, T. J. R.: On the variational foundations of Assumed Strain Methods,
Journal of Applied Mechanics, Trans. ASME 53, 1986, 51-54.
7. References 173
439 Simo, J.C. & Rifai, S.: A class of mixed assumed strain methods and the method of
incompatible modes, Int. Journal for Numerical Methods in Engineering 29, 1990, 1595-1638.
440 Simo, J. C., Rifai, M. S. & Fox, D. D.: On a stress resultant geometrically exact shell model.
Part IV: Variable thickness shells with through-the-thickness stretching, Computer Methods in
441 Simo, J. C, Wriggers, P., Schweizerhof, K. H. & Taylor, R. L.: Finite deformation
postbuckling analysis involving inelasticity and contact constraints, Int. Journal for Numerical
442 Sing, G., Venkateswara Rao, G. & Iyengar, N. G. R.: Geometrically nonlinear flexural
response characteristics of shear deformable unsymmetrically laminated plates, Computers &
Structures 53, 1994, 69-81.
443 Snell, M. & Morley, N.: The compression buckling behaviour of highly curved panels of
carbon fibre reinforced plastic, Proceedings of 5th Int. Conf. on Composite Materials, 1985,
1327-1354.
444 Soldatos, K. P. & Watson, P.: A method for improving the stress analysis performance of one-
and two-dimensional theories for laminated composites, Acta Mechanica 123, 1997, 163-186.
445 Somashekar, B.R., Prathap, G. & Ramesh Babu, C.: A field-consistent four noded laminated
anisotropic plate/shell element, Computers & Structures 25, 1987, 345-353.
446 Stander, N., Matzenmiller, A. & Ramm, E.: An assessment of assumed strain methods in finite
rotation shell analysis, Engineering Computations 6, 1989, 58-66.
447 Stockton, S. L.: Engineering and Design: Composite Materials for Civil Engineering
Structures, Technical Letter No. 1110-2-548, U.S. Army Corps of Engineers Washington, DC
1997.
448 Stolarski, H.: Some concepts of the Assumed Strain Method, Computational Mechanics ’91,
Theory and Applications. Proceedings of ICCES’91. S.N. Atluri, D.E. Beskos, R. Jones and G.
Yagawa (Eds), ICES Publications, W.G. Wolfe Associates, Alpharetta, Georgia/USA, 1991,
1032-1037.
449 Stolarski, H. & Belytschko, T.: Shear and membrane locking in curved C0elements, Computer
450 Sun, C.T. & Chin, H.: Analysis of asymmetric composite laminates, AIAA Journal 26, 1988,
714-718.
451 Surana, K. S.: Geometrically nonlinear formulation for the axisymmetric shell elements, Int.
452 Surana, K. S.: A generalized geometrically nonlinear formulation with large rotations for finite
elements with rotational degrees of freedom, Computers & Structures 24, 1986, 47-55.
453 Sutyrin, V. G. & Hodges, D. H.: On asymptotically correct linear laminated plate theory, Int.
454 Sze, K. Y., He, L.-W. & Cheung, Y. K.: Predictor-corrector procedures for analysis of
laminated plates using standard Mindlin finite element models, Composite Structures 50, 2000,
171-182.
455 Sze, K. Y., Liu, X. H. & Lo, S. H.: Popular benchmark problems for geometric nonlinear
analysis of shells, Finite Elements in Analysis and Design 40, 2004, 1551-1569.
456 Sze, K. Y., Yao, L.-Q. & Pian, T.H.H.: An eighteen-node hybrid-stress solid-shell element for
homogenous and laminated structures, Finite Elements in Analysis & Design 38, 2002, 353-
374.
457 Sze, K. Y. & Zheng, S.-J.: A stabilized hybrid-stress solid element for geometrically nonlinear
homogeneous and laminated shell analyses, Computer Methods in Applied Mechanics &
Engineering 191, 2002, 1945-1966.
458 Takano, N., Ohnishi, Y., Zako, M. & Nishiyabu, K.: The formulation of homogenization
method applied to large deformation problem for composite materials, Int. Journal of Solids &
Structures 37, 2000, 6517-6535.
459 Tanov, R. & Tabiei, A.: A note on finite element implementation of sandwich shell
homogenization, Int. Journal for Numerical Methods in Engineering 48, 2000, 467-473.
460 Tanov, R. & Tabiei, A.: A simple correction to the first-order shear deformation shell finite
element formulations, Finite Elements in Analysis and Design 35, 2000, 189-197.
461 Tarn, J.-Q. & Wang, Y.-B.: A refined asymptotic theory and computational model for
multilayered composite plates, Computer Methods in Applied Mechanics & Engineering 145,
1997, 167-184.
462 Taylor, R.L.: Ritz and Galerkin: the road to the Finite Element Method, IACM Expressions 12,
2002, 2-5.
463 Telang, N. M., Dumlao, C., Mehrabi, A. B., Ciolko, A.T. & Gutierrez J.: Field Inspection of
In-Service FRP Bridge Decks, Transportation Research Board, NCHRP Report 564, 2006.
464 Teng, J.G. & Hong, T.: Nonlinear thin shell theories for numerical buckling predictions, Thin-
Walled Structures 31, 1998, 89-115.
465 Tessler, A. & Hughes, T. J. R.: An improved treatment of transverse shear in the Mindlin-type
four-node quadrilateral element, Computer Methods in Applied Mechanics & Engineering 39,
1983, 311-335.
466 Toledano, A. & Murakami, H.: A composite plate theory for arbitrary laminate configurations,
467 Toledano, A. & Murakami, H.: A High-Order Laminated Plate Theory with Improved In-Plane
Responses, Int. Journal of Solids & Structures 23, 1987, 111-131.
468 Tong, L., Mouritz, A. P. & Bannister, M. K.: 3D Fibre Reinforced Polymer Composites,
Elsevier Science Ltd., Oxford 2002.
469 Toorani, M. H. & Lakis, A. A.: General equations of anisotropic plates and shells including
transverse shear deformations, rotary inertia and initial curvature effects, Journal of Sound and
Vibration 237, 2000, 561-615.
470 Tsai, C. T., Palazotto, A. N. & Dennis, S. T.: Large-rotation snap-through buckling in
laminated cylindrical panels, Finite Elements in Analysis and Design 9, 1991, 65-75.
471 Valid, R.: The Nonlinear Theory of Shells Through Variational Principles: From Elementary
Algebra to Differential Geometry, John Wiley & Sons, Chichester, 1995.
472 Valid, R.: Duality in nonlinear theory of shells, Int. Journal of Engineering Science 37, 1999,
1521-1547.
473 Van Den Einde, L., Zhao, L. & Seible, F.: Use of FRP composites in civil structural
applications, Construction & Building Materials 17, 2003, 389-403.
474 Vassilev, V. M.: Symmetry Groups and Equivalence Transformations in the Nonlinear
Donnell–Mushtari–Vlasov Theory for Shallow Shells, Journal of Theoretical and Applied
Mechanics (Bulgarian Academy of Sciences) 27, 1997, 43-51.
475 Vasiliev, V. V. & Morozov, E. V.: Mechanics and Analysis of Composite Materials, Elsevier
Science Ltd, Oxford 2001.
476 Vinson, J. R. & Chou, T.-W.: Composite Materials and Their Use in Structures, Applied
Science Publishers Ltd, London, 1975.
477 Vlachoutsis, S.: Shear correction factors for plates and shells, Int. Journal for Numerical
478 Vu-Quoc, L., Deng, H. & Tan, X. G.: Geometrically-exact sandwich shells: The static case,
7. References 175
479 Vu-Quoc, L. & Tan, X.G.: Optimal solid shells for non-linear analyses of multilayer
composites. I. Statics, Computer Methods in Applied Mechanics & Engineering 192, 2003,
975-1016.
480 Wagner W.: Zur Formulierung eines geometrisch nichtlinearen Finite Elementes für
zylindrische Faserverbundschalen (On formulation of geometrically nonlinear finite elements
for fiber reinforced cylindrical shells) (in German), Statik und Dynamik in Konstruktiven
Ingenieurbau, Festschrift Wilfried B. Krätzig, SFB 151 – Berichte nr 23, 1992, B3-B10.
481 Wagner, W. & Gruttmann, F.: FE–Modeling of Fiber Reinforced Polymer Structures, WCCM
V: Proceedings of the Fifth World Congress on Computational Mechanics, July 7-12, 2002,
Vienna 2002.
482 Wang, C. M., Lim, G. T., Reddy, J. N. & Lee, K. H.: Relationships between bending solutions
of Reissner and Mindlin plate theories, Engineering Structures 23, 2001, 838-849.
483 Wang, L. & Thierauf, G.: Finite rotations in non-linear analysis of elastic shells, Computers &
Structures 79, 2001, 2357-2367.
484 Wanji, C. & Cheung, Y. K.: Refined quadrilateral element based on Mindlin/Reissner plate
theory, Int. Journal for Numerical Methods in Engineering 47, 2000, 605-627.
485 Waszczyszyn, Z.: Stability problems and methods of analysis of nonlinear FEM equations, in
[245], 1998, 253-323.
486 Waszczyszyn, Z., Cichoń, C. & Radwańska, M.: Stability of structures by finite element
method, Elsevier, Amsterdam, 1994.
487 Website of ACMA: Composites Basics: Materials, FRP Educational Campus, American
Composites Manufacturers Association, 2004, http://www.mdacomposites.org/mda/, Last
visited on November 20, 2006.
488 Website of Composites IQ: Construction and Civil Engineering, State of the Art, Knowledge
bank, Composites Intelligence Exchange Limited at the http://www.compositesiq.com/, Last
visited on November 25, 2006.
489 Wempner, G.: Mechanics and finite elements of shells, Applied Mechanics Reviews 42, 1989,
129-142.
490 Wempner, G. A, Oden, J. T. & Kross, D. A.: Finite-element analysis of thin shells, Journal of
Engineering Mechanics ASCE 94, 1968, 1273-1294.
491 Wempner, G., Talaslidis, D. & Hwang, C. M.: A simple and efficient approximation of shells
via finite quadrilateral elements, Journal of Applied Mechanics, Trans. ASME 49, 1982, 115-
120.
492 Whitney, J. M.: Shear correction factors for orthotropic laminates under static load, Journal of
Applied Mechanics, Trans. ASME 40, 1973, 302-303.
493 Whitney, J. M. & Pagano, N. J.: Shear deformation in heterogeneous anisotropic plates,
494 Williams, T. O. & Addessio, F. L.: A general theory for laminated plates with delaminations,
Int. Journal of Solids & Structures 34, 1997, 2003-2024.
495 Wisnom, M. R.: The effect of fibre rotation in ± 45º tension tests on measured shear properties,
Composites 26, 1995, 25-32.
496 Wiśniewski, K.: A shell theory with independent rotations for relaxed Biot stress and right
stretch strain, Computational Mechanics 21, 1998, 101-122.
497 Wittrick W. H.: Analytical, three-dimensional elasticity solutions to some plate problems, and
some observations on Mindlin's plate theory, Int. Journal of Solids & Structures 23, 1987, 441-
464.
498 Woo, K. S., Hong, C. H. & Basu, P. K.: Materially and geometrically nonlinear analysis of
laminated anisotropic plates by p-version of FEM, Computers & Structures 81, 2003, 1653-
1662.
499 Woźniak, C. (ed.): Mechanika sprężystych płyt i powłok (Mechanics of elastic plates and
shells), (in Polish), Polish Scientific Publishers PWN Warszawa, 2001.
500 Wriggers, P. & Gruttmann, F.: Thin shells with finite rotations: Theory and Finite Element
formulation, in [328] CED-Vol.3, 1989, 135-159.
501 Wriggers, P. & Gruttmann, F.: Thin shells with finite rotations formulated in Biot stresses:
Theory and Finite Element formulation, Int. Journal for Numerical Methods in Engineering
36, 1993, 2049-2071.
502 Yang, H. T. Y., Saigal, S., Masud, A. & Kapania, R. K.: A survey of recent shell finite
elements, Int. Journal for Numerical Methods in Engineering 47, 2000, 101-127.
503 Yeo, S. T. & Lee, B. C.: Equivalence between enhanced assumed strain method and assumed
stress hybrid method based on the Hellinger-Reissner principle, Int. Journal for Numerical
504 Yu, G., Guang-Yau, T., Chaturvedi, S., Adeli, H. & Shao, Q.Z.: A finite element approach to
global-local modeling in composite laminate analysis, Computers & Structures 57, 1995, 1035-
1044.
505 Yu, W. & Hodges, D. H.: A geometrically nonlinear shear deformation theory for composite
shells, Journal of Applied Mechanics, Trans. ASME 71, 2004, 1-9.
506 Yu, W., Hodges, D. H. & Volovoi, V. V.: Asymptotic generalization of Reissner-Mindlin
theory: accurate three-dimensional recovery for composite shells, Computer Methods in
507 Zahrouni, H., Cochelin, B. & Potier-Ferry, M.: Computing finite rotations of shells by an
asymptotic-numerical method, Computer Methods in Applied Mechanics & Engineering 175,
1999, 71-85.
508 Zhang, W. C. & Evans, K. E.: Numerical prediction of the mechanical properties of anisotropic
composite materials, Computers & Structures 29, 1988, 413-422.
509 Zhang, Y.X. & Kim K.S.: Two simple and efficient displacement-based quadrilateral plate
elements for the analysis of composite laminated plates, Int. Journal for Numerical Methods in
Engineering 61, 2004, 1771-1796.
510 Zienkiewicz, O. C. & Taylor, R. L.: The Finite Element Method, 5th ed. Volume 1: The Basis
& Volume 2: Solid Mechanics, Butterworth-Heinemann, Oxford, 2002.
511 Zienkiewicz, O. C., Taylor, R. L. & Too, J. M.: Reduced integration technique in general
analysis of plates and shells, Int. Journal for Numerical Methods in Engineering 3, 1971, 275-
290.
512 Ziyaeifar, M. & Elwi, A. E.: Degenerated plate-shell elements with refined transverse shear
strains, Computers & Structures 60, 1996, 1079-1091.
GEOMETRICALLY NON-LINEAR ANALYSIS OF LAYERED
COMPOSITE PLATES AND SHELLS
An appropriate computational representation of laminated composites and sandwich panels is a

matter of a current concern in many fields of structural engineering. The key objective of the research
portrayed in this report was to develop a numerical model for geometrically non-linear analysis of
moderate thick multi-layered panels within the range of the large rotation shell theory.
An essential part of the present report is devoted to a state-of-the-art review on the modeling of
laminated composite and sandwich panels. Almost five hundred texts have been included in the
survey with the focus put on three major topics: theoretical models for multi-layered plates and shells,
geometrically non-linear large rotation analysis, and FEM implementation of various plate and shell
theories. As a result of the review, one could notice a lack of a single numerical model capable for a
universal representation of all layered composite and sandwich panels. Usually, with the increase of
the range of rotations considered in the particular model, one can observe the decrease of the degree
of complexity of the through-the-thickness representation of deformation profiles. It is also quite
interesting that a very limited number of examples available in the literature for geometrically non-
linear analysis of laminated shells accounts for the large rotations.
The next part of the report contains a systematic construction of a numerical model for a
geometrically non-linear analysis of elastic laminated shells undergoing large rotations. A smooth
geometry of the middle surface has been assumed together with the Equivalent Single Layer approach
within the scope of the First Order Shear Deformation theory. The strain-displacement relations
established for the large rotation shell theory (LRT) have been consistently examined according to the
assumption of the director inextensibility with an appropriate treatment of finite rotations. The series
of large deformation formulations for laminated shells have been constructed with gradually increased
and clearly distinguished levels of geometrical non-linearity, from the Refined von Kármán Theory
(RVK5), through the Moderate Rotation Theory (MRT5) and the (Simplified) Large Rotation Theory
with 5 degrees of freedom (LRT5) to the Large Rotation Theory (LRT56) formulation based on the
application of Euler angles.
The incremental strategy of the large deformations analysis has been established basing on the
Total Lagrangian approach. The Newton-Raphson equilibrium iterations and the Riks-Wempner-
Ramm arc-length control method have been applied for tracing of the non-linear equilibrium paths.
Standard quadrilateral finite elements with the Lagrange and Serendipity interpolation schemes have
been considered for the possible application. After a set of numerical tests performed within the range
of small displacements for various integration and interpolation schemes (including the Assumed
Natural Strain approach), the 8-node Serendipity shell element with a uniformly reduced integration
has been selected as the most promising choice for applications in the large deformation analysis. The
proposed finite element formulation for the linear and non-linear analysis of multi-layered shells has
been implemented in a family of computer programs written in FORTRAN.
A comparative finite element analysis of various sample problems of a non-linear, large rotation
response of composite laminated plate and shell structures composes the largest part of the present
report. Ten examples selected from the literature have been analyzed with the LRT56 formulation and
in all cases this model was positively validated by confronting the obtained results with the available
reference solutions. Additionally, the results calculated with the simplified non-linear models, like the
RVK5, MRT5 and LRT5 have been included in the comparison. A detailed examination of the
obtained results allows one to make several significant observations, with the main conclusion that
the proper updating procedure of the rotations in the geometrically non-linear analysis of shells is of
extreme importance as soon as the range of moderate rotations is exceeded. Moreover, the proper
updating of the rotational degrees of freedom appears to be more important for the accuracy of the
large rotation solution for moderately thick composite shells than the refined representation of the
transverse shear strains incorporated in the Higher Order Shear Deformation models.
GEOMETRYCZNIE NIELINIOWA ANALIZA WARSTWOWYCH
PŁYT I POWŁOK KOMPOZYTOWYCH
Zauważalne ostatnio zwiększenie skali wykorzystania dźwigarów warstwowych i laminatów

kompozytowych w różnych konstrukcjach inżynierskich powoduje, że podstawowego znaczenia
nabiera kwestia właściwego doboru modeli obliczeniowych dla tych elementów konstrukcyjnych.
Zagadnienie poprawnego modelowania matematycznego statyki wielowarstwowych płyt i powłok
sprężystych w zakresie dużych obrotów stanowi główny temat niniejszej monografii.
Praca zawiera wyniki badań prowadzonych przez autora w ciągu ostatnich 15 lat, jednak
intensywność pracy nad zagadnieniami będącymi przedmiotem niniejszej rozprawy w całym tym
okresie była bardzo zróżnicowana. Wszystko zaczęło się od udziału autora w projekcie badawczym
"Teoria i analiza nieliniowa MES sprężystych i sprężysto-plastycznych konstrukcji anizotropowych z
uwzględnieniem zniszczenia” finansowanym przez Deutsche Forschungsgemeinschaft. Projekt ten
był realizowany w latach 1991-1993 na Uniwersytecie w Wuppertalu pod kierunkiem prof. Dietera
Weicherta i Dr-Ing. Rüdigera Schmidta. Równolegle, aczkolwiek do pewnego stopnia niezależnie od
głównego tematu projektu DFG została zainicjowana współpraca badawcza z Dr-Ing. Schmidtem
poświecona komputerowej implementacji teorii umiarkowanych obrotów dla powłok warstwowych,
którą R. Schmidt i J. N. Reddy zaproponowali w swojej wspólnej pracy z 1988 roku. Pierwsze
działające wersje programów komputerowych do geometrycznie nieliniowej analizy powłok
warstwowych autor zbudował jeszcze w Wuppertalu. Wyniki badań z tego okresu zostały
zaprezentowane pod koniec 1992 roku w dwóch referatach konferencyjnych oraz w artykule (autorzy:
Kreja, Schmidt & Reddy) z 1997 roku. Po powrocie do Gdańska, autor kontynuował badania
poszukując możliwości poprawienia dokładności otrzymywanych wyników poprzez wzbogacenie
związków odkształcenie-przemieszczenie, zastosowanie założonych pól odkształceń oraz
poprawienie procedur uaktualniania dużych obrotów. W 1997 roku autor otrzymał pierwsze wyniki w
zakresie analizy skończonych obrotów powłok warstwowych wykorzystując własny program
zbudowany na podstawie teorii zaproponowanej dla powłok anizotropowych przez Librescu w pracy
z 1987 roku. Po dłuższej przerwie spowodowanej zaangażowaniem w innych projektach badawczych,
autor powrócił do zagadnień związanych z analizą dużych obrotów powłok warstwowych podczas
miesięcznego pobytu w RWTH Aachen latem 2003 roku. Rezultatem tego wyjazdu był wspólny
artykuł z R. Schmidtem, opublikowany drukiem w 2006 roku. Właściwe prace nad przygotowaniem
niniejszej rozprawy rozpoczęły się w połowie 2004 roku, jednak ich intensywność była ograniczona
innymi obowiązkami autora. Ostatecznie powstała monografia, którą można traktować jako pewną
klamrę spinającą wieloletni okres badań autora, jednak wiele analiz prezentowanych w niniejszym
tekście nie było jeszcze nigdzie publikowanych.
Znacząca część monografii jest poświecona studiom dostępnych materiałów źródłowych w
zakresie: modeli teoretycznych dla płyt i powłok wielowarstwowych, analizy geometrycznie
nieliniowej ze szczególnym uwzględnieniem opisu skończonych obrotów oraz numerycznych
implementacji różnych teorii płyt i powłok. Przedstawiony przegląd literatury obejmuje blisko pięćset
pozycji, z czego ponad połowa traktuje bezpośrednio o problemach analizy konstrukcji warstwowych.
Jako podstawową konkluzję z tego przeglądu, należy chyba przyjąć spostrzeżenie, że nie istnieje
jeden uniwersalny model matematyczny zdolny do efektywnego reprezentowania wszystkich płyt i
powłok warstwowych. Praktyczna przydatność różnych sformułowań teoretycznych zależy istotnie od
specyfiki analizowanego problemu. Przy oczywistych ograniczeniach zdolności obliczeniowych,
głównie natury sprzętowej i ekonomicznej, całkiem naturalnym zjawiskiem jest zauważalna tendencja
do zmniejszania rzędu dokładności reprezentacji deformacji profilu poprzecznego wraz ze wzrostem
stopnia nieliniowości analizowanych zagadnień. Można też zauważyć, że zbiór dostępnych w
literaturze przykładów testowych dla powłok laminowanych w zakresie dużych obrotów jest bardzo
ograniczony, jeżeli go porównamy z bogactwem przykładów dotyczących powłok izotropowych.
Summary 179
W dalszej części monografii przedstawiono systematyczną konstrukcję modelu obliczeniowego

dla geometrycznie nieliniowej analizy sprężystych powłok laminowanych podlegających dużym
obrotom. Przyjęto koncepcję zastępczego modelu jednowarstwowego w ramach teorii ścinania
pierwszego rzędu. Istotnym ograniczeniem jest założenie o gładkiej powierzchni środkowej powłoki,
bez załamań i rozgałęzień. Związki odkształcenie-przemieszczenie wyprowadzone dla teorii powłok
w zakresie skończonych obrotów zostały odpowiednio zmodyfikowane poprzez pominiecie zmiany
grubości powłoki podczas deformacji. Następnie rozważono kilka przybliżonych wariantów
związków odkształcenie-przemieszczenie i porównano je z relacjami proponowanymi przez innych
autorów. W rezultacie powstał zestaw matematycznych modeli statyki dużych deformacji powłok
warstwowych charakteryzujący się stopniowym wzrostem stopnia geometrycznej nieliniowości, od
najprostszego modelu odpowiadającego teorii von Kármána (RVK5), poprzez teorię umiarkowanych
obrotów (MRT5) i 5-parametrową (uproszczoną) teorię dużych obrotów (LRT5), aż do teorii dużych
obrotów z opisem skończonych obrotów przy zastosowaniu kątów Eulera (LRT56).
Zgodnie z zasadami Stacjonarnego Opisu Lagrangea skonstruowano przyrostową strategię
analizy dużych deformacji powłok warstwowych bazującą na technice sterowania parametrem ścieżki
Riksa-Wempnera-Ramma oraz iteracjach równowagi metodą Newtona-Raphsona. W algorytmie
Metody Elementów Skończonych przewidziano wykorzystanie standardowych elementów
izoparametrycznych ze schematami interpolacji Lagrangea i Serendipity. Po serii testów
numerycznych przeprowadzonych w zakresie analizy liniowej dla różnych schematów interpolacji i
całkowania (łącznie z techniką założonych pól odkształceń) wytypowano 8-węzłowy element
Serendipity z równomiernie zredukowanym całkowaniem, jako najbardziej obiecujący do
zastosowania w analizie dużych deformacji.
Najobszerniejsza część rozprawy zawiera porównawczą analizę nieliniowej odpowiedzi płyt i
powłok w zakresie dużych obrotów wyznaczonej metodą elementów skończonych dla szeregu
reprezentacyjnych przykładów zaczerpniętych z literatury. Każdy z dziesięciu przykładów był
analizowany przy zastosowaniu modelu LRT56 i za każdym razem uzyskano potwierdzenie
poprawności otrzymanych wyników poprzez skonfrontowanie ich z dostępnym rozwiązaniem
odniesienia. Dodatkowo w porównaniach tych ujęto wyniki otrzymane przy zastosowaniu modeli
uproszczonych, takich jak RVK5, MRT5 oraz LRT5. Szczegółowa analiza uzyskanych rezultatów
pozwoliła na sformułowanie następujących wniosków:
a) w momencie przekroczenia zakresu umiarkowanych obrotów dokładność otrzymanego
rozwiązania istotnie zależy od prawidłowego uaktualniania dużych obrotów; właściwe
traktowanie obrotowych stopni swobody ma czasami większy wpływ na dokładność
wyników niż uwzględnienie wszystkich członów w związkach przemieszczenie-
odkształcenie;
b) zastosowanie uproszczonego wariantu teorii dużych obrotów (LRT5) jest właściwie
ograniczone do zakresu małych i umiarkowanych obrotów, bogatsza reprezentacja
odkształceń w modelu LRT5 względem sformułowania MRT5 rzadko przekłada się na
większą dokładność uzyskanych rozwiązań;
c) wzbogacona interpolacja odkształceń poprzecznego ścinania w modelach TOSD jest mniej
istotna dla dokładności wyników analizy dużych obrotów umiarkowanie grubych powłok
kompozytowych niż poprawne uaktualnianie obrotowych stopni swobody;
d) zmodyfikowane modele pomijające człony nieliniowe dla odkształceń poprzecznego
ścinania (m-LRT5 oraz m-LRT56) nie zapewniają uzyskania poprawnych rozwiązań;
e) paradoksalnie, najbardziej uproszczony z rozpatrywanych modeli nieliniowych,
odpowiadający teorii von Kármána model RVK5 zaskakująco często zapewniał lepsze
rozwiązanie niż bardziej od niego zaawansowane sformułowania MRT5 i LRT5;
f) dzięki porównaniu rezultatów uzyskanych przy zastosowaniu różnych modeli
obliczeniowych o stopniowo wzrastającym stopniu geometrycznej nieliniowości możliwe
było wskazanie przykładów, które kwalifikują się do zastosowania jako przykłady testowe
dla analizy powłok warstwowych w zakresie dużych obrotów;
g) w analizie utraty stateczności cylindrycznych utwierdzonych powłok kompozytowych
poddanych osiowemu ściskaniu z pewnym zaskoczeniem zaobserwowano, że w przypadku
180 Summary
określonych schematów uwarstwienia, nastąpić może wzrost wielkości obciążenia

krytycznego mimo redukcji globalnej sztywności konstrukcji.
Oryginalnymi elementami zawartymi w monografii, zdaniem jej autora, są:
a) systematyczna konstrukcja modelu obliczeniowego dla geometrycznie nieliniowej analizy
sprężystych powłok wielowarstwowych w zakresie dużych obrotów, ze specjalnym
uwzględnieniem jego praktycznej implementacji w autorskich programach komputerowych;
b) metodyczne zbadanie konsekwencji wprowadzenia różnych założeń upraszczających
stosowanych w opisie dużych obrotów;
c) wskazanie przykładów, które kwalifikują się na przykłady testowe dla analizy powłok
warstwowych w zakresie dużych obrotów;
d) bogaty przegląd literatury w zakresie dostępnych metod analizy płyt i powłok
wielowarstwowych.
Przedstawiona w pracy analiza uzyskanych rezultatów wykazała, że zaproponowany model
LRT56 może być skutecznym narzędziem obliczeniowym w geometrycznie nieliniowej analizie
umiarkowanie grubych sprężystych powłok wielowarstwowych w zakresie dużych obrotów, można
jednak wskazać szereg pomysłów na rozszerzenie zakresu analizy, przykładowo:
a) uwzględnienie imperfekcji geometrycznych i efektów termicznych, szczególnie w aspekcie
analizy stateczności;
b) skonstruowanie specjalnych elementów przejściowych umożliwiających na połączenie
modeli przestrzennych i powierzchniowych w jednej analizie;
c) zastosowanie elementów skończonych niskiego rzędu bazujących na koncepcji założonych
pól odkształceń;
d) uwzględnienie efektów nieliniowości materiałowej;
e) detekcja uszkodzeń kompozytów (w formie rozwarstwienia, pękania matrycy lub zrywania
włókien zbrojenia);
f) analiza kompozytów o funkcyjnie zmiennej strukturze;
g) badanie sprzężenia elektro-mechanicznego w dźwigarach z warstwami piezoelektrycznymi.

Geometrically Non-Linear Analysis of Layered Composite Plates and Shells

Uploaded by

Copyright:

Available Formats

You might also like

Geometrically Non-Linear Analysis of Layered Composite Plates and Shells

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geometrically Non-Linear Analysis of Layered Composite Plates and Shells

Uploaded by

Copyright:

Available Formats

IRENEUSZ KREJA

LIST OF SYMBOL AND ABBREVIATIONS.................................................................................. 9

1.1. Plates and shells – from Nature to space industry ......................................................... 13

1.2. Laminated composites and sandwich panels................................................................... 13

1.3. Scope and objectives of this report .................................................................................. 16

2. LITERATURE REVIEW AND MODELING CONSIDERATIONS ....................................... 18

2.1. Development of theoretical models for plates and shells................................................ 18

2.2. Geometrically non-linear analysis of plates and shells .................................................. 22

2.3. Theoretical models for layered thin-walled structures .................................................. 25

2.4. Numerical implementation of plate and shell theories .................................................. 35

2.5. Modeling considerations resulted from the presented review ...................................... 51

3. INCREMENTAL FORMULATION OF NONLINEAR SHELL ANALYSIS ........................ 52

3.1. Incremental formulation .................................................................................................. 52

3.2. Shell geometry .................................................................................................................. 53

3.3. Shell deformation ............................................................................................................. 57

3.4. Strain-displacement relations .......................................................................................... 59

3.5. Virtual work principle ..................................................................................................... 72

3.6. Constitutive relations ....................................................................................................... 77

4. FINITE ELEMENT METHOD IMPLEMENTATION............................................................. 82

4.1. Finite element discretization of the problem ................................................................... 82

4.2. Incremental equilibrium equations of FE model ............................................................ 84

4.3. Solving of incremental equilibrium equations ............................................................... 86

4.4. Finite elements used in this study .................................................................................... 90

4.5. Computer implementation of the proposed FEA algorithm .......................................... 92

4.6. Assessment of selected elements in linear analysis ......................................................... 93

5. LARGE DEFORMATION ANALYSIS EXAMPLES ............................................................ 103

5.1. Instability of clamped-hinged circular arches............................................................... 103

5.2. Clamped laminated shallow arch under point load...................................................... 108

5.4. Glass-epoxy cylinder under internal pressure ............................................................. 116

5.5. Asymmetric cross-ply simply supported plate strip .................................................... 117

5.7. Stretching of an open cylinder ....................................................................................... 125

5.8. Pinched hemispherical shell with 18° hole .................................................................... 129

5.9. Axial compression of composite cylindrical panel ........................................................ 133

6. CONCLUSIONS AND FUTURE PERSPECTIVES ................................................................ 148

6.1. Concluding Remarks....................................................................................................... 148

6.2. Original Contribution ..................................................................................................... 150

6.3. Recommendations and Future Perspectives.................................................................. 150

7. REFERENCES ............................................................................................................................ 151

SUMMARY IN ENGLISH ............................................................................................................. 177

SUMMARY IN POLISH ................................................................................................................ 178

LRT – Large Rotation Theory

1.1. Plates and shells – from Nature to space industry

1.2. Laminated composites and sandwich panels

Fig. 1.1: An example of a FRC three-layer laminate

A typical fiber reinforced composite laminate is made of a number of unidirectional

Aluminum alloy Composite laminates composed with 60% fibre and

Considering its light weight, a lamina of fiber reinforced composite is remarkably

Fig. 1.3: Sandwich plate with a honeycomb core

It is easy to find a prototype of all modern composite laminates in a traditional layered

1.3. Scope and objectives of this report

The following fundamental assumptions have been adopted in this work:

The content is organized as follows. Chapter 2 provides a state-of-the-art in the

LITERATURE REVIEW AND MODELING CONSIDERATIONS

2.1. Development of theoretical models for plates and shells

Fig. 2.1: Simplified representation of the FOSD model

for a linear through-the-thickness distribution of the transverse normal displacement

Fig. 2.2: Simplified representation of the Kirchhoff-Love model

where the skew-symmetric tensor ( m« known as a spin tensor is introduced as