Professional Documents
Culture Documents
by
Fiorenzo A. Fazzolari.
February 2014
I herewith declare that I have produced this work without the prohibited assistance
of third parties and without making use of aids other than those specified; notions
taken over directly ()r indirectly from other sources have been identified as such by
appropriate references. This thesis has not been presented previously in identical or
similar form to any other English or foreign examination board.
The Ph.D. thesis wasco~ducted between January 2011 to February 2014 under the
supervision of Prof. J. R. Banerjee in the School of Engineering and Mathematical
Sciences at City University London.
11
Acknowledgements
First and foremost I wish to thank deeply my advisor, Prof. J. R. Banerjee for
his technical advices and many useful comments and criticisms provided during the
course of the PhD. Since the beginning, he supported me not just providing me re-
search assistantship over three, years, but also academically and emotionally through
tlfe bumpy road which led me to successfully accomplish my PhD. I will always b'e '
thankful, for the rest of my life, for the understanding and forgiveness shown during
these years and I will always be grateful for what he did and for his unselfish, unbiased
~ I
III
Abstract
In this thesis both static and dynamic analyses of composite thin-walled structures
are carried out. Most notably, the Dynamic Stiffness Method (DSM) has been exten-
sively exploited to develop advanced formulations for plates and shells. In particular,
the. Dynamic Stiffness (DS) matrices have been developed for laminated composite
plates and shells using Higher-order Shear Deformation
~ . Theory (HSDT) in order to
investigate their free vibration behavior and buckling characteristics. First, the Gov-
erning Differential Equations (GDEs) of motion and associated natural Boundary
Conditions (BCs) (Neumann-type) for the given displacement field are derived via
Hamilton's principle for both composite plate and shell structures. In the case of
composite plates, the DS matrices are formulated for both out-of-plane and in-plane
deformations. The GDEs for each of the two cases are solved in Levy's form sep-
arately. Next the problems for both plates and shells are reduced to a system of
ordinary differential equations which are then solved by using the classical exponen-
tial solution procedure. This led to the formation of the DS matrices. It is worth
highlighting that the solution of the GDEs by means of Levy's procedure has been
applied in this thesis by using a symbolic approach. Mathematica software has been
extensively used to achieve this aim. The Wittrick-Williams algorithm is used as a
solution technique to compute the natural frequencies and the critical buckling loads.
The results from the developed DS elements have been validated, for both free vi-
bration and buckling analysis, against published results and by using Finite Element
(FE) models for a number of carefully chosen cases with the help of MSC Nastran
and ANSYS commercial software. Furthermore, extension of the method to com-
plex structures like stepped plates and stringer panels with arbitrary stringer shape
lV
reinforcements has been undertaken using the development of the transformation ma-
trix. Extensive results for a wide range of case studies are provided and commented
on in detail. The effects of significant parameters such as thickness-to-Iength ratio,
orthotropy ratio, number of layers, lay-up and stacking sequence and boundary con-
ditions on the natural frequencies arid critical buckling loads are investigated and
discussed.
v
Contents
List of Figures ix
1 Introduction 1
1.1 Literature review . . . . . 2
1.1.1 Vibration of Plates 3
1.1.2 Buckling of Plates. 8
1.1.3 Vibration of shells 10
1.2 Thesis organization . . 13
vi
2.2.4 Omega stringer panels 51
2.3 Conclusions .. 52
3 Buckling of Plates 55
3.1 Theoretical formulation . 55
3.1.1 Governing differential equations 55
3.1.2 Dynamic stiffness method . . . 57
3.1.3 Levy-type closed form exact solution and DS development 58
3.2 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . 67
3.2.1 Buckling analysis of cross-ply composite stiffened plates. 69 .
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . 77
5 Conclusions 120
Appendix A 128
Appendix B 130
Appendix C 133
Vll
Appendix D 135
Bibliography 141
viii
List of Figures
2.7 Fundamental natural frequency and mode shape of a simple and stepped
composite plate, with boundary condition S-F-S-F and stacking se-
quence [0°/90°]5' f; = 10, ~ = 2. . . . . . . . . . . 45
2.8 Sixth natural frequency and mode shape of a simple and stepped com-
posite plate, with boundary condition S-F-S-F and stacking sequence
[0° /90°]5' f; = 10, ~ = 2 . . . . . . . . . . . . . . . . . . . . . . . .. 46
2.9 Ninth natural frequency and mode shape of a simple and stepped com-
posite plate, with boundary condition S-F-S-F and stacking sequence
[0° /90°]5' f; = 10, ~ = 2 . . . . . . . . . . . . . . . . . . . . .. 46
IX
2.10 Comparison between FSDT and HSDT first three modes and circular
frequency parameters w= whb p,. . . . . . . . . . . . . . . . . . .. 48
2.11 Comparison between FSDT and HSDT first three modes and circular
frequency parameters w= whb p,. . . . . . . . . . . . . . . . . . .. 49
2.12 Comparison between FSDT and HSDT first three modes and circular
f requency paramet ers w -A - whb VfL.
E2 . . . . . . . . . . . . . . . . . .. . . 50
2.13 Comparison between FSDT and HSDT first three modes and circular
frequency parameters w= whb P, . 53
x
4.3 First six mode shapes of a symmetric cross-ply cylindrical shell with sese
boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 104
4.4 First six mode shapes of a symmetric cross-ply cylindrical shell with SFSe
boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 105
4.5 First six mode shapes of a symmetric cross-ply cylindrical shell with SFSF
boundary condition. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 106
4.6 First six mode shapes of a symmetric cross-ply spherical shell with sese
boundary condition. . 112
4.7 Fist six mode shapes. 115
Xl
List of Tables
xii
3.1 Dimensionless uniaxial buckling load (along x direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates with b/h = 10, Ed E2 =
open, G 1d E2 = G13 / E2 = 0.6, G23 / E2 = 0.5, 1/12 = 1/13 = 0.25. . . .. 69
3.2 Dimensionless uniaxial buckling load (along x direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [0° /90° /90° /0°]
and Ed E2 = open, G 12 / E2 = G 13 / E2 = 0.6, G23 / E2 = 0.5, 1/12 =
1/13 = 0.25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 70
3.3 Dimensionless uniaxial buckling load (along y direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [0° /90° /90° /0°]
and Ed E2 = open, G 12 / E2 = G 13 / E2 = 0.6, G23 / E2 = 0.5, 1/12 =
1/13 = 0.25. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 71
3.4 Dimensionless biaxial buckling load Ncr = Ncr E~2h3' for simply sup-
ported cross-ply square plates, stacking sequence [0°/90°/90 %
°] and
Ed E2 = open, G12 / E2 = G 13 / E2 = 0.6, G23 / E2 = 0.5, 1/12 = 1/13 = 0.25. 72
3.5 Dimensionless uniaxial bucklingload (along x direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [0° /90° /90° /0°]
and = open, G 12 / E2 = G 13 / E2 = 0.6, G23 / E2 = 0.5, 1/12 =
Ed E2
xiii
4.4 First five circular natural frequency parameters w= W a: It, of
square cylindrical shells with stacking sequence [0°/90° /OO]length-to-
thickness ratio a/h = 10 and varying the radius-to-length ratio Rf3/a. 107
4.5 First five circular natural frequency parameters w= W ~ It, of
square cylindrical shells with stacking sequence [0°/90° /OO]length-to-
thickness ratio a/h = 100 and varying the radius-to-Iength ratio Rf3/a. 107
4.6 First five circular natural frequency parameters w= W ~ It, of
square cylindrical shells with stacking sequence [0°/90° /90° /0°] length-
to-thickness ratio a/ h = 10 and varying the radius-to-Iength ratio Rf3/ a.lOS
4.7 First five circular natural frequency parameters w= W ~ It, of
square cylindrical shells with stacking sequence [0°/90° /90° /0°] length:.
to-thickness ratio a/h = 100 and varying the radius-to-Iength ratio R f3 /a.10S
4.S First six natural frequencies Wmn rad/ s,. of square isotropic spherical
shells. 109
4.9 First three circular natural frequency parameters w= W ~ It, of
%
square spherical shells with staking sequence [0°/90 °] and varying
the radius-to-Iength R/a and the length-to-thickness a/h ratios. . .. 110
4.10 Fundamental circular natural frequency parameter w= W ~ It, of
square spherical shells with staking sequence [0°/90°/90 %
°], R/ a =
50 and varying orthotropic Ed E2 and the length-to-thickness a/h ratios. 111
4.11 Dimensionless fundamental circular natural frequency parameter w=
W (l~;h) It, of circular cylindrical shells with staking sequence [0° /90° /0°]'
R/h = 5 and varying the boundary conditions and the length-to-radius
ratio L/ R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 114
4.12 Circular frequencies (rad/s) of one-layer orthotropic simply-supported
circular cylindrical shells with m = 1 and varying the half-wave number
n, L = 5 m, R = 1 m, h = 0.01 m, Ex = 120 G Pa, Ey = 10 G Pa,
G xy = 5.5 GPa, !ly = 0.27 and p = 1700 kg/m 3 . ............ 116
4.13 Circular frequencies (rad/s) of one-layer orthotropic simply-supported
circular cylindrical shells varying the half-wave numbers m and n. 116
XIV
4.14 Dimensionless circular frequency parameters w, of simply-supported
circular cylindrical shells with symmetric cross-ply stacking sequence,
R/ L = 1 and material Ed E2 = 40, G I 2/ E2 = 0.6 G 13 / E2 = G 23 / E2 =
0.5. lII2 = 0.25. .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. 117
4.15 Dimensionless fundamental circular natural frequency parameter w=
W C~;h) It, of circular cylindrical shells varying the stacking se-
quence, the radius-to-thickness ratio R/h, the boundary conditions
and the length-to-radius ratio L/ R. . . . . . . . . . . . . . . . . . .. 118
xv
List of abbreviations, nomenclature
and symbols
Abbreviations
ID One dimensional
2D Two dimensional
3D Three dimensional
CPT Classical plate theory
CLT Classical lamination theory
CST KRF CST for shell structures proposed by Khdeir, Reddy and Fredrick
CST GanSiv Exact solution based on CST by Ganesan and Sivadas
CST GreStk Exact solution based on CST by Greenberg and Stavsky
DS Dynamic Stiffness
DSM Dynamic Stiffness Method
ESL Equivalent single layer
Exact FSDT Exact solution based on FSDT by Liu et al.
F-DSM FSDT shell element based on DSM given by Thinh and Nguyen
FEM Finite Element Method
FSM Finite Strip Method
FSDT First order shear deformation theory
FSDT 6 FSDT for circular cylindrical shell by Viswanathan et al.
FSDT KRF FSDT for shell structures proposed by Khdeir, Reddy and Fredrick
HSDT High order shear deformation theory
XVI
HSDT KRF HSDT for shell structures proposed by Khdeir, Reddy and Fredrick
LW Layer-wise
SDPT Shear deformation plate theory
S-DQFEM Semi-analytical differential quadrature finite element method
Ye-Soldatos Three-dimensional vibration analysis by Ye and Soldatos
ZZ Zig-Zag
Nomenclature
A Area·
A,B,D,E,F,H Membrane, coupling, flexural-torsional and
high order stiffnesses
b Width of the place
C Constitutive matrix 'of the laminate
E Elastic modulus
xvii
Tp Roots out-of-plane case
s Elements of the dynamic stiffness matrix
T Kinetic Energy
t Time
U Potential Energy
U, V, W Displacements amplitudes
u, v, w Displacements in the x, y, and z direction respectively
x, y, z Coordinate system
Symbols
<P Rotations
<I> Rotation amplitudes
p Density
0",C7' Stress and stress vector
v Poisson's ratio
o Membrane
1,2 Fibre direction and orthogonal direction
11, 12, etc Matrix element positions
k Layer number
m N umber of semi sin waves in the y directions
n Number of semi sin waves in the x directions
x, y, z Directions
XVlll
Chapter 1
Introduction
Composite materials consist of two or more components which together produce de-
sirable proprieties that cannot be achieved with any of the constituents alone. Fibre-
reinforced composite materials, for example contain high strength and high modulus
fibres in a matrix material. In laminated composite materials, fibres are the principal
load-carrying members, and the matrix material keeps the fibres together, acting as
a load-transfer medium between fibres, and also protecting fibres from being exposed
to the environment. Fibre reinforced composite materials for structural applications
are often made in the form of a thin layer, called lamina. A lamina is a macro unit of
material whose material proprieties are determined through appropriate laboratory
tests.
A lamina represents a fundamental building block. A laminate is a collection of
laminae stacked to achieve a desired stiffness and thickness. For example unidirec-
tional fiber-reinforced laminae can be stacked so that the fibers in each lamina are
oriented in the same or different directions. The sequence of various orientations of
fiber-reinforced composite layer in a laminate is termed as the lamination scheme or
stacking sequence or simply lay up. The layers are usually bonded together with the
same matrix materials as that in a lamina. The lamination scheme and material prop-
erties of individual laminae provide an added opportunity to designers to tailor the
stiffness and strength of the laminate to match the structural stiffness and strength
requirements.
1
Structures composed of composite materials offer lower weight, higher stiffness and
strength than those composed of most isotropic materials. This coupled with ad-
vances in manufacturing of composite materials and structures, give them a com-
petitive edge over normal engineering materials. In particular, composite plates and
shells components constitute a large percentage of aerospace and submarine struc-
tures. They are also increasingly being used in areas like automotive engineering,
amongst other applications. Without doubt, the use of laminated composite plates
and shells in many engineering applications has been expanding rapidly over the past
decades. This has resulted in considerably more research activities to determine their
dynamic behaviour. The main aim of this thesis is to develop a new refined DS (Dy-
namic Stiffness) element based on HSDT (Higher-order Shear Deformation Theory)
and symbolic computation in order to cope with the aforementioned problems. The
proposed advanced formulations have been entirely developed by harnessing the nu-
merical and symbolic computational power of Matlab® and Mathematica® softwere.
In section 1.1 below, the history of metallic and composite plates and shells is traced
back with a focus on the contributions made in free vibration and buckling analysis.
This section gives a general overview of the historical development of plate and shell
theories for free vibration and buckling analysis. Most notably, the literature survey is
carried out starting from the early pioneering works in the first half of the nineteenth
century and finishing up with the most advanced plate and shell theories developed
recently to deal with many of the challenging theoretical and practical engineering
problems. The review cites some of the most significant contributions in the free
vibration and buckling analysis of plates and shells with a particular emphasis on the
application of the Dynamic Stiffness Method (DSM).
2
1.1.1 Vibration of Plates
Classical methodologies
The first accurate treatment of plates can be attributed to Germain [1] and Lagrange
[2] early in the 19th century. They used a pure bending concept of plates in the devel-
opment of the governing differential equations, where normals to the middle surface
remain straight and normal. The assumption is of course, valid for small deforma-
tion of thin plates. A good historical review of the subsequent developments can be
found in the books of Soedel [3] and Timoshenko [4]. This old but pioneering theory
is now referred to as the classical plate theory (CPT). Later, the inclusion of shear
deformation to the fundamental equations of-plates is due to Reissner [5] and Mindlin
[6]. Theories that account for shear deformation are now referred to as thick plate
theories or first order shear deformation plate th~ories (FSDT).
With regard to the analysis of composite structures, Hearmon [7] presented what
could be possibly the first attempt to analyse composite plates. Among the first few
to work on composite plates was also Smith [8]. A consistent theory for symmetri-
cally laminated plates was subsequently presented by Reissner and Stavski [9]. There
is evidence that some Russian scientists and engineers might have considered the
problem earlier but there is insufficient published literature in English journals. Am-
bartsumian [10] and Lekhnitskii [11] published probably the first book in the area of
composite plates and shells. Paradoxically the original contribution by Lekhniskii has
been almost ignored in the subsequent literature, even though the very first method
which he used was able to describe the zig-zag effect 1 and interlaminar continuous
transverse normal and shear stresses was provided by Lekhniskii [12]. On the contrary
the works of Ren [13, 14], had considerable coverage in the literature. In these two
papers Ren has, in fact, extended Lekhnitskii's theory to orthotropic and anisotropic
plates. A further pioneering analysis was presented by Yu [15] where the in-plane
zig-zag effect and interlaminar continuous transverse normal and shear stresses were
both fulfilled in the interfaces when analysing sandwich plates. Ashton and Whitney
lzig-zag trend of the displacement components u, v, w through-the-thickness direction due to the
transverse anisotropy of the composite structures.
3
[16] presented fundamental equations of laminated plates. On the other hand, Vinson
and Sierakowski [17] presented analysis of composite beams, plates and shells, while
Whitney [18] presented various structural analyses including free vibration of lami-
nated anisotropic plates, and in [19] he applied an extended Abartasumian theory to
generally anisotropic and symmetrical and nonsymmetrical plates. The subject was
also researched by Mohan and Kingsbury [20] as well as by Noor [21] among other
researchers in the field. In addition to these previously published articles and books,
more recent literature on the free vibration analysis of composite structures are due to
Qatu [22-26] and Du et al. [27] , Leissa and Narita [28], and other similar works can
be found in various conference proceedings and international journals. A complete
historical review of the so called zig-zag theories for laminated structures has been
provided by Carrera [29]. l
~
During the last two decades, a variable kinematics 2D-model approach with hier-
archical capabilities for composite laminated plates and shells has been extensively
developed by Carrera of which the primary contribution is provided in [30], where a
generalization by proposing a systematic use of Reissner's Mixed Variational Theorem
(RMVT) [31, 32] as a tool to furnish a class of two dimensional theories for multi-
layered plate analysis was presented. Attention was focused on approximate solution
techniques, and the resulting governing equations were written as a system of algebraic
equations. A weak form of Hooke's law was also introduced in this work to reduce the
mixed displacement and stress formulation into a displacements one. Further details
about hierarchical theories can be found in [33], which also gives an overview of finite
element methods that have been developed for multilayered, anisotropic composite
plates and shells. An assessment and benchmarking were performed in [34] to vali-
date the 2D hierarchical models. Applications of what is reported in [30, 35] to derive
GDEs in strong forms have been given in several other papers [36-42] in which both
Navier-type closed form and finite element solutions were given. Finite element for-
mulation was also extensively developed in [43, 44], where different loadings as well as
boundary conditions were treated. The variable kinematics modeling technique based
on Carrera's Unified Formulation (CUF) offers a systematic procedure to obtain re-
4
fined structural models by considering the order of the theory as a free parameter of
the formulation. The CUF has been applied in conjunction with analytical, FEM and
other approximated solution techniques in order to predict in an accurate manner the
vibration response of laminated composite plates. In particular, advanced Galerkin
and Ritz formulations were derived by Fazzolari and Carrea [45-52]. Radial basis so-
lutions were provided by Ferreira [53-58] and the applications of refined CUF-based
FEs were given by Cinefra et. at [59, 60].
One powerful method of analysis, other than the ones mentioned above, is that of the
Dynamic Stiffness Method (DSM) ([61]). Application of this method involves develop-
ing the dynamic stiffness (DS),matrix for each individual element in the structure and
then assembling them into a global DS matrix for subsequent free vibration analysis.
This method is, in many ways, analogous to the conventional finite element method
([62]). The main difference between the two methods is that the FEM discretizes a
structural element based on assumed shape functions to derive the mass and stiffness
matrices separately, whereas the DSM uses a single element matrix containing both
mass and stiffness proprieties, which are derived from the exact frequency-dependent
shape functions obtained from the solution of the governing differential equations of
the element in free vibration. The assembly procedure for the two methods is essen-
tially the same, but the solution techniques are different in that the FEM generally
leads to a linear eingenvalue problem in sharp contrast to the non-linear (transcen-
dental) eigen-solution encountered in the DSM, which is generally solved by applying
the well-established algorithm of Wittrick and Williams [63]. For structures consist-
ing of beam elements there is no restriction on the application of the DSM and there
is a well known software based on the method to analyze plane or space frames [64].
Another important difference between the FEM and the DSM is that, the number of
natural frequencies that can be computed using the FEM is restricted to the number
of chosen degrees of freedom of the structure and the accuracy of results diminishes
with higher order modes. This can be a serious limitation in modal analysis. By con-
5
trast, the DSM has no such limitation and any number of natural frequencies can be
computed to any desired accuracy using the DS matrix without the need to increase
the number of elements to achieve higher accuracy. Moreover, when fast iterative
matrix solvers are used, the DSM will be much more efficient than the FEM. With
regard to plate elements the DSM gives exact results because the equations of motion
are solved in Levy-type closed form to obtain the element properties and no other
approximation is made en route during the analysis. Witt rick and Williams [65] are
known to be the first who attempted the extension of DSM to plate elements. Their
pioneering formulation is interesting and relies on extensive use of complex algebra.
In 1972, Williams [66] presented two computer programs, GASVIP and VIPAL to
compute the natural frequencies, based on DSM. Essentially GASVIP was used to
set up the overall stiffness matrix for the structure, and VIPAL demonstrated the
use of substructuring. A couple of years later, Wittrick and Williams reported the
computer code VIPASA [65] for free vibration analysis of prismatic plate assemblies,
which was a significant development at the time. VIPASA code allowed free vibration
analysis of isotropic or anisotropic plate assemblies and had many additional features.
The complex stiffnesses described in [67] were incorporated, as well as allowances for
eccentric connections between the component plates were accounted for, but more im-
portantly, the code used a powerful algorithm, developed by Wittrick and Williams
[63] as solution technique to compute natural frequencies of plated structures. The
algorithm is robust and it ensures that no natural frequencies of the structure are
missed. (A brief discussion of the Wittrick-Williams algorithm is presented in section
(2.1.5)). In 1983, Williams and Anderson [68] showed modifications to the eigenvalue
algorithm described in [63]. They made use of Lagrangian multipliers to apply point
constraints at any location of plate edges. Each sinusoidal mode of the freely vibrating
plate in the longitudinal direction was included within the dynamic stiffness matrix.
These modifications formed the basis for the enhanced computer code VICON (VIpasa
with CONstraints) which was a significant improvement, over the previous code (see
[69]). However, the analysis capability of VICON was based on classical plate theory
(CPT), and particularly for composite plates, attention was focused on symmetric
6
laminates. A later version of the code included plates on Winkler foundations [70].
Next, a major enhancement of the program took place in the early nineties when
the optimum design features were added and the new program VICONOPT (VICON
with OPTimization) [71, 72] came into existence. A few years later, Anderson and
Kennedy [73] incorporated the effect of the shear d~formation into VICONOPT using
a numerical approach. The general purpose application of VICONOPT was further
enhanced by them [64, 73] to allow for analysis of angle-ply laminates. An interesting
historical review of the DSM procedure for plates can be found in [74]. It should be
noted that DSM has been extensively researched by Banerjee [61, 75-79], amongst
others for modal analysis of structures idealized by beam elements based on Euler-
Bernoulli, Timoshenko and associated coupled beam theories. The extension of the
DSM to plate elements is no doubt difficult, but indeed, essential to model complex
structures. Following the earlier research on DS theories of isotropic and composite
plates; Boscolo and Banerjee [80-83] advanced the state of the art on these topics by
including the effects of shear deformation and rotatory inertia and thereby providing
a detailed modal analysis procedure through the application of symbolic computation
and Matlab. They used the first order shear deformation theory (FSDT) for which
the introduction of a user specified shear correction factor was necessary. The current
thesis is partly motivated by these earlier developments and are of the most impor-
tant contributions made here is the inclusion of higher order shear deformation theory
(HSDT), for the first time, when developing the DS matrix for laminated composite
plates. This useful extension is of considerable theoretical and computational com-
plexity as will be shown later. The research is particularly relevant when analysing
thick composite plates for their free vibration characteristics. It should be recognised
that Reddy and co-authors [84-86] used HSDT in a different context in free vibration
analysis of composite plates without resorting to the development of the DSM. From
a historical prospective HSDT, can be traced back to third order plate bending theory
originally proposed by Vlasov [87] in the late 1950s. His theory was substantiated
and extended to laminated composite plates many years later by Reddy [84] using a
variational approach. This is sometimes referred to as Vlasov-Reddy theory (VRT).
7
Further improvements of this theory can be found in the work of Jemielita [88, 89J.
Inclusion of HSDT in the DSM framework will enable free vibration analysis of plates
with moderate to high thickness to width ratio, in an accurate and computationally
efficient manner. One of the great advantages of using HSDT as opposed to FSDT
is that the former accounts for the effects of the shear deformation in a judicious
manner without using a fictitious (and often controversial) shear correction or shape
factor that is prevalent in the latter. The usefulness of HSDT becomes apparent when
analyzing composite structures, particularly of thicker dimensions, because fiber re-
inforced ·composites have generally very low shear modulii. Both the in-plane and
out-of-plane free vibration analyses of plates are considered in this thesis.
Several methodologies have been developed over the years to solve the elastic stabil-
ity problem of beams, plates and shells. A simplified approach to calculate the i-th
critical buckling load, is to consider it as the load at which more than one infinites-
imally adjacent equilibrium configurations exist that can be identified with the i-th
bifurcation point (Euler'S method) [90J. In a linearized structural stability analysis,
the determination of the critical load leads to a linear eigenvalue problem. The bi-
furcation method can be successfully used particularly for plates, when the critical
equilibrium configuration shows a geometry change as the critical buckling load is
reached. However, as explained by Leissa [91], linearized stability analysis is mean-
ingful, if and only if, the initial in-plane loading does not produce an out-of-plane
deformation. Furthermore, there are many cases in which Euler's method may fail,
particularly when thin-walled structures like shells exhibit the snap-buckling phe-
nomenon. In such cases, the most general approach, based on the solution of the
complete equilibrium and stability equations [92, 93J is preferred.
Amongst a wide class of methodologies employed by researchers to analyze the elastic
stability of advanced composite structures, the DSM is probably the most accurate
and computationally efficient option. The DSM based on Levy-type closed form so-
lution for plates [52J is indeed an exact approach to the solution. Wittrick [94J laid
8
the groundwork of the DSM for plates. The basic assumption in this work is that
the deformation of any component plate varies sinusoidally in the longitudinal di-
rection. Using this assumption, a stiffness matrix may be derived that relates the
amplitudes of the edge forces and moments to the corresponding edge displacements
and rotations for a single component plate. For the exact DSM, this stiffness matrix
is derived directly from the equations of equilibrium that describe the buckling be-
havior of the plate. Essentially, Wittrick [94] developed an exact stiffness matrix for
a single isotropic, long fiat plate when subjected to uniform axial compression. His
analysis basically used classical plate theory (CPT). Wittrick and Curzon [95] later
extended this analysis to account for the spatial phase difference between the pertur-
bation forces and displacements which occur at the edges of the plate during buckling
due to the presence of in-plane shear loading. This phase difference was accounted for
by defining the magnitude of these quantities using complex quantities. Wittrick [96]
then extended his analysis further to consider fiat isotropic plates under any general
state of stress that remains uniform in the longitudinal direction (Le., combinations
of bi-axial direct stress and in-plane shear). A method very similar to that described
by Wittrick in [94] was also presented by Smith [97] for the bending, buckling, and
vibration of plate-beam structures. Following these developments, Williams [66] pre-
sented two computer programs, GASVIP and VIPAL to compute the initial buckling
stress of prismatic plate assemblies subjected to uniform longitudinal stress or uni-
form longitudinal compression, respectively. GASVIP was used to set up the overall
stiffness matrix for the structure, and VIPAL demonstrated the use of substructuring.
Following this work, Wittrick and Williams [65] reported on the VIPASA computer
code for the buckling analysis of prismatic plate assemblies. This code allowed for
analysis of isotropic or anisotropic plates using a general state of stress (including
in-plane shear). The complex stiffnesses described by Witt rick and Williams in [67]
were incorporated in VIPASA, as well as allowances were made for eccentric connec-
tions between component plates. This code also implemented an algorithm, referred
to as the Wittrick-Williams algorithm [98] for determining any critical buckling load
for any given wavelength. The development of this algorithm was necessary because
9
the complex stiffnesses described above are transcendental functions of the load factor
and half wavelength of the buckling modes of the structure which make a determinant
plot cumbersome and unfeasible. Viswanathan and Tamekuni [99, 100J presented an
exact DSM based upon CPT for the elastic stability analysis of composite stiffened
structures subjected to biaxial in-plane loads. The structure was idealized as an as-
semblage of laminated plate elements (flat or curved) and beam elements. Tamekuni,
and Baker extended this analysis in [101J considering long curved plates subject to
any general state of stress, together with in-plane shear loads. Anisotropic material
properties were also allowed. This analysis utilized complex stiffnesses as described in
[67J. The works described in [97, 98, 101J are more or less similar. The differences are
discussed in [65J.Williams and Anderson [68J presented modifications to the eigen-
value algorithm described in [98J. Further modifications presented in [68J allowed the
recovery of buckling modes corresponding to a general loading to be represented as a
series of sinusoidal modes in combination with Lagrangian multipliers to apply point
constraints at any location on the edges. The DS matrix for laminated composite
plates for buckling analysis by using the HSDT is a significant extension in this the-
sis. The development is of considerable theoretical and computational complexity as
will be shown later. This contribution is particularly relevant when analysing thick
composite plates for their buckling characteristics.
Tracing back the history of the shell theories makes interesting reading. According to
Novozhilov [102J the first known effort properly channeled to shell theory was given by
Aron [103J, who attempted, for the first time, to extend Kirchhoff's hypothesis, which
is valid for flat plates, to extend it to shell structures. Despite this pioneering work
Aron's development was not strictly correct· and some inaccuracies were later recti-
fied by Love [104, 105J. Nevertheless the development of the shell theory proposed by
Love [104, 105J was also scrutinized due to some mathematical inconsistencies aris-
ing from the fact that some small terms were retained whereas some others terms of
the same order of magnitude were discarded. The first set of GDEs completely free
10
from inconsistencies were provided by Lur'e [106]. By contrast Gol'denveizer [107],
for the first time, provided the compatibility conditions of strains for shells starting
from the Gauss-Codazzi conditions. In his work, Gol'denveizer introduced pioneer-
ing insights into the possibility of identically satisfing the GDEs, written in terms of
forces and moments, by making use of stress functions. At about the same time and
in a completely independent manner from Gol'denveizer [107], Lur'e [106] provided
a similar solution. More specifically both authors, considering a shell loaded only
at the edge, proved that the GDEs made up of ten unknowns including forces and
moments could have been expressed in terms of only four arbitrary stress functions.
This solution was somehow an extension to shell structures of the solution provided
by Airy [108] in the plane theory ofelasticity. Afterwards, Novozhilov [102], by setting
the Poisson's ratio to zero, provided several complex forms of the GDEs written in
a compact and concise manner, showing that the advantages of this new formulation
lie in the simplification of their solution. A deeper understanding of the usefulness
of the application of the complex transformations was given by Mushtari [109, 110].
Subsequently, in order to provide solutions of the GDEs of practical interest, consid-
erable efforts were expended for their simplification. As a result, the GDEs of shallow
shells were satisfactorily derived. In this regard, it worth highlighting the articles by
Donnell [111, 112], Mushtari [109, 110] and Vlasov [113, 114]. Most notably, the turn-
ing point of the development came from Donnell [111], who wrote: "Much confusion
seems to exist as to what simplification can be made, and the condition under which
they can be made. One author considers items which another rejects and vice-versa.
An attempt is made, in the following discussion, to clarify this question and to ob-
tain the greatest simplification possible, under the condition of the present problem;
the results are applicable to a large class of problems". His efforts led to the shell
theory which is universally adopted to this day for the investigation of shallow shell
structures. Afterwards, independently of each other, Fliigge [115, 116], Lur'e [106]
and Byrne [117] developed shell theories discarding the hypothesis of thinness and
considering higher order expansion of the reciprocal of the Lame parameters. Other
refinements of the shell theories were proposed by Sanders [118] amongst others. Ad-
11
ditional effects when developing shell theories were taken into account by Whitney
and Sun [119], Librescu [120], Gulati and Essemberg [121], Zucas and Vinson [122]
and Ambartsumian [123-130]. Additional references on the subject can be found in
Naghdi [131]' Ambartsumian [132] and Bert [133-135]. Reddy [84] proposed a gen-
eralization of Sander's theory to anisotropic doubly-curved shells. The application of
layer-wise theories for shell structures can be found in the papers presented by Hsu
and Wang [136], Cheung and Wu [137], Barbero et al. [138] and Carrera [139-141].
Reviews on FE shell formulations have been given by Denis and Palazzotto [142] and
Di and Ramm [143]. Exhaustive reviews on classical theories can be found in the
work of Librescu [120]. With regards to the use of approximation methods, Qatu and
Asadi [144] addressed the free vibration analysis of doubly-curved shallow shells with
arbitrary boundary conditions by using the Ritz method with algebraic polynomial
displacement functions. Asadi et al. [145] employed a 3D and several shear defor-
mation theories in order to carry out static and free vibration analysis of thick deep
laminated cylindrical shells. Ferreira et al. [54] used a wavelet collocation method for
the analysis of laminated shells. The same authors [56] combined a sinusoidal shear
deformation theory with the radial basis functions collocation method to deal with
the static and free vibration analyses of laminated composite shells. Tornabene et
al. [146, 147] studied the free vibration behavior of doubly-curved anisotropic lam-
inated composite shells and shell of revolution panels by means of the Generalized
Differential Quadrature (GDQ). Within the framework of static and dynamic anal-
ysis of composite shell structures the application of asymptotic methods must not
be underrated. In particular the works of Fettahlioglu and Steel [148], Widera and
Logan [149], Wider a and Fan [150], Spencer et al. [151] and Cicala [152] deserve to be
highlighted. A complete overview of different problems related to multilayered shells
modelling has been provided by Kapania [153] and Noor and Burton [154]. Recently
a dynamic stiffness procedure for the free vibration analysis of circular cylindrical
shells has been proposed by El-Kaabazi and Kennedy [155]. Their analysis is inter-
esting and focused on isotropic shells. The shell theories employed were based on
Donnel [111, 112], Timoshenko [156] and Fliigge [115, 116] assumptions. Thinh and
12
Nguyen [157] dealt with the free vibration of laminated composite circular cylindrical
shells by using the dynamic stiffness method and first order shear deformation theory
(FSDT). From the voluminous literature available on shell theories, it is clear that
only a few attempted the solution of the governing differential equations in an exact
sense. In this respect, Forsberg [158] endeavoured to solve the problem by using the
Donnell and Fliigge [115, 116] assumptions for orthotropic circular cylindrical shells
with simply supported (shear diaphragms) boundary condition. Soedel [159] is an-
other researcher of this type who attempted the exact solution amongst a few others
[160, 161]. On the other hand, a systematic procedure for obtaining the closed-form
eigensolution was given by Callahan and Baruh [162] by using Matlab®. Although
most of the above contributions rely on substantial simplifications, the efforts ex-
pended are nevertheless important and timely due to the high variability with which
the approximate solutions generally match the exact ones.
1. Chapter one
Chapter one gives a brief introduction to composite materials and presents a
literature review accounting for the main contributions made in the past few
decades on the free vibration and buckling analysis of composite plates and
shells with a focus on the Dynamic Stiffness Method (DSM).
2. Chapter two
Chapter two describes the application of the DSM in combination with a Higher-
order Shear Deformation Theory (HSDT) in order to analyze the free vibra-
tion behavior of laminated composite simple and stepped plates, and composite
stringer panels with general shape.
13
3. Chapter three
Chapter three provides the development of an exact method of solution for the
buckling analysis of composite plates. The solution technique is always based
on the application of the DSM incorporating HSDT.
4. Chapter four
Chapter four describes the development of an advanced DS formulation for the
investigation of doubly-curved laminated composite shells.
5. Chapter five
Chapter five provides the principal conclusions of the investigation carried out
in this thesis.
6. Chapter six
Chapter six shows the scope for future ambitious work that could be carried
out as a direct extension of the work proposed in this thesis.
14
Chapter 2
An exact dynamic stiffness method based on higher order shear deformation theory
is developed for the first time using symbolic computation in order to carry out
the free vibration analysis of composite plate assemblies. Hamilton's principle is
applied to derive the governing differential equations of motion and natural boundary
conditions. Then by imposing the geometric boundary conditions in algebraic form
the dynamic stiffness matrix is developed. The Wittrick-Williams algorithm is used
as a solution technique to compute the natural frequencies and mode shapes for a
range of laminated composite plates, stepped composite panels and composite plate
with stringers.
The effects of significant parameters such as the length-to-thickness ratio, orthotropy
ratio, step ratio, number of layers, lay-up, stringer sections and stacking sequence
and boundary conditions on the natural frequencies and mode shapes are critically
examined and discussed. The accuracy of the method is demonstrated by comparing
results with those available in the literature and by using commercially available
software.
15
2.1 Theoretical formulation
In the derivation that follows , the hypotheses of straightness and normality of a trans-
verse normal after deformation are assumed to be no longer valid for the displacement
field which is now considered to be a cubic function in the thickness coordinate; and
hence the use of higher order shear deformation theory (HSDT). This is in sharp
contrast to earlier formulations based on CPT and FSDT. For a composite plate, the
kinematics of deformation of a transverse normal using both first order and higher
order shear deformation are schematically shown in Fig. 2.1. The laminate is assumed
X,Y
Figure 2.1: Kinematic descriptions of displacements using FSDT and HSDT for a
multilayered plate.
W (x, y , z, t) = Wo (x, y , t)
16
where u, v, ware general displacements within the plate in the x, y, and z directions,
respectively, whereas uo, va, Wo are the corresponding displacements of the reference
surface (mid-plane D). Hamilton's principle is now applied. The variational statement
is:
£; It2 5
Nl
tl £k dt =0 (2.2)
where £k is the Lagrangian for the kth layer of the composite plate. The first variation
can be expressed as:
(2.3)
where 5U k is the virtual strain energy, 5Tk is the virtual kinetic energy, and they
assume the following form:
(2.4)
the stresses (u), the strains (e) and the displacements ("1) vectors are expressed as
follows:
uk ={ ak
xx
k
a yy
k
a xy
k
a xz a;z }T
k k
e = { c xx k
Cyy ~xy
k k
~xz ~;z }T (2.5)
"1={ u v W }T
In Eq. (2.4) pk denotes mass density while an over dot denotes differentiation with
respect to time. The subscript T signifies an array transposition and 5 the variational
operator. Constitutive and geometrical relationships (deformation) are respectively
defined as:
(2.6)
-k
where C is the plane stress constitutive matrix and V is the differential matrix (see
Appendix A for details). Substituting Eq. (2.6) into Eq. (2.4) and imposing the
17
condition in Eq. (2.2), the equations of motion are obtained after extensive algebraic
manipulation as:
c5'uo: All uO,xx + A12 VO,yx + A 16 (UO,yx + vo,xx) + Bll cPx,xx + B12 cPy,yx + B 16 (cPx,yx + cPy,xx)
+ Ell C2 cPx,xx + Ell C2 wO,xxx + E12 C2 cPy,yx + E12 C2 WO,yyx + E 16 C2 cPx,yx + E 16 C2 cPy,xx
+ 2 E 16 C2 WO,xyx + A16 UO,xy + A 26 VO,yy + A66 (UO,yy + VO,xy) + B16 cPx,xy + B26 cPy,yy
+ B66 (cPx,yy + cPy,xy) + E12 C2 (cPx,xy + WO,xxy) + E66 C2 (cPx,yy + cPy,xy + 2 WO,xyy)
+ E26 C2 (cPy,yy + WO,yyy) = 10uo +h ¢x +h C2 ¢x + 13 C2 WO,x
c5vO: A 16 uO,xx + A 26 VO,yx + A66 (UO,yx + vo,xx) + B16 cPx,xx + B 26 cPy,yx + B66 (cPx,yx + cPy,xx)
+ E16 C2 cPx,xx + E16 C2 wO,xxx + E26 C2 cPy,yx + E 26 C2 WO,yyx + E66 C2 cPx,yx + E66 C2 cPy,xx
+ 2 E66 C2 WO,xyx + A12 UO,xy + A22 VO,yy + A 26 (UO,yy + VO,xy) + B12 cPx,xy + B22 cPy,yy
+ B 26 (cPx,yy + cPy,xy) + E12 C2 (cPx,xy + WO,xxy) + E22 C2 (cPy,yy + WO,yyy) + E26 C2 (cPx,yy
+ cPy,xy + 2wO,xyy) = 10 Vo +h ¢y + 13 C2 ¢y + h C2 WO,y
c5wO: A44 (cPy,y + WO,yy) + A45 (cPx,y + WO,xy) + D44 C1 (cPy,y + WO,yy) + D45 C1 (cPx,y + WO,xy)
+ A45 (cPy,x + WO,xy) + A55 (cPx,x + Wo,xx) + D45 C1 (cPy,x + WO,xy) + D55 C1 (cPx,x + Wo,xx)
+ D44 C1 (cPy,y + WO,yy) + D45 C1 (cPx,y + WO,xy) + F44 C~ (cPy,y + WO,yy) + F45 C~(cPx,y
+ WO,xy) + D45 C1 (cPy,x + WO,xy) + D55 C1 (cPx,x + Wo,xx) + F45 C~ (cPy,x + WO,xy)
+ F55 C~ (cPx,x + Wo,xx) - Ell C2 'ito,xxx - E12 C2 VO,xxy - E16 C2 (uo,xxy + vo,xxx)
- Fll C2 cPx,xxx - F12 C2 cPy,xxy - F16 C2 (cPx,xxy + cPy,xxx) - Hll c~ (cPx,xxx + wo,xxxx)
- H12 c~(cPx,xxy + WO,xxyy) - H16 c~ (cPx,xxy + cPy,xxx + 2wO,xxxy) - 2 E16 C2 UO,xxy
18
8cPx: Bu uO,xx + B12 VO,yx + B16 (UO,yx + VO,xx) + Du cPx,xx + D12 cPy,xy +D16 (cPx,yx + cPy,xx)
+ Fu C2 (cPx,xx + WO,xxx) + F12 C2 (cPy,yx + WO,yyx) + F16 C2 (cPx,yx + cPy,xx + 2 WO,xyx)
+ B16 UO,xy + B 26 VO,yy + B66 (UO,yy + VO,xy) + D16 cPx,xy + D26 cPy,yy + D66 (cPx,yy + cPy,xy)
+ F16 C2 (cPx,xy + WO,xxy) + F26 C2 (cPy,yy + WO,yyy) + F66 C2 (cPx,yy + cPy,xy + 2 WO,xyy)
+ Eu C2 UO,xx + E12 C2 VO,yx + E 16 C2 (UO,yx + VO,xx) + Fu C2 cPx,xx + F12 C2 cPy,xy
+ F 16 C2 (cPx,yx + cPy,xx) + Hu c~ (cPx,xx + wo,xxx) + H12 C~ (cPy,yx + WO,yyx)
+ H 16 C~ (cPx,yx + cPy,xx + 2 WO,xyx) + E 16 C2 UO,xy + E26 C2 VO,yy + E66 C2 (UO,yy + VO,xy)
+ F 16 C2 cPx,xy + F26 C2 cPy,yy + F66 C2 (cPx,yy + cPy,xy) + H16 C~ (cPx,xy + WO,xxy)
+ H 26 C~ (cPy,yy + WO,yyy) + H66 C~ (cPx,yy + cPy,xy + 2 WO,xyy) - A45 (cPy + 2 WO,y)
- A55 (cPx + 2wo,x) - 2 D45 C1 (cPy + 2wO,y) - 2 D55 C1 (cPx + 2wo,x) - F45 c~ (cPy + 2wO,y)
- - F55 c~ (cPx + 2wo,x) = (h + C2 13) UO + (h + 2C2 14 + C~ 16) ¢x + (14 + C~ 16) WO,x
8cPy: B16 uO,xx + B26 VO,yx + B66 (UO,yx + VO,xx) + D16 cPx,xx + D26 cPy,xy + D66 (cPx,yx + cPy,xx)
+ F16 C2 (cPx,xx + WO,xxx) + F26 C2 (cPy,yx + WO,yyx) + F66 C2 (cPx,yx + cPy,xx + 2wO,xyx)
+ B12 UO,xy + B22 VO,yy + B 26 (UO,yy + VO,xy) + D12 cPx,xy + D22 cPy,yy + D26 (cPx,yy + cPy,xy)
+ F12 C2 (cPx,xy + WO,xxy) + F22 C2 (cPy,yy + WO,yyy) + F26 C2 (cPx,yy + cPy,xy + 2 WO,xyy)
+ E16 C2 uO,xx + E26 C2 VO,yx + E66 C2 (UO,yx + VO,xx) + F16 C2 ¢x,xx + F26 C2 cPy,xy
+ F66 C2 (cPx,yx + cPy,xx) + H16 C~ (cPx,xx + WO,xxx) + H26 C~ (cPy,yx + WO,yyx)
+ H66 C~ (cPx,yx + cPy,xx + 2 WO,xyx) + E12 C2 UO,xy + E22 C2 VO,yy + E26 C2 (UO,yy + VO,xy)
+ F12 C2 cPx,xy + F22 C2 cPy,yy + F26 C2 (cPx,yy + cPy,xy) + H12 c~ (cPx,xy + WO,xxy)
+ H22 c~ (cPy,yy + WO,yyy) + H 26 c~ (cPx,yy + cPy,xy + 2 WO,xyy) - A44 (¢y + 2 WO,y)
- A45 (cPx + 2wo,x) - 2D44 C1 (cPy + 2wO,y) - 2D45 C1 (¢x + 2wo,x) - F44C~ (¢y + 2wO,y)
- F45cI (cPx + 2wo,x) = (h + C2 h) Vo + (h + 2C2 14 + c~ h) ¢y + (14 + c~ h) WO,y
(2.7)
19
The natural boundary conditions are:
t5uo: N xx = Au uO,x + Bn cPx,x + Eu C2 cPx,x + Ell C2 wO,xx + Al2 VO,y + Bl2 cPy,y
+ El2 C2 cPy,y + El2 C2 WO,yy + A16 UO,y + A16 vO,x + B16 cPx,y + Bl6 cPy,x
+ Ell:) C2 cPx,y + El6 C2 cPy,x + 2 E16 C2 WO,xy
t5vo: N xy = Al6 uO,x + Bl6 cPx,x + El6 C2 cPx,x + El6 C2 wO,xx + A 26 VO,y + B26 cPy,y
+ E26 C2 cPy,y + E26 C2 WO,yy + A66 UO,y + A66 vO,x + B66 cPx,y + E66 C2 cPy,x
+ E66 C2 cPx,y + E66 C2 cPy,x + 2 E66 C2 WO,xy
- F45 ci WO,y - A55 cPx - A55 wO,x - D55 CI cPx - 2 CI wO,x - F55 ci cPx
- F55 ci wO,x
t5cPx: Mxx = Du cPx,x + Hn c~ cPx,x + Hn c~ wO,xx + Bll uO,x + Ell C2 uO,x + 2 Fu C2 cPx,x
+ Fu C2 wO,xx + Fu C2 wO,xx + Bl2 VO,y + Dl2 cPy,y + Fl2 C2 cPy,y + Fl2 C2 WO,yy
+ El2 C2 VO,y + Fl2 C2 cPy,y + Hl2 c~ cPy,y + Hl2 c~ WO,yy + B16 UO,y + Bl6 vO,x
+ D16 cPx,y + Dl6 cPy,x + F16 C2 cPx,y + Fl6 C2 cPy,x + 2 Fl6 C2 WO,xy + E16 C2 UO,y
+ El6 C2 vO,x + Fl6 C2 cPx,y + Fl6 C2 cPy,x + Hl6 c~ cPx,y + H16 c~ cPy,x + 2 Hl6 c~ WO,xy
20 (2.8)
8cPy: Mxy = D16 cPx,x + H16 c~ cPx,x + H16 c~ wO,xx + B16 uO,x + E16 C2 uO,x + 2 F16 C2 cPx,x
+ F16 C2 wO,xx + F16 C2 wO,xx + B26 VO,y + D12 cPy,y + F26 C2 cPy,y + F26 C2 WO,yy
+ E26 C2 VO,y + F26 C2 cPy,y + H 26 c~ cPy,y + H26 c~ WO,yy + B66 UO,y + B66 vO,x
+ D66 cPx,y + D66 cPy,x + F66 C2 cPx,y + F66 C2 cPy,x + 2 F66 C2 WO,xy + E66 C2 UO,y
+ E66 C2 vO,x + F66 C2 cPx,y + F66 C2 cPy,x + H66 c~ cPx,y + H66 c~ cPy,x + 2 H66 c~ WO,xy
8wo,x: P xx = Hll c~ cPx,x + Hll c~ wO,xx + Ell C2 uO,x + Fll C2 cPx,x + El2 C2 VO,y + F12 C2 cPy,y
+ H12 c~ cPy,y + H12 c~ WO,yy + E16 C2 UO,y + E16 C2 vo,x + F16 C2 cPx,y + F16 C2 cPy,x
+ H16 c~ cPx,y + H16 c~ cPy,x + 2 H16 c~ WO,xy
where the suffix after the comma denotes the partial derivative with respect to that
variable, and
are laminate stiffnesses and inertia terms, respectively with i and j varying from 1 to
6.
Once the equations of motion and the natural boundary conditions, i.e., Eqs. (2.7)
and (2.8) are obtained, the classical method to carry out exact free vibration analysis
of a plate consists of (i) solving the system of differential equations in Navier or Levy
type closed form in an exact manner, (ii) applying particular boundary conditions on
the edges and finally (iii) obtaining the frequency equation by eliminating the inte-
gration constants [165-168]. This method, although extremely useful for analysing an
individual plate, lacks generality and cannot be easily applied to complex structures
for which researchers usually resort to approximate methods such as the FEM. In this
21
respect, the dynamic stiffness method (DSM), which is analogous to FEM, is more
powerful as it always retains the exactness of the solution even when it is applied to
complex structures. The dynamic stiffness matrix of a structural element used in the
DSM has many other advantages. It can be offset and/or rotated and assembled in
a global DS matrix in the same way as the FEM. This global DS matrix contains
implicitly all the exact natural frequencies of the structure which can be computed
by using the well established algorithm of Wittrick and Williams [63].
A general procedure to develop the dynamic stiffness matrix of a structural element
can be summarized as follows:
(i) Seek a closed form analytical solution of the governing differential equations of
m'otion of the structural element undergoing free vibration.
)
(ii) Apply a number of general boundary conditions in algebraic forms that are
equal to twice the number of integration constants; these are usually nodal
displacements and forces.
(iii) Eliminate the constants by relating the amplitudes of the harmonically varying
nodal forces to those of the corresponding displacements which essentially gen-
erates the frequency-dependent dynamic stiffness matrix, providing the force-
displacement relationship between nodes.
Referring to the equations of motions given by Eqs. (2.7), an exact solution can
be sought in Levy's form for symmetric, cross ply laminates. For such laminates
B = E = 0, and Gf6 = G~6 = Gj5 = 0 and the out-of-plane displacements are uncou-
pled from the in-plane ones.
22
2.1.3 Levy-type closed form exact solution and DS develop-
ment
00 00
wO(x, y, t) = L Wm(x) eiwt sin(a y), ¢x(x, y, t) = L <I>X1n (x) eiwt sin(a y), (2.10)
m=l m=l
00
where w is the arbitrary circular frequency, ex = ~7r and m = 1,2, .... ,00. This is
the so-called Levy's solution which assumes that the two opposite sides of the plate .
are simply supported (S-S), i.e. w = cPx = 0 at y = 0 and y = L. Substituting Eq.
(2.10) into Eqs. (2.7) a set of five ordinary differential equations that are uncoupled
between in-plane and out-of-plane deformations, is obtained which can be written in
two different matrix forms one for the in-plane and the other for the out-of-plane, as
follows:
where .cPij (i,j = 1,2) and .c Oij (i,j = 1,2,3) are differential operators. For the
in-plane free vibration case, .c Pij
(i,j = 1,2) are given by:
23
For the out-of-plane case, LOij (i,j = 1,2,3) are given by:
L011 = (_0
2
(A44 + 2 C2 D44 + c~ F44 + 0 2 ci H 22 ) + (Ia + 0 2 ci h) w2) + (A55
2
L021 = (-A55 - C2 (2 D55 + C2 F55 ) - 0 Cl (F12 + 2 F66 + Cl H12 +.2 Cl H66 )
+ Cl (I4 + Cl h) w2) Dx + Cl (Fu + Cl Hu) D;
24
2
L033 = (-A44 - C2 (2 D44 + C2 F 44 ) - 0. (D22 + Cl (2 F22 + Cl H 22 )) + (/2 + Cl ( 2/4
(2.13)
already been defined in Eq. (2.9). Expanding the determinant of the matrices in Eq.
(2.11) the following differential equations for the in-plane and out-of-plane cases are
respectively obtained as follows:
where
(2.15)
Using a trial solution eA in Eq. (2.14) yields the following auxiliary equations for the
two cases:
(2.16)
Substituting jjp = >.; and jjo = >.~) the fourth and eighth order polynomials of Eqs.
(2.16) become
(2.17)
The two roots for the in-plane case i.e. the quadratic equation of the left) are given
25
by:
(2.18)
whereas the four roots for the out of plane case, i.e. the quartic equation on the right
are given by:
(2.19)
where
(2.20)
The explicit form of the coefficients aj with j = 1,2,3,4,5 and bj with j = 1,2 can
be found in Appendix B. Note that when computing J-tpj (j = 1,2) and J-toj (j =
1,2,3,4), some roots may turn out to be complex, but the amplitudes of the dis-
placements Um (x), Vm (x), Wm (x) , 4> xm (x), <Pym (x) will always be real, whilst the
associated constant coefficients can be complex. As complex roots occur in conjugate
pairs, the associated coefficients will also occur in conjugate pairs. Setting P'oj = ...;rr:;;;
with j = 1,2,3,4 and J-tpj = JJipj with j = 1,2, the solution for out-of-plane and
26
in-plane free vibration can thus be written as:
+ A5 e+ JLo3X
+ A6 e- JLo3X
+ A7 e+ JLo4X
+ As e- JLo4X
(2.21)
<fly"" (x) = C1 e+ JLo1X + C2 e- JLo1X
+ C3 e+ JLo2X
+ C4 e- JLo2X
+ C5 e+ JLo3X
+ C6 e- JLo3X + C7 e+ JLo4X
+ Cs e- JLo4X
Um (x) = D1 e+ JLP1X
+ D2 e- JLplX + D3 e+ JLp2X + D4 e- JLp2X
where Al -As, B1 -Bs, C1-Cs', D1 -D4' E1 -E4 are integration constants. For both
in-plane and out-of-plane cases, the constants are not all independent. Thus a set
of four independent constants, for the in-plane case, and a set of eight independent
constants, for the out-of-plane case, can be chosen and then related to the others.
Constants E1 - E4 for the in-plane case, and B1 - Bs for the out-of-plane case are
respectively chosen here to be the base and independent. By substituting Eqs. (2.21)
into (2.11) the following relationships can be obtained for the in-plane case:
D1 = f31 E 1, D2 = -f31E2
(2.22)
D3 = f32 E 3, D4 = -f32E4
27
where
8i = -( _01. 2 (D12 + D66 + CI (2 Dl2 + 2 D66 + CI (H12 + H66)))2 ~~i + (-Ass - C2 (2Dss + C2 Dss)
- 01.
2 (D66 + CI (2 D66 + CI H66)) + (12 + CI (214 + CI Is)) w 2 + (Du + CI (2 Du + CI Hu)) ~~i) (A44
+ c~ D44 + CI (01.
2 (D22 + CI H22) - (14 + CI 16) w 2 - (D12 + 2 D66 + CI H12 + 2 CI H66) ~~i)))
+ D66 +CI (2F12 + 2F66 +CI (H12 + H66)))2 ~~i(-Ass - C2 (2Dss +C2 Fss) + CI (_01.
2 (H2 + 2F66.
+ CI (H12 + 2 H 66 )) + (14 + CI 16) w 2 + (Fu + CI Hu) ~~i)))/( -~oi (A44 + C2 (2 D44 + C2 F44) + 01.
2 (D22 .
+ CI (2 F22 + CI H22)) - (12 + CI (214 + cIIs)) w 2 - (D66 + CI (2 F66 + CI H66))~~i) (Ass + 2 C2 Dss + c~ Fss
+ CI (01.
2 (F12 + 2 F66 + CI (H12 + 2 H66)) - (14 + CI Is) w2 - (Fu + CI Hu) ~~i)) + 01.
2 (D12 + D66 + CI (2 Fl2
+ 2 F66 + CI (H12 + H66))) ~oi (A44 + 2 C2 D44 + c~ F44 + CI 2 (F22 + CI H ) - (14 + CI Is) w 2 - (F12
(01. 22
+ CI H22)) - (12 + CI (214 + CI Is)) w 2 - (D66 + CI (2 F66 + CI H66 )) ~~i)( -AS5 - C2 (2 Dss + C2 F55 )
+CI(-0I. 2 (F12 + 2F66 + CI (H12 + 2H66)) + (14 + cIIs)w 2 + (Fu +CI HU)~~i)))/(-~oi (A44 +C2 (2D44
+ C2 F44) + 2 (D22 + CI (2 F22 + CI H22)) - (h + CI (214 + CI Is)) w 2 - (D66 + CI (2 F66 + CI H6 )) ~~i) (Ass
01. 6
+ 2C2 Dss + c~ Fss + CI (01.
2 (H2 + 2 F66 + CI (H12 + 2 H66)) - (14 + CI 16)W 2 - (Fu + CI Hu) ~~i))
+ a 2 (D12 + D66 + CI (2H2 + 2 F66 + CI (H12 + H66))) ~oi (A44 + 2 C2 D44 + c~ F44 + CI (01.
2 (F22 + CI H22)
28
case and only four for the in-plane case can be respectively formulated. Thus
(2.25)
The expressions for forces and moments can also be found in the same way by sub-
stituting Eqs. (2.25) into Eqs. (2.8). In this way
. el'p2 X (E3 + E4e- 2 1'p2X) (-A12 0 + All I-Ip1 .82)) sin (ay) =Nxx sin(oy)
2
Qx (x,y) = (el'Ol X (B1 + B2e- l'ol X)(Ass + Ass 'h /-Lo1 +2C2 (Dss +DSS"l 1-101) +C~ (Fss + "1 FSSI-Iod
2
+ C1 (a ')'1 (F12 + 2 F66 + C1 H12 + 2 C1 H66) /-Lo1 - 11'~1 (Fll + C1 Hll + C1 "1 Hll 1-101) + a (2 F66
el'o2 X(B3 + B4 e- 2 1'o2 X)(Ass + Ass "2 1-102 + 2 C2 (Dss + Dss "2 1-102) + C~ (Fss + "2 Fss /-L02)
2
+ C1 (a ''"/2 (F12 + 2 F66 + q H12 + 2 q H66) /-L02 - /-L~2 (Fl1 + q Hl1 + C1 "2 Hl1 /-L02) + a (2 F66
el'o3 X(Bs + B6e-21'03 X)(Ass + Ass "3 1-103 + 2 C2 (Dss + Dss "3 /-L03) + C~ (Fss + "3 Fss /-L03)
2
+ q (0 "13 (F12 + 2 F66 + C! H12 + 2 q H66) /'03 - /'~3 (Fl1 + C! Hl1 + C1 "3 Hl1 /-L03) + 0 (2 F66
el'o4 X(B3 + B4 e- 2 1'o4 X)(Ass + Ass "41-104 + 2 C2 (Dss + Dss "41-104) + C~ (Fss + 84 Fss 1-104)
2
+ C1 (0 "14 (F12 + 2 F66 + q H12 + 2 q H66) 1-104 - 1-I~4 (Fl1 + q Hl1 + C1 84 Hll /-L04) + 0 (2 F66
29
Mxx (x, y) = (el'Ol X(Bl + B2 e- 2 l'ol X)(O? C! 'h (H2 + C! H12) + a ')'1 (D12 + C! (2 F12 + Cl H 12» - /Lol (Dll
+ Cl (2 Fll + Cl Hll + 64 Fll /Lo4 + C! 84 Hll /L04»») sin (a y) = Mxx sin (ay)
Mxy (x,y) = (el'OlX (Bl +B2e-2I-'OlX)(,),1 (D66 + q (2F66 +Cl H 66»/Lol + a (D66 +C! (2F66 + C! H66
el'o2 X (Bl + B2 e- 2I-'02 X )(,),2 (D66 + Cl (2 F66 + Cl H66» /Lo2 + a (D66 + C! (2 F66 + C! H66
P xx (x,y) = (el'Ol x (-Bl + B2e- 2 1'Ol X) (a2c181 H12 + a ')'1 (H2 +C! H12) - /Lol (Fll +C! Hll (1 +81 /Lol»)+
el'o2 x (-Bl + B2 e- 2 1'o2 X) (a 2 q 82 H12 + a ')'2 (F12 + Cl H 12) - /Lo2 (Fll + Cl Hll (1 + 82 /Lo2»)+
el'o3 X (-Bl + B2e-2I'o3X) (a 2 q 63H12 + a ')'3 (F12 +C! H12) - /Lo3 (Fll +C! Hll (1 + 83/Lo3»)+
el'o4 X
(-Bl + B2 e- 2 1'o4 X) (a 2 q 84H12 + a ')'4 (F12 +Cl H 12) - /Lo4 (Fll +Cl Hll (1 +84 /L04»»)
sin (a y) = P xx sin (a y)
(2.26)
At this point, zero boundary conditions are generally imposed to eliminate the con-
stants in the classical method in order to establish the frequency equation for a single
plate element. By contrast, the development of the dynamic stiffness matrix entails
imposition of general boundary conditions in algebraic form. Thus in order to develop
the two dynamic stiffness matrices for in-plane and out-of-plane cases (which will be
subsequently combined), the following boundary conditions are applied next.
30
In-plane case:
x = 0:
x = b:
(2.27)
x= 0 : N xx = - NXXI , N xy = - N XY1
x= b: N xx = NXX2 ' N xy = N XY2
Out-oj-plane case:
x =0 :
x =b:
(2.28)
x =0 :
x = b:
By substituting Eqs. (2.27), (2.28) into Eq.{2.25), the following relations are obtained:
(2.29)
Equations (2.29) relate the in-plane and out-of-plate displacement vectors to the
constant vectors by virtue of Ap and Ao matrices, which assume the following form
and
31
01 .
11 e b 1-'01 - f I c - bpo ! 12 c b 1-'02 -/2 e- b J..lo2 fa eb ILo3 - f3 e- b P-03 /4 c b J..lotl -i4 e- b .uo4
(2.31)
where
Ii = 5i ~loi, with i = 1,2,3,4
By applying the same procedure for forces and moments, i.e. substituting Eqs. (2.27),
(2.28) into Eq. (2.8) the following relationships are obtained:
(2.32)
The above equation relate the in-plane and out-of-plate force vectors to the constant
vectors by means of Rp and Ro matrices, which assume the following form
tl tl t2 t2
-gl gl -g2 g2
Rp= (2.33)
_e b jJ.P1 h _e- b jJ.P1 tl _e b jJ.P2 t2 _e- b jJ.p2 t2
eb jJ.p1 gl _e- b jJ.P1 gl e b jJ.p2 g2 _e- b jJ.P2 g2
where
32
and
Q, Q, Q2 Q2 Q3 Q3 Q4 Q4
7, -7, 72 -72 73 -73 74 -74
-I, -I, -I2 -I2 -I3 -I3 -I4 -I4
£, -£, £2 -£2 £3 -£3 YL -£4
Ro Ql eb ILol -Ql e-b~ol Q2 e b ""02 -Q2 c-l.o ~o2 Q3 e L flo3 -Q3 e- b 1-'03 Q4 e b J.Lo4 _Q4 C - L J.Lo4
-TIC blLol TIC-bJJ.ol -T2 Cb IL02 T2 e - b J.J.o2 -Ta eh '~o3 T3 e - b 1'-03 -T4e b J..lo4 T4 e - b 11-04
IIe h ILoi II e- b flol Z2ebPo2 I2 e - b JJ-02 Laeb ""·03 I3 e - b J.4-o3 I4 eb flo4 I4 e - b J.Lo4
-LICbP.ol CIC- b 1"01 -.ea cb ""02 £2 C - b J.Lo2 -e3 cb J.Lo3 C3e-bllo3 _£4ebPo4 £4C-blLo4
(2.35)
where
(2.36)
Ii = 1'1 (D66 + c1(2F66 + c1H 66 »l-toi - a(D66 + c1(2F66 + c1H 66 + 20i F66 l-toi + 2C10iH66l-toi»
with i = 1,2,3,4
By eliminating the constants vectors C p and Co the two dynamic stiffness matrices
for the in-plane and out-of-plane cases are respectively formulated as follows:
i.e.
Snn Snl Inn Inl
Sym Stl
33
Sqq Sqm Sqt Sqh fqq fqm fqt fqh
Stt -Sth
Shh
Finally the in-plane DS matrix Kp and the out-of-plane DS matrix Ko are combined'
together, to give the complete dynamic stiffness matrix as:
F=K~ (2.40)
or more explicitly
Stt -Sth
Shh
(2.41)
where the related force and displacement vectors assume the following forms:
F = [NXXI N XY1 QXl MXXI M XY1 P XX1 NXX2 N XY2 QX2 MXX2 M XY2 PXX2] T
(2.42)
34
(2.43)
The above dynamic stiffness matrix will now be used in conjunction with the Wittrick-
Williams algorithm [63] to analyze assemblies of composite plates to investigate their
free vibration characteristics based on HSDT. Explicit expressions for each element of
the DS matrix were obtained via symbolic computation, but they are far too extensive
and voluminous to report. The correctness of these expressions was further checked
by implementing them in a Matlab program and then carrying out a wide rage of
numerical simulations.
Once the DS matrix of a laminate element has been developed, it can be rotated
and/or offset if required and thus can be assembled to form the global DS matrix of
the final structure. The assembly procedure is schematically shown in Fig. 2.2 which
is similar to that of FEM. Although like the FEM, a mesh is required in the DSM,
/~,
~ ,...,......,.-,
Global Dynamic Stiffness Matrix
U
•••1ft)
0 0
0
- - . -.- ..
.
• ~
~ ~
it should be noted that the latter is mesh independent in the sense that additional
elements are required only when there is a change in the geometry of the structure. A
single DS laminate element is enough to compute any number of its natural frequencies
35
to any desired accuracy, which, of course, is impossible in the FEM. However, for the
type of structures under consideration DS plate elements do not have nodal points,
but have nodal lines instead. In this particular case, no change in geometry along the
longitudinal direction is admitted. This is in addition to the assumed simple support
boundary conditions on two opposite sides, inherent in DSM for plate elements at
present. The other two sides of the plate can have any boundary conditions. The
application of the boundary conditions of the global dynamic stiffness matrix involves
the use of the so-called penalty method. This consists of adding a large stiffness to
the appropriate leading diagonal term which corresponds to the degree of freedom of
the node that needs to be suppressed. It is thus possible to apply free (F), simple
. support (S) and clamped (C) boundary conditions on the structure by penalizing the
appropriate degrees of freedom. Clearly for simple support boundary condition, V, W
aild <Py are penalized. On the other hand, for clamped boundary condition U, V, W,
l " •
<P y, <P x , W, x will have to be penalized. Of course for the free-edge boundary condition
no penalty will be applied. Because of the similarities between DSM and FEM, DS
elements can be implemented in FEM codes and thus the accuracy of results, when
analyzing complex structures, can be enhanced considerably.
In order to compute the natural frequencies of a structure by using the DSM, an effi-
cient way to solve the eigenvalue problem is to apply the Wittrick-Williams algorithm
[63] which has featured in literally hundreds of papers. For the sake of completeness
the procedure is briefly summarized as follows.
First the global dynamic stiffness matrix of the final structure K* is computed for
an arbitrarily chosen trial frequency w*. Next, by applying the usual form of Gauss
elimination to the global stiffness matrix, it is transformed into its upper triangular
K*6 form. The number of negative terms on the leading diagonal of K*6 is now
defined as the sign count s(K*) which forms the fundamental basis of the algorithm.
In its simplest form, the algorithm states that j, the number of natural frequencies
(w) of a structures that lie below an arbitrarily chosen trial frequency (w*) is given
36
by:
]=]o+s(K*) (2.44)
where]o is the number of natural frequencies of all single elements within the structure
which are still lower than the trial frequency (w*) when their nodes are fully clamped.
It is necessary to account for these clamped-clamped frequencies because exact free
vibration analysis using DSM allows an infinite number of natural frequencies to be
accounted for when all the nodes of the structures are fully clamped, i.e. in the overall
formulation K ~ = 0, these natural frequencies correspond to ~ = °modes. Thus
]0 is an integral part of the algorithm and not really a peripheral issue. However,
unless exceptionally high frequencies are needed, ]0 is usually zero and the dominant
term of the algorithm is the sign-count s(K*), of Eq. (2.44). One way of avoiding the
. .
computation of the troublesome ]0 is to split the structure into sufficient number of
elements so that the clamped-clamped natural frequencies of an individual element
in the structure are never exceeded. Once s(K*) and ]0 ofEq. (2.44) are known,
any suitable method, for example, the bi-section technique, can be devised to bracket
any natural frequency within any desired accuracy. The mode shapes are routinely
computed by using a standard eigenvector recovery procedure in which the global
dynamic stiffness matrix is computed at the natural frequency and the force vector
is set to zero whilst deleting one row of the DS matrix and giving one of the nodal
displacement component an arbitrarily chosen value and then determining the rest of
the displacements in terms of the chosen one.
The main purpose of the following section is to highlight the refinement introduced by
the developed HSDT [169J in relation to the FSDT [170J. To this aim, both thin and
thick simple, stepped and stri!lger composite panels are analyzed. The section pro-
vides a detailed investigation of the free vibration behavior of several configurations
of composite stringer panels. In particular, the effect of the L and Omega stringer
sections on the free vibration characteristics of the reinforced composite panels are
37
examined and commented on.
The first set of results was obtained to validate the dynamic stiffness theory using
HSDT. For the fundamental natural frequency, Table 2.1 shows representative results
for the fundamental natural frequency in non-dimensional form for a cross-ply com-
posite square plate simply supported on all edges using the present theory along side
the published results from the literature. Of particular significance, is the inclusion of
the 3D elasticity solution and numerical results using ANSYS [171] which show close
agreement with the results obtained by the present theory. Note that the ANSYS
results were obtained by using SHELL181 element.
3 20 40
~1D%
0
3D-Elasticity [172] 6.6185 9.5603 ~3D% 10.7515 ~3D%
0
[0 /90 ]5
[0 0 /90 0 fools 3D-Elasticity [172] 6.6468 ~3D% 9.948 ~3D% 11.3435 ~3D%
ANSYS+ 6.5780 (-1.04) 9.7363 (-2.13) 11.051 (-2.58)
Classical Levy's solution Reddy [172] 6.5850 (-0.93) 9.8413 (-1.07) 11.2617 (-0.72)
[0 0 /90 0 /0 0 /90 0 /0 0 ]5 3D-Elasticity [172] 6.66 ~3D% 10.1368 ~3D% 11.6698 ~3D%
ANSYS+ 6.5879 (-1.08) 9.9986 (-1.36) 11.4926 (-5.74)
Classical Levy's solution Reddy [172] 6.5959 (-0.96) 10.0598 (-0.76) 11.6198 (-0.43)
Results in Table 2.1 cover a broad range of laminate lay-ups and stacking se-
quences. It is evident that the DS theory using HSDT predicts natural frequencies
38
of composite plate in an accurate manner. The maximum error incurred in the fun-
damental natural frequency when compared to 3D elasticity solution is 4.54% for an
artificially large value of the orthotropic ratio Ed E2 = 40. For realistic orthotropic
ratios, the error is expected to be much less. (Note that for carbon-epoxy and glass-
epoxy composite structures the ratio Ed E2 is around 10.) The next set of results was
obtained to examine the effects of the thickness to length ratio and the orthotropic
ratio on the fundamental natural frequencies of the square plate, simply supported
on all edges, but with stacking sequence [0°/90 %
°/90° /OO]s' The results using the
current DSM based on HSDT are shown in Table 2.2 together with the ones obtained
by using the DSM based on FSDT. Some interesting observations can be made from
these results. Clearly, the difference in natural frequencies when using the more accu-
rate HSDT as opposed to the relatively less accurate FSDT, increases when the plate
becomes progressively thicker', as expected. One of the anomalies in using FSDT
arises from the difficulty to select an appropriate shear correction factor (X), which
is generally introduced on an ad-hoc basis in an attempt to account for the correct
shear stress distribution which in reality is not uniform through the cross section.
Strictly speaking, the FSDT can never achieve zero shear stress distribution at the
free boundaries. Thus there is an element of uncertainty in choosing the shear correc-
tion factor and different authors have used different values (see Mindlin [6] , Reissner
[5]). The problem of choosing the shear correction factor is even more troublesome
for composites. However, this factor is taken to be 5/6 (see [5]) in the FSDT results
shown in Table 2.2. By contrast the HSDT results based on a refined displacement
field do not rely on such a fictitious (and quite often arbitrarily chosen) shear correc-
tion factor because the HSDT intrinsically accounts for the parabolic shear stresses
distribution. To confirm the predictable accuracy of the current method, 3D elas-
ticity solution has been used for comparison purposes. Both the influence of the
thickness-to-length ratio and the orthotropic ratio on results are also shown in Table
2.2. The next set of results are focused on the effect of boundary conditions. For two
representati ve values of thickness ratio (b / h ), the results in Table 2,3 show the effects
of the boundary conditions on the first four natural frequencies of the above plate. It
39
should be noted that both HSDT and FSDT results are included in the table. Clearly,
when the plate is simply supported on to opposite sides and clamped on the other
two sides, the natural frequencies assume higher values as expected. For this case,
the maximum error encountered is in the third natural frequency when FSDT is used
instead of the more accurate HSDT. The absolute values of these errors are around
7.5% and 4.2% when the thickness ratios are 5 and 10 respectively, as can be seen
in Table 2.3. It also evident from the results that on occasions, the FSDT results
are lower than the HSDT ones. The reason for this can be attributed to the fact
that the choice of the shear correction factor (which is non-existent and of course,
unnecessary in HSDT), influences the FSDT results in some unpredictable way. Such
discrepancies are not uncommon and can be found in the literature. In order to
40
Figure 2.3: A stepped composite plate
41
lar reference to Fig. 2.3. The ratios bI/b, b2 lb, b3 /b are taken to be liS, 1120, 1110,
respectively which are representative from a practical standpoint. The results were
obtained for different boundary conditions, and for a wide range of thickness ratios
between the stiffened plate and parent plate (tdtl) ranging from 2 to 6. The first
ten natural frequencies using the present theory with S-C-S-C and S-F-S-F boundary
conditions, are shown in Table 2.4 for different values of blh. The results shown are
exact and cannot be found in the existing literature, neither can they be obtained in
an exact sense using other methods. The following comments about these results are
relevant. Understandably, the natural frequencies are higher for'S-C-S-C boundary
conditions compared to S-F-S-F ones, as expected, but more importantly, for thick
plates, e.g. blt l = 2, increasing tdtl' decreases all the natural frequencies signifi-
cantly. By contrast, for relatively thin plates with bit 1 ~ 10 the natural frequencies
increase with increasing tdh ratio for this particular problem. The reason for this
can be attributed to the fact that for higher bit 1 ratio the effect of mass of the stiff-
ened plate appears to be more pronounced than its stiffness, yielding lower natural
frequencies as a consequence. The final set of results was obtained to demonstrate
the mode shapes of the composite plate and the stepped panel using HSDT based
DSM. In Figs. 2.4 to 2.6, a direct comparison of the first, fifth and ninth modes be-
tween the simple cross-ply laminated composite plate and the stepped panel has been
made for the boundary conditions S-C-S-C, when the step ratio tdtl = 2, whilst the
overall dimensions for the two configurations are kept the same. These figures reveal
some interesting features. For the fundamental mode, see Fig. 2.4, there is hardly
any difference in the natural frequency and mode shape between the simple plate
and stepped panel. This is in sharp contrast to the fifth and ninth modes shown in
Figs. 2.5 and 2.6 respectively, where appreciable differences in the natural frequencies
and mode shapes are noticeable. It is clear from these two figures that significant
alteration in the mode shapes is possible when required as a result of using stepped
panel. The corresponding results for S-F-S-F boundary condition are shown in Figs.
2.7, 2.8 and 2.9. Figure 2.7 shows that the fundamental natural frequency changes
significantly, but the mode shape follows more or less the same pattern. The sixth
42
and ninth modes shown in Figs. 2.8 and 2.9, fortuitously reveal the same picture as
the fundamental one shown in Fig. 2.7. It is interesting to note that for S-C-S-C and
S-F-S-F boundary conditions, results and trends are markedly different. These obser-
vations are important when solving frequency attenuation problems to avoid certain
undesirable natural frequencies and mode shapes of complex composite structures.
Table 2.4: Dimensionless natural frequency parameter w - Ii: E2' of a simply Wb/t
supported cross-ply stepped square composite plate, using a HSDT for stacking
sequence [0°/90°/90 °] , EdE2 - 40, G 12 /E 2 = G 13 /E2 = 0.6, G 23 /E 2 - 0.5,
%
bit, 10 50 bit, 2 10 50
Mode m n
1 1 1 6.097 1 1 18.473 1 1 30.046 1 1 3.403 1 1 7.244 1 1 8.002
2 2 1 9.227 2 1 30.000 2 1 41.947 1 2 3.946 1 2 8.217 1 2 9.586
3 1 2 12.115 1 2 37.404 . 3 1 71.210 2 1 .7.616 2 1 23.303 2 1 30.356
4 3 1 13.372 22 44.296 12 77.876 13 7.736 22 23.444 22 30.863
2 5 22 14.055 3 1 47.505 22 84.531 22 8.301 1 3 27.773 13 35.785
6 1 3 16.099 1 3 56.002 32 102.922 23 9.803 23 36.539 23 49.778
7 1 4 16.920 32 57.595 4 1 115.004 14 9.886 31 42.134 3 1 62.796
8 32 17.025 23 61.073 42 136.311 3 1 12.160 32 42.253 32 64.236
9 1 5 17.289 41 66.859 13 149.614 24 12.173 33 51.799 33 82.335
10 4 1 18.240 33 71.819 23 153.671 32 12.957 14 52.953 14 86.661
Mode mn mn mn mn mn
1 1 1 5.922 1 1 18.783 1 1 30.145 1 1 2.705 1 1 10.633 1 1 15.947
2 1 2 7.725 1 2 33.663 2 1 71.096 1 2 3.762 1 2 11.327 12 16.435
3 2 1 8.750 2 1 34.431 1 2 71.258 2 1 5.371 1 3 26.287 13 36.993
4 22 8.750 22 44.351 2 2 97.426 13 6.107 2 1 27.064 2 1 48.923
4 5 1 3 9.403 1 3 47.389 1 3 124.620 3 1 8.544 22 28.594 22 50.370
6 3 1 11.022 3 1 53.139 3 1 135.590 1 4 8.952 23 37.628 23 71.117
7 23 13.713 23 56.002 23 141.442 1 5 8.952 3 1 45.638 3 1 85.170
8 4 1 14.039 32 60.239 32 148.635 22 9.824 32 46.783 32 85.414
9 1 4 14.076 33 69.624 33 186.566 23 9.824 1 4 49.484 1 4 85.546
10 1 5 15.088 14 71.520 4 1 208.478 24 11.007 33 54.682 24 106.295
Mode mn
1 1 1 4.444 1 1 19.266 1 1 34.798 1 1 1.948 1 1 11.953 1 1 29.947
2 1 2 4.444 1 2 29.658 1 2 69.167 2 1 4.277 12 13.722 1 2 29.948
3 2 1 5.869 2 1 36.346 2 1 96.522 1 2 6.748 1 3 23.874 1 3 42.819
4 22 5.869 22 42.039 1 3 110.864 22 7.103 2 1 28.197 2 1 72.821
6 5 3 1 8.500 13 42.082 22 114.264 31 7.197 22 30.669 22 72.821
6 1 3 8.913 14 44.622 23 144.387 32 8.510 23 36.323 1 4 103.419
7 1 4 10.091 23 51.451 3 1 175.783 1 3 9.027 1 4 40.146 15 103.524
8 32 10.263 3 1 56.107 32 184.041 14 9.027 1 5 44.438 1 6 109.667
9 23 10.393 1 5 58.325 1 4 199.545 23 9.457 3 1 46.155 23 114.436
10 1 5 10.405 16 59.233 33 211.700 24 9.457 24 47.365 24 114.437
In the L stringer case the skin of the structures is a square plate 1 m x 1 m and
the stringer is placed in the middle of the plate. The web of the stringer is 0.2 m
high and its flange is 0.1 m wide. The thickness of each plate is kept constant for the
43
7 x
vZ«z7 .x
7 x
Zdllz x
Figure 2.5: Fifth natural frequency and mode shape of a simple and stepped composite
plate, with boundary condition S-C-S-C and stacking sequence [0°/90°]8' t
= 10,
!A
tl = 2 .
44
7 x
zvmz x
Figure 2.6: Ninth natural frequency and mode shape of a simple and stepped compos-
ite plate, with boundary condition S-C-S-C and stacking sequence [0°/90°]5' = 10, t
!2.
tl = 2.
7 x
zvZm x
Figure 2.7: Fundamental natural frequency and mode shape of a simple and stepped
composite plate, with boundary condition S-F-S-F and stacking sequence [0°/90°]5'
~ - 10 ' h
tl -
!2. -- 2 .
45
7 z~ x
x
Figure 2.8: Sixth natural frequency and mode shape of a simple and stepped compos-
ite plate, with boundary condition S-F-S-F and stacking sequence [0 0/90 0]8' t
= 10,
~ =2
tl
z x
z~ x
Figure 2.9: Ninth natural frequency and mode shape of a simple and stepped compos-
ite plate, with boundary
.
f
condition S-F-S-F and stacking sequence [0 0/90 0]8' 1 = 10,
!1.
tl
=2
46
whole structure and two different thickness ratios have been used, namely b/ h = 10
and b/h = 100. The material properties are: Ed E2 = 40, G 12 = G 13 = 0.6 E 2 ,
G23 = 0.5 E 2 , = 0.25, k = 5/6. The laminate lay-up is [0°/90°/90 °] and each
1/12
%
layer is 0.25 mm thick. With regard to the thick plate of ratio b/h = 10 a comparison
between the first three modes when computed using the FSDT and the HSDT is
shown in Fig. 2.10. Understandably, as expected the HSDT introduces a refinement,
leading to lower circular frequency parameters. However, the mode shapes remain
more or less unchanged for both cases. When thin stringer panels are studied, the
first three modes computed by using both FSDT and HSDT lead to similar results
(see Fig. 2.11). The effect of the refinement is not so prominent. Once more the
mode shapes related to the first three frequencies do not appear to be affected by the
kinematics. By contrast, as can be observed comparing Fig. 2.10 with Fig. 2.11 the
mode shapes undergo noticeable changes when varying the length-to-thickness ratio.
To underpin how the length-to-thickness ratio affect the dynamic behaviour of the
stringer panels the first twenty circular frequency parameters were computed by using
the developed HSDT and varying the ratio b/h from 5 to 100. The results are shown
in Table 2.5. It can be noticed how the circular frequency parameters increase when
increasing the b/ h ratio. Furthermore the presence of the mode shifting or modal
interchange phenomenon is observed.
In Fig. 2.12 the effect of the L stringers on the first three mode shapes is shown.
Geometrical and material data are similar to those used in the case of a single L
stringer and the length-to-thickness ratio is fixed at b/h = 100. The first three modes
are compared using bothe the FSDT and HSDT theories. First of all, it is useful to
highlight how in both cases (FSDT and HSDT) the introduction of two more stringers
leads to higher natural frequencies due to the increase in the global stiffness of the
structure. Again the mode shifting phenomenon is observed. Both the variation of the
length-to-thickness ratio and the introduction of a different number of stringers can be
successfully used to shift or defer a particular mode in solving frequency attenuation
47
0.1 0.1
0.09 0.09
0.08 0.08
0.07 0,(17
0.08 0.00
0.05 0.05
0.04 0.04
0.03 0.03
0.02 0.02
0. 0.'
02
02
,.m 0 0
O.ot
,.m 0 0
0.01
0.09 0.09
02
0.01
,.m o 0
02
0.1 0.1
0.09 0.09
0.08
o 0 ,.m o 0
(e) FSDT 3rd Mode w= 32.096 (f) HSDT 3rd Mode w= 31.531
Figure 2.10: Comparison between FSDT and HSDT first three modes and circular
frequency parameters w = whb Iii
48
0.1 0 .1
0.00 0."
0.08 0.00
om 0,07
0.00 0.00
0 .06 0 .06
0." 0."
0.03 0.03
0 .02 0.'"
0 .01 0.01
0.1 0 .1
0.00 0.00
0.'
Y, m
" 0 0
0.2 0 .01
(c) FSDT 2nd Mode w= 83.427 (d) HSDT 2nd Mode w= 83.362
0." 0.00
0.00 0.00
0 .07 0 .07
0.00 0.00
0 .05 0 .06
0." 0."
0.03 0.03
0.'" 0.'"
0.01 0.01
,.m ,.m o 0
(e) FSDT 3rd Mode w= 86.694 (f) HSDT 3rd Mode w= 86.694
Figure 2.11: Comparison between FSDT and HSDT first three modes and circular
frequency parameters w = whb ji;.
49
0.1 0.1
0.09 0.09
0.00 0.08
0.1ll 0.07
0.00 0.00
0.06 0.06
0." 0."
0.03 0.03
0.02 0.02
,.m ,.m
0.01 '"1
0.1 0.1
0.00 0.00
e
"
02
02 0 .01 0 .2 0.01
,.m ,.m o 0
(c) FSDT 2nd Mode w = 104.50 (d) HSDT 2nd Mode w= 103.93
0.12 0. 12
0.'
0 .2 02
0.2 02
,.m ,.m o 0
(e) FSDT 3rd Mode w= 141.56 (f) HSDT 3rd Mode w = 141.22
Figure 2.12: Comparison between FSDT and HSDT first three modes and circular
frequency parameters w = whb It
50
Table 2.5: Dimensionless natural frequency parameter w= whb -Iii, of a simply
supported cross-ply square composite plate, using a HSDT for stacking sequence
[0 /90 /90 /0 Ed E2 = 40, G l 2/ E2 = G13 / E2 = 0.6, G23 / E2 = 0.5, V12 = V13 =
0 0 0 0
]'
V23 = 0.25.
b/h 5 10 50 100
Mode mn mn mn mn
1 11 10.250 11 16.343 11 49.902 11 74.662
2 12 11.112 12 21.403 12 63.881 12 83.362
3 21 19.103 21 31.531 21 80.279 21 86.694
4 13 20.654 13 34.215 22 86.273 22 103.902
5 22 22.091 22 40.620 13 91.099 31 105.424
6 14 24.114 14 41.365 31 99.954 32 121.612
7 23 26.401 23 44.461 32 109.767 13 124.826
8 15 29.000 31 50.118 14 118.454 41 l41.104
9 31 29.122 15 50.599 15 127.454 42 154.086
10 24 29.541 24 53.105 41 134.472 14 166.251
11 32 32.748 32 56.178 42 143.192 51 193.893
12 25 32.997 25 57.339 23 145.868 52 203.781
13 33 34.219 33 61.899 24 171.935 15 230.134
14 34 38.634 34 69.334 51 183.425 23 234.371
15 41 39.318 41 69.710 52 190.316 24 261.779
16 35 39.829 42 73.026 33 201.898 61 261.944
17 42 42.128 35 74.775 25 226.658 16 266.774
18 43 44.137 43 82.123 16 234.562 33 269.493
19 44 47.472 16 85.950 34 235.527 62 269.499
20 51 49.560 44 86.186 26 240.074 43 290.674
51
constant for the whole structure and the thickness ratio is b/h = 100. The material
properties are: Ed E2 = 40, G 12 = G 13 = 0.6 E 2, G 23 = 0.5 E 2, 1I12 = 0.25, k = 5/6.
2.3 Conclusions
An exact dynamic stiffness theory for composite plate elements using higher order
shear deformation theory is developed for the first time using Hamiltonian mechanics
and symbolic algebra. The theory is implemented in a computer program to carry out
the free vibration analysis of composite structures modelled as plate assemblies. The
proposed theory is a significant refinement over recently developed dynamic stiff-
ness method using classical and first order shear deformation plate theories. The
developed DSM model is particularly useful when analyzing thick composite plates
with moderate to high orthotropic ratios for which the FEM may become unreliable,
particularly at high frequencies. A detailed parametric study has been carried out
by varying significant plate parameters and boundary conditions. The results have
been critically examined and the theory has been assessed using existing theories,
FEM simulation and in particular, three-dimensional mathematical theory of elas-
ticity. Stepped composite plates and composite plate with stringers have also been
analyzed for its dynamic behaviour. Based on the computed results the following
comments can be made:
1. The proposed exact dynamic stiffness composite plate element based on HSDT is
52
0.1 0.1
0.09 0.'"
0.08 0.08
0.07 0.Q7
0.08 0.00
0.'" 0.05
0.04 0.04
0.00 0.00
0.02 0.02
0.0 1 0.01
(a) FSDT 1"t Mode w= 58.961 (b) HSDT l,t Mode w= 58.894
0.1 0.1
0.09 0.09
0.01 0.08
0.07 0.01
0.00 0.08
0.06 0.05
0.04 0.04
0.00 0.00
0.02 0.02
0.5 0.5
(c) FSDT 2nd Mode w= 93.963 (d) HSDT 2nd Mode w= 93.944
0.1 0.1
0.'" 0.09
0.08 0.08
0.Q7 0.07
0.08 0.08
0.06 0.05
0." 0."
0.00 0.00
0.02 0.02
0.5 0.5
Figure 2.13: Comparison between FSDT and HSDT first three modes and circular
frequency parameters w= whb Ii
53
shown to be more accurate in terms of results and computational efficiency when
compared with FEM in free vibration analysis of composite plate assemblies.
3. The boundary conditions do not seem to affect the error incurred using FSDT
as opposed to the more accurate HSDT.
4. The dynamic behaviour of stepped composite plates and composite plate with
stringers are very different from those of simple uniform composite plates de-
pending on the boundary conditions, but significant alteration in mode shapes
(modal intercha~ge) is possibie by using stepped panels. This could be useful
in solving frequency attenuation problems.
54
Chapter 3
Buckling of Plates
An exact dynamic stiffness element based on higher order shear deformation theory
and extensive use of symbolic algebra is developed to carry out buckling analysis of
composite plate assemblies. The principle of minim)lm potential energy is applied to
derive the governing differential equations and natural boundary conditions. Then by
imposing the geometric boundary conditions in algebraic form the dynamic stiffness
matrix, which includes contributions from both direct stiffness and initial pre-stress
terms, is developed. The Wittrick-Williams algorithm [98] is used as the solution
technique to compute the critical buckling loads and mode shapes for a range of
laminated composite plates. The effects of significant parameters such as thickness-
to-length ratio, orthotropy ratio, number of layers, lay-up and stacking sequence and
boundary conditions on the critical buckling loads and mode shapes are investigated.
The accuracy of the method is demonstrated by comparing results wherever possible
with those available in the literature.
The displacement field employed in the following stability analysis is the one used
in chapter 2 and explicitly given in Eq. (2.1). The principle of minimum potential
55
energy is now applied. The variational statement at multilayer level is:
(3.1)
where rr k is the total potential energy for the kth layer of the composite plate. The
first variation can be expressed as:
(3.2)
where, 8U k is the virtual potential strain energy, 8V k is the virtual potential energy
due to external loadings, and assume the following form:
(3.3)
O-xo and o-yO denote the in-plane initial stresses. The non-linear strains c~; and c~t are
approximated with the Von Karman's non-linearity:
)2
knl
c xx ="21 (
W,X (3.5)
The subscript T in Eq. (3.4) signifies an array transposition and 8, in Eqs. (3.1),
(3.2) and (3.3), the variational operator. Constitutive and geometrical relationships
are defined respectively as:
e='DTJ (3.6)
where i/ is the plane stress constitutive matrix and 'D is the differential matrix (see
Appendix A for details). Substituting Eq. (3.6) into the Eq. (3.3) and imposing the
condition in Eq. (3.1), the stability equations are obtained. Their explicit form is
similar to the Eqs. (2.7), in which the dynamic terms in the right-hand side are ne-
glected. Moreover in the equation related to transverse displacement w the following
function N (w) = Nxo wo,xx + NYo WO,yy is introduced on the right-hand side of the
56
equation in order to include the dead loads. The in-plane loadings can be defined as
Nxo = >.. Nxo and Nyo = >.. N yO ' where N xo ' NyO are the initial in-plane loadings and >..
is a scalar load factor.
Once the stability equations and the natural boundary conditions are obtained, the
classical method to carry out an exact buckling analysis of a plate consists of (i)
solving the system of differential equations in Navier or Levy-type closed form in an
exact manner, (ii) applying particular boundary conditions on the edges and finally
(iii) obtaining the stability equation by eliminating the integration constants [165-
168]. This method, although extremely useful for analysing an individual plate, lacks
generality and cannot be easily applied tb complex structures assembled from plates·
for which researchers usually resort to approximate methods such as the FEM. In this
respect, the dynamic stiffness method (DSM), which is, in many ways, analogous to
FEM has no such limitations and importantly it always retains the exactness of the
solution even when applied to complex structures. This is because once the dynamic
stiffness matrix of a structural element is obtained from the exact solution of the
governing differential equations and it can be offset and/or rotated and assembled in
a global DS matrix in the same way as the FEM. This global DS matrix thus contains
implicitly all the exact critical buckling loads of the structure which can be computed
by using the well established algorithm of Wittrick and Williams [98].
A general procedure to develop the dynamic stiffness matrix of a structural element
is generally summarized as follows:
(i) Seek a closed form analytical solution of the governing differential equations of
the structural element.
(ii) Apply a number of general boundary conditions in algebraic forms that are
equal to twice the number of integration constants; these are usually nodal
displacements and forces.
57
(iii) Eliminate the integration constants by relating the amplitudes of the har-
monically varying nodal forces to those of the corresponding displacements
which essentially generates the dynamic stiffness matrix, providing the force-
displacement relationship at the nodes of the structural element.
Referring to the stability equations, an exact solution can be found in Levy's form
for symmetric, cross ply laminates. For such laminates B = E = 0, and Of6 = 0~6 =
0: = 0 and the out-of-plane displacements are uncoupled from the in-plane ones.
5
00 00
(3.8)
58
where Lij with i, j = 1,2,3, are differential operators given by:
£13 = -a (A44 + c2(2D44 + C2 F44) + a 2c1 (F22 + C1 H 22)) + V; (ac1F12 + 2ac1F66 + acr H12
+ 2acrH66)
£21 = c1V;(Fu + C1 H U) + Vx ( - A55 - c2(2D55 + C2 F5fi) - a 2c1(F12 + 2F66 + C1 H 12
+ 2c1H 66 ))
,
£22 = -A55 - c2(2D 55 + C2 F55) +'V; (Du + 2C1FU + crHu) - a 2 (D66 + 2C1F66 + Cr H 66)
£23 = Vx (- aD12 - aD66 - 2ac1F12 - 2ac1F66 - aCrH12 - aCrH66)
(3.9)
(3.10)
where
(3.11)
59
Using a trial solution e A in Eq. (3.10) yields the following auxiliary equation:
(3.12)
Substituting /-L = ).2, the 8th order polynomial of Eq. (3.12) can be reduced to a
quartic as:
(3.13)
The four roots for the quartic equation are given by:
(3.14)
where
The explicit form of the polynomial coefficients aj with j = 1,2,3,4,5 are given in
Appendix C. Some pair or pairs of complex roots may occur when computing /-Lj with
j = 1,2,3,4, but the amplitudes of the displacements Wm (x) , <PXm (x) , <Pym (x) are
all real, whilst the associated constant coefficients can be complex. As complex roots
always occur in conjugate pairs, the associated constant coefficients will also occur
as conjugates. Setting /-Loj = ..;rr; with j = 1,2,3,4, the solution of the system of
60
ordinary differential equations shown in Eq. (3.8) can thus be written as:
61
where
2 2 2 4 2 2
6i = - [ - Asset D22 - 2As5C2D44 - 20' C2D22DS5 - 4c2D44D55 - a: D22D66 - 20: C2D44D66 - 2A5SQ: cIF22
+ A44(Dll + q (2Fll +q Hll »/L~ + (A55(D66 + q (2F66 + qH66» + 2C2(DllD44 + D55D66 + q (2D44Fll
- D22(2Fll + qHll) + q(8F12F66 + 2D66H12 - 2F,,(2F22 + cIH22) + q(-2F22Hll + H12(4(F12 + F66) + qH12)
'"Yi = [o:( A44AS5 + 2A55C2D44 + 2A44C2D55 + 4c~D44D55 + A44Q2 D66 + 2a?c2D44D66 + A5SQ: 2 CIF22 + 2Q2CIC2D55F22
+ cr4cID66F22 + A55C~F44 + 2c~D55F44 + 0'2c~D66F44 + A44C~F55 + 2c~D44F55 + Q?CIC~F22F55 + c~F44F55
+ 2A440?CIF66 + 4Q2C}C2D44F66 + 2a:4ci'F22F66 + 2o?CIC~F44F66 + A55a?c~H22 + 2a:2c~C2D55H22 + a:4c~D66H22
+ a:2cic~F55H22 + 2ct4C~F66H22 + A44ct2C~H66 + 2a?cic2D44H66 + ct4C~F22H66 + ct2cic~F44H66 + a:4c1H22H66"
+ (A5S(D12 + D66) - A44(Dl1 + q(2F" + qH,,» - 2C2(D"D44 - D55(D12 + D66) + q(2D44F" - D5SF12 + qD44Hl1
+ q D55H66» + c~( -D" F44 + (D12 + D66)FS5 - q (2Fll F44 - F12FSS + q F44H" + q FS5H66» + q (AS5(F12 - q H66)
+ 02( -D"F22 + D12(F12 + 2F66 + q(H12 + 2H66» + q(2F~2 - D"H22 - 2F,,(F22 + qH22) + F12(4F66 + 3qH12
+ 4qH66) + q(H12(2F66 + qH12) - H,,(F22 + qH22) + 2qH12H66»»)I'~ + q(D,,(F12 + 2F66) - D12(F" + qH,,)
- D66(F" + qHl1) + q(D,,(H12 + 2H66) + qH,,( -F12 + qH66) + F,,( 2F66 + q(H12 + 3H66»»I'1) I
/ [I'; (A4.A55 - A44,,2 D12 + A5502 D22 + 2As5C2D.4 - 2,,2c2D12D44 + 2A44C2DS5 + 2,,2c2D22D55 + 4c~D44D5S
- A44a:2D66 - 2a:2C2D44D66 - A440?qF12 + a: 4 CID22F12 - 2a:2cIC2D44F12 + 2A55a:2qF22 4
- a: CID12F22 + 4a: 2c IC2 D 55 F 22
- ct4CID66F22 + A55C~F44 - 0'2c~D12F44 + 2c~D55F44 - ct2c~D66F44 - ct2CIC~F12F44 + A44C~F55 + ct2C~D22F55
+ 2c~D44F55 + 20?CIC~F22F55 + C~F44F55 + 2ct 4 q D22F66 + 2ct 4 c i F22F66 + a: 4 c i D22H12 + a: 4c~ F22H12 + A55ct2ci H22
- a:4ciD12H22 + 2a'?cic2D55H22 - a:4ciD66H22 - a:4c~F12H22 + o?cic~F55H22 + A44a:2ciH66 + 2a4ciD22H66
+ 20 22 43 222 44 (
cl C2D44H66 + 30 c1F22H66 + ex cl C2F44H66 + 0 c1H22H66 - A5S(D66 + q(2F66 + qH66» + c2(D66F5S
2
62
Thus
(3.19)
<I>ym (x) = B l 1'l e+l'lX - B 2 1'l e-l'lX + B 3 1'2 e+ JL2X - B 4 1'2 e-1'2 X
+ B 5 1'3 e+1'3 x - B 6 1'3 e-1'3 x + B 7 1'4 e+1'4 x - B 8 1'4 e- P,4 x
where 5i with i = 1,2,3,4 have already been defined in Eq. (3.18). The expressions
for forces and moments can also be found'in the, same way" by substituting Eqs. (3.19)
into the boundary conditions and using symbolic computation. In this way
2
Qx (x, y) = (eM x (Bl + B2 e- 1', X)(A55 + A55 'h 1'1 + 2 C2 (D55 + D55 'h I'll + c~ (F55 + 'h F55 /ll)
+ Cl (0 'Yl (F12 + 2 F66 + Cl H12 + 2 Cl H66) 1'1 - I'~ (Fll + Cl Hll + Cl 'h Hll 1'1) + 0 2 (2 F66
+ 2 Cl H66 + C! 'h H12 1'1 + 4 Cl 'h H66 1'1)))+
el'2 X(B3 + B4 e- 2 1'2 X)(A55 + A55 82 1'2 + 2 C2 (D55 + D55 82 1'2) + c~ (F55 + 82 F55 Jl2)
2
+ Cl (0"/2 (F12 + 2 F66 + Cl H12 + 2 Cl H66) 1'2 - I'~ (Fll + Cl Hll + Cl 82 Hll 1'2) + 0 (2 F66
el'3 X(B5 + B6e-21'3 X)(A55 + A55 83 1'3 + 2 C2 (D55 + D55 83 JL3) + c~ (F55 + 83 F55 1'3)
+ Cl (0"/3 (F12 + 2F66 +Cl H12 + 2Cl H 66 ) 1'3 - 1'5 (Fll +Cl Hll + c183Hll 1'3) +0 2 (2F66
el" X(B3 + B4 e- 2 1" X)(A55 + A55 84 1'4 + 2 C2 (D55 + D55 84 1'4) + c~ (F55 + 84 F55 1'4)
+ C! (0"/4 (F12 + 2F66 +Cl H12 + 2C! H66) 1'4 - I'~ (Fll +Cl Hll + Cl84Hll 1'4) +0 2 (2F66
63
Mxx (x, y) = (el'l X(BI + B2 e- 2 1'1 x)(02 Cl 'h (F12 + Cl H 12) + 0"11 (D12 + Cl (2 F12 + q H I 2)) - J.£1 (Dll
eM X(B3 + B4 e- 2 1'2 X)(02 q 82 (F12 + q H 12) + 0"12 (D12 + q (2 F12 + Cl H I 2)) - J.£2 (Dll
el'3 X(B5 + B5 e- 2 1'3 x)(02 C! 83 (F12 + q H12) + 0"13 (D12 + q (2 F12 + C! HI2)) - J.£3 (Dll
+ Cl (2 Fll + Cl Hll + 84 Fll J.£4 + Cl 84 Hll J.£4)))) sin (oy) = Mxx sin (a y)
Mxy (x, y) = (el'lX (Bl + B2 e- 21'1x)hl (D66 + Cl (2 F66 + q H66)) J.£1 + a (D66 + q (2 F66 + C! H66
el'2X (Bl' + B2 e- 21'2 X)("I2 (D66 + q (2 F66 + q H66)) J.£2 + a (D66 + q (2 F66 + Cl H66
Pxx (x, y) = (el'l x (-Bl + B2 e- 2 1'1 X) (0 2 q 81 H12 + 0"11 (F12 + Cl H12) - J.£1 (Fll + Cl Hll (1 + 81 J.£tl))+
el'2X (-Bl + B2e-21'2X) (0 2 q 82 H12 + 0"12 (Ft2 +q H12) - J.£2 (Fll +q Hll (1 +82J.£2)))+
el'3 X (-Bl + B2e-21'3X) (0 2 q 83H12 + 0"13 (Ft2 + q H12) - J.£3 (Fll +Cl Hll (1 +83J.£3)))+
el'4X (-Bl +B2e-21'4X) (0 2 q 84H12 +0"14 (Ft2 +q H12) - J.£4 (Fll + Cl Hll (1 +84 J.£4))))
sin (a y) = P xx sin (a y)
(3.20)
Zero boundary conditions are generally used to eliminate the constants when using
the classical method which establishes the stability equation for a single individual
plate. By contrast, the development of the dynamic stiffness matrix entails imposi-
tion of general boundary conditions in algebraic form and widens the possibility of
the analysis to multi-plate systems. The following boundary conditions are applied:
Generalized displacements
x=o:
x =b :
64
Generalized forces
x=o:
(3.21)
x= b:
By substituting Eq. (3.21) into Eq. (3.19), the following relations for the displace-
ments are obtained:
d=AC (3.22)
Equation (3.23) relates the displacement vector to the constants vector by virtue of
the matrix A, which assumes the following form
6.
By applying the same procedure for forces and moments, i.e. substituting Eq. (3.21)
into Eq. (??) the following relationship is obtained:
F=RC (3.24)
Equation (3.24) relates the force vector to the constants vector by means of the matrix
65
R, which can be written as follows
Q, Q, Q2 Q2 Q3 Q3 Q. Q4
T, -T, T2 -T, T3 -T3 T. -T4
-II -I, -I2 -I2 -I3 -I3 -I4 -I.
.c, -.c, .c2 - .c2 .c3 -.c3 YL -£,
R - Ql cbl.l. o1 -'21 e-
b
J1-ol Q2 ch.uo2 -Q2 e- b l-'-o2 Q3 e bfLo3 _Q3c-bfLo3 Q4 e b 1-'-04 -Q4 e- bPo4
-TIC b 1-'-01 TIC-bf£ol -T2 Cb 1-'-02 T2 C - b 1-'-02 -Ta cb 1.1.03 Ta c - b J.l-03 _T4CbfLo4 T4 C - b ""'o4
-C 1 c b l-'ol CIC-b,J.ol _C2 cbl J. o 2 C2 C - b J1-o2 - C3 cb jJoo3 C3 e - b J..Lo3 -C4 eb J1-o4 C4 C - b !J.o4
(3.25)
where
Qi = -A55(1 + c5i J.lol) - 2C2(D55 + D 55 c5i J.loi) - C~(F55 + c5i F 55J.loi) - cl(a1'i(F12 + 2F66 + c1H 12
2
+ 2c1H 66 )J.loi - J.l~i(Fll + c1Hl l + c 1c5i H ll J.loi) + a ( 2F66 + 2c 1H 66 + c 1c5i H 12 J.loi + 4c1c5i H 66 J.loi))
with i = 1,2,3,4
(3.26)
By eliminating the constants vector C from Eq. (3.22) the dynamic stiffness matrix
is formulated as follows:
(3.27)
or more explicitly
8 mm -8 m t 8m h
8tt -8th
8hh
66
Finally the dynamic stiffness matrix relating the force' and displacement vectors can
be written as follows:
P XX2 Shh W 2 ,x
The above dynamic stiffness matrix will now be used in conjunction with the Wittrick-
Williams algorithm [98] to analyze composite simple and stiffened plates for their
buckling behaviour based on HSDT. Explicit expressions for each element of the DS
matrix were obtained via symbolic computation, but they are far too extensive and
voluminous to report here. The correctness of these expressions was further checked
by implementing them in a Matlab program and carrying out a wide rage of numerical
simulations. The assembly procedure and the Wittrick-Williams algorithm follow the
same procedure provided in Chapter 2, see section 2.1.4 and section 2.1.5, respectively.
Results are shown accounting for the boundary conditions depicted in Fig. 3.1 and
the loadings shown in Fig. 3.2. A preliminary validation of the critical buckling
load analysis for moderately thick (a/h = 10) simply-supported cross-ply square
plates uniaxially loaded in the x direction is carried out and the results are shown
in Table 3.1 for different orthotropy ratio. The dimensionless critical buckling load,
obtained using HSDT within the framework of the DSM are in excellent agreement
when compared with the 3D elasticity solution and the results also lead to the same
67
y y y
x x x
y y y
~ ---------------------------
~
SFSF ~
~ SCSF
~
~ ---------------------------
x x x
findings as the classical Levy-type closed form solution. Note that for all practical
purposes,. it is only the first buckling load that matters. Therefore only the first
. critical load is presented in this paper. As expected the percentage error, with
x
b
respect to the 3D elasticity solution, increases when increasing the orthotropy ratio.
In Table 3.2 the dimensionless critical buckling load for the same case study of Table
3.1 is computed but taking into account the effects of the length-to-thickness and
orthotropy ratios and boundary conditions (see Fig. 3.1). At a fixed length-to-
thickness ratio, the dimensionless critical buckling load increases when increasing the
68
2
Table 3.1: Dimensionless uniaxial buckling load (along x direction) Ncr = NCT E: h 3 '
for simply supported cross-ply square plates with b/ h = 10, Ed E2 = open, G 12/ E2 =
G 13 / E2 = 0.6, G 23 / E2 = 0.5, 1I12 = 1I13 = 0.25.
Classical Levy's solution i·iSDT 5.393 (1.68) 15.298 (1.86) 23.340 (2.01)
FSDT 5.399 (1.79) 15.351 (2.21) 23.453 (2.50)
CLPT 5.754 (8.48) 19.712 (31.2) 36.160 (58.0)
orthotropic ratio for' all the considered boundary conditions. A similar behaviour can
be observed when varying the length-to-thickness ratio but by fixing the orthotropy
ratio. Understandably, the largest dimensionless critical buckling load is given by the'
boundary condition S-C-S-C and the lowest by S-F-S-F. In Table 3.3 the results are
given for composite plates that are uniaxially loaded in the y direction, instead of
the x direction for different values of length-to-thickness and orthotropy ratios. The
dimensionless critical buckling load is generally lower for all the boundary conditions,
but for the case with one or two sides free, namely, S-S-S-F and S-F-S-F, it decreases
significantly. The biaxial compression effect is examined and the results in Table 3.4
show as expected, a notable reduction in the critical buckling load, with respect to
the uniaxial load considered along x and y axes, respectively. In Table 3.5, results
for moderately thick plate (a/h=10) with the effect of the in-plane ratio (L/b) are
presented for different values of orthotropy ratio. As can be seen from this table, the
critical buckling load decreases when increasing the in-plane ratio, independently for
all boundary conditions.
A particular feature of the DSM is that it allows the buckling analysis of stiffened
plates in an exact sense. Two different stiffened composite plate configurations shown
69
Table 3.2: Dimensionless uniaxial buckling load (along x direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [0°/90°/90 °] and
%
70
Table 3.3: Dimensionless uniaxial buckling load (along y direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [00 /90 0 /90 0 /0 0 ] and
Ed E2 = open, G 12 / E2 = G 13 / E2 = 0.6, G 23 / E2 = 0.5, ZlI2 = ZlI3 = 0.25.
71
2
Table 3.4: Dimensionless biaxial buckling load Ncr = Ncr Eb h3 , for simply sup-
ported cross-ply square plates, stacking sequence [0 /90 /90 /ooJ and Ed E2 = open,
0 0 0
72
Table 3.5: Dimensionless uniaxial buckling load (along x direction) Ncr = Ncr E~2h3'
for simply supported cross-ply square plates, stacking sequence [0°/90°/90° /ooJ and
Ed E2 = open, G 12 / E2 = G 13 / E2 = 0.6, G 23 / E2 = 0.5, 1I12 = 1I13 = 0.25 and
b/h = 10.
73
in Fig. 3.3 and Fig. 3.4 respectively, are analyzed and the results are discussed in
this section. The geometrical parameters of the two stiffened composite plates are
given below:
0.10 m; t2 = 0.20 m;
In addition to the above stiffened pal).els, a simple uniform panel of thickness 0.002
m and with the same of the overall dimensions of Figs. 3.3 and 3.4, has also been
analyzed for comparison purposes. In Fig. 3.S the dimensionless critical buckling
load for different boundary conditions are shown against the length-to-thickness ratio
for the two stiffened composite plates alongside the results of the simple panel of
uniform thickness. The panels were loaded in the x direction. As can be seen from the
figure, the introduction of the stiffeners increases the buckling loads very considerably,
particularly for thin composite plates. On the contrary, when thick composite plates
are analyzed the increase in the critical buckling load is not so prominent. Clearly,
74
b
Figure 3.4:. 2nd stiffened composite plate configuration.
the results are dependent on the applied boundary conditions. Indeed, in the case
of S-S-S-S boundary condition the use of stiffeners increases the critical buckling
load for both of the stiffened plate configurations. By contrast, using a S-C-S-C
boundary condition does not affect the critical buckling load so significantly. It is
interesting to note that the relative advantages of using the first or the second stiffened
composite plate configurations depend on the applied boundary conditions, although
the differences in the dimensionless buckling loads between the two configurations
are sometimes rather small. In particular, when applying S-S-S-S, S-S-S-C and S-
C-S-C boundary conditions, the first stiffened composite plate configuration leads to
higher buckling load. Whilst, applying S-S-S-F, S-F-S-F and S-C-S-F the highest
buckling loads are reached with the second stiffened panel configuration. Finally
in Figs. 3.6, 3.7 and 3.8 the representative mode shapes corresponding to simple
composite plates, first and second configurations of stiffened composite plates are
respectively shown. When presenting modes, square symmetric cross-ply plates made
of five layers and uniaxially loaded along the y direction are considered. It is now
possible to provide an overview of how the mode shapes change when changing the
geometrical characteristics of the composite plate assemblies. It should be noted that
the mode shape of a S-S-S-S simple composite plate of uniform thickness related to
75
14,------------------------------,
12
10
. 8
?o
r 20
.
40
Y..
blh
60
.-.
7
80 100 20 40
blh
60 80 100
~~---2~0----~
40-----6~
0 ----8~
0----~
IOO
20 40 60 80 100
blh blh
14
25
12
20
l5 ~ j
8
I !
1
10' •
~~---2~0----~
40~--~6~
0 ----8~0--~100 ~~---2~0----~
40~--~6~
0 ----8~
0--~100
blh blh
Figure 3.5: Dimensionless uniaxial buckling loads Ncr = Ncr E~2h3 along the x direction
varying the length-to-thickness ratio for simple and stiffened cross-ply plates, stacking
sequence [0° /90°/0° /90° /0°]' step ratio t 2 /t 1 = 2 and Ed E2 = 5, G 12 / E2 = G 13 / E2 =
0.6 , G 23 / E2 = 0.5, V12 = V13 = 0.25.
76
its critical buckling load is made up of two half-waves in the x direction and one
half-wave in the y direction. On the contrary, in either of the two stiffened composite
plates the mode shape is characterized by only one half-wave in both direction x and
y. Other changes in the mode shapes related to different boundary conditions can be
observed in the same figures. The results are particularly useful when controlling the
mode shapes which generally have significant impact on response.
3.3 Conclusions
An exact dynamic stiffness theory for composite plate elements using higher order
shear deformation theory is developed using the principle of minimum potential en-
.,
ergy and symbolic algebra to carry out buckling analysis in an exact sense. The theory
is implemented in a computer program to carry out buckling analysis of complex com-
posite structures modelled as plate assemblies. The proposed theory is a significant
refinement over other dynamic stiffness theories using classical plate theory and/or
first order shear deformation theory. The developed DSM model is particularly use-
ful when analyzing thick composite plates with moderate to high orthotropic ratios
for which the FEM may become unreliable. A detailed parametric study has been
carried out by varying significant plate parameters and boundary conditions. The
results have been critically examined and the theory has been assessed using existing
theories. Two different stepped composite plate configurations have been analyzed
for their stability behavior. Based on the computed results the following conclusions
can be drawn:
1. The exact HSDT plate element has been shown to be extremely accurate in
terms of results and computational efficiency when carrying out buckling anal-
ysis of composite plate assemblies.
2. The exact HSDT plates element provides a significant refinement over to the
FSDT element particularly when thick plate with a high orthotropy ratio are
analyzed.
77
0.9
0.08
0.06
0.9 ~~~ 0.09
0.08
0.8 0.8
0.04 0.07
0.7 0.7
0.08
0.06
111_ _
0.9
0.09
0.08
0.8
0.0.76 ~~
0.04
0.07
0.6
0.02
~ 0.06
>- 0.5
o >0 0.5 0.05
0.4~~
0.4 - 0.02 0.04
11
-0.06 0.02
0.1
°0L-~0.-2--~0.4~~0
.6--~0~
.8--~
-0.08
0·:. . . . 0.01
0.09 0.08
0.9 0.9
0.08 0.06
0.8 0.8
0.3 0.3
0.03 - 0.04
0.2
0.02 - 0.06
O:L~-==:::===~~~~
0.1
0.01 - 0.08
O~~--~--~~~~
78
0.09 0.09
0.9
0.08 0.08
0.8
0.07 0.07
0.7
0,08
0,9 _ _ _ _ 0,09
0.9
0,06
0,08
0,8
0,04
0,07
0.7
0,02
0,6 0,06
0.09 0,08
0,9 0,9
0.08 o,OS
0,8
0.07 0.04
0.7
Figure 3.7: First buckling modes for the first stiffened cross-ply plate configurations
Fig. 3.3, lamination scheme [00/900/00/900/00] under uniaxial compression along the
y direction, length-to-thickness ratio b/t 1 = 10 and orthotropic ratio EI/E2 =25.
79
0.09 0.09
0.08 0.06
0.6
0.07 0.07
0.7
1
0.06
0.09
0.9 0.9
0.06
0.06
0.8 0.6
0.04
-
-
0.07
0.7
0.02
0.7
- -
-~
0.6 0.6 0.06
~-
>- 0.5 ,.. 0.5 0.05
0.8
0.09
0.06
0.9
--- 0.08
0.06
0.8
0.3 0.3
0.03 -0.04
0.2 0.2
0.02 - 0.06
'-...~
0.1
O:L~~::::===~g~~~
o 0.2 0.4 0.6 0.8
0.01
0
0 0.2 0.4 0.6 0.6
- 0.08
Figure 3.8: First buckling modes for the second stiffened cross-ply plate configurations
Fig. 3.4, lamination scheme under [0°/90 °/90° /ooJ uniaxial compression along the
%
80
3. The boundary conditions affect significantly the buckling modes.
5. The buckling load of stiffened composite plates changes prominently with re-
spect to the simple composite plate depending on the stiffeners position and on
the applied boundary conditions.
81
Chapter 4
82
4.1 Theoretical formulation
In the derivation that follows, the hypotheses of straightness and normality of a trans-
verse normal after deformation are assumed to be no longer valid for the displacement
field which is now considered to be a cubic function in the thickness coordinate and
hence the use of higher order shear deformation theory (HSPT). The composite cir-
cular cylindrical shell is assumed to be composed of Nl layers so that the theory is
sufficiently general. The geometry of the shell and the coordinate system are shown in
Fig. 4.1. The integer k is used as a superscript denoting the layer number where the
numbering starts from the bottom. After imposing the transverse shear stress homo-
. geneous conditions [163, 164J at the top/bottom surface of the shell, the displacements
field in the usual form is given below:
z) 3 4 ( 1 awo (a,/3,t))
v(a,/3,z,t)= ( 1+ R{:J vo(a,/3,t)+z¢{:J(a,/3,t)-z 3h2 ¢(:J(a,/3,t) + A{:J 0/3
83
XI XI
defined as:
(4.2)
(4.3)
(4.4)
with
(4.5)
are referred to as Lame parameters and A~ and A~ are the first fundamental magni-
tudes of the first fundamental form dD k . Attention here is focused on shells with a
constant curvature, i.e. , doubly-curved shells (cylindrical, spherical, toroidal geome-
•
tries) for which A~ = A~ = 1. In the case of a shallow shell the approximations
84
(1 + ~O<) ~ 1 and (1 + ~f3) ~ 1 are generally valid and leaa to reasonably reli-
able results. The strain-displacement relationships with reference to an orthogonal
curvilinear coordinate system give rise to the following deformation field [54, 164]:
where
1
k fJz = C2
(
<PfJ wO)
+ {)(3
. (4.7)
The stress-stain relationship of the k-Iayer is
(4.8)
where t/ is the plane-stress constitutive matrix (see [164]) and uk and c k are the
stress and strain vectors, respectively, and defined as
(4.10)
85
where £k is the Lagrangian for the kth layer of the composite shell. The first variation
can be expressed as
(4.11)
where bU k is the virtual strain energy, bT k is the virtual kinetic energy, and assume
the following form.
where stresses (uk) and strains (gk) are defined in Eq. (4.9) and 1] is the displacement
vector given by
1] = { uv w }T (4.13)
pk denotes mass density while an over dot denotes differentiation with respect to
time. The symbol 8 represents the variational operator. Imposing the condition in
Eq. (4.10), the equations of motion are derived as follows:
8uo:
(4.14)
86
The natural boundary conditions are
8uO:
8vO:
8wO:
(4.15)
8¢0I: Mn = Mn + Ci Pn
8¢(3: Mns = Mns + Cl Pns
8wO,0I: Pn = Cl Pn
8wo,(3: Pns = Cl Pns
where
and Cl =- 3 h2, C2 = - ;2' Note that, nOi. and n(3 are the direction cosines of the unit
normal on the boundary of the laminate. The' force and moment resultants appearing
in Eqs. (4.14) and (4.15) are defined as follows:
Nt .zk+l
~ lk
Nt zk+l
87
(4.18)
and
zk+l Nt
(IO,!1, h,!3,!4, h) = 1k 2: l (1 Z, Z2, z3, Z4, z6) dz (4.19)
k=1 z
The resultants in Eq. (4.17) can be expressed in terms of strain components
(Aij,Bij,Dij,Eij,Fij,Hij) = 2:1
Nt
k=1 zk
zk+l
Ci~ (1
'
Z, Z2, z3, Z4, z6) dz with i,j = 1,2,6
k=1 zk
C~ (1 Z2, Z4) dz with i,j = 4,5
(4.21)
{N} = {Naa, N(3(3, Na(3}; {M} = {Maa, M(3(3, Ma(3}; {P} = {Paa, P(3(3, Pa(3};
{Q} = {Qaz' Q(3z}; {K} = {Kaz' K(3z}
(4.22)
88
For cross-ply stacking sequences an exact solution can be sought in Navier's or Levy's
form. For such laminates Cf6 = 6~6 = C!5 = O. Moreover, if the stacking sequence is
symmetric the coupling elastic coefficients Bij and Eij (see Eq. (4.21)) are zero.
00 00
<PfJ(a,,B,t) = L <I>fJm(a)eiwtcos(O,B)
m=1
£ll = A66 a
2
- All V; - fo w 2 ; £12 = (A12 + A 66 ) a Va;
£13 = (-All ~a - A12 ~fJ) Va;
89
£14 = £15 = £24 = £25 = £41 = £42 = £51 = £52 = 0;
where Va. = dda.' C1 and C2 have already been given previously, and A ij , B ij , Gij , D ij ,
E ij , Fij , Hij have already been defined in Eq. (4.21). Expanding the determinant of
the matrix in Eq. (4.25) the following ordinary differential equation is obtained:
(4.27)
where
(4.28)
90
Using a trial solution eA in Eq. (4.27) yields the following auxiliary equation:
(4.29)
The polynomial coefficients aj with j = 1, ... ,7 have been derived in symbolic form.
The details are not reported here for brevity. Numerically some roots of Eq. (4.29)
may turn out to be complex. As complex roots occur in conjugate pairs, the asso-
ciated constant coefficients will also occur in conjugate pairs. The solution of the
system of ordinary differential equations can thus be written as:
12 12 . 12
I Um (a) = L~ e AiQ
; Vm (a) = L@i e~iQ; Wm (a) = L'Gi e
AiQ
;
i=l i=l, i=l
, 12 (4.30)
<T> f3 (a) /= L @i e Ai Q
i=l i=l
where ~, @i, 'Gi, ~A, 0i with i = 1, ... ,12 are integration constants.' Note that as a
results of Eq. (4.25) the sets ofthe constants are not all independent. Thus a set of
only twelve independent constants can be chosen and they can be related to the other
sets. This choice is completely arbitrary. In the present case the set, !!)i, is chosen to
be the base. By substituting Eqs. (4.30) into (4.25) the following relationships can
be obtained:
(4.31)
where
Ai (AI2Q2 Re> +All (-A22Q 2 + Iow 2 + A66A;) Re> + A22A6602 R/3 + A12 (A66Q2Re> + Iow 2 R/3 + A66A; R/3))
'i = (A2202 - Iow 2 ) (A66 Q2 - Iow 2 ) + ((AI2 - AllA22 + 2A12A66) Q2 + (All + A66)Iow2) A;
+ AllA66A{
(4.32)
8i =
Q (-(A12+ A66)A;(AllRe> + A12R/3) - (A66 Q2 - Iow 2 - AllAn (A12Re> + A22R/3))
(A12QAi + A660Ai)2 - (A66 Q2 - Iow 2 - AllA;) (-A2202 + Iow 2 + A66An
(4.33)
91
Xi = -( - a 2(D12 +D66 +cl(2FI 2 +2F66 +q(HI2 +H66)))2Af + ( - A55 -c2(2D55 +C2F55) - a (D66 +q(2F66
2
2
+ qH66)) + (h + q(214 + qI6))w2 + (Dll + cl(2Fll + C1Hll))At) (A44 + c2(2D44 + C2F44) + a (D22
+ q(2F22 + qH22)) - (12 + q(2h + qh))w 2 - (D66 + q( 2F66 + qH66))At))/ ( - Ai (A44 + c2( 2D 44
+ C2F44) + a 2 (D22 + q (2F22 + qH22)) - (h + q(214 + qI6))w2 - (D66 + q (2F66 + qH66))At) (A55
+ 2C2D55 + C~F55 + Cl 2
(a 2 (FI2 + 2F66 + cl(H l 2 + 2H66)) - (14 + cII6)W 2 - (Fll + c1Hll)A;) ) + a (D12
+ D66 + q(2H2 + 2F66 + q(H12 + H66)))Ai (A44 + 2C2D44 + C~F44 + q (a 2 (F22 + qH22) - (14 + qh)w 2
- (H2 + 2F66 + Cl (H12 + 2H66))At)))
+ cl(H l 2 + H66)))2 At ( - A55 - c2(2D55 + C2F55) + Cl ( - a 2(H2 + 2F66 + cl(Hl2 + 2H66)) + (14 + clh)w 2 + (Fll
+ qHll)At)) )/( - Ai (A44 + c2(2D44 + C2F44) + a 2 (D22 + q(2F22 + cIH22)) - (h + Cl(214 + CIh))w 2
- (D66 + cl(2F66 + c 1 H66))At) (A55 + 2C2D55 + C~F55 +q (a 2 (Fb + 2F66 + q(H12 + 2H66)) - (14 + q I6)w 2
- (Fll + CIHll)At)) + a 2 (D12 + D66 + q(2H2 + 2F66 + q(H12 + H66)))Ai (A44 + 2C2D44 + ·C~F44 + Cl (a (F22
2
+ qH22) - (14 + q I 6)w 2 - (F12 + 2F66 + cl(H l 2 + 2H66)) At) )) - (Ai (A55 + c2(2D55 + C2F55) + a 2 (D66 + q(2F66
+ c 1H66)) - (12 + cl(2i4 + qh))w 2 - (Dll + q (2Fll + qHll))At) (A44 + c2(2D44 + C2F44) + a 2 (D22 + q (2F22
+ qH22)) - (12 + q(2I4 + cII6))W 2 - (D66 + cl(2F66 + qH66))At) ( - A55 - c2(2D55 + C2F55) + Cl ( - a 2 (F12
2 2
+2F66 +cl(HI2 +2H66)) + (14 +qh)w + (Fll +qHll)At)))/( - Ai (A44 + c2(2D44 + C2F44) + a (D22
2
+ q(2F22 + qH22)) - (12 + q(214 + cII6))W - (D66 + cl(2F66 + c1H66))At) (A55 + 2C2D55 + C~F55 + Cl (a 2 (FI2
+ 2F66 + q(H 12 + 2H66 )) - (h + cII6)W 2 - (Fll + qHll)At)) + a 2 (D12 + D66 + q(2F12 + 2F66 + q(H 12
+ H66)))Ai (A44 + 2C2D44 + C~F44 + Cl (a 2 (F22 + c1H22) - (14 + qI6)w2 - (H2 + 2F66 + q(H12 + 2H66))At))))
with i = 1,··· ,12. When Eqs. (4.31) are substituted into Eqs. (4.30) a solution in
terms of only twelve integration constants is obtained as follows:
12 12
Urn (a) = L 'Yi 'i!ii e
Aia
= L 'Yi Xi ~i e
Aia
i=1 i=1
12 12
Vrn(a) = L6i 'i!ii e Aia
= L6i Xi ~i e Aia
i=1 i=1
12 12 12
Wrn(a) = LXi ~i <Pa (a) = L ~i <PiJ (a) = L <Pi ~i
i=1 i=1 i=1
(4.34)
92
The expressions for the generalized forces can then be found in the same way by
substituting Eqs. (4.34) into Eqs. (4.15). Thus, retaining the terms according to the
symmetric cross-ply composite shallow shells the following expressions are derived:
00 12 12
Naa (X, (3) = L [An L Ai Ii Xi !!Ji e Ai
a - A12 eL Di Xi !!Ji e Ai a
m=l i=l i=l
12 12
+ ~12 L Ai Xi !!Ji e Ai a+ ~n L Ai Xi !!Ji e a] sin(e(3) = Naa sin( e(3) Ai
f3 i=l . a i=l
93
00 12 12 12
00 12 12 12
M~~ (a, (3) = L [D12 L Ai ¢i ~i'eA;o. + D66 L ~i eAio. + D66 L ¢i ~i eAio.
m=l i=l i=l i=l
12 12 12
+ D12 L Ai ¢i ~i eA;o.,+ 2 F66 L ~2 ~i eA;o. + 2 F66 L C2 ¢i ~i eA;o.
i=l i=l i=l
12 12 12
+ 2 F66 L C2 Ai Xi ~i eA; 0. + H66 L C~ ~i eAi 0. + H66 L c~ ¢i ~i e A
; 0.
(4.35)
00 12 12 12
Po.o. Ca, (3) = L [Hll L C~ Ai ~i e A; 0. + Hll L c~ A; Xi ~i e A; 0. Fn L C2 Ai ~i e A; 0.
Zero boundary conditions are then generally imposed to eliminate the constants in
the classical method in order to establish the frequency equation for a single shell ele-
ment. By contrast, the development of the dynamic stiffness matrix entails imposition
of general boundary conditions in algebraic form which has much wider implications.
Thus in order to develop the dynamic stiffness matrix the following boundary condi-
tions are applied:
94
Generalized displacements
a=O
a=L
(4.36)
Generalized forces
a=O
(4.37)
a=L
The imposition of Eq. (4.36) into Eq. (4.34) leads to the following relationship:
t5=AD (4.38)
where
t5={ Um1 Vm1 W m1 <PQ1 <P /31 Wm1,Q Um2 Vm2 W m2 <PQ2 <P /32 W m2 ,Q
}T
D={ ~l ~2 ~3 ~4 ~5 ~6 ~7 ~8 ~9 ~10 ~ll ~12 }T
(4.39)
95
and
Atl A~l A~l Ail Afl A~l Ail A~l A~l Alo
11
All
11
A12
·11
At2 A~2 A~2 Ai2 Af2 A~2 Ai2 A~2 A~2 A 12 12 A1212
lO All
where:
Aio = e(a Ai), Ail = (_1)i+l ¢i e(a Ai), Ai2 = Xi Ai e(a Ai)
(4.41)
with i = 1,··· ,12. By applying the same procedure for boundary conditions of
the generalized forces and thus imposing Eq. (4.37) into Eq. (4.35), the following
relationship is obtained:
F=RD (4.42)
96
where
F= {NOI.OI.l Nf3f31 QOI.l MOI.OI.l MOI.f31 P ocq NOI.0I.2 Nf3f32 Q0I.2 MOI.0I.2 MOI.f32 P"''''2 } T
D= { ~l ~2 ~3 ~4 ~5 ~6 ~7 ~8 ~9 ~IO ~Il ~12 }T
(4.43)
and
R~o R~o R~o Rio Rfo R~o Rio R~o Rio RIO
10
Rll
10
Rl2
10
R~l R~l R~l Ril Rfl R~l RIl R~l Ril RIO
11
Rll
11
R12
11
R~2 R~2 R~2 Ri2 Rf2 R~2 RI2 R~2 Ri2 RIO 12 Rl2
12 Rll 12
(4.44)
with
+ ci Xi Hll A~;
97
R~ = (_1)Hl (a Cl Xi F12
2 + a 2 ci Xi H12 + a D12 cPi +2a Cl F12 cPi +a ci H12 cPi
R~ = (_1)Hl ( - A12 a Xi lSi e(a Ai) + Au Xi e(a Ai) 'YiAi + A12 Xi e(a Ai) ~(3
R~ = e(a Ai) (A55 + C2 (2 D55 + C2 F55 )) + 2 a 2 Cl e(a Ai) F66 + 2 a 2 ci e(a Ai) H66
+ Xi e(a Ai) (A55 + C2 (2 D55 + C2 F55 )) Ai + a 2 ci Xi e(a Ai) H12 Ai + 4 a 2 ci Xi e(a Ai)
x H66 Ai +a Cl e(a Ai) F12 cPi Ai +2a (;1 e L Ai F66 cPi Ai +a ci e(a Ai) H12 cPi Ai
+2a ci e(a Ai) H66 cPi Ai - Cl e(a.Ai) Fu A; - ci e(a Ai) Hu A; - ci Xi e(a Ai) Hu Ar;
Rio = (-1) HI ( - a 2 Cl Xi e(a Ai) F12 - a 2 ci Xi e(a A;) H12 - a D12 e(a Ai) cPi
- 2 a Cl e(a Ai) F12 cPi - a ci e(a Ai) H12 cPi + Du e(a Ai) + 2 Cl e(a Ai) Fu Ai
+ ci e(a Ai) Hu Ai + Cl Xi e(a A;) (Fu + Cl Hu) A;);
Ril = a D66 e(a Ai) +2a Cl e(a Ai) F66 +a ci e(a Ai) H66 +2a Cl Xi e(a Ai) F66 Ai
+2a ci Xi e(a Ai) H66 Ai + e(a Ai) (D66 + Cl (2 F66 + Cl H 66 )) cPi Ai;
Ri2 = (-1 )Hl (Cl (_a 2 Cl Xi e(a Ai) H12 - a e(a Ai) F12 cPi - a Cl e(a Ai) H12 cPi
(4.45)
with i = 1,··· ,12. Now by eliminating the constants vectors D from Eqs. (4.38)
and (4.42) the dynamic stiffness matrix which links the forces and moments vector
F with the generalized displacements vector ~ is derived:
F=K~, (4.46)
98
i.e.
K}1 K}1 I(~ 1(4 I(f 1(61 1(71 I(~ 1(91 1(10
1
1(11 1(12
1 1
1
1(1 I(~ I(~ I(~ I(~ I(g 1(72 1(82 1(92 1(10 1(11 1(12
2 2 2 2
I(~ I(~ I(~ I(j 1(3 1(63 1(7 1(8 I(~ I(~o 1(11 1(12
3 3 3 3
I(! 1(2 1(34 1(4 1(5 I(~ I(I I(~ I(~ I(!O 1(!1 1(!2
4 4 4
1(1 1(2 1(3 1(4 1(5 1(6 1(7 1(8 1(91(10 1(11 1(12
5 5 5 5 5 5 5 ·5 5 5 5 5
1(1 1(2 1(3 1(4 I(~ I(~ 1(7 1(8 1(91(10 1(11 1(12
6 6 6 6 6 6 6 6 6 6
K= (4.47)
1(1 1(2 1(37 1(4 1(57 1(67 1(7 1(87 1(91(10 1(11 1(12
7 7 7 7 7 7 7 7
1(1 1(2 1(3 1(4 1(5 1(6 I(~ I(~ 1(91(10 1(11 1(12
8 8 8 8 8 8 8 8 8 8
1(1 1(2 1(3 1(4 1(5 1(6 1(7 1(8 1(91(10 '1(11 1(12
9 9 9 9 9 9 9 9 9 9 9 9
I(to I(io I(~o I(io I(fo I(~o I(io I(~o I(io 1(10
10
1(11 1(12
10 10
I(tl I(il 1(~1 I(il I(fl 1(~1 I(il 1(~1 I(il 1(10 1(11. 1(12
11 11 11
I(t2 l(i2 1(~2 l(i2 I(f2 1(~2 l(i2 1(~2 l(i2 1(10 1(11 1(12
12 12 12
The above dynamic stiffness matrix will now be used in conjunction with the Wittrick-
Williams algorithm [63] to analyze assemblies of laminated composite cylindrical and
spherical shallow shells to investigate their free vibration characteristics based on
HSDT. Explicit expressions for each shell element of the DS matrix were obtained
via symbolic computation using Mathematica®. They are far too large and lengthy
to report in this thesis. The correctness of these expressions was further checked by
implementing them in a MATLAB® program and then carrying out a wide range of
numerical simulations.
Once the DS matrix of a laminated composite element has been developed, it can be
assembled to form the global DS matrix of the final structure (see Fig 4.2). Although
like the FEM, a mesh is required in the DSM, but it should be noted that the latter
is mesh independent in the sense that additional elements are required only when
99
Global dynamic
stiffness matrix
.. l..aral
Igl ~ I· I_uu :
,,
i"-------
''
there is a change in the geometry of the structure. With a single DS laminate ele-
ment it is possible to compute any number of its natural frequencies to any desired
accuracy, which, of course, is impossible in the FEM. However, for the type of struc-
tures under consideration DS shell elements do not have point nodes, but have line
nodes instead. In this particular case, no change in geometry along the longitudinal
direction is admitted. This is in addition to the assumed simple support boundary
conditions on two opposite sides, inherent in DSM for shell elements at present. The
other two sides of the shell can have any boundary conditions. The application of
the boundary conditions of the global dynamic stiffness matrix involves the use of
the so-called penalty method. This consists of adding a large stiffness to the appro-
priate leading diagonal term which corresponds to the degree of freedom of the node
that needs to be suppressed. It is thus possible to apply free (F), simple support (S)
and clamped (C) boundary conditions on the structure by penalizing the appropriate
degrees of freedom. Clearly for simple support boundary condition, the generalized
100
displacement amplitudes Vi, Wi and <I>.Bi are assigned zero values. On the other hand,
for clamped boundary condition Ui , Vi, Wi, <I>O:i' <I>.Bi and Wi,o: on the boundary are
all assigned zero values. Of course for the free-edge boundary condition stress resul-
tants are assigned zero values and then no penalty will be applied at the generalized
displacement amplitudes. Because of the similarities between DSM and FEM, DS
elements can be implemented in FEM codes to enhance the accuracy of results in
FEM considerably.
Using the above theory free vibration characteristics of laminated composit~ cylin-
drical and spherical shallow shells with symmetric cross-ply lamination schemes are
now numerically investigated. Results have been obtained in an exact sense by using
the Levy-type closed form solution within the framework of the DSM. The material
properties used in all of the analysis carried out are
unless otherwise stated. At first a validation exercise is undertaken and the results
are shown in Table 4.1, the fundamental circular frequency parameter of three-layer
symmetric cross-ply and moderately thick spherical and cylindrical shallow shells is
shown. The results obtained by using the present DSM formulation have been com-
pared with those proposed by Khdeir and Reddy [85] by using the classical Levy-type
solution. As can be seen in Table 4.1 the results match almost exactly for all of
the considered boundary conditions. In Tables 4.2 and 4.3 further assessments of the
fundamental natural frequency have been carried out. Most notably, the fundamental
circular frequency parameter has been computed for three and four-layer symmetric
cross-ply and moderately thick spherical and cylindrical shallow shells, respectively.
101
Table 4.1: Fundamental circular natural frequency parameter w= w~ /ii, of
square cylindrical and spherical shells with stacking sequence [0 0 /90 0 /OO]length-to-
thickness ratio a/h = 10 and radius-to-length ratio R(3/a = 20.
Cylindrical shell
Khdeir and Reddy [85] 11.793 13.825 3.789 15.999 4.322 6.089
Present H~DT DSM 11.793 13.825 3.789 15.999 4.322 6.089
Results are compared with several meshless formulations proposed by Ferreira el.
RfJla
5 10 20 50 100 Plate
Cylindrical
HSDT RBF [54J 11.851 11.808 11.797 11.794 11.793 11.793
FSDT WLC [55J 12.214 12.176 12.166 12.163 12.163 12.163
SSDT RBFC [57J 11.923 11.915 11.913 11.912 11.912 11.912
HSDT DSM 11.846 11.804 11.793 11.790 11.790 11.790
all [54, 55, 57], in particular a radial basis formulation (RBF) based on a HSDT,
a wavelet collocation (WLC) formulation based on a FSDT and a radial basis col-
location (RBFC) solution procedure based on a sinusoidal shear deformation theory
(SSDT) have been used. As can be seen in Tables 4.2 and 4.3 the present formulation
leads to results which are in an excellent agreement with those obtained by using the
other theories, for different values of the radius-to-length ratio (R/a). Results of Ta-
102
Table 4.3: Dimensionless fundamental circular frequency parameter w = W a: ft, of
square spherical and cylindrical shells with stacking sequence [0 0 /90 0 /90 0 /0 0 ] ' length-
to-thickness ratio a/h = 10 and varying the radius-to-Iength ratio R(3/a and R/a.
Rrda
5 10 20 50 100 Plate
Cylindrical
HSDT RBF [54] 11.838 11.794 11.783 11.780 11.779 11.779
FSDT WLC [55] 12.279 12.240 12.230 12.228 12.227 12.227
SSDT RBFC [57] 11.901 11.887 11.884 11.883 11.883 11.883
HSDT DSM 11.832 11.790 11.780 11.777 11.777 11.776
bles 4.4 to 4.7 focus on symmetric cross-ply cylindrical shells for which two different
stacking sequences are investigated. Both moderately thick and thin shallow shells
have been taken into account. The first five circular natural frequency parameters
are computed for several boundary conditions. In Figs. 4.3, 4.4 and 4.5 the first six
mode shapes of sese, SFSC and SFSF square cylindrical shallow shells are depicted.
The mode shapes correspond to a symmetric cross-ply [00 /90 0 /0 0 ] stacking sequence
with a/h = 100 and R/a = 30. An assessment of the present formulation is carried
out in Table 4.8 for isotropic spherical panels. The first six natural frequencies
are computed and compared with the 3D elasticity solution and many other theories,
which are refined in the displacement field or in the curvature description. As can
be seen from Table 4.8 the proposed DSM formulation provide the best accuracy for
the fundamental natural frequency with respect to the 3D elasticity solution, but
for higher frequencies the error increases (the maximum error incurred is well within
1.5%). This marginal loss of accuracy is due to the fact that approximated curvature
descriptions have been employed in the analysis. In Table 4.9 the first three circular
frequency parameters of three-layer symmetric cross-ply shallow spherical shells are
103
(a) Mode 1, W = 31.916 (b) Mode 2, W = 36.277
Figure 4.3: First six mode shapes of a symmetric cross-ply cylindrical shell with sese
boundary condition.
104
(a) Mode 1, is; = 7.031 (b) Mode 2, is; = 17.257
Figure 4.4: First six mode shapes of a symmetric cross-ply cylindrical shell with SFSC
boundary condition.
105
(a) Mode 1, W = 3.919 (b) Mode 2, W = 6.334
Figure 4.5: First six mode shapes of a symmetric cross-ply cylindrical shell with SFSF
boundary condition.
106
Table 4.4: First five circular natural frequency parameters w= w ~ #' of square
cylindrical shells with stacking sequence [0°/90° /OO]length-to-thickness ratio a/h =
10 and varying the radius-to-Iength ratio Rf3/a.
Table 4.5: First five circular natural frequency parameters w= w ~ #' of square
cylindrical shells with stacking sequence [0°/90° /OO]length-to-thickness ratio a/h =
100 and varying the radiu8-to-Iength ratio Rf3/a.
107
Table 4.6: First five circular natural frequency parameters w= w~ It, of square
cylindrical shells with stacking sequence [0°/90°/90 %
°] length-to-thickness ratio
a/h = 10 and varying the radius-to-Iength ratio Rf3/a.
Table 4.7: First five circular natural frequency parameters w= w ~ It, of square
cylindrical shells with stacking sequence [0°/90°/90 °] length-to-thickness ratio
%
108
Table 4.8: First six natural frequencies w~n rad/ s, of square isotropic spherical shells.
Natural frequencies
Theory Wl (1,1) W2 (2,1) W3 (1,2) w4(2,2) W5 (3,1) W6 (1,3)
I-" CST [174J 0.53263 ( -1.370)t 0.59041 ( -1.063) 0.59080 ( -1.014) 0.68486 (-1.197) 0.76020 ( -1.065) 0.76260 (-1.383)
0 FSTD [175J 0.50211 (+4.438) 0.56247 (+3.720) 0.56248 (+3.828) 0.65706 (+2.911) 0.73915 ( +1.734) 0.74035 (+1.575)
CD
HSTD [175J 0.50223 (+4.415) 0.56276 (+3.670) 0.56277 (+3.779) 0.65788 (+2.790) 0.73966 (+1.666) 0.74081 ( +1.514)
FSTD (DT+)[176J 0.52864 (-0.611) 0.58954 ( -0.914) 0.58954 (-0.798) 0.68370 ( -1.025) 0.75974 ( -1.004) 0.75974 ( -1.002)
FSTD (ST' )[176J 0.52830 ( -0.546) 0.58853 (-0.741) 0.58853 (-0.626) 0.68232 (-0.822) 0.75818 ( -0.796) 0.75818 (-0.795)
HSDT DSM 0.52795 ( -0.480) 0.58899 (-0.820) 0.58982 ( -0.846) 0.68562 (-1.309) 0.75989 ( -1.023) 0.76032 ( -1.080)
Material and geometric properties: a - b - 1.0118, h - 0.0191, R - 1.91, E - 1, p - 1, II - 0.3
t Error% = w~~~w x 100; j: Donnell approximation; * Sanders approximation
shown. The boundary conditions SSSS, sess and sese are used. The investiga-
tion is carried out for different values of the radius-to-Iength ratio and for moderately
thick and thin spherical panels. In Table 4.10 the effect of the orthotropic ratio on the
dimensionless fundamental frequency parameter is illustrated. The boundary condi-
Table 4.9: First three circular natural frequency parameters w = w ~ It, of square
spherical shells with staking sequence [0°/90 %
°] and varying the radius-'to-Iength
R/a and the length-to-thickness a/h ratios.
tions SSSS, sess and sese are employed and a symmetric cross-ply [0°/90°/90 %
°]
square spherical shallow shell (R/a = 50) is analyzed. As expected the fundamental
natural frequency parameter increases when increasing both the orthotropy ratio and
the length-to-thickness ratio, for all the boundary conditions considered. Finally, in
Fig. 4.6 the first six mode shapes of sese square spherical shallow composite shells
are represented. The geometrical characteristics and the lamination scheme are same
as the ones used for the cylindrical shallow shells shown in Figs. 4.3, 4.4 and 4.5.
110
Table 4.10: Fundamental circular natural frequency parameter w= w~ Ii, of
square spherical shells with staking sequence [0 /90 /90 /0
0 0 0 0
]' R/ a = 50 and varying
orthotropic Ed E2 and the length-to-thickness a/h ratios.
111
(a) Mode 1, W = 32.361 (b) Mode 2, W = 36.905
Figure 4.6 : First six mode shapes of a symmetric cross-ply spherical shell with sese
boundary condition.
112
Dimensionless circular frequency parameters of circular cylindrical shells
Finally some results are proposed in the case of cross-ply circular cylindrical shells.
The results of Table 4.11 show excellent agreement with those available in literature.
In the set of results the dimensionless fundamental circular frequency parameters have
been calculated for a three-layer circular cylindrical shell made up of cross-ply stacking
sequence and material E1 = 40 GPa, E2 = 1 GPa, /112 = 0.25, G 12 = G 13 = 0.6
GPa, G 23 = 0.5 GPa and p = 1500 Kg/m 3 , which will be used in all the numerical
analysis carried out. Results are shown for different boundary conditjons as can
be seen in Table 4.11. Understandably, as expected the non-dimensional frequency
parameter is higher for clamped boundary conditions and it increases, when increasing
the cylinder length L as a results of using L2 in the numerator of the dimensionless
parameter. In Table 4.12 results are shown by setting the half-wave number m =1
and then computing the circular frequencies of one-layer orthotropic simply-supported
circular cylindrical shell for n = 1,2"" ,9. As can be seen results are in excellent
agreement with those reported by other researchers. Comparative results are wide
ranging which include a semi-analytical differential quadrature finite element method
(S-DQFEM), and exact procedure using classical shell theory (CST) and FSDT (see
Table 4.12). Focusing on the same sample problem, the calculation of the circular
frequencies has been extended to higher values of the half-wave number m and the
results are shown in Table 4.13. Some mode shapes corresponding to the results of
Table 4.13 are shown in Figs. 4.7(a) to 4.7(f). They are self-explanatory in terms
of displacement patterns and they reveal some interesting features, and of course,
they conform to the simply-supported (shear diaphragm) boundary condition. Two
different circular cylindrical shells made up of symmetric cross-ply stacking sequences
are now considered. Table 4.14 shows results for several values of the thickness-to-
radius ratio hi R. As can be observed the dimensionless circular frequency parameters
increase when increasing the ratio hi R, for both staking sequences. The introduction
of the 90° plies as outer layers and 0° plies as inner-layers leads to lowering the
dimensionless circular frequency parameter. Results match quite well with those
113
Table 4.11: Dimensionless fundamental circular natural frequency parameter w= w C~; h) It, of circular cylindrical shells
with staking sequence [0°/90 %
°]' R/h = 5 and varying the boundary conditions and the length-to-radius ratio L/ R.
ss se ee Fe
Theory L/R= 1 L/R= 2 L/R= 1 L/R= 2 L/R= 1 L/R= 2 L/R= 1 L/R= 2
..q<
,....,
,....,
HSDT KRF [177J 0.1007 0.1777 0.1087 0.1972 0.1192 0.2191 0.0506 0.0995
FSDT KRF [177J 0.1004 0.1779 0.1036 0.1945 0.1093 0.2129 0.0495 0.0988
CST KRF [177J 0.1479 0.2073 0.1850 0.2662 0.2049 0.3338 0.0669 0.1099
F-DSM [157J 0.0996 0.1722 0.1025 0':1950 0.1083 0.2083 0.0483 0.0914
(c) Mode n = 1, m = 3 with circular fre- (d) Mode n = 2, m = 1 with circular fre-
quency w = 1849.4 rad/ s quency w = 416.20 rad/ s
(e) Mode n = 2, m = 2 with circular frequency (f) Mode n = 2, m = 3 with circular fre-
w = 909.39 rad/ s quency w = 1302.4 rad/ s
115
Table 4.12: Circular frequencies (rad/s) of one-layer orthotropic simply-supported
circular cylindrical shells with m = 1 and varying the half-wave number n, L = 5 m,
R = 1 m, h = 0.01 m, Ex = 120 CPa, Ey = 10 CPa, C xy = 5.5 CPa, Vy = 0.27 and
p = 1700 kg/m 3 .
Theory n
2 3 4 5 6 7 8 9
S-DQFEM [178] 741 416 258 198 209 266 350 452 570
CST GanSiv [179] 743 417 257 194 204 260 344 447 567
CST GreStk [180] 765 430 266 202 211 270 358 464 587
Exact FSDT [178] 741 416 258 198 209 266 350 452 570
Present 741.35 416.20 258.02 197.82 208.81 265.98 349.77 452.00 569.86
available in the literature, particularly with the ones given by Ye and Soldatos [181J
which include a three-dimensional vibration analysi~ and the ones obtained by the
dynamic stiffness formulation based on FSDT, see Thinh and Nguyen [157J and a
classical free vibration analysis accounting for shear deformation effect proposed by
Viswanathan [99J. Finally, in Table 4.15, re~ults for fundamental circular frequency
n m
1 2 3 4 5 6 7 8 9
parameters of circular cylindrical shells for several boundary conditions and length-
to-radius ratio (L/ R) and with varying the radius-to-thickness ratio (R/h) are shown.
Different symmetric cross-ply lamination schemes have been analyzed.
116
Table 4.14: Dimensionless circular frequency parameters W, of simply-supported circular cylindrical shells with symmetric
cross-ply stacking sequence, R/ L = 1 and material Ed E2 = 40, G 12 / E2 = 0.6 G 13 / E2 = G23 / E2 = 0.5. V12 = 0.25.
FSDT.6 [99J Ye-Soldatos [181J F-DSM [157J Present. FSDT.6 [99J Ye-Soldatos [181J F-DSM [157J Present
0.1 0.074118 0.064600 0.06397 0.065543 0.051555 0.052748 0.05310 0.055205
0.076309 0.066335 0.06566 0.066730 0.054261 0.059130 0.05907 0.063499
""""'
""""'
-.J 0.082831 0.079277 0.07888 0.072155 0.062899 0.070738 0.07089 0.071120
4.3 Conclusions
Exact free vibration analysis of laminated composite shallow cylindrical and spher-
ical shells has been carried out by combining the dynamic stiffness method and a
higher order shear deformation theory (HSDT). The effects of several parameters
such as length-to-thickness ratio and radius-to-length ratio, orthotropy ratio, stack-
ing sequence and number of layers on the dimensionless circular natural frequency
parameters have been investigated in details. Results have been compared with those
available in the literature including advanced meshless methods such as radial basis
functions, Ritz and FEM, and the three-dimensional elasticity solution. Different
boundary conditions have been considered. Representative mode shapes have been
presented and discussed.
The investigation has shown that the DSM allows computing all of the natural fre-
quencies of thin and thick laminated composite shallow cylindrical and spherical shells
with high accuracy and low computational cost. The implementation of the HSDT
within the framework of the DSM permits considerable improvements in results ac-
curacy when compared to other formulations based on FSDT.
119
Chapter 5
Conclusions
The research carried out during this PhD thesis has been focused on the investiga-
tion of the free vibration and buckling behavior of laminated composite structures
r
120
significant refinement over FSDT elements, particularly when thick plates with high
orthotropy ratios are analyzed. The boundary conditions do not seem to affect the
error incurred using FSDT as opposed to more accurate HSDT. It has been noted
that the dynamic behaviour of stepped composite plates is very different from that of
simple composite plates depending on the boundary conditions, but significant alter-
ation in mode shapes is possible by using stepped panels. Next, the investigation has
been extended to cover complex stringer panels. Based on the computed results it
is concluded that the dynamic behaviour of stepped composite plates and composite
stringer panels can be very different from those of simple composite plates of similar
size and dimensions. Significant alteration in natural frequencies and mode shapes is
observed. This could be useful in solving frequency attenuation problems. The re-
search is followed up to develop exact dynamic stiffness composite plate elements for
buckling analysis of simple and stepped laminated composite plates. It is concluded
that the boundary conditions affect significantly the buckling modes. Stiffeners, if
properly introduced increase the buckling load considerably. The buckling load of
stiffened composite plates changes remarkably with respect to the simple composite
plate depending on the stiffeners position and on the applied boundary conditions.
Next an exact free vibration analysis of laminated composite circular cylindrical shells
was undertaken using the DSM. The investigation has shown that the DSM allows
computing all of the natural frequencies of thin and thick circular cylindrical shells
with high accuracy and importantly with much less demanding computational cost
which otherwise would be required by the FEM. The implementation of the HSDT
within the framework of the DSM permits considerable improvements in accuracy
in comparison with other DSM formulation based on FSDT. The analysis has then
been extended to laminated composite doubly-curved shallow shells for free vibration
analysis showing significant results for several configurations.
It has been highlighted on several occasions that the DSM represents a valuable tool
to cope with free vibration and buckling problems of composite structures providing
significant enhancements in the accuracy of results when compared to other approx-
imation methods. The advantages of using the DSM can be underpinned by means
121
of several considerations.
For instance, aircraft wings and fuselages can be idealized as assemblies of thin skin
panels reinforced by stringers. These structures are often investigated by using the
finite element method (FEM). One of the main advantages of using the FEM is that
it can handle structures with complex geometry and the results generally converge
to the exact ones with increasing number of elements. However, FEM is an approx-
imate method, requiring considerable computational power and modelling efforts.
The accuracy of the FEM solution cannot be always guaranteed. A better alterna-
tive method for free vibration analysis as demonstrated in this thesis, is the dynamic
stiffness method (DSM). In the FEM, the number of natural 'frequencies that can be
computed is restricted to the number of chosen degrees of freedom of the strllcture
and the accuracy of results diminishes with higher order modes. This can be a serious
limitation in modal analysis. By contrast, the DSM has no such limitation and any
number of natural frequencies can be computed to any desired accuracy by using the
DSM without the need to increase the number of elements to achieve higher accuracy.
Furthermore, when fast iterative matrix solvers are used, the DSM will be much more
efficient than the FEM. The DSM is particularly effective when it is used in conjunc-
tion with the Wittrick-Williams algorithm. The DSM gives exact results because
the equations of motion are solved in closed analytical form to obtain the element
properties and no approximation is made en route during the analysis. Results and
conclusions presented in this thesis are significant in this context.
The proposed advanced formulations have been entirely developed by harnessing the
numerical and symbolic computational power of Matlab® and Mathematica® soft-
were.
122
Chapter 6
The research carrie4_ out in this thesis has open up potential possibilities to pave the
way for future work. The proposed dynamic stiffness formulation can be extended,
with considerable efforts,' to cover the following topics:
123
that are characterized by a smooth and continuous variation of properties through the
thickness of the structure, i.e. from the bottom to the top layer. The FGM has lot of
advantages over other types of advanced materials like fibre-reinforced composites in
that problems like delamination, fibre failure, adverse hygroscopic effects due to mois-
ture content etc are effectively eliminated or non-existent. The properties of FGM
are controlled by the variation of the volume fraction of the constituent materials
which can be suitably adapted to become ultrahigh temperature-resistant, and thus
making them appropriate materials for aerospace applications, such as aircraft, space
vehicles, barrier coating and propulsion systems. Commonly used laminated com-
posite materials such as carbon or glass fibre reinforced plastics are characterized by
. ~a strong discontinuity of mechanical properties across the interfaces of layers (plies)
consisting of fibres and resins, and because of the low ratio of the transverse shear
modulus to the in-plane elastic modulus, the failure due to delamination can occur
even at a relatively low load level. It is well known that fibre-reinforced composites
have low shear strength due mainly to their low shear moduli. In some circumstances,
this serious drawback can be overcome by employing FGM which no doubt has much
superior shear strength and importantly, it has the ability to withstand harsh tem-
perature environment. Against this background, future work could be focused on the
development and application of the dynamic stiffness method (DSM) for plates and
shells made of functionally graded materials with the important inclusion of thermal
effects and higher-order shear deformation theory (HSDT) to investigate the free vi-
bration characteristics of aerospace structures. The proposed DSM formulations will
be more accurate and computationally efficient than conventional FEM and other
approximate methods.
Another possible extension of the work carried out in this thesis could be based
on piezo-electric material applications. Piezoelectric materials are able to generate
positive or negative electric charges when subjected to mechanical pressure which
124
can be taken advantage of in structural design work. When dealing with piezoelectric
materials, a charge is generated when molecular electrical dipoles are caused by a
mechanical loading: that is, the direct effect (sensor configuration). Conversely, when
an electric charge is applied, a slight change in the shape of the structure occurs:
that is, the inverse effect (actuator configuration). It has been demonstrated that
piezoelectric materials can be used at the same time as actuators and sensors, resulting
in the so-called self-sensing piezoelectric actuator. Piezoelectricity is a feature of
some natural crystals (such as quartz and tourmaline) or synthetic crystals (lithium
sulfate), and several kinds of polymers and polarized ceramics. The most common
piezoelectric materials are the piezoceramic barium titanate (BaTi03) and piezo lead
zirconate titanate (PZT). The application of piezoelectric materials leads to the well-
known smart structures. Smart structures have been used in sensing, actuating,
diagnosing, and assess,ingthe health of structures, depending on the external stimuli.
Sensors and actuators should be integrated into the complete structures and this leads
to unusual design solutions, compared to traditional structural design solutions. In
the most advanced design concepts, smart structures could have the ability to save
and analyze information in order to perform a learning process. Nowadays, smart
structures are applied in many different domains, but they all share the common
feature of having a highly cross-disciplinary design. Among other applications, the
following are highly pertinent:
2. Vibration control
3. Shape morphing
125
6.3 High frequency wave propagation and response
analysis
126
obtained by CLT, FSDT and HSDT is probably acceptable when macro or overall
behaviour of a structure is sought, but for primary structures that are intended to
carry large loads and are thus thicker, the error will be much higher. Recent research
has shown that for some structures the error can be over 30% on the fundamental
natural frequency. In aerospace industry where safety factors are generally low to
achieve a lighter structure, an error of 30% is unacceptable and it may lead to struc-
tural failures during laboratory or flight tests. For this reason, improved modelling
techniques are to be developed. Novel modelling techniques could indeed, be compu-
tationally demanding bt1-t, on the other hand, they deliver the much needed accuracy.
One important recent development in this area is the so-called Layer-Wise technique
in which each single layer is modelled as an individual plate by using appropriate
displacement assumptions and adopting an assembly procedure.· The main drawback
of the LW model using FEM is that it leads to a large number of unknowns which
depends on the number of layers. Clearly, very many finite elements are required to
model a structure, and thus the number of DOF for a LW FE model may become
excessive, making the use of conventional LW theory practically impossible. A major
break-through would be to use LW theory in conjunction with the dynamic stiffness
method (DSM) to make this application realistically possible. For free vibration of
plate assemblies, Dynamic stiffness (DS) elements based on CLT, FSDT and HSDT
have already been developed as is evident from the literature, showing huge superi-
ority over conventional finite elements. One of the potential benefits of the DSM in
sharp contrast to FEM, is that one single element is enough to model any part of the
structure with uniform geometry in an exact sense without losing any accuracy, and
thus reducing the number of unknowns of a given problem drastically. This is pos-
sible in DSM because instead of discretising the structure, the differential equation
of motion is solved in closed-form and the solution is generalised to develop element
properties which can then be rotated, offset, assembled to model a complex structure
such as wing boxes. Thus the DSM can make the application of a LW theory feasible
when investigating real composite structures. With particular reference to composite
. shell structures, DSM coupled with layer-wise theory is surely a potential candidate.
127
Appendix A
The geometric relation for a lamina in the local or lamina reference system can be
written as:
[
::: ] [:x ~y ~
"Yxy
'"Yyz
Vy
0
1)x
'D z
0
Vy
j (1)
"Yxz 'D z 0 'Ox
(2)
OJ['' j
C'2 0 0
[:.q
T23
T13
[ C'2
Cu
0
0
0
C22
0
0
0
C66
0
0
0
C44
0
0
0 C55
o
o
o
<2
')'12
')'23
')'13
(3)
where the Gij are expressed in terms of stiffness coefficients Gij , as:
(4)
128
where E1 is the elastic modulus in the fibre direction, E2 the elastic moduli in per-
pendicular to the fibre, Zl12 and Zl21 = Zl12 Ed E1 the Poisson's ratios, G 12 = G 13 and
G 23 the shear modulus of each single orthotropic lamina. If the lamina is placed
at an angle 0 in the laminate or global reference system, the equation needs to be
transformed as follows:
- - 4
C u =CuC + 2(C- 12 + 2C- 66 )S22 -
C + C 22 S
4
- -
C 12 =(C11 + C- 22 - -
4C66 )S C
22
+ C- 12 (S 4+4
C)
- -
C 22 =C11 S
4
+ 2(C- 12 + 2C- 66 )S22 -
C + C 22 C
4
-
C 26 =(C11
-
-
-
C 12 -
- 3
2C66 )S C + (C12
-
-e- 22 + 2C- 66 )SC3 (5)
- - 2
C 55 =C44 S + C- 55C2.
C 45 =(055 - 044)CS
where C = cos (0) and S = sin (0). This leads to the constitutive equation for the
k-th lamina in the laminate or global reference system:
O"xx 0 11 0 12 0 16 0 0 Cxx
O"yy 0 12 C 22 0 26 0 0 Cyy
Txz 0 0 0 0 45 0 55 "/xz
129
Appendix B
The polynomial coefficients for out-of-plane and in-plane cases are following defined:
a2 = -c~ (A44 + 0,2 D22) Fi1 'I- Ass Dll (D66 + C1 (2F66 + c1 H66)) + c~ (Dll D66 FSS + 2 CJ Dll FSS F66
+ c~ (-Fi1 F44 + Dll F44 Hll + Dll FSS H66)) + 2 c2 (-c~ D •• Fi1 + Dll (c~ D44 Hll + Dss (D66
+ 2 CJ F66 + C~ H66))) + c~ (A •• D1; Hll - 0,2 (D~2 Hll - 2 D12 (Fll (F12 + 2 F66 + CJ H12 + 2 CJ H66)
- Hll (D66 + c1 (F12 - c1 H66))) + c1 (Fi1 (2 F22 + CJ H22) - 2 Fll (Fi2 - 4 Fi6 + c1 (F12 + 2 F66) H12
+ 2 D66 (H12 + 2 H66)) + Hll (CJ (F12 + 2 F66)2 + 4 D66 (F12 - CJ H66))) + Dll (Fi2 + 4 Fi6 - D22 Hll
- 2 D66 (H12 + 2 H66) + 2 F12 (2 F66 + CJ (H12 + 2 H66)) + CJ (-2 F22 Hll + q (H~2 - Hll H22 + 2 H12
H66)))) - (Hll (Dll + D66 + 2 CJ F66 + c~ H66) 12 + 2 q DllHll I. - 2 Fll (D66 + c1 (2F66 + CJ H66)) I.
044 D66Hll +204c~ 022 F12 H11 - 204c~ 012 F22 Hll - 40:4c~ 066 F22 Hll +20
4
c1 F12F22Hll - 2o? cic~ 012F44 Hll
066H12 - 2Q4C~ D22 Fu H12 +40'4c~ D12F12 H12 +80'4c~ D66 F12 H12 - 20:
4
c1 F;2 H12 - 204c~ 011 F22 H12 - 20:
4
c1
4 4
F11 F22 H12 - 8a c1 F12 F66 H12 - 80 c1 Fi6H12 +20 c1 D12 H'f2 +4a ci D66H'f2 - 2O' cic2 Dll D55 H22 -
4 4 2
o4 c i Du
4
066H22 - 20'4c~ D12 FI1 H22 - 40'4c~ 066Fll H22 + 2Q4C~ Dll F12 H22 +2o: c1 Fl1 F12H22 - o:2cic~Dll F55 H22 + 2
4 4
a4c~
4
Dll F66 H22 + 40 c1 Fu F66H22 - 20: c1 D12 H11 H22 - 40: c1 D66 H11 H22 - A44 0'2 ci DII H66 +40:4c~ D~2 H66
4 2
- 4 Q
2
ci Dll D22 H66 - 20. ci c2 011 D44 H66 - 4o? ci C2 D12 D55 H66 + 80
4
c~ D12 D66 H66 - 8 0 c~ c2 055 D66 H66 - 4
04c~ D22 FI1 H66 + 804c~ DI2 F12 H66 + 1604c~ D66F12 H66 _4Q4C~ Dll F22H66 - 20
4
c1 Fl1 F22 H66 - 0:2cic~ D11 F44
130
H66 - 2o:2C~C~ D12 FSS H66 -4o:2C~c~D66 FS5 H66 - o:4 c 1 D22H11 H66 +40
4
C1 D12 H12H66 +8o:
4
c1 D66H12 H66 _ 0
4
c1 Dl1 H22 H66 + A55 ,,2(D~2 - Dl1(D22 + q(2F22 + q H22» +2D12(D66 + q(F12 - q H66» + q(q(F12 + 2F66)2
+ 4 D66( F12 - q H66») + (D66( A55 + c2(2 D55 + C2 F55» 12 + 2 q (D66 Fl1 10 + (A55 + C2(2 D55 + c2 F55» F66 12)+
C~( D66 Hl1 10 - ,,2( F12 + 2 F66)2 12 + A44 Hl1 12 + ,,2 D22 Hl1 12 + 2 C2 D44 Hl1 12 + n 2 D66 Hl1 12 + C~ F44 Hl1 12 + 2 ",2
D66 H 1212 + A55 H66 12 + 2 C2 D55 H66 12 + 4,,2 D66 H66 12 + C~ F55 H66 12 + 2 n 2 D12 F12 14 + 4 "'.2 D12 F66 14 + 2 Fl1 (2 F66
(10 - ",2 12 ) - (A44 + 2 C2 D44 + c~ F44) 14 + ,,2(_ F12 12 + (D12 - D22 + D66) 14» - ,,2 D12( D12 + 2 D66) 16) + Dl1
«A55 + C2(2D55 + c2 F55»12 +2q(F66 10 + (A55 + c2(2D55 + C2F55»)I4) + ",2 ci(H22 + H66)I6 + D66(IO + ",2C~ 16)
+ 2 ",2 C~( H12 14 + 2 H66 14 - (F12 - F22 + F66) 16) + C~( H66( 10 + 4 ",2 12 ) + (A44 + A55 + C2(2 D44 + 2 D55 + C2( F44
+ F55»)16 + ",2(2H1212 - 2(F12 +2F66)I4 + D2216») +2C~(-4,,2 Fi6 14 + F66(Hl1(I0 - ",2 12 ) - 6,,2 Fl1 14) + n 2
(F22 Hl1 12 + Fl.214 + (D12 + 2 D66)( Hl1 + H12 + 2 H66) 14 - F12« Hl1 + H12 + 2 H66) 12 + (D12 + 2 D66) 16» + Fl1
2
(H66( 10 + 2 ",2 12 ) + n ( H12 12 - 2( F12 + F22) 14 + (D12 + 2 D66) 16») + c1(- ,,2( Hl.2 12 + 2 H12 H66 12 - 2( Fl1 + F12
+ 2 F66) H12 14 + 2 Fl1( H22 - H66) 14 + «F12 + 2 F66)(2 Fl1 + F12 + 2 F66) - 2( D12 + 2 D66) H66) 16) + Hl1( H66( 10
+ ",2 12 ) + ",2( H22 12 - 2( F12 + 2 F66) 14 + 2( D12 + 2 D66) 16)))) w 2 - C~(- 14(2 Fl1 12 + (D66 + cl (4 Fl1 + 2 F66 + q
H66»14) + «Dl1 + D66 +2q F66 + C~ H66) 12 + 2q(Dl1 - q Fl1)14)I6 + C~ Dl1 Ii + Hl1 12(12 + q(214 + q 16»)W 4
4
a4 = _20 c2 D~2 D'1'1 + 20 4 c2 Du D22 D44 + 8 a? c~ D12 D44 D5S - 40
4
c2 D12 D44 D66 + 2 0 4 c2 D22 Dss D66 + 16 0
2
c~ D44
4
D55 D66 + 2 A44 A55 ",2( D12 + 2 D66) + 4 A55 ,,2 c2 D44( D12 + 2 D66) + 4 A44 ,,2 c2 D55( D12 + 2 D66) + 4" ~l c2 D22 D44
4 4
F~1 - 40: c1 c2 D12 D4'1 F12 - 40: Ci c2 D22 DS5 F12 - 80:
4
c1 c2 D'14 D66 F12 - 0:
6
c~ D22 F't2 - 20:
4
c~ c2 D'14 F't2 + 40:
4
C1 c2
4 6 4
D12 Dss F22 + 80: c1 c2 D5S D66 F22 + 2 0 c~ D1.2 F12 F22 - 40 c~ c2 D5S F12 F22 + 2 0: c~ Fi.2 F22
6
- 0
6
c~ Du Fi2 - 20:
6
4
c~ c~ D~2 c~ 2
c~ 4
c~ c~ c~
4 2
F11 Fi2 - 0:: F4'1 + 0 Du D22 F'14 + 4 0 D12 DSS F44 - 2 a D12 D66 F44 + 8 a? Dss D66 F4'1 + 2 ASS 0
4 4 4
( D12 + 2 D66) F44 + 2 0 c1 c~ D22 F11 F44 - 2 Q ci. c~ D12 F12 F44 - 40: c1 c~ D66 F12 F44 - 0:
4
c~ c~ Fi.2 F44 + 4 0 2 c~ D12 D4'1
F55 + ",4 c~ D22 D66 F55 + 8 ",2 c~ D44 D66 F55 + 2 A .. ,,2 c~( D12 + 2 D66) F55 - 2 ",4 ci c~ D22 F12 F55 + 2 ",4 q c~ D12 F22 F55.
+ 4 ~ 4 c1 c~ D66 F22 FSS -: 20: 4 c~ c~ F12 F22 Pss + 2 a? c~ D12 F44 FS5 + 4 a? c~ D66 F44 Fss - 40 4 c1 c2 D22 Dss F66 - 4 0: 6 c~
4
D22 F12 F66 - 8 0: c~ c2 D44 F12 F66 + 4 0: 6 c~ D12 F22 F66 - 8 0:
4
c~ 4
c2 Dss F22 F66 - 40: cf c~ F12 F44 F66 - 20:
4
c1 c~ D22 Fss
422 62 2 42 2 63 2 422 2
F66 - 4 a: c 1 c2 F22 FSS F66 - 40: c1 D22 F66 - 8 Q c 1 C2 D44 F66 - 8 Q c1 F22 F66 - 40 c 1 c2 F44 F66 + 20: 42 •
c 1 c2 D22 D44 H11
6 4 6
- 0 c1 Fi2 H11 + 0: cf c~ D22 F44 H11 +20 cf D22 D66 H12 - 2 0:
6
c~ D22 F12 H12 + 2 0 6 c~ D12 F22 H12 + 4 0 c~ D66 F22 H12
6
6 6
+ 2 0: c1 F12 F22 H12 + 4 0 c~ F22 F66 H12 - 0:
6
c~ D22 H~2 - 0
6
cf D~2 H22 + 0:
6
cf Du D22 H22 + 40: 4 cf c2 D12 Dss H22 - 20:
6
4 6
cf D12 D66 H22 + 80. cf c2 DS5 D66 H22 + 2 0 c~ D22 Fll H22 - 2 0
6
c~ D12 F12 H22 - 4 0
6
c~ D66 F12 H22 - 0
6
ci F't2 H22 + 2 0: 4
4,,6 ci D66 H22 H66 - A44 '" 4( D~2 - Dl1 D22 + q (-2 D22 Fl1 + 4 D66 F12 + q (F12 + 2 F66)2 - q D22 Hl1 - 4 q D66 H66)+
2 D12( D66 + q (F12 - q H66»)) + A55 ",4(2 q (F22( D12 + 2 D66 - q ( F12 + 2 F66)) + q ( D12 + 2 D66) H22) + D22( D66+
q (-2( F12 + F66) + q H66»)) + (-( A55( D66 + cd2 F66 + q H66» + 2 C2( Dl1 D44 + D55 D66 + q (2 D44 Fl1 + 2 D55 F66
+ q D44 Hl1 + q D55 H66)) + c~( Dl1 F44 + D66 F55 + q (2 Fl1 F44 + 2 F55 F66 + q F44 Hl1 + q F55 H66») 10 - c2(2 D44
+ C2 F44)( A55 + C2(2 D55 + C2 F55» 12 - ",2( - D~2 10 - 2 D12 D66 10 + 2 q D22 Fl1 10 - 4 q D66 F12 10 - 4 c~ Fl.2 10 + 4 c~
Fl1 F22 10 - 8c~ F12 F66 10 + c~ D22 Hl1 10 + 2 c~ F22 Hl1 10 - 2 c~ D66 H12 10 - 4c~ F12 H12 10 - 4 c~ F66 H12 10 - c1 H~2 10
+ 2 c~ Fl1 H22 10 + c1 Hl1 H22 10 + Dl1( D22 + q (2 F22 + Ci H22» 10 - 4c~ Fl2 H66 10 - 2 c1 H12 H66 10 - 2 Ci D122 F12
+ 2 F66 + Ci (H12 + H66» 10 + A55 D22 12 + 2 C2 D22 D55 12 + A55 D66 12 + 2 C2 D44 D66 12 + 2 C2 D55 D66 12 + 4 q C2 D44 Fl1
12 - 2 A55 c1 F12 12 - 4 c1 c2 D44 F12 12 - 4 C1 c2 Dss F12 12 + 2 ASS C1 F22 12 + 4 Ci c2 Dss F22 12 + c~ D66 F44 12 + 2 c1 c~ Fu
F44 12 - 2 q c~ F12 F . . 12 + C2 Dl1 (2 D44 + c2 F44) 12 + c~ D22 F55 12 + c~ D66 F55 12 - 2 q c~ Fl2 F55 12 + 2 q c~ F22 F55 12
- 2 AS5 c1 F66 12 - 4 c1 c2 D44 F66 12 - 4 c1 c2 DSS F66 12 - 2 C1 c~ F44 F66 12 - 2 C1 c~ Fss F66 12 + 2 ci c2 D44 Hu 12 + cf c~ F44
H11 12 + A5S ci H22 12 + 2 cf C2 DssH22 12 + c~ c~ Fss H22 12 + Ass cf H66 12 + 2 cf c2 D44 H66 12 + 2 cf C2 Dss H66 12 + cf c~
F44 H66 12 + ci c~ FSS H66 12 + 4 Ass q D66 14 + 8 C1 C2 D44 D66 14 + 8 C1 C2 Dss D66 14 - 2 Ass ci F12 14 - 4 c~ c2 D44 F12 14-
4 cf c2 Dss F12 14 + 4 c1 c~ D66 F44 14 - 2 cf c~ F12 F44 14 + 4 c1 c~ D66 FSS 14 - 2 cf c~ F12 FS5 14 + 2 C1 D12( ASS + c2(2 D44 +2
D55 + C2( F44 + F55») 14 - 4 A55 c~ F66 14 - 8 c~ C2 D44 F66 14 - 8 c~ C2 D55 F66 14 - 4 c~ c~ F44 F66 14 - 4 c~ c~ F55 F66 14 + 2
c~( Dl2 + 2 D66)( A55 + C2(2 D44 + 2 D55 + C2( F44 + F 55 ») 16 ) - A44« A55 + 2 c2 D55 + ",2 D66 + c~ F55) 12 + Dl1 ( 10 + ",2
131
12) + 2Ci( Fll( 10 + a? 12) + ,,2( -( F'2 + F66) 12 + (D12 + 2D66) 14)) + Ci( Hl1( 10 + ,,2 12) + ,,2( H66 12 - 2( F'2 + 2 F66)
14 + 2( D'2 + 2 D66) 16))) - ,,4 Ci( -4 Fi6 12 + 2 D22 H'2 12 + 2 D66 H'2 12 - Ci H~2 12 + Dl1 H22 12 + D66 H22 12 + 2 c, Fll
H22 12 - 2 Ci F66 H22 12 + ci Hl1 H22 12 + 4 D22 H66 12 + 4 D66 H66 12 - 2 ci H'2 H66 12 + Ci H22 H66 12 + 4 D'2 F66 14 - 4 D22
F66 14 - BCi Fi6 14 + 2 C, D'2 H'2 14 + 2c, D22 H'2 14 + 4c, D66 H'2 14 + 4ci F66 H'2 14 + 2c, D'2 H22 14 + 4c, D66 H22 14
- 4Ci F66 H22 14 +4Ci D'2H66 14 +4c, D22 H66 14 +Bc, D66H66 14 + (- Di2 + Dl1 D22 +2D'2(- D66 + ci(H22 + H66))
+ 4 ci( - Fi6 + D66( H22 + H66)) + D22( D66 + Ci(2 Fl1 - 2 F66 + Ci( Hl1 + H66)))) 16 + 2 F,2( -( F22 + 2 F66 + C, (H'2 + H22
+ 2 H66)) 12 + (D'2 - D22 + Ci( -2 F22 + C, H'2 - C, H22)) 14 - c,( D'2 + D22 + 2 D66 + Ci F22 + 2 Ci F66) 16) - F'f2( 12+
c,(-2 14 + Ci 16)) + 2 F22«- Dll + D'2 + D66) 14 + Ci(- Hl1 + H'2 + H66) 14 + c,( H'2 12 + 2 H66 12 - 2 Fl1 14 + D'2 16
+ 2 D66 16) - 2F66( 12 + c,(3 14 + Ci 16))))) w 2 + (Dl1 10 12 + D66 10 12 + 2Ci Fl1 10 12 + 2c, F66 10 12 + ci Hl1 10 12 + ci
H66 10 12 + Ass I~ + 2 C2 DSS I~ + c~ FSS I~ + 2,,2 ci H'2 I~ + 4,,2 Ci H66 I~ + 2 C, Dl1 10 14 + 2 C, D66 10 14 + 4 ci Fl1 10 14
(D44 + DSS)) 12 + 2 ,,2 D66 12 + 2,,2 Ci Fl1 12 - 4,,2 C, F'2 12 + 2,,2 C, F22 12 + c~ F44 12 + c~ Fss 12 - 4,,2 Ci F66 12 + ,,2 ci
Hl1 12 + ,,2 ci H22 12 + 2 ,,2 ci H66 12 + Dll ( 10 + ,,2 12 ) + 4 ,,2 Ci (D'2 + 2 D66 - C, ( F'2 + 2 F66)) 14) 16 -i- 2 ,,2 c1( D'2 + 2
4 6
D66) Ii)w + ci( I~ - 1216)( 12 + c,(2 14 + C, 16))W
as = -( Ass + 2 C2 'Dss t c~ Fss + ,,2( D66 + 2 C, F66 + ci H66) - ( 12 + 2 C, 14 + ci 16) w 2 )(" 4( A44 D22 + 2 C2 D22 D44 + c~ D22
F44 + ,,2 ci (- F;2 + D22 H22)) --. (C2(2 D44 + c2 F44) 10 + A44( 10 + ,,2 12) + ,,2« D22 + 2 Ci F22 + ci H22) ,10 + C2(2 D44+
c2 F44) 12) + ,,4 ci( H22 12 - 2 F22 14 + D22 16))W 2 + (,,2 ci( - I~ + 12 16) + IO( 12 + c,(2 14 + Ci 16))) w 4 )
(8)
132
Appendix C
a2 = -C 1
2( A44 + a 2 ) 2 22
D22 Fu + 2Q c1D12FUF12 - Q
22
cIDUF12
2
+ 20 23 2
c1FUF12 -
232
20: cIFUF22 + 40 22
c 1 D12
133
D12H12H66 + 804C1D66H12H66 - o.4C1DuH22H66 + a? D~2)'N'XO - o? DUD22)'NxO - 2C2DuD44ANxO + 20 2 D12D66
)"NxO - 2C2DSsD66ANxO - 2o:2CID22FI1ANxO - 4CIC2D44FuANxO + 402CID12F12>.Nxo + 402q D 66 P 12A
Nxo + 40?c~F~2)'NxO - 202qDUF22)..Nxo - 40?c~FllF22ANxO - C~DllF44>'NxO - 2CIC~FllF44>'
NxO - c~D66F55>'NxO + 402qD12F66"\NxQ - 4CIC2D55F66ANxO + 802c~F12F66ANxO - 2CIC~F55F66>'
NxO - o:2c~D22Hl1)"NxO - 2c~C2D44Hll)'NxO - 2Ct2C~F22Hll>'NxO - c~c~F44Hll>'NxO + 2o.2c~D12H12>'
Nxo + 2a?c~D66H12)'NxO + 4a?c~F12H12>'NxO + 4o:2c~F66H12>'NxO + Ct2c1H~2>'NxO - o:2c~ DUH22A
NxO - 202c~FllH22)'NxO - a2c1HuH22>.Nxo + 2a2c~D12H66>'NxO - 2c~C2D55H66>'NxO + 4o?c~ Fl2H66
>.NxO - c 22
1 c2FSSH66>'NxO + 2a 24
c 1 H12H66>'NxO - A •• ( AssDll 2
+ 2C2DllDSS + c2DllFSS + '" 2 (2q (Fll(D12 + 2D66
- q(F12 + 2F66» + q (D12 + 2D66)Hll) + Dll(D66 + q( -2(F12 + F66) + qH66») + (Dll + q(2Fll + qHll»>'Nxo)
+ Ass ( - C2Dll(2D •• + C2F•• ) + ",2 (D~2 - Dll(D22 + q(2F22 + qH22» + 2D12(D66 + q(F12 - qH66»
+q (CdF12 + 2F66)2 + 4D66(F12 - qH66»)) - (D66 + q(2F66 + qH66»>'Nxo) - ",2(Dll + q(2Fll + qHll»
(D66 +q (2F66 +q H66))>'Ny o
- 4ct
6
c1 Fi6H22 + o:6 c i D22Hll H22 + A5Sll: 4 c i D22H66 + 40: 4cic2D12D44H66 + 2o:4cic2D22D5SH66 + 40 6 ci D22
D66H66 + 80:4cic2D44D66H66 - 406crD22F12H66 +40:6cID12F22H66 + 8a:6crD66F22H66 + 20:4cic~D12F44H66
+ 4a:4cic~D66F44H66 + a::4cic~D22FssH66 - 2a:6c1D22H12H66 + 2o:6c1D12H22H66 + 4o:6c1D66H22H66 + AS50?
D22)'Nx O + 2ASSC2D44)'NxO + 2a::~C2D22Dss)'NxO + 4c~D44Dss)'NxO + a/I D22D66>'NxO + 2o:2c2D44D66>'Nxo
+ 2AS50?CIF22)'Nxo + 4o?C}C2DSSF22)'Nxo + 20:4C}D66F22>'NxO + A5SC~F44)'NxO + 2c~DssF44>'Nxo + 0: 2
C~D66F44)'Nxo + O?C~D22F55)'Nxo + 2c~D44FS5>'NxO + 2o:2CIC~F22Fss)'NxO + c~F44Fss)'Nxo + 2o:4cl
D22F66ANxO + 4o?C}C2D44F66ANxo + 40:4ciF22F66>'Nxo + 2a2clc~F44F66>'NxO + AssCt CiH22>'N x o
2
(Ass(D66 + q(2F66 + qH66» + 2C2(DllD •• + DssD66 + q(2D •• Fll + 2DssF66 + qD •• Hll + qDSSH66»
2
+ c2(DllF•• + D66FSS + q(2FllF4' + 2FsSF66 + qF4.Hll + qFsSH66» - " 2 ( D12
2 - Dll(D22 + q(2F22
+ qH22» + 2D12(D66 + q(2F12 + 2F66 + q(H12 + H66») + q(4F12(D66 + qF12) - D22(2Fll + qHll)
+ q(8F12F66 + 2D66H12 - 2Fll(2F22 + q H 22) + q(-2F22Hll + H12(4(F12 + F66) + qH12) - c1HllH22
+ 4F12H66 + 2qH12H66»») )>'Nyo + A •• (2Ass,,2(D12 + 2D66) + a' ( - D~2 + DllD22 + q (2D22Fll - 4D66
F12 - q (F12 + 2F66)2 + qD22Hll + 4qD66H66) - 2D12(D66 + q(F12 - qH66») + Ass>.Nxo + c2(2Ds5 + C2
Fss».Nxo + ",2 ( 4C 2 D SS(D12 + 2D66) + 2c~(D12 + 2D66)Fss + (D66 + q(2F66 + q H66»>'Nx o + (Dll + q
(2Fll +qHll»>'Nyo))
134
Appendix D
135
5. F. A. Fazzolari, J. R. Banerjee. Dynamic Stiffness Method for Modal
Analysis of Composite Circular Cylindrical Shells. Structural Engineer-
ing and Mechanics - An International Journal, Submitted for publication 2013.
136
12. F. A. Fazzolari, J. R. Banerjee, M. Boscolo. Buckling of Composite Plate
Assemblies using Higher Order Shear Deformation Theory - An Exact
Method of Solution. Thin- Walled Structures, Vol. 71 (2013), pages 18-34.
Conference Proceedings
137
4. F. A. Fazzolari, J. R. Banerjee. A Refined Dynamic Stiffness Element
for Free Vibration Analysis of Composite Circular Cylindrical Shells
Based on Higher Order Shear Deformation Theory. 54th AIAA/ASME/
ASCE/AHS / ASC Structures, Structural Dynamics, and Materials Conference,
Boston, Massachusetts 8-11 April 2013.
138
9. F. A. Fazzolari, M. Boscolo, J. R. Banerjee. Dynamic Stiffness Formula-
tion and Free Vibration Analysis of Composite Plates using Higher
Order Shear Deformation Theory. 53rd AIAAIASMEIASCEIAHS IASC
Structures, Structural Dynamics, and Materials Conference, Sheraton Waikiki
Honolulu, Hawaii 23-26 April 2012.
Reports
139
5. F. A. Fazzolari, J. R. Banerjee, M. Boscolo. Dynamic Stiffness Modeling
of Composite Plate Assemblies. Report 5/6. EOARD Grant No. FA8655-
10-1-3084.
140
Bibliography
[1] S. Germain. Researches sur la theorie des surfaces elastiques. Courcier, Librarie
Pour Les Sciences, Paris, 1821.
[3] W. Soedel. . Vibrations of Shells and Pla~es. Marcel Dekker, New York, 2n &
edition, 2004.
[5] E. Reissner.The effect of transverse shear defor~ation 0)1 the bending of elastic
plates. Journal ofApplied Mechanics, 67:A67-A77, 1945.
[6] R.D. Mindlin. Influence of rotary inertia and shear on flexural motions of
isotropic elastic plates. Journal of Applied Mechanics, 18(10):31-38, 1951.
[8] C.B. Smith. Some new types of orthotropic plates laminated of orthotropic
materials. J. Appl. Mech, 20:286-288, 1953.
[9] E. Reissner and Y. Stavsky. Bending and stretching of certain types of heteroge-
neous anisotropic elastic plates. Journal of Applied Mechanics, 28(3):402-408,
1961.
[13] J.G. Ren. A new theory of laminated plates. Compos. Sci. Tech., 26:225-239,
1986.
141
[14] J.G. Ren. Bending theory of laminated plates. Compos. Sci. Tech., 27:225-248,
1986.
[15] W. Yu. A new theory of elastic sandwich plates one dimensional case. ASME
J. Appl. Mech., 37:1031-1036, 1959.
[17] J.R. Vinson and R.L. Sierakowski. The Behaviour of Structures Composed of
Composite Materials. Martinus Nijhoff Publishers, Dordrecht, The Netherlands,
1986.
[19] J.M. Whitney. The effects of 'transverse shear deformation on the bending of
laminated plates. J. Compos. Mater., 3:534-547, 1969.
[20] D. Mohan and H.B. Kingsbury. Free vibrations of generally laminated or-
thotropic plates. J. Acoust. Soc. Am., 50:266-269, 1971.
[22] M. S. Qatu. Accurate equations for laminated composite deep thick shells.
International Journal of Solids and Structures, 36(19):2917-2941, 1999.
[24] M. S. Qatu, W. S. Rani, and W. Wenchao. Recent research advances on the dy-
namic analysis of composite shells: 2000?-2009. Composite Structures, 93(1):14-
31, 2010.
[27] J. Du, Z. Liu, W. L. Li, X. Zhang, and W. Li. Free in-plane vibration analysis
of rectangular plates with elastically point-supported edges. J. Vib. Acoust.,
132:DOI: 10.1115/1.4000777, 2010.
[28] A.W. Leissa and Y. Narita. Vibration studies for simply supported symmetri-
cally laminated rectangular plates. Composite Structures, 12: 113-132, 1989.
142
[29] E. Carrera. Historical review of zig-zag theories for multilayered plates and
shells. Appl Mech Rev, 56:287-308, 2003.
[33] E. Carrera. Theories and finite elements for multilayered anisotropic composite
plates and shells. Archives of Computational Methods in Engineering, 9(2):87-
140,2002.
"[34] E. Carrera. Theories and finite elements for multilayered plates and shells:
a unified compact formulation with numerical assessment and benchmarking.
Archives of Computational Methods in Engineering, 10(3):216-296, 2003.
[36] E. Carrera. A refined multilayered finite element model applied to linear and
nonlinear analysis of sandwich plate. Compo Sci. Technol., 58: 1553-1569, 1998.
[37] E. Carrera. Mixed layer-wise models for multilayer plate analysis. Compos.
Struct., 43:57-70, 1998.
[38] E. Carrera. Evaluation of layer-wise mixed theories for laminated plate analysis.
AIAA J., 26:830-839, 1998.
[43] E. Carrera and L. Demasi. Classical and advanced multilayered plate elements
based upon pvd and rmvt. partl: Derivation of finite element matrices. Int. J.
Numer. Meth. Engng., 55:191-231, 2002.
143
[44] E. Carrera and L. Demasi. Classical and advanced multilayered plate elements
based upon pvd and rmvt. part2: Numerical implementation. Int. J. Numer.
Meth. Engng., 55:253-291, 2002.
[45] F. A. Fazzolari and E. Carrera. Advanced variable kinematics Ritz and Galerkin
formulations for accurate buckling and vibration analysis of laminated compos-
ite plates. Composite Structures, 94(1):50-67, 2011.
[51] F. A. Fazzolari and E. Carrera. Advances in the Ritz formulation for free
vibration response of doubly-curved anisotropic laminated composite shallow
and deep shells. Composite Structures, 101:111-128, 2013.
144
[55] A. J. M. Ferreira, C. M. C. Roque, and R. M. N. Jorge. Static and free vibration
analysis of composite shells by radial basis fucntions. Engineering Analysis with
Boundary Elements, 30:719-733, 2006.
[59] M. Cinefra, E. Carrera, L. Della Croce, and C. Chinosi. Refined shell ele-
ments for the analysis of functionally graded structures. Composite Structures,
94(2):415-422, 2012.
[60] C. Chinosi, M. Cinefra, L. Della Croce, and E. Carrera. Reissners mixed vari-
ational theorem toward mitc finite elements for multilayered plates. Composite
Structures, 99:443-452, 2013.
[63] W. H. Witt rick and F. W. Williams. A general algorithm for computing natural
frequencies of elastic structures. Quarterly Journal of Mechanics and Applied
Mathematics, 24(3):263-284, 1970.
145
[67J RJ. Plank and W. H. Wittrick. Buckling under combined loadings of thin flat-
walled structures by a complex finite-strip method. International Journal for
Numerical Methods in Engineering, 8:323-339, 1974.
[75J J. R Banerjee. Free vibration analysis of a twisted beam using the dynamic stiff-
ness method. International Journal of Solids and Structures, 38(38-39):6703-
6722, 2001.
[76J J. R Banerjee. Free vibration of sandwich beams using the dynamic stiffness
method. Computers and Structures, 81(18-19):1915-1922, 2003.
146
[77] J. R. Banerjee. Development of an exact dynamic stiffness matrix for free vibra-
tion analysis of a twisted Timoshenko beam. Journal of Sound and Vibration,
270(1-2):379-401, 2004.
[80] M. Boscolo and J. R. Banerjee. Dynamic stiffness elements and their appli-
cations for plates using first order shear deformation theory. Computers &
Structures, 89:395-410, 2011.
[81] M. Boscolo and J. R. Banerjee. Dynamic stiffness method for exact in-plane
free vibration analysis of plates and plate assemblies. Journal of Sound and
Vibrations, 330(12):2928-2936, 2011.
[84] J.N. Reddy. A simple higher order theory for laminated plates. Journal of
Applied Mechanics, 51:745-752, 1984.
[85] J.N. Reddy and D. Phan. Stability and vibration of isotropic, orthotropic and
laminate plates according to a higher-order shear deformation theory. Journal
of Sound and Vibration, 98(2):157-170, 1985.
[87] B.F. Vlasov. On the equations of bending of plates. Dokla Ak Nauk Azerbei-
janskoi SSR (in Russian)., 3:955-959, 1957.
[88] G. Jemielita. Technical theory of plates with moderate thickness. Rozprawy Inz
(in Polish), 23:483-499, 1975.
[90] S. P. Timoshenko. Theory of elastic stability. McGraw Hill, New York, 1961.
147
[91] A.W. Leissa. Condition for laminated plates to remain flat under inplane load-
ing. Composite Structures, 6:261-270, 1986.
[95] W. H. Wittrick and Curzon P. L. V. Stability functions for the local buckling of
thin flat-walled structures with the walls in combined shear and compression.
Aeronautical Quarterly, 19:327-351, 1968.
[96] W. H. Wittrick. General sinusoidal stiffness matrices for buckling and vibra-
tion analysis of thin flat-walled structures. International Journal of Mechanical
Sciences, 10:949-966, 1968.
[102] V. V. Novozhilov. The theory of thin elastic shells. P. Noordhoff Lts (Groningen,
The Netherlands), 1964.
[103] H. Aron. Das gleichgewicht und die bewegung einer unendlich dunnen,beliebig
gekrummten elastischen schale. J. fur Math., pages 136-174, 1874.
148
[105] A. E. H. Love. The small free vibrations and deformations of a thin elastic shell.
Phil. Trans. Roy. Soc. (London), ser. A(179):491-549, 1888.
[106] A. I. Lur'e. General theory of elastic shells. Prikl. Mat. Mekh., 4(1):7-34, 1940.
[107] A. L. Gol'denveizer. Theory of Thin Shells. Pergamon Press, New York, 1961.
[110] K. M. Mushtari. Certain generalizations of the theory of thin shells, (in Rus-
sian). Izv. Fiz. Mat. Odd. pri Kazan. Univ., Vol. 11 (8), 1938.
[111] L. H. Donnell. Stability of thin walled tubes under .torsion. Technical Report
479, NACA, 1933.
[115] W. Fliigge. Statik und Dynamik der Schalen. Julius Springer, Berlin (Reprinted
by Edwards Brothers Inc., Ann Arbor, Mich., 1943), 1934.
[117] R. Byrne. Theory of small deformations of a thin elastic shell. Seminar Reports
in Math., Univ. of Calif. Pub. in Math., N. S., 2(1):103-152, 1944.
149
[121] S. T. Gulati and F. Essemberg. Effect of anisotropy in axisymmetric cylindrical
shells. Journal fo Applied Mechanics, 34:659-666, 1967.
[123] S. A. Ambartsumian. Some main equations of the theory of thin layered shell.
DAN ArmSSR, 8(5), 1948.
[135] C. W. Bert. Dynamic of composite and sandwich panels - part II. Shock fj
Vibration Digest, 8(11):15-24, 1976.
150
[137] Y. K. Cheung and C. I. Wu. Free vibration of thick, layered cylinders having
finite length with various boundary conditions. Journal of Sound and Vibration,
24:189-200, 1972.
151
[151] A. J. M. Spencer, P. Watson, and T. G. Rogers. Stress analysis of laminated
circular cylindrical shells. In Recent developments in composite materials struc-
tures; Proceedings of the Symposium, ASME Winter Annual Meeting, Dallas,
TX, pages 57-60, Nov. 25-30, 1990.
[152] P. Cicala. Systematic approach to linear shell theory. Levrotto & Bella, Turin,
Italy, 1965.
[156] S. P. Timoshenko. Theory of Plates and Shells. McGraw Hill, New York, 1959.
[159] W. Soedel. Simplified equations and solutions for the vibration of orthotropic
cylindrical shells. J. Sound. Vib., 87(4):555-566, 1983.
[160] Y. C. Das. Vibrations of orthotropic cylindrical shells. Appl. Sci. Res., 12(4-
5):17-26,1964.
[162] J. Callahan and H. Baruh. A closed-form solution procedure for circular cylin-
drical shell vibrations. International Journal of Solids and Structures, 36:2973-
3013, 1999.
[163] E. Carrera. On the use of transverse shear stress homogeneous and non-
homogeneous conditions in third-order orthotropic plate theory. Composite
Structures, 77:341-352, 2007.
[164] J. N. Reddy. Mechanics of laminated composite plates and shells. Theory and
Analysis. CRC Press, 2nd edition, 2004.
[165] Y. F. Xing and B. Liu. Exact solutions for the free in-plane vibrations of
rectangular plates. International Journal of Mechanical Sciences, 51(3):246-
255,2009.
152
[166] C. I. Park. Frequency equation for the in-plane vibration of a clamped circular
plate. Journal of Sound and Vibration, 313(1-2):325-333, 2008.
[167] D.J. Gorman. Exact solutions for the free in-plane vibration of rectangular
plates with two opposite edges simply supported. Journal of Sound and Vibra-
tion, 294(1-2):131 - 161, 2006.
[168] Y. F. Xing and B. Liu. New exact ·solutions for free vibrations of thin orthotropic
rectangular plates. Composite Structures, 89(4):567-574, 2009.
[171] ANSYS v10.0 theory manual. ANSYS Inc., Southpointe, PA, 2006.
[174] S. C. Fan and M. H. Luah. Free vibration analysis of arbitrary thin shell
structures by using spline finite eJement. Journal of Sound and Vibration,
179(5):763-776, 1995.
[175] R. K. Khare, T. Kant, and A. K. Garg. Free vibration of composite and sand-
wich laminates with a higher-order facet shell element. Composite Structures,
65(3-4):405-418, 2004.
153
[180] J. B. Greenberg and Y. Stavsky. Vibration of laminated filament-wound cylin-
drical shells. AIAA Journal, 9: 1055-1062, 1981.
154