Optimal Designs of Submerged Domes

OPTIMAL DESIGNS OF SUBMERGED DOMES
VO KHOI KHOA
NATIONAL UNIVERSITY OF SINGAPORE
2007
VO KHOI KHOA
(B. Eng, University of Technology, Vietnam)
A THESIS SUBMITTED
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
DEPARTMENT OF CIVIL ENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
2007
ACKNOWLEDGMENTS
First and foremost I wish to express my sincere gratitude to my supervisor, Professor
Wang Chien Ming (Engineering Science Programme and Department of Civil
Engineering, National University of Singapore) for his highly valuable supervision
throughout my course of study. His constant inspiration, kind encouragement,
extensive knowledge, serious research attitude and enthusiasm have extremely
assisted me in completion of this thesis. Also special thanks go to Professor Rob Y.H.
Chai (Department of Civil and Environmental Engineering, University of California,
Davis) for his valuable suggestions, discussions and help in the research work.
I want to express my gratitude to the National University of Singapore for
providing the Research Scholarship during this doctoral study in the Department of
Civil Engineering.
My parents and sisters have been extraordinary sacrificial for providing me with
whatever requirements for my education opportunity. For this, I am thankful. Finally,
I am also grateful to my girlfriend, Ms Le Nguyen Anh Minh, and to my friends Mr.
Dang The Cuong, Mr. Nguyen Dinh Tam and Mr. Tun Myint Aung for their kind help
and encouragement.
i
TABLE OF CONTENTS
Acknowledgements ........................................................................................................ i
Table of Contents .......................................................................................................... ii
Summary....................................................................................................................... vi
Nomenclature ............................................................................................................... ix
List of Figures .............................................................................................................. xi
List of Tables .............................................................................................................. xiv
CHAPTER 1. Introduction......................................................................................1
1.1 Submerged dome ideas ........................................................................................3
1.2 Rotational shells...................................................................................................7
1.3 Buckling of rotational shells ................................................................................8
1.4 Optimal design of domes against buckling........................................................11
1.5 Objectives and scope of study............................................................................12
1.6 Layout of thesis..................................................................................................14
CHAPTER 2. Uniform Strength Designs Of Submerged Spherical Domes .....16
2.1 Introduction........................................................................................................17
2.2 Membrane theory ...............................................................................................18
2.2.1 Basic assumptions of classical thin shell theory ........................................18
2.2.2 Geometrical properties of rotational shells ...............................................19
ii
2.2.3 Membrane analysis ....................................................................................20
2.3 Problem definition and basic equations .............................................................23
2.3.1 Problem definition .....................................................................................23
2.3.2 Basic equations ..........................................................................................24
2.4 Results and discussions......................................................................................29
2.4.1 Analytical solution using power series method .........................................29
2.4.2 Accuracy of analytical solution for dome thickness ..................................30
2.4.3 Critical value of subtended angle...............................................................32
2.4.4 Effect of water depth on thickness variation..............................................33
2.4.5 Minimum weight design ...........................................................................34
2.5 Concluding remarks ...........................................................................................39
CHAPTER 3. Constant Strength Designs Submerged General Domes ..............40
3.1 Introduction........................................................................................................41
3.2 Problem definition and basic equations .............................................................42
3.2.1 Problem definition .....................................................................................42
3.2.2 Governing equations for membrane analysis of submerged domes ..........43
3.2.3 Boundary conditions for membrane actions in fully stressed submerged
domes .........................................................................................................48
3.3 Results and discussions......................................................................................50
3.3.1 Weightless constant strength submerged general domes...........................50
3.3.2 Constant strength of submerged general domes ........................................55
3.4 Concluding remarks ...........................................................................................66
iii
Table of Contents
CHAPTER 4 Energy Functionals and Ritz Method for Buckling Analysis of
Domes ..............................................................................................................67
4.1 Introduction......................................................................................................68
4.2 Governing eigenvalue equation .......................................................................69
4.2.1 Geometrical properties of domes .............................................................69
4.2.2 Mindlin shell theory.................................................................................71
4.2.3 Strain-displacement relations...................................................................73
4.2.4 Stress-strain relations ...............................................................................74
4.2.5 Derivation of energy functionals .............................................................75
4.3 Ritz method for buckling analysis ...................................................................80
4.3.1 Introduction..............................................................................................80
4.3.2 Ritz formulation .......................................................................................82
4.3.3 Boundary conditions ................................................................................88
4.3.4 Mathematica for solving eigenvalue problem..........................................89
4.4 Concluding remarks .........................................................................................95
CHAPTER 5 Buckling Of Domes Under Uniform Pressure ..........................91
5.1 Problem definition ...........................................................................................92
5.2 Geometrical parameters ...................................................................................92
5.3 Results and discussions....................................................................................94
5.3.1 Spherical domes .......................................................................................94
5.3.2 Parabolic domes .....................................................................................103
5.4 Concluding remarks .......................................................................................108
CHAPTER 6 Buckling Of Submerged Domes................................................109
6.1 Problem definition .........................................................................................110
iv
Table of Contents
6.2 Governing equations and Ritz method...........................................................111
6.2.1 Geometrical and loading properties .....................................................111
6.2.2 Energy functionals and Ritz method....................................................115
6.3 Results and discussions..................................................................................118
6.3.1 Spherical domes ...................................................................................120
6.3.2 Parabolic domes ...................................................................................126
6.4 Concluding remarks ........................................................................................131
CHAPTER 7 Optimal Designs of Submerged Domes....................................132
7.1 Problem definition .........................................................................................133
7.2 Method of Optimization.................................................................................135
7.3 Results and Discussions.................................................................................138
7.3.1 Spherical domes ...................................................................................138
7.3.2 Parabolic domes ...................................................................................142
7.4 Concluding remarks .......................................................................................146
CHAPTER 8. Conclusions And Recommendations ...........................................147
8.1 Summary and conclusions……………………………………………….. ...147
8.2 Recommendations for Future Studies ……………………………….……..150
8.2.1 Domes with very large thickness ........................................................150
8.2.2 Non-axisymmetric domes ....................................................................150
8.2.3 Vibration of submerged domes............................................................151
8.2.4 Other design loads on submerged domes.............................................151
References.................................................................................................................152
Appendix...................................................................................................................165
List of Author’s Publications ..................................................................................172
v
SUMMARY
So far, little research has been done on submerged large dome structures. This
prompted the present study on the optimal design of submerged domes for minimum
weight as well as for maximum buckling capacity.
The first part of the thesis presents the membrane analysis and minimum
weight design of submerged spherical domes. By adopting a uniform strength design
as governed by the Tresca yield condition, an analytical expression in the form of a
power series for the thickness variation of a submerged spherical dome was derived.
Further, based on a family of uniform strength designs associated with a given depth
of water and base radius of the dome, the optimal subtended angle 2α and the
optimal dome height for the minimum weight design of submerged spherical domes
were determined.
Extending the research on spherical domes, membrane analysis and optimal
design of submerged general shaped domes were treated. By adopting a constant
strength design, equations governing the meridional curve and thickness variation of
submerged domes were derived with allowance for hydrostatic pressure, selfweight
and skin cover load. The set of nonlinear differential equations, which correspond to a
two-point boundary problem, was solved by the shooting-optimization method. A
notable advantage of the equations derived in this part is the parameterization of the
vi
equations using the arc length s as measured from the apex of the dome. Such
parameterization allows the entire shape of the submerged dome to be determined in a
single integration process whereas previous methods that made used of the Cartesian
coordinates gave problems when vertical or infinite slope was encountered in the
meridian curve. For the special case of a weightless dome without skin cover load,
the thickness of the dome was found to be constant when subjected to hydrostatic
pressure only. The shape of the dome was also found to agree well with the shape
currently reported in the literature. Further, parametric studies of dome shapes under
different water depths and selfweight also led to a better understanding of the optimal
shape of submerged domes. Numerical examples indicated that the airspace enclosed
by the optimal dome reduces in the presence of large hydrostatic pressure. The
reduced airspace is accompanied by a significant increase in the dome thickness,
which in turn results in an increased overall weight of the dome.
In the second part of the thesis, the optimal design of domes against buckling is
focused. Although buckling of shells under compressive loading is of practical
significance in the design of these structures, most of the studies thus far have focused
on spherical domes using a thin shell theory. This study presents the formulation and
solution technique to predict the critical buckling pressure of moderately thick
rotational shells generated by any meridional shape under external pressure. The
effect of transverse shear deformation is included by using Mindlin shell theory so
that the critical buckling pressure will not be excessively overestimated when the shell
is relatively thick.
The critical buckling pressure of moderately thick shells under uniform pressure,
formulated as an eigenvalue problem, is derived using the well accepted Ritz method.
vii
Summary
One feature of the proposed method is the high accuracy of the solutions by using an
adequate number of terms in the Ritz functions. The formulation is also capable of
handling different support conditions. This is made possible by raising the boundary
equations to the appropriate power so that the geometric boundary conditions are
satisfied a priori. The validity of the developed Ritz method as well as the
convergence and accuracy of the buckling solutions are demonstrated using examples
of spherical domes (a special case of generic dome structures) where closed-form
solutions exist. Based on comparison and convergence studies, the Ritz method is
found to be an efficient and accurate numerical method for the buckling of dome
structures. New solutions for the buckling pressure of moderately thick spherical and
parabolic shells of various dimensions and boundary conditions are presented and,
although these results are limited by the material properties assumed, they are
nonetheless useful for the preliminary design of shell structures.
Upon establishment of the validity of method and its ability to furnish accurate
results for the buckling of dome structures under uniform pressure, the research was
extended to submerged domes. In addition to hydrostatic pressure, loads acting on the
dome include the selfweight. New solutions for the buckling pressure of moderately
thick spherical and parabolic shells of various dimensions and boundary conditions
are presented. Further, based on a family of spherical and parabolic domes associated
with a given dome height submerged under a given water depth, we determine the
Pareto optimal design for maximum enclosed airspace and minimum weight dome
design.
This thesis should serve as a useful reference source for vast optimal dome design
data for researchers and engineers who are working on analysis and design of shell
structures.
viii
NOMENCLATURE
D water depth
E s , Eθ , Young’s moduli in the direction of the meridian and parallel circle,
respectively
Gsζ shear modulus in the s − ζ plane
H dome height
h dome thickness
κ2 Mindlin’s shear correction factor
L dome base radius
l curve length of one-half of the meridian
N rs , N zs horizontal and vertical components of the meridian forces N s
N s , Nφ membrane force in the meridian direction
Nθ membrane force in the circumference direction
ph , pc , pa hydrostatic pressure, skin cover load and self-weight
ps , pn loads normal and tangential to the middle surface
R radius of spherical domes
r0 the distance of one point on the shell to the axis of rotation
r1 , r2 principal radii of curvature of the dome
ix
Nomenclature
s arc length along the meridian as measured from the apex of the
dome
U elastic strain energy functional
W work done functional
W0 dome weight
u ,w middle-surface displacement along the meridional and normal
directions, respectively
z vertical coordinate
α subtended angle
γa specific weight of dome material
γ sζ transverse shear strain associated with rotation of the shell in the
meridian direction
γw specific weight of water
ε s , εθ normal strain in the direction of the meridional and circumference
direction, respectively
λ buckling pressure parameter
ν s , νθ Poisson’s ratios
ξ normalized thickness
Π total potential energy functional
σ0 the allowable compressive stress
σφ , σθ the meridian and circumferential stress
φ meridian angle
ψ rotation of the middle-surface in the meridional direction
x
LIST OF FIGURES
Fig. 1.1 Pantheon domes .............................................................................................2
Fig. 1.2 Hagia Sophia of Constantinople ....................................................................2
Fig. 1.3 Yumemai floating bridge ...............................................................................4
Fig. 1.4 Mega-Float in Tokyo Bay..............................................................................5
Fig. 1.5 Floating oil storage facility ............................................................................5
Fig. 1.6 Author’s impression of a submerged dome complex ....................................6
Fig. 2.1 Rotational shells (Domes)............................................................................19
Fig. 2.2 Meridian of a dome......................................................................................20
Fig. 2.3 Shell element................................................................................................21
Fig. 2.4 Submerged spherical dome ..........................................................................23
Fig. 2.5 Tresca yield condition..................................................................................24
Fig. 2.6 Free body diagram of dome above horizontal plane a-a..............................24
Fig. 2.7 Thickness variation obtained by series and numerical methods..................31
Fig. 2.8 Variation of α cr with respect to water depth D ........................................32
Fig. 2.9 Thickness variations of submerged domes for various water depths ..........34
Fig. 2.10 Family of uniform strength designed domes for a given base radius L .......35
Fig. 2.11 Variations of weight W0 with respect to subtended angle α

for L = 0.01, 0.02 and 0.04 .......................................................................36
Fig. 2.12 Variations of minimum weight W0 and α opt with respect to base radius L
......................................................................................................................37
xi
List of Figures
Fig. 2.13 Variation of optimal dome height H opt / L with respect to water depth
D / L ............................................................................................................38
Fig. 3.1 Calcareous shell of a sea urchin...................................................................41
Fig. 3.2 Coordinate systems and parameters defining the shape of submerged dome
......................................................................................................................42
Fig. 3.3 Load components on submerged domes ......................................................44
Fig. 3.4 Horizontal and vertical components of the meridian force Ns acting on the
ring foundation.............................................................................................48
Fig. 3.5 Coordinate system for the Runge-Kutta forward integration ......................53
Fig. 3.6 Weightless fully stressed submerged dome shapes under various water
depths ...........................................................................................................54
Fig. 3.7 Submerged dome shapes under selfweight and skin cover load for various
water depths .................................................................................................58
Fig. 3.8 Fully stressed submerged dome shapes with different selfweight parameter
β ...............................................................................................................60
Fig. 3.9 Variation of submerged dome weight respect to subtended base angle φb .62
Fig. 3.10 Optimal shapes of submerged domes with respect to water depths.............64
Fig. 4.1 Coordinate systems and parameters defining the shape of dome structures...
....................................................................................................................70
Fig. 4.2 Membrane forces in an axisymmetrically loaded domes.............................77
Fig. 4.3 Boundary conditions ....................................................................................88
Fig. 5.1 Dome under uniform pressure .....................................................................92
Fig. 5.2 Spherical domes under uniform pressure.....................................................94
Fig. 5.3 Parabolic domes under uniform pressure...................................................103
Fig. 5.4 SAP2000 model of parabolic dome (50x50 elements) ..............................107
Fig. 6.1 Domes under selfweight and hydrostatic pressure.....................................110
xii
List of Figures
Fig. 6.2 Hydrostatic pressure components ..............................................................111
Fig. 6.3 Selfweight of the dome ..............................................................................114
Fig. 6.4 Spherical dome under its own selfweight and hydrostatic pressure ..........120
Fig. 6.5 Variations of critical water depth Dcr = D / H with respect to normalized
thickness ξ = h / H of a hemispherical dome.............................................125
Fig. 6.6 Parabolic dome under its own selfweight and hydrostatic pressure ..........127
Fig. 6.7 Variations of critical water depth Dcr = D / H with respect to normalized
thickness ξ = h / H of a parabolic dome ....................................................130
Fig. 7.1 Dome under selfweight and hydrostatic pressure .....................................133
Fig. 7.2 Family of spherical domes for a given dome height H ..............................134
Fig. 7.3 Family of parabolic domes for a given dome height H .............................134
Fig. 7.4 Variations of performance index J of spherical domes with respect to

normalized base radius L in case of αˆ = 0 and αˆ = 1 . ...........................139
Fig. 7.5 Trade-off curve of normalized dome weight Ŵa and normalized enclosed
airspace parameter Ŝ' a of spherical domes................................................140

normalized base radius L in case of α̂ = 0.25; 0.5 and 0.75 ....................141
Fig. 7.7 Variations of performance index J of parabolic domes with respect to

normalized base radius L in case of αˆ = 0 and αˆ = 1 ............................143
Fig. 7.8 Trade-off curve of normalized dome weight Ŵa and normalized enclosed
airspace parameter Ŝ' a of parabolic domes ...............................................144

normalized base radius L in case of α̂ = 0.25; 0.5 and 0.75 ....................145
xiii
List of Tables
LIST OF TABLES
Table 3.1 Optimal values of base angle φb opt , apex thickness h0 opt , and curved
length lopt ................................................................................................65
Table 5.1 Convergence of critical buckling pressure parameter λ of a clamped

hemispherical dome .................................................................................97
Table 5.2 Convergence of critical buckling pressure parameter λ of a simply

supported hemispherical dome ................................................................97
Table 5.3 Comparison of critical buckling pressure ratio pcr/pcl of a 900 clamped
spherical dome (R/h = 25)........................................................................99
Table 5.4 Effect of transverse shear deformation on the buckling pressures pcr/E of
simply supported hemispherical domes .................................................101
Table 5.5 Buckling pressures pcr/ E of orthotropic hemispherical domes ...........102

parabolic dome with normalized base radius L = 1 ............................ 105
Table 5.7 Convergence of critical buckling pressure parameter λ of a simply

supported parabolic dome with normalized base radius L = 1 ..............105
Table 5.8 Buckling pressure parameter λ of isotropic parabolic domes ..............107
Table 5.9 Buckling pressure parameter λ of orthotropic parabolic domes ..........107
Table 6.1 Convergence of critical buckling pressure parameter λ1 of clamped

hemispherical domes under hydrostatic pressure only ..........................123
Table 6.2 Convergence of critical buckling pressure parameter λ1 of a simply

supported hemispherical domes under hydrostatic pressure only .........123
Table 6.3 Convergence of critical buckling pressure parameter λ1 of a clamped

spherical hemispherical under its own selfweight and hydrostatic pressure
...............................................................................................................124
xiv
List of Tables

supported hemispherical domes under its own selfweight and hydrostatic
pressure ..................................................................................................124

parabolic domes with normalized base radius L = 1 under hydrostatic
pressure only ..........................................................................................128

supported parabolic domes with normalized base radius L = 1 under
hydrostatic pressure only .......................................................................128

parabolic domes with normalized base radius L = 1 under its own
selfweight and hydrostatic pressure.......................................................129

supported parabolic domes with normalized base radius L = 1 under its
own selfweight and hydrostatic pressure ...............................................129
xv
CHAPTER 1
INTRODUCTION
Shell structures have been widely used since ancient times as one of the most common
types of structural form. One of the earliest applications of the shell as a structural
form is represented by beautiful domes that have been constructed as roofs for temples,
mosques, monuments and other buildings. A small dome was even discovered inside
the Bent Pyramid which was built during the Fourth Egyptian Dynasty in about 2900
B.C. (Cowan, 1977). However, domes were not widely used until the Roman Empire. A
good example of the dome construction during the Roman Empire is the Pantheon
dome, which had the longest span (43 m) prior to the 19th century and is still in use
today as a church. The Hagia Sophia of Constantinople (now Istanbul) was built
approximately 1500 years ago, St Peter’s Cathedral in Rome was designed by
Michelangelo in about 1590. In the modern shell applications, many domes were
constructed all over the world for different purposes such as the Millennium Dome (in
England) for exhibition purposes and the Georgia Dome (in USA) for sporting events.
1
Chapter 1: Introduction
Fig 1.1 Pantheon domes
(Source: 1897 Encyclopaedia Britannica)
Fig 1.2 Hagia Sophia of Constantinople
(Source: 1911 Encyclopaedia Britannica)
2
1.1 Submerged dome ideas
As the population and urban development expand in coastal cities, city planners and
engineers resort to land reclamation and construction on and under the sea to create
additional space so as to ease the pressure on existing land use. In recent times, we
have seen very large floating structures being constructed on the coast of densely
populated cities. For example, Japanese engineers have constructed a floating steel
arch bridge that spans 410m across the Yumemai channel in Osaka (Watanabe and
Utsunomiya, 2003), floating oil storage facilities at Shirashima and Kamigoto, a
floating amusement facility at Onomichi and floating emergency rescue bases in Osaka
Bay, Ise Bay and Tokyo Bay. Based on the knowledge gained from the Mega-Float
which measures 1000m x 60m x 3m test model for a floating runway (Yoshida, 2003),
the Japanese are considering the construction of a floating runway of 3.6km x 500m x
20m in the expansion programme for the Haneda International Airport. Other countries
having floating structures include Norway with its famous floating Bergsøysund bridge
and Nordhordland bridge (Watanabe and Utsunomiya, 2003), Hong Kong with its
floating restaurant <http://www.jumbo.com.hk/eng/main.php>, Saudi Arabia with its
floating desalination plant (Abdul Azis et al., 2002), North Korea with its floating
hotel, Canada with its floating heliport and piers, Brazil with its floating pulp plant and
Singapore with its floating performance platform.
Many submerged tunnels have been constructed to join two parts of cities across a
river or to connect two countries over a channel (for example the Channel Tunnel
Crossing between France and England and the Oresund Link between Sweden and
Denmark). These tunnels enhance greater connectivity, and help to redistribute the
population concentrations and generate more economic activities. Research studies on
3
seeking optimal shapes of these submerged tunnels in the form of funicular arches
have been carried out by Gavin and Reilly (2000), Wang and Wang (2002), Fung
(2003), Wang and Ler (2003) and Chai and Kunnath (2003).
Offshore activities are also increasing as mankind seeks to tap the riches of the seas
and oceans. In addition to drilling for oil and natural gas in deep water, there has been
recent interest among engineers to mine methane hydrate (Komai 2003; Ichikawa and
Yonezawa 2003) scattered over the seabed for a cleaner source of fuel. This 21st
century will also likely see the construction of floating and underwater cities, for
example, the Hydropolis project which is an underwater complex featuring a luxurious
hotel with 220 underwater suites in Dubai, the United Arab Emirates
<http://www.hydropolis.com/project/>. For submerged cities, a dome complex may be
used to create the living environment suitable for sustaining human activities for a long
time (see Fig. 1.6). This vision prompted the author to study the optimal design of
submerged domes. Before tackling the aforementioned problem herein, a literature
survey on design of rotational shells is presented.
Fig. 1.3 Yumemai floating bridge
(Source: http://www.tokyo-wankou.com/)
4
Fig. 1.4 Mega-Float in Tokyo Bay
(Photo courtesy of Prof E. Watanabe - Kyoto University)
Fig. 1.5 Floating oil storage facility
(Photo courtesy of Dr Namba - Shipbuilding Research Centre of Japan)
5
Fig. 1.6 Author’s impression of a submerged dome complex
6
1.2 Rotational shells
In thin shell structures used in engineering practice, rotational shells or domes have the
widest application because of their elegance and strength. Large span vaults of
revolution, chiefly as the roofs of sacred buildings, were built in ancient times without
any strength calculations being used. Of course, the domes of stone or brick
constructed those days were many times thicker than the thin shells of buildings,
aircraft and naval structures built over the past forty years based on suitable analytic
methods.
The classical thin shell theory was firstly developed by Aron (1874). However, in
1888, Love (1888) noticed Aron’s inaccuracies and proposed a shell theory that is
analogous to the plate theory proposed by Kirchhoff (1876). Galerkin (1942) also
played an important part in the development of the theory of thin shell by his work.
Goldenweizer (1946) and Mushtari (1949) gave the basis for a general principle for
simplification of the equations of theory of shells.
The above general thin shell theory of shells was preceded by the momentless or
membrane theory. Membrane theory was firstly used in 1833 by Lame and Claperon
(1833). In this work, Lamé and Claperon (1833) considered the symmetrical loading of
shells of revolution. Beltrami (1881) and Lecornu (1938) established the general form
of the equations of membrane theory. Sokolovskii (1938) made a significant
contribution by reducing the equations of the problem to canonical sform and revealed
a number of their characteristic properties. Moreover, Vlasov (1939) Sokolovskii
(1938) investigated the shell of revolution under arbitrary loads. So far, a brief mention
of thin shell theory and membrane theory for thin shell structures is given. In this next
part, a literature review on buckling analysis of the rotational shells will be presented.
7
1.3 Buckling of rotational shells
Shell structures are efficient three-dimensional entities that are capable of resisting
high compressive stresses with essentially little or no bending deformation. Their
inherent efficiency, coupled with elegant shapes and geometry, often results in
thicknesses that are small compared to their span length. Owing to their relatively
small thickness when compared to the length dimensions, the design strength of these
structures is commonly governed by their buckling capacities. Buckling is a
phenomenon in which a structure undergoes visibly large transverse deflection in one
of the possible instability modes. Buckling of a structural component may affect the
strength or stiffness of the whole structure and even triggers unexpected global failure
of the structure. Therefore, it is important to know the buckling capacities of structures
in order to avoid premature failure.
The first notable buckling analysis of shell structures was carried out by Zoelly in
1915 for spherical caps under uniform external pressure. While earlier investigations
mainly centered on the provision of analytical solutions, later approaches relied more
on numerical techniques as facilitated by the advent of modern computers. Bushnell
(1976, 1984) developed a general-purpose computer program for the analysis of shells
of revolution based on the finite-difference method. At about the same time, Cohen
(1981) developed a computer code FASOR, based on a numerical integration method
called the field method, for the analysis of stiffened, laminated axisymmetric shells.
By using the Kalnins and Lestingi (1967) method of multi-segment integration,
Uddin (1987) solved the governing differential equations for axisymmetric buckling of
spherical shells. In Uddin’s (1987) paper, numerical results were presented for
spherical shells with various subtended angles and these results were in good
agreement with those obtained by Huang (1964), Budiansky (1959), Thurston (1961)
8
and Dumir (1984). Chao et al. (1988) presented a semi-analytical solution for
axisymmetric buckling of thick, orthotropic, complete spherical shells and
hemispherical shells with various boundary conditions. Their solutions were derived
from the Ritz method with the displacement functions approximated by Legendre
polynomials. Muc (1992) presented the buckling analysis of axisymmetric composite
shells of revolution such as spherical caps, torispheres and hemispheres. Its first part is
devoted to linear buckling analysis in order to determine the appropriate divisors for
buckling pressures. Uddin and Haque (1994) also investigated the buckling behavior of
semi-ellipsoidal shells, where the critical buckling pressure was found to increase with
increasing ratio of minor axis to major axis lengths of the ellipsoidal shell, and the
critical pressure was found to increase with increasing thickness-to-radius ratios.
Other notable contributions on this subject were made by Ross and his colleagues.
In 1981, Ross and Mackney (1983) presented a constant meridional curvature element
for the buckling of hemi-ellipsoidal domes under uniform external pressure. In this
study, only linear variations were assumed for the meridional and circumferential
displacements along the meridian of these elements. Ross (1990) presented a varying
meridional curvature element to extend this study. Furthermore, Ross (1996) extended
this work to a cubic and a quadratic variation being assumed for the meridional and the
circumferential displacements along the meridian of these elements. In this study,
comparisons were made between experiment and theory for both buckling and
vibration of hemi-ellipsoidal shell domes, which varied from very flat oblate vessels to
very long prolate vessels. In general, agreement between experiment and theory was
good for the hemi-spherical dome and the prolate vessels, but not very good for the flat
oblate vessels. Ross et al. (2001) conducted many experiments on buckling, post-
buckling and plastic collapse of spherical shells subjected to external pressure. Ross et
9
al. (2003) reported on a theoretical and an experimental investigation into six GRP
hemi-ellipsoidal dome shells, which were tested to destruction under external
hydrostatic pressure.
As another attempt, Redekop (2005) developed a new method to predict the
buckling characteristics of an orthotropic shell of revolution with an arbitrary meridian
subjected to a normal pressure. The solution was given within the context of the
linearized Sanders–Budiansky shell buckling theory and makes use of the differential
quadrature method. Dumir et al. (2005) investigated the axisymmetric buckling
analysis of moderately thick laminated shallow annular spherical cap under transverse
load. Buckling under central ring load and uniformly distributed transverse load,
applied statically or as a step function load, was presented.
Recently, applying the boundary element formulation, Baiz and Aliabadi (2007)
presented the buckling analysis of shear deformable shallow shells. The boundary
element formulation is presented as an eigenvalue problem, to provide direct
evaluation of critical load factors and buckling modes.
However, their studies were confined to the treatment of spherical shells, and their
formulations were based on either classical thin shell theory or shallow shell theory. A
literature survey conducted as part of this study indicated that previous treatments of
moderately thick rotational shells had all assumed the specific shape of spherical
shells, limiting their general applications. The methodology developed herein for
buckling analysis is applicable to rotational shells of any meridional shape.
10
1.4 Optimal design of domes against buckling
Over the past four decades, structural optimization has widened considerably, but
optimization to enhance the elastic buckling resistance of structures remains an active
area of research. Structural optimization for problems with buckling constraints is
complicated because calculation of buckling loads is generally an involved process that
requires the solution of two boundary value problems (static analysis and eigenvalue
solution) at each optimization step. While earlier investigations mainly centered on the
provision of analytical solutions, later approaches have relied more on numerical
techniques as facilitated by the advent of modern computers. Buckling of general
rotational shells depends on many variables, such as the geometric properties of the
shell, the material properties and the type of the applied loads. The various parameters
change the buckling behavior of shells, making it difficult to achieve a general optimal
design. Many techniques have been used for optimal design of shells under stability
constraints. A detailed survey of these problems was given by Krużelecki and
Życzkowski (1985) and Życzkowski (1992). The monograph by Gajewski and
Życzkowski (1988) was devoted to structural optimization under stability constraints.
The largest number of papers is concerned with optimization of cylindrical shells.
A lateral pressured cylindrical shell was considered by Hyman (1971), Sun and Hansen
(1988), Sun (1989), Levy and Spillers (1989) and Gajewski (1990). More complex
optimization problems are presented in shells with a double curvature. In this case, a
single loading already causes a combined state of stress. Parametrical optimization of
barrel shaped shells under stability constraints was presented by Blachut (1987),
Krużelecki and Trzeciak (2000). As another attempt, the monograph by Hinton et al.
(2003) was devoted to the buckling analysis and optimization of plates and shells.
11
Research on optimization of shell structures has been extended to multi-objective
optimization. Multi-objective optimizations of cylindrical shells under torsional, axial,
external and internal pressure have been carried out by Sun and Hansen (1988) and
Tennyson and Hanse (1983). Walker et al. (1995) studied the Pareto optimal design of
a symmetrically laminated shell with the objectives defined as the maximization of the
axial and torsional buckling loads.
So far, little work has been done on the multi-objective optimization of submerged
domes against buckling. Prompted by this fact, we focus our study on the Pareto
optimal designs of submerged domes with allowance for selfweight.
1.5 Objectives and scope of study
This thesis investigates the optimal designs of submerged dome structures. First we
consider the least weight design of rotationally symmetric shells. In particular, we
consider
• Submerged spherical domes of uniform strength design governed by the
Tresca yield condition - Based on a family of uniform strength designs associated
with a given depth of water and the dome’s base radius, the optimal subtended angle,
the optimal dome height and optimal thickness variation for the minimum weight
design of submerged spherical domes are determined.
• Submerged general domes adopting constant strength design - Based on a
family of constant strength designs associated with a given water depth and dome
height, the optimal dome shape and the optimal thickness variation for minimum
weight are determined.
12
In the second part of the thesis, we focus our attention on the optimal design of
rotational shells against buckling. For this research study, we first formulate the
buckling problem and derive the governing eigenvalue equation using the Ritz method.
The Ritz computer code for the buckling analysis is developed which can readily
handle any support edge condition. The buckling analysis and problems considered are
given below.
• Buckling analysis of moderately thick domes under uniform pressure using
the Ritz method - The Ritz method was applied to determine the critical uniform
buckling pressures of moderately thick, rotational orthotropic shells that include
spherical, parabolic shells.
• Buckling analysis of moderately thick submerged domes using the Ritz method
- The Ritz method was applied to determine buckling load of submerged domes or the
maximum water depth that a rotational shell can sustain before buckling occurs. Next
we solve the optimal design problem of submerged domes against buckling as well as
for minimum weight and maximum enclosed airspace.
• Optimal design of submerged domes – The Pareto optimal design of
submerged domes for minimum weight as well as maximum enclosed airspace
whereby the dome will not buckle under the hydrostatic pressure and its own weight is
investigated.
Results of the present study are useful in providing a basic knowledge for
constructing a submerged dome that will be used to create a living environment under
the sea. Moreover, the study may contribute to a better understanding of the buckling
behaviour of shell structures under rotationally axisymmetric loads and hydrostatic
pressure.
13
The thesis focuses on identifying the optimal design of submerged spherical and
general dome structures. It is recognized that there are many criteria in designing a
submerged dome structure such as strength, buckling, vibration, wave, current and
blast effect. The current study only investigates the first two criteria, namely strength
and buckling criteria. Moreover, during the analysis, since the bending stress in thin
shell structures is negligibly small, we consider only domes under membrane stress
conditions. Future studies may be carried out to investigate the other criteria for
optimal design and also to investigate the bending of submerged domes under wave
and current loads.
1.6 Layout of thesis
The background information on shell structures, literature review on buckling of shells
of revolutions, the objectives and scope of study have been presented in this chapter.
In Chapter 2, the membrane analysis and minimum weight of the submerged
spherical domes are investigated. In addition to the hydrostatic pressure, loads acting
on the dome include the selfweight and a skin cover load. Based on a family of
uniform strength designs associated with a given depth of water and the dome’s base
radius, we determine the optimal subtended angle 2α (and the optimal dome height)
for the minimum weight design of submerged spherical domes.
In Chapter 3, membrane analysis and optimal design of submerged domes is
considered. In addition to hydrostatic pressure, the domes are also subjected to
selfweight and skin cover load, which are invariably present in this type of structure.
Based on a family of constant strength designs associated with a given water depth and
dome height, the optimal dome shape for minimum weight is determined.
14
By adopting Mindlin shell theory, the energy functionals and governing equations
are derived in Chapter 4 for the elastic buckling analysis of moderately thick rotational
shells under any rotational symmetric loading. Moreover, detailed formulations of the
Ritz method for the buckling analysis are also presented.
Chapters 5 and 6 present buckling analyses of domes of various shapes (such as
spherical and parabolic domes) and under different loading conditions. In Chapter 5,
the buckling problem of rotational shells under uniform pressure is treated whilst
Chapter 6 considers the buckling problem of submerged rotational shells. The validity,
convergence and accuracy of the Ritz solutions are demonstrated using spherical shells
(a special case of rotational shells), where closed-form solutions exist for some cases.
A parametric study is conducted to study the buckling behaviour of spherical and
parabolic domes with respect to the base-radius-to-height ratios, thickness-to-height
ratios and different support conditions. The buckling solutions are presented for the
first time for these shells.
In Chapter 7, the optimal design of the submerged rotational shells (such as
spherical and parabolic domes) against buckling is investigated. Based on a family of
spherical and parabolic domes associated with a given dome height, we investigated
the Pareto optimal dome shape for minimum weight as well as maximum enclosed
airspace whereby the dome will not buckle under the hydrostatic pressure and its own
weight.
Chapter 8 summarizes the findings of this research study and presents some
recommendations for future studies.
15
CHAPTER 2
UNIFORM STRENGTH DESIGNS OF
SUBMERGED SPHERICAL DOMES
This chapter is concerned with the membrane analysis and minimum weight design of
submerged spherical domes. In addition to hydrostatic pressure, loads acting on the
dome include the selfweight and a skin cover load. By adopting a uniform strength
design as governed by the Tresca yield condition, the variation of the shell thickness of
spherical domes can be accurately defined by a power series. Based on a family of
uniform strength designs associated with a given depth of water and the dome’s base
radius, we determine the optimal subtended angle 2α (and the optimal dome height)
for the minimum weight design of submerged spherical domes.
16
Chapter 2: Uniform Strength Designs of Submerged Spherical Domes
2.1 Introduction
In 1958, Ziegler (1958) investigated the uniform strength design of spherical cupolas
under their own weight. Using the Tresca yield hexagon, he found that if the stress
point is restricted to the sides AB and AC of the Tresca hexagon (see Fig. 2.5), the
cupola uses less material than when the stress point is confined only to the Tresca side
AB. Issler (1964) considered membrane shell designs based on the Tresca hexagon as
well as on the von Mises ellipse. He treated shells under constant vertical dead load
per unit projected area. Schumann and Wuthrich (1972) and Sayir and Schumann
(1972) studied membrane shells without rotational symmetry. Prager and Rozvany
(1980) investigated the optimal design of spherical cupolas of a given base radius. The
cupolas are assumed to be constructed from a material with negligible tensile strength.
The combined action of the weights of the cupola proper and a cover of uniform
thickness was considered and the minimum weight design was examined. Nakamura et
al. (1981) extended Prager and Rozvany’s (1980) work to include the weight of the
roof cover, snow load, external and internal pressure. Moreover, Pesciullesi et al.
(1997) obtained the shape of a uniform strength shell subjected to selfweight by
solving the eigenvalue problem associated with the integral equilibrium equations. So
far, the aforementioned studies on spherical domes do not include hydrostatic pressure.
This prompted us to study the membrane and minimum weight design of spherical
domes under hydrostatic pressure.
17
2.2 Membrane Theory
2.2.1 Basic assumptions of classical thin shell theory
In formulating the classical thin shell theory, the following assumptions are made
(Love 1888)
• The shell thickness h is negligibly small in comparison with the smallest
radius R of curvature of the middle surface. According to Novozhilov (1943), the
thickness to radius ratio h / R ≤ 1 / 20 should be satisfied for the shell to qualify as a
thin shell.
• Strains and displacements that arise within the shell are small. This implies
that the products of deformation quantities occurring in the development of the theory
may be neglected, ensuring that the system is described by a set of geometrically linear
equations. This also makes it possible to formulate the equilibrium conditions of the
deformed middle surface with reference to the original position of the middle surface
prior to deformation.
• Straight line that are normal to the middle surface prior to deformation remain
straight and normal to the middle surface during deformation, and experience no
change in length. It implies that the direct strain in the direction normal to the middle
surface, and the shearing strains in planes perpendicular to the middle surface and due
to transverse shear forces, are all zero. This assumption is valid for thin shells.
However, when the shell is thick, it is necessary to incorporate the effect of transverse
shear deformation.
• The normal stresses σ z transverse to the middle surface are small, and can be
neglected.
18
2.2.2 Geometrical properties of rotational shells
Consider a surface of revolution generated by rotation of a plane curve about an axis in
its plane as shown in Fig. 2.1. A point on the shell can be located by the θ - φ - r
coordinate system where r0 is the distance of one point on the shell to the axis of
rotation, and
r0 = r2 sin φ (2.1)
O
Shell axis
θ Parallel
r dθ
A
C
B D r 2
r1
Meridian dφ
Fig. 2.1 Rotational shells (Domes)
Referring to Fig. 2.2, s is the arc length along the meridian as measured from the
apex of the dome, r1 is the radius of curvature of the meridian. The principal radius r2
generates the middle surface of the dome in the direction perpendicular to the tangent
on the meridian. Referring again to Fig. 2.2, for the line element ds of the meridian, we
have
19
ds = r1 dφ (2.2a)
dr0 = ds cos φ (2.2b)
dz = ds sin φ (2.2c)
O
s
Generating (meridional) curve
φ
r
φ r2 dz
ds
r1 dr
dφ
Fig. 2.2 Meridian of a dome
2.2.3 Membrane analysis
For membrane theory to be valid in the analysis, the following conditions must be
satisfied:
• The middle surface of the shell is continuously curved and the curvatures are
slowly varying.
• The thickness of the shell is small and constant or varies continuously.
• Surface loadings are distributed continuously.
• The boundary forces and reactions of the boundary constraints are oriented
tangentially to the middle surface.
20
• The components of the state of displacement determined from the respective
equation are finite.
O
θ dθ
Nφ
Nθ Nφθ
Nθφ
pθ
pφ ΟN
Nθ + φ θ dφ
pn Ο
ΟN
Nθφ + θ θφ dθ
Ο
ΟN ΟNφθ r2
N φ + φφ dφ Nφθ + dθ
Ο Οθ
φ
dφ
Fig. 2.3 Shell element
In general, the shell element is bounded by two meridional lines and two
circumferential lines, each pair of aforementioned curves is close together as shown in
Fig. 2.3. The conditions of its equilibrium will furnish three equations, which are
necessary for solving the three unknown stress resultants, namely, the meridian force
N φ , the hoop force N θ and the shear force N φθ . These three equilibrium equations
are given by Kraus (1967)
∂ (rN φ ) ∂N θφ
+ r1 − r1 N θ cos φ + pφ rr1 = 0 (2.3)
∂φ ∂θ
∂ (rN φθ ) ∂N θ
+ r1 + r1 N θφ cos φ + pθ rr1 = 0 (2.4)
∂φ ∂θ
21
Nφ Nθ
+ = − pn (2.5)
r1 r2
By noting the relationship between s and φ , i..e. ds = r1 dφ , the equilibrium
equations may also be expressed as
∂ (rN s ) ∂N θs
+ − N θ cos φ + p s r = 0 (2.6)
∂s ∂θ
∂(rN θs ) ∂N θ
+ + N θs cos φ + p s r = 0 (2.7)
∂φ ∂θ
N s Nθ
+ = − pn (2.8)
r1 r2
In the case of axisymmetry loads, the stresses are independent of θ . Therefore,
we have two equations to evaluate the two unknown stress-resultants N φ and N θ
∂ (rNφ )
− r1 Nθ cos φ + pφ rr1 = 0 (2.9a)
∂φ
Nφ Nθ
+ = − pn (2.9b)
r1 r2
or
∂ (rN s )
− Nθ cos φ + ps r = 0 (2.10a)
∂s
N s Nθ
+ = − pn (2.10b
r1 r2
22
2.3 Problem definition and basic equations
2.3.1 Problem definition
Consider a submerged, spherical dome of radius R, subtended angle 2α and specific
weight γ a as shown in Fig. 2.4.
h(φ ) O
D
H
φ α
L L
Fig. 2.4 Submerged spherical dome
From geometrical considerations, the base radius L = R sin α and the dome
height H = R (1 − cos α ) = L(1 − cos α ) / sin α . The dome is subjected to hydrostatic
pressure, its own selfweight and skin cover load. The loads are assumed to be
transmitted through the dome structure to the supporting ring foundation via membrane
forces only. By adopting a uniform strength design governed by the Tresca yield
condition (see Fig 2.5), the problem at hand is to seek the variation of the dome
thickness h. From a family of such uniform strength designs associated with a
23
prescribed value of base radius L, we determine the optimal value of α (and hence the
optimal shape) which gives the minimum weight of the submerged dome.
σθ
D E
σ0
C σφ
F
σ0
A B
σ0 σ0
Fig. 2.5 Tresca yield condition
2.3.2 Basic equations
Consider a uniform strength design of a spherical shell under hydrostatic pressure,
selfweight and skin cover load. Assuming the dome to carry the load to the foundation
via membrane forces, we seek the variation of the shell thickness h with respect to the
meridian angle.
F
dF dF
f f
a a
Nφ Nφ
R
dφ
φ α
Fig. 2.6 Free body diagram of dome above horizontal plane a-a
24
Referring to Fig. 2.6, the vertical resultant force F on a free body of the spherical
dome above the horizontal plane a-a (defined at angle φ ) is given by
F = 2πN φ Rsin 2φ (2.11)
where N φ = −σ φ h is the meridian force per unit length and σ φ the meridian stress.
The incremental vertical force dF corresponding to the incremental angle dφ is
given by
dF = f 2πR 2 sinφdφ (2.12)
where f is the vertical force acting on an elemental strip of the dome (see Fig. 2.6)
and is given by
f = phv + pc + pa = γ w (D + Rcosα − Rcosφ )cosφ + pc + γ a h (2.13)
where p hv , p c , and p a are the vertical components of the hydrostatic pressure, the
skin cover load, and the selfweight, respectively, γ a is the specific weight of the dome
material, γ w the specific weight of water and D the depth of water.
By taking total differential of Eq. (2.11) and noting that the meridian force is
N φ = −σ φ h , Eq. (2.12) can be re-written as
( )
− σ φ h / sinφ + 2σ φ hcosφ = fR (2.14)
25
where the prime denotes differentiation with respect to φ . By substituting f given in
Eq. (2.13) into Eq. (2.14), one obtains
( )
− σ φ h / sinφ + 2σ φ hcosφ = [γ a h + pc + γ wcosφ (D + Rcosα − Rcosφ )]R (2.15)
Consider the fully stressed state of the dome material without tensile strength, i.e.
the stress point lying on side AC of the Tresca hexagon. The meridian and
circumferential stresses at this stress state are
σ φ = −σ 0 ; 0 ≥ σ θ ≥ σ φ (2.16)
where σ 0 is the allowable compressive stress and σ θ the circumferential stress.
By applying Eq. (2.16) to Eq. (2.15) and dividing by (σ 0 h0 ) , one obtains
h / sinφ = h (R − 2cosφ ) + p c R + γ w R cosφ (D + R cosα − R cosφ ) (2.17)
where the non-dimensional terms with over-bars are given by
σφ σ
σφ = , σ θ = θ = βσ φ , (2.18a-b)
σ0 σ0
h γ σ
h= , γ w = w2 0 , (2.18c-d)
h0 γ a h0
Rγ a
R= , (2.18e)
σ0
26
Dγ a pc
D= and p c = (2.18f-g)
σ0 γ a h0
Note that Eq. (2.17) contains the non-dimensional pressure p c which can be
eliminated as follows. The condition of equilibrium in the normal direction furnishes
(Ugural, 1999)
N φ + N θ = − Rp n (2.19)
where p n is the normal load component per unit area of the middle surface and is
positive when acting inwards. The normal load consists of components from the
hydrostatic pressure, dome selfweight and skin cover load and is given by
p n = (γ a h + p c )cosφ + γ w (D + Rcosα − Rcosφ ) (2.20)
From Eqs. (2.16), (2.18), (2.19) and (2.20), the ratio of the circumferential stress
to the meridian stress is given by
σ θ (h + p c )R cosφ + γ w R (D + R cosα − R cosφ )

β= = −1 (2.21)
σφ h
where 0 ≤ β ≤ 1 in order to ensure that the stress state condition (2.16) is satisfied.
In axisymmetric domes of revolution, the stresses σ φ and σ θ at the apex (φ = 0)
must approach the same limit, i.e.
27
− σ θ = −σ φ = 1 , at the apex φ = 0 where h = 1 (2.22)
By imposing the above limiting condition, Eq. (2.20) can be written as
p c R = 2 − R [1 + γ w (D + R cosα − R )] (2.23)
and by substituting p c given in Eq. (2.22) into Eq. (2.17), we have
( )
h / sinφ + h (2cosφ − R ) + γ w R (D + R cosα )(1 − cosφ ) − γ w R 2 1 − cos 2φ + R − 2 = 0 (2.24)
Equation (2.24) is the governing differential equation for determining the
thickness variation of the submerged spherical dome of uniform strength.
Note that for the special case of zero hydrostatic pressure (i.e. γ w = 0 ), Eq. (2.24)
reduces to
h / sinφ + h (2cosφ − R ) + R − 2 = 0 (2.25)
which corresponds to the expression derived by Prager and Rozvany (1980).
28
2.4 Results and discussions
2.4.1 Analytical solution using power series method
In order to solve the foregoing first-order ordinary differential equation (2.24), we
represent the function h (φ ) in the form of a power series:
n
h (φ ) = c0 + c1φ 2 + c 2φ 4 + c3φ 6 + ... + c nφ 2 n = ∑ ciφ 2i (2.26)
i =0
The differentiation of the normalized dome thickness h (φ ) with respect to the
meridian angle φ gives
n
h / (φ ) = 2c1φ + 4c 2φ 3 + 6c3φ 5 + .... + 2nc nφ 2 n −1 = ∑ (2i )ciφ 2i −1 (2.27)
i =1
In order to facilitate the analytical solution, trigonometric functions of Eq. (2.24)
are also expressed in the form of power series, i.e.
φ3 φ5 φ7
sinφ = φ − + − + ... ; (2.28a)
3! 5! 7!
φ2 φ4 φ6
cosφ = 1 − + − + ... (2.28b)
2! 4! 6!
By substituting the power series given in Eqs. (2.26) to (2.28) into Eq. (2.24) and
then comparing the coefficients, one obtains the following recursive formula for the
coefficients ci
29
⎞ ⎡ (− 1) γ w (D + R cosα )R (− 1) 2 γ w R i 2(− 1) j +1 (i + j + 1)c ⎤

i i +1 2 i −1 2
⎛ 1
+∑
i− j
ci = ⎜ ⎟⎢ + ⎥
⎝ 2i + 2 − R ⎠ ⎢⎣ (2i )! (2i )! j =1 (2 j + 1)! ⎥⎦
(2.29)
with c 0 = 1 . With these known coefficients ci , the variation of normalized dome
thickness h (φ ) is fully described by Eq. (2.26). For given values of σ 0 , pc, γ a , γ w , R,
D and α , and noting that p c = pc (γ a h0 ) , we can use Eq. (2.23) to calculate the
thickness of the dome at the apex h0 , as
p c R + γ w R (D + Rcosα − R ) p c R + γ w R (D − H )
h0 = = (2.30)
2σ 0 − Rγ a 2σ 0 − Rγ a
2.4.2 Accuracy of Analytical Solution for Dome Thickness
The analytical solution for the normalized shell thickness is furnished by Eqs. (2.26)
and (2.29). In order to check the correctness of the analytical solutions, we can solve
the problem independently by integrating numerically the first-order ordinary
differential equation (2.24) using the Runge-Kutta method (Kreyszig 1993). For this
test, a dome with dimensions R = 500cm, D = 4000cm, α = 2.5 radians and material
properties σ 0 = 75 kgf/cm 2 , γ a = 0.0024 kgf/cm 3 , and the skin cover load
p c = 0.5 kgf/cm 2 are assumed. The analytical solution is computed using exponents of
n = 5, 7 and 9 while a very small step size of Δφ = 0.0001 rad is used for the fourth-
order Runge-Kutta method to ensure a high accuracy.
30
Runge-Kutta solution
1.8 Series solution (n=9)
Series solution (n=7)
1.6 Series solution (n=5)
h
1.4
1.2
1
0.5 1 1.5 2 2.5
φ (rad)
Fig. 2.7 Thickness variation obtained by series and numerical methods
The variations of normalized shell thickness h with respect to the meridian
angle φ , obtained from both the analytical and numerical methods, are shown in Fig.
2.7. It can be seen that the thickness of the submerged dome is characterized by a fairly
rapid increase in the large meridian angle range (e.g. φ > 2 radians). The shell
thickness at φ = 2.5 radians is almost twice as thick as the shell at the apex. A
comparison between the analytical and numerical solutions also indicates that the
analytical solution agrees well with the numerical solution for all three exponents in
the small meridian angle range i.e. 0 ≤ φ < 2 radians. A slight divergence of the
analytical solution is noted in the large meridian angle range. However, an increasing
order of the exponent gives rise to a better agreement with the numerical solution. For
this example, an exponent of n = 9 gives an approximate solution that is sufficiently
close to the numerical solution. This test establishes the correctness of the analytical
solutions and for practical applications, it will be assumed that a power series with 9
terms is sufficient for estimating the shell thickness. This number of terms will thus be
used for all calculations following.
31
2.4.3 Critical Value of Subtended Angle
In order to satisfy the stress state condition (2.16), we have to ensure that the condition
0 ≤ β ≤ 1 is observed for the entire range 0 ≤ φ ≤ α . This implies that for a given water
depth D and base radius L (or dome radius R ), there is a critical subtended angle α cr
(or a critical dome height H cr = R(1 − cos α cr ) = L(1 − cos α cr ) / sin α cr ). The α cr value
is evaluated by setting β = 0 , φ = α cr and α = α cr in Eq. (2.24). This results in the
transcendental equation:
⎛ n ⎞
(R cos α cr − 1)⎜ ∑ ciα cr2i ⎟ + p c R cos α cr + γ w R D = 0 (2.31)
⎝ i =0 ⎠
from which α cr is to be evaluated.
2.4
(α > α cr) region
(α = α cr) curve
2.2
α (rad) 2
D
1.877
H
1.8
αcr
1.6
1.4
0.02 0.04 0.06 0.08 0.1 0.12
D = Dγ σ s 0
Fig. 2.8 Variation of α cr with respect to water depth D
32
As an example in finding α cr , consider a spherical dome of R = 500 cm under
various water depths D ranging from 625 cm to 4000 cm. The assumed material
properties are σ 0 = 75 kgf/cm 2 , γ a = 0.0024 kgf/cm 3 and the skin cover load is
taken as p c = 0.5 kgf/cm 2 . By solving Eq. (2.31), we obtain the variation of α cr with
respect to the non-dimensional water depth D as shown in Fig. 2.8. For a given water
depth D , the subtended angle α of the dome has a critical value α = α cr to ensure
that there is no tensile stress region in the entire dome structure. Domes shapes with
α > α cr in shaded region of Fig. 2.8 have the tensile stress region in the lower base of
domes. For example, if the dome is submerged in a water depth of D = 0.04 , the
maximum subtended angle that the dome can have is α cr = 1.877 radians. Beyond this
α cr value in the shaded area, a tensile stress region will appear in the lower base of the
dome. It can be seen that as α cr increases with water depth D, the dome is restricted to
a flatter profile.
2.4.4 Effect of water depth on thickness variation
It is clear that the thickness variation of the submerged domes depends on the water
depth, the selfweight and the skin cover load. In this section, we study the effect of the
water depth D on the thickness variation when the dome shape is defined by a given
subtended angle α . Figure 2.9 shows the normalized thickness h = h / h0 variations of
submerged hemispherical domes with R = 500 cm and α = π / 2 radians for various
values of D . It can be seen that the normalized thickness at the base of the dome is
relatively sensitive to the water depth. A larger normalized thickness ratio at the base is
associated with a shallow water depth.
33
1.7
1.6
D = 0.03
D = 0.05
1.5
D = 0.128
h 1.4
1.3
1.2
1.1
1
0.2 0.4 0.6 0.8 1 1.2 1.4
φ (rad)
Fig. 2.9 Thickness variations of submerged domes for various water depths
2.4.5 Minimum weight design
For a given water depth D and a base radius L, there is a family of uniform strength
designed domes. Each dome is associated with a subtended angle α (or dome height
H = L(1 − cos α ) / sin α ) as shown in Fig. 2.10. However, there is a minimum weight
solution within this family of solutions that we want to seek because of its practical
importance. This optimal solution is associated with the optimal subtended angle α opt .
For a given base radius, the non-dimensional optimal height H opt = H opt γ a / σ 0 of the
submerged dome is related to α opt by
L
H opt = (1 − cos α opt ) (2.32)
sin α opt
34
D
H
L L
Fig. 2.10 Family of uniform strength designed domes
for a given base radius L
The objective function of this optimization problem is defined by the selfweight
of the dome. The selfweight W0 can be calculated directly by integrating the product of
h and the surface area of the dome, i.e.
α
W0 = γ a ∫ h.2πRsinφ .Rdφ (2.33)
0
or in normalized form as
α
W0 2
πpc L2 p c sin 2α ∫0
W0 = = h sinφdφ (2.34)
35
0.4
0.3 L = 0.01
L = 0.02
L = 0.04
W0 0.2
W0min
0.1
0
0.25 0.5 0.75 1 1.25 1.5 1.75
α (rad)
Fig. 2.11 Variations of weight W0 with respect to subtended angle α

for L = 0.01, 0.02 and 0.04
The variations of W0 with respect to the subtended angle α are shown in Fig.
2.11 for a given water depth D = 1500 cm for various base radii L = 0.01, 0.02 and
0.04. The variation of W0 is rather small over a wide range of α , especially when
the base radius of the dome is small. The insensitivity of the dome weight for
1 radian ≤ α ≤ 1.25 radians (which contains the optimal of subtended angle) is good
news for engineers as it means that there is some flexibility when designing the dome
shape without compromising too much on the optimum weight. It can be seen that the
optimal subtended angle α opt is about 1.1 radians.
Using the thickness variation h , as given by Eq. (2.26), one can obtain the
optimal value of the subtended angle α opt for a minimum value of W0 by a simple
minimization technique such as the Golden Section Search technique (Kreyszig 1993).
Figure 2.12 shows the values of α opt and Womin with D = 5000 cm for a wide range
36
of practical base radii L (i.e. 0.01 ≤ L ≤ 0.055). It can be seen that α opt varies in a
narrow range of 1.04 radians ≤ α opt ≤ 1.14 radians.
(rad) W0min
opt
1.14
0.4
1.12
0.3
1.1 W0min
αopt
0.2
1.08
1.06 0.1
1.04
0
L
0 0.01 0.02 0.03 0.04 0.05
Fig. 2.12 Variations of minimum weight W0 and α opt with respect to base radius L
The relationship between the optimal dome height and water depth is also of
interest in the minimum weight design of submerged domes. Figure 2.13 shows the
variations of the optimal dome height to base radius ratio H opt / L with respect to the
water depth to base radius ratio D / L . It can be seen that the optimal shape of the
spherical dome gets flatter with increasing water depth, but the optimal height to base
radius ratio varies within a small range ( 0.58 ≤ H opt / L ≤ 0.61 ) for a wide range of
practical water depths ( 5 ≤ D / L ≤ 30 ). For very deep water, the optimal height of the
dome is approximately H opt / L = 1 / 3 or α opt = π / 3 radians.
37
0.61
0.605
0.6
Hopt /L 0.595
0.59
0.585
0.58
5 10 15 20 25 30
D/L
Fig. 2.13 Variation of optimal dome height H opt / L with respect to water depth D / L
38
2.5 Concluding Remarks
In this chapter, we have derived an analytical expression in the form of a power series
for the thickness variation of a submerged spherical dome in a uniform strength design
as governed by the Tresca yield condition. Numerical examples show that the sum of
the first 9 terms in the power series is sufficiently accurately for practical applications.
Further, the optimal subtended angle αopt (and the optimal dome height H opt ) for the
minimum weight design of these domes have been determined. For very deep water, it
was found that the optimal height of the dome is approximately H opt / L = 1 / 3 or
α opt = π / 3 radians.
Although the present chapter considers only spherical domes, the next chapter
will treat non-spherical domes where the aim is to determine the optimal thickness
variation as well as the shape of fully stressed submerged domes for minimum weight.
39
CHAPTER 3
CONSTANT STRENGTH DESIGNS OF
SUBMERGED GENERAL DOMES
This chapter is concerned with the membrane analysis and optimal design of constant
strength submerged domes. In addition to hydrostatic pressure, the domes are also
subjected to selfweight and skin cover load, which are invariably present in this type of
structure. Using membrane theory for thin shells and by adopting a fully stressed
design, equations governing the meridional curve of submerged domes are derived
with allowance for selfweight and skin cover load. The set of nonlinear differential
equations, which correspond to a two-point boundary problem, is solved by the
shooting-optimization method. Based on a family of fully stressed (constant strength)
designs associated with a given water depth and dome height, the optimal dome shape
for minimum weight is determined.
40
Chapter 3: Constant Strength Designs of Submerged General Domes
3.1 Introduction
In 1959, Timoshenko and Woinowsky-Krieger (1959) presented the optimal shape for
domes under hydrostatic pressure only. Royles et al. (1980) pointed out that the
optimal shape of submerged domes is similar to the shape of a sea urchin, which is a
member of marine invertebrates in the phylum Echinodemata. Figure 3.1 shows the
calcareous shell of a sea urchin after its spines have been removed. Due to their
similarity, the optimal shape of fully stressed submerged domes under hydrostatic load
has been referred to as an Echinodome by Royles et al. (1980). So far, little work has
been done on the optimization of fully stressed submerged domes with allowance for
selfweight. Prompted by this fact, we focus our study on the membrane analysis and
the optimal shape of fully stressed domes under selfweight, hydrostatic pressure upon
submergence in water and skin cover load arising from attachments on the domes.
Fig. 3.1 Calcareous shell of a sea urchin
41
3.2 Problem definition and basic equations
3.2.1 Problem definition
Consider a shell of revolution of height H and submerged under water at a depth D.
Figure 3.2 shows the geometry of the submerged dome as defined by its meridian.
Referring to the figure, r0 is the distance from a point on the meridian to the vertical
axis z, which is pointed in the gravity direction, and r1 is the radius of curvature of the
meridian. The principal radius r2 generates the middle surface of the dome in the
direction perpendicular to the tangent on the meridian. Only dome shapes involving
positive values of r1 and r2 are considered. A second coordinate system, defined by the
arc length s along the meridian angle φ , is also shown in Fig 3.2.
r0
s
z h(φ)
D
r2 r1 H
φ
L L
Fig. 3.2 Coordinate systems and parameters defining the shape of submerged dome
42
By adopting a fully stressed design, the problem at hand is to determine the
thickness variation of the submerged dome under hydrostatic, selfweight and skin
cover load. The optimal shape of submerged domes for least weight condition is sought
as part of the solution.
3.2.2 Governing equations for membrane analysis of submerged domes
Consider a dome subjected to three types of loads: (i) hydrostatic pressure
p h = γ w (D − H + z ) , (ii) selfweight p a = γ a h where h is thickness that varies with
respect to the angle φ and (iii) skin cover load pc . The positive direction of these
loads and their distributions are shown in Fig. 3.3. Note that the skin cover load pc in
Fig 3.3(c) is defined as force per unit surface area and is assumed to be constant in this
chapter. However, the skin cover load is indicated as varying in Fig. 3.3c due to the
projection of the skin cover load on the horizontal plane. In deriving the governing
equations for submerged domes, resolution of the forces normal and tangential to the
middle surface is appropriate. In this case, net components of the load’s normal pn and
tangential p s to the middle surface are given by
p n = (γ a h + p c ) cos φ + γ w (D − H + z ) (3.1)
p s = (γ a h + p c )sin φ (3.2)
respectively where γ a is the specific weight of the dome material and γ w is the
specific weight of water.
43
CL γa (D-H+z) CL CL
γa h pc
L L L
(a) Hydrostatic pressure (b) Selfweight (c) Skin cover load

Fig. 3.3 Load components on submerged domes
By considering equilibrium of forces in the shell of revolution in the normal φ -
direction, one obtains the well known equation
Nφ Nθ
+ = − pn (3.3)
r1 r2
where N φ = σ φ h and N θ = σ θ h are the membrane forces in the meridional and
circumferential directions, respectively. Note that unlike typical analyses, which
involve the determination of membrane forces for a defined geometry of the dome, the
task at hand corresponds to the inverse of the problem where the geometry of the dome
is to be determined for prescribed stresses in the two principal directions.
To this end, consider the special case of a fully compressed dome where the
meridian and circumferential stresses take on the same stress value
σ φ = σ θ = −σ 0 (3.4)
44
in which σ 0 is the allowable compressive stress. It is thus implicitly assumed that
buckling of the dome will not occur and that the stress condition defined by Eq. (3.4)
can be achieved under the load combination.
By substituting Eq. (3.4) and Eq. (3.1) into Eq. (3.3), one obtains the condition for
curvature variation of the dome i.e.
1 (γ a h + p c )r2 cos φ + γ w r2 (D − H + z ) − σ 0 h
= (3.5)
r1 σ 0 hr2
In order to facilitate the solution of Eq. (3.5), the following geometrical relations
for the shell of revolution are required
ds
r1 = (3.6a)
dφ
r0 = r2 sin φ (3.6b)
dr0 = ds cos φ (3.6c)
dz = ds sin φ (3.6d)
For generality, the following non-dimensional terms (denoted with over-bars) are
introduced for the geometric and stress parameters
D L
D= , L= , (3.7a-b)
H H
r0 h
r0 = ,h= , (3.7c-d)
H H
45
l z s
l = , z= , s= , (3.7e-g)
H H l
pc γ wH γ H
pc = ,α = , β = a (3.7g-i)
σ0 σ0 σ0
where l is the curve length for one-half of the meridian and s is the arc length along the
meridian as measured from the apex of the dome (see Fig. 3.2). Using the geometrical
relations of Eq. (3.6) and definitions in Eq. (3.7), the following differential equation
governing the shape of the meridian in normalized form can be obtained
dφ
=
[ ]
(β h + pc )r0 cos φ + α r ( D − 1 + z ) − h sin φ l (3.8)
ds hr
In this case, the shape of the meridian, which is defined by the angular change of
the middle surface with respect to the arc length, depends on all three load
components.
Although Eq. (3.8) involves only a first-order differential equation, it must be
solved in combination with the equilibrium condition of the shell in the meridian
direction. To this end, the equation for equilibrium of forces in the s-direction, which is
well known, is given by
d
(r0 N s ) − N θ cos φ = − p s r0 (3.9)
ds
By substituting Eqs. (3.2), (3.4), (3.6) and (3.7) into Eq. (3.9), one obtains the
variation of the meridian thickness in normalized form as
46
= (β h + p c )l sin φ
dh
(3.10)
ds
Thus Eqs. (3.8) and (3.10) correspond to the equilibrium condition for the shell of
revolution subjected to hydrostatic pressure, selfweight and skin cover load. The two
equations may be solved in conjunction with the following normalized geometrical
relations
dr0
= l cos φ (3.11)
ds
dz
= l sin φ (3.12)
ds
which are obtained from Eqs. 3.6(c) and (d) using the definitions in Eq. (3.7). Note that
the shape of the fully stressed submerged dome is characterized in terms of the
Cartesian coordinates i.e. normalized r0 and z , and is parameterized in terms of the
normalized arc length s , which is measured from the apex. An auxiliary result that
forms a part of the solution includes the variation of the subtended angle φ and the
variation of the dome thickness with respect to the normalized arc length s .
47
3.2.3 Boundary conditions for membrane actions in fully stressed submerged
domes
Because substantial forces are developed in the dome due to the combined effects of
hydrostatic pressure, dome selfweight and skin cover load, an adequate foundation
must be provided for the dome in order to ensure its integrity. Typically, domes are
supported by a ring foundation at the base of the dome, where the loads are assumed to
be transmitted to the ring foundation via membrane actions only. However, for the
membrane theory to be valid for the aforementioned problem, the forces acting on the
dome must be in equilibrium with the forces acting on the ring foundation, and the
resulting deformation of the dome and ring foundation must be compatible at their
boundary. In order to eliminate bending in the dome, which is pre-requisite for the
membrane theory used here, the circumferential lengthening of the dome at the base
must be equal to that of the supporting ring foundation (due to the horizontal
component Nrs of meridian force Ns, see Fig. 3.4).
C
L
Nrs
Nzs Ns
φb
Fig. 3.4 Horizontal and vertical components of the meridian force Ns

acting on the ring foundation
48
The stress in the supporting ring is given by Novozhilov (1970)
N rs r0 N L
σk = = − s cos φb (3.13)
Ωk Ωk
where Ω k is the cross-sectional area of the support ring, φb is the subtended angle at
the dome base, L is the radius of the support ring, approximately equal to the radius at
the dome base. The hoop stress at the dome base is equal to
Nθ
σb = (3.14)
hb
where hb is the thickness at the dome base. By equating the hoop strain of the dome
base to the extensional strain of the supporting ring, one obtains
NsL N
− cos φb = θ (3.15)
Ek Ω k Ehb
where E denotes the Young’s modulus of the dome and E k denotes the Young’s
modulus of the supporting ring and the Poisson ratio has been assumed to be equal to
zero for both materials for simplicity. Therefore, the required cross-sectional area of
the ring foundation may be calculated by the formula of Novozhilov (1970)
E Ns
Ωk = − Lhb cos φb (3.16)
Ek Nθ
49
Since the dome is assumed to be fully stressed, the circumferential stress and the
meridian stress are equal to σ 0 and will have the same sign. Since the required area of
the cross-section of the support ring has to be positive, Eq. (3.16) implies that in order
for the dome to be in a fully stressed membrane state
cos φb < 0 (3.17)
The inequality in Eq. (3.17) means that the subtended base angle φb of a fully
stressed submerged dome has to be larger than π 2 , or conversely, a dome with a
subtended base angle φb of less than π / 2 cannot be under a fully compressive stress
state. Note that this condition is independent of the hydrostatic pressure, dome
selfweight or imposed skin cover load.
3.3 Results and Discussions
In this section, numerical solutions for the dome thickness and shape are determined
for two load cases. The first load case corresponds to purely hydrostatic pressure
whereas the second load case consists of hydrostatic pressure, selfweight and skin
cover load.
3.3.1 Weightless constant strength submerged domes
The thickness of the fully stressed dome can be represented as a function of elevation z
by substituting the geometrical relation dz = ds sin φ into Eq. (3.10). This leads to
50
dh
= β h + pc (3.18)
dz
For the case of hydrostatic pressure only, the selfweight and skin cover load are
zero leading to β = p c = 0 . Therefore, Eq. (3.18) reduces to
dh h
= 0 ⇒ h = hc = c = constant (3.19)
dz D
Equation (3.19) implies that, for a fully stressed condition, a momentless dome
has a constant thickness when subjected to hydrostatic pressure only.
In order to obtain the shape of such a submerged dome, one needs only to solve Eqs.
(3.8), (3.11) and (3.12) since the thickness was determined to be constant. Therefore,
the governing equations for this problem are
[ ]
dφ α r0 ( D − 1 + z ) − hc sin φ l
= (3-20 a)
ds hc r
dr0
= l cos φ (3-20 b)
ds
dz
= l sin φ (3.20 c)
ds
For a given water depth D , specific weight of water γ w , dome height H, constant
thickness hc , and an allowable compressive stress σ 0 (note that α = γ w H σ 0 ), there
is a unique shape for the fully stressed submerged dome (Timoshenko and Woinowsky-
Krieger, 1959 and Royles et al., 1980). In order to determine this dome shape, the
foregoing equations 3.20(a-c) are solved together with these boundary conditions that
51
φ (0) = 0 (3.21a)
r0 (0) = 0 (3.21b)
z (0) = 0, z (1) = 1 (3.21c)
The set of nonlinear differential equations and boundary conditions i.e. Eqs.
(3.20)-(3.21) constitute a two-point boundary value problem that can be solved using
the shooting-optimization method as proposed by Wang and Kitipornchai (1992). In
this method, the two-point boundary value problem is first converted into a set of
initial value problems and the differential equations integrated forward by using the
fourth-order Runge-Kutta algorithm (Kreyszig 1993). The only terminal boundary
condition z (1) = 1 to be satisfied can be taken care of by minimizing the objective
function Φ with respect to the curved length l , where Φ is defined by
min Φ = z (1) − 1 (3.22)

l
However, this optimization problem must be subjected to the inequality constraint
φ (1) ≥ π 2 (3.23)
where z (1) and φ (1) = φb are obtained from forward integration of the system of first
order differential equations Eqs. 3.20(a-c). Note that the subtended angle at the base
needs to satisfy the inequality φb ≥ π / 2 , as noted earlier, in order to ensure that the
deformation of the dome at the base is compatible with the circumferential lengthening
52
of the ring foundation. The foregoing optimization problem can be solved using the
generalized reduced gradient code GRG2 (Ladson et al. 1978). In this chapter, the
accuracy of the solution is ensured by taking a very small step size Δs = 0.001 in the
forward integration in the Runge-Kutta algorithm (Kreyszig 1993)
r0
O s
z D
r2
H
φ
Fig. 3.5 Coordinate system for the Runge-Kutta forward integration
As an example, consider a dome of height H = 1000 cm and submerged at
different water depths of D = 2000 cm, 2500 cm, 3000 cm and 3062 cm. The
following values are assumed in the calculation: hc = 10 cm, γ w = 1.10 −3 kgf/cm3 and
σ 0 = 75 kgf/cm2. These values give rise to α = 0.004. Using the shooting optimization
method, the shape of the submerged dome at different depths is plotted in the
normalized coordinates r and z in Fig. 3.6. The final results indicated that the
normalized curve length are l = 1.7240, 1.5355, 1.5071, 1.5732 for these depths. It can
be seen from Fig. 3.6 that the submerged dome changes from a shape that is relatively
53
flat for shallow water to one that is highly curved for deep water. Note that the
maximum water depth for this dome is D = 3.062 and is associated with the limiting
φb = π at its base. This limiting dome shape is often referred to as the Echinodome
shape, which is easily understood by its remarkable resemblance to the shape of the sea
urchin in Fig. 3.1 with the submerged dome shape at D = 3.3062 in Fig. 3.6. Also
note that for the case of D = 2 the slope of the meridian is nearly vertical at the base
of the dome with the angle φb slightly greater than π / 2 .
0
z
0.2
D=2
0.4 D = 2.5
D=3
0.6 D = 3.062
0.8
1.0 r
0
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0 1.2
Fig. 3.6 Weightless fully stressed submerged dome shapes

under various water depths
Although the above problem has been solved and well documented (see for
example Timoshenko and Woinowsky-Krieger, 1959 and Royles et al., 1980), it should
be pointed out that our present formulation and solution technique have the following
advantage. Owing to the complicated shape of the dome, Timoshenko and
Woinowsky-Krieger (1959), Royles et al. (1980) and Sofoluwe et al. (1981) divided
their calculations into three segments when integrating the dome shape in the x (or z)
direction. The three segments, which depend on the value of the angle φ , were
54
necessary in order to avoid an infinite slope being encountered in the numerical
solution. In the method proposed in this chapter, the entire shape of the dome can be
determined without dividing the meridian into different segments since the integration
is carried out using the arc length coordinate s and there is no difficulty in the solution
for infinite slope. Although not explicitly shown in this chapter, the numerical solution
for the shape of a submerged dome under hydrostatic pressure alone compares well
with that presented by Timoshenko and Woinowsky-Krieger (1959) and Sofoluwe et
al. (1981).
3.3.2 Constant strength of submerged general domes
For the second load case, the fully stressed dome is subjected to hydrostatic pressure,
selfweight and a skin cover load. Unlike the case of hydrostatic pressure only, the
dome thickness, which is defined by Eq. (3.10), cannot be uncoupled from the other
three nonlinear equations. Consequently, one has to solve a set of four nonlinear
ordinary differential equations as given by Eqs. (3.8), (3.10), (3.11) and (3.12)
dφ
=
[ ]
(β h + pc )r cos φ + α r ( D − 1 + z ) − h sin φ l (3.24a)
ds hr
= (β h + p c )l sin φ
dh
(3.24b)
ds
dr0
= l cos φ (3.24c)
ds
dz
= l sin φ (3.24d)
ds
where φ , h , r0 , z and l are the unknowns. This system of ordinary differential
equations is solved in conjunction with these boundary conditions:
55
φ (0 ) = 0 (3.25a)
h (0) = h0 (3.25b)
r (0 ) = 0 (3.25c)
z (0 ) = 0, z (1) = 1 (3.25d)
where h0 is the non-dimensional thickness at the apex of the dome. It should be noted
that for this problem, the thickness is no longer constant but varying along the
meridional curve.
Although the number of equations to be solved increases by one for the case of
combined hydrostatic pressure, selfweight and skin cover load, these equations can
also be treated as a two-point boundary value problem and solved using the same
numerical technique described above for hydrostatic pressure only. The terminal
boundary condition is satisfied by minimizing the objective function with respect to
the curved length l
min Φ = z (1) − 1 (3.26)

l
and the optimization problem is subjected to the inequality constraint given in (3.23).
The values z (1) and φ (1) are obtained from forward integration of the system of first
order differential equations Eq. 3.24(a-d). Since the shape of submerged domes
subjected to combined hydrostatic pressure, selfweight and skin cover load has not
been hitherto investigated, it is instructive to examine the influence of pertinent
parameters on the shape of submerged domes.
56
3.3.2.1 Influence of water depth on dome shapes
Studies on submerged arches (Gavin and Reilly, 2000 and Chai and Kunnath, 2003)
indicated that the shape of a momentless arch changes for a changing water depth. For
deep water, the large hydrostatic pressure results in a funicular shape that tends to be
circular, resulting in a small span length of the arch. On the other hand, shallow water
results in a funicular shape that tends to be parabolic, resulting in a relatively long
span. Although the observation was made on the basis of 2D structures, a shape change
under different water depth is nonetheless expected of submerged domes.
In this section, the shape of a fully stressed submerged dome is investigated for
three water depths, namely D = 2000 cm, 2500 cm and 3000 cm. The height of the
dome is taken as H = 1000 cm and the thickness of the dome is taken as h0 = 10 cm at
the apex. The specific weight of water is taken as γ w = 1.10 −3 kgf/cm3 while the
specific weight of the dome material is taken as γ a = 0.0024 kgf/cm3. The uniform
skin cover load is assumed to be pc = 0.1 kgf/cm2, which is significant compared to the
weight of the dome. The dome material is assumed to be uniformly compressed to an
allowable stress of σ 0 = 75 kgf/cm2. The shape of the meridian as obtained using the
shooting optimization technique is plotted in Fig. 3.7 using the normalized Cartesian
coordinates r0 and z . The same step size of Δs = 0.001 is used in this example.
57
0
z
0.2
0.4 D = 2.5
D=3
0.6 D = 3.5
0.8
1.0 r0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Fig. 3.7 Submerged dome shapes under selfweight and skin cover load
for various water depths
It can be seen from Fig. 3.7 that the meridional shape of the dome changes under
different water depths. Although the general shape of the dome under combined
hydrostatic pressure, selfweight and skin cover load is similar to that under hydrostatic
pressure alone, the base horizontal coordinate is smaller in the case of combined
loading. For example, for the same normalized water depth of D = 3, the base
horizontal coordinate is rb = 0.6636 for the combined loading in Fig. 3.7, which is
considerably larger than the base horizontal coordinate of rb = 0.4291 for hydrostatic
pressure alone. The expansion in the base horizontal coordinate indicates the
importance of including the effect of selfweight and skin cover load for determining
the membrane (momentless) shape of the dome. It is also of interest to compare the
dome shapes under different water depths but in combination with selfweight and skin
58
cover load. For the combined loads shown in Fig. 3.7, the base horizontal coordinate of
the meridian reduces from rb = 0.8461 to rb = 0.6636 at the base as the water depth
increases from D = 2.5 to D = 3 . In contrast, the dome exhibits a more significant
reduction in the base horizontal coordinate when subjected to hydrostatic pressure
alone. For the same increment of water depths from D = 2.5 to D = 3 , the horizontal
coordinate at the base reduces from rb =0.7990 to rb =0.4291 as shown in Fig. 3.6. For
the combined loads of hydrostatic pressure, selfweight and skin cover load, the change
in the base horizontal coordinate is 21.6% as compared to 44.9% for the case of
hydrostatic pressure alone. This comparison indicates that water depth exerts a lesser
influence on the shape of fully stressed domes in the presence of selfweight and skin
cover load. It can also be seen that the dome in Fig. 3.7 approaches an Echinodome-
like shape with increasing water depth.
Although not explicitly shown in Fig. 3.7, the thickness of the dome increase
from the apex to the base. The thickness variation with respect to z , which is governed
by Eq. (3.15), depends only on selftweight, skin cover load and dome height, but does
not depend on hydrostatic pressure.
3.3.2.2 Effect of selfweight on dome shapes
In the application of submerged domes, various materials may be used for their
construction. Since the weight of these materials is expected to have an influence on
the membrane shape of the dome, the influence of selfweight is examined through a
numerical example using the normalized selfweight parameter β which is defined as
β = γ a H /σ 0 .
59
0
z
0.2
0.4 β=0
β = 0.1
0.6 β = 0.5
0.8
1.0 r
0
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8

Fig. 3.8 Fully stressed submerged dome shapes with different selfweight parameter β
Consider a fully stressed dome of height H = 1000 cm having an apex thickness
h0 = 10 cm submerged at the water depth D = 3000 cm. The allowable compressive
stress of the dome material is assumed to be σ 0 = 75 kgf/cm2. Three values are
selected for the selfweight parameter for comparison, namely β = 0 which
corresponds to a weightless condition, β = 0.1 and β = 0.5 . Note that the skin cover
load skin is not included in this example i.e. pc = 0 kgf/cm2 so that the effect of
selfweight on the dome shapes can be readily observed. The resulting dome shapes are
plotted in Fig. 3.8, which indicates that the dome is characterized by a slight reduction
in curvature of the meridian for increasing selfweight. In particular, the dome base
radius increases with increased selfweight, which means that the dome shape deviates
from that of the Echinodome shape as the selfweight parameter β increases. It is also
evident from Fig. 3.8 there is a cross-over point where the coordinates of the meridian
remains relatively constant despite the changing values of β . For this example, the
60
cross-over point occurs in the vicinity of r0 = 0.673 and z = 0.690.
3.3.2.3 Optimization of Submerged Domes
In characterizing the shape of fully stressed submerged domes, it is important to note
that, for a given water depth and dome height, the shape of the domes is not unique but
rather consists of a family of curves each of which is associated with a different value
of the subtended base angle φb and dome apex thickness ho. Since each shape in the
family of curves gives rise to a different overall weight of the dome, the variation of
the dome weight with respect to base angle φb is important especially when the
optimal shape of the dome is to be determined. To this end, the weight of the
submerged dome is numerically integrated using the following expression in
dimensionless form
1
W
W = = l ∫ h r0 ds (3.27)
2πγ a H 3 0
61
0.2
D=5
D = 10
0.16
D = 15
Dome weight W
0.12
0.08
0.04
0 π 1.6
1 1.2 1.4 1.8 2
2
Subtended angle φb
Fig. 3.9 Variation of submerged dome weight respect to subtended base angle φb
As an example, consider a dome height of H = 1000 cm, selfweight of
γ a = 0.0024 kgf/cm3, skin cover load of pc = 0.5 kgf/cm2 and an allowable stress of
σ 0 = 75 kgf/cm2. The overall normalized weight of the dome is calculated for three
normalized water depths of D = 5, 10 and 15 and is plotted against the subtended base
angle φb in Fig. 3.9. Note that, in order to ensure compatible deformation between the
ring foundation and the base of the dome, the subtended angle at the base must satisfy
φb ≥ π / 2 as discussed earlier. The feasible solution space for the overall dome weight
therefore lies to the right of the vertical line φb = π / 2 in Fig. 3.9. It can be seen from
the figure that the normalized dome weight decreases monotonically with increasing
values of the subtended base angle φb . Thus the problem of determining the optimal
shape, which is defined by the minimum weight of the submerged dome, is equivalent
to the problem of maximizing the subtended base angle φb . Note that even though the
62
size of the dome decreases with increasing water depth, as seen earlier in Fig. 3.7, the
overall weight of the dome actually increases with increased water depth. The increase
in overall dome weight is due to the increased thickness of the dome.
For a given water depth and dome height, the minimization of dome weight under the
combined hydrostatic pressure, selfweight and skin cover load may be stated
mathematically as
max Φ = φb = φ (1) (3.28)

h0 ,l
and subjected to the inequality constraints
φ (1) ≥ π 2 (3.29)
z (1) − 1 ≤ 10 −7 (3.30)
The inequality in Eq. (3.29) is equivalent to Eq. (3.17) which is to ensure
compatible deformations between the dome at the base and the ring foundation. The
inequality condition in Eq. (3.30) ensures the satisfaction of the terminal boundary
condition z (1) = 1 . It should be noted that the decision variables in the optimization
problem are the apex thickness h0 and the dome’s curve length l . The variation of the
thickness is determined once these two decision variables are known.
63
0
z
0.2
D=2
0.4
D=5
D = 15
0.6
0.8
1.0 r
-0.6 -0.4 -0.2 0 0.2 0.4 0.6
Fig. 3.10 Optimal shapes of submerged domes with respect to water depths
As an example of the optimal shape of submerged domes, the following
parameters are assumed: dome height H = 1000 cm, selfweight γ a = 0.0024 kgf/cm3,
skin cover load pc = 0.5 kgf/cm2 and allowable stress σ 0 = 75 kgf/cm2. The optimal
shape of the dome is shown in Fig. 3.10 for three normalized water depths of D = 2, 5
and 15. It can be seen that the water depth affects the lower half of the dome i.e.
z ≥ 0.5 more so than the upper half of the dome. The optimal shape of the submerged
dome is also characterized by an increased curvature in the lower half of the dome for
increased water depth. The increased curvature for domes submerged in deep water is
accompanied by a reduced base radius. The increased water depth, however, has a
diminishing influence of the optimal shape of the dome. For example, for a change of
water depth from D = 2 to 5 (2.5-folded increase), the base radius reduces from rb =
0.4764 to 0.2808, which represents a 41.1% reduction. On the other hand, for the
change in water depth from D = 5 to 15 (threefold increase), the base radius reduces
64
from rb =0.2808 to 0.1698, which represents a smaller reduction of 39.5%. For
completeness in the presentation of results, Table 3.1 shows the optimal value of the
base angle φb opt , apex dome thickness h0 opt , and curve length l opt where the subscript
opt is used to denote the optimal value. It can be seen that the optimal dome thickness
is significantly increased for a large water depth. For the normalized water depth of
D = 15, the normalized dome thickness at the apex is h0 opt = 0.004364 as compared to
the thickness of h0 opt = 0.000361 for the water depth of D = 2. The increase in
thickness is 12.1 times that of the thickness associated with D = 2. The increased
dome thickness however is accompanied by a slight reduction in the curve length of
the dome. Although the variations of dome thickness may be different for different
water depths, the thicker shells needed for fully stressed domes in deep water
nonetheless results in an increase in the overall weight of the dome.
Table 3.1 Optimal values of base angle φb opt , apex thickness h0 opt ,
and curved length lopt
Water depth Subtended base Apex thickness Curve length

D angle φb opt h0 opt lopt
2 1.6942 0.000361 1.2446
5 2.1044 0.001225 1.2775
15 2.4610 0.004364 1.3603
65
3.4 Concluding remarks
Motivated by recent studies of funicular arches in submarine applications, this chapter
extends the analysis to submerged domes where pure membrane actions are assumed.
Equations governing the geometry of fully stressed submerged domes under combined
hydrostatic pressure, selfweight and skin cover load are derived. These equations
describe the curvature and thickness variation of the dome as well as the Cartesian
coordinates of its meridian. For the special case of a weightless dome without skin
cover load, the thickness of the dome was found to be constant when subjected to
hydrostatic pressure only. The shape of the dome was also found to agree well with the
shape currently reported in the literature.
Although the set of governing equations for submerged domes is highly nonlinear,
the shooting optimization technique currently available in the literature was found to
be well suited for solving this problem. A notable advantage of the equations derived in
this chapter is the parameterization of the equations using the arc length s as measured
from the apex of the dome. Such parameterization allows the entire shape of the
submerged dome to be determined in a single integration process whereas previous
methods cannot determine the Cartesian coordinates of the dome once vertical or
infinite slope is encountered in the meridian. Parametric studies of dome shapes under
different water depths and selfweight also led to an investigation of the optimal shape
of submerged domes. Numerical examples indicated that the airspace enclosed by the
optimal dome reduces in the presence of large hydrostatic pressure. The reduced
airspace is accompanied by a significant increase in the dome thickness, which in turn
results in an increased overall weight of the dome
66
CHAPTER 4
ENERGY FUNCTIONALS AND RITZ METHOD
FOR BUCKLING ANALYSIS OF DOMES
This chapter is concerned with the elastic, axisymmetric buckling analysis of
moderately thick domes under rotational loads. The domes have orthotropic properties
which include the isotropic case as a specialized case. In order to capture the effect of
transverse shear deformation, which is significant for moderately thick domes, Mindlin
shell theory is used. Based on Mindlin shell theory, the energy functional is derived
first. By using the Ritz method, the total potential energy functionals are minimized
with respect to the parameterized admissible displacement functions to yield a system
of homogenous equations. These equations forms the governing eigenvalue equation.
With the aid of the commercial software package Mathematica (Wolfram, 1999), a
computer code was written to solve the eigenvalue equation for the critical buckling
pressure.
67
Chapter 4: Energy Functionals and Ritz method for Buckling Analysis of Domes
4.1. Introduction
The literature survey as reported in Chapter 1 provided the previous treatments of
moderately thick domes. Most of these studies considered spherical domes and
adopted shallow shell theory. The buckling formulation and analysis developed in this
chapter are, however, applicable to domes of any meridional shape. Since the critical
buckling pressure for moderately thick domes may be sensitive to the transverse shear
deformation which depresses the buckling capacity, the effect of transverse shear
deformation is included in the energy formulation.
The buckling analysis is carried out using the well accepted Ritz method,
primarily for its simplicity and ease of implementation. The automation of the Ritz
method for any boundary condition is achieved by approximating the shell
displacement components as the product of one-dimensional polynomial functions
with the boundary equations raised to appropriate powers so that the geometric
boundary conditions are satisfied at the outset. By taking an appropriate number of
Ritz function terms in the solution, the critical buckling pressure of rotational shells
can be obtained accurately. This convergence characteristic of the solutions is
demonstrated by comparing the results of spherical shells with existing solutions.
Using a computer code developed in this study, new buckling solutions for moderately
thick spherical and parabolic domes of various dimensions and boundary conditions
are presented. These solutions are deemed useful to engineers engaged in the design of
dome structures.
68
4.2 Governing Eigenvalue Equation
4.2.1 Geometrical properties of domes
In order to define the dome geometry, two principal radii of curvature: r1 and r2 have
to be specified. The radius r1 is the principal radius of curvature of the meridian
whereas the principal radius r2 generates the middle surface of the dome in the
direction perpendicular to the tangent on the meridian. The principal radii of
curvature, r1 and r2 , which subsequently appear in the governing equations of shell
buckling, may be determined from the generating curve r0 = f ( z ) using the following
well known relations (Dym, 1974)
3
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
⎨1 + ⎢ ⎬
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
r1 = − (4.1a)
d 2 f (z )
dz 2
1
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
r2 = f ( z )⎨1 + ⎢ ⎬ (4.1b)
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
Although the preceding equations are expressed in terms of the vertical
coordinate, z, the deformation of rotational shells is often expressed in terms of the
coordinate s, which is the arc length measured from the apex of the dome (see Fig.
4.1). In order to facilitate the transformation of functions associated with the problem,
which include strains, one observes the following geometric relation between ds and
dz, as indicated in Fig. 4.1,
69
⎡ ⎛ dr0 ⎞ 2 ⎤ ⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪
(ds ) 2
= (dz ) + (dr0 )
2 2
= ⎢1 + ⎜ ⎟ ⎥ (dz ) = ⎨1 + ⎢
2
⎥ ⎬(dz )
2
(4.2)
⎢⎣ ⎝ dz ⎠ ⎥⎦ ⎪⎩ ⎣ dz ⎦ ⎪⎭
1
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
which leads to ds = ⎨1 + ⎢ ⎬ dz (4.3)
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
r=f(z)
0
2 1/2
z
ds={1+[f'(z)] } dz r2
dz H
r1
dr=f'(z)dz
0
φ
L L
Fig. 4.1 Coordinate systems and parameters defining the shape of dome structures
Thus, for any function g(z), one may express its derivatives in either the z or s
coordinates by the following transformation equation:
∂g ( z ) ∂g ( z ) ∂z 1 ∂g ( z ) 1 ∂g ( z )
= = = (4.4)
∂s ∂z ∂s 1
∂z η ∂z
⎧⎪ ⎡ df ( z ) ⎤ 2
⎫⎪ 2
⎨1 + ⎢ ⎬
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
1
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
where η = ⎨1 + ⎢ ⎬ (4.5)
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
70
4.2.2 Mindlin Shell Theory
Generally, shell analyses are based on the classical thin shell theory (Aron 1874, Love
1888), in which the effect of transverse shear deformation is normally neglected.
However, when dealing with moderately thick shells, the classical thin shell theory
under-predicts the deflections and over-predicts the buckling loads and natural
frequencies due to the effect of transverse shear deformations. As we are dealing with
moderately thick shells, it is necessary to adopt a more refined shell theory such as
Mindlin shell theory that will allow for the effects of transverse shear deformation.
4.2.2.1 Assumptions
In Mindlin shell theory, the following assumptions are made (Mindlin 1951, Reddy
2004):
• The transverse normal is inextensible.
• Normals to the reference surface of the shell before the deformation remain
straight but not necessarily normal after the deformation.
• The shell deflections are small so that strains may be treated as infinitesimal.
• The transverse normal stress is negligible so that the plane stress assumptions
can be invoked.
• The normals during bending undergo constant rotations about the middle
surface while maintaining the straightness and thereby admitting a constant shear strain
through the shell thickness. The constant rotations of the normals to the middle surface
now become unknown independent variables and are denoted by ψ ( z ) .
The first four assumptions are the same as their classical thin shell counterparts.
The last assumption that allows the constant rotation of normal is the main difference
between Mindlin shell theory and the classical thin shell theory. The allowance of
71
constant rotation implies that transverse shear strain is constant through the thickness
of the shell. This, however, contradicts the fact that the actual transverse shear strain
distribution is parabolic through the thickness. As the constant strain (stress) violates
the statical requirement of vanishing shear stress at the surface of the shell, a shear
correction factor κ 2 was proposed by Mindlin (1951) to compensate for the error. He
pointed out that for an isotropic plate, the shear correction factor κ 2 depends on
Poisson’s ratio v and it may vary from κ 2 = 0.76 for v = 0.3 to κ 2 = 0.91 for v = 0.5.
On the other hand, by comparing the constitutive Mindlin shear force with the one
proposed by Reissner (1945), who assumed a parabolic shear stress distribution at the
outset of his plate theory formulation, the implicit shear correction factor becomes
κ 2 = 5 / 6 . This value of the shear correction factor has been commonly used for the
analyses of Mindlin plates and shells (see for example, Liew et al. 2004 and Hou et al.
2005) and κ 2 = 5 / 6 will also be used in this study.
4.2.2.2 Displacement Components
Based on Mindlin shell theory, the displacement components of an arbitrary point at a
distance ζ from the shell mid-surface are given by (Chao et al. 1988)
u~ ( z , ζ ) = u (z ) + ζψ ( z ) (4.6a)
~ ( z , ζ ) = w( z )
w (4.6b)
where u is the meridional displacement, w is the radial displacement of the middle
surface and ψ is the rotation of the middle-surface in the meridional direction. It
should be apparent that Eq. (4.6a) assumes that the meridional displacement varies
72
~
∂w
linearly across the thickness of the shell. Note that by setting ψ (z ) = , one recovers
∂z
the displacement fields of classical thin shell theory.
4.2.3 Strain-Displacement Relations
According to Kraus (1967), the strain-displacement relations for small deformation of
linearly elastic axisymmetric rotational shells are given by:
1 ⎛ ∂r0 u w ∂r ψ ⎞
εθ = ⎜⎜ + + ζ 0 ⎟⎟ (4.7a)
1 + (ζ / r2 ) ⎝ ∂s r0 r2 ∂s r0 ⎠
1 ⎛ ∂u w ∂ψ ⎞
εs = ⎜⎜ + + ζ ⎟⎟ (4.7b)
1 + (ζ / r1 ) ⎝ ∂s r1 ∂s ⎠
1 ⎛ u ∂w ⎞
γ sζ = ⎜⎜ − + + ψ ⎟⎟ (4.7c)
1 + (ζ / r1 ) ⎝ r1 ∂s ⎠
where εθ is the normal strain in the direction of the parallel circles, ε s is the normal
strain in the meridional direction, and γ sζ is the transverse shear strain associated with
rotation of the shell in the meridian direction.
By invoking Eq. (4.2), the kinematic equations of Eqs. (4.7a-c) may be re-written as:
1 ⎛ ∂r0 u w ∂r ψ ⎞
εθ = ⎜⎜ + +ζ 0 ⎟⎟ (4.8a)
1 + (ζ / r2 ) ⎝ η∂z r0 r2 η∂z r0 ⎠
1 ⎛ ∂u w ∂ψ ⎞
εs = ⎜⎜ + +ζ ⎟ (4.8b)
1 + (ζ / r1 ) ⎝ η∂z r1 η∂z ⎟⎠
1 ⎛ u ∂w ⎞
γ sζ = ⎜⎜ − + + ψ ⎟⎟ (4.8c)
1 + (ζ / r1 ) ⎝ r1 η∂z ⎠
73
1
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
where η = ⎨1 + ⎢ ⎬ (4.5)
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
4.2.4 Stress-Strain Relations
In order to accommodate a more general orthotropic shell, the stress-strain relation is
assumed to follow the orthotropic Hooke’s law in the form given by (Chao et al. 1988)
⎧σ θ ⎫ ⎡Q11 Q12 0 ⎤⎧ ε θ ⎫
⎪ ⎪ ⎢ ⎪ ⎪
⎨σ s ⎬ = ⎢Q12 Q22 0 ⎥⎥ ⎨ ε s ⎬ (4.9)
⎪τ ⎪ ⎢ 0 0 Q44 ⎥⎦ ⎪⎩γ sζ ⎪⎭
⎩ sζ ⎭ ⎣
where σ θ is the normal stress in the direction of the parallel circles, σ s is the normal
stress in the meridional direction, and τ sζ is the transverse shear stress, and
Eθ Es
Q11 = ; Q22 = ; (4.10a)
1 − v s vθ 1 − vφ vθ
Q12 = Q11ν s and Q44 = κ 2 G sς (4.10b)
where E s , Eθ , ν s and ν θ are the Young’s moduli and Poisson’s ratios in the direction
of the meridian and parallel circle, respectively, and Gsζ is the shear modulus in the
s − ζ plane. The parameter κ 2 is a shear correction factor introduced to compensate
for the error inherent in the assumption of a constant shear strain (stress) in Mindlin
shell theory. The commonly accepted value of κ 2 = 5/6 is adopted for the correction
factor in this chapter.
74
4.2.5 Derivation of Energy Functionals
For an assumed kinematically admissible displacement field for the middle surface, the
elastic strain energy functional U of the rotational shell is defined as (Chao et al. 1988)
1
(σ θ ε θ + σ s ε s + τ sζ γ sζ )dV
2 ∫V
U= (4.11)
In view of the stress and strain relations Eq. (4.9), one obtains
U=
1
2 ∫V
( )
Q11ε θ2 + 2Q12 ε θ ε s + Q22 ε s2 + Q44 γ s2ς dV (4.12a)
where the incremental volume dV is given by
⎛ ζ ⎞⎛ ζ ⎞ ⎛ ζ ⎞⎛ ζ ⎞
dV = ⎜⎜1 + ⎟⎟⎜⎜1 + ⎟⎟r0 dsdθdζ = ⎜⎜1 + ⎟⎟⎜⎜1 + ⎟⎟ηr0 dzdθdζ (4.12b)
⎝ r1 ⎠⎝ r2 ⎠ ⎝ r1 ⎠⎝ r2 ⎠
In view of Eqs. (4.8) and (4.12b) and after integrating Eq. (4.12) from θ = 0 through
to 2π , and using the following parameters:
⎛ ζ ⎞
⎜ 1+ ⎟
r1
A11 = ⎜ ⎟Q ;
⎜ ζ ⎟ 11
⎜1+ ⎟
⎝ r2 ⎠
⎛ ζ ⎞
⎜1+ ⎟
r2
A22 = ⎜ ⎟Q ; (4.13a-d)
⎜ ζ ⎟ 22
⎜ 1+ ⎟
⎝ r1 ⎠
75
A12 = Q12
⎛ ζ ⎞
⎜1+ ⎟
r2
A44 = ⎜ ⎟Q
⎜ ζ ⎟ 44
⎜ 1+ ⎟
⎝ r1 ⎠
one obtains the following expression for the elastic strain energy of the rotational shell
h/2 H ⎧⎪ ⎛ ∂r u w ∂r ψ ⎞
2
U =π ∫ ∫ ⎨ A11 ⎜⎜ + +ζ 0 ⎟⎟
0
−h / 2 0 ⎪⎩ ⎝ η∂z r0 r2 η∂z r0 ⎠
⎛ ∂r u w ∂r ψ ⎞⎛ ∂u w ∂ψ ⎞
+ 2 A12 ⎜⎜ 0 + +ζ 0 ⎟⎟⎜⎜ + +ζ ⎟ (4.14)
⎝ η∂z r0 r2 η∂z r0 ⎠⎝ η∂z r1 η∂z ⎟⎠
⎛ ∂u w ∂ψ ⎞
2
⎛ u ∂w ⎞
2
⎫⎪
+ A22 ⎜⎜ + +ζ ⎟⎟ + A44 ⎜⎜ − + + ψ ⎟⎟ ⎬ηr0 dzdζ
⎝ η∂z r1 η∂z ⎠ ⎝ r1 η∂z ⎠ ⎪⎭
In applying the Ritz method, the work done by the external forces in moving from
one configuration to another configuration must be included in the estimation of the
total potential energy. To that end, the work done by the buckling pressure, according
to Kawai (1974) and Chao et al. (1998), is given by
2π
∫ ∫ [Nθ (l ) [ (
H
W =
1
2
2
11 + l 212 + l 312 + N s l122 + l 222 + l 322 )
0 0
]
+ N θs (l12 l11 + l 21l 22 ) ηr0 dθdz (4.15)
where N θ , N s and N θs are taken as the initial membrane forces due to the critical
buckling pressure (see Fig. 4.2). Note that by virtue of axisymmetry displacements in
the assumed buckling mode of the dome, the in-plane shear force N sθ vanishes i.e.
76
N sθ = 0 (4.16)
θ
dθ
p
Νθ Νs r
0
Νθ
r2
r
0
Νs + d Νs d φ φ
dφ dφ
r
1 Νs Νs
Fig. 4.2 Membrane forces in an axisymmetrically loaded domes
Furthermore, the initial membrane forces due to the critical buckling pressures
can be defined in the form of:
N θ = pnθ , N s = pn s , (4.17)
where p depends on the external loading conditions, and n s , nθ are parameters
dependent on the geometry of the rotational shell, and
∂r0 u w
l11 = + l12 = 0
η∂z r0 r2
∂u w
l21 = 0 l 22 = + (4.18a-f)
η∂z r1
77
u ∂w
l31 = 0 l 32 = −
r1 η∂z
By substituting Eqs. (4.16), (4.17) and (4.18a-f) into Eq. (4.15), one obtains the
following functional for the work done by the buckling pressure
⎧⎪ ⎛ ∂r u w ⎞ 2
H ⎡⎛ ∂u w ⎞ 2 ⎛ u ∂w ⎞ 2 ⎤ ⎫⎪
W = pπ ∫ ⎨nθ ⎜⎜ 0
+ ⎟⎟ + n s ⎢⎜⎜ + ⎟⎟ + ⎜⎜ − ⎟⎟ ⎥ ⎬ηr0 dz (4.19)
0⎪ ⎝ η ∂z r r ⎠ ⎢ ⎝ η ∂z r ⎠ ⎝ r η∂z ⎠ ⎥⎦ ⎪⎭
⎩ 0 2 ⎣ 1 1
For convenience, the coordinates and length parameters are normalized by a
reference length H, which is the height, or by h , which is the thickness of the shell,
and the critical pressure and material properties are normalized by an effective
Young’s modulus E , which according to Tsai and Pagano (1968) may be taken as
E = (3 / 8)E s + (5 / 8)Eθ for orthotropic composites, i.e.
z r r r
z= ; r0 = 0 ; r1 = 1 ; r2 = 2 ;
H H H H
u w Hψ
u= ; w = ;ψ = ;
h h h
h ζ
ξ= ;ζ = ; (4.20a-n)
H h
ns n
ns = ; nθ = θ and
H H
pH (1 − ν sν θ )
λ=
Eh
78
By adopting the foregoing non-dimensional terms, the total potential energy
functional Π of the rotational shell may be written in normalized form as
Π = U +W (4.21)
where
U (1 − v s vθ ) 1 / 2 1 ⎧⎪ ⎛ ∂r u w 2 2
∂r ψ ⎞ ⎛ ∂u w ∂ψ ⎞
U = = ∫ ∫ ⎨ A11 ⎜⎜
0
+ + ξζ 0 ⎟⎟ + A22 ⎜⎜ + + ξζ ⎟
πE h 3 −1 / 2 0 ⎪⎩ ⎝ η∂z r0 r2 η∂z r0 ⎠ ⎝ η∂z r1 η∂z ⎟⎠
⎛ ∂u w ∂ψ ⎞⎛ ∂r0 u w ∂r ψ ⎞
+ 2 A12 ⎜⎜ + + ξζ ⎟⎟⎜⎜ + + ξζ 0 ⎟⎟ (4.22)
⎝ η∂z r1 η∂z ⎠⎝ η∂z r0 r2 η∂z r0 ⎠
⎛ u ∂w ⎞
2
⎫⎪
+ A44 ⎜⎜ − + +ψ ⎟⎟ ⎬ηr0 dz dζ
⎝ r1 η∂z ⎠ ⎪⎭
and
1⎧
⎪ ⎡⎛ ∂u w ⎞ ⎛ u ∂w ⎞ ⎤ ⎛ ∂r0 u w ⎞ ⎫⎪
2 2 2
W (1 − ν sν θ )
W = = λ ∫ ⎨n s ⎢⎜⎜ + ⎟ +⎜ − ⎟ ⎥ + nθ ⎜⎜ + ⎟ ⎬ηr0 dz
πE h 3 0⎪ ⎢ ⎝ η∂z r1 ⎟⎠ ⎜⎝ r1 η∂z ⎟⎠ ⎥ ⎝ η∂z r0 r2 ⎟⎠ ⎪
⎩ ⎣ ⎦ ⎭
(4.23)
where non-zero terms Aij are related to the material properties by:
⎛ ζ ⎞ ⎛ ζ ⎞
⎜1+ ξ ⎟ ⎜1+ ξ ⎟
⎜ r1 ⎟ Q11 (1 − v s vθ ) ⎜ r1 ⎟ Eθ
A11 = ⎜ ⎟ =⎜ ⎟ E (4.24a)
ζ E ζ
⎜⎜ 1 + ξ ⎟⎟ ⎜⎜ 1 + ξ ⎟⎟
⎝ r2 ⎠ ⎝ r2 ⎠
ν s Eθ
A12 = (4.24b)
E
⎛ ζ ⎞ ⎛ ζ ⎞
⎜1+ ξ ⎟ ⎜1+ ξ ⎟
⎜ r2 ⎟ Q22 (1 − v s vθ ) ⎜ r2 ⎟ Es
A22 = ⎜ ⎟ =⎜ ⎟E (4.24c)
ζ E ζ
⎜⎜ 1 + ξ ⎟⎟ ⎜⎜ 1 + ξ ⎟⎟
⎝ r1 ⎠ ⎝ r1 ⎠
79
⎛ ζ ⎞
⎜1+ ξ ⎟ 2
⎜ r1 ⎟ κ G sζ (1 − v s vθ )
A44 = ⎜ ⎟ (4.24d)
ζ E
⎜⎜ 1 + ξ ⎟⎟
⎝ r2 ⎠
The total potential energy, as expressed in terms of unknown normalized
displacement and rotation components u , w and ψ in Eqs. (4.22) to (4.24), is in a
suitable form for buckling analysis where its stationary condition is sought to yield the
critical buckling pressure.
4.3 Ritz method for buckling analysis
4.3.1 Introduction
In 1909, Walter Ritz published a paper that demonstrates his method for minimizing a
functional, and determining the frequencies and mode shapes of structures. Since then,
the Ritz method has been widely used because of its simplicity in implementation.
Two years after Ritz’s paper (1909), Rayleigh (1911) published a book where he
complained that Ritz had not recognized his similar work (Rayleigh, 1877). Therefore
it is sometime referred to as the Rayleigh-Ritz method. However, Leissa (2005)
investigated carefully the historical works of Rayleigh and Ritz and arrived at the
conclusion that Rayleigh’s name should not be attached to the Ritz method.
In the Ritz method, the displacement function, ℜ(z ) is approximated by a finite
linear combination of trial functions in the form
80
N
ℜ( z ) ≈ ∑ ci pi ( z ) (4.25)
i =1
in which pi (z ) are the approximate functions which individually satisfy at least the
geometric boundary conditions to ensure convergence to the correct solutions. The
static boundary conditions need not be satisfied by these approximate functions. By
minimizing the energy functional ∏ with respect to each of the unknown coefficients
ci, a set of homogeneous equations is obtained as follows
∂∏
= 0; i = 1,2,......, N (4.26)
∂ci
For buckling and vibration problems, the above set of homogeneous equations is
reduced to eigenvalue and eigenvector problems.
The exact solution is obtained if infinite terms are adopted in Eq. (4.25).
However, it is impractical to use an infinite number of terms and so the number of
terms is usually truncated to N terms in applications. The choice of the approximate
functions is very important in order to simplify the calculations and to guarantee
convergence to the exact solution. Some of the commonly used trial functions in the
Ritz method for plates and shells analysis are orthogonal characteristic beam
polynomials (Bhat 1985), spline and B-spline functions (Mizusawa 1986; Vermeulen
and Heppler 1998); pb-2 Ritz formulation (Lim and Liew 1994, Liew et al. 1995 and
Liew and Lim 1995); trigonometric functions (Lim et al. 2003) and two dimensional
polynomial functions with appropriate basic functions (Liew 1990; Liew and Wang
1992, 1993; Geannakakes 1995). Among them, the latter trial functions can be
81
modified and used for the analysis of axisymmetric rotational shells with general shape
and boundary conditions while the others may not be so convenient. Moreover, the
computational accuracy may also be increased since the polynomial functions admit
exact calculations of differentiation and integration of the functions (Liew et al. 1998).
Therefore, in this study, mathematically complete, one-dimensional polynomial
functions are adopted together with basic functions comprising boundary equations
that are raised to appropriate powers in order to ensure the satisfaction of the geometric
boundary conditions.
4.3.2 Ritz formulation
As noted earlier, buckling analysis of moderately thick shells of revolution when
treated as an eigenvalue problem may be solved by the Ritz method where the method
lends itself to yield reasonably accurate results. In using the Ritz method,
kinematically admissible Ritz functions are assumed for the deflection and rotation
components of the middle surface of the rotational shell. To that end, the normalized
displacement and rotation components are approximated by polynomials as follows:
N1
u ( z ) = ∑ ci pi (4.27a)
i =1
N2
w (z ) = ∑c p i i (4.27b)
i = N1 +1
N3
ψ (z ) = ∑c p i i (4.27c)
i = N 2 +1
82
where N1, (N2 – N1) and (N3 – N2) correspond to the number of the polynomial terms
and ci the unknown coefficients for the displacements and rotation and the functions pi
can be expressed in the following forms
pi = η u z (i −1) for i = 1 to N1, (4.28a)
pi = η w z (i − N1 −1) for i = N1 to N2, (4.28b)
pi = ηψ z (i − N 2 −1) for i = N2 to N3. (4.28c)
The terms ηu , η w and ηψ are the product of the boundary equations raised to an
appropriate power so that the selected Ritz functions satisfy the geometric boundary
conditions. More specifically, η u , η w , and ηψ are given below for the following
boundary conditions of domes.
In view of equations (4.22), after integrating Eq. (4.13) over the shell thickness h ,
one obtains the following coefficients
⎛ ζ ⎞
1
2
⎜1+ ξ
1
2
⎟
~ ⎜ r1 ⎟ Eθ
A11 = ∫ A11 dζ = ∫ ⎜ ⎟ E dζ (4.29a)
ζ
− ⎜1+ ξ ⎟⎟
1 1
−
2⎜
2
⎝ r2 ⎠
⎛ ζ ⎞
1
2
⎜1+ ξ
1
2
⎟
~ ⎜ r2 ⎟ Eθ
A22 = ∫ A22 dζ = ∫ ⎜ ⎟ E dζ (4.29b)
ζ
− ⎜1+ ξ ⎟⎟
1 1
−
2⎜
2
⎝ r1 ⎠
⎛ ζ ⎞
1
⎜1+ ξ
1
⎟ 2
⎟ κ G sζ (1 − vφ vθ )
2 2
~ ⎜ r2
A44 = ∫ A44 dζ = ∫ ⎜ ⎟ dζ (4.29c)
ζ E
− ⎜1+ ξ ⎟⎟
1 1
−
2⎜
2
⎝ r1 ⎠
83
1 1
~ 2 2
ν θ Es
B12 = ∫A1
12 ζ dζ = ∫ 1 E
ζ dζ (4.29d)
− −
2 2
⎛ ζ ⎞
1
2
⎜1+ ξ 1
2
⎟
~ ⎜ r1 ⎟ Eθ 2
D11 = ∫ A11ζ 2 dζ = ∫ ⎜ ⎟ E ζ dζ (4. 29e)
ζ
− ⎜1+ ξ ⎟⎟
1 1
−
2⎜
2
⎝ r2 ⎠
1 1
~ 2 2
ν θ Es
D12 = ∫A
1
12 ζ 2 dζ = ∫
1 E
ζ 2 dζ (4. 29f)
− −
2 2
⎛ ζ ⎞
1
2
⎜1+ ξ 1
2
⎟
~ ⎜ r2 ⎟ Eθ 2
D22 = ∫ A22ζ 2 dζ = ∫ ⎜ ⎟ E ζ dζ (4. 29g)
ζ
− ⎜1+ ξ ⎟⎟
1 1
−
2⎜
2
⎝ r1 ⎠
In view of Eqs. (4. 29a-g) and (4.28a-c), the energy functional U given by Eq.
(4.22) may be written in matrix form as
∫ ∫ {[A ]
N3 N3 1/ 2 1
i w j + B12 (wi q j + q i w j ) +
~ ~ ~~ ~ ~~ ~~
U = ∑∑ q~ q~ j + D11 w
11 i
i =1 j =1 −1 / 2 0
[
~ ~ ~ ~ ~ ~ ~~
+ A22 e~i e~j + D11 ~
~
]
ri r j + B12 (ei r j + ri e j ) + A12 (q~i e~j + q~ j e~i ) + (4.30)
+ D12 (w }
i r j + w j ri ) + B12 (q i r j + q j ri ) + B12 (wi e j + w j ei ) + A44 ti t j ci c jηr0 dz dζ
~ ~ ~ ~ ~ ~ ~~ ~ ~ ~ ~ ~ ~ ~ ~~
where the terms in Eq. (4.30) can be expressed as
∂r0 u w N 3 ~
(a) + = ∑ ci q i (4.31)
η∂z r0 r2 i =1
84
∂r p p
and q~i = 0 i for i = 1 to N1, q~i = i for i = N1 +1 to N2 and q~i = 0 for i = N2 +1
η∂z r0 r2
to N3
∂r0 ψ N3
~
(b) ζ = ∑ ci w (4.32)
η∂z r0 i =1
i
~ = 0 for i = 1 to N , w
and w ~ = ξζ ∂r0 pi
~ = 0 for i = N +1 to N and w for i =
1 1 2
η∂z r0
i i i
N2 +1 to N3
∂u w N3
(c) + + = ∑ ci ~
ei (4.33)
η∂z r1 i =1
1 ∂pi p
and e~i = for i = 1 to N1, e~i = i for i = N1 +1 to N2 and e~i = 0 for i =
η ∂z r1
N2+1 to N3
∂ψ N3
(d) ξ = ∑ ci ~
ri (4.34)
η∂z i =1
~ ξ ∂pi
and ri = 0 for i = 1 to N1, ~
ri = 0 for i = N1 +1 to N2 and ~
ri = for i = N2 +1
η ∂z
to N3
u ∂w N3
(e) − + + ψ = ∑ ci ~
ti (4.35)
r1 η∂z i =1
and ~
p ~ 1 ∂pi for i = N +1 to N and ~
ti = − i for i = 1 to N1, ti = ti = pi for i =
1 2
r1 η ∂z
N2+1 to N3
85
Similarly, the work done by the external forces (Eq. 4.23) can be expressed as
N3 N3 1
W = ∑∑ λ ∫ {n s (u~i u~ j + g~i g~ j ) + nθ ~ y j }ηr0 dz
yi ~ (4.36)
i =1 j =1 0
where the terms in Eq. (4.36) may be expressed as
∂r0 u w N 3 ~
(f) + = ∑ ci y i (4.37)
η∂z r0 r2 i =1
1 ∂r0 p i p
and ~
yi = y i = i for i = N1 +1 to N2 and ~
for i = 1 to N1, ~ y i = 0 for i =
η ∂z r0 r2
N2+1 to N3
∂u w N 3 ~
(g) + = ∑ ci u i (4.38)
η∂z r1 i =1
1 ∂pi p
and u~i = for i = 1 to N1, u~i = i for i = N1 +1 to N2 and u~i = 0 for i = N2 +1
η ∂z r1
to N3
u ∂w N3
(h) − = ∑ ci g~i (4.39)
r1 η∂z i =1
p 1 ∂pi
and g~i = i for i = 1 to N1, g~i = − for i = N1 +1 to N2 and g~i = 0 for i = N2+1
r1 η ∂z
to N3
86
The total potential energy functional given by Eqs. (4.21) (with (4.30) and (4.36))
is expressed in terms of displacements, the material stiffness, and external loads.
Following the standard procedure for the Ritz method, the unknown coefficients ci are
obtained by extremizing the total potential energy functional Π , i.e.
∂Π
= 0 ; i = 1, 2,..., N 3 (4.40)
∂ci
which yields a set of homogeneous equations that can be conveniently expressed in a
matrix form containing the unknown coefficients ci
([K ] + λ[ M ]){c} = {0} (4.41)
where [K] and [M] are (N3 x N3) square matrices and {c} is a column vector consisting
of the coefficients ci.
The elastic buckling pressure parameter λ is obtained by solving eigenvalue of
the governing equation. With the aid of the commercial software package Mathematica
(Wolfram, 1999), this eigenvalue problem is solved using built-in function
Eigenvalues.
87
4.3.3 Boundary conditions
Two edge supporting conditions are considered, namely the clamped support and the
simply supported as shown in Fig. 4.3.
(a) Clamped (b) Simply supported

Fig. 4.3 Boundary conditions
For a clamped edge, the boundary conditions are u (1) = w (1) = ψ (1) = 0 . In view
of these boundary conditions, the basis functions are given by
η u = (z − 1)z
η w = ( z − 1) (4.42)
η ψ = ( z − 1 )z
For a simply supported edge, the boundary conditions are u (1) = w (1) = 0 but
ψ (1) ≠ 0 . In view of these boundary conditions, the basis functions are given by
η u = ( z − 1)z
η w = ( z − 1) (4.43)
ηψ = z
88
4.3.4 Mathematica for solving eigenvalue problem
Modern computer algebra systems especially Mathematica (Wolfram 1999) are
very powerful not only in symbolic computations, but in numerical computations as
well. Among other things, Mathematica (Wolfram 1999) allows its user to manipulate
symbols, numbers, data, and graphics. Such computing environments are already used
quite extensively by researchers for a wide range of serious scientific calculations. In
1990, Beltzer (1990) gave a comprehensive review of symbolic computation packages
and analytical applications in engineering analysis. Ioakimidis (1992a, 1992b, 1992c)
has demonstrated the use of Mathematica(Wolfram 1999) in semi-analytical numerical
structural applications, particularly those involving energy methods.
A system such as Mathematica (Wolfram 1999) is ideally suited for many
analytical applications in small engineering energy problems. The eigenvalue problem
is solved using built-in function Eigenvalues in the software package Mathematica
(Wolfram, 1999). In Mathematica, Eigenvalues function used the function DSYEVR
in LAPACK<www.netlib.org/lapack/> routines to calculate the numerical eigen values
and vectors of a real and symmetric matrice. The Mathematica code to obtain buckling
strength of rotational shells is presented in the Appendix.
89
Although buckling of shells under compressive loading is of practical significance in
the design of these structures, most of the studies thus far focused on rotational shells
of spherical shape using a thin shell theory. An attempt is made in this chapter to
formulate a methodology for predicting the critical buckling pressure of moderately
thick rotational shells generated by any meridional shape under external uniform
pressure. The effect of transverse shear deformation is included in the formulation
using Mindlin shell theory so that the critical buckling pressure will not be excessively
overestimated when the shell is relatively thick.
The critical buckling pressure of thick shells under uniform pressure, formulated
as an eigenvalue problem, is derived using the well accepted Ritz method. Numerical
results, obtained from a computer program, were shown to be in close agreement with
existing buckling solutions for isotropic and orthotropic spherical shells. One feature
of the proposed method is that highly accurate solutions can be ensured by including
an appropriate number of terms in the Ritz functions. The formulation is also capable
of handling different support conditions, by raising the boundary equations to the
appropriate power so that the geometric boundary conditions are satisfied a priori.
New solutions for the buckling pressure of moderately thick spherical and parabolic
shells of various dimensions and boundary conditions are presented and, although
these results are limited by the material properties assumed in this chapter, they are
nonetheless useful for preliminary design of shell structures.
90
CHAPTER 5
BUCKLING OF DOMES UNDER UNIFORM
PRESSURE
This chapter is concerned with the elastic, axisymmetric buckling of moderately thick,
orthotropic domes under a uniform external pressure. For the buckling analysis, we
apply the Ritz method presented in Chapter 4. The validity of the developed Ritz
method as well as the convergence and accuracy of the buckling solutions are
demonstrated using examples of spherical domes (a special case of generic dome
structures) where closed-form solutions exist. Upon establishment of the validity of the
method and its ability to furnish accurate results, we generate extensive buckling
solutions for moderately thick spherical and parabolic domes of various dimensions
and boundary conditions. These new results, presented in tabulated form, are deemed
useful to engineers engaged in the design of shell structures.
91
Chapter 5: Buckling of Domes under Uniform Pressure
5.1 Problem definition
Consider a dome, also known as a synclastic shell of revolution, of height H, base
radius L, and uniform thickness h. The dome is formed by rotating a curve defined by
r0 = f ( z ) with (df / dz )z =0 = 0 , about the vertical z axis as shown in Fig. 5.1. The
dome is subjected to a static uniform external pressure p, and is free of geometric and
material imperfections. The problem at hand is to determine the critical pressure pcr for
axisymmetric buckling of domes under the uniform external pressure p.
r0
p
z
r0 =f(z)
H
L L
Fig 5.1 Dome under uniform pressure
5.2 Geometrical parameters
To consider the aforementioned buckling problem, one needs to first evaluate the
parameters for general energy functionals given by Eqs. (4.21), (4.22) and (4.23) in
Chapter 4. From statical considerations, the membrane force N s acting in the meridian
direction is given by
92
φ
1
Ns = −
r2 sin 2 φ ∫ p cos φ sin φ r r dφ
0
1 2 (5.1)
In view of the following geometric relations for rotational shells
r0 = r2 sin φ ; (5.2a)
dr0 = ds cos φ = r1 cos φdφ (5.2b)
Eq. (5.1) may be expressed as
r0
1 pr02 pr
Ns = −
r2 sin 2 φ ∫ pr0 dr0 = −
0 2r2 sin φ
2
=− 2
2
(5.3)
Similarly, from statical considerations, the membrane force N θ acting in the
circumferential direction is given by
⎛ Nφ ⎞ ⎛ r ⎞
N θ = −r2 ⎜⎜ p + ⎟⎟ = − pr2 ⎜⎜1 − 2 ⎟⎟ (5.4)
⎝ r1 ⎠ ⎝ 2r1 ⎠
By comparing the definitions of Eq. (4.17) with Eqs. (5.1) and (5.4), one can
deduce that
r2 ⎛ r ⎞
ns = − ; nθ = −r2 ⎜⎜1 − 2 ⎟⎟ (5.5)
2 ⎝ 2r1 ⎠
93
5.3. Results and discussions
5.3.1 Spherical domes
For the spherical dome, the meridional curve is defined by (see Fig. 5.2)
r0 = f ( z ) = 2 Rz − z 2 (5.6)
In view of the non-dimensional parameters in Eq. (4.20), one can obtain
r0 = 2 R z − z 2 (5.7)
where R = R / H and z = z / H .
r=f(z)
0 p
z
z
R
H
α
L L L L
(a) (b)
Fig 5.2 Spherical domes under uniform pressure
By substituting Eqs. (5.5) and (5.7) into Eq.(4.1) and (4.20), one can obtain the
geometrical properties of the spherical dome
r1 = r2 = R (5.8)
R
n s = nθ = − (5.9)
2
94
For a given spherical dome with a subtended angle α (see Fig. 5.2a) and thickness
to radius ratio h/R, the non-dimensional terms R , ξ that appear in the energy
functionals are defined as
R 1
R= = (5.10a)
H (1 − cos α )
h h R T
ξ= = . = (5.10b)
H R H (1 − cos α )
By substitution of the foregoing geometrical parameters into the eigenvalue
equation (4.41) in Chapter 4, and upon solving the equation, one obtains the critical
p cr H (1 − ν sν θ )
buckling pressure parameter λ = of spherical domes. Results for
Eh
spherical domes with different thickness-to-radius ratios will be given in the next
section.
Convergence and comparison studies
To study the convergence of the Ritz solutions with respect to the number of
polynomial terms used to approximate the displacements, we consider hemispherical
domes with radius-to-thickness ratios ranging from R/h = 10 to 1000 and with different
boundary conditions, namely, clamped and simply supported. Tables 5.1 and 5.2 show
the convergence studies of the critical buckling pressure parameter λ . It can be seen
that the Ritz solutions converge monotonically with increasing degrees of the
polynomials for all radius-to-thickness ratios. In order to obtain accurate buckling
solutions (within 0.05% error), the number of polynomial terms required for each
95
displacement function is 40. This number of terms will be assumed to suffice for
generating accurate results for other spherical dome shapes in this chapter.
For comparison purposes, numerical solutions for clamped domes obtained by
Uddin (1987, 1993) who used the multi segment method of integration of Kalnins and
Lestingi (1967) to solve the governing shell equations are included in Table 5.1. By
comparing Uddin’s results with the present Ritz solutions, it can be seen that for the
large radius-to-thickness ratios (e.g. R/h = 300), the converged results are found to be
in good agreement. However, for relatively small radius-to-thickness ratios (i.e. thicker
shells), the Ritz solutions are slightly lower than those presented by Uddin (1987). For
example, for R/h = 25, the difference between the converged Ritz solution and that of
Uddin (1987) is about 1.2% lower. This difference is due to neglect of the effect of
transverse shear deformation in Uddin’s analysis.
96
Table 5.1 Convergence of critical buckling pressure parameter λ

of a clamped hemispherical dome
λ x103 Radius over thickness ratio R/h

10 25 100 300 1000
3 222.41039 194.03803 188.27774 187.92796 187.88811
6 118.54395 55.67167 27.17262 25.13063 24.89573
9 112.83077 46.57039 14.33367 7.43311 6.39213
12 110.93703 45.69036 11.66922 5.23093 2.64581
15 109.84001 45.26218 11.27922 4.23114 1.77093
Number 18 109.15098 44.95244 11.19395 3.88616 1.46390
of 21 108.69171 44.71732 11.16818 3.75452 1.31493
terms 24 108.36999 44.53604 11.15321 3.70952 1.20982
27 108.13490 44.39430 11.14080 3.69693 1.16335
30 107.95671 44.28166 11.12910 3.69448 1.13873
33 107.81719 44.19041 11.11740 3.69317 1.12215
36 107.70465 44.11488 11.10527 3.69199 1.11328
39 107.61129 44.05082 11.09222 3.69081 1.10788
40 107.58348 44.03143 11.08757 3.69039 1.10670
Uddin 1987) …….. 44.6469 11.0423 3.6364 …..
Table 5.2. Convergence of critical buckling pressure parameter λ

of a simply supported hemispherical dome
λ x103 Radius over thickness ratio R/h

10 25 100 300 1000
3 159.45770 102.99050 88.55159 87.63990 87.53578
6 106.57882 46.48155 16.72819 9.38351 8.46736
9 105.95376 43.91228 11.96595 5.71284 2.42462
12 105.63408 43.55971 11.23972 4.10419 1.77645
15 105.38689 43.41658 11.03587 3.83347 1.38606
18 105.19826 43.31454 11.00392 3.72957 1.22685
Number
21 105.05506 43.23428 11.00054 3.69173 1.15762
of terms
24 104.94536 43.16959 10.99926 3.67563 1.13439
27 104.85989 43.11685 10.99842 3.67253 1.11983
30 104.79192 43.07337 10.99769 3.67146 1.11122
33 104.73669 43.03703 10.99696 3.67090 1.10711
36 104.69077 43.00614 10.99616 3.67050 1.10441
39 104.65169 42.97931 10.99521 3.67014 1.10293
40 104.63986 42.97107 10.99484 3.67002 1.10249
97
Although a clamped spherical dome with a 90o meridian angle differs
geometrically from that of a complete spherical shell, the critical buckling pressure of
the clamped spherical shell is nonetheless of the same order of magnitude when
compared to that of a complete shell with the same radius. For comparison of the
results obtained from the various methods, it is convenient to express the critical
buckling pressure of incomplete shells as a fraction of that of complete shells. To that
end, it is instructive to note that the classical buckling pressure of complete isotropic
spherical shells, first established by Zoelly (1915), is
2
2E ⎛h ⎞
pcl = ⎜ ⎟ (5.11)
3 (1 − v 2 ) ⎝ R ⎠
Recall from the definition in Eq. (4.20) that,
p cr H (1 − ν 2 )
λ= (5.12)
Eh
and in view of Eqs. (5.11) and (5.12), the solution for incomplete shells can be
expressed as a fraction of the classical buckling pressure of a complete spherical shell
by the following ratio
p cr 2 1 R R
=λ (5.13)
3 (1 − ν 2 ) h H
3 / 2
pcl
98
The ratio of critical buckling pressures in Eq. (5.13) will be used as the basis for
comparison of available solutions in the literature. Table 5.3 shows the comparison of
critical buckling pressures of the 900 clamped spherical domes with solutions by
various researchers for a small radius-to-thickness ratio of R/h = 25. Among these
researchers, only Uddin (1987) calculated the buckling pressure without adopting the
shallowness assumptions of shell structures. It can seen from Table 5.3 that the present
result, expressed as a ratio in Eq. (5.13), is less than that of Uddin (1987) due to the
effect of transverse shear deformation. Note that the current formulation may be used
to furnish the critical buckling pressure associated with the corresponding result based
on classical thin shell theory by setting a large value for the shear correction factor, say
κ 2 =1000. By doing so, the current formulation yields a ratio of pcr/pcl =1.120 which is
relatively close to Uddin’s ratio of pcr/pcl =1.127.
Table 5.3 Comparison of critical buckling pressure ratio pcr/pcl of

a 900 clamped spherical dome ( R / h = 25)
Present Uddin Huang Archer Dumir Budiansky Thurston

result (1987) (1964) (1958) et al. (1959) (1961)
(pcr/pcl) (1984)
1.076 1.127 1.057 0.750 1.035 1.058 1.067
It can be seen from Table 5.3 that the prediction of the critical buckling pressure of
hemispherical domes with R/h = 25 varies among different researchers. The value
varies from a low ratio of 0.75 by Archer (1958) to a high ratio of 1.067 by Thurston
(1961). It should be noted that a shallow shell theory was assumed in the studies by
Huang (1964), Archer (1958), Dumir et al. (1984), Budiansky (1959) and Thurston
(1961). Their results are therefore expected to be lower than the results based on the
99
non-shallow shell theory, which is adopted herein. Also interestingly, their analyses,
which neglect the effect of transverse shear deformation and employ shallow shell
theory, yield critical buckling pressures that are comparable with the present result for
moderately thick shells. The good prediction is, in part, because the shallow shell
assumption lowers the critical buckling pressure whilst the neglect of transverse shear
deformation raises the critical buckling pressure. These are compensating effects that
have effectively canceled each other out for the considered shell problem.
Effects of radius-to-thickness ratio and transverse shear deformation
As the critical buckling pressure may be sensitive to the amount of transverse shear
deformation that occurs during buckling, the influence of shear deformation is
examined in moderately thick shells by varying the radius-to-thickness ratio. The
sensitivity study is made for isotropic hemispherical shells that are simply supported at
their edges. Table 5.4 presents the critical buckling pressure normalized by the Young’s
modulus i.e. pcr/E for radius-to-thickness ratios R/h ranging from 10 to 1000. The
assessment is made for hemispherical domes with simply-supported boundary
conditions. Critical buckling pressures obtained using classical thin shell theory are
also included in the table so as to observe the effect of transverse shear deformation on
the buckling pressure.
100
Table 5.4 Effect of transverse shear deformation on the buckling pressures pcr/E
of simply supported hemispherical domes
Thin shell theory Difference

R/h Mindlin shell theory
(Muc 1992) (%)
10 12104.5507 11507.0859 -5.1921
25 1936.7281 1890.5567 -2.4422
50 484.1820 480.5784 -0.7499
100 121.0455 120.8434 -0.1673
200 30.2614 30.2318 -0.0977
300 13.4495 13.4441 -0.0400
400 7.5653 7.5637 -0.0220
500 4.8418 4.8412 -0.0119
600 3.3624 3.3622 -0.0064
700 2.4703 2.4705 0.0071
800 1.8913 1.8917 0.0198
900 1.4944 1.4951 0.0475
1000 1.2105 1.2115 0.0901
It can be seen from the results in Table 5.4 that, for a small shell thickness, as
reflected by a large radius-to-thickness ratio e.g. R/h > 100, the buckling pressure pcr/E
is close to the result obtained using thin shell theory (Muc 1992). However, for a large
shell thickness e.g. R/h < 100, the buckling pressure pcr/E based on Mindlin shell theory
is somewhat lower than their thin shell theory counterparts due to the shear
deformation effect. At a radius-to-thickness ratio of R/h = 10, the difference between
the Mindlin and thin shell theories is approximately 5%.
Orthotropic Spherical Domes under uniform pressure p
The versatility of the proposed method in this chapter will now be demonstrated using
an orthotropic dome when subjected to uniform external pressure. According to an
earlier study by Muc (1992), the critical buckling pressure of a simply supported
orthotropic spherical shell is given by
101
2
⎛t⎞ A11 A22 − A122
pcr = 4⎜ ⎟ (5.14)
⎝R⎠ 12
We shall assess the accuracy of this formula by comparing it with our Ritz results.
For the subsequent numerical calculations, the orthotropic dome is assumed to be made
from a graphite/epoxy material with values of E s = 120 GPa, Eθ = 4.8 GPa, G sζ = 2.4
GPa and ν sθ = 0.25 taken from Chao et al (1988). Table 5.5 furnishes the results by
the present Ritz results based on Mindlin shell theory and its comparison with the thin
shell theory by Muc (1992). The results indicate that the normalized buckling pressure
as obtained by Mindlin shell theory may be noticeably smaller than that of thin shell
theory, depending on the thickness of the shell. In the case of thin shells (R/h > 200),
both results are in good agreement, which is expected. However, the difference
between the two theories increases in thick shells as characterized by R/h <200. In the
case of R/h = 25, the difference in critical buckling pressure is as high as 8%. The
tendency of Eq. (5.14) to over-estimate the critical buckling pressure of moderately
thick shells is likely due to the neglecting of the transverse shear deformation.
Table 5.5 Buckling pressures pcr/ E of orthotropic hemispherical domes
Thin shell theory Difference

R/h Mindlin shell theory
(Muc 1992) (%)
25 924.917 855.6495 -8.0953
100 57.8073 57.2088 -1.0461
200 14.4518 14.4346 -0.1197
300 6.42304 6.4159 -0.1118
500 2.31229 2.3112 -0.0482
1000 0.57807 0.5780 -0.0085
102
5.3.2 Parabolic Domes
For parabolic domes, the meridional curve is defined by (See Fig. 5.3)
r0 = 4az (5.15)
where a = L2 / (4 H ) .
In view of the non-dimensional parameters in Eq. (4.20), one can obtain
r0 = 4a z (5.16)
a L2 L2 z
where a = = 2
= and z =
H 4H 4 H
r=h(z)
0
p
z z
r2 H
r1
L L L L
Fig 5.3 Parabolic domes under uniform pressure
By substituting Eqs. (5.15) and (5.16) to Eq. (4.1) and (4.20), one can obtain the
geometry properties of parabolic domes
3
2a ( a + z ) 2
r1 = (5.17)
a2
r2 = 2 a (a + z ) (5.18)
103
n s = − a (a + z ) (5.19)
a (a + 2 z )
nθ = − (5.20)
a (a + 1)
By substituting of the foregoing geometrical parameters into the eigenvalue
equation (4.41) in Chapter 4, and upon solving the equation, one obtains the critical
p cr H (1 − ν sν θ )
buckling pressure parameter λ = of parabolic domes. Results for
Eh
parabolic domes with different height-to-base radius ratios and base radius-to-
thickness ratios will be given in the next section.
Convergence study
Table 5.6 shows the convergence of the critical buckling pressure for an isotropic
parabolic dome of equal base radius and height, and clamped at the base. Results were
generated for domes with base-radius-to-thickness ratio from L/h = 50 to 300. It can be
seen from the table that the solutions from the Ritz method converge monotonically
when the number of terms in the polynomials is increased. The convergence criterion
for parabolic domes was also taken as 0.05%. It is of interest to note that the rate of
convergence differs for different base-radius-to-thickness ratios. For the case of
parabolic domes with a smaller L/h ratio, the critical buckling pressure converges
faster than that of large L/h ratios. Unlike the case for clamped hemispherical domes
where 40 terms are needed in the power series, only 30 terms are needed to satisfy the
convergence criterion of 0.05% in clamped parabolic domes. This number of terms is
assumed to be sufficient for yielding accurate results in parabolic domes of different
height-to-base-radius ratios, base-radius-to-thickness ratios and support conditions.
104

parabolic dome with normalized base radius L = 1
Base radius over thickness ratio L/h

λ x103
10 25 100 300 1000
3 306.24159 268.34711 260.36352 259.87675 259.82128
6 160.04376 65.69875 37.59424 35.75372 35.54293
9 151.57371 57.11352 15.62226 8.55188 7.69817
Number 12 151.52923 56.52362 12.71975 5.10587 2.78505
of 15 151.52875 56.51417 12.27600 4.07852 1.62731
terms 18 151.52852 56.51386 12.21165 3.80622 1.43103
21 151.52839 56.51366 12.20937 3.72199 1.16553
24 151.52831 56.51353 12.20936 3.70724 1.09714
27 151.52826 56.51343 12.20935 3.70623 1.06598
30 151.52822 56.51336 12.20935 3.70617 1.04336
Table 5.7 Convergence of critical buckling pressure parameter λ of a simply supported

parabolic dome with normalized base radius L = 1

λ x103
10 25 100 300 1000
3 163.11313 117.33128 106.67237 106.00877 105.93304
6 110.63181 48.44707 15.19676 9.54045 8.85377
9 110.08303 43.41861 11.15899 4.92409 2.22540
Number 12 110.06928 43.34664 10.31764 3.64115 1.47504
of 15 110.06330 43.34458 10.21054 3.39915 1.22387
terms 18 110.05964 43.34344 10.20759 3.28913 1.02073
21 110.05722 43.34270 10.20750 3.26611 0.98596
24 110.05554 43.34218 10.20746 3.26337 0.96253
27 110.05432 43.34180 10.20743 3.26322 0.94995
30 110.05340 43.34151 10.20740 3.26321 0.94706
The methodology is readily adaptable to parabolic domes of different height, thickness,
material properties and support conditions. Tables 5.8 presents the critical buckling
pressures for parabolic domes of isotropic properties for both clamped and simply
supported edge conditions.
105
Orthotropic Parabolic Domes under Uniform Pressure p
For the case of orthotropic parabolic domes, the same material properties with Es = 120
GPa, Eθ = 4.8 GPa, Gsξ = 2.4 GPa and νsθ = 0.25 for a graphite/epoxy composite are
assumed. Base radius-to-height ratios in the range of 1 / 3 ≤ L / H ≤ 3 and base radius-
to-thickness ratios of 10 ≤ L / h ≤ 1000 are considered.
The critical axisymmetric buckling pressure pcr are presented in Table 5.9 for the
domes with simply supported edge and fixed edge. It can be seen from this two tables
that the critical buckling pressure pcr is sensitive to the edge support conditions. For
instance, for a base radius-to-thickness ratio of L/h =1/100 and base radius-to-height
ratio of L/H = 1 the critical buckling pressure increases from pcr = 5.6758 for the
simply-supported edge condition to 7.5879 for the fixed edge condition. This shows
that the critical buckling pressure increases by 1.34 times from the case of simply
supported domes to fixed edge domes in this particular dome dimensions. Similarly,
large increases in the critical buckling pressure are observed for other base radius-to-
thickness and base radius-to-height ratios.
In order to check our results, we employ the finite element package SAP2000
(Computers and Structures, Inc, 2007) to analyze the above parabolic dome example.
The type of shell element used is the thick shell element and the mesh design adopted
for the analysis is shown in Fig. 5.4. The critical buckling pressures furnished by
SAP2000 are pcr = 5.60892 for simply supported dome and pcr = 7.7375 for fixed
edge dome. These finite element results are in good agreement with our results (within
2%), thereby confirming the correctness of our solutions.
106
Fig 5.4 SAP2000 model of parabolic dome (50x50 elements)
Table 5.8 Buckling pressure parameter λ of isotropic parabolic domes

λ .103 10 25 100 1000
1/3 882.07863 407.69691 105.74235 10.42014
1/2 437.41458 188.47243 46.71385 4.45216
Simply
L/H 1 110.05340 43.34151 10.20740 0.94706
supported
2 17.56449 6.70454 1.56722 0.14638
3 4.52483 1.75368 0.41569 0.03946
1/3 1085.90030 498.99744 122.36784 12.11705
1/2 567.40922 238.32228 55.02010 4.98820
Clamped L/H 1 151.52822 56.51336 12.20935 1.04336
2 22.46878 8.30881 1.81417 0.15776
3 5.29120 2.03606 0.46225 0.04177
Table 5.9 Buckling pressure parameter λ of orthotropic parabolic domes

λ .103 10 25 100 1000
1/3 223.31297 154.46551 53.15067 5.48792
1/2 133.90936 81.35266 24.64222 2.39820
Simply
L/H 1 45.74201 22.25162 5.67579 0.51788
supported
2 8.93617 3.75169 0.87637 0.07961
3 2.41126 0.99661 0.22899 0.02125
1/3 231.13316 189.75079 66.17285 6.18249
1/2 150.08675 107.14368 31.80184 2.73826
Clamped L/H 1 62.41928 32.28857 7.58789 0.59871
2 12.82840 5.34474 1.11203 0.08978
3 3.39914 1.31463 0.27230 0.02323
107
The applicability of the Ritz method for various kinds of domes is demonstrated by
solving the buckling problems of spherical and parabolic domes. The results are
verified by comparing them with limited existing solutions. Based on these examples,
it can be concluded that the developed Ritz method can be readily applied for the
buckling analyses of arbitrarily shaped domes. In the next chapter, results for buckling
of domes under hydrostatic pressure and selfweight will be presented.
108
CHAPTER 6
BUCKLING OF SUBMERGED DOMES
This chapter is concerned with the elastic, axisymmetric buckling of submerged
moderately thick domes. In addition to the water pressure, the domes are also subjected
to selfweight, which is invariably present in this type of structure. Applying the Ritz
method presented in Chapter 4, new buckling solutions for moderately thick spherical
and parabolic domes with various dimensions and boundary conditions are presented.
The validity of the method, convergence and accuracy of solutions are also
demonstrated.
109
Chapter 6: Buckling of Submerged Domes
Consider a dome of height H, base radius L, and uniform thickness h. The dome is
formed by rotating a curve, defined by r0 = f ( z ) with f ′(0 ) = 0 , about the vertical z
axis as shown in Fig. 6.1. The dome is subjected to hydrostatic pressure
p h = γ w (D − H + z ) where γ w is the specific weight of water and its selfweight
p a = γ a h where h is thickness of the dome and γ a the specific weight of dome
material.
r0 r0
z p=
a
γa h z
ph =γw (D-H+z)
H D H
L L L L
(a) Hydrostatic pressure (b) Selfweight
Fig 6.1 Domes under selfweight and hydrostatic pressure
The dome is also assumed to be free of geometrical and material imperfections.
For a given dome height H, the problem at hand is to determine the critical pressure
pcr, for axisymmetric buckling of this submerged dome. This critical pressure will
110
provide information on the maximum height of water depth (or critical water depth)
Dcr for which the dome can be constructed without premature buckling failure.
6.2 Governing equations and Ritz method
6.2.1 Geometrical and loading properties
Consider a dome subjected to hydrostatic pressure p h = γ w (D − H ) + γ w z and its
selfweight p a = γ a h . The positive direction of these loads and their distributions are
shown in Figs. 6.2 and 6.3.
ph =γw (D-H+z)
C
L CL C
L
p1 = γw (D-H) p2= γw z
z z z
D H
L L L
(a) Hydrostatic pressure (b) Pressure p1 (c) Pressure p2
Fig. 6.2 Hydrostatic pressure components
In the case of hydrostatic pressure (see Fig. 6.2a), we can separate the hydrostatic
pressure ph into two components p1 and p 2 . The pressure p1 is constant over the
surface of the dome (see Fig. 6.2b) and is calculated by the product of the distance
from the water level to the apex of the dome (D − H ) and the specific weight of the
water γ w , i.e.
p1 = γ w (D − H ) (6.1)
111
According to Eqs. (5.3) and (5.4) in Chapter 5, we obtain the internal forces of the
dome under uniform pressure p1 as
p1 r2
N s1 = − (6.2)
2
⎛ N ⎞
N θ 1 = − r2 ⎜⎜ p1 + s1 ⎟⎟ (6.3)
⎝ r1 ⎠
Recall Eq. (4.17)
N s = pn s
N θ = pnθ (4.17)
From Eqs. (6.1), (6.2) and (6.3), one obtains
p h1 = p1 = γ w ( D − H ) (6.4)
r2
n s1 = − (6.5a)
2
⎛ n ⎞
nθ 1 = −r2 ⎜⎜1 + s1 ⎟⎟ (6.5b)
⎝ r1 ⎠
On the other hand, the pressure p 2 is linearly dependent on the distance z from
the apex of the dome (see Fig. 6.2c). We can see that z only varies for 0 to the dome
height H. Therefore, for a given dome height, the pressure p 2 does not depend on the
112
water depth D and is given by
p2 = γ w z (6.6)
From statical considerations, the membrane force N s acting in the meridian
direction is given by
φ
1
N s2 = −
r2 sin 2 φ ∫p
0
2 cos φr1 r0 dφ (6.7)
In view of the following geometric relations for rotational shells
r0 = f ( z )
r0 = r2 sin φ (6.8)
dr0 = ds cos φ = r1 cos φdφ
Eq. (6.7) may be expressed as
z
r2
N s 2 = −γ w
r02 ∫ z f (z ) f ' (z )dz
0
(6.9)
In view of Eq.(5.4), one obtains
⎛ N ⎞
N θ 2 = −r2 ⎜⎜ γ w z + s 2 ⎟⎟ (6.10)
⎝ r1 ⎠
113
Similar to the above case, one can obtain the loading parameters of rotational
shells
ph 2 = γ w H (6.11)
z
r2
ns 2 = −
Hr02 ∫ z f (z ) f ' (z )dz
0
(6.12a)
⎛ z n ⎞
nθ 2 = − r2 ⎜⎜ + s 2 ⎟⎟ (6.12b)
⎝H r1 ⎠
In the case of selfweight (see Fig. 6.3), the internal membrane force is given by
φ
1
N sa =−
r2 sin 2 φ ∫ p r r dφ
0
a 1 0 (6.13)
r0
p=
a
γa h z
r=f(z)
0
L L
Fig. 6.3 Selfweight of the dome
In view of the geometric relations for rotational shells given in Eq. (6.8), Eq.
(6.13) may be expressed as
114
z
r
N sa = −γ a h 22
r0 ∫ f (z )ηdz
0
(6.14)
1
⎧⎪ ⎡ df ( z ) ⎤ 2 ⎫⎪ 2
where η = ⎨1 + ⎢ ⎬ . In view of Eq. (5.4), one obtains
⎪⎩ ⎣ dz ⎥⎦ ⎪⎭
⎛ f ' ( z ) N sa ⎞
N θa = − r2 ⎜⎜ γ a h + ⎟ (6.15)
⎝ η r1 ⎟⎠
In view of Eq. (4.17) , one can obtain the geometrical and loading parameters of
the domes as
pa = γ a h (6.16)
z
r
n sa = − 22
r0 ∫ f (z )ηdz
0
(6.17a)
⎛ f ' ( z ) n sa ⎞
nθa = − r2 ⎜⎜ + ⎟⎟ (6.17b)
⎝ η r1 ⎠
6.2.2 Energy functionals and Ritz method
The elastic strain energy of the dome was derived in Eq.(4.22) as
115
1/ 2 1 ⎧⎪ ⎛ ∂r u w ∂r ψ ⎞
2
⎛ ∂u w ∂ψ ⎞
2
U = ∫∫ ⎨ A11 ⎜⎜ + + ξζ 0 ⎟⎟ + A22 ⎜⎜ + + ξζ ⎟
0
−1 / 2 0 ⎪⎩ ⎝ η∂z r0 r2 η∂z r0 ⎠ ⎝ η∂z r1 η∂z ⎟⎠

⎛ ∂u w ∂ψ ⎞⎛ ∂r0 u w ∂r ψ ⎞
+ 2 A12 ⎜⎜ + + ξζ ⎟⎟⎜⎜ + + ξζ 0 ⎟⎟ (4.22)
⎝ η∂z r1 η∂z ⎠⎝ η∂z r0 r2 η∂z r0 ⎠
⎛ u ∂w ⎞
2
⎫⎪
+ A44 ⎜⎜ − + +ψ ⎟⎟ ⎬ηr0 dz dζ
⎝ r1 η∂z ⎠ ⎪⎭
Note that the work done by the hydrostatic pressure is the sum of the work done
by each load component p1 and p 2 . The work done by the hydrostatic pressure and
selfweight is thus given by
W = W1 + W2 + Wa (6.18)
In view of Eq.(4.23), one obtains
1 ⎧⎪ ⎡⎛ ∂u w ⎞ 2 ⎛ u ∂w ⎞ 2 ⎤ ⎛ ∂r0 u w ⎞ ⎫⎪
2
W1 = λ1 ∫ ⎨n s1 ⎢⎜⎜ + ⎟ +⎜ − ⎟ ⎥ + nθ 1 ⎜⎜ + ⎟ ⎬ηr0 dz (6.19)

0⎪ ⎢ ⎝ η∂z r1 ⎟⎠ ⎜⎝ r1 η∂z ⎟⎠ ⎥ ⎝ η∂z r0 r2 ⎟⎠ ⎪
⎩ ⎣ ⎦ ⎭
1 ⎧⎪ ⎡⎛ ∂u w ⎞ 2 ⎛ u ∂w ⎞ 2 ⎤ ⎛ ∂r0 u w ⎞ ⎫⎪
2
W2 = λ 2 ∫ ⎨n s 2 ⎢⎜⎜ + ⎟ +⎜ − ⎟ ⎥ + nθ 2 ⎜⎜ + ⎟ ⎬ηr0 dz (6.20)

0⎪ ⎢ ⎝ η∂z r1 ⎟⎠ ⎜⎝ r1 η∂z ⎟⎠ ⎥ ⎝ η∂z r0 r2 ⎟⎠ ⎪
⎩ ⎣ ⎦ ⎭
1 ⎧⎪ ⎡⎛ ∂u w ⎞ 2 ⎛ u ∂w ⎞ 2 ⎤ ⎛ ∂r0 u w ⎞ ⎫⎪
2
Wa = λ a ∫ ⎨n sa ⎢⎜⎜ + ⎟ +⎜ − ⎟ ⎥ + nθa ⎜⎜ + ⎟ ⎬ηr0 dz (6.21)

0⎪ ⎢ ⎝ η∂z r1 ⎟⎠ ⎜⎝ r1 η∂z ⎟⎠ ⎥ ⎝ η∂z r0 r2 ⎟⎠ ⎪
⎩ ⎣ ⎦ ⎭
where the non-dimensional terms are defined as
nφ 1 nθ 1
n s1 = ; nθ 1 =
H H
116
p h1 (1 − ν sν θ ) γ w (D − H )(1 − ν sν θ )
λ1 = = ; (6.22a-c)
Eξ Eξ
ns 2 n
ns 2 = ; nθ 2 = θ 2
H H
p h 2 (1 − ν sν θ ) γ w H (1 − ν sν θ )
λ2 = = (6.23a-c)
Eξ Eξ
n sa n
n sa = ; nθa = θa
H H
p a (1 − ν sν θ ) γ a h(1 − ν sν θ )
λa = = (6.24a-c)
Eξ Eξ
The total potential energy functional Π of the dome under uniform pressure may
be written as
Π = U +W (6.25)
Following the standard procedure for the Ritz method in Chapter 4, the unknown
coefficients ci are obtained by extremizing the total potential energy functional Π , i.e.
∂Π
= 0 ; i = 1,2,..., N 3 (6.26)
∂ci
which yields a set of homogeneous equations that can be conveniently expressed in a

matrix form containing the unknown coefficients ci
([ K ] + λ1[ M 1 ] + λ2 [ M 2 ] + λa [ M a ]){c} = {0} (6.27)
where [K] is stiffness matrix; [M1], [M2] and [Ma] are work matrices by the pressure
p1 , p 2 and p a and {c} is a column vector consisting of the coefficients ci . It is to be
117
noted that λ2 and λa are known from the information given on the dome height H,
dome thickness h, specific weight of water γ w and specific weight of dome material
γ a . Therefore, λ1 becomes the unknown eigenvalue of the governing equation. The

eigenvalue problem is solved using built-in function Eigenvalues in the software
package Mathematica (Wolfram, 1999). Recall the critical buckling pressure
parameter λ1 is defined as (see Eq. 6.13c)
γ w (D − H )(1 − ν sν θ )
λ1 = (6.28)
Eξ
The critical water depth in which the submerged dome will buckle is given by
Eξ
D = λ1 +H (6.29)
γ w (1 − ν sν θ )
In the subsequent sections, we present new buckling solutions for moderately
thick spherical and parabolic domes with various dimensions and boundary conditions
under their own selfweight and hydrostatic pressure. The validity of the method,
convergence and accuracy of solutions are also demonstrated.
6.3 Results and discussions
As an example, consider a dome of height H=3000cm submerged at the water depth D.
For the subsequent numerical calculations, the dome is assumed to be made from a
material with values of E s = Eθ =30.104 kgf/cm2, ν sθ = 0.3 and
γ a = 2.4 x 10 -3 kgf/cm 3 . The specific weight of water is assumed in the calculation
γ w = 1 x 10 −3 kgf/cm3 . These values give rise to
118
γ w H (1 − ν sν θ ) 4.55 x10 −7
λ2 = = (6.30)
Eξ ξ
γ a h(1 − ν sν θ ) γ a H (1 − ν sν θ )
λa = = = 1.092 x10 −6 (6.31)
Eξ E
With the given value of normalized thickness ξ , substituting Eqs. (6.30) and
(6.31) to Eq.(6.16), one obtains the critical buckling parameter λ1 . Therefore, the
critical buckling water depth can also obtained from Eq. (6.29)
Eξ
D = λ1 +H (6.32)
γ w (1 − ν sν θ )
Note that, in case of γ a = 0, one obtains the critical buckling of domes under
hydrostatic force only.
119
6.3.1 Spherical Domes
Similarly to part 5.3.1 of chapter 5, for spherical domes, the meridional curve is
defined by (see Fig. 6.4)
r0 = 2 R z − z 2 (6.33)
R z
where R = and z =
H H
r=f(z)
0
γ(D-H+z) p=
a
γa h
z
z
R D H
α
L L L L
Fig 6.4 Spherical dome under its own selfweight and hydrostatic pressure
By substituting Eq. (6.30) to Eq.(4.1), one can obtain the geometric properties of
spherical shells
r1 = r2 = R (6.34)
In view of Eqs. (6.4) and (6.5), the parameters of the hydrostatic component p1
are given by
R
n s1 = nθ 1 = − (6.35)
2
120
By substituting Eq. (6.34) to Eqs. (6.11) and (6.12), one obtains the parameters of
hydrostatic component p 2
R (3R − 2 z )z
ns 2 = − (6.36)
12 R − 6 z
R z (4 z − 9 R )
nθ 2 = − (6.37)
12 R − 6 z
Similarly, by substituting Eq. (6.34) to Eqs. (6.16) and (6.17), the selfweight
parameters is given by
R2
n sa = − (6.38)
2R − z
R 2 − 3z R + z 2
nθa = − (6.39)
2R − z
By substituting all the above geometric parameters into the eigenvalue equation
(6.27), one can obtain the buckling pressure parameters λ1 and the critical water depth
in which the submerged dome will buckle.
121
Convergence study
Tables 6.1 and 6.2 show the convergence of the critical buckling pressure parameter λ1
for weightless hemispherical domes under hydrostatic forces by setting the specific
weight of the dome material as γ a = 0.
Table 6.3 and 6.4 show the convergence of the critical buckling pressure parameter
λ1 of hemispherical domes under their own selfweight and hydrostatic pressure with
radius-to-thickness ratio from R/h = 10 to R/h = 300 with different boundary
conditions: clamped, simply supported.
The same as for the uniform pressure case, in order to obtain an accurate solution,
the number of terms in the power series is increased until the difference in the result is
less than or equal to 0.05%. It can be seen from the tabulated results that the critical
buckling pressure parameter λ1 achieves the required 0.05% accuracy when 40 terms
were taken in the polynomial functions. This number of terms is assumed to be
sufficient for accurate results of other spherical shells generated in this chapter.
122

hemispherical domes under hydrostatic pressure only
λ1 x103 Radius over thickness ratio R/h

10 25 100 300
3 222.39639 194.00132 188.12904 187.48133
6 118.52774 55.62063 26.93537 24.41495
9 112.81965 46.53730 14.11276 6.62415
12 110.92726 45.66285 11.46389 4.44392
15 109.83091 45.23730 11.11006 3.42276
Number 18 109.14228 44.92945 11.04762 3.14028
of 21 108.68325 44.69574 11.03112 3.03240
terms 24 108.36170 44.51550 11.02182 3.02392
27 108.12673 44.37456 11.01425 3.02328
30 107.94863 44.26252 11.00714 3.02328
33 107.80918 44.17175 11.00002 3.02328
36 107.69669 44.09661 10.99257 3.02328
39 107.60338 44.03287 10.98448 3.02328
40 107.57558 44.01358 10.98157 3.02328

supported hemispherical domes under hydrostatic pressure only

10 25 100 300
3 159.42869 102.91870 88.26668 86.78558
6 106.55644 46.41693 16.43394 8.50931
9 105.93322 43.85519 11.68065 4.81828
12 105.61462 43.50903 10.97080 3.21768
15 105.36832 43.36872 10.79344 2.96056
Number 18 105.18038 43.26883 10.77757 2.86980
of 21 105.03771 43.19032 10.77643 2.85274
terms 24 104.92840 43.12707 10.77605 2.85023
27 104.84323 43.07552 10.77583 2.85018
30 104.77550 43.03301 10.77566 2.85017
33 104.72046 42.99749 10.77551 2.85017
36 104.67470 42.96728 10.77537 2.85017
39 104.63575 42.94105 10.77523 2.85016
40 104.62397 42.93300 10.77517 2.85016
123

spherical hemispherical under its own selfweight and hydrostatic pressure

10 25 100 300
3 222.37027 193.97458 188.10213 187.45440
6 118.50084 55.59072 26.90317 24.38265
9 112.79475 46.51126 14.08138 6.58786
12 110.90279 45.63755 11.43356 4.40858
15 109.80665 45.21236 11.08176 3.38647
Number 18 109.11814 44.90478 11.02027 3.10635
of 21 108.65919 44.67126 11.00410 2.99859
terms 24 108.33768 44.49116 10.99500 2.99022
27 108.10275 44.35033 10.98760 2.98961
30 107.92467 44.23838 10.98065 2.98960
33 107.78525 44.14767 10.97368 2.98960
36 107.67278 44.07258 10.96640 2.98960
39 107.57948 44.00889 10.95849 2.98960
40 107.55168 43.98961 10.95564 2.98960

supported hemispherical domes under its own selfweight and hydrostatic pressure

10 25 100 300
3 159.39456 102.88405 88.23202 86.75092
6 106.52590 46.38259 16.39416 8.46966
9 105.90364 43.82325 11.64241 4.77710
12 105.58557 43.47845 10.93461 3.17736
15 105.33969 43.33871 10.75953 2.92116
Number 18 105.15208 43.23924 10.74441 2.83112
of 21 105.00965 43.16109 10.74334 2.81446
terms 24 104.90052 43.09812 10.74298 2.81209
27 104.81549 43.04679 10.74278 2.81204
30 104.74787 43.00448 10.74263 2.81204
33 104.69291 42.96911 10.74249 2.81203
36 104.64722 42.93903 10.74236 2.81203
39 104.60833 42.91292 10.74223 2.81203
40 104.59657 42.90490 10.74218 2.81203
124
Figure 6.5 shows the variation of the critical water depth Dcr = D / H for which
that the dome can be constructed without buckling failure. Results were generated for a
spherical dome with a normalized base radius L = 1,2 and 3. It can be seen that, for a
same value of normalized thickness ξ , the normalized critical water depth Dcr of the
dome has the smaller base radius L much larger. For example, for the same
normalized thickness ξ = 0.01, the critical water depth Dcr =13.039 for the dome with
a normalized base radius L =3, which is considerably larger than the critical water
depth Dcr =1.19 of the dome with a normalized base radius L =1.
200
L=3
160
L=2
120 L=1
Dcr = D/H
80
40
0
0 0.01 0.02 0.03 0.04
ξ= h / H
Fig 6.5 Variations of critical water depth Dcr = D / H with respect to
normalized thickness ξ = h / H of a hemispherical dome
125
6.3.2 Parabolic Domes
Similar to part 5.3.2 of chapter 5, for parabolic domes, the meridional curve is defined
by (see Fig. 6.6)
r0 = 4a z (6.40)
a L2 L2
where a = = =
H 4H 2 4
r=h(z)
0
γ(D-H+z) p= γa h
a
z z
r2 D H
r1
φ
L L L L
Fig 6.6 Parabolic dome under its own selfweight and hydrostatic pressure
Similarly the above spherical case, by substituting Eqs.(6.41) to Eqs (6.4), (6.5),
(6.11), (6.12), (6.16) and (6.17), one can obtain the geometric properties of parabolic
domes
• Meridional and circumference radii
3
2a (a + z ) 2
r1 = (6.41)
a2
r2 = 2 a (a + z ) (6.42)
• Hydrostatic component p1
n s1 = − a (a + z ) (6.43)
126
a (a + 2 z )
nθ 1 = − (6.44)
a (a + z )
• Hydrostatic component p 2
1
n s 2 = − z a (a − z ) (6.45)
2
⎛ az ⎞
nθ 2 = −2 z a (a − z )⎜⎜ z − ⎟ (6.46)
⎝ 4(a + z ) ⎟⎠
• Selfweight parameters:
n sa = −
(
2 a (a + z ) − a 2 + z a (a + z ) + a 3 (a + z ) ) (6.47)
3a z
nθa = −
( ( )
2a a 2 − z + a (a + z ) a − 2 z 2 ) (6.58)
3z (a + z )
Tables 6.5 and 6.6 show the convergence of the critical buckling pressure parameter
λ1 for weightless parabolic domes with L = 1 under hydrostatic forces by setting
specific weight of the dome material as γ a = 0. Tables 6.7 and 6.8 show the
convergence of the critical buckling pressure parameter λ1 of parabolic domes with
L = 1 under their own selfweight and hydrostatic pressure with height-to-thickness
ratios from H/h = 10 to H/h = 300 with different boundary conditions: clamped, simply
supported.
The same for the uniform pressure case, in order to obtain an accurate solution, the
number of polynomial terms is increased until the difference in the result is less than or
equal to 0.05%. It can be seen from the tabulated results that the critical buckling
pressure parameter λ1 achieves the required 0.05% accuracy when 30 terms were
127
taken in the polynomial function. This number of terms is assumed to be sufficient for
accurate results of other parabolic shells generated in this chapter.

parabolic domes with normalized base radius L = 1 under hydrostatic pressure only
Height-over-thickness ratio H/h

λ1 x103
10 25 100 300
3 306.20544 268.25757 260.00589 258.80373
6 160.01226 65.61563 37.26095 34.75325
9 151.54276 57.03116 15.24487 7.40797
Number 12 151.49825 56.44285 12.35274 3.89063
of 15 151.49777 56.43338 11.91667 2.90081
terms 18 151.49754 56.43308 11.85564 2.65044
21 151.49741 56.43287 11.85336 2.57042
24 151.49733 56.43274 11.85334 2.55510
27 151.49727 56.43264 11.85333 2.55401
30 151.49724 56.43257 11.85333 2.55399

supported parabolic domes with normalized base radius L = 1
under hydrostatic pressure only
Height-over-thickness ratio H/h

λ1 x103
10 25 100 300
3 163.07327 117.23260 106.27888 104.82850
6 110.59318 48.35053 14.77523 8.28281
9 110.04449 43.32113 10.75227 3.61208
Number 12 110.03074 43.24899 9.91127 2.36965
of 15 110.02476 43.24692 9.80371 2.12465
terms 18 110.02109 43.24579 9.80068 2.01849
21 110.01868 43.24504 9.80060 2.00114
24 110.01699 43.24452 9.80055 1.99960
27 110.01577 43.24414 9.80052 1.99952
30 110.01485 43.24386 9.80049 1.99951
128

parabolic domes with normalized base radius L = 1
under its own selfweight and hydrostatic pressure
λ1 x103 Height-over-thickness ratio H/h

10 25 100 300
3 306.17307 268.22526 259.97358 258.77143
6 159.98111 65.58308 37.22836 34.72065
9 151.51145 56.99835 15.21022 7.37320
Number 12 151.46692 56.41022 12.31829 3.85500
of 15 151.46644 56.40074 11.88241 2.86550
terms 18 151.46621 56.40044 11.82147 2.61528
21 151.46608 56.40023 11.81918 2.53525
24 151.46600 56.40010 11.81916 2.51992
27 151.46595 56.40000 11.81916 2.51882
30 151.46591 56.39993 11.81915 2.51880

supported parabolic domes with normalized base radius L = 1 under its own
selfweight and hydrostatic pressure
λ1 x103 Height-over-thickness ratio H/h

10 25 100 300
3 163.04084 117.20001 106.24627 104.79588
6 110.55973 48.31640 14.73944 8.24706
9 110.01104 43.28662 10.71679 3.57560
Number 12 109.99729 43.21445 9.87573 2.33351
of 15 109.99131 43.21238 9.76814 2.08843
terms 18 109.98764 43.21125 9.76511 1.98229
21 109.98522 43.21050 9.76502 1.96499
24 109.98354 43.20998 9.76498 1.96345
27 109.98232 43.20960 9.76494 1.96337
30 109.98139 43.20931 9.76492 1.96337
129
Figure 6.7 shows the variation of the critical water depth Dcr = D / H for which
the dome can be constructed without buckling failure. Results were generated for
parabolic domes with normalized base radius L = 1,2 and 3. It can be seen that, for a
same value of the normalized thickness ξ , the normalized critical water depth Dcr of
the dome has the smaller base radius L much larger. For example, for the same
normalized thickness ξ = 0.01, the critical water depth Dcr =13.398 for the dome with
normalized base radius L =3, which is considerably larger than the critical water depth
Dcr =1.137 of the dome with a normalized base radius L =1.
250
L=3
200
L=2
150 L=1
Dcr = D/H
100
50
0
0 0.01 0.02 0.03 0.04
ξ= h / H
Fig 6.7 Variation of critical water depth Dcr = D / H with respect to normalized
thickness ξ = h / H of parabolic dome
130
The applicability of the Ritz method for various kinds of domes is demonstrated by
solving the buckling problems of spherical and parabolic domes under selfweight and
hydrostatic pressure. Based on these examples, it can be concluded that the developed
Ritz method can be readily applied for the buckling analyses of arbitrarily shaped
rotational shells under their own selfweight and hydrostatic pressure.
131
CHAPTER 7
AGAINST BUCKLING
This chapter is concerned with the optimal design of moderately thick submerged
domes. In addition to the water pressure, we also take into consideration the
selfweight, which is a significant load for such long span structures. For a given dome
height, based on a family of domes that is defined by the meridian curve r0 = f ( z )
with f ′(0 ) = 0 and submerged in a given water depth, we seek the dome design for
minimum weight as well as maximum enclosed airspace whereby the dome will not
buckle under the hydrostatic pressure and its own weight. The performance index of
the optimization is formulated as the weighted sum of individual objectives in order to
obtain Pareto optimal solutions. The buckling analysis of the submerged dome is
carried out using the Ritz method that was presented in Chapter 6.
132
Chapter 7: Optimal Designs of Submerged Domes Against Buckling
Consider a dome of height H and uniform thickness h. The dome is formed by
rotating a curve, defined by r0 = f ( z ) with f ′(0 ) = 0 , about the vertical z axis as
shown in Fig. 7.1. The dome is deployed under a water depth of D and thus it is
subjected to a hydrostatic pressure p h = γ w (D − H + z ) as shown in Fig. 7.1a where
γ w is the specific weight of water. The loading due to its selfweight is p a = γ a h as
shown in Fig. 7.1b where h is the thickness of the dome and γ a the specific weight of
the dome material. The dome is also assumed to be free of geometrical and material
imperfections.
r=f(z)
0
γ(D-H+z) p= γa h
a
z z
r2 D H
r1
L L L L
(a) Hydrostatic pressure (b) Selfweight
Fig 7.1 Dome under selfweight and hydrostatic pressure
For a given dome height H and water depth Drc0, there is a family of domes
defined by r0 = f ( z ) (see Fig. 7.2) that will not buckle when deployed in a given
water depth Dcr0. Based on this family of domes, we seek the dome with the
maximum enclosed airspace S a and minimum weight Wa . The problem will be
specialized for the optimization of a family of spherical domes (Fig. 7.2) and
parabolic domes (Fig 7.3) and extensive results are presented in this chapter.
133
r=f(z)
0
H D
L L
Fig. 7.2 Family of spherical domes for a given dome height H
r=f(z)
0
H D
L L
Fig 7.3 Family of parabolic domes for a given dome height H
134
7.2 Method of Optimization
The weight Wa of a uniform thickness dome, defined by r0 = f ( z ) with f ′(0 ) = 0 ,
can be calculated as follows
H
Wa = 2γ a πh ∫ f ( z ) 1 + f ' ( z ) dz
2
(7.1)
0
The enclosed airspace S a of the dome is given by
H
S a = π ∫ [ f (z )] dz
2
(7.2)
0
Although Eqs. (7.1) and (7.2) are valid for any function f ( z ) , we first restrict our
study to a family of spherical domes for which the meridional curves is defined by
r0 = f ( z ) = 2 Rz − z 2 (7.3)
In view of the non-dimensional parameters in Eq. (4.20), Eq. (7.3) may be
expressed as
r0 = 2 R z − z 2 (7.4)
where R = R / H and z = z / H .
135
By substituting Eq. (7.4) into Eq. (7.1) and in view of Eq. (4.20), one obtains the
weight of a spherical dome as
Wap
Wap = 3
= πξ (L 2 + 1) (7.5)
H
Similarly, by substituting Eq. (7.5) into Eq. (7.2), one obtains the non-dimensional
enclosed airspace of a spherical dome as
S ap ⎛ L2 1 ⎞
S ap = = ⎜ + ⎟π (7.6)
H 3 ⎜⎝ 2 6 ⎟⎠
For a given dome height H, we can apply the Ritz method presented in Chapter 5
and the Bisection method (Kreyszig 1993) to seek for a family of spherical domes that
have the critical water depth D = Dcr 0 . In view of Eqs. (7.1) to (7.6), it can be easily
seen that, the spherical dome shape is defined by the dome height H and dome base
radius L. The dome weight and enclosed airspace increase with increasing base radius
L for a given dome height.
However, the objective function of this optimal design involves the maximum
enclosed airspace and the minimum the material weight. In general, for a given dome
height and water depth, these aforementioned objectives are in conflict with each
other. We have a bi-criterion optimization problem to deal with.
As the dome height is prescribed, the dome base radius L (which will be
H
confined to ≤ L ≤ 3H from practical considerations) is taken as the decision
3
variable. The performance index is formulated as the weighted sum of individual
136
objectives in order to obtain Pareto optimal solutions for the bi-criterion optimization
problem.
1
One introduces S ' a = as the objective function for the enclosed space. Thus,
Sa
seeking the maximum of enclosed space Sa is tantamount to seeking the minimum
enclosed space parameter Sa’.
In order to get a better result in the Pareto optimization, one restricts each
criterion to be bounded in range [0,1] by using the following normalizations
Wa − Wa min
Wˆ a = (7.7a)
Wa max − Wa min
S ' a − S ' a min

Sˆ ' a = (7.7b)
S ' a max − S ' a min
where Wa min , Wa min , S 'a min , S 'a max are the maximum and minimum values of the
weight and enclosed airspace parameters of the dome in prescribed range of L , i.e.
1
≤L ≤3
3
( )
The performance index J αˆ , βˆ ; L of the problem is given by
J (αˆ , L ) = αˆWˆ a + (1 − αˆ ) Sˆ ' a (7.8)
where 0 ≤ αˆ ≤ 1 is the weighting factor of the material weight Ŵa and 0 ≤ (1 − αˆ ) ≤ 1
denotes the weighting factor of the enclosed airspace parameter Sˆ 'a .
The bi-criterion optimization problem can be mathematically stated as
137
L L
[
J = min[ J (αˆ , L )] = min αˆWˆ a + (1 − αˆ ) Sˆ ' a ] (7.9a)
1
subject to ≤ L ≤ 3 and Dcr = Dcr 0 (7.9b)
3
A simple minimum weight optimization problem is obtained by setting αˆ = 1 for
minimum weight whereas αˆ = 0 corresponds to the optimization problem which
maximizes the enclosed airspace. For 0 < αˆ < 1 , the base radius L minimizing
J (α̂ , L ) gives the Pareto optimal solution.
In the optimization stages, one applied the Golden Section Search technique
(Kreyszig 1993) to determine the minimum value of the performance index J.
7.3 Results and Discussions
7.3.1 Spherical domes
As an example, one considers a spherical dome of height H = 3000cm. For the
subsequent numerical calculations, the dome is assumed to be made from a material
with E s = Eθ = 30x104 kgf/cm2, ν sθ = 0.3 and γ a = 2.4 x 10 -3 kgf/cm 3 . The
specific weight of water is assumed to be γ w = 1 x 10 −3 kgf/cm3 .
Results for the single objective optimization are presented first. Figure 7.4 shows
the variations of performance index J in the case of αˆ = 1 and αˆ = 0 . In the case of
αˆ = 1 , one obtains J (αˆ , L ) = Wˆ a , i.e. the performance index is the normalized dome
~
weight Wa . It is clear that the performance index J reaches the minimum value at the
boundary value of L = 1 / 3 , i.e. the minimum weight of the dome is obtained at the
lowest value of the normalized base radius L .
138
In the case of α = 0 , one obtains J (αˆ , L ) = Sˆ ' a , i.e. the performance index is the
normalized enclosed airspace parameter S ' a . It is clear that the performance index J
reaches the minimum value at the boundary value of L = 3 , i.e. the maximum
enclosed airspace is obtained at the largest value of the normalized base radius L in
1
the given practical range ≤ L ≤ 3.
3
0.8
α =1 α =0
0.6
J
0.4
0.2
0
1/3 0.5 1 1.5 2 2.5 3
L

normalized base radius L in case of αˆ = 0 and αˆ = 1 .
139
0.8
0.6
Wa
0.4
0.2
0
0.2 0.4 0.6 0.8 1
S'a
Fig 7.5 Trade-off curve of normalized dome weight Ŵa and

normalized enclosed airspace parameter Ŝ ' a of spherical domes
Figure 7.5 shows the trade-off curve of normalized dome weight Ŵa and
normalized enclosed airspace parameter Ŝ ' a . The contributions these criteria are
plotted against each other to give the trade-off between the weightings α̂ =1 and
α̂ =0.
140
0.8
α = 0.25
α = 0.5
0.6
α = 0.75
J
0.4
0.2
0
1/3 0.5 1 1.5 2 2.5 3
L
Fig 7.6 Variations of performance index J of spherical domes with respect to
normalized base radius L in case of α̂ = 0.25; 0.5 and 0.75
Figure 7.6 shows the variations of the performance index J in the case of α =0.25;
0.5 and 0.75. In the case of α̂ = 0.25, one obtains the minimum value of the
performance index J at L =1.2. For α̂ =0.5 and 0.75, the performance index reaches
the minimum value at L =1.5 and L = 1.9. It can be seen that the Pareto optimal result
is highly dependent on the weight coefficient α̂ which the design engineer has to
decide in consultation with the client.
141
7.3.2 Parabolic domes
Extending our research on optimal designs of submerged domes, we investigate the
optimal design of parabolic domes for which the meridional curve is defined by
r0 = 4az (7.10)
where a = L2 / (4 H ) .
In view of the non-dimensional parameters in Eq. (4.20), Eq. (7.10) may be
expressed as
r0 = 4a z (7.11)
a L2 L2 z
where a = = 2
= and z = . By substituting Eq. (7.11) into Eq. (7.1) and
H 4H 4 H
in view of Eq. (4.20), one obtains the weight of a parabolic dome as
Wap =
Wap
H 3
=
πξL
6
(L ( L
2 2
)
+ 4 − L + 4 L2 + 4 ) (7.12)
Similarly, by substituting Eq. (7.5) into Eq. (7.2), one obtains the non-dimensional
enclosed airspace of a parabolic dome as
S ap πL 2
S ap = = (7.13)
H3 2
142
As an example, one considers a parabolic dome of height H = 3000cm. For the
subsequent numerical calculations, the dome is assumed to be made from a material
with E s = Eθ = 30x104 kgf/cm2, ν sθ = 0.3 and γ a = 2.4 x 10 -3 kgf/cm 3 . The
specific weight of water is assumed to be γ w = 1 x 10 −3 kgf/cm3 .
Results for the single objective optimization are presented first. Figure 7.7 shows
the variations of performance index J in the case of αˆ = 1 and αˆ = 0 . Similarly for
the spherical domes, the minimum weight of the dome is obtained at the lowest value
of the normalized base radius L = 1 / 3 and the maximum enclosed airspace is
obtained at the largest value of the normalized base radius L = 3
0.8
α =1 α =0
0.6
J
0.4
0.2
0
1/3 0.5 1 1.5 2 2.5 3
L
normalized base radius L in case of αˆ = 0 and αˆ = 1 .
143
0.8
0.6
Wa
0.4
0.2
0
0.2 0.4 0.6 0.8 1
S'a
Fig 7.8 Trade-off curve of normalized dome weight Ŵa and

normalized enclosed airspace parameter Ŝ ' a of parabolic domes
Figure 7.8 shows the trade-off curve of normalized dome weight Ŵa and
normalized enclosed airspace parameter Ŝ ' a The contributions these criteria are
plotted against each other to give the trade-off between the weightings α̂ =1 and
α̂ =0.
144
0.8
α = 0.25
α = 0.5
0.6
α = 0.75
J
0.4
0.2
0
1/3 0.5 1 1.5 2 2.5 3
L
Fig 7.9 Variations of performance index J of parabolic domes with respect to
normalized base radius L in case of α̂ = 0.25; 0.5 and 0.75
Figure 7.9 shows the variations of performance index J in the case of α =0.25; 0.5
and 0.75. In the case of α̂ = 0.25, one obtains the minimum value of the performance
index J at L =0.8. For α̂ =0.5 and 0.75, the performance index reaches the minimum
value at L =1.1 and L = 1.4. It can be seen that the Pareto optimal result is highly
dependent on the weight coefficient α̂ which the design engineer has to decide in
consultation with the client.
145
This chapter is concerned with the Pareto optimization of a submerged moderately
thick dome. The performance index for the optimization problem is formulated as the
weighted sum of the dome weight and the enclosed airspace of the dome. It can be
seen that the Pareto optimal solutions are highly dependent on the weighting
coefficient α which has to be decided by the design engineer in consultation with the
client.
146
CHAPTER 8
CONCLUSIONS AND RECOMMENDATIONS
8.1 Summary and Conclusions
This thesis was concerned with the optimal design of submerged domes where both
strength and buckling criteria have been taken into consideration. Membrane analysis
was carried out and minimum weight designs of submerged domes of uniform and
constant strength were investigated with the view to provide better designs for the
construction of submerged domes. The elastic buckling problem of moderately thick
domes was also studied. By using the Ritz method, the buckling capacities of
moderately thick domes under their own selfweight and hydrostatic pressure were
obtained. Moreover, based on a family of spherical and parabolic domes submerged
under a given water depth, the optimal dome shapes for maximum enclosed airspace
and minimum weight were determined.
147
Chapter 8: Conclusions and Recommendations
The first part of the thesis presented the membrane analysis and minimum weight
design of submerged spherical domes. An analytical expression, in the form of power
series, for the thickness variation of a submerged spherical dome of uniform strength
design as governed by the Tresca yield condition was presented. Numerical examples
showed that 9 terms in the power series sufficed for accurate solutions. Further, the
optimal subtended angle αopt and the optimal dome height H opt for the minimum
weight design of spherical domes were determined. It was found that α opt varies
within a narrow range of 1 radian ≤ α opt ≤ 1.25 radians. The insensitivity of the dome
weight over this range, which contains the optimal subtended angle, is a good feature
for engineers as it means that there is some flexibility when designing the dome shape
without compromising too much on the optimum weight.
Next, we extended the formulation to submerged domes of general shapes. The
equations governing the geometry of constant strength domes under combined
hydrostatic pressure, selfweight and skin cover load were derived. These equations
described the curvature and thickness variation of the dome as well as the Cartesian
coordinates of its meridian. The equations were purposely expressed in terms of the
arc length s as measured from the apex of the dome instead of using the Cartesian
coordinate system. This allowed the entire shape of the submerged dome to be
determined in a single integration process even in the presence of vertical or infinite
slope that may be encountered in the meridian curve. Based on parametric studies of
dome shapes under different water depths and selfweight, one may understand better
the optimal shape of submerged domes.
In the second part of the thesis, the optimal design of domes against buckling was
investigated. Although buckling of shells under compressive loading is of practical
148
significance in the design of these structures, most studies found in the open literature
thus far have focused on spherical domes modelled by classical thin shell theory. In
the present study, we developed the model and solution technique to predict the
critical buckling pressure of moderately thick rotational shells generated by any
meridional shape under external pressure. In order to capture the effect of transverse
shear deformation, which is significant for moderately thick shells, Mindlin shell
theory was used. Based on Mindlin shell theory, the energy functional was first
derived and the Ritz method was used to derive the eigenvalue equation. The Ritz
method was automated to handle any boundary conditions. This was made possible by
adopting Ritz functions formed by taking the product of mathematically complete
polynomial functions and boundary equations raised to appropriate powers; the latter
ensured the satisfaction of the geometric boundary conditions at the outset. The
desired accuracy of the results can be achieved by taking appropriate degree of
polynomials for approximating the displacement functions. The presented buckling
results were more accurate than those reported earlier (such as by Uddin 1989, and
Muc 1992) since the effect of transverse shear deformation was incorporated.
Moreover, the Ritz method developed in the thesis was simple to understand and to
code. The Ritz results should be useful benchmark data for analysts developing
numerical techniques for shell analysis.
Upon establishing the validity of the Ritz formulation and computer code and its
ability to furnish accurate buckling results for dome structures under uniform
pressure, we extended the work to submerged domes with allowance for the effect of
selfweight. New solutions for the buckling pressure of moderately thick spherical and
parabolic shells of various dimensions and boundary conditions were presented.
Further, based on a family of spherical and parabolic domes associated with a given
149
dome height submerged under a water depth, we determined the Pareto optimal dome
designs for the maximum enclosed airspace and minimum weight.
The vast optimal dome design data presented in this thesis should serve as a rich
reference source for researchers and engineers who are working on analysis and
design of shell structures.
8.2 Recommendations for Future Studies
The analysis and design of submerged domes involves the consideration of many
factors which expand the scope of this study for future research. Below are some
recommendations for future studies.
8.2.1 Domes with very large thickness
In this thesis, we dealt with moderately thick domes. The Mindlin shell theory is
adequate for the treatment of such shell structures. However, when the dome has a
very large thickness for deep sea deployment, it is necessary to use three-dimensional
elasticity theory to account for the thickness effect. Preliminary research along this
line has been initiated by Kang and Leissa (2005).
8.2.2 Non-axisymmetric domes
This thesis only dealt with axisymmetric dome structures as axisymmetric domes are
one of the most popular dome shapes. However, research should be extended to
investigate other non-axisymmetric domes such as a dome with a projected
rectangular or square plan area. Least weight designs of such domes approximated by
archgrids were investigated by Rozvany et al. (1982), Alwis and Wang (1985) and
150
Thevendran and Wang (1986). However, the buckling capacities of such archgrids
have yet to be studied.
8.2.3 Vibration of submerged domes
This thesis focused on identifying the optimal design of submerged spherical and
general dome structures. During the analysis of submerged dome structures, we
considered the buckling problem which is the most important criterion in designing
thin shell structures. However, it is also important to consider the vibration behaviour
of such shell structures so as to avoid the resonant frequencies which may be excited
by wave induced vibrations. In such problems, the computational method used by
Kang and Leissa (2005) on the free vibration of domes can be applied to solve the
vibrations of domes under hydrostatic pressure.
8.2.4 Other design loads on submerged domes
In this thesis, we considered selfweight, skin cover and the hydrostatic pressure as the
design loads on submerged domes. Although, in deep water, the hydrostatic pressure
is the largest load acting on submerged domes, other environmental loads such as
wave and current loads as well as incidental loads such as blast loads must be taken
into account.
151
REFERENCES
Abdul Azis, P. K., Al-Tisan, I., Al-Daili, M., Green, T.N., Dalvi, A.G.I. and Javeed,
M.A. (2002). “Effects of environment on source water for desalination plants
on the eastern coast of Saudi Arabia.” Desalination, 132(1-3), 29-40.
Alwis, W.A.M. and Wang, C.M. (1985). “On optimal archgrids.” Engineering
Optimization, 8(4), 315-331.
Archer, R.R. (1956). On the post buckling behavior of thin spherical shells, PhD thesis,
Massachusetts Institute of Technology.
Aron, H. (1874). “Das Gleichgewicht und die Bewegung einer unendluch dünnen,
beliebig gekrümmten, elastischen Shale.” Journ. für reine und ang. Math, 78.
Baiz, P.M., Aliabadi, M.H. (2007).“Buckling analysis of shear deformable shallow

shells by the boundary element method.” Engineering Analysis with Boundary
Elements, 31, 361–372
Barski M, Krużelecki J.(2005). “Optimal design of shells against buckling by means of

the simulated annealing method.” Structural and Multidisciplinary
Optimization, 29, 61–72.
Barski, M., Krużelecki J. (2005). “Optimal design of shells against buckling under
overall bending and external pressure.” Thin-Walled Structures, 43, 1677–1698
152
Beltramy, E. (1881). “Sull equibrio delle supperficie flessibili ed inestendibili.” Mem.
R. Acad. Sci di Bologna.
Beltzer, A. I.(1990). “Engineering analysis via symbolic computation-a breakthrough.”

Applied Mechanics Review. 43, 127.
Bhat, R.B.(1985). “Natural frequencies of rectangular plates using characteristic

orthogonal polynomials in the Rayleigh–Ritz method.” Journal of Sound and
Vibration, 102, 493–499.
Błachut J. (1987). “Combined axial and pressure buckling of shells having optimal
positive gaussian curvature.” Computers and Structures, 26(3), 513–9.
Błachut J.(1987). “On optimal barrel-shaped shells under buckling constraints.” AIAA
Journal, 25, 186–8.
Budiansky, B. (1959). “Buckling of clamped shallow spherical shells.” Proceeding of

the Symposium on the Theory of Thin Elastic Shells, Amsterdam, North Holland.
Bushnell, D. (1976). “BOSOR5-Program for buckling of elastic plastic complex shells

of revolutions including large deflections and creep.” Computers and structures,
6, 221.
Bushnell, D. (1984). “Computerized analysis of shells-governing equations.”

Computers and Structures, 18(3), 471-536.
Chai, Y.H. and Wang, C.M. (2005). “Approximate solution for the shape of submerged
funicular arches with selfweight.” Journal of Structural Engineering, ASCE, 131
Chai, Y.H. and Kunnath, S.K. (2003). “Geometry of submerged funicular arches in
Cartesian coordinates.” Journal of Structural Engineering, 129(9), 1087-1092.
Chao C.C., Tung T.P., Chern Y.C. (1988). “Buckling of thick orthotropic spherical
shells.” Composites Structures, 9, 113-137.
Cohen, G.A. (1981). “FASOR- A program for stress, buckling, and vibration of shells
of revolutions.” Advanced Engineering Software, 3(4), 155-162.
153
References
Computers and Structures, Inc. (2007). SAP2000 Linear and Nonlinear, Static and
Dynamics Analysis Design of Three Dimensional Structures, Berkeley,
California, USA.
Cowan, H.J. (1977). “A history of masonry and concrete domes in building

construction.” Building and Environment, 12, 1-24.
Dumir, P.C., Dube, G.P. and Mallick A. (2005). “Axisymmetric buckling of laminated
thick annular spherical cap.” Communications in Nonlinear Science and
Numerical Simulation, 10, 191–204
Dumir, P.C., Gandhi, M.L. and Nath, Y. (1984). “ Axisymmetric static and dynamic
buckling of orthotropic shallow spherical caps with flexible supports.” Acta
Mechanica, 52, 93.
Dym, C. L. (1974). Introduction to the Theory of Shells. Pergamon Press, Oxford.
El-Deeb, K.M. and Royles, R. (1993). “Dynamic and static buckling assessment of an
echinodome.” Computer and structures, 46(5), 899-903.
Fu, Y. (1997). “Some asymptotic results concerning the buckling of a spherical shell of
arbitrary thickness.” International Journal Of Non-Linear Mechanics, 33(6),
1111-1122.
Fung, T.C. (2003). “Shapes of submerged funicular arches.” Journal of Engineering

Mechanics, 129(1), 120-125.
Gajewski A, Życzkowski M. (1988). Optimal Structural Design under Stability

Constraints. Kluwer Academic Publishers. The Netherlands.
Gajewski A. (1990). “Multimodal optimization of uniformly compressed cylindrical

shells.” Engineering Optimization in Design Processes. New York: Springer,
275–82.
Garlerkin, B.G. (1942). “Equilibrium of an elastic spherical shell.” Prikl. Mat. Mekh.
Akademiya Nauk. S.S.S.R., IV, 6.
154
References
Gavin, H.P. and Reilly, K.J. (2000). “Submerged funicular arches.” Journal of
Structural Engineering, 126(1), 120-125.
Geannakakes, G.N. (1995). “Natural frequencies of arbitrarily shaped plates using the
Rayleigh-Ritz method together with natural co-ordinate regions and normalized
characteristic orthogonal polynomials.” Journal of Sound and Vibration, 182(3),
441-478.
Gol’denweiger, A.L. (1946). “Certain examples of the integration of the equations of

the theory of thin shells.” Prikl. Mat. Mekh. Akademiya Nauk. S.S.S.R, X, 3.
Haftka, R.T. and Gürdal, Z. (1992). Elements of Structural Optimization. Kluwer,

Dordrecht.
Hinton, E., Sienz, J. and Ozakca, M. (2003). Analysis and Optimization of Prismatic
and Axisymmetric Shell Structures - Theory , Practice and Software. Springer,
London.
Hou, Y., Wei, G.W., Xiang, Y. (2005). “DSC Ritz method for the free vibration
analysis of Mindlin plates.” International Journal for Numerical Methods in
Engineering, 62, 262-288.
Huang, N.C. (1964). “Unsymmetrical buckling of thin shallow spherical.” Journal of

Applied Mechanics, 31, 447.
Hyman B.I. and Lucas Jr A.W. (1971). An optimum design for the instability of
cylindrical shells under lateral pressure. AIAA Journal, 9, pp. 738–40.
Ichikawa, Y. and Yonezawa, T. (2003). “The outline of the MH21 program and the
R&D plan of methane hydrate development system for offshore Japan.”
Proceedings of International Symposium on Ocean Space Utilization, National
Maritime Research Institute, 28 Jan – 1 Feb 2003, Tokyo, Japan, 398-404.
Imam, M.H.(1982). “Three dimensional shape optimization.” International Journal of

Numerical Methods in Engineering, 18, 661–73.
155
References
Infante Barbosa, J., Mota Soares, C.M. and Mota Soares, C.A. (1991). “Sensitivity
analysis and shape optimal design of axisymmetric shell structures.” Computing
Systems in Engineering, 2(5,6), 525-533.
Ioakimidis, N.O. (1992a). “Applications of MATHEMATICA to the direct semi-

numerical solution of finite element problems.” Computers and Structures, Vol.
45 No. 5/6, 833-9.
Ioakimidis, N.O. (1992b). “Applications of MATHEMATICA to the direct solution of

torsion problems by the energy method.” Computers and Structures, Vol. 43,
No. 4, 803-7.
Ioakimidis, N.O. (1992c). “Semi-numerical iterative series solution of linear algebraic

equations with MATHEMATICA.” Communications Applied Numerical
Methods, Vol. 8, 421-9.
Issler, W. (1964). “Membranschalen gleicher Festigeit.” Ing. Arch., 33, 330-345.
Kang, J.H., Leissa, A.W. (2005). “Three-dimensional vibrations of thick spherical

shell segments with variable thickness.” International journal of Solids and
Structures, 37 (2000), 4811-4823.
Kawai, T. (1974). Analysis of Buckling Problems, Baifukan, Tokyo
Kirchhoff, G. (1876). Vorlesungen über Mathematische Physik, Mechanik, Vol. 1.
Koiter, W.T. (1969). “The nonlinear buckling problem of a complete spherical shell
under uniform external pressure.” Proc. Kon. Nederl. Acad. Wet. Amsterdam B.
Komai, T. (2003). “Replacement and sequestration techniques of CO2 hydrate in the

developing gas hygrates.” Proceedings of International Symposium on Ocean
Space Utilization, National Maritime Research Institute, 28 Jan – 1 Feb 2003,
Tokyo, Japan, 381-384.
Kraus, H. (1967). Thin Elastic Shells, John Wiley & Sons. Inc, New York.
Kreyszig, E. (1993). Advanced Engineering Mathematics, 7th ed., John Wiley & Sons
Inc., Canada.
156
References
Kreyszig, E. (1993). Advanced Engineering Mathematics, 7th ed., John Wiley & Sons
Inc., Canada, 1040-1043.
Krużelecki J. and Trzeciak P. (2000). “Optimal design of axially symmetrical shells

under hydrostatic pressure with respect to their stability.” Structural and
Multidisciplinary Optimization, 19(2), 148–54
Krużelecki, J. (1988). “Optimal design of a cylindrical shell under overall bending

with axial force.” Bull Pol Acad Sci Tech Sci, 36(3–4), 141–50.
Krużelecki, J. and Trzeciak, P.( 2000). “Optimal design of axially symmetrical shells
under hydrostatic pressure with respect to their stability.” Structural and
Multidisciplinary Optimization, 19(2), 148–54.
Krużelecki, J. and Życzkowski, M. (1984). “Optimal design of an elastic cylindrical

shell under overall bending with torsion.” Solid Mechanics Archive, 9, 269–306.
Krużelecki, J. and Życzkowski, M. (1985). “Optimal design of shells—a survey.” Solid

Mechanics Archives, 10, 101–70.
Kukreti, A.R., Farsa, J. and Bert, C.W. (1996). “Differential quadrature and Rayleigh–
Ritz methods to determine the fundamental frequencies of simply supported
rectangular plates with linearly varying thickness.” Journal of Sound and
Vibration, 189, 103–122.
Ladson, L.S., Waren, A.D., Jain, A., and Ratner, M. (1978). “Design and testing a
generalized reduced gradient code for nonlinear programming.” ACM
Transaction on Mathematical Software, 4(1), 34-50.
Lamé, G. and Claperon, B.P.E. (1833). “Mémoire sure l’ équilibre interieur des corps
solides.” Mém. prés. par div. savants, 4.
Lecornu, Z. (1938). “Sur l’équilibre des surfaces flexibles et inextendibles.” Journ. de

l’école Polytechnique, c.XLVIII.
Leissa, A.W. (2005). “The historical bases of the Rayleigh and Ritz methods.” Journal
of Sound and Vibration, 287, 961-978.
157
References
Levy, R. and Spillers, W.R. (1989). “Optimal design for axisymmetric cylindrical shell
buckling.” Journal of Engineering Mechanics, 115, 1683-1690.
Li, Q.S., Liu, J. and Tang, J. (2003).. “Buckling of shallow spherical shells including
the effects of transverse shear deformation.” International Journal of
Mechanical Sciences, 45, 1519-1529.
Liew, K.M. and Lim, C.W. (1995). “A Ritz vibration analysis of doubly-curved
rectangular shallow shells using a refined first-order theory.” Computer Methods
in Applied Mechanics and Engineering, 127, 145–162.
Liew, K.M., Lim, M.K., Lim, C.W., Li, D.B. and Zhang, Y.R.(1995). “Effects of
initial twist and thickness variations on the vibration behaviour of shallow
conical shells.” Journal of Sound and Vibration, 180 (2), 271–296.
Liew, K.M., Wang, C.M., Xiang, Y. and Kitipornchai, S. (1998). Vibration of Mindlin
Plates, Elsevier, Amsterdam.
Liew, K.M. (1990). The Development of 2-D Orthogonal Polynomials for Vibration of
Plates, Ph.D. Thesis, National University of Singapore, Singapore.
Liew, K.M. and Wang, C.M. (1992). “Vibration analysis of plates by pb-2 Rayleigh-
Ritz method: mixed boundary conditions, reentrant corners and curved internal
supports.” Mechanics of Structures and Machines, 20(3), 281-292.
Liew, K.M. and Wang, C.M. (1993). “pb-2 Rayleigh-Ritz method for general plate
analysis.” Engineering Structures, 15(1), 55-60.
Liew, K.M., Chen, X.L. and Reddy, J.N. (2004). “Mesh-free radial basis function
method for buckling analysis of non-uniformly loaded arbitrarily shaped shear
deformable plates.” Computer Methods in Applied Mechanics and Engineering,
193, 205-224.
Lim, C.W. and Liew, K.M.(1994). “A pb-2 Ritz formulation for flexural vibration of
shallow cylindrical shells of rectangular planform.” Journal of Sound and
Vibration, 173 (3), 343–375.
158
References
Lim, C.W, Ma, Y.F., Kitipornchai, S., Wang, C.M. and Yuen, R.K.K. (2003).
“Buckling of vertical Cylindrical Shells Under Combined End Pressure and
Body Force.” Journal of Engineering Mechanics ASCE, 876-884
Love, A.E. (1888). “The small free vibrations and deformations of a thin elastic shell”.
Phil. Trans. Roy. Soc., London. Ser. A, 179, 491-546
Magnucki, K., Lewinski, J. and Stasiewicz, P.(2004). “Optimal sizes of a ground-based

horizontal cylindrical tank under strength and stability constraints.”
International Journal of Pressure Vessels and Piping, 81(12), 913–917.
Mindlin, R.D. (1951). “Influence of rotatory inertia and shear on flexural motions of
isotropic, elastic plates.” Transaction of ASME, Journal of Applied Mechanics, 18,
31-38.
Mizusawa, T. (1986). “Natural frequencies of rectangular plates with free edges.”

Journal of Sound and Vibration, 105, 451-459.
Muc, A. (1992). “On the buckling of composite shells of revolution under external
pressure.” Composite Structures, 21(2), 107-119.
Mushtari, K. M. (1949). “A qualitative study if the state of stress of an elastic shell

subject to small deformations and arbitrary displacements.” Prikl. Mat. Mekh.
Akademiya Nauk. S.S.S.R, XIII, 2.
Nakamura, H., Dow, M. and Rozvany, G.I.N. (1981). “Optimal spherical cupola of
uniform strength: allowance for selfweight.” Ing. Arch, 51, 159-181.
Narita, Y. and Leissa A.W. (1990). “Buckling studies for simply supported
symmetrically laminated rectangular plates.” International Journal of
Mechanical Sciences, 32, 909–924.
Neut, A. van der (1932). The elastic stability of thin-walled sphere. Dissertation, Delft.
Novozhilov, V.V. (1970). The Theory of Thin Shells, Wolters-Noordhoff Publishing,

The Netherlands, Groningen, 138-147.
159
References
Ohga, M., Shigematsu, T. and Kawaguchi, K. (1996). “Buckling analysis of thin-

walled members with variable thickness.” Journal of Structural Engineering,
121, 919-924.
Parlett, B.N., Dhillon, I.S.(2000). “Relatively robust representations of symmetric

tridiagonals.” Linear Algebra and its Applications, 309, 121–151
Pesciullesi, C., Rapallini, M., Tralli, A. and Cianchi, A. (1997). “Optimal spherical
masonry domes of uniform strength.” Journal of Structural Engineering, 123,
203-209.
Prager, W. and Rozvany, G.I.N. (1980). “Optimal spherical cupola of uniform

strength.” Ing. Arch., 49, 287-293.
Rayleigh, J.W. (1977). Theory of Sound, Macmillan Vol 1 (reprinted by Dover

Publications. 1945).
Rayleigh, L. (1911). “On the calculation of Chladni’s figures for a square plate.”
Philosophical Magazine Sixth Series, 22, 225-229.
Reddy, J.N. (2004). Mechanics of Laminated Composite Plates and Shells: Theory and
Analysis. Second edition. CRC Press, Florida.
Redekop, D. (2005). “Buckling analysis of an orthotropic thin shell of revolution using

differential quadrature.” International Journal of Pressure Vessels and Piping,
82, 618–624
Reiss, E.L. (1958). “Axially buckling of shallow spherical shells under external
pressure.” Journal of Applied Mechanics, Trans. ASME, 25, 556.
Reissner, E. (1945). “The effect of transverse shear deformation on the bending of

elastic plates.” Journal of Applied Mechanics, 12, A69-A77.
Ritz, W. (1909). “Uber eine neue Mehode zur Lösung gewisser Variationprobleme der
mathematischen Physik.” Journal fur Reine and Angewandte Mathematik 135,
1-61.
160
References
Ross, C. T. F. (1990). Pressure Vessels under External Pressure. Chapman & Hall
Ross, C. T. F. (1996). “Vibration and elastic instability of thin-walled domes under

uniform external pressure.” Thin-Walled Structures, 26(3), 159-177.
Ross, C. T. F., Mackney, M. D. A., (1983). “Deformation and stability studies of thin-
walled domes under uniform pressure, J. Strain Analysis, 18, 167-172
Ross, C.T.F., Youster, P. and Sadler, R. (2001). “The buckling of plastic oblate hemi-
ellipsoidal dome shells under external hydrostatic pressure.” Ocean Engineering,
28(7), 789-803.
Ross, C.T.F., Huat B.H., Chei T.B., Chong C.M. and Mackney M.D.A.(2003). “The
buckling of GRP hemi-ellipsoidal dome shells under external hydrostatic
pressure.” Ocean Engineering, 30, 691–705.
Royles, R. and Llambias, J.M. (1985). “Buckling behavior of an underwater storage

vessel.” Experimental Mechanics, 25, 421-428.
Royles, R., Sofoluwe, A.B., Baig, M.M. and Currie, A.J. (1980). “Behavior of
underwater enclosures of optimum design.” Strain, 16(1), 12-20.
Rozvany, G.I.N., Wang, C.M. and Dow, M. (1982). “Prager-structures: arch-grids and
cable networks of optimal layout.” Computer Methods in Applied Mechanics
and Engineering, 31, 91-133.
Rysz, M. and Życzkowski, M. (1989). “Optimal design of a cylindrical shell under

overall bending with axial force with respect to creep stability.” Structural
Optimization, 1(1), 29–36.
Sayir, M. and Schumann, W. (1972). “Zu den anissotropen Membranschalen mit

gegebenefalls gleicher Festigkeit.” A. Angew. Math. Phys., 23, 815-827.
Schumann, W. and Wüthrich, W. (1972). “Über Schalen gleicher Festigkeit.” Acta

Mechanica, 14, 89-197.
Smith, S.T., Boyle, J.M., Garbow, B.S. Ikebe, Y., Klema, V.C. and Moler, C.B.
(1974). Matrix Eigensystem Routines-EISPAC Guide. Springer-Verlag, New York.
161
References
Sofiyev, A.H. and Aksogan, O. (2004). “Buckling of a conical thin shell with variable
thickness under a dynamic loading.” Journal of Sound and Vibration, 270, 903-
915.
Sofoluwe, A.B., Royles, R. and Ibidapo-Obe, O. (1981). “An improved numerical

approach to the analysis of the echinodome.” Mechanics Research
Communications, 8(4), 237-243.
Sokolovskii, V.V. (1938). “On membrane shells of revolution.” Prikl. Mat. Mekh.
Akademiya Nauk. S.S.S.R., I, 3.
Sun, G. (1989). “A practical approach to optimal design of laminated cylindrical shells

for buckling.” Composites Science and Technology, 36, 243-253.
Sun, G. and Hansen, J. S.(1988). “Optimal design of laminated composite circular-

cylindrical shells subjected to combined load.” Journal of Applied Mechanics.
55, 136-142.
Tennyson, R. C. and Hansen J. S. (1983). “Optimum design for buckling of laminated

cylinders.” The buckling of Structures in Theory and Practice (Edited by J. M.
T. Thomnson and G. W. Hunt). Cambridge University Press, Cambridge.
Thevendran, V. and Wang, C.M. (1986). “On the optimality criteria for archgrid.”
Journal of Structural Engineering ASCE, 112(1), 185-189.
Thurston, G.A. (1961). “A numerical solution of the nonlinear equations for

axisymmetric bending of shallow spherical shells.” Journal of Applied
Mechanics, 28, 557
Timoshenko, S.P. and Woinowsky-Krieger, W. (1959). Theory of Plates and Shells,

McGraw-Hill, New York, 2nd Edition, 442-445.
Tsai, S.W. and Pagano, N.J. (1968). “Invariant properties of composite materials.”
Composite Materials Workshop, Technomic Publishing Co., Stamford,
Connecticut, 233-253.
Uddin, M. W. and Haque, M.M. (1994). “Instability of semi-ellipsoidal shells.”

International Journal of Pressure Vessels and Piping, 58, 65-74.
162
References
Ugural, A.C. (1999). Stresses in Plates and Shells, 2nd Edition, McGraw-Hill, USA.
Uysal, H., Gula, R. and Uzmanb, U. (2007). “Optimum shape design of shell
structures.” Engineering Structures, 29, 80–87
Vermeulen, A. H. and Heppler, G. R.(1998). “Structural analysis of shells by the b-

spline field approximation method.” Computers and Structures, 68 (1-3), 167-
179.
Vlasov, V.Z. (1939). “Membrane theory of thin shells, formed by second order
surfaces.” Plates and Shells, Gosstroiizdat.
Walker, M., Reisst, T. and Adalit, S. (1997). “Multiobjective design of laminated

cylindrical shells for maximum torsional and axial buckling loads.” Computers
and Structures, 62(2), 231-242.
Wang, C.M. and Kitipornchai, S. (1992). “Shooting-optimization technique for large

deflection analysis of structural members.” Engineering Structures¸ 14(4), 231-
240.
Wang, C.M. and Ler, C.W. (2003). “Optimization of submerged funicular arches.”
Mechanics Based Design of Structures and Machines, 31(2), 181-200.
Wang, C.M. and Wang, C.Y. (2002). “Funicular shapes of submerged arches.” Journal
of Structural Engineering, 128(2), 266-270.
Watanabe, E. and Utsunomiya, T. (2003). “Analysis and design of floating bridges.”

Progress in Structural Engineering and Materials, 5, 127-144.
Wolfram, S. (1999). Mathematica book, 4th Edition. Cambridge University Press, New
York.
Yang, M. F., Liang, C. C. and Chen, C. H. (1992). “A rational shape design of

externally pressurized torispherical dome ends under buckling constraints.”
Computers & Structures 43(5), pp. 839-851.
163
References
Yoshida, K. (2003). “A brief review of recent activities of VLFS in Japan.”

Proceedings of International Symposium on Ocean Space Utilization, National
Maritime Research Institute, Tokyo, Japan, 21-28.
Zbigniew, E.M. and Roman, T.N. (1991). Shells of Revolution. PWN- Polish Scientific
Publishers, Warszawa.
Ziegler, H. (1958). “Kuppelm gleicher festigkeit.” Ing. Arch., 26, 378-382.
Zingoni, A. (1997). Shell structures in civil and mechanical engineering - Theory and
closed-form analytical solutions, Thomas Telford Publishing, London.
Zoelly, R. (1915). Über ein Kinckungsproblem an der Kugelschale. Dissertation,

Zurich.
Życzkowski, M., Krużelecki J. and Trzeciak P. (2001). “Optimal design of rotationally

symmetric shells for buckling under thermal loadings.” Journal of Theory
Applied Mechanics, 39(2), 443–55.
Życzkowski, M. and Krużelecki, J. (1973). “Optimal design of shells with respect to

their stability.” IUTAM Symposiums on Optimization in Structural Design. New
York, Springer. 229–47.
Życzkowski, M. (1992). “Recent advances in optimal structural design of shells.”

European Journal of Mechanics - A/Solid, 11, 5–24.
164
APPENDIX
This part details the use of Mathematica (Wolfram 1999) to obtain the buckling
strength of rotational shells according to the Ritz method and the formulations
presented in Chapter 4.
• Material properties
νθs = 3 ê 10;
νsθ = 3 ê 10;
Eθ = 200 104 ;
Es = 200 104;
Gsz = Es ê 2 ê H1 + νsθL;
• Calculation of the material parameters for the stiffness matrix [K] (Eq.
4.24)
Eθ
Q11 = ;
Eb
Es
Q22 = ;
Eb
νφθ Eθ
Q12 = ;
Eb
H1 − νθs νsθL;
5 Gsz
Q44 =
6 Eb
165
ζ
r1@zD
A11@z_D = Q11 IntegrateA , 8ζ, −1 ê 2, 1 ê 2<E;
1+ ξ
r2@zD
ζ
1+ ξ
A12@z_D = Q12 Integrate@ 1, 8ζ, −1 ê 2, 1 ê 2<D;

ζ
r2@zD
1+ ξ
r1@zD
ζ
1+ ξ
ζ
r2@zD
1+ ξ
r1@zD
ζ
1+ ξ
B12@z_D = Q12 Integrate@ ζ, 8ζ, −1 ê 2, 1 ê 2<D;

ζ
r1@zD
D11@z_D = Q11 IntegrateA ζ2, 8ζ, −1 ê 2, 1 ê 2<E;
1+ ξ
r2@zD
ζ
1+ ξ

ζ
r2@zD
1+ ξ
r1@zD
ζ
1+ ξ
• Geometrical properties of spherical dome
è!!!!!!!!!!!!!!!!!
f@z_D = 2 z R − z2 ;
r0@z_D = f@zD;
r1@z_D = R;
r2@z_D = R;
η@z_D =
è!!!!!!!!!!!!!!!!!!!
R
H2 R − zL z
;
ns1@z_D = − ;
R
2
nθ1@z_D = − ;
R
2
166
• Geometrical properties of parabolic domes
L2
A= ;
è!!!!!!!
4
f@z_D = 2 A z ;
r0@z_D = f@zD;
2 HA HA + zLL 2
r1@z_D =
3
;
è!!!!!!!!!!!!!!!!
A2
r2@z_D = 2 A HA + zL ;
η@z_D = $%%%%%%%%%%%% ;
A +z
z
è!!!!!!!!!!!!!!!!!
ns1@z_D = − A HA + zL ;
è!!!!!!!!!!!!!!!!! j
nθ1@z_D = −2 A HA + zL i j
y
+ 1z
z;
k 2 H A + zL
A
{
• Basic functions need to be defined for different boundary functions
¾ Clamped edge (Eq. 4.42)
ηu = Hz − 1L z;
ηw = Hz − 1L;
ηψ = Hz − 1L z;
¾ Simply supported edge (Eq. 4.43)
ηu = Hz − 1L z;
ηw = Hz − 1L;
ηψ = z;
167
• The mathematically complete polynomial functions are formed as a list. (Eq.
4.28)
TableApi = ηu zi−1, 8i, 1, N1<E;

TableApi = ηw zi− N1−1, 8i, N1 + 1, N2<E;
TableApi = ηψ zi− N2−1, 8i, N2 + 1, N3<E;
• The matrix elements of the [K] matrix (Eqs. (4.31-35)
∂z r0@zD
TableAqi = pi , 8i, 1, N1<E;
η@zD r0@zD
TableAqi = , 8i, N1 + 1, N2<E;
r2@zD
pi
Table@qi = 0, 8i, N2 + 1, N3<D;

K1@z_D = Table@A11@zD qi qj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
Table@wi = 0, 8i, 1, N1<D;

Table@wi = 0, 8i, N1 + 1, N2<D;
∂z r0@zD
TableAwi = ξ pi, 8i, N2 + 1, N3<E;
η@zD r0@zD
K2@z_D = Table@D11@zD wi wj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
TableAei = , 8i, 1, N1<E;

∂z pi
η@zD
TableAei = , 8i, N1 + 1, N2<E;
r1@zD
pi
Table@ei = 0, 8i, N2 + 1, N3<D;

K3@z_D = Table@A22@zD ei ej r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
168
Table@ri = 0, 8i, 1, N1<D;
Table@ri = 0, 8i, N1 + 1, N2<D;
TableAri = ξ z i , 8i, N2 + 1, N3<E;
∂ p
η@zD
K4@z_D = Table@D22@zD ri rj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
TableAti = − , 8i, 1, N1<E;

r1@zD
pi
TableAti = z i , 8i, N1 + 1, N2<E;

∂ p
η@zD
TableAti = pi, 8i, N2 + 1, N3<E;
K5@z_D = TableA A44@zD ti tj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<E;
K6a@z_D = Table@B12@zD qi wj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K6b@z_D = Table@B12@zD qj wi r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
K7a@z_D = Table@ B12@zD ei rj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K7b@z_D = Table@B12@zD ej ri r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
K8a@z_D = Table@A12@zD qi ej r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K8b@z_D = Table@ A12@zD qj ei r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
K9a@z_D = Table@D12@zD wi rj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K9b@z_D = Table@D12@zD wj ri r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
K10a@z_D = Table@B12@zD qi rj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K10b@z_D = Table@B12@zD qj ri r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
K11a@z_D = Table@B12@zD wi ej r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;

K11b@z_D = Table@B12@zD wj ei r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
• The matrix [K] can be obtained as follow (Eq. 4.30)
K@z_D = K1@zD + K2@zD + K3@zD + K4@zD + K5@zD + K6a@zD + K6b@zD + K7a@zD + K7b@zD +
K8a@zD + K8b@zD + K9a@zD + K9b@zD + K10a@zD + K10b@zD + K11a@zD + K11b@zD;
169
• The elements of matrix [M] can be obtained as Eq. (4.37-39)
∂z r0@zD
TableAyi = pi , 8i, 1, N1<E;
η@zD r0@zD
TableAyi = , 8i, N1 + 1, N2<E;
r2@zD
pi
Table@yi = 0, 8i, N2 + 1, N3<D;

M1@z_D = Table@nθ1@zD yi yj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
TableAui = , 8i, 1, N1<E;

∂z pi
η@zD
TableAui = , 8i, N1 + 1, N2<E;
r1@zD
pi
Table@ui = 0, 8i, N2 + 1, N3<D;

M21@z_D = Table@ns1@zD ui uj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
TableAgi = , 8i, 1, N1<E;

r1@zD
pi
TableAgi = − z i , 8i, N1 + 1, N2<E;

∂ p
η@zD
Table@gi = 0, 8i, N2 + 1, N3<D;
M31@z_D = Table@ns1@zD gi gj r0@zD η@zD, 8i, 1, N3<, 8j, 1, i<D;
• The matrix [M] can be obtained as follow (Eq. 4.36)
M@z_D = M1@zD + M2@zD + M3@zD;
For elastic buckling, the aforementioned eigenvalue problems can be expressed in the
standard form of a generalized eigenvalue problem and a standard eigenvalue routine,
for e.g. EISPAC (Smith et al., 1974), can be used to solve the problem. However, it
was found that the built-in function “Eigenvalues” in Mathematica (Wolfram, 1999)
can be used instead of Gaussian elimination. The use of built-in function
“Eigenvalues” is found to be faster than using the Gaussian elimination in the
determination of the eigenvalues for this class of problems. Eigenvalues function in
170
Mathematica used the function DSYEVR in LAPACK<www.netlib.org/lapack/>
routines to calculate the numerical eigen values and vectors of a real and symmetric
matrice.
LAPACK, the Linear Algebra PACKage, is a software library for numerical
computing written in Fortran 77. It provides routines for solving systems of
simultaneous linear equations, least-squares solutions of linear systems of equations,
eigenvalue problems, and singular value problems.
DSYEVR function computes selected eigenvalues and, optionally, eigenvectors
of a real symmetric matrix A. Eigenvalues and eigenvectors can be selected by
specifying either a range of values or a range of indices for the desired eigenvalues.
DSYEVR first reduces the matrix A to tri-diagonal form T with a call to DSYTRD.
Then, DSYEVR calls DSTEMR to compute the eigen spectrum using Relatively
Robust Representations (Parlett and Dhillon, 2000)
171
LIST OF AUTHOR’S PUBLICATIONS
Journal Papers
1. Vo, K.K., Wang, C.M. and Chai, Y.H. (2007), “Buckling analysis of moderately
thick rotational shells under uniform pressure using Ritz method.” Journal of
Structural Engineering, ASCE, 134(4), 593-607.
2. Vo, K.K., Wang, C. M. and Chai, Y. H. (2006), “Membrane analysis and
optimization of submerged domes with allowance for selfweight and skin cover
load.” Archive of Applied Mechanics, 75, 235-247.
3. Wang, C.M., Vo, K.K. and Chai, Y. H. (2006), “Membrane analysis and
minimum weight design of submerged spherical domes.” Journal of Structural
Engineering, ASCE, 132(2), 253-259.
172
List of Author’s Publications
Conference Papers
1. Vo, K.K., Wang, C.M. and Chai, Y.H. (2005) “Membrane analysis and
optimization of submerged domes with allowance for selfweight and skin cover
load.” Proceedings of the Tenth International on Civil, Structural and
Environmental Engineering Computing, ed. B.H.V. Topping, Civil-Comp Press,
30 Aug - 2 Sep 2005, Rome, Italy, 1-12.
2. Wang, C.M., Aung, T.M. and Vo, K.K. (2005) “Ritz Method for Plastic
Buckling Analysis of Thick Plates.” Proceedings of the Tenth International on
Civil, Structural and Environmental Engineering Computing ed. B.H.V.
Topping, Civil-Comp Press, 30 Aug - 2 Sep 2005, Rome, Italy, 1-13.
3. Wang, C.M. and Vo, K.K. (2004) “Least weight design of submerged spherical
domes.” Computational Mechanics, The Sixth World Congress on
Computational Mechanics (WCCM) in conjunction with the Second Asian-
Pacific Congress on Computational Mechanics (APCOM), 5-10 Sep 2004,
Beijing, China.
173

Optimal Designs of Submerged Domes

Uploaded by

Copyright:

Available Formats

Optimal Designs of Submerged Domes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Optimal Designs of Submerged Domes

Uploaded by

Copyright:

Available Formats

OPTIMAL DESIGNS OF SUBMERGED DOMES

NATIONAL UNIVERSITY OF SINGAPORE

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF CIVIL ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

First and foremost I wish to express my sincere gratitude to my supervisor, Professor

Wang Chien Ming (Engineering Science Programme and Department of Civil

Engineering, National University of Singapore) for his highly valuable supervision

throughout my course of study. His constant inspiration, kind encouragement,

extensive knowledge, serious research attitude and enthusiasm have extremely

Chai (Department of Civil and Environmental Engineering, University of California,

I want to express my gratitude to the National University of Singapore for

whatever requirements for my education opportunity. For this, I am thankful. Finally,

I am also grateful to my girlfriend, Ms Le Nguyen Anh Minh, and to my friends Mr.

Table of Contents .......................................................................................................... ii

List of Figures .............................................................................................................. xi

List of Tables .............................................................................................................. xiv

1.1 Submerged dome ideas ........................................................................................3

1.2 Rotational shells...................................................................................................7

1.3 Buckling of rotational shells ................................................................................8

1.4 Optimal design of domes against buckling........................................................11

1.5 Objectives and scope of study............................................................................12

1.6 Layout of thesis..................................................................................................14

CHAPTER 2. Uniform Strength Designs Of Submerged Spherical Domes .....16

2.2 Membrane theory ...............................................................................................18

2.2.1 Basic assumptions of classical thin shell theory ........................................18

2.2.2 Geometrical properties of rotational shells ...............................................19

2.3 Problem definition and basic equations .............................................................23

2.3.1 Problem definition .....................................................................................23

2.3.2 Basic equations ..........................................................................................24

2.4 Results and discussions......................................................................................29

2.4.1 Analytical solution using power series method .........................................29

2.4.2 Accuracy of analytical solution for dome thickness ..................................30

2.4.3 Critical value of subtended angle...............................................................32

2.4.4 Effect of water depth on thickness variation..............................................33

2.4.5 Minimum weight design ...........................................................................34

2.5 Concluding remarks ...........................................................................................39

CHAPTER 3. Constant Strength Designs Submerged General Domes ..............40

3.2 Problem definition and basic equations .............................................................42

3.2.1 Problem definition .....................................................................................42

3.2.2 Governing equations for membrane analysis of submerged domes ..........43

3.2.3 Boundary conditions for membrane actions in fully stressed submerged

3.3 Results and discussions......................................................................................50

3.3.1 Weightless constant strength submerged general domes...........................50

3.3.2 Constant strength of submerged general domes ........................................55

3.4 Concluding remarks ...........................................................................................66

CHAPTER 4 Energy Functionals and Ritz Method for Buckling Analysis of

4.2 Governing eigenvalue equation .......................................................................69

4.2.1 Geometrical properties of domes .............................................................69

4.2.2 Mindlin shell theory.................................................................................71

4.2.3 Strain-displacement relations...................................................................73

4.2.4 Stress-strain relations ...............................................................................74

4.2.5 Derivation of energy functionals .............................................................75

4.3 Ritz method for buckling analysis ...................................................................80

4.3.2 Ritz formulation .......................................................................................82

4.3.3 Boundary conditions ................................................................................88

4.3.4 Mathematica for solving eigenvalue problem..........................................89

4.4 Concluding remarks .........................................................................................95

CHAPTER 5 Buckling Of Domes Under Uniform Pressure ..........................91

5.1 Problem definition ...........................................................................................92