Historical Analysis of Arches and Modern Shells

by

R. Georg

B.S., University of Wisconsin - Platteville, 2012

A thesis submitted to the

Faculty of the Graduate School of the

University of Colorado in partial fulfillment

of the requirements for the degree of

Master of Science

Department of Civil, Environmental and Architectural Engineering

2015
 This thesis entitled:
Historical Analysis of Arches and Modern Shells
written by R. Georg
has been approved for the Department of Civil, Environmental and Architectural Engineering

Prof. Victor Saouma

Prof. George Hearn

Prof. Abbie Liel

Date

The final copy of this thesis has been examined by the signatories, and we find that both the
content and the form meet acceptable presentation standards of scholarly work in the above
mentioned discipline.
 iii

Georg, R. (M.S. Civil Engineering)

Historical Analysis of Arches and Modern Shells

Thesis directed by Prof. Victor Saouma

Prior 20th century, the study of masonry arches and domes consumed years of work with

some of the greatest minds in history. However, with the advancements of materials, such as steel

and concrete, and the use of computer analysis programs, the art of masonry arch and shell design

has become stagnant. With various design methods seen throughout literature, an appreciation for

the development of these methods, through basic hand calculations, must be understood to first,

ensure correct design principles are applied and second, aid in the further development of these

methods.

This thesis starts with an extensive literature review of historical analysis and design method-

ology, starting in the 16th century and continuing on through the mid-20th century and today’s

current practices. The review focuses first on masonry arch design, including principles of geo-

metric design, wedge theory, line of thrust and the ultimate load theorem. The second part views

the design and analysis of domes and vaults, concluding with a case study of St. Peter’s Dome in

Rome.

The thesis continues by reviewing the derivations of a beam and plate subjected to flexure,

prior to the thin shell derivation. In all three cases, equilibrium, compatibility and stress-strain

relationships are considered to develop the differential equation relating transverse displacement to

the load. This methodology is chosen in order to introduce the shell gradually by building upon

the initial derivations of the beam and plate.

Tying the historical design methods and derivation of the shell equation, the design and

analysis of a circular cylindrical shell will be conducted. The derived shell equation will first be

simplified to membrane theory, followed by the derivations of the governing equations for shells

through the theory of shallow shells. The analysis of the cylindrical shell will hold similarity to the
 iv

analysis of statically indeterminate beams.

Finally, the methodology of shallow shells will be incorporated into the development of a

reinforced concrete design and analysis program. The development of this program will simplify

future analyses of circular cylindrical shells and improve design efficiency. The resulting design

methodology will be recorded to aid in the future design of shells and the inspection of current

structures. The thesis concludes by offering future studies to further develop the field of masonry

arch and dome design.

 Dedication

This thesis is dedicated Professor Saouma due to his passion for the topic and continuous

guidance. May the work I have completed help fuel his future work and development of his structural

analysis book.
 vi

Acknowledgements

I would first like to thank Professor Saouma for selecting myself to perform this thesis and

work under him. He has continually guided my efforts and offered aid and insight when necessary.

Also, I would like to thank Prof. Hearn and Prof. Liel for agreeing to participate as members of

the thesis committee and so graciously evaluating the work that I have performed.

Second, I would like to thank my thesis work group members Trupti Sonavane and Kyle

Prusinski. We spent many meetings together collaborating and discussing our work. Both individ-

uals have offered much aid and support in the development and finalization of my thesis.

Finally I would like to the University of Colorado and the administration, specifically Pamela

Williams, and all her continuous help and guidance through the process of completing my master’s

thesis.
 Contents

Chapter

1 Introduction 1

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Evolution of Modern Era Shell Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Historical Review 5

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2.2 The Origin of Masonry Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2.1 Structural Theory of Masonry . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2.2 Modern Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.3 Masonry Arch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.1 Geometric Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3.2 Wedge Theory/ Collapse Analysis . . . . . . . . . . . . . . . . . . . . . . . . 16

2.3.3 Line of Thrust Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.3.4 Stress Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.3.5 Elastic Theory - Winkler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.6 Ultimate Load Theory - Heyman . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.4 Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
 viii

2.4.2 The Hen’s Egg - Viviani . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.3 Structural Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2.4.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2.4.5 Vault Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.5 St. Peter’s Basilica, Rome - Case Study . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.5.2 Three Mathematicians Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

2.5.3 Poleni’s Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

2.5.4 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

2.5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

3 Beams, Plates and Shells 63

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

3.2 Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.2.1 Curvature Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.2.2 Differential Equation of the Elastic Curve . . . . . . . . . . . . . . . . . . . . 66

3.3 Plates (Adapted from Pilkey & Wunderlich) . . . . . . . . . . . . . . . . . . . . . . . 67

3.3.1 Fundamental Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.3.2 Plate Theories: Kirchhoff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3.4 Thin Shell Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3.4.1 Definitions and Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

3.4.2 Derivation of Governing Differential Equation . . . . . . . . . . . . . . . . . . 83

3.4.3 Simplifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

3.4.4 Shallow Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

3.4.5 Membrane Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

 ix

3.4.6 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

4 Analysis of Circular Cylindrical Shells 111

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4.2 Circular Cylindrical Shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4.3 Simply Supported Shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

4.3.1 Membrane Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4.3.2 Theory of Shallow Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

4.3.3 Shallow Shells vs. Membrane Theory . . . . . . . . . . . . . . . . . . . . . . . 121

4.3.4 Bending Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

4.3.5 Edge Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

4.3.6 Prestressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

5 Circular Cylindrical Shell Design Tool 138

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

5.1.1 Preliminary Design Example . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

5.2 Matlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

5.2.1 Cylindrical Shell Roof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

5.2.2 Edge Beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6 Conclusions and Recommendations for Future Studies 149

6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

6.2 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150

 x

Bibliography 153

Appendix

A Design Code 155

A.1 Shell Design - Part One . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

A.1.1 Shell Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

A.1.2 Shell Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

A.2 Edge Beam - Part Two . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

 Tables

Table

3.1 Comparison of Governing Equations in Elasticity and Plate Bending . . . . . . . . . 78

3.2 Summary of major governing equations in thin shells . . . . . . . . . . . . . . . . . . 102

4.1 Comparison of Analysis Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

 Figures

Figure

2.1 The semicircular arch under its own weight a) minimum abutment thrust; b) Maxi-

mum abutment thrust, (Heyman, 1997) . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 Collapse of a circular arch under a point load, (Heyman, 1997) . . . . . . . . . . . . 9

2.3 Fr. Derands’s rule, (Derand, 1743) (Adapted by (Benvenuto, 1991)) . . . . . . . . . 12

2.4 Hooke’s hanging chain, (Heyman, 1997) . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.5 Fabri’s Arch Thrust Model, (Benvenuto, 1991) . . . . . . . . . . . . . . . . . . . . . 15

2.6 De la Hire’s construction for determining the size of the voussoirs of an arch, (de la

Hire, 1679) (Adapted by (Benvenuto, 1991)) . . . . . . . . . . . . . . . . . . . . . . . 16

2.7 Mechanism of the semi-circular arch, (de la Hire, 1712) (Adapted by (Heyman, 1998)) 17

2.8 The statics of the arch, (de la Hire, 1712) (Adapted by (Heyman, 1998)) . . . . . . . 18

2.9 Belidor’s Variant of de la Hire’s method, (Benvenuto, 1991) . . . . . . . . . . . . . . 20

2.10 Belidor’s Model, (Benvenuto, 1991) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.11 Couplet’s voussoir size and thrust determination, (Couplet, 1731) (Adapted by (Ben-

venuto, 1991)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.12 Couplet’s Hypothesis for the collapse mechanism of an arch, (Couplet, 1732) (Adapted

by (Benvenuto, 1991)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.13 Line of resistance and line of pressure, (Moseley, 1839) (Adapted by (Benvenuto,

1991)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2.14 Significant equilibrium conditions after Moseley and Scheffler, (Kurrer, 2012) . . . . 29
 xiii

2.15 Stress distribution over the cross section after Young, (Huerta, 2005) (Adapted by

(Kurrer, 2012)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.16 Navier’s stress distribution, (Navier, 1826) (Adapted by (Benvenuto, 1991)) . . . . . 32

2.17 Distributed stress analysis in masonry arches, (Navier, 1826) . . . . . . . . . . . . . 33

2.18 Stresses at the extreme fibers in the masonry arch cross-section according to Band-

hauer and Young/Navier, (Kurrer, 2012) . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.19 Winkler’s determination of the position of the line thrust in masonry arch using

elastic theory, (Kurrer, 2012) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

2.20 The yield surface for masonry, (Heyman, 1998) . . . . . . . . . . . . . . . . . . . . . 39

2.21 Hemispherical shell under its own weight, (Heyman, 1997) . . . . . . . . . . . . . . . 41

2.22 Meridians and parallels defining an element of the shell, (Heyman, 1997) . . . . . . . 43

2.23 Equilibrium of a small element of shell, (Heyman, 1997) . . . . . . . . . . . . . . . . 43

2.24 Hoop stress resultants necessary for the equilibrium of a hemispherical shell, (Hey-

man, 1997) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

2.25 Bossut’s ideal vault formulation, (Benvenuto, 1991) . . . . . . . . . . . . . . . . . . . 45

2.26 Infinitesimal vault, (Benvenuto, 1991) . . . . . . . . . . . . . . . . . . . . . . . . . . 48

2.27 Bossut’s figure for his studies of domes, (Benvenuto, 1991) . . . . . . . . . . . . . . . 49

2.28 Bouguer’s analysis of domes and oblique forces, (Benvenuto, 1991) . . . . . . . . . . 50

2.29 Mascheroni’s simplified analysis of finite domes, (Benvenuto, 1991) . . . . . . . . . . 51

2.30 Salimbeni’s graphical analysis of force transfer in a dome, (Salimbeni, 1787) (Adapted

by (Benvenuto, 1991)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

2.31 Cracked section and structural analysis of the dome of Saint Peter’s, (Poleni, 1748)

(Adapted by (López, 2006)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

2.32 Poleni’s analysis model, (Poleni, 1748) (Adapted by (López, 2006)) . . . . . . . . . . 56

2.33 Cross sections of the dome of Saint Peter’s, (Poleni, 1748) (Adapted by (López, 2006)) 57

2.34 Illustrations of the mechanics of masonry arches, (Poleni, 1748) (Adapted by (Hey-

man, 1998)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
 xiv

2.35 The hanging chain applied to the analysis of the dome of St Peter’s, (Poleni, 1748) . 61

2.36 The location of the six new rings placed around the dome of Saint Peter’s, (Poleni,

1748) (Adapted by (López, 2006)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.1 Curvature of a Flexural Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

3.2 Finite Element Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

3.3 Stresses in a Plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

3.4 Free Body Diagram of an Infinitesimal Plate Element . . . . . . . . . . . . . . . . . 69

3.5 Displacements in a Plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3.6 Positive Moments and Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.7 Differential Shell Element, Forces, (Billington, 1965) . . . . . . . . . . . . . . . . . . 79

3.8 Differential Shell Element, Stresses, (Billington, 1965) . . . . . . . . . . . . . . . . . 80

3.9 Differential Shell Element, Vectors of Stress Resultants, (Billington, 1965) . . . . . . 80

3.10 Differential Shell Element, Vectors of Stress Couples, (Billington, 1965) . . . . . . . 81

3.11 Differential Shell Element, Qx Stress Resultants, (Billington, 1965) . . . . . . . . . . 86

3.12 Differential Shell Element, Principal Curvatures, (Billington, 1965) . . . . . . . . . . 87

3.13 Differential Shell Element, Initial and Deformed State, (Billington, 1965) . . . . . . . 92

3.14 Differential Shell Element, Initial and Deformed State; Shear Strain Caused by In-

crease in Length of Positive Sides, (Billington, 1965) . . . . . . . . . . . . . . . . . . 94

3.15 Differential Shell Element, Initial and Deformed State; Shear Strain Caused by Twist

of Surface, (Billington, 1965) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

3.16 Differential Shell Element, Initial and Deformed State; Rotation, (Billington, 1965) . 96

3.17 Middle surface of differential element, (Billington, 1965) . . . . . . . . . . . . . . . . 98

3.18 (a) Shell of positive gaussian curvature. (b) Shell of zero gaussian curvature. (c)

Shell of negative gaussian curvature, (Billington, 1965) . . . . . . . . . . . . . . . . . 109

4.1 Circular Cylindrical Shell Structure,(Billington, 1965) . . . . . . . . . . . . . . . . . 112

4.2 Typical Cylindrical Shell Section, (Billington, 1965) . . . . . . . . . . . . . . . . . . 113

 xv

4.3 Sine Series Plot, (Billington, 1965) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

4.4 Shell with Vertical Edge Beams, (Billington, 1965) . . . . . . . . . . . . . . . . . . . 130

4.5 Shell with Horizontal Edge Beams, (Billington, 1965) . . . . . . . . . . . . . . . . . . 131

4.6 Corrective Line Loads Applied at Edge Members, (Billington, 1965) . . . . . . . . . 132

5.1 Pennsylvania Warehouse Complex, (Billington, 1965) . . . . . . . . . . . . . . . . . . 139

5.2 Pennsylvania Warehouse Roof Layout, (Billington, 1965) . . . . . . . . . . . . . . . . 139

5.3 Single Barrel used for Design, (Billington, 1965) . . . . . . . . . . . . . . . . . . . . . 140

5.4 Pennsylvania Warehouse Roof Edge Beam Dimensions, (Billington, 1965) . . . . . . 145
 Chapter 1

Introduction

1.1 Motivation

The motivation for performing this thesis lies first with the interest of architecture and second

with the necessity to understand the theory behind basic engineering principles. The beauty of

basic structures that are utilized for daily use can often be overlooked. This thesis will re-enlighten

the topic of architecture and its importance in the field of structural engineering.

In performing this particular thesis, one can mix the passion for architecture with structural

design. Arches and shells have been some of the staples of architectural design due to the beauty

of a curved structure that still exhibits great strength. Being able to study and analyze both the

masonry arch and the shell allows for great insight into how many of the great cathedrals were

constructed. These cathedrals are works of art architecturally, and yet structures that have stood

the test of time due to key design features such as arches, flying buttresses and shells.

In addition to the passion of architecture and old age structures, this thesis allows for the

study of the theory behind these structures. The arch and shell are both topics that are no longer

discussed in today’s concrete cities. Instead the focus lies in beams and columns of concrete and

steel girders and joists. The theory of the arch and shell found in almost all historic structure now

lay waste and are forgotten.

The theory of arches and shells requires no more than pen and paper and the patience to

perform hand calculations. Rather than running to the computer, the design of these structures can

be performed entirely with the use of basic equations derived years ago. Performing this analysis
 2

and re-deriving the equations of the shell allows the designer to appreciate the work of past theorists

and gain an understanding for the complex theory itself.

Ultimately this thesis will aid in the understanding of a forgotten theory and staple in ar-

chitecture. Hopefully the following will bring an enlightenment to the structural and architectural

field and remind those of the beauty and simplicity of these structures.

1.2 Evolution of Modern Era Shell Analysis

It was difficult to secure resources for the shell analytical study, especially current versions,

since the topic has been neglected for many years. However, as of recent, a few authors have

addressed the topic, Peerdeman (2008) and Krivoshapko and Hyeng (2014). Within their work a

discussion of the evolution of shells in the modern era has been discussed and prepared by Prusinski

(2015).

It was not until the early 20th century when German engineers Dischinger and Finsterwalder

teamed up with Bauersfeld, an engineer of the Carl Zeiss Company, that the first thin-walled

reinforced concrete shell structure was built. Bauersfeld wanted to build a planetarium and required

a large hemisphere for the projection of the starry sky. Dischinger developed equations that would

take into account the shearing forces and moments within a curved structure.

In 1923, Dischinger was the first to attempt the design of a shell that would cover a rectangular

floor. The attempt failed as a consequence of the difficulties of the mathematical equations, but a

successful second attempt was made the following year after implementing simplified equations. It

was in 1925 that the Zeiss planetarium was built. At the same time, Finsterwalder was working in

parallel to Dischinger to improve the design of concrete shells. Finsterwalder worked to improve

Dischinger’s equations and theories by including measurable displacements, which were later proved

through experimentation. Finsterwalder eventually published this work in his doctoral thesis, dated

1930.

Over the course of the next six years, an American engineer by the name of Schorer worked

to improve Finsterwalder’s equations, publishing his findings 1936. It was at this time that the
 3

material and physical demands of Second World War shifted the focus of study concerning modern

technological advancements to studies that would benefit war-time necessities. This caused a brief

interruption in shell development, but later resulted in the ideal conditions required for shell design

and construction. With the advances of reinforcement within concrete, an opportunity presented

itself in which shells could be redesigned to be thinner, resulting in a lower consumption of per-

manent materials compared to other structural designs of the time. Low labor cost in conjunction

with a limited supply of steel paired perfectly with the low consumption of steel and high labor

and formwork hours such shell designs called for.

With Schorer’s equations in hand, shell structures could be designed with a measurable

degree of confidence. Although Schorer’s equations were proven sufficient for design, the actual

computations called for by such equations were tedious and required an innumerable amount of man-

hours to complete. With the demand for shell structures growing, The American Society of Civil

Engineers appointed a committee to streamline the design process. The results of this committee

included numerical tables and practical formulas for various cylindrical shell designs. This data

was published in a manual titled ”Design of Cylindrical Concrete Shells Roofs”, dated 1952. For

a span of 20 years, shell construction and design bloomed all across the globe. Shell construction

and design became the flagship structural model for structural efficiency and architectural design.

It was not until the early 1970’s that the blooming era of shells ended abruptly. The low

permanent material consumption of the shell structure could no longer outweigh the rising cost of

labor and formwork associated with such designs.

Within the context of this thesis, the books of Billington (1965) and ASCE design manual

(Whitney, 1952) were used exclusively for the shell theories presented within the thesis. Even

though books such as Theory of Plates and Shells by Timoshenko, and Stresses in Shells by Flugge

were available, neither go into the length of detail seen within Billington (1965) and Whitney

(1952).

The ASCE design manual was the inspirations for majority of the work found within Billing-

ton (1965). Both sources present the topic of membrane theory in great detail. However, the
 4

theory of shallow shells is seen exclusively by Billington (1965) while the ASCE Manual no. 31

(Whitney, 1952) relies solely on the membrane theory along with an undefined correction method.

Even though this corrective method is not mentioned specifically, the results of various examples

within the manual are duplicated with the theory of shallow shells presented by Billington (1965).

Chapter two of the thesis presents a discussion of pre-modern era shell analysis focusing on

the sources of Heyman (1997), Heyman (1998), Benvenuto (1991), and Kurrer (2012). Various

other sources and papers have been cited within the thesis to aid in discussions and are presented

within the context of the bibliography. Buckling was not considered for the context of this thesis

as provided within multiple sources. Likewise, the focus on numerical methods is also neglected.

1.3 Thesis Organization

The first chapter of the thesis dives into a lengthy discussion of the historical theories of arch

and shell design, and the founders of these theories. Topics presented within this discussion include

geometry and art based design, wedge theory, line of thrust and stress analysis. Application of the

theory is found within the case study at the end of the chapter.

Chapter three begins the discussion of shells and the modern age derivation of the design

equations for a shell. The differential equations of a beam and plate in flexure are first derived,

followed by the equation for thin shell.

Chapter four focuses on the analysis of a simply supported shell circular cylindrical shell with

edge beams. The stress resultants, stress couples, and displacements are derived by membrane

theory and the theory of shallow shells. A brief discussion is included comparing the results of the

two theories.

Chapter five discusses the development of a design tool, utilizing the theory of shallow shells,

along with a brief preliminary design. A discussion of the code and results follows.

The final chapter concludes the thesis and presents a discussion of future work to be conducted

in connection with this thesis.

 Chapter 2

Historical Review

2.1 Introduction

The architectural masterpieces of the arch and dome can be seen in various examples through-

out historical landmarks. Structures like Notre-Dame, St. Peter’s Basilica in Rome, and Hagia

Sofia in Constantinople are only a few examples of such landmarks. These structures serve both

the artistic sense along with the structural integrity required for stability.

This chapter dives into a lengthy discussion of the historical theories of arch and shell design,

and the founders of these theories. This is not meant to be an exhaustive literature review. Rather,

the historical review is meant to focus on important events related to arches and shells and the

motivation of the shell theory development. As stated by Isaac Newton, ”If I have seen further

it is by standing on the shoulders of giants”. Likewise, this chapter places the emphasis on the

origination of the theory and how it progressed throughout history to our current methodologies.

The topic of the masonry design is first explored in brief detail. The structural theory of masonry

is discussed along with a brief discussion of modern design.

The chapter then shifts focus to the masonry arch beginning in the age of art with the likes of

Da Vinci, Viviani and Derand. All three visionaries of the arch viewed the structure as art. Though

all three methodologies differed, they all shared the same belief that the arch could be designed by

geometry and trial and error. Often these structures were over designed to ensure stability.

Drifting slightly from the context of geometry and art, Hooke, Gregory, and Bernoulli brought

about the idea of the catenary shape and the definition of the perfect arch. This would eventually
 6

lead to the idea of thrust that is known and studied in all arches today.

The chapter continues with a new topic, being the wedge theory, and its counterpart collapse

analysis. Designers such as La Hire, Belidor, and Coulumb spent many years developing the wedge

theorem and the distribution of forces throughout the arch. With a concept of force distribution

discovered, collapse analysis presented itself and opened the door for the ultimate load theorem.

Improving upon these methodologies, Gerstner, Moseley, and Scheffler brought about the

idea of the line of thrust within the structure. This line of thrust would be the line of force transfer

discovered by the wedge theorem. Though little improvement to the wedge theory came about

from this new analysis theory it did enlighten the next step of arch analysis.

Stress analysis was soon introduced by Young, Huerta and Navier. Rather than having a

single line of force distribution within the structure, the forces are distributed throughout its entire

area and the stresses can then be analyzed at any point. Eventually the modern methodologies of

elastic theorem and ultimate load theorem were created and used exclusively till the creation of

finite element aided computer programming.

The chapter finishes by discussing the shell. Before discussing the theorists, once again the

basic structural theory and analysis is discussed. This section concludes by discussing the work of

Bossut, Mascheroni, and Salimbeni and their contributions to the field of shell design and analysis.

The chapter concludes with the case study of St. Peter’s Basilica in Rome. In this case study,

we review the historical task of confirming the safety of a structure prior to the use of computers.

In this case, Poleni is assigned the task of determining the risk of collapse of the St. Peter’s dome.

Utilizing the historical methodology of the catenary and Hooke’s hanging chain, Poleni successfully

provides an analysis and confirmation of safety.

2.2 The Origin of Masonry Design

Masonry design often brings to mind words such as beautiful, or historical, or even strong.

The Gothic and Renaissance eras used masonry exclusively due to the lack of better materials

and ease of availability. Even with such a simple structural component, masonry structures were
 7

designed with a strong focus of art and beauty. Key components, like the flying buttresses and

towering spires of Notre-Dame, represented the essence of Gothic architecture. While the Renais-

sance showcased historical structures such as St. Peter’s Basilica in Rome and Hagia Sofia in

Constantinople, displaying the focus of both beauty and art.

Yet, the fundamental structure that undergirds such masonry design is the arch. The inven-

tion of the arch is impossible to date. Some of the earliest arches can be seen in the Mesopotamian

burial chambers, and throughout the Etruscan civilization, about 3000 years ago. The stability

and strength of these ancient structures fueled the advancement of masonry arch design well into

the 20th century.

2.2.1 Structural Theory of Masonry

Strength, stiffness, and stability are required for the design of any structure. Even though

strength and stability control the design for the majority of all structural materials, stability is the

only focus of masonry design. Strength can be ignored due to masonry’s high crushing capacity.

The compressive stress experienced by masonry is often less than 1/10 of the crushing capacity.

Also stiffness is ignored due to the low mean stresses and negligible deflection.

The majority of the focus of stability is placed on the compressive stresses formed through

compaction under gravity. This compaction accounts for the three assumptions required for sta-

bility: masonry has no tensile strength, masonry has effectively unlimited compression strength,

and sliding failure does not occur. Even though an individual masonry stone has the ability to

transfer tensile stresses, the structure as a whole has no tensile strength capacity. The compressive

stresses create friction forces between the stone wedges (voussoirs) and prevent the development of

tensile stresses and sliding failure. Even though compressive stresses are important to stability, the

location of these stresses and the resulting collapse mechanisms are the two main points of interest

in stability design.

The thrust line is the line of force (compressive stresses) or load transfer line found within the

structure. When the line of thrust remains entirely within the structure, the structure is considered
 8

to be in a state of equilibrium. ”The safe theorem states that if any one such position can be found

for the line of thrust, then this is an absolute proof that the structure is stable, and indeed that

collapse can never occur under the given loading” (Heyman, 1997, pg. 22). Thrust lines occurring

at the edge of a structure indicate the development of a hinge. When a hinge develops, the thrust

must be transferred through the point of contact between the two resulting bodies.

Figure 2.1: The semicircular arch under its own weight a) minimum abutment thrust; b) Maximum
abutment thrust, (Heyman, 1997)

Hinges are found in the form of cracks. Cracking is inevitable due to shifting of the external

environment and the necessity of the structure to respond. Cracks are often developed through

settlement, leading to changes in the geometry of the structure. Most structures are designed to

allow for some settlement and cracking. However, differential settlement is a major concern that
 9

develops over time and causes hinging and often collapse mechanisms.

Collapse mechanisms develop only when a sufficient number of hinges form. A common

collapse mechanism is the four-bar chain. This occurs when four hinges develop transforming ”the

stable arch into a mechanism of collapse” (Heyman, 1997, pg. 18) as seen in Fig. 2.2 (Heyman,

1997, pg. 19). However, hinging is not an indication of collapse until the load reaches a failure

state. In fact, a three hinge arch is statically determinate as the thrust line is fixed by the hinges.

Eventually, the load applied reaches the failure state, and the thrust line can no longer be contained

within the structure. In turn, a collapse mechanism develops resulting in failure.

Figure 2.2: Collapse of a circular arch under a point load, (Heyman, 1997)

Since collapse mechanisms are dependent on the thrust line location and the resulting hinge

formation, proper proportioning of the thickness of masonry will dictate the design and behavior.

From this proportioning a geometric factor of safety can be developed. The factor of safety considers
 10

the shape and size of the structure being analyzed. A ”load bearing structure will not collapse when

an equilibrium condition can be found that does not infringe the hinge condition” (Kurrer, 2012,

pg. 236). The value of safety is the ratio of the actual size of the structure compared to the

minimum size able to hold the thrust line without the development of a failure mechanism. In

order to contain the line of thrust within the inner third of the structure, a geometric factor of

safety of three must be implemented. Even though the computation of the true line of thrust is

impossible to know, it is ”also not important because the safety of the loadbearing structure can

be calculated without having to make assumptions regarding its actual state” (Kurrer, 2012, pg.

236).

2.2.2 Modern Design

The development of modern age masonry design consumed many years and various theories

to get to its finalized state. The earliest theories based design on the geometry and proportioning of

the masonry structure, which was eventually determined to be outdated. Galileo found that ”if the

dimensions...were doubled, the strength was very much more than doubled” (Heyman, 1997, pg.

6). This discovery led to the emergence of mechanics and stresses in the determination of design.

Modern engineering began to move from the idea of proportioning, focusing rather on internal

stress and thrust lines within the structure. Eventually the elastic theory followed for the analysis

of the arch and thrust line. However, ideal conditions required for the elastic method do not occur,

jeopardizing the solution. This brought about the creation of the plastic theory, also utilized for

masonry by finding an equilibrium state where forces remain within the boundary of the material.
 11

2.3 Masonry Arch

2.3.1 Geometric Theory

2.3.1.1 Geometry and Art - Da Vinci

Leonardo da Vinci once said, ”An arch is nothing but a strength caused by two weaknesses;

that is why an arch in buildings is composed of two quarter-circles; these quarter-circles, each very

weak in itself, wish to fall, and opposing each other’s ruin, convert weakness into a single strength”

(Benvenuto, 1991, pg. 309). Da Vinci envisioned the arch as man’s artistry trying to overcome

nature. He found interest in the arch not only as a beautiful work of art but also as a useful tool

for structural applications. With no predecessors, he hinted at ideas that would be developed some

three centuries later, such as the concepts of internal arch thrust, horizontal pier thrust, and his

self-created wedge theorem.

Leonardo da Vinci envisioned the arch as a system made up of machines, including wedges,

ropes and pulleys. He divided the arch into wedge-shaped, discrete elements matching the voussoirs.

His wedge theorem became the first idealized approach to arch equilibrium and static analysis via

simple machines. However, he ”did not contradict the common conviction that geometry, not

statics, could provide the simplest, most harmonious, and safest proportions for making arches”

(Benvenuto, 1991, pg. 311).

2.3.1.2 Geometry and Art - Viviani

Da Vinci was not alone in the belief of geometry controlled design. In 1692, Vincenzo Viviani

published a transcript on the formation and size of any regular arch. The work was intended to

”teach the ’expert turner geometricians’ how to use ’chisels, drills, and gimlets’ to make any sort of

vault or to bore out certain solids ’with highly usable rules’” (Benvenuto, 1991, pg. 311). Viviani’s

main focus dealt with the Florentine rib-and-panel vault and the Roman haul shaped vault. Though

it strayed from the masonry arch, the sole basis of geometry and tracing design continued to be

exhibited. Viviani attributes the shape of the domed vault to the observation of an eggshell placed
 12

upside down and its immense strength. Once again the focus is not on the material properties but

rather the shape of the structure.

Figure 2.3: Fr. Derands’s rule, (Derand, 1743) (Adapted by (Benvenuto, 1991))

2.3.1.3 Geometry and Art - Derand

Jesuit François Derand was one of the last visionaries of the 17th century to base arch design

solely off of geometric tracing. In 1643, he transcribed his work entitled L’Architecture des voûtes,

ou l’art des traits et coupes des voûtes which was not published until Derand (1743). The book

detailed answers for various problems requiring expert geometric tracing, nearly neglecting all

structural dimensions. In fact, only one chapter of the book dealt with thrusts and dimensioning.

Within this chapter he developed Derand’s Rule for determining abutment thickness, becoming the

first theory developed not utilizing statics. Instead, he relied extensively on geometry and tracing.

The method can be seen in Fig. 2.3. The circular arch P can be broken into thirds ABCD.
 13

Between points C and D a line is drawn and extended beyond D. A compass is then opened to the

distance of chord CD with D at the center. A vertical line may be drawn where the new traced arc

strikes the extended line CDF. This represents the outside wall of the abutment with the interior

being placed at D. The accuracy of the method was based on experience and practice and regarded

as acceptable. This eventually changed with the advancement of statics when the method was

disproved.

2.3.1.4 The Catenary - Hooke

Original arch design determined by Leonardo Da Vinci, Vincenzo Viviani, and Jesuit François

Derand relied heavily on geometry and tracing (Fig. 2.3). This continued into the early 17th

centuries through the work of Robert Hooke.

Hooke worked as the Curator of Experiments to the Royal Society. During this time he

experimented on model arches without the use of mathematical analysis. In 1675, Robert Hooke

published a book aiding the theory of geometry by determining the perfect shape of an arch to be

the shape of a catenary or hanging chain. ”As hangs the flexible line, so but inverted will stand

the rigid arch” (Kurrer, 2012, pg. 214).

2.3.1.5 The Catenary - Gregory and Bernoulli

Nearly twenty years later, in 1697, David Gregory expanded upon Hooke’s theory. He did

not limit his thinking to only arches of catenary shape but to other figures as well by adding the

concept of catenary shape containment within the arch. ”When an arch of any other figure is

supported, it is because in its thickness some catenaria is included...if any thrust line can be found

lying within the masonry, the arch will stand” (Heyman, 1998, pg. 80). Jakob Bernoulli furthered

the hypothesis by calculating the curvature of the catenary.

Gregory also defined abutment thrust stating that ”the same force that a chain exerts inward,

an arch of equal form exerts outward” (Benvenuto, 1991, pg. 327). However, by the end of the late

1600’s, geometry lost ground to the scientific and mathematical theory of statics.
 14

Figure 2.4: Hooke’s hanging chain, (Heyman, 1997)

2.3.1.6 Introduction of Thrust - Fabri

Honoré Fabri began the transition from geometric analysis to thrust line. Fabri analyzed a

semicircular arch as a 3-pin system loaded at C by assuming the thrust lines to be straight lines

from the springing to the crown (Fig. 2.5). If the thrust line lied completely within the arch, the

arch remained stable. Fabri used the model to geometrically determine the arch and abutment

thickness by circumscribing a semicircular arch at the tangent of the thrust line (assumed thrust

line at 45 deg).

Dead = 2R = 2(CB)
(2.1)
Horizontal Thrust = R = BD (1/2 Dead Load)
 15

Figure 2.5: Fabri’s Arch Thrust Model, (Benvenuto, 1991)

The forces calculated are found to be about 5 percent higher than computed via the elastic the-

ory. Fabri’s geometric analysis was a conservative approach to calculating the arch and abutment

thickness.

√
d1 = N D = 2R(3 − 2 2) = 0.343R (Thickness) (2.2)
√
r = R − d1 = BN = R(4 2 − 5) = 0.657R (Radius) (2.3)

Though Fabri still utilized geometric analysis, he was one of the first theorists to introduce

the concept of thrust line into arch and abutment design.

 16

2.3.2 Wedge Theory/ Collapse Analysis

2.3.2.1 Static Analysis - de La Hire

Philippe de la Hire’s work is recognized as the first globally accepted scientific and static

analysis of an arch. In 1695, as one of the leading members of the Académie Royale des Sciences

de Paris (Benvenuto, 1991, pg. 321), he presented his Traité de méchanique. His arch theory

developed from the need of a scientific verification for masonry arch design. In his text, de la

Hire examined a semicircular arch assembled from voussoirs with predetermined weights. The

voussoirs were assumed perfectly smooth with no friction implying the line of thrust must act

perpendicular to the joints. The keystone was assumed a wedge resting on neighboring wedges

implying that the weight of the keystone was supported by forces perpendicular to the joints. De la

Hire determined that the forces within each individual voussoir were proportional to the sides of a

triangle perpendicular to the force. For example, we can look at the voussoir next to the keystone

in Fig. 2.6.

Q2 : F 1 : F r = LO : LC : CO (2.4)

Figure 2.6: De la Hire’s construction for determining the size of the voussoirs of an arch, (de la
Hire, 1679) (Adapted by (Benvenuto, 1991))

This method can be performed for all voussoirs except the last one located at the springing.

According to the theory, the weight at the springing would be infinite since the resulting weight
 17

and thrust forces are parallel along with the corresponding sides of the force triangle. The issue

only arises due to the assumption of infinitely smooth voussoirs.

Figure 2.7: Mechanism of the semi-circular arch, (de la Hire, 1712) (Adapted by (Heyman, 1998))

On February 27, 1712, Philippe de la Hire submitted a memoir, entitled Sur la construction

des voûtes dans les edifices, to the Académie Royale (de la Hire, 1712). In his memoir, he revisited

the arch and determined the voussoirs were no longer frictionless but rather friction was assumed

so large, sliding failure could not occur. This meant the thrust lines were no longer fixed as

perpendicular at the joints.

He also introduced the wedge and lever to the static behavior of the arch. The new parameters

were used to determine arch thrust for abutment design, utilizing mechanics and graphical analysis.

De la Hire developed an ultimate load theory proposing three ideas regarding the analysis of a

round arch. First, ”an arch breaks in an intermediate section between the impost and the keystone,

at about 45 degrees”. The newly formed wedge drops, due to its weight, pushing the abutments
 18

Figure 2.8: The statics of the arch, (de la Hire, 1712) (Adapted by (Heyman, 1998))

and causing rotation. Second, ”the three zones bounded by the disconnections of the voussoirs are

so stable that they form a single body”. Third, the thrust at the hinge was determined to act

tangentially to the intrados. Knowing the weight of the upper portion of the arch, the resulting

thrust could be determined.

The component of the thrust, D, acted perpendicular to the lever arm, LH 2.7, which was

then traced to the base of the abutment determining the abutment width. The weight of the

abutment and base were represented on the other end of the lever arm by Q acting at T. Rather

than focusing on geometry, the focus is placed on the weight of these stabilizing parts. Through

the initial graphical analysis, de la Hire developed a mathematical proof and equation to determine

the necessary abutment width. From Fig. 2.7 the following relationship were made.
 19

LG : CG = Pc : D (2.5)

If we set LE = f (half chord), CE = e, SA = a, CF = r, EF = s, LA = g, TO = h, OS = t, the

abutment height HB = b, and the width HS = y, the following equations can be derived

f
CG = e − (y − a) (2.6)
g
s
f2
LG = f 2 + 2 (y + a)2 (2.7)
g
eg − f y − f a
D = Pc p (2.8)
f g 2 + (y + a)2

Since the design assumed constant thickness of the abutment, the weight can be considered

proportional to the area. Therefore, the weight of the abutment equaled b*y. De la Hire accounted
0
for the weight, P , of segment ILM with an imaginary increase of the abutment height BV. The

total weight, P, to be applied at a distance of HT = y/2 from H is given by

( 21 y + h)P 0
P = by + 1 (2.9)
2y

The law of levers can then be applied

D ∗ HL = P ∗ HT (2.10)
p
HL = (y + a) + g 2 (2.11)

Finally the abutment width is computed by the following relationship

1 1
bf y 2 + P 0 f y + P 0 f h = Pc eg − Pc f y − Pc f a (2.12)
2 2
 20

Figure 2.9: Belidor’s Variant of de la Hire’s method, (Benvenuto, 1991)

2.3.2.2 Wedge Theory - Belidor

In 1729, Bernard Forest de Belidor took a similar approach to de la Hire except changed the
√
critical hinging point to 45 degrees, computing a thrust of 2W (W is the weight of the voussoir

above the hinge). He also changed the location of the thrust to act perpendicular at the center of

the joint rather than the intrados (Fig. 2.9). The new, easier determined thrust value compounded

upon the theory to develop an equation for abutment width.

From Fig. 2.10, we can derive an equation for determining the abutment width, y, where l

is the height, b is the distance between the intrados of the abutment and the vertical line passing

through center of gravity of MmKk, and c is the distance between the intrados and the vertical line

from L. We can simplify Fig. 2.10 by setting LY = y + c, HY = l + OC, and HN = HY - NY =

HY - LY = l + OC - (y+c). The figure can be further simplified by setting f = OC + l - c and HN

 21

Figure 2.10: Belidor’s Model, (Benvenuto, 1991)

= f - y, designating the lever arm

√
2(f − y)
HΩ = (2.13)
2

Using these assumption, Belidor derived the equilibrium equation

y2
Pc (f − y) = l + Pc (y − b) (2.14)
2

and the final equation for the required abutment width

s
2Pc 4Pc 2 + 2Pc l(f + b)
y=− + (2.15)
l l2

Similar to de La Hire, Belidors wedge theory neglected friction. However, Belidor came to

the conclusion that mortar or a bond would be required. ”The theory of La Hire and Belidor
 22

was accepted throughout continental Europe almost without question during the rest of the 18th

century” (Kurrer, 2012, pg. 208).

2.3.2.3 Collapse Analysis - Couplet

The work of Claude Antoine Couplet (Couplet, 1731) and (Couplet, 1732) laid the basis of

future arch analysis. His two memoirs to the Académie Royal des Science, De la poussée des voûtes

and Seconde partie de l’examen de la poussé des voûtes expanded and improved upon the theories

of de la Hire and Belidor. Couplet determined the three key postulates about masonry behavior.

First, masonry has no tensile strength. Second, masonry has infinite compressive strength. Third,

sliding failure cannot occur (Heyman, 1998, pg. 83).

Utilizing equilibrium (statics) and deformation (mechanics), Couplet became the first to

develop techniques for determining the size and thrust of voussoirs. With a known keystone weight,

the said keystone can be decomposed into equivalent forces perpendicular to the joints of the

adjoining voussoirs. The equivalent forces were extended to the center of gravity of the neighboring

voussoir. A vertical line could then be drawn from the equivalent force to the perpendicular force

of the voussoir. The vertical component determined the weight of the wedge and the perpendicular

component determined the thrust (Fig. 2.11).

Couplet’s final major achievement was the development of a ratio, dependent on the arch

radius, to determine the least thickness of a semicircular arch carrying only self weight. Utilizing

collapse mechanism, in particular the four-bar chain, he broke the arch into four segments at 45

degrees. The points of rotation were regarded as hinges through which the thrust must pass. The

failure mode was analyzed to determine a t/R=0.101 (Fig. 2.12).

tmin,c = 0.101 × R (2.16)

All masonry arches less than the minimum thinkness were considered unstable. The theory

was later refined by Jacques Heyman realizing the critical hinging did not occur at the intrados at
 23

Figure 2.11: Couplet’s voussoir size and thrust determination, (Couplet, 1731) (Adapted by (Ben-
venuto, 1991))

45 degrees but rather 58.9 degrees. The value was refined to

tmin,h = 0.106 × R (2.17)

Figure 2.12: Couplet’s Hypothesis for the collapse mechanism of an arch, (Couplet, 1732) (Adapted
by (Benvenuto, 1991))
 24

2.3.2.4 Voussoir Rotation Theorem - Coulomb

In 1773, Charles Auguste Coulomb published his work (Coulomb, 1773) determining the

dimensions of vaults and arches, the rupture locations, and the limit state of rest with cohesion and

friction. The first three corollaries contributed to vault dimensioning. His first corollary focused

on an infinitely thin vault as seen in Fig. 2.13(c). He defined P and Q as the loads located at M.

Given the relationship of similar triangles, it can be determined that

dx P
= (2.18)
dy Q

Assuming the horizontal thrust is constant and equal to P at the crown location, a, and that

the resultant vertical forces equal the weight of the vault, the relationship derived above can be

modified to
dx P
=R (2.19)
dy qds

with the curve of the vault calculated by the left hand side of the equation and the weight being

attained by the other. From one known element, the other can be easily attained.

In the second corollary, Coulomb determined the vault width at any point with only the self-

weight load condition. Through rigorous derivations, Coulomb developed the following equation
r
ds 2
r+h= r2 + 2P ( ) (2.20)
dx

where r equals the radius of curvature of the intrados of the vault and h is the width at the

location of interest. However, Coulomb found two problems with his theory. First, the resultants

of the forces must always be perpendicular to the joints for the method to be valid. Second, the

formulation is not completely reliable due to the infinite growth of h as θ approaches 0.

Coulomb’s third corollary looked to determine the direction of the joints between the vous-

soirs, given the known intrados and extrados. The corollary determines the joints for the vaults

and arch case (Fig. 2.13(a)), along with the platband condition seen in Fig. 2.13(b). Through

derivations that we will overlook, the location and direction of the joint is determined for each case.
 25

(a) Circular arch joint direction determination, (b) Platband joint direction determination,
(Coulomb, 1773) (Adapted by (Benvenuto, 1991)) (Coulomb, 1773) (Adapted by (Benvenuto, 1991))

Coulomb’s most memorable contribution to masonry arch design comes in his final corol-

lary involving cohesion and friction. By calculating the rupture location and limit states of the

arch, Coulomb determined the maximum value of arch thrust through extreme value calculation

of differential calculus. He determined the arch thrust for four cases (Fig. 2.13(c)): keystone

sliding downward (1), keystone sliding upward (2), rotation about M (3), and rotation about m

(4). Coulomb used statically determinate collapse mechanisms rather than the four-pin collapse

mechanisms of Couplet. In cases 1 and 2, Coulomb first developed his laws of friction. He stated

that the friction force must be applied in the direction opposite of the tendency to slide. Applying

the friction coefficient to the equation of equilibrium, the horizontal thrust at the crown, P, was

determined. The maximum and minimum value of P was calculated to prevent collapse of the arch

by sliding.

Coulomb’s voussoir rotation theory looked mainly upon the cases of 3 and 4, rotation about

M and m (Fig. 2.13(d)). He determined a half arch was maintained in equilibrium by a horizontal

thrust acting at the capstone. Coulomb’s analysis introduced the idea of maximum and minimum

allowable thrust to keep the arch in equilibrium, dependent upon the location of the hinge. By
 26

(c) Four limit cases for considering friction and co- (d) Coulomb’s voussoir rotation principle, (Coulomb,
hesion, (Coulomb, 1773) (Adapted by (Benvenuto, 1773) (Adapted by (Benvenuto, 1991))
1991))

assuming the hinge location about the intrados, M, the equivalent thrust was determined. The

location of the hinge was moved until the maximum thrust was found. This was the minimum

value required to maintain stability of the arch. The process was then repeated by assuming

hinging about the extrados, m. Once again the hinge was relocated until a minimum value was

found, or the maximum value for arch stability. The maximum and minimum values of all four

cases control the design.

2.3.2.5 Voussoir Rotation Theorem Advancement

Though Coulomb was not correct in determining the critical hinging location in the arch,

his theory was a stepping stone for future research. In 1820, Audoy improved upon Coulomb’s
 27

method and determined the location of the ”fracture joints at the intrados which correspond to

the maximum horizontal thrust at the key in the case of sliding or rotating” (Benvenuto, 1991, pg.

429). Barlow later solidified the theory of maximum and minimum thrust in 1846 by creating a 6

voussoir arch model with wood acting as mortar. Removing select pieces of wood, he developed

hinging and thrust lines, determining the line of resistance and line of impression (limits of least

and greatest horizontal abutment thrust) (Heyman, 1998, pg. 92).

2.3.3 Line of Thrust Theory

The line of thrust theory was introduced through recognizing the need to achieve ”load

bearing system synthesis” (Kurrer, 2012, pg. 216) in order to analyze the arch as a whole. The

chain models introduced by Hooke and Gregory analyzed a whole system, however, lacked the

ability to analyze individual elements. The load bearing system analysis of de la Hire and Couplet

was limited to individual load bearing elements and their addition to form the model arch. Neither

the chain models nor the addition of individual wedges could properly analyze the arch and saw no

progress until Gerstner in the early 1800s.

2.3.3.1 Introduction of Line of Thrust Theory - Gerstner

Franz Joseph Ritter von Gerstner brought about the ”merger of the loadbearing system syn-

thesis with loadbearing system analysis” (Kurrer, 2012, pg. 216). In 1831, Gerstner accomplished

this through the introduction of the line of thrust theory and formulation of the three prime tasks.

First task: ”Determine the loading case for given loading case”. Second task: ”Determine loading

case for a given arch center of gravity axis such that said axis coincides with the line of thrust”.

Third task: ”Take into account the line of thrust for given loading case and masonry arch center of

gravity axis” (Kurrer, 2012, pg. 217). The line of thrust theory favored the third task through the

recognition that a masonry arch’s stability is dependent on the number of statically possible lines

of thrust. Eventually the focus changed from the infinite number of statically possible thrust lines

to the true line of thrust.

 28

2.3.3.2 True Line of Thrust - Moseley

The true line of thrust was considered as the true acting thrust found within the arch for a

given loading. The topic of true thrust presented great debate in the 1800’s. The first exploration

of the topic came from Henry Moseley in 1833. Moseley proposed the principle of the line of

resistance and line of pressure, stating that the line of minimum resistance would identify the true

line of thrust. The line of resistance was created by connecting the intersections of the thrust

resultants with the joints. The line of pressure was created from the direction of the resultant

pressures in the joints (Fig. 2.13).

Figure 2.13: Line of resistance and line of pressure, (Moseley, 1839) (Adapted by (Benvenuto,
1991))

”For equilibrium to exist, the line of resistance must be entirely inside the interior of the

arch; if it crosses the extrados or intrados below a certain angle, the arch will break near its point
 29

Figure 2.14: Significant equilibrium conditions after Moseley and Scheffler, (Kurrer, 2012)

of intersection with the line” (Benvenuto, 1991, pg. 432). Moseley determined the true line of

thrust must pass through the extrados at the key and the intrados at the springing, since this was

the minimum arch thrust. However, he did not examine of the case of maximum thrust. Rotation

occurs at the locations where the line of resistance is tangent to the intrados or extrados.

Moseley also included the concept of friction in his analysis. He stated that the angle, ρ, at
 30

which the line of pressure crosses the joint must be connected with the friction in the joint.

φ = tan−1 f friction angle

ρ > φ sliding will occur (2.21)

ρ < φ no sliding

2.3.3.3 True Line of Thrust - Scheffler

An infinite number of lines of pressure can be drawn within the arch, but which is the correct,

true line of thrust? This was main topic of opposition to the viewpoint of Moseley. Scheffler quickly

counters by stating that it’s unclear to determine which of the infinite force systems occur in nature.

”According to Scheffler, the minimal line of thrust is the true line of thrust only for masonry arches

with a rigid mass of voussoirs. However, as the material of the voussoirs is not rigid, but rather

elastic, the true line of thrust lies between the minimum and maximum lines” confirming the idea of

limit analysis presented by Coulomb (Kurrer, 2012, pg. 219). Scheffler led the initial push toward

the idea of elastic material properties and elastic analysis.

2.3.4 Stress Analysis

2.3.4.1 Uniform Stress Distribution - Young and Huerta

Thomas Young worked on his masonry arch theory from 1801 to 1816. His theory consisted

of six parts, however, we will only focus on the first part: the resistance of materials. In this portion

of Young’s masonry arch theory, he developed a law of distribution of stress over the arch cross

section when an axial force is applied eccentrically to the neutral axis position. The neutral axis

position is described by Young as ”the distance of the neutral point from the axis is to the depth,

as the depth to twelve times the distance of the force, measured in the transverse section” (Huerta,

2005, pg. 202) and can be seen below.

d2
z= (2.22)
12y
 31

Figure 2.15: Stress distribution over the cross section after Young, (Huerta, 2005) (Adapted by
(Kurrer, 2012))

Huerta advances Young’s theory with the following mathematical explanation. The axial stress, σ,

at the extreme right (Fig. 2.15) can be written as

 
d + 6y
σ = σm (2.23)
d

where the average stress,

N
σm = (2.24)
db

When the applied force neared the edge of the cross-section, the force distribution exhibited

compressive forces near the applied force and tension away from the force. Indirectly, Young made

the determination of the middle third rule. ”If the axial force N is applied in the middle-third of

the depth d of the arch, then all the stresses are compressive” (Kurrer, 2012, pg. 88). Ensuring

compression throughout the arch instills confidence in the overall stability and safety of the arch.

As stated previously, a ”load bearing structure will not collapse when an equilibrium condition can

be found that does not infringe the hinge condition” (Kurrer, 2012, pg. 236). If no tension is found
 32

within the structure, no hinges will form and the structure will remain stable.

2.3.4.2 Middle Third - Navier

Claude - Louis Navier introduced stress analysis to masonry arch theory, which laid the

foundation to what became the basis of strength analysis. Navier’s masonry arch theory observed

the voussoir rotation theory of Coulomb. In addition he ”permits horizontal loads, assuming a

triangular distribution for the compressive stress in the joints under consideration” (Kurrer, 2012,

pg. 220). He considered the stress distributed over every point in the transverse joints which allows

for a more realistic distribution of the compressive forces which the material resists (Fig. 2.16).

Figure 2.16: Navier’s stress distribution, (Navier, 1826) (Adapted by (Benvenuto, 1991))

Even though Navier introduced stress analysis into masonry arch theory and elastic theory,

he could not relate the elastic theory developed for timber and iron construction to masonry arches.

This was due more to the time frame and lack of experimentation and material knowledge than
 33

a lapse on Navier’s part. The ”force-deformation behavior of masonry arch materials in the ser-

vice condition was not yet researched experimentally until the final third of the 19th century, in

particular by Johann Bauschinger”. Also ”the small compressions under service conditions could

not be quantified reliably with the testing apparatus available at that time”. Finally, ”deformation

measurements on the generally oversized masonry arch structures could not be meaningful because

the effects of arch settlement, dimensional stability, etc. were in the same order of magnitude as

the deformations under service conditions” (Kurrer, 2012, pg. 220).

Figure 2.17: Distributed stress analysis in masonry arches, (Navier, 1826)

However, Navier did come to the conclusion that ”the resultant of normal pressure at the

joint must pass at a distance from the most compressed edge equal to a third of the actual width

of that joint” and that ”the pressure on this edge is twice as much as the one that would occur

in the hypothesis of a uniform repartition on the whole surface of the joint” (Benvenuto, 1991, pg.

431). Navier, along with Young discussed previously, noticed the importance of thrust occurring

within the middle third of the member to maintain uniform compressive stress across the member

face (Fig. 2.17).

 34

2.3.4.3 Advancement of Stress Analysis - Bandhauer

Without the creation of elastic theory, alternatives for analyzing the stress distributions in the

joints of masonry arches continued development. In 1831, Bandhauer introduced a new approach of

stress distribution at voussoir joints and the idea of a factor of safety. Bandhauer determined that

the thrust line of a catenary arch with a factor of safety of one must follow the given calculation

or failure may occur. However, as the factor of safety grows average compression line can deviate

from the given calculation without the fear of failure. ”It is this and only this condition that we

have to thank for the stability of all our free-standing masonry arches designed according to the

catenary” (Kurrer, 2012, pg. 221).

Bandhauer’s stress concept developed from this new approach resulted in the compressive

stress equation seen below

N 1
σB (v) = × (2.25)
bd 1 − v

where d is the thickness, b is the width, N is the normal force applied at the eccentricity, e

and
2e
v= (2.26)
d

Converting Eq. 2.25 to the dimensionless form allowed the creation of Bandhauer’s hyperbolic

function of compressive stress distribution.

σB (v) 1
= (2.27)
σm 1−v

For comparison, Young’s equation (Eq. 2.23) is also converted to the dimensionless form

σ(v)
= 3v + 1 (2.28)
σm

In Fig. 2.18 seen below, Bandhauer’s equation differs from the compressive stress distribution

developed by Young and Navier for N applied in the middle third of the cross section.

The equations of Young, Navier and Bandhauer all equate equal values at v=0 and v=2/3.

The difference is found as the equations approach v=1. As Bandhauer’s hyperbolic function ap-
 35

Figure 2.18: Stresses at the extreme fibers in the masonry arch cross-section according to Bandhauer
and Young/Navier, (Kurrer, 2012)

proaches 1, it goes to infinity in accordance with the conditions set by Couplet and Heyman of

infinite compressive strength. In comparison, the equation of Young/ Navier continues linearly as it

approaches 1. Bandhauer’s hyperbolic functions mimics today’s equation of load carrying capacity.

NT 1 NT
σB (v = 2/3) = BR = × 2 =3 (2.29)
bd 1− 3
bd

or
1
NT = bdBR (2.30)
3

where BR equals the standardized compressive strength of masonry.

2.3.5 Elastic Theory - Winkler

In 1860, Saavedra was the first individual to try elastic theory with masonry arches. However,

his approach was far too complex and not practical for the design of masonry arch structures. With
 36

minimal results, his theory was placed on the back burner to make room for the theorems of Rankine

and Winkler. Rankine ”postulated the theorem that the stability of a masonry arch is guaranteed

when a line of thrust due to a given loading case can be drawn through the middle-third of all arch

cross sections” (Kurrer, 2012, pg. 223). Rankine combined the middle third rule with thrust line

theory which was used in analysis until the 1960’s.

From 1867 to 1868, Wrinkler developed the influence line concept for 3-pin, 2-pin, and fixed

arches. However, his more notable achievement comes about in 1879 with his presentation on the

concept of elastic arch theory. Winkler adapted his elastic arch theory to the analysis of masonry

arches. He first defined the line of thrust as ”the geometrical position of the point at which the

resultants intersect the masonry joints” (Kurrer, 2012, pg. 225). He next determined the position

of the line of thrust. In the final step, he differentiated the normal state from the disrupted state.

Winkler also identified disruptions such as ”incompletely cured mortar, temperature changes,

yielding centering during construction and, first and foremost, sinking abutments after striking the

centering, which lead to visible cracks and considerable changes to the course of the line of thrust”

(Kurrer, 2012, pg. 225). He proposed that the correct line of thrust can be determined through

elastic analysis when the arch is subjected to normal state conditions.

In Winkler’s elastic theorem he derived three elasticity conditions to determine the position

of the line of thrust in a masonry arch with three degrees of static indeterminacy. No rotation

(∆φ = 0) and no horizontal (∆u = 0) or vertical (∆v = 0) movement.

1
R R
∆φ = 0 = dγ = EIz × M (s)ds
1 (2.31)
R R
∆u = 0 = ∆dsx = EIz × M (s)yds
1
R R
∆v = 0 = ∆dsy = EIz × M (s)xds

where M is the bending moment with the arc coordinate s, E is the elastic modulus, and Iz

is the second moment of area about the z-axis.

The Winkler Theorem states that ”for a constant thickness, the line of thrust close to the

[correct] one [is] the one for which the sum of the squares of the deviations from the center-of-gravity

axis is a minimum” (Winkler, 1880, pg. 128).

 37

Figure 2.19: Winkler’s determination of the position of the line thrust in masonry arch using elastic
theory, (Kurrer, 2012)

Z
I= [z(s)]2 × ds = Minimum (2.32)

Eq. 2.32 must satisfy the following conditions

 38

dI
= 0 (2.33)
da
dI
= 0 (2.34)
db
dI
= 0 (2.35)
dc

where a, b, and c are unknown position parameters. The vertical support reaction, X1 , the

horizontal thrust, X2 and the moment, X3 from Fig. 2.19 can be used for the position parameters.

2.3.6 Ultimate Load Theory - Heyman

Even though Drucker is credited as first suggesting the use of ultimate load analysis for

investigating the equilibrium and failure of voussoir arches, the memorable contribution to this

topic comes from Jacques Heyman. In 1966, Heyman published a discussion of the ultimate load

theory for all masonry load bearing structures, including plane arches, domes, fan vaults, groined

vaults, towers, and spires (Kurrer, 2012, pg. 233). Heyman required three principles be met in order

to conduct the ultimate load analysis of masonry construction. All three principles aligned with

those of Couplet, infinite compressive strength, zero tensile strength, and no sliding occurs. ”When

the masonry material satisfies these conditions, the component of the resultant of the effective

stresses acting perpendicular to the cross-sectional area must be a compressive force N for each

cross-section whose intersection point lies within the cross-section” (Kurrer, 2012, pg. 233). If

N acts at the edge of the cross-section, a hinge forms leading to a yield surface bounded by two

straight lines.

M = Ne (2.36)

−h< e <h (2.37)

The normal force must act within the yield surface of AOB. If N passes outside the yield

surface, the thrust passes outside the masonry arch section and the structure is deemed unstable.
 39

Figure 2.20: The yield surface for masonry, (Heyman, 1998)

In today’s design principles for masonry arches, the ultimate load theorem of Heyman, first hinted

by Coulomb through voussoir rotation, is most often utilized to analyze stability.

If we replace the stone of infinite strength (Fig. 2.20) with a stone of finite crushing strength,

the yield zone is reduced to the curved boundary OCDEO. Even though the zone narrows to point

D, this has little effect on the majority of all masonry design. ”A typical value of permitted stress

used in the nineteenth-century design...is 10 [percent] of the crushing strength; nominal stresses are

likely to be less than this, but even at 10 [percent] the portion of the yield surface is the slightly

curvilinear triangle OCE” (Heyman, 1998, pg. 94).

To aid his ultimate load analysis, Heyman proposed a geometric factor of safety. ”If the real

arch has twice the thickness of the limiting arch, then the geometrical factor of safety is 2” (Kurrer,

2012, pg. 237). The limiting arch must be determined from the most unfavorable loading case.

Elaborate calculations, often including virtual displacements, were required for the exact value.

However, from the concept of simply passing a thrust line through the arch, it was easy to show.

Heyman also proved that the ultimate load analysis enabled the equilibrium approach for the

analysis of loadbearing structures. This tool was made useful for determining the structural stability

of the masonry. ”It is not the task of the structural engineer to determine the true equilibrium

condition for a particular loadbearing structure, but rather sensible equilibrium conditions” (Kurrer,

2012, pg. 237).

 40

2.4 Shells

2.4.1 Introduction

”A dome is a rounded vault forming a roof over a large interior space” (Heyman, 1997, pg.

27). This characteristic alone brought interest to dome and vault research as theorists searched for

new and aesthetically pleasing forms of shelter. The majority of the properties of the masonry arch

hold true for the dome, resulting in far more research in the field of arches. However, some unique

properties are seen only in the dome and require special attention.

2.4.2 The Hen’s Egg - Viviani

The basic shape of a dome is a shell of revolution where every horizontal section is circular,

similar to that of the hen’s egg. Vincenzo Viviani spent immense time studying the egg’s shape

and its extraordinary strength. Even with its thin and fragile shell, the shell cannot be broken with

pressure applied between the thumb and finger when held in the longitudinal axis. Instead a high

local pressure is required to crack the shell. Given the high strength with respect to the thin shell,

the egg formed the basis of experimentation and development of the dome.

2.4.3 Structural Theory

The shell is considered as a curved surface or plate with a small thickness compared to the

size of the shelled structure. The interior forces of the shell resist self weight and external loads

and are able to carry a wide range of loadings. The major concern of the shell’s carrying capacity

lies not with the thickness, as in an arch, but solely with local compressive buckling. However, as

discussed previously with arches, masonry has an extremely high compressive strength, ensuring

no local buckling danger. Similar the masonry arch, the masonry dome exhibits low compressive

stresses. Fig. 2.21 displays a uniform, thin walled hemisphere solely supporting self weight. The

shell is supported at its base by a uniform compressive stress, σ. The radius, a, thickness, t, and

unit weight, ρ, will help aid the illustration.

 41

Figure 2.21: Hemispherical shell under its own weight, (Heyman, 1997)

Volume = 2πa2 t (2.38)

Diametral ring area = 2πat (2.39)

σ(2πat) = ρ(2πa2 t) (2.40)

σ = ρa (2.41)

The proof shows that the compressive stress is unreliant on the dome thickness but rather

the radius.

Another unique characteristic comparing the structural theory of the arch to the dome deals

with the keystone. The masonry arch requires the keystone to stand where as the crown of the

dome is not required. This is due to the multi-directional distribution of stress. Since the hoop

stresses are always in compression, a full circle around the dome is enough to support itself. The

segmental dome allowed for a oculus (skylight) along with large bell towers and spires to extend

from the dome at the opening. Famous examples of segmental, incomplete hemisphere, domes are
 42

the Santa Maria del Fiore and the Pantheon in Rome.

2.4.4 Analysis

Initial methods of dome analysis followed that of the masonry arch, utilizing statics or Hooke’s

hanging chain. Hooke’s hanging chain represented the thrust line within the dome. This static

approach allowed the dome to carry a wide range of loadings. This methodology was performed

by Poleni in the study of the cracking of the St. Peter’s dome in Rome and will be discussed in

a later section. Similarly to the arch, no matter the shape, an outward thrust was observed at

the supports. However, domes often require buttresses due to the lack of lateral strength from the

supporting walls.

The second and more recognized method of dome analysis is the membrane theory. The

membrane theory assumes ”that the surface has no stiffness against bending, so that the forces

in the shell are purely tensile or compressive” (Heyman, 1997, pg. 28). However, since masonry

has no tensile properties, only compression is considered. The theory analyzes the transmission

of stresses through the shell to the base. Using differential equations and an infinitesimally small

element cut from the shell, the stress resultants within the dome can be found (see Fig. 2.22 and

2.23).

The stress resultants Nφ and Nθ act on the cut edges of the element. Nφ represents the

stresses along the meridian which increase from the crown to the base. Nθ accounts for the parallel

circles, or hoop stresses. The meridian stresses are compressive throughout the shell which are the

same stresses accounted for by the hanging chain theory. However, Mascheroni, discussed in a later

section, discovered that the hoop stresses switch from compression to tension between the crown

and the base. From the crown (90 degrees) to 51.82 degrees the stresses are in compression and

then switch to tension increasing toward the base (Fig. 2.24). Steel rings or buttresses are required

to prevent the tensile forces from controlling at the base of the dome and limits the outward thrust

at the supports.
 43

Figure 2.22: Meridians and parallels defining an element of the shell, (Heyman, 1997)

Figure 2.23: Equilibrium of a small element of shell, (Heyman, 1997)

2.4.5 Vault Research

2.4.5.1 Ideal Vault - Bossut

The French mathematician, Abbe Charles Bossut, set his sight on the formulation of the ideal

vault, neglecting friction and cohesion between the vousoirs. Bossut assumed smooth, frictionless
 44

Figure 2.24: Hoop stress resultants necessary for the equilibrium of a hemispherical shell, (Heyman,
1997)

blocks with joints perpendicular to the intrados. The weights of the blocks along with all non-

vertical forces were assigned to F1 and F2 (Fig. 2.25). The forces were then decomposed into the

internal forces F1l , F1r , F2l and F2r . From equilibrium

F1r = F2l (2.42)

Utilizing the theorem of sines

sin α
Fl = F (2.43)
sin β
sin γ
Fr = F (2.44)
sin β
(2.45)
 45

Figure 2.25: Bossut’s ideal vault formulation, (Benvenuto, 1991)

Eq. 2.42 becomes

F1 sin α2 sin β1
= (2.46)
F2 sin γ1 sin β2

The angles α, β and γ were written as functions of φ which described the rotation of the intrados

curve, and θ which gave the direction of the exterior force (Benvenuto, 1991, pg. 378). If we set

∆φ1 = φ2 − φ1 and ∆φ2 = φ3 − φ2

 46

β1 = ∆φ1 (2.47)

β2 = ∆φ2 (2.48)
π
− γ 1 = φ 1 − θ1 (2.49)
2
π
− α 2 = φ 3 − θ2 (2.50)
2

Substituting into Eq. 2.46

F1 sin ∆φ1 cos(φ3 − θ2 )

= (2.51)
F2 sin ∆φ2 cos(φ1 − θ1 )
F1 sin ∆φ1 cos φ3 cos θ2 + sin φ3 sinθ2
= (2.52)
F2 sin ∆φ2 cos φ1 cos θ1 + sin φ1 sin θ1

Bossut’s formulation dictated that if the path of the intrados and the direction of the forces were

known, the ratio of forces could be found throughout the geometry.

Expanding upon his original formulation, Bossut converted his analysis into differential terms.

Bossut assumed the vault, of basic arch shape, was composed of an infinite number of infinitesimal

blocks. The variables of Eq. 2.52 are modified to their differential form. Let F1 and F2 become

continuous

F1 = f ds (2.53)

F2 = (f + df )ds (2.54)

cos θ1 = cos θ(s) (2.55)

sin θ1 = sin θ(s) (2.56)

cos θ2 = cos θ(s) + d[cos θ(s)] (2.57)

sin θ2 = sin θ(s) + d[sin θ(s)] (2.58)

(2.59)
 47

From the curve of Fig. 2.26, the following relations can be made

dx(s)
cos φ1 = (2.60)
ds
dy(s)
sin φ1 = (2.61)
ds
dx(s) d2 x(s)
cos φ3 = +2 ds (2.62)
ds ds2
dy(s) d2 y(s)
sin φ3 = +2 ds (2.63)
ds ds2
(2.64)

Substituting the relations above into Eq. 2.52 we obtain

d2 x dr dx d2 y dr dy
   
f cos θ 2r 2 + + f sin θ 2r 2 +
ds ds ds ds ds ds
dx d dy d
+r (f cos θ) + r (f sin θ) = 0 (2.65)
ds ds ds ds

Bossut’s new form of analysis allowed him to determine the solution to two problems. First,

the figure of the vault can be determined from the known forces. Second, the forces can be deter-

mined from the known vault figure.

Bossut put his formulation of the best vault figure to practice through the analysis of four

different cases. He first looked at the homogeneous, uniform arch, subject only to self-weight

loading. Through formulating equations and slight derivation, the homogeneous catenary was

determined, confirming the results of Gregory. The second case included a variable vertical loading

condition. The results once again presented the catenary but this time was proportional to the

increasing vertical force. The third case examined the arch loaded normal to the axis, similar to

hydrostatic pressure. The results matched that of a suspended rope. Finally, Bossut looked at the

case of the dome vault. Each groin is considered separately as an arch of variable thickness (Fig.

2.27)

2.4.5.2 Domes of Finite Thickness - Mascheroni

In 1785, Lorenzo Mascheroni published his methodology for the principle problems in con-

structing arches and domes and the calculation of domes of finite thickness. In this review, we will
 48

Figure 2.26: Infinitesimal vault, (Benvenuto, 1991)

focus on the later. In calculating the shape of a dome of finite thickness, Mascheroni first reviewed

the work of Bouguer. Bouguer stated that the resultant of the horizontal thrust and vertical force

from weight must be perpendicular to the joint between the voussoirs of an arch. However, the

resultant force must be at an oblique angle to the joint in the case of a dome. Bouguer derived the

relationship of

Q
> tan φ (2.66)
P
dy
Q = P (2.67)
dx
 49

Figure 2.27: Bossut’s figure for his studies of domes, (Benvenuto, 1991)

to ensure an oblique resultant force angle. He determined that if this condition held true in all

joints around the dome, the forces would ”hinder and annul each other” (Benvenuto, 1991, pg.

420). Even though his realization was wrong, the idea of a possible oblique resultant is correct.

The resultant ”can be oblique to Mn [the joint], because the stresses created by each element of

the dome include not only compressive forces along the meridians, but also lateral internal forces

along the other principal direction” (Benvenuto, 1991, pg. 421). Given the use of masonry, only

compression can be considered because voussoirs cannot exert tensile stress on each other.
 50

Figure 2.28: Bouguer’s analysis of domes and oblique forces, (Benvenuto, 1991)

Mascheroni analyzed the ideal conditions to simplify the analysis. He assumed the joints

were perpendicular to the intrados and that the thickness of the dome was very small compared to

its radius. The simplifications allowed the reduction of the element volume M n0 (Fig. 2.29) to
 
0 1 2
Vol(M n ) = hds + h dφ x (2.68)
2

Since 21 h2 dφ is negligible compared to hxds, Eq. 2.68 is reduced to

Vol(M n0 ) = hxds (2.69)

Since the groin volume, M n0 , is proportional to the wieght, Q, Eq. 2.67 can be rewritten as
Z
dy
hxds > P (2.70)
dx

Through logarithms and differentiating, Eq. 2.71 is derived and can be used to determine
 51

Figure 2.29: Mascheroni’s simplified analysis of finite domes, (Benvenuto, 1991)

the size of the wedge from a known intrados curve.

hxds dy 0
R > 0 (2.71)
hxds y

dy
where y 0 = dx

The final contribution of Mascheroni came about through expanding upon the suggestion of

Bouguer to determine the scale of teh voussoir size from a known arch curve. But instead of the

voussoir proportions, Mascheroni shifts the view to a round dome of constant thickness and the
p
conditions that must satisfy equilibrium. Assuming a arch intrados curve of x = 2Ry − y 2 and

basic calculation, Eq. 2.71 becomes

(R − y)(2R − y) > R2 (2.72)

Solving for y the following relationship is derived

√
3− 5
y< R = 0.382R (2.73)
2
 52

Mascheroni realized that a ”dome of uniform thickness cannot be hemispheric, because its

keystone cannot exceed 0.382R - that is, the generating arch of the dome cannot be more than

51◦ 490 5000 . In the lower zones of such a structure, the lateral stress between groins changes from

compression to tension, and the structure would need to be hooped” (Benvenuto, 1991, pg. 425).

2.4.5.3 Salimbeni

Figure 2.30: Salimbeni’s graphical analysis of force transfer in a dome, (Salimbeni, 1787) (Adapted
by (Benvenuto, 1991))

Mascheroni’s final realization of tensile forces found in the dome was confirmed by the

Veronese engineer and mathematician, Leonardo Salimbeni. In Salimbeni’s published work, he

studied the component, N, transverse to the intrados. As N progresses through the groin of the
 53

dome and away from the keystone, the value decreases until the sign changes. It is observed that

the springing yields outward rather than inward as observed in the arch. Salimbeni sought the

solution to the location of the transfer of forces but could only develop a graphical solution rather

than an exact formulation.

2.5 St. Peter’s Basilica, Rome - Case Study

2.5.1 Introduction

In 1742, the dome of Saint Peter’s Basilica, at Rome, showed severe cracking. Pope Benedict

XIV deemed it necessary to conduct a study of the dome with a detailed analysis and report of

the conclusion. The Pope assigned three mathematicians to review the state of the dome. Frs.

Ruggiero Guiseppe Boscovich, Thomas le Seur, and Francois Jacquir analyzed the dome utilizing

virtual work (Benvenuto, 1991, pg. 352). They broke the dome into a simple static scheme for the

laws of mechanics to be applied. The conclusions presented a case of total separation between the

drum and buttresses and between the interior and exterior portion of the base. Given the results,

the three mathematicians deemed the dome unsafe and susceptible to collapse. They advised the

dome be reinforced with iron rings to prevent a possible collapse.

Unsatisfied with the results, the pope wanted confirmation of the hypothesis of collapse. In

1743, Giovanni Poleni was appointed by Pope Benedict XIV to review the work of the mathemati-

cians and provide his own conclusion. In Poleni’s report of the dome, he focused on the cracks, the

restoration and tension rings (Poleni, 1748).

2.5.2 Three Mathematicians Model

Frs. Ruggiero Guiseppe Boscovich, Thomas le Seur, and Francois Jacquir developed a geo-

metric model to explain the domes damage, Fig. 2.31. The model presented the movement of the

dome and allowed for the evaluation of its stability. Due to sagging of the drum and buttresses,

each groin exhibited subsidence at its upper end and opening or separation at the lower end (López,
 54

2006, pg. 1958). The three mathematicians also modeled the pilasters of the dome, concluding

that they bent outward aiding in the sagging of the dome.

Poleni was asked to review the model of the mathematicians, concluding that a dome sub-

jected to the magnitude of movement presented in the model could not possibly stand. Disproving

their model, Poleni stated that the movement could not have occurred without separating the sec-

tions, as seen from his own developed model, Fig 2.32. The damage to the actual dome should

coincide with that of his own geometric model, which it did not.

Poleni also went into detail on the supporting structural components. Poleni analyzed the

drum as a whole, rather than just the pilasters like the mathematicians, along with the piers

and buttresses. In fact, majority of the damage viewed in the dome, occurred over the piers and

were entirely neglected by the mathematicians. Poleni first looked at the theory of the drum

pilasters developed by the mathematician’s model. He concludes that since the pilasters stood

perpendicular, in order for outward movement to occur, the pilaster’s initial design would have

required an inward bend which is highly doubtful. However, the drum did subside, which Poleni

attributed to a couple factors. He focused on the material bond of the drum, stating that the main

reason for damage is due to the constant pressure of the full weight of the dome resting on the

faulty masonry wall sections. He also attributed damage to the buttresses inability to support the

forces distributed from the main arches (López, 2006, pg. 1964). These two factors contributed to

a slight spreading and sagging of the dome, but nothing to the extent of the movement modeled

by the three mathematicians which would result in total separation of the sections.

2.5.3 Poleni’s Analysis

After disproving the model of the 3 mathematicians, Poleni dove into the theories of arches,

vaults, and domes. He viewed principles related to the stability of vaulted structures and found

that the same type of damage was observed in the drum and buttresses. Since the dome consisted

of wedges and vousoirs, ”whatever is said of the parts of a [vousoir] arch, must equally apply to

the vault and dome” (López, 2006, pg. 1967). In other words, the wedges support one another
 55

Figure 2.31: Figure

Cracked2. Cracked section
section and and structural
structural analysis
analysis of theofdome
the dome of Saint
of Saint Peter’s
Peter’s, (Poleni, 1748)
(Adapted by (López, 2006)) (Le Seur, Jacquier and Boscovich 1743)

Neither did Poleni agree that the double shell vault had suffered the same displacement as the
ribbing. In other
preventing words,due
a collapse thetomodel would be valid for a section of the dome between two planes
gravity.
close together, but not for larger sections. He reached this conclusion not only through a process of
thought but:

Following my inclination to experiment... I have ordered a small model of the Drum and
 Poleni adds in this first manuscript the contributions of Blondel, Parent, Frézier, Dulacq and Stirling
(later he will mention Gregory, but he never quotes Hooke, even in his Memorie, López 1998b,
p.438. About the state of knowledge on the theory of arches, vaults and domes see Benvenuto 1991;
Heyman 1995, 1999 (1995), 2004 (1998); Huerta 1990, 1996, 2004; López 1998b).
56

Figure 3. Analysis of the structural model proposed by Le Seur, Jacquier and

Figure 2.32: Poleni’sBoscovich
analysisinmodel,
their Parere (Riflessioni
(Poleni, 1748) 1743, pl. I) by (López, 2006))
(Adapted

Regarding the resistance of the tension rings, Poleni talked of the contributions of Musschenbroek
and his experiments on iron, and also cites Borelli and Mead. In terms of the capacity for resistance
of ring-shaped
Poleni firstiron, the name
referenced of Johann
Coulomb andBernoulli
his workappears, but distribution
with the above all it is
of his own forces.
thrust writings,He
dating from 1724, that he uses to “explain the tension of the cells forming the fibres of
differentiated the dome’s pressure into its horizontal and perpendicular components. As the thrust
muscles”(Riflessioni 1743, p.26 art.28. See fig.5). As in the Parere, Poleni states that the
relationship
neared between
the base, the load
it more supported
closely bywith
aligned a straight rod and another
the vertical in ring
with less shape thrust
outward is approximately
(Fig 2.33).
thatrealization
This existing between
allowedthePoleni
radiusto
and the perimeter
conclude of the
that the circumference.
gothic dome configuration is better than the

semicircle. He attributed this gothic shape of the dome as the main factor for stability. Referencing
1961
Stirling and Gregory on the relationship between the catenary and the geometric shape of an arch,

Poleni applied Stirling’s theory of spheres to the form of the dome (Fig. 2.34).
 57

Figure 7. The stability of vaulted structures. Cross sections of the dome of Saint Peter’s
Figure 2.33: Cross sections of the dome of Saint Peter’s,
(Riflessioni (Poleni,
1743, pl. II) 1748) (Adapted by (López, 2006))

STABILITY OF THE DOME

For stability, the line of thrust should be contained everywhere within the masonry. He
In order to assess the damage to the dome proper, Poleni posits a series of principles relating to the
divided the dome into 50 half spherical lunes which were subdivided into 16 wedges. He then
stability of the vaulted structures, while pointing out that the damage is the same as that sustained
selected
by the drumone
and2D arch formed
buttresses. by the
Poleni lunesbyforidentifying
begins analysis. The position
those and
parts of theweight of each
structure wedge
which can be
supported by those in
was determined placed
orderimmediately
to accurately below,
interpret perpendicularly
the weight and and those supported
accompanying bythe
thrust of other means,
dome.
in Fig.6 (fig.7). As can be observed, the relationship between some parts and others favours those
Using weighted balls proportional to the recorded segments, a flexible string was hung confirming
which are self-supporting, separated by the GX and DS plans from those exerting outward thrusts,
the shape of the dome as acceptable (Fig. 2.35) for supporting the weight and thrust of the
which means that weight is still exerted on the main arches, pendentives and piers. The upper
structure.
sections, GNAMX and DRAHS, are therefore among those whose vaulted structures remain to be
studied:

The more they incline towards a horizontal position, the more the tendency would be for
them to fall, although this should never actually happen: this inclination also leads to
 58

Poleni also analyzed the affects of cold, heat, humidity, and drying on the various construction

materials, along with the influence of friction and imperfections due to craftmen’s work and external

forces such as earthquakes.

2.5.4 Recommendations

Poleni attributed the damage seen in the dome to defective building materials and methods,

rather than an unstable structure. The vertical cracking of the dome came about from the yielding

of the drum resting on main arches due to the immense weight that the latter have to bear when

compared to the piers. This was confirmed by the fact that the ribbing, ”essential elements of

vault” (López, 2006) were in nearly perfect condition. The damage was not viewed as serious due

to the lack of horizontal cracks. Poleni recommended restoring the structure to its original state

as much as possible. He suggested repairing the cracks, especially in the main arches, by filling the

cracks with bronze wedges, specially cut stone, and fine plaster.

The final recommendation was to reinforce the dome with six new iron rings. Though the

dome was in no danger of collapsing, Poleni felt it fitting to support against any further damage.

Iron rings were to be placed in the drum to reinforce the cylindrical surface. Also two new rings

were to be added to the dome in addition to monitoring of the old rings.

2.5.5 Conclusion

Poleni ultimately determined the dome of St. Peter’s in Rome to be safe and unsusceptible to

collapse. The conclusion came about through detailed static analysis, physical structural inspection

and the theory of Hooke’s hanging chain. Poleni utilized the dome’s thrust components to determine

that the stability of the structure was highly dependent on its geometry. He successfully confirmed

this hypothesis through use of the weighted hanging chain, finding that the chain remained within

the thickness of the entire structural shell. This became the first case that statics and structural

mechanics were applied successfully to a real architectural problem.

 59

2.6 Conclusion

In this chapter, we focused on the historical development of the arch and dome theories of

analysis along with the structural theory of masonry design. The emphasis was placed on the

origination of the theory and how it progressed throughout history to our current methodologies.

Though the review was not extensive, it allows for proper insight into multiple examples of pre-

emodern era analysis arch and shell analysis.

 60

Figure 2.34: Illustrations of the mechanics of masonry arches, (Poleni, 1748) (Adapted by (Heyman,
1998))
 61

Figure 2.35: The hanging chain applied to the analysis of the dome of St Peter’s, (Poleni, 1748)
 62

Figure 13.2.36:
Figure Location of the six
The location of new ironnew
the six rings placed
rings around
placed thethe
around dome ofofSaint
dome SaintPeter’s (Memorie
Peter’s, (Poleni, 1748)
1748) (Adapted by (López, 2006))

CONCLUSIONS

It is well known that Poleni´s analysis of the dome based on the catenary principle, included in his
Memorie (1748), was the first time that the safe theorem of the Limit Analysis was applied to a
masonry structure. However, his manuscripts are not so well known as the Memorie but they are
 Chapter 3

Beams, Plates and Shells

3.1 Introduction

The previous chapter focused exclusively on the historical theory of masonry arch and shell

design. In hopes of grasping an appreciation for hand calculations, we reviewed the historical

analysis of arch and shell design in an age where computers were obsolete. We will now continue

with the idea of hand calculations by introducing the derivations of the differential equations of a

beam and plate subject to flexure. Before we can continue with the discussion of shells, we must

first understand the basics. We will build up from the basic beam, progress to the plate, and finally

begin the discussion of thin shells.

Chapter three begins by viewing a beam in flexure and determining the equation of curva-

ture through equilibrium, compatibility and stress-strain relations. After deriving the differential

equation governing a beam, we advance to the plate. Once again, we utilize our three fundamen-

tal relations of equilibrium, compatibility, and stress-strain to derive the differential equation of a

beam in flexure. The section is summed up by the development of table comparing the elasticity

and plate theory equations.

The chapter continues with the derivation of the equation for thin shells. Following the

previous two methodologies, utilizing our fundamental relations, we develop a set of simultaneous

equations with 11 equations and 11 unknowns containing five stress resultants, three stress couples,

and three displacements. The section is, once again, summed up in a table displaying the equations

for each fundamental relation. The chapter concludes with a discussion of the methodology of the
 64

theory of shallow shells and the membrane theory. The governing assumptions and equations for

each theory are derived.

It must be noted that in order for the reader to fully understand the derivations of this chapter,

they must understand the figures representative of the equations. All equations and derivations are

produced directly from the provided figures found within.

3.2 Beam

3.2.1 Curvature Equation

In order to properly understand the shell theory discussed in the next chapter, we must first

understand the derivation of a simple beam subjected to flexural loading. This required deriving

the basic curvature equation for a beam in flexure utilizing properties and relations defined through

mechanics of materials. Let us consider a segment of a beam (between point 1 and point 2), Fig.

ρ dθ
dθ
ε=φy
ρ dθ
2 dθ=θ2-θ1
ε
u
θ2
φ
θ1 y
1 N.A.

θ1
θ2
dx ds≈dx Linear Strain Distribution

Figure 3.1: Curvature of a Flexural Element

3.1. The slope is denoted by θ, the change in slope per unit length is the curvature φ, the radius

of curvature is ρ. From mechanics of material we have the following relationship

1 dθ
φ= = (3.1)
ρ dx
 65

dy
assuming the flexure only induces a small displacement, ds ≈ dx and θ = dx . The approximate

equation of curvature, Eq. 3.1, can be rewritten as

1 dθ d2 y
φ= = = 2 (3.2)
ρ dx dx

Next, we shall derive the exact expression for the curvature. From Fig. 3.1, we have

dy
tan θ = (3.3)
dx

Defining t as
dy
t= (3.4)
dx

and combining with Eq. 3.3 we obtain

θ = tan−1 t (3.5)

dθ
Applying the chain rule to φ = ds we have

dθ dt
φ= (3.6)
dt ds

ds can be rewritten as 
p
ds = dx2 + dy 2




r  2 
 p
dy ds = 1 + t2 dx (3.7)
= 1 + dx dx 



dy 
t = dx


Next combining Eq. 3.6 and 3.7 we obtain

 
dθ √ dt
φ = dt 1+t2 dx 
 

 
d2 y

 

 
−1 1 √ 1 dt dx2
θ = tan t φ = 1+t2 1+t2 dx
φ=  (3.8)
   2  23
  dy
dθ 1

 dt
= d2 y

 1+ dx
=
 
dx2
2
 dx 
dt 1+t

Thus the slope θ, curvature φ, radius of curvature ρ are related to the y displacement at a point x

along a flexural member by

d2 y
1 dx2
φ= =  (3.9)
ρ  2  32
dy
1+ dx
 66

dy
If the displacements are very small, we will have dx << 1, thus Eq. 3.9 reduces to

d2 y 1
φ= 2
= (3.10)
dx ρ

3.2.2 Differential Equation of the Elastic Curve

We will next derive the basic curvature equation for a beam in flexure utilizing the differential

equation of the elastic curve. Referencing Figure 3.1, a positive dθ at the upper fibers will cause a

differential shortening represented by du.

du = −ydθ (3.11)

Dividing both sides by dx,

du dθ
= −y (3.12)
dx
|{z} dx
ε

Combining this with Eq. 3.2, we derive the fundamental relationship between curvature (φ), elastic

curve (i.e. displacement) (y), and linear strain (ε).

1 ε
=φ=− (3.13)
ρ y

Note that so far we have made no assumptions about the material properties (i.e. it can be elastic

or inelastic). However, we will focus solely on the elastic case:


σ
ε = E

 My
ε=− (3.14)
EI
σ = − MI y 


Combining Eq. 3.14 with Eq. 3.13 yields

1 dθ d2 y M
φ= = = 2 = (3.15)
ρ dx dx EI

This is the fundamental differential equation governing for beams.

 67

3.3 Plates (Adapted from Pilkey & Wunderlich)

Building off the basis of beam flexure, Sec. 3.2, we will next discuss the transverse deformation

of plates. The approach followed will be consistent with the finite element formulation, Fig. 3.2.

3.3.1 Fundamental Relations

3.3.1.1 Equilibrium

Consider an arbitrary plate element with the given stresses, Fig.3.3.

The resultants, per unit width, are given by

 Z t
 2


 Nxx = σxx dz


 − 2t
Z t 
 Z t
2 2
Membrane Force N = σdz Nyy = σyy dz
− 2t 
 − 2t

 Z t

 2
 Nxy =

 σxy dz
− 2t
 Z t
 2


 Mxx = σxx zdz
 − 2t
(3.16-a)


Z t 
 Z t
2 2
Bending Moments M = σzdz Myy = σyy zdz
t
−2 
 − 2t

 Z t

 2
 Mxy = σxy zdz


− 2t
 Z t
2
Vx = τxz dz


Z t 
− 2t
2

Transverse Shear Forces V = τ dz Z t
− 2t  2
V = τyz dz

 y

− 2t

Note that in plate theory, we ignore the effect of the membrane forces. Those in turn will be

accounted for in shells.

The equation of equilibrium is derived by considering an infinitesimal element tdxdy subjected

to an applied transverse load pz . We would have to consider three equations of equilibrium, Fig.

3.4:
 68

Fundamental Relations

Kinematics Constitutive Equilibrium

?
Dierential Equation

?
Variational Formulation

?
Finite Element Discretization

Figure 3.2: Finite Element Formulation

Y
Z

τ xz
X
Qyz
Mx y
Y
Z Qxz M
xx
Myy

∂Mxy
Mxy Mxy + dx
∂Mxx ∂x
Mxx +
∂x
dx

∂Qxz
σxx
∂M Qxz +

τ xy
dx
Myy + yy dy ∂x
∂y

Mxy +
∂Mxy
∂y
dy
Qyz +
∂Qyz
∂y
dy τ yz σyy τ yx

Figure 3.3: Stresses in a Plate

 69
t
τ xy 2
X
σxx t
τ xz
τ xy 2
σyy τ yz σxx
τ xy
τ xy τ xz
Y σyy τ yz

Figure 3.4: Free Body Diagram of an Infinitesimal Plate Element

Summation of Forces in the z direction

∂Vx ∂Vy
dxdy + dxdy + pz dxdy = 0 (3.17)
∂x ∂y

or
∂Vx ∂Vy
+ + pz = 0 (3.18)
∂x ∂y

Summation of Moments about the x axis

∂Mxy ∂Myy
dxdy + dxdy − Vy dxdy = 0 (3.19)
∂x ∂y

or
∂Mxy ∂Myy
+ − Vy = 0 (3.20)
∂x ∂y

Summation of Moments about the y axis

∂Myx ∂Mxx
+ − Vx = 0 (3.21)
∂y ∂x
 70

Since Mxy = Myx , those equations can be expressed in matrix form as

 
Mxx

 

 
 



    
 
∂ ∂
0 −1 0 Myy 0   0 

 
 
   
∂x ∂y 
  
 
 
    
  
 
 
    
 ∂ ∂ 
+ = (3.22)
 0
 ∂y ∂x 0 −1 
 Mxy   0   0 
  
 
 

 
 
 
∂ ∂
    
0 0 0 ∂x ∂y Vx p  0 
   
z

 
   
 
| {z }




LT
 
 Vy 
| {z }
M

Note that the left matrix corresponds to LT where the 1 term has been substituted by −1

3.3.1.2 Kinematic Relations

From Fig. 3.5 we have five displacements u, v, w, θxx and θyy . However, two of the three

displacements, u, v can be expressed in terms of the θxx and θyy which are the rotations of the plate

6
t
2

- x
t
2

?
yy

;; w ?
xx

;
;; -
;;
y u

v ?
w
?
z

Figure 3.5: Displacements in a Plate

middle surface, with the third displacement being the transverse one w. We will use a notation

consistent with the traditional one adopted for plate bending, rather than one consistent with the

coordinate directions as used in finite element. Since we are focusing on thin plates, the middle

surface will be assumed to remain without strain, and the plane sections remain plane. Based on
 71

the plane section remaining plane assumption, we have,

    
 u    z 0 0   θxx

   

   


 
    

= (3.23)
 
 v   0 z 0   θyy
 


 
   

    
 w 
  0 0 1  w
 


However, from mechanics of material, we know the relation  = Lu or

   
∂
εxx   ∂x 0 0 

 


 

 
  
 

 εyy 
   ∂
  0 ∂y 0  
  

u 

 
   
    
∂ 
   
 εzz 
    0 0 ∂z 


= 
 v 
 (3.24)

 γ xy

  ∂ ∂
0  

 
  ∂y ∂x  

w

 
   
 

 
   ∂ ∂ 


 γxz 
  ∂z 0 ∂x  | {z }
u

 
  
   
∂ ∂
 
γyz 0 ∂z ∂y

 

| {z } | {z }
ε L

Since w corresponds to the transverse deflection of the middle surface, and does not vary with z,

then εzz = 0. Substituting Eq. 3.23, we obtain

     
εxx  zθxx,x zκxx

    
 
 


 
 
 
 
 


 
 
 
 
 


  
  
 
 




 ε yy







 zθyy,y 






 zκyy 




 
 
 
 
 

     
 εzz 
    0

  0
 

= = (3.25)
γxy  z(θxx,y + θyy,x ) 2zκxy

    
 
 


 
 
 
 
 


 
 
 
 
 


 
 
 
 
 




 γ xz



 
 (θxx + w,x )









 γxz 




 
 
 
 
 

     
 γyz  
   (θyy + w,y ) 
  γyz
 


Note that this equation assumes that the displacement w is small compared to the thickness of the

plate, and the rotation is small. Because the rotation is small, its square is negligible with respect

to unity, and hence the curvature

∂θ
∂x ∂θ
κ= ≈ (3.26)
(1 + θ2 )3/2 ∂x
 72

is equal to the rate of change of rotation. Hence, the kinematic relation for the transverse displace-

ment of a plate is    
∂
κxx 0 0 

 






  ∂x

 
   
∂
κyy  0 0  θxx

 
   

∂y

 
  

 
   

  
= ∂ ∂ (3.27)
 2κxy   ∂y ∂x 0  θyy 

 
   
   
∂ 
 

γxz  1 0 w 

 
  





  ∂x 

  
 | {z }

 γ  ∂ u
yz
 0 1 ∂y
| {z } | {z }
κ L

3.3.1.3 Constitutive Relations

For three dimensional continuum, the strain-stress relation is

    
εxx   1 −ν −ν σxx

 
 
 


   


 
   


 
   

 εyy 



  −ν 1 −ν 0  σyy 


 
   


 
   

 
 εzz  −ν −ν 1 σzz
   
 1   
=   (3.28)
  E   


 γxy 

  2(1 + ν) 0 0 
 τxy 



 
   


 
   




 γxz 




 0 0 2(1 + ν) 0 


 τxz 




 
   

   
 γyz 
  0 0 2(1 + ν)

 τyz



However from Eq. 3.25, εzz = 0. Therefore we can neglect σzz which is much smaller than the

other stresses. Inverting the previous equation yields

    
σ  1 ν 0 εxx
   
xx 

  
 


 
  

 
   

0  
σyy   ν 1 0 εyy

    

 
  

    

  E   0 0 1−ν
 
τxy =  γxy  (3.29)

 
 1 − ν2 
 2 
 
   
1−ν
    
 τxz  0 γxz 

    
 2  
0

 
   

 
  




  1−ν  
 τyz  0 2
 γyz 
| {z } | {z }| {z }
σ D ε
E
Note that the shear modulus is µ = 2(1+ν) .
 73

Figure 3.6: Positive Moments and Rotations

We now seek to write the moments in terms of the curvatures, Fig. 3.6. Introducing the

stresses from Eq. 3.29 into Eq. 3.16-a, and using Eq. 3.25 we derive the first term
Z t Z t Z t
2 2 E 2 E
Mxx = σxx zdz = 2
(εxx + νεyy )zdz = 2
(zκxx + νzκyy )zdz
− 2t − 2t 1 − ν − 2t 1 − ν
Z t
E 2
2 Et3
= (κ xx + νκyy ) z dz = (κxx + νκyy ) (3.30-a)
1 − ν2 −t 12(1 − ν 2 )
2

Following a similar procedure for the other two terms, we obtain the following moment-curvature

relation     
Mxx  1 ν 0 κxx

 
 
 


 
  

Et3

 
   

= (3.31)
 
 Myy  12(1 − ν  ν 1
2)  0 
 κyy 

 
   

1−ν
   
 Mxy
 
 0 0

 2κxy 

2
| {z } | {z }| {z }
M D κ
Et3
The 12(1−ν 2 )
term is referred to as the flexural rigidity and is analogous to the flexural stiffness EI

of a beam (if the plate has unit width, and ν = 0, then EI = Et3 /12).

3.3.2 Plate Theories: Kirchhoff

In the following section the derived equations from our fundamental relations will be applied

to the plate theory of Kirchhoff. This theory, is primarily applicable to thin plates in which

shear deformations can be neglected as assumed previously. This formulation is analogous to the

conventional Euler-Bernoulli beam theory.

 74

3.3.2.1 Fundamental Relations

Kinematic Relations : Since shear deformations are neglected, γxz = γyz = 0 and thus the last

two relation of Eq. 3.25 reduce to

θxx = −w,x and θyy = −w,y (3.32)

and the first three strains become

∂2w ∂2w ∂2w

xx = −z ; yy = −z ; γxy = −2z ; (3.33)
∂x2 ∂y 2 ∂x∂y

or    
∂2
κxx − ∂x

 
 
 


 
 
 2 


 
 
 
 
= ∂2 (3.34)
 κyy   − ∂y 2
 w

 
 
 
 | {z }
∂ ∂
   
 2κxy −2 ∂x
 
 
  u

∂y
| {z } | {z }
κ L

Constitutive Relation : The constitutive relation of Eq. 3.31 still applies

    
Mxx   1 ν 0  κxx 

 
 
 

   

3
   
  Et    
= (3.35)
 
 M yy  12(1 − ν 2 ) 
 ν 1 0 
 κ yy 

 
    

1−ν
   
 Mxy 
 
0 0  2κxy 
 
2
| {z } | {z } | {z }
M D κ

Equilibrium : The equilibrium equation, as expressed by Eq. 3.22 is also still valid. If we were to

substitute the second and third relations into the first one, we would obtain the following

equilibrium relation in terms of the moments

∂ 2 Mxx ∂ 2 Mxy ∂ 2 Myy

+ 2 + + pz = 0 (3.36)
∂x2 ∂x∂y ∂y 2

d2 M
(Note the similarity with the corresponding equations for beam flexure − Vx = 0)
dx2
 75

3.3.2.2 Differential Equation

If we combine the kinematic and constitutive relation equations, 3.34 and 3.35, we obtain
   
∂ 2 ∂ 2
Mxx   ∂x2 + ν ∂y2  

 


 
3
  
  Et 
 ∂2 2

=− ∂ (3.37)

 Myy  12(1 − ν 2 )  ∂y2 + ν ∂x2  w
 
 
   
(1 − ν) ∂ ∂
 
 Mxy 
 
∂x ∂y

Finally, we substitute the equilibrium equation, 3.36, into the previous one,

∂4w ∂4w ∂4w pz

+ 2 + = Et3
(3.38)
∂x4 ∂x2 ∂y 2 ∂y 4
12(1−ν 2 )

or
pz
∇4 w = Et3
(3.39)
12(1−ν 2 )

Note the similarity with the corresponding equation for beams

∂2 ∂2w ∂4 pz
2
EI 2
= p z or 4
= (3.40)
∂x ∂x ∂x EI

3.3.2.3 Stresses

Combining the stress-strain relation of Eq. 3.29, with Eq. 3.33, and 3.37, the stresses can be

expressed in terms of the moments

Mxx Myy Mxy

σxx = t3
z; σyy = t3
z; τxy = t3
z (3.41)
12 12 12

My
Again we note the analogy with the flexural stress expression in beams σ = I . Using the three

dimensional equilibrium equation:

∂σxx ∂τxy ∂τxz

+ + = 0
∂x ∂y ∂z
∂τyx ∂σyy ∂τyz
+ + = 0 (3.42-a)
∂x ∂y ∂z
∂τzx ∂τzy ∂σzz
+ + = 0
∂x ∂y ∂z
 76

integrating the first equation yields

Z z   Z t/2  
∂σxx ∂τxy 12z ∂Mxx ∂Mxy
τxz = − + dz = + dz
t/2 ∂x ∂y z t3 ∂x ∂y
12 t/2
Z
= 3 zVx dz (3.43-a)
t z

Finally "  2 # "  2 #

3Vx 2z 3Vy 2z
τxz = 1− and τyz = 1− (3.44)
2t t 2t t
Thus, the shear stress distribution across the plate thickness is parabolic (though they tend to be

very small compared to τxy ) and the peak shear stresses occur at the middle surface (z = 0) where

3 Vx 3 Vy
τxz |max = and τyz |max = (3.45)
2 t 2 t

Finally, it can be shown that σzz varies cubically.

3.3.2.4 Variational Formulation

Prior to the finite element discretization, we seek to obtain from the previously derived

relations a variational formulation of the problem. The internal virtual work is given by
Z Z
T
δWi = − δ σdA = − δκT M dA (3.46)
A A

where M = Dκ is obtained from Eq. 3.35, and κ = Lu is obtained from Eq. 3.34. Accounting for

the external virtual work, we obtain

Z Z Z
δW = δWi + δWe = − δuT (LT DL)udA + δuT pv dA + uT pdΓ = 0 (3.47)
| A {z } | A
{z
Γp
}
where pv contains the applied transverse loading pz , and p the edge loading ps . Substituting for

the LT DL terms (Eq. 3.34) we obtain

  
 − ∂2
 1 ν 0
 

 ∂x2
 

Et3 
  

K e = LT DL = b − ∂ 22 ∂ 2 ∂ ∂ c ∂2
 
∂x
− ∂y 2 −2 ∂x ∂y 12(1 − ν 2 )  ν 1
 0 
 − ∂y 2

  

1−ν  −2 ∂ ∂
 
0 0
 

2 ∂x ∂y
= K
∂2 ∂2 ∂2 ∂2 ∂2 ∂2
      
∂ ∂ ∂ ∂
+ ν + 2(1 − ν) + 2 + ν (3.48-a)
∂x2 ∂x2 ∂y 2 ∂x ∂y ∂x ∂y ∂y ∂y 2 ∂x2
 77

Finally, noting that u = w, the virtual work becomes

Z Z
δWe = −K δw (Lwpz ) dA + δwps dΓ = 0 (3.49)
A Γp

where K = Et3 /12(1 − ν 2 ).

3.3.3 Summary

Table 3.1 summarizes some of the major equations governing plate bending, and contrasts

them with the equivalent elasticity ones. Now that we have fully reviewed the derivation of the

beam and plate subjected to flexure, we can increase the difficulty one step farther and begin the

discussion of thin shells.

3.4 Thin Shell Theory

As discussed previously in Sec. 2.4.3, a thin shell is a curved surface or plate whose thickness

is small compared to the dimensions and radii of curvature, rx and ry , of the shelled structure.

This section will focus on the derivation of the differential equation of thin shells. This will be

achieved by first establishing equilibrium of a differential element cut from the shell, and next

by ensuring that each element remains continuous with the element adjacent after deformation,

through achieving strain compatibility.

3.4.1 Definitions and Assumptions

3.4.1.1 Definitions

Before we begin, a few terms must be defined to allow for clarity of the subject. The stress

resultants and stress couples are integrated over the shell thickness and are defined as the total

forces and moments acting per unit length at the middle surface (the face created by bisecting the

thin shell thickness).

In order to properly determine the stress resultants and couples, we will first consider the

infinitesimal segment shown in Fig. 3.7 and more particularly the edge along the y axis. From
 Elasticity Plate Theory (Kirchho )
Kinematic
 ∂ 0 0

  ∂x


εxx 
  0 ∂ 0  
2


 εyy 
  ∂y     
 − ∂2 


 
 
0 0 ∂   ux   κxx  
 ∂x 

εzz 
∂z

∂ 2 
= ∂ ∂
 uy κyy = − w
εxy  0 ∂y 2
∂y ∂x
       

 
   uz 2κxy 
 
 | {z }

   ∂ 0 ∂   −2 ∂ ∂ 
u
 εxz 
  ∂z ∂x
| {z } | {z } ∂x ∂y
εyz ∂ ∂ u κ | {z }
| {z } 0 ∂z ∂y L
ε | {z }
L
   ∂ 0

" # 
 κxx  ∂x     ∂ 1 0  w 
 0 ∂  θxx γxz ∂x
κyy = ∂y  ; = ∂ θxx
 2κ  ∂ ∂
θyy γyz
∂y
0 1  θ 
xy yy
| {z } ∂y ∂x | {z } | {z } | {z }| {z }
| {z } θ γ Ls
κ u
Lf
Equilibrium
 
  σxx     Mxx 
∂ 0 0 ∂ ∂ 0


 σyy


   ∂ 0 ∂ −1 0 
 
    
∂x ∂y ∂z  
  bx  ∂x ∂y    0 

0 ∂ 0 ∂ 0 ∂  σzz 
0 ∂ ∂ 0
  Myy   0 

∂y ∂x ∂z
 +ρ by =0 
∂y ∂x
−1 
  σxy      Mxy  +  0  =  0 
∂ ∂ ∂ ∂ ∂
 b 0
0 0 0

 σxz

 z 0 0 0

  pz
∂z ∂x ∂y 




 | {z } ∂x ∂y  Vx 
Vy

| {z } σyz | {z }
ρb | {z }
LT | {z } LT M
σ
   Mxx 
∂ 0 ∂ −1 0 
 
    
∂x ∂y    0 

0 ∂ ∂ 0
  Myy   0 
M

 ∂y ∂x
−1 
 xy + 0 = 0
∂ ∂
  
0 0 0

   pz  0
∂x ∂y  Vx  
| {z } Vy
| {z }
LT =b LT
f LT
s c M
Constitutive
    
 σxx  1 β β 0 0 0  εxx 
   


 σyy 
  β 1 β 0 0 0  εyy 


  
1 ν 0
 
      Mxx   κxx 
σzz  β β 1 0 0 0  εzz 1 0
= α  Myy =K ν  κyy
 σ xy   0 0 0 γ 0 0  εxy    1−ν  

 
    Mxy 0 0 2κxy


 σ yz 

 0 0 0 0 γ 0 

 εyz 


 | {z }
2
{z }
| {z }|
σzx 0 0 0 0 0 γ εzx κ
M D
| {z } | {z }| {z }
σ D ε   
  1 ν 0
 
 Mxx   κxx 


 Myy


  K ν 1 0  0  


 1−ν   κyy
Mxy =
 0 0 2
 2κ xy
     

 
  1 0 
 

 Vx
Vy
 0 ζ  γxz
γyz

0 1
| {z } | {z }| {z }
M D κ
E(1−ν) ν ; γ = 1−2ν Et3
α = (1+ν)(1−2ν) ; β = 1−ν 2(1−ν)
K = ; 5 µt
ζ = 6
12(1−ν 2 )
Virtual Work Z Z
R R R
δ = δ(Lu)T Dd − δuT bd − δuT ^
td =0 δW = − δw (KLwpz ) dA + δwps d = 0
t A
Z Z p Z Z
T e 6ks (1 − ν) T e T T
δW = −K δθ Kf θdA + δu Ks udA + δu pV dA + δu pd =0
A t2 A A

Table 3.1: Comparison of Governing Equations in Elasticity and Plate Bending

78
 79

equilibrium we have
Z +h
2
Nxx ry dαy = σ xx (ry − z)dαy dz (3.50)
| {z } −h
2
Resultant force on neutral axis | {z }
internal force

Dividing both sides by ry dαy , we obtain the membrane force acting along the x axis
Z +h  
2 z
Nxx = σ xx 1 − dz (3.51)
−h ry
2

 
z
The 1 − ry term, which was not present in Eq. 3.50, accounts for the fact that our section is

really a wedge rather than rectangular.

The differential element of consideration can be seen in Fig. 3.7. along with the stress

Figure 3.7: Differential Shell Element, Forces, (Billington, 1965)

resultants and stress couples in Fig. 3.8 The vectors of stress resultants are seen in Fig. 3.9 and

the vectors of stress couples in Fig. 3.10. The derivation of the following stress resultants and

stress couples follow the same formulation as the membrane force Nxx from Eq. 3.51.
 80

Figure 3.8: Differential Shell Element, Stresses, (Billington, 1965)

Figure 3.9: Differential Shell Element, Vectors of Stress Resultants, (Billington, 1965)
 81

Figure 3.10: Differential Shell Element, Vectors of Stress Couples, (Billington, 1965)
 82

Membrane Force
+h
 Z  
2 z
σxx 1 −


 Nxx = dz


 −h ry
 2
+h
 Z  
z

 2
 Nyy = σyy 1 − dz


+h

rx
Z
2
 z −h
N = σ 1− dz 2
+h
r
Z  
−h  2 z
2
Nxy = σxy 1 − dz


ry

−h


 2
+h

 Z  

 2 z
 Nyx =

 σxy 1 − dz
−h rx
2

Bending Moments
+h
 Z  
2 z
σxx z 1 −


 Mxx = dz (3.52)


 −h ry
 2
+h
 Z  
z

 2
 Myy = σyy z 1 − dz


+h

rx
Z
2
 z −h
M = σz 1 − dz 2
Z +h
r
 
−h  2 z
2
Mxy = − σxy z 1 − dz


ry

−h


 2

 Z +h  

 2 z
 Myx

 = σxy z 1 − dz
−h rx
2

Transverse Shear Forces

 Z +h  
 2 z
Q = τxz 1 − dz

 x

+h

ry
Z
2
 z −h
Q = τ 1− dz 2
+h
r
Z  
−h  2 z
2
Q = τyz 1 − dz

 y

rx

−h2

3.4.1.2 Assumptions

For the purpose of simplification, a number of assumptions are considered:

Thin shell: If the terms z/rx and z/ry are neglected when they appear with unity, and with

τxy = τ yx then

Nxy = Nyx (3.53-a)

Mxy = −Myx (3.53-b)

 83

Small-Deflection: changes in geometry will not affect the static equilibrium.

Linear Elastic Behavior: linear relationships between stress and strain (or moment and curva-

ture).

Conservative System: Points on lines normal to the middle surface before deformation remain

on lines normal to the middle surface after deformation.

Neglect Shear Deformations: Radial shears Qx and Qy will not be considered.

3.4.2 Derivation of Governing Differential Equation

Based on preliminary definitions/assumptions, we can now derive a general shell theory

through the following steps:

(1) Determine the equilibrium of forces and moments on a differential element (5 equations,

8 unknowns).

(2) Establish the strain-displacement relationships (6 equations, 3 unknowns).

(3) Establish a stress-strain relationship by assuming a set of elastic properties (3 equations,

6 unknowns).

(4) Transform the the force-strain relationships into force-displacement equations (6 equa-

tions 3 unknowns).

(5) Obtain a complete formulation by combining the force-displacement equations with the

equilibrium equations (11 equations with 11 unknowns).

 84

3.4.2.1 Equilibrium

Considering the differential element of a shell, Fig. 3.7, we now consider the 6 equations of

static equilibrium
ΣFx = 0 ΣMx = 0

ΣFy = 0 ΣMy = 0 (3.54)

ΣFz = 0 ΣMz = 0
In shell theory we usually neglect Mz because τxz and τyz are neglected. So we have to satisfy 5

equations of equilibrium, Eq. 3.52, in terms of 8 generalized forces, Eq. 3.54.

Equilibrium Forces

ΣFx , ΣFy , ΣFz , ΣMx , ΣMy Nx , Ny , Nxy , Qx , Qy , Mx , My , Mxy

(a) ΣFx = 0:

We will consider the contribution of each term separately, and then combine all components together

to form the completed equilibrium equation. Contributions from the stress resultants, Nxx , Nyy ,

Nxy , Nyx , two radial shear forces, Qx , Qy , and the external pressures will be considered. The stress

couples attribute no force component in the x direction. Each component will be determined by

multiplying the stress resultant, Fig. 3.9, by the length of the element side on which it acts.

From Fig. 3.7, a difference in geometry of the differential element can be noted. The sides

of the element which intersect the origin have curved lengths of ax dαx and ay dαy , whereas the
∂ax ∂ay
opposing edges have increased lengths of (ax + ∂αy )dαx and (ay + ∂αx )dαy .

Contribution from Nx :
  
∂Nx ∂αy
ΣFxNx = −Nx ay dαy + Nx + dαx ay + dαx dαy (3.55-a)
∂αx dαx
∂αy ∂Nx ∂Nx ∂ay
= Nx dαx dαy + ay dαx dαy + dαx dαx dαy (3.55-b)
∂αx ∂αx ∂α ∂αx
| x {z }
nd
2 order effect
By combining the first two terms and eliminating the third due to second order effects

∂(Nx ay )
ΣFxNx = dαx dαy (3.56)
∂αx
 85

Contribution from Ny : First we note that in general Ny does not act perpendicular to the x

axis but at some angle

(∂ay /∂αx )dαx dαy

∆x = (3.57-a)
ax dαx
∂ay dαy
= (3.57-b)
∂αx ax

thus in the x direction, the component of the total force Ny ax dαx is

Ny ∂ay dαy
ΣFx = −Ny ax dαx (3.58-a)
∂αx ax
∂ay
= −Ny dαx dαy (3.58-b)
∂αx

Contribution from Nxy : As for Nx y, this force also has a non perpendicular component and is

determined along the angle ∆y :

(∂ax /∂αy )dαy dαx

∆y = (3.59-a)
ay dαy
∂ax dαx
= (3.59-b)
∂αy ay

and the component of the total force Nxy ay dαy is

Nxy ∂ax dαx

ΣFx = Nxy ay dαy (3.60-a)
∂αy ay
∂ax
= Nxy dαy dαx (3.60-b)
∂αy

Contribution from Nyx : is similar to the one of Nx , Eq. 3.56

  
N ∂Nyx ∂ax
ΣFx yx = −Nyx ax dαx + Nyx + dαy ax + dαy dαx (3.61-a)
∂αy ∂αy
∂(Nyx ax )
= dαx dαy (3.61-b)
∂αy
 86

Figure 3.11: Differential Shell Element, Qx Stress Resultants, (Billington, 1965)

 87

Contribution from Qx : The horizontal component of the total force on the sloping side, Fig.

3.11, is
  
∂Qx ∂ay ax
ΣFxQx = − Qx + dαx ay + dαx dαy dαx (3.62)
∂αx ∂αx rx

where dαx ' sin dαx . We note that Qx has no x component on the negative side. Neglecting

higher order terms, this reduces to

ax
ΣFxQx = −Qx ay dαy dαx (3.63)
rx

Contribution from Qy : The stress resultant will contribute to the equilibrium in the x direction

only if ax and ay are not principle radii of curvature. In Fig. 3.12 the differential element

with curvatures, which result in a total distance of zx + zy + zxy at the corner opposite to

Figure 3.12: Differential Shell Element, Principal Curvatures, (Billington, 1965)

the origin, is shown. zx and zy are due to the slopes ∂z/∂αx and ∂z/∂αy . The change in

slope results in

1 ∂2z
zxy = ax ay dαx dαy (3.64-a)
ax ay ∂αx ∂αy
ax ay
= dαx dαy (3.64-b)
rxy
 88

where
1 1 ∂2z
= (3.65)
rxy ax ay ∂αx ∂αy

and corresponds to the twist of the surface with respect to the x and y axes. Hence, the

slope of Qy is
zxy ay dαy
= (3.66)
ax dαx rxy

and the total component of Qy in the x direction is

  
Q ∂Qy ∂ax ay
ΣFx y = Qy + dαx ax + dαy dαx dαy (3.67-a)
∂αx ∂αy rxy
ay dαy
= Qy ax dαx (3.67-b)
rxy

This expression vanishes where the element is bounded by lines of principal curvature rx

and ry , zxy = 0, the twist rxy = ∞.

Contribution from the external load: Assumed to be a pressure with components px , py and

pz acting on the differential element of an area ax dαx ay dαy :

ΣFxp = px ax dαx ay dαy (3.68)

Finally, we can now combine the x components of Nx , Ny , Nxy , Nyx , Qx , Qy and the load from

Eqs. 3.56, 3.58-b, 3.60-b, 3.61-b, 3.63, 3.67-b, and 3.68. After cancelling the common multipliers

dαx dαy , we obtain

∂(Nx ay ) ∂ay ∂ax ∂(Nyx ax ) ax ay ax

− Ny + Nxy + − Qy − Qx ay + px ax ay = 0 (3.69)
∂αx ∂αx ∂αy ∂αy rxy rx

(b) ΣFy = 0:

This second equation is exactly the same as the one corresponding to ΣFx = 0 except that we

switch the x and y terms yielding

∂(Ny ax ) ∂ax ∂ay ∂(Nxy ay ) ay ax ay

− Nx + Nyx + − Qx − Qy ax + py ax ay = 0 (3.70)
∂αy ∂αy ∂αx ∂αx ryx ry

(c) ΣFz = 0:
 89

The z direction is always defined to be perpendicular to the middle surface; hence it is always in

the direction of principal radii of curvature, Fig. 3.11. Thus the assumption for small angles can

be made that sin θ ' θ.

Similar to ΣFx = 0, contributions from the stress resultants, Nxx , Nyy , Nxy , Nyx , two radial

shear forces, Qx , Qy , and the external pressures will be considered. The stress couples attribute no

force component in the z direction. Each component will be determined by multiplying the stress

resultant, Fig. 3.9, or radial shear force, Fig. 3.11, by the length of the element side on which it

acts.

Contribution from Nx :
ax
ΣFzNx = Nx ay dαy dαx +2nd order terms (3.71)
rx
| {z }
sin θ'θ

All of the following expressions will be similar to the layout of the Nx formulation by

altering the x and y variables.

Contribution from Ny :
Ny ay
ΣFz = Ny ax dαx dαy (3.72)
ry

Contribution from Nxy :

Nxy ax
ΣFz = Nxy ay dαy dαx (3.73)
rxy

Contribution from Nyx :

Nyx ay
ΣFz = Nyx ax dαx dαy (3.74)
rxy

Contribution from Qx :
    
∂Qx ax ∂ay
ΣFzQx = −Qx ay dαy + Qx + dαx cos dαx ay + dαx dα(3.75-a)
y
∂αx rx ∂αx
| {z }
'1
∂(Qx ay )
= dαx dαy (3.75-b)
∂αx
 90

Contribution from Qy : By analogy to Qx

Qy ∂(Qy ax )
ΣFz = dαy dαx (3.76)
∂αy

Contribution from the external load: From Eq. 3.68

ΣFzp = pz ax dαx ay dαy (3.77)

Finally, we can now combine the z components of Nx , Ny , Nxy , Nyx , Qx , Qy and the load

from Eq. 3.71, 3.72, 3.73, 3.74, 3.75-b, 3.76, and 3.77 respectively. Again after canceling the

common multipliers dαx dαy , we obtain

∂(Qx ay ) ∂(Qy ax ) ax ax ay ay
+ + Nx ay + Nxy ay + Nyx ax + Ny ax + pz ax ay = 0 (3.78)
∂αx ∂αy rx rxy rxy ry

(d) ΣMx = 0:

The solutions for moments have components which are similar to the ones of forces seen above, ex-

hibiting differences primarily with the contributors. No contributions are from the stress resultants

and only one radial shear forces, Qy , is considered. However, now the stress couples, Mx , My , Mxy ,

Myx , attribute force components in the x, y, and z directions. Each component will be determined

by multiplying the stress couples, Fig. 3.10, or radial shear force, Fig. 3.11, by the length of the

element side on which it acts.

Contribution from Mx : Has a component about the x axis which is analogous to the one of Nx

in the y direction (Eq. 3.56 with x and y reverted)

∂ax dαx
ΣMxMx = Mx ay dαy (3.79-a)
∂αy ay
∂ax
= Mx dαy dαx (3.79-b)
∂αy

Contribution from My : As in Eq. 3.56, and with reference to Fig. 3.10

  
My ∂My ∂ax
ΣMx = My ax dαx − My + dαy ax + dαy dαx (3.80-a)
∂αy dαy
∂(My ax )
= − dαy dαx (3.80-b)
∂αy
 91

Contribution from Mxy : is analogous to Eq. 3.80-b

  
M ∂Mxy ∂ay
ΣMx xy = −Mxy ay dαy + Mxy + dαx ay + dαx dαy (3.81-a)
∂αx dαx
∂(Mxy ay )
= dαx dαy (3.81-b)
∂αx

Contribution from Myx :

Myx ∂ay dαy

ΣMx = −Myx ax dαx (3.82-a)
∂αx ax
∂ay
= −Myx dαx dαy (3.82-b)
∂αx

Contribution from Qy :
Q
ΣMx y = Qy ax dαx ay dαy (3.83)

Finally, we can now combine the effect of all the components Mx , My , Mxy , Myx , and Qy

Eq. 3.79-b, 3.80-b, 3.81-b, 3.82-b and 3.83 respectively. After canceling the common multipliers

dαx dαy , we obtain

∂(My ax ) ∂ax ∂ay ∂(Mxy ay )

− + Mx − Myx + + Qy ax ay = 0 (3.84)
∂αy ∂αy ∂αx ∂αx

(e) ΣMy = 0:

Moments about the y axis give the same expression as in Eq. 3.84 except that the x and y subscripts

flip, and the sign of the twisting-stress couples are reversed.

∂(Mx ay ) ∂ay ∂ax ∂(Myx ax )

− + My + Mxy − + Qx ax ay = 0 (3.85)
∂αx ∂αx ∂αy ∂αy

(f ) Complete Set of Equilibrium Equations:

 92

The 5 equilibrium equations, Eq. 3.69, 3.70, 3.78, 3.84 and 3.85 are summarized below

∂a ∂ax ax ay
ΣFx = ∂
∂αx (Nx ay ) − Ny ∂αyx + Nxy ∂α y
+ ∂
∂αy (Nyx ax ) − Qy rxy − Qx ay arxx + px ax ay = 0
∂ ∂ax ∂a ∂ ay ax a
ΣFy = ∂αy (Ny ax ) − Nx ∂α y
+ Nyx ∂αyx + ∂αx (Nxy ay ) − Qx ryx − Qy ax ryy + py ax ay = 0
a a
ΣFz = ∂
∂αx (Qx ay ) + ∂
∂αy (Qy ax ) + Nx arxx ay + Nxy raxy
x y
ay + Nyx rxy ax + Ny ryy ax + pz ax ay = 0
∂ax ∂a
ΣMx = − ∂α∂ y (My ax ) + Mx ∂α y
− Myx ∂αyx + ∂
∂αx (Mxy ay ) + Qy ax ay = 0
∂a
∂ax
ΣMy = − ∂α∂ x (Mx ay ) + My ∂αyx + Mxy ∂α y
− ∂
∂αy (Myx ax ) + Qx ax ay = 0
(3.86)

3.4.2.2 Compatibility

(a) εx0 :

The strain-displacement relations will next be derived with reference to Fig. 3.13. The linear

Figure 3.13: Differential Shell Element, Initial and Deformed State, (Billington, 1965)

element of length ax dαx undergoes three changes in length:

∂u
Axial extension x axis of ∂αx dαx
 93

Lateral extension y: The original undeformed shape of the differential element is not rectangular

(as a result of dαx and dαy ), and the angle between the positive y side and the vertical is
∂ax dαx
(Eq. 3.59-b) ∆y = ∂αy ay (rather than 0); Thus, a translation along the y axis, will result

in a translation along the x axis equal to v∆y .

Decrease in radius of curvature: −w arxx dαx

Summing all those terms, and dividing by the original length ax dαx yields

1 ∂u v ∂ax w
εx0 = + − (3.87)
ax ∂αx ax ay ∂αy rx

where the 0 subscript indicates that the middle surface is being considered.

(b) εy0 :

The strain of the middle surface along the y axis, by analogy, is obtained by reversing the x and y

variables along with v and u in the preceding equation.

1 ∂v u ∂ay w
εy0 = + − (3.88)
ay ∂αy ax ay ∂αx ry

(c) γxy0 :

The shearing strain, representing the total angular change between ay dαy and ax dαx , is also some-

what complex, and its three components will be derived separately.

∂v
Difference in displacements: For the v term, Fig. 3.13, we have (v + ∂α x
dαx ) − v. If we divide

by the elemental length, ax dαx , and repeat the same operation for the u displacement, we

obtain:
I 1 ∂v 1 ∂u
γxy0 = + (3.89)
ax ∂αx ay ∂αy

Increase in length of positive sides: We start by determining ∆s, Fig. 3.14

 
∂u
∆s = u + dαx ∆y (3.90)
∂αx
 94

Figure 3.14: Differential Shell Element, Initial and Deformed State; Shear Strain Caused by Increase
in Length of Positive Sides, (Billington, 1965)
 95

where ∆y was derived in Eq. 3.59-b as

∂ax dαx
∆y = (3.91)
∂αy ay

hence,
 
∂u ∂ax dαx
∆s = u+ dαx (3.92-a)
∂αx ∂αy ay
∂ax dαx ∂u ∂ax dαx
= u + dαx (3.92-b)
∂αy ay ∂αx ∂αy ay
| {z }
Higher order term

By dividing ∆s by the element length, ax dαx and repeating the same operation for the

other side, we obtain

II u ∂ax v ∂ay
γxy0 =− − (3.93)
ax ay ∂αy ax ay ∂αx

where the negative sign reflects the fact that the angle between the x = 0 and y = 0 faces

has decreased (originally was a right angle).

Twist of Surfaces: With reference to Fig. 3.15, we have

Figure 3.15: Differential Shell Element, Initial and Deformed State; Shear Strain Caused by Twist
of Surface, (Billington, 1965)
 96
ax dαx
w∗ = −w (3.94)
rxy

Dividing by original length ax dαx and recognizing that we have a similar term for the plane

x = 0, we obtain
III w
γxy0 = −2 (3.95)
rxy

Thus, summing all three components, Eq. 3.89, 3.93 and 3.95, we obtain

1 ∂v 1 ∂u u ∂ax v ∂ay w
γxy0 = + − − −2 (3.96)
ax ∂αx ay ∂αy ax ay ∂αy ax ay ∂αx rxy
| {z } | {z } | {z }

(d) Changes in Curvature: Considering the x − z plane, Fig. 3.16, and assuming w = 0,

Figure 3.16: Differential Shell Element, Initial and Deformed State; Rotation, (Billington, 1965)

the normals to the differential element on the positive face before and after deformation
u ∂w ∂w
make an angle rx . The conventional term due to transverse displacement is ∂sαx = ax ∂αx .

Finally accounting also for the twisting term we obtain

u ∂w v
φx = + + (3.97)
rx ax ∂αx rxy

Similarly we obtain
v ∂w u
φy = + + (3.98)
ry ay ∂αy rxy
 97

The change in curvature will thus be equal to the change in rotation ∂φx per arc length

ax ∂αx
1 ∂φx
χx = (3.99)
ax ∂αx

Furthermore, due to the non parallel sides of the element, because of the changing values of

ax and ay , the rotation φy in the y direction will produce a component in the x direction.

φy ∆y φy ∂ax
= (3.100)
ax ∂αx ax ay ∂αy

Hence the complete expressions for changes in curvature become

1 ∂φx φy ∂ax
χx = + (3.101-a)
ax ∂αx ax ay ∂αy
1 ∂φy φx ∂ay
χy = + (3.101-b)
ay ∂αy ay ax ∂αx
1 ∂φx 1 ∂φy φx ∂ax φy ∂ay
2χxy = ++ − − (3.101-c)
ay ∂αy ax ∂αx ax ay ∂αy ay ax ∂αx

Summary:

We now summarize all six strain expressions in terms of the three unknown displacements, Eq.

3.87, 3.88, 3.96 and the preceding set, we have

1 ∂u v ∂ax w
εx0 = ax ∂αx + ax ay ∂αy − rx
1 ∂v u ∂ay w
εy0 = ay ∂αy + ax ay ∂αx − ry
1 ∂v 1 ∂u u ∂ax v ∂ay
γxy0 = ax ∂αx + ay ∂αy − ax ay ∂αy − ax ay ∂αx − 2 rwxy
(3.102)
1 ∂φx φy ∂ax
χx = ax ∂αx + ax ay ∂αy
1 ∂φy φx ∂ay
χy = ay ∂αy + ay ax ∂αx
1 ∂φx 1 ∂φy φx ∂ax φy ∂ay
2χxy = ay ∂αy + ax ∂αx − ax ay ∂αy − ay ax ∂αx

where
u ∂w v
φx = rx + ax ∂αx + rxy
(3.103)
v ∂w u
φy = ry + ay ∂αy + rxy
 98

3.4.2.3 Stress-Strain

The strain εx at a distance z from the middle surface, Fig. 3.17,

Figure 3.17: Middle surface of differential element, (Billington, 1965)

is composed of an axial component εx0 caused by extension of the middle surface, and another
 99

caused by bending.

L2 − L1
εx = (3.104-a)
L1
 
z
L1 = ds 1 − (3.104-b)
rx
 
z
L2 = ds(1 + εx0 ) 1 − 0 (3.104-c)
rx
0
(1 + εx0 )(1 − z/rx ) − (1 − z/rx )
εx = (3.104-d)
1 − z/rx
1 − z/rx + εx0 (1 − z/rx0 ) − 1 + z/rx
0
= (3.104-e)
1 − z/rx

we can drop the small term z/rx0 and z/rx when they appear with unity,
 
1 1
εx = εx0 − z − (3.105)
rx0 rx

Again χ, referring to the change in curvature, we find

εx = εx0 − zχx (3.106-a)

εy = εy0 − zχy (3.106-b)

γxy = γxy0 − 2zχxy (3.106-c)

At this point, we will no longer ignore the material properties of the system. In order to derive the

stress-strain relationship, we will assume that we have a linear elastic, isotropic and homogeneous

material. Since we have already assumed σz = 0, we have a plane stress condition and from

mechanics of materials:

E
σx = (εx + νεy ) (3.107-a)
1 − ν2
E
σy = (εy + νεx ) (3.107-b)
1 − ν2
τxy = Gγxy (3.107-c)

E
where the modulus of rigidity, G = 2(1+ν) .

Substituting the actual expressions for the strains, Eqs. 3.106-a, 3.106-b, and 3.106-c, we
 100

obtain

E
σx = [(εx0 + νεy0 − z(χx + νχy )] (3.108-a)
1 − ν2
E
σy = [(εy0 + νεx0 − z(χy + νχx )] (3.108-b)
1 − ν2
τxy = G(γxy0 − 2zχxy ) (3.108-c)

Substituting the previous set of equations into the expressions for the force-stress relations,

Eq. 3.52, integrating over h and neglecting the z/rx and z/ry terms when they appear next to

unity, we obtain

Nx = K(εx0 + νεy0 ) (3.109-a)

Ny = K(εy0 + νεx0 ) (3.109-b)

Nxy = Nyx0 = Ghγxy0 (3.109-c)

Mx = −D(χx + νχy ) (3.109-d)

My = −D(χy + νχx ) (3.109-e)

Gh3
Mxy = −Myx = χxy = D(1 − ν)χxy (3.109-f)
6

where

Eh
K = (3.110-a)
1 − ν2
Eh3
D = (3.110-b)
12(1 − ν 2 )

K is axial rigidity and D the flexural rigidity, corresponding to EA and EI for the equivalent one

dimensional problem. We note that K and D are larger than the corresponding one-dimensional

constants by a factor 1 − ν 2 which represents the increase in rigidity caused by restriction on lateral

strains.
 101

3.4.2.4 Final Equations

We now substitute the axial and flexural strains given by Eq. 3.102 into Eqs. 3.109-a through

3.109-f to obtain
h  i
1 ∂u v ∂ax w 1 ∂v u ∂ay w
Nx = K ax ∂αx + ax ay ∂αy − rx +ν ay ∂αy + ax ay ∂αx − ry
h  i
1 ∂v u ∂ay w 1 ∂u v ∂ax w
Ny = K ay ∂αy + ax ay ∂αx − ry +ν ax ∂αx + ax ay ∂αy − rx
 
u ∂ax v ∂ay
Nxy = Nyx = Gh a1x ∂α∂v
x
+ 1 ∂u
ay ∂αy − ax ay ∂αy − ax ay ∂αx − 2w
rxy
h  i (3.111)
∂φx φ ∂φ ∂a
Mx = −D a1x ∂α x
∂ax
+ ax ay y ∂α y
+ ν a1y ∂αyy + aφx axy ∂αyx
h  i
∂φ ∂a φy ∂ax
My = −D a1y ∂αyy + aφx axy ∂αyx + ν a1x ∂α ∂φx
x
+ ax ay ∂αy
 
D(1−ν) 1 ∂φx 1 ∂φ y φx ∂ax φy ∂ay
Mxy = −Myx = 2 ay ∂αy + ax ∂αx − ax ay ∂αy − ax ay ∂αx

where we recall that from Eq. 3.97

u ∂w v
φx = + + (3.112-a)
rx ax ∂αx rxy
v ∂w u
φy = + + (3.112-b)
ry ay ∂αy rxy

These six equations contain three stress resultants (Nx , Ny , and Nxy ), three stress couples (Mx ,

My , and Mxy ) and three displacements (u, v and w). The five equations of equilibrium, Eq. 3.86,

in turn contain eight unknowns: five stress resultants (Nx , Ny , Nxy , Qx and Qy ), and three stress

couples (Mx , My , and Mxy ). When those two sets of equations are comnined together, we have a

total of 11 equations with 11 unknowns (Nx , Ny , Nxy , Qx , Qy , Mx , My , and Mxy , u, v and w).

Thus the problem is properly defined and a complete solution is possible.

3.4.2.5 Summary

Table 3.2 summarizes some of the major equations governing thin shells.

3.4.3 Simplifications

The derivation of the analysis discussed in the previous section (11 equations - 11 unknowns)

can be used to solve the stresses and displacements for any type of thin elastic shell. However,
 Kinematic
 ∂ ∂ax a ay 
  ay ∂α − x
r  
x0 x ∂αy x
  1 
 ∂ay ∂ a ay  u 

 =  ∂αx
ax ∂α − x  v
 γ y0  ax ay  y ry  w 
xy0 ∂ ∂ax ∂ ∂a 2a a
ax ∂α − ∂α ay ∂α − ∂α)y − rx y
| {z } y y x x xy | {z }
ε | {z } u
L
∂
 ∂ax 
  ay ∂α
x ∂αy 
 χxx  1 
 ∂ay ∂

 θxx
χyy =  ∂αx
ax ∂α 
  ax ay  y  θyy
2χxy ∂ ∂ax ∂ ∂a | {z }
ax ∂α − ∂α ay ∂α − ∂αy
| {z } y y x x
χ | {z } θ
L
Equilibrium
 ∂ (a ) ∂a ∂ax ∂ (a )

∂αx y − ∂αy + ∂α x 0 0 0  
 x ∂αx y 
Nxx

1 0 0 − r1 − r1  
∂ax ∂ay x xy
 
 ∂ (a ) ∂ (a ) 0 0 0   Pxx
 − ∂α x + ∂α y  Nyy 
 
1 0
 
 

y ∂αy ∂αx x   0 − r1 − r1   

 ax ay ax a 


Nxy
  xy y   Pyy 
a a a + r y ax 0 0 0 +ax ay   Pzz =0

 rx y ry x rxy y xy

 Mxx   0 0 1 ∂ (a )
∂αx y
∂ (a )
∂αy x   
 ∂ax ∂ ∂a  
   
 

 0 0 0 ∂αy
∂ (a )
− ∂α x (ay ) − ∂αy  Myy 
  0 0 0 0 1   Qxx 
 y ∂αx x   Qyy
∂ay ∂ax
Mxy 0 0 0 1 0
0 0 0 ∂ (a ) ∂ (a ) b| {z }
− ∂α y − ∂α x | {z }
ρ
x ∂αx ∂αy y
| {z } M
LT
Constitutive
 
 xx0 

   
 σxx  1 ν 0 −z −zν 0  yy0 





1 0 −zν −z 0 γxy0
σyy = β ν 
 1 −2z  χ xx 
 τ
xy 0 0 β
0 0 β





| {z }  χyy 

| {z }
σ χxy
D | {z }
κ
β = E
1−ν 2

Table 3.2: Summary of major governing equations in thin shells

102
 103

even an analysis of this sort can be extremely complicated for even the simplest thin shell systems.

Thus, shell design has often utilized two types of simplifications to ease the analysis.

The first type of simplification is of the mathematical formulations. Often we assume or

prove certain terms of relative unimportance and remove these terms from the equation of interest.

This has already been performed in Sec. 3.4.2 by neglecting the terms z/r when they appear with

unity, along with the assumption of plane stress, σz = 0. The purpose of this simplification is to

reduce the number of terms while still obtaining a reasonable mathematical solution.

The second type of simplification is of the physical formulations. Often we can assume a

specific structural action to ease the analysis. This assumption can be used to replace the shell

system with an equivalent simpler structure. This new, simpler structure is then utilized for analysis

of the complex one.

These simplifications are desirable for RC design where precise analysis is not justified due

to a lack of construction accuracy. Even so, it is vital for the designer to understand the real

structural action obtained only through the physical features of the system and the corresponding

mathematical formulation.

Two simplified theories of analysis are considered, the theory of shallow shells and membrane

theory, and will be presented in the sections to follow.

3.4.4 Shallow Shells

Shallow shells is the first of two simplified shell theories we will review. The following will

utilize both simplifications of mathematical and physical system formulations discussed in the

previous section.

3.4.4.1 Assumptions

Shallow shell theory will utilize the following assumptions:

(1) The shell’s slope is small compared to some reference plane (often the horizontal plane for

roofs).
 104

(2) The surface curvature is small.

(3) The shell boundaries are such that the surface loads are carried by in-plane stress resultants

(Nx , Ny and Nxy ). Therefore the transverse deflections will be much greater than the in-

plane deflections.

(4) The changes in curvature of the surface are small.

3.4.4.2 Application of Assumptions

We will now apply these assumptions to Eqs. 3.86 and 3.111. Assumption 1 allows us to

neglect the radial component of loading. From assumption 2, the terms containing Q in the first

two of Eq. 3.86 and N in the third of Eq. 3.86 are small. This also applies to the terms with w in

the first three of Eq. 3.111 and the u and v terms in the last three of 3.111. However, we cannot

just drop these terms without first considering the third assumption.

Assumption 3 relates the shell to a flat arch, which, under uniform loading, the axial stress

and vertical displacements control. Meaning, the small transverse components of the in-plane stress

resultants, in the third of Eq. 3.86, are not negligible in comparing the change in the transverse

shear resultant.

Combining assumptions 2 and 3 together, we can drop Q from the first two terms of Eq.

3.86 but not N from the third. From Eq. 3.111 we are allowed to drop the effect of in-plane

displacements (u and v terms) on stress couples but not the effect of displacement (w terms) on

in-plane stress resultants. Assumption 4 allows us to take ax and ay as constants.

Eqs. 3.86 and 3.111 can be rewritten as:

 105

1 ∂Nx 1 ∂Nyx
+ + px = 0 (3.113-a)
ax ∂αx ay ∂αy
1 ∂Ny 1 ∂Nxy
+ + py = 0 (3.113-b)
ay ∂αy ax ∂αx
1 ∂Qx 1 ∂Qy Nx 2Nxy Ny
+ + + + + pz = 0 (3.113-c)
ax ∂αx ay ∂αy rx rxy ry
1 ∂My 1 ∂Mxy
− + + Qy = 0 (3.113-d)
ay ∂αy ax ∂αx
1 ∂Mx 1 ∂Myx
− − + Qx = 0 (3.113-e)
ax ∂αx ay ∂αy
  
1 ∂u w 1 ∂v w
Nx = K − +ν − (3.113-f)
ax ∂αx rx ay ∂αy ry
  
1 ∂v w 1 ∂u w
Ny = K − +ν − (3.113-g)
ay ∂αy ry ax ∂αx rx
 
1 ∂v 1 ∂u 2w
Nxy = Nyx = Gh + − (3.113-h)
ax ∂αx ay ∂αy rxy
2 1 ∂2w
 
1 ∂ w
Mx = −D +ν 2 (3.113-i)
a2x ∂αx2 ay ∂αy2
1 ∂2w 1 ∂2w
 
My = −D +ν 2 (3.113-j)
a2y ∂αy2 ax ∂αx2
∂2w
 
1
Mxy = −Myx = D(1 − ν) (3.113-k)
ax ay ∂αx ∂αy

We observe that we have separate deep-beam and slab equations, paired by the N terms in

Eq. 3.113-c and the w terms in 3.113-f through 3.113-h. Eqs. 3.113-a and 3.113-b are expressions

for a plate loaded in its plane (deep-beam equations). Eqs. 3.113-f through 3.113-h are also deep-

beam equations except for the w terms. Eqs. 3.113-i through 3.113-k are expressions for a laterally

loaded plate (bent slab). Eqs. 3.113-c through 3.113-e are also slab equations except for the N

terms in 3.113-c.
 106

3.4.4.3 Deep Beam Equation

We will now combine the above expressions into a single shallow shell expression. But first,

we must derive the deep beam and slab equations. From Eqs. 3.113-f, 3.113-g and 3.113-h
 
2 1 ∂u w
Nx − νNy = K(1 − ν ) − (3.114-a)
ax ∂αx rx
 
2 1 ∂v w
Ny − νNx = K(1 − ν ) − (3.114-b)
ay ∂αy ry
 
1 ∂v 1 ∂u 2w
2(1 + ν)Nxy = K(1 − ν 2 ) + − (3.114-c)
ax ∂αx ay ∂αy rxy

Differentiated twice and combined, we obtain

1 ∂ 2 (Nx − νNy ) 2(1 + ν) ∂ 2 Nxy 1 ∂ 2 (Ny − νNx )

− +
a2y ∂αy2 ax ay ∂αx ∂αy a2x ∂αx2
(3.115)
1 ∂2 ∂2 1 ∂2
      
2 w 2 w w
= −K(1 − ν ) 2 − + 2
ax ∂αx2 ry ax ay ∂αx ∂αy rxy ay ∂αy2 rx

When rx , ry and rxy are assumed constant, we set

1 ∂2 ∂2 1 ∂2
 
2 2
∆R w = − + w (3.116)
ry a2x ∂αx2 rxy ax ay ∂αx ∂αy rx a2y ∂αy2

Replacing the stress resultants with the following stress functions

1 ∂2F
Z
Nx = 2 − px ax dαx (3.117-a)
ay ∂αy2
1 ∂2F
Z
Ny = 2 − py ay dαy (3.117-b)
ax ∂αx2
1 ∂2F
Nxy = − (3.117-c)
ax ay ∂αx ∂αy

and substituting into Eq. 3.115, we arrive at our differential equation for deep beams

1 ∂ 2 px 1 ∂ 2 py
Z Z
1 ∂px 1 ∂py
∆4 F + K(1 − ν 2 )∆2R w = a x dα x + ay dαy − ν −ν (3.118)
a2y ∂αy2 a2x ∂αx2 ax ∂αx ay ∂αy

where
1 ∂4 ∂4 1 ∂4
 
4 2
∆ F = + + F (3.119)
a4x ∂αx4 a2x a2y ∂αx2 ∂αy2 a4y ∂αy4
If rx = ry = rxy = ∞, ax dαx = dx, and ay dαy = dy, and px = py = 0, Eq. 3.118 reduces to

∆4 F = 0 (3.120)
 107

3.4.4.4 Slab Equation

In deriving the slab equation, we will first substitute the values of Qx and Qy from Eqs.

3.113-d and 3.113-e into Eq. 3.113-c to create the combined equation

1 ∂ 2 Mx 1 ∂ 2 Mxy 1 ∂ 2 My Nx 2Nxy Ny
2 2
−2 + 2 + + + + pz = 0 (3.121)
ax ∂αx ax ay ∂αx ∂αy ay ∂αy2 rx rxy ry

We next substitute Eqs. 3.113-i, 3.113-j and 3.113-k into 3.121 to obtain
 
4 1 Nx 2Nxy Ny pz
−∆ w + + + =− (3.122)
D rx rxy ry D

Once again, we will insert the stress functions of Eqs. 3.117-a, 3.117-b and 3.117-c into our differ-

ential equation 3.122 to generate the bent slab differential equation

 Z Z 
1 2 pz 1 1 1
∆4 w − ∆ F = − px ax dαx + py ay dαy (3.123)
D R D D rx ry

If rx = ry = rxy = ∞, Eq. 3.123 reduces to

pz
∆4 w = (3.124)
D

3.4.4.5 Equation for Shallow Shells

The differential equations of deep beam (3.118) and bent slab (3.123) are required to solve

the problem of shallow shells. The simultaneous differential equations can be rewritten in the form

of

∆4 F + K(1 − ν 2 )∆2R w = f (p) (3.125)

1 2
∆4 w − ∆ F = f 0 (p) (3.126)
D R

where f (p) and f 0 (p) are the right sides of Eqs. 3.118 and 3.123. If we operate on Eq. 3.125 by ∆4

and on Eq. 3.126 by K(1 − ν 2 )∆2R , and combine we obtain

12(1 − ν 2 4
∆8 F + ∆R F = ∆4 f (p) + K(1 − ν 2 )∆2R f 0 p (3.127)
h2
 108

However, if we operate on Eq. 3.125 by (1/D)∆2R and on Eq. 3.126 by ∆4 and combine, we obtain

the solution for shallow shells

12(1 − ν 2 ) 4 1
∆8 w + 2
∆R w = ∆2R f (p) + ∆4 f 0 (p) (3.128)
h D

3.4.5 Membrane Theory

Membrane theory is the second simplified method we will review. However, we will focus

more on the assumptions and theory than the methodology.

3.4.5.1 Assumptions

Membrane theory will utilize the following assumptions:

(1) All bending in the shell is neglected. Therefore Eq. 3.86 becomes

0 a )
∂(Nx0 ay ) ∂a ∂(Nyx
− Ny0 ∂αyx + Nxy
0 ∂ax + x
∂αx ∂αy ∂αy + px ax ay = 0
∂(Ny0 ax ) ∂a 0 a )
∂(Nxy
− Nx0 ∂α
∂ax 0 y
+ Nyx y (3.129)
∂αy y ∂αx + ∂αx + py ax ay = 0
0 0 Ny0
Nx0 Nxy Nyx
rx + rxy + rxy + ry + pz = 0

(2) Since we assume Nxy = Nyx , the three equations above only have three unknowns Nx0 , Ny0
0 = N 0 . The prime marks indicate the values are approximate.
and Nxy yx

(3) All membrane stresses act in the plane of the shell.

3.4.5.2 Theory

The membrane theory can provide a reasonable design basis if the following conditions are

fullfilled:

(1) The displacement due to membrane stress resultants does not increase shell bending sub-

stantially. Previous assumptions state that all bending in the shell is neglected.
 109

(2) A uniform load distribution must be seen over entire shell surface.
32 THIN SHELL CONCRETE STRUCTURES

A very detailed discussion of the limitations on the membrane

(3) The boundaries cancan
theory supply the
be found forces
in ref. permit the displacements required by theexcept
and474.
18, page mem- just at the boundaries
demonstrated for spherical dom
Either a hemispherical roof or a cylindrical tank on free-sliding bases For shells of zero curvature
may be analyzed for uniform loading solely on the basis of the membrane
brane stress resultants. extend further into the shell th
theory. Usually, however, the membrane theory is insufficient. Finally, for shells of negativ
than for the others. Thus the
cant over large portions of the
Usually membrane 1-11 theory is insufficient
CLASSIFICATIONS OF SHELLbutSYSTEMS
may be found useful for hemispherical roofs or useful classificatio
Another
lational systems. Domes and
The most general classification of thin shells is by gaussian curvature, w?ereas cylindrical barrels, elli
cylindrical tanks. illustrated in Fig. 1-12. Shells of positive gaussian curvature, sometimes 101ds are surfaces of translation.
called synclastic shells, are formed by two families of curves both with the ing the analysis and the constru
same direction. Spherical domes and elliptic paraboloids are examples. regard to structural action. F
Shells of zero gaussian curvature or singly curved shells are formed by one spherical domes normally used f
3.4.6 Classification
family of curves only; some examples are cylinders and cones. Shells of low elliptic paraboJoids. In fac
negative gaussian curvature, sometimes called anticlastic, are formed by monly used roof dimensions the
two families of curves each in opposite directions. Hyperbolic parabo- of rot ation and of translation wi
Thin shells can loids
be classified intoof three
and hyperbolas main
revolution categories. The basis of this classification
are examples. comes
of the shell itself (less than 1 in.)
Mathematically the gaussian curvature of a surface is defined as the a sma~l constructional deviation
product of the principal curvatures: ences m geometry between the
from gaussian curvature. Synclastic shells are those of positive gaussian curvature and are formed
K= - -
11
r,. r'll
by two families of curves both with the same direction. Singly curved shells are those1-12
where r,. and r'll are the principal radii of curvature.
of zero
STABILITY

One important characteristic difference among these shell types is

I Much has .been written in recent
gaussian curvature and
the are formedof by
propagation edgeone family
effects into theof curves.
shell. 16 •18 Finally, anticlastic shells are bu~
those
thereof
is as yet not much ex
For shells of positive curvature the edge effects tend to damp rapidly which to develop specific design
andare
are formed
usually confined to afamilies
narrow zone at the edge. Thus in these
. parameters strongly affe
certam
negative curvature and by tow of curves of opposite direction.
shells the membrane theory will often be valid throughout the entire shell page 517) gives the critical or b
2E
q•• = v3(1 - v 2)
(h)2
a
or we may write

Experiments show that C is much

such results given by Schmidt 20
high as 0.32 but none near 2/ y3(1
(a) (b) (c) proposes (l-31) be written as
Fig. 1-12 (a) Shell of positive gaussian curvature. (b) Shell of zero gaussian h2
q.., = CE -
curvature. (c) Shell of negative gaussian curvature. r,.r11

Figure 3.18: (a) Shell of positive gaussian curvature. (b) Shell of zero gaussian curvature. (c) Shell
of negative gaussian curvature, (Billington, 1965)

The main property difference between the three categories is the propagation of edge effects

into the shells. In synclastic shells, the edge effects often damp rapidly and are restricted to a

narrow near edge zone. This property allows for valid use of the membrane theory throughout the

entire shell except at the boundaries. In singly curved shells, the edge effects are damped similar to

the synclastic shell, but propagate farther into the shell away from the edge. In anticlastic shells,

the damping is much smaller than the other two allowing boundary effects to control over majority
 110

of the shell.

Another type of classification defines the shell systems as either rotational or translational.

Domes and tanks are often classified as rotational systems, while cylindrical barrels, elliptical

paraboloids and hyperbolic paraboloids are considered translational.

3.4.7 Conclusion

We have now introduced the derivations of the differential equations of a beam and plate

subject to flexure, along with the discussion of shells. Building up from the basics of a beam, we

progressed to the plate, and finally began the discussion of thin shells.

Now that we fully understand thin shells and have derived some of the basic differential

equations, we can now shift our focus to the actual analysis of a system. In the upcoming section,

we will begin our discussion on circular cylindrical shells, including an in depth discussion of the

analysis and design.

 Chapter 4

Analysis of Circular Cylindrical Shells

4.1 Introduction

This chapter focuses on the analysis of a circular cylindrical shell using both the theories

of shallow shells and membrane theory. The analysis focuses exclusively on a simply supported

shell with edge beams. The stress resultants, stress couples, and displacements are first derived

via membrane theory, and repeated utilizing the theory of shallow shells. Given the results from

both shallow shells and membrane theory, a discussion comparing the two theories and the results

follows.

The chapter continues with the methodology of bending theory. In this section, our eighth

order differential equation is derived and solved to determine all eight arbitrary constants. These

constants can then be substituted into the equations for the stress resultants, couples, and displace-

ments. Given the finalized stresses and displacements, the chapter concludes with the addition of

the edge beam and the derivation of the equation of compatibility required to combine the shell

and beam. Both the vertical and horizontal beam cases are discussed.

4.2 Circular Cylindrical Shell

Circular cylindrical shells or barrel vaults are often ”defined as a curved slab...cut from a full

cylinder” (Billington, 1965). The slab is curved in only one direction and thus may be classified

as a singly curved shell as discussed in the previous chapter. The shell consists of two straight

longitudinal edges and two curved transverse edges. The basic structure and shape of the shell can
 112

be seen in the 156

Figs. 4.1 andSHELL
THIN 4.2. CONCRETE STRUCTURES

Circular
cylindrical
shell

------ y

Figure 4.1: Circular Cylindrical Shell Structure,(Billington, 1965)

Three categories are available for describing cylindrical shells. The class is dependent upon

the ratio of transverse radius and longitudinal length, r/L. The ratio categories can vary depending

upon the source. Common classification consists of long, intermediate and short.

4.3 Simply Supported Shell

The analysis of the circular cylindrical shell will be conducted for the case of a simply sup-

ported shell with all boundary conditions along the straight longitudinal edge satisfied. The pro-

cedure for analysis focuses on four key items and are as follows:
N' + 1m;" de/>
(b) +x aq,
Primary System : This is obtained by reducing the general theory to membrane theory. Through

membrane theory, the surface loads are resisted solely by stress resultants.
N'
•
Errors : These correspond to the incompatible edge effects, or forces required by membrane theory
(a)
at the free edges.
Fig. 5-2
Corrections : Line loads (unit edge effects) are applied along the free edges.
supported shell, in which the boundary conditions along the straight
longitudinal edges are satisfied; and second, for a shell built together
with the transverse frames, in which the boundary conditions along the
curved transverse edges are studied. The first phase has been studied
 ce

Fig. 5-1 re
113
------ y
ed

b
d

n
S
c
d

d
E
d
d
so
N' + 1m;" de/>
(b) +x aq, in

•
N'
o
le
c
(a)

supported shell, in which the boundary conditions along the straight

longitudinal edges
Figure 4.2: are satisfied;
Typical and
Cylindrical second,
Shell for(Billington,
Section, a shell built together
1965)
with the transverse frames, in which the boundary conditions along the
curved transverse edges are studied. The first phase has been studied n
extensively (see Sec. 5-5) and many methods are available for its solution. io

Compatibility : Equilibrium can be achieved by determining the number of corrections required

to remove the errors of membrane theory. In our example, we will set the line loads equal

to the membrane stress resultants along the free edge.

4.3.1 Membrane Theory

Prior to examining a full analysis conducted through shallow shells, we will first examine

the membrane theory. However, we will only focus on the derivations of the stress resultants and

displacements.
 114

4.3.1.1 Stress Resultants

The general stress resultants are derived from Eq. 3.129. The equations reduce to
0
∂Nθ0 ∂Nyθ
∂θ +r ∂y + pθ r = 0
0
∂Ny0 ∂Nθy
r + + py r = 0 (4.1)
∂y ∂θ
Nθ0
r + pz = 0

with the following circular cylindrical shell definitions:

αx = θ ax = r rx = r 0 = N0
Nxy θy
(4.2)
αy = y ay = 1 ry = ∞ Nx0 = Nθ0

where r is a constant. The equations of 4.1 are rewritten to the form of

Nφ0 = −pz r
0 = −1
R ∂Nφ0
(4.3)
R
Nxφ r ∂φ dx − pφ dx + f1 (φ)
0
R ∂Nφx
Nx0 = − 1r
R
∂φ dx − px dx + f2 (φ)

when the following definitions are prescribed:

αx = x ax = 1 rx = ∞ rxy = ∞ 0 = N0
Nxy xφ
(4.4)
αy = φ a y = r ry = r Ny0 = Nφ0 py = pφ

where r is a constant and φ is measured from the longitudinal edge.

The stress resultant expressions are now written in the form of three equations with three

unknowns Nφ0 , Nx0 and Nxφ

0 = N 0 and the primary system is established. It must be noted, as per
φx

the errors stage of the analysis, that longitudinal edge reactions must equal the membrane values

of Nφ0 and Nxφ

0 in order for the membrane theory to produce the correct internal stress resultants.

However, at the free edges of the shell, these reactions are zero.

The required edge reaction forces can be seen in Fig. 4.3(a). In order to account for the lack

of free edge reaction forces, we must apply corrections or line loads, Fig. 4.3(b). In order to find

the true resultant reactions and achieve equilibrium, line loads TL = −Nφ0 and SL = −Nxφ
0 must

be applied.
 The first three terms of a sine series ar
Fourier series: close correspondence to the uniform lo
4 ~ 1 . n'll"X sufficient to use just the first term of th
(TL),.=:;;: TL ~
n=l,3,5 .. .
nsID -y; short shells it is sometimes advisable to

The surface loads must also be expressed by a Fourier series to provide

compatible forces and deformations along the length of the rec edges. 5-3 DISPLACEMENTS FROM THE ME
or example, the Fourier series for the uniform shell dead load is
N' N' N' From (1-4):
4 ~
"'
./>:&
" (a)
(pd),. = :;;: Pd ~ n1 SID
.
(b) r
n'll"X
OU
Fig. 5-4
Fig. 5-5 n=l,3,5 ••. 115 =OXE:z;

The first three terms of a sine series are plotted in Fig. 5-7 and show a E=E= - - - -
av w
close correspondence to the uniform load. For long shells it is usually 1/ "' r aq, r
sufficient to use just the first term of the series (Fig. 5-6b), although for au av
short shells it is sometimes advisable to include the second term as Iwell:
) 'Y:z;.p = r aq, +ax
The strains may be written in terms o
where (Sec. 1-5):
5-3 DISPLACEMENTS FROM THE MEMBRANE THEORY
OU 1 (N' N')
From (1-4): Ex = OX = Eh "' - 11 </>

(a)
~ ~ __!__ (N~ - vN')
(b)
OU f.p = - =
E:z; =OX r aq, r Eh "'
av w _ au + av _ 2(1 + 11) N'
E=E= - - - -
edge reactions in Fig. 5-5. Where the shell has free edges, these reactions
1/ "' r aq, r
(5-8) 'Yz<1> - r aq, ox - Eh "'"'
(a) Required Edge Reactions, (Billing- (b) Applied
are zero. Line
Therefore, auLoad,
correction av (Billington,
forces or line loads TL = 1965)
- N~ and For concrete 11 is usually about l and ha
= r aq, +ax
ton, 1965) SL = - N~,. must be 'Y:z;.p
applied I)
to give the true resultant reactions. The mations. It seems reasonable, therefo
displacements and internal forces due to these line loads are considered
The strains may be written in terms of the membrane stress resultants expressions may be obtained. The fi
in the bending theory
wherein(Sec.
Sec.1-5):
5-4. directly for the longitudinal displaceme
To overcome mathematical difficulties in the bending theory, it is
desirable to express the uniformly
OU
= OX TL
loads. A uniform lineEx load = Eh
distributed
1 (N'
(Fig. "'5-6a)
line loads as sums of partial
N')
- 11 may
</>
be represented by a
u = .ih JN~dx +fa(!/>)
f.p = ~ - ~ = __!__ (N~ - vN') (5-9)
Fig. 5-6 r aq, r Eh "'
_ au + av _ 2(1 + 11) N'
r aq, ox - Eh "'"'
BendingareTheory
edge must
reactions in Fig. bethe considered
5-5. Where shell has free edges, theseto
zero. Therefore, correction forces or line loads TL
determine the displacements
reactions
N~ and
= -
and internal forces pro-
'Yz<1> -

For concrete 11 is usually about l and has only a smalLlnJluence on defor-

SL = - N~,. must be applied to give the true resultant reactions. The mations. It seems reasonable, therefore, to neglect v so that simpler
displacements and internal forces due to these line loads are considered expressions may be obtained. The first of Eqs. (5-9) is then solved
duced from the line loads in the corrections phase of the analysis. For simplicity, uniformly dis-
in the bending theory in Sec. 5-4.
To overcome mathematical difficulties in the bending theory, it is
directly for the longitudinal displacement where
desirable to express the uniformly distributed line loads as sums of partial
loads. A uniform line load TL (Fig. 5-6a) may be represented by a
u = .ih JN~dx +fa(!/>) (5-10)

tributed line loads can be assumed as a sum of partial loads and represented by a Fourier series.
∞
4 X 1 nπx
(TL )x = TL sin (4.5)
π n L
n=1,3,5...

The first three terms of the series are plotted and can be seen in Fig. 4.3. The three waves

together closely assimilate a uniform load. The first term is often considered sufficient for long

shells. However, short shells often require the addition of the second term. This methodology will

be discussed in more detail in Sec. 4.3.2.

I 64 THIN SHELL CONCRETE STRUCTURES 116

4 . 11'%
1i'" smT

(a)

f
..!..
37r
(b) .__,___.__.__.___.__,~---r---,---..--...,.........,,1--J~---L.....I.,_~_l_

(d)

(e)

Figure 4.3: Sine Series Plot, (Billington, 1965)

 117

4.3.1.2 Displacements

The derivation of the displacement equations from membrane theory requires recalling the

middle surface strains from Sec. 3.4.2.2 (Eqs. 3.87, 3.88 and 3.89). Utilizing the circular cylindrical

shell definitions (4.4), the strains can be rewritten as

∂u
x = ∂x
∂v w (4.6)
y = φ = r∂φ − r
∂v ∂u
γxφ = ∂x + r∂φ

The strains may be rewritten once more to accommodate membrane stress resultants
 
x = ∂u
∂x = 1
Eh Nx0 − νNφ0
 
y = φ = ∂v
r∂φ − wr = Eh
1
Nφ0 − νNx0 (4.7)
∂v ∂u 2(1+ν) 0
γxφ = ∂x + r∂φ = Eh Nxφ

Since ν has minimal impact on the deformations, it is neglected to simplify the displacement

equations. The first term of Eq. 4.7 is solved for the longitudinal displacement
Z
1
u= Nx0 dx + f3 (φ) (4.8)
Eh

The third term is solved for the tangential displacement equation

Z Z
1 ∂u 2 0
v=− dx + Nxφ dx + f4 (φ) (4.9)
r ∂φ Eh

and the radial displacement is obtained from the second

∂v rNφ0
w= − (4.10)
∂φ Eh

In order to determine the vertical and horizontal displacements, the following relations must

be considered.

∆V = −v sin(φk − φ) + w cos(φk − φ)
(4.11)
∆H = v cos(φk − φ) + w sin(φk − φ)
where ∆V can be written in the form of
 118

( " 2 #)
L4 2r 2  r 4
∆V = 3 pr + 4+ cos2 (φk − φ) sin kx (4.12)
r hE πL π L

and ∆H is
pr2
∆H = sin(φk − φ) cos(φk − φ) sin kx (4.13)
Eh

where ∆V is positive for downward displacement and ∆H is positive for inward displacement.

4.3.2 Theory of Shallow Shells

Before performing the analysis through the theory of shallow shells, we must first rederive

the equations of the stress resultants and couples along with the equations of equilibrium to accom-

modate the properties of the circular cylindrical shell. The equations of equilibrium (3.86) were

previously derived for shallow shell theory in Sec. 3.4.4. Utilizing the definition for a cylindrical

shell (4.4), the equations of equilibrium become

∂Nx ∂φx
∂x r + ∂φ + px r = 0
∂Nφ ∂Nxφ
∂φ + ∂x r − Qφ + p φ r = 0
∂Qx ∂Qφ (4.14)
∂x r + ∂φ + Nφ + px r = 0
∂Mφ ∂Mxφ
− ∂φ + ∂x r + Qφ r = 0
∂Mφx
− ∂M
∂x r −
x
∂φ + Qx r = 0

The stress resultants and stress couples, Eq. 3.111, were also previously derived in Sec. 3.4.4.

Once again, the equations are modified to account for the cylindrical shell properties

Nx = Eh ∂u
∂x
 
∂v w
Nφ = Eh r∂φ − r
 
Eh ∂v ∂u
Nxφ = Nφx = 2 ∂x + r∂φ
(4.15)
3 ∂2w
Mx = − Eh
12 ∂x2
3
 
∂2w
Mφ = − Eh
12
∂v
r2 ∂φ
+ r2 ∂φ2
 
Eh3 ∂v ∂2w
Mxφ = −Mφx = 12 2r∂x + r∂x∂φ
 119

where
∂w v ∂w
φx = ∂x φy = r + r∂φ ν=0

Now that we have rederived all 11 equations, we can begin the derivation of the single eighth-

order partial differential equation using the theory of shallow shells. We will ultimately reduce the

11 equations to one equation with one unknown.

We will first recall the assumptions for shallow shells presented in Sec. 3.4.4.1. All stress

couples must be expressed in terms of the radial displacement. This is achieved by neglecting the

terms with v in Eq. 4.15

3 ∂2w
Mx = − Eh
12 ∂x2
3
 
∂2w
Mφ = − Eh
12 r2 ∂φ2
(4.16)
 
Eh3 ∂2w
Mxφ = −Mφx = 12 r∂x∂φ

We will next rewrite the radial shear stress resultants from Eq. 4.14 in terms of w

3
 
∂3w ∂3w
Qφ = − Eh
12 r3 ∂φ3
+ r∂x2 ∂φ
  (4.17)
3 ∂3w ∂3w
Qx = − Eh
12 ∂x3
+ r2 ∂x∂φ2

where h and r are constants. All other stress resultants can now be written in terms of w by

substituting Eq. 4.17 into the third equation of 4.14, obtaining

Eh3
 4
∂ w 2 ∂4w 1 ∂4w

Nφ = r 4 + + − pz r (4.18)
12 ∂x r ∂x2 ∂φ2 r3 ∂φ4

In turn, we substitute Nφ into the second equation of 4.14. If we drop Qφ , due to the assumption

2, we obtain
Eh3 ∂5w 2 ∂5w 1 ∂5w
 
∂Nxφ ∂pz
=− + + + − pφ (4.19)
∂x 12 ∂x4 ∂φ r2 ∂x2 ∂φ3 r4 ∂φ5 ∂φ
Finally, the first equation of 4.14 is differentiated with respect to x

∂ 2 Nx 1 ∂ 2 Nφx ∂px
= − − (4.20)
∂x2 r ∂φ∂x ∂x

By differentiating Eq. 4.19 with respect to φ and substituting the new derived term

∂ 2 Nφx
(4.21)
∂φ∂x
 120

into Eq. 4.20, we obtain

∂ 2 Nx Eh3 1 ∂ 6 w 2 ∂6w 1 ∂6w ∂ 2 pz

 
∂pφ ∂px
2
= 4 2
+ 3 2 4
+ 5 6
− 2
+ − (4.22)
∂x 12 r ∂x ∂φ r ∂x ∂φ r ∂φ r∂φ r∂φ ∂x

Now that we have rederived our stress couples and stress resultants in terms of radial dis-

placements, we begin the formation our compatibility equation with the remaining stress resultants

of Eq. 4.15. The first three equations are rearranged to the form of

∂u 1
∂x = Eh Nx
∂v r (4.23)
∂φ = Eh Nφ +w
 
Eh ∂v 1 ∂u
Nxφ = 2 ∂x + r ∂φ

Once again, we differentiate to obtain common terms

∂5u 1 ∂ 4 Nx
∂x3 ∂φ2
= Eh ∂x2 ∂φ2
4
∂5v r ∂ Nφ ∂4w (4.24)
∂x4 ∂φ
= Eh ∂x4 + ∂x4
∂ 4 Nxφ
 
Eh ∂5v 1 ∂5u
∂x3 ∂φ
= 2 ∂x4 ∂φ
+ r ∂x3 ∂φ2

and substitute to obtain one equation of compatibility

∂ 4 Nxφ
 4
∂ Nφ 1 ∂ 4 Nx Eh ∂ 4 w

1
3
= r + + (4.25)
∂x ∂φ 2 ∂x4 r ∂x2 ∂φ2 2 ∂x4

The stress resultants derived in Eqs. 4.18, 4.19 and 4.22 are now substituted into the newly

derived compatibility equation (Eq. 4.25).

r2 ∂ 8 w ∂8w 1 ∂8w 1 ∂8w 1 ∂8w

+ + + +
2 ∂x8 ∂x6 ∂φ2 2r2 ∂x4 ∂φ4 2r2 ∂x4 ∂φ4 r4 ∂x2 ∂φ6
8
1 ∂ w ∂8w 2 ∂8w 1 ∂8w
+ 6 + + + (4.26)
2r ∂φ8 ∂x6 ∂φ2 r2 ∂x4 ∂φ4 r4 ∂x2 ∂φ6
6 ∂4w 1 ∂ 3 pφ ∂ 3 pφ
 2 4
1 ∂ 4 pz 1 ∂ 3 px ∂ 4 pz

12 r ∂ pz
+ 2 = + 2 − 2 + + −
h ∂x4 Eh3 2 ∂x4 2r ∂φ4 2r ∂φ3 2r ∂x∂φ2 ∂x2 ∂φ2 ∂x2 ∂φ

After simplification, we arrive at the eighth order partial differential equation.

 121

4 "  2 #
3p 3p
∂2 ∂2 12 ∂ 4 w 2 2 3p

12 ∂ ∂ ∂ φ 1 ∂ φ 1 ∂ x
r2 + w+ 2 = r2 + pz − 2 2 − +
∂x2 r2 ∂φ2 h ∂x4 Eh3 ∂x2 r2 ∂φ2 ∂x ∂φ r2 ∂φ3 r ∂x∂φ2
(4.27)

If we recall our previously derived equation for shallow shells, Eq. 3.128, our new equation

is of the same form when ν equals zero.

12(1 − ν 2 ) 4 1
∆8 w + 2
∆R w = ∆2R f (p) + ∆4 f 0 (p) (4.28)
h D

where, for circular cylindrical shells,

∂4w
∆4R w = r2 ∂x4
 4
∂2 ∂2
∆8 w = ∂x2
+ r2 ∂φ2
w
R ∂ 2 pφ (4.29)
1
R ∂ 2 px
f (p) = r2 ∂φ2
dx +r ∂x2
dφ
pz
f 0 (p) = 1
R
D − D pφ dφ

4.3.3 Shallow Shells vs. Membrane Theory

Now that we have reviewed the analysis methods of the theory of shallow shells and membrane

theory, we will briefly discuss the difference between the two methodologies and the results.

4.3.3.1 Methodology

Membrane theory is an approximate method that is often practiced to avoid the more rigorous

approach of shallow shells. Membrane theory allows for the computation of the stress resultants and

displacements without the use of bending, since all bending in the shell is neglected. The theory

also assumes that all membrane stresses act in the plane of the shell. This is similar to the theory

of shallow shells which requires the surface loads to be carried by the in-plane stresses. However,

contrary to membrane theory, the theory of shallow shells accounts for bending within the shell.

The theory assumes that the shell’s surface curvature, along with the change in curvature, is small

but is still required for analysis purposes.

 122

4.3.3.2 Results

We will now view two examples, first with membrane theory, and second with the theory of

shallow shells, to compare the results of both methods.

Membrane Theory : Lets first consider the stress resultants derived from membrane theory

given a uniform shell load p = (4/π)pd . The loading components are as follows:

pz = p cos(φk − φ) sin kx pφ = p sin(φk − φ) sin kx px = 0

where k = nπ/L and n = 1. Substituting into Eq. 4.3 we obtain the stress resultant

Nφ0 = −pr cos(φk − φ) sin kx (4.30)

When we substitute Nφ0 into the second term of Eq. 4.3 we obtain

∂[−pr cos(φk − φ) sin kx]

Z Z
0 1
Nxφ =− dx + p sin(φk − φ) sin kxdx + f1 (φ) (4.31)
r ∂φ

0 becomes
After differentiation and integration, Nxφ

0 p
Nxφ = 2 sin(φk − φ) cos kx + f1 (φ) (4.32)
k

where for a simply supported shell f1 (φ) = 0.

We will next substitute the newly derived term into the third term of Eq. 4.3. Again

through differentiation and integration we obtain

p
Nx0 = −2 cos(φk − φ) sin kx + f2 (φ) (4.33)
k2 r

where for a simply supported beam Nx0 = 0 at x = 0 and x = L. f2 (φ) can be computed

as zero and dropped from the above equation.

Next we will derive the displacements from membrane theory using the above derived stress

resultants and Eqs. 4.8, 4.10, and 4.9.

 123

After substitution, we find

2 p
u= EH k3 rcos(φk − φ) cos kx
2 p (4.34)
1 + 2k 2 r2 sin(φk − φ) sin kx


v = EH k4 r2
 
2 pk4 2 r 2 + r4 k4 cos(φ − φ) sin kx
w = EH r 2 1 + 2k 2 k

Theory of Shallow Shells : The loading components will be the same as those used for mem-

brane theory.

pz = p cos(φk − φ) sin kx pφ = p sin(φk − φ) sin kx px = 0

Using the right side of the partial differential equation (Eq. 4.27), we determine the par-

ticular solution.

12
r2 k 4 p cos(φk − φ) sin kx + 2p(− cos(φk − φ))(− sin kx)k 2
Eh3
1
+ 2 p sin kx cos(φk − φ) − 2k 2 p cos φk (− sin kx)
r   
1 12 2 4 2 2
− 2 p sin kx(− cos(φk − φ)) r k + 4k + 2 p cos(φk − φ) sin kx
r Eh3 r
(4.35)

The particular solution is determined as

w = Cp cos(φk − φ) sin kx (4.36)

When substituted into the left side of the partial differential equation, we find
 2 4 4
∂2 k −φ) sin kx]
∂
r2 ∂x 2 + r 2 ∂φ2 Cp cos(φk − φ) sin kx + h122 ∂ [Cp cos(φ
∂x4

4 (4.37)
r2 k 2 + r12 Cp cos(φk − φ) sin kx + h122 k 4 Cp cos(φk − φ) sin kx

Equating the left and right hand sides of the equation, we solve for the constant C of our

particular solution
r2 k 4 + 4k 2 + 2/r2
 
1
C= (4.38)
Eh (r2 h2 /12)(k 2 + 1/r2 )4 + k 4
Substituting into Eq. 4.36, we compute w at any point in terms of the load and dimen-

sions of the shell. The stress couples, stress resultants and displacement are obtained by

substituting w into Eq. 4.16 through Eq. 4.24.

 124

4.3.3.3 Comparison

It is helpful to compare the results of the membrane theory to that of shallow shells. The

results are presented in the table below.

Membrane Theory Theory of Shallow Shells

Stress Couples
0 3
Mφ =0 Mφ = Eh 2 pC cos(φk ) sin kx
12r
0 3
Mxφ =0 Mxφ = Eh
12r
kpC sin(φk ) cos kx
0
Mx =0 Mx = Eh3 k2 pC cos(φk ) sin kx
12

Stress Resultants
0
Nφ = −pr cos(φk ) sin kx Nφ = −pr(1 − 2P ) cos(φk ) sin kx
0 p p
Nxφ = 2k sin(φk ) cos kx Nxφ = 2 k (1 − P ) sin(φk ) cos kx
0
Nx = − 2p cos(φk ) sin kx Nx = − 2p (1 − P ) cos(φk ) sin kx
k2 r k2 r
Displacements

u = Eh2 p
cos(φk ) cos kx 2
u = Eh p
(1 − P ) cos(φk ) cos kx
k3 r k3 r
2 2
2
v = Eh p
4 r2 (1 + 2k r ) sin(φk ) sin kx 2
v = Eh p
(1 + 2k2 r 2 )(1 − P ) sin(φk ) sin kx
k k4 r 2
4 4 r 4 k4 (1 − 2P ) cos(φ ) sin kx
  
2 p 2 2 2 p 2 2
w = Eh 4 2 1 + 2k r + r 2k cos(φk ) sin kx w = Eh 4 2 (1 + 2k r )(1 − P ) − 2 k
k r k r

Table 4.1: Comparison of Analysis Results

Note that the displacement computed via the particular solution can be rewritten as

u = u0 (1 − P )

v = v 0 (1 − P ) (4.39)

w = w0 B+1
1

where
Eh3
P = 24r4
(1 + 2k 2 r2 + k 4 r4 )C
(4.40)
h2
B= 12r6 k4
(k 2 r2 + 1)4
and from the comparison table, φk = (φk − φ). P and B are used to measure the accuracy of the

membrane approach compared to the shallow shell method.

Let us consider the following variables:

h 1
h = 3 in = 0.25 ft r = 106
r
r = 26.6 ft L = 0.4
π
L = 66.5 ft k= L

φk = 45 deg E = 4.32x105 ksf

 125

The long shell example allows the use of membrane theory in comparison to the theory of shallow

shells based on the following values:

C = 0.0284 ft3 /kip

P = 0.00011 (4.41)

B = 0.000131

Now lets consider the case of a short shell where

h 1
h = 4 in = 0.33 ft r = 420
r
r = 140 ft L = 4.67

L = 30 ft φk = 45 deg

E = 4.32x105 ksf

The results of the short shell case are still reasonable and would allow the use of membrane theory.

C = 0.135 ft3 /kip

P = 0.011 (4.42)

B = 0.024

Even though membrane theory may seem like the easy choice to avoid the rigorous eighth order

differential equation, it can only provide reasonable results if it meets certain criteria. First, the dis-

placement results must not increase the shell bending substantially since the methodology neglects

all shell bending. Next, the loading on the shell must be uniform over the entire surface. Finally,

the boundaries must supply the forces and permit the displacements required by the theory. Often

membrane theory is not permitted for design use, but in the case of a circular cylindrical shall, it

appears useful to accelerate the design process.

4.3.4 Bending Theory

After determination of the primary system and errors by either membrane theory or the

theory of shallow shells, we continue the analysis by determining the displacements and internal

forces produced from the line loads in the corrections phase of the analysis. We must first determine
 126

the homogeneous solution to our general formulation. In this case, we will view the eighth order

differential equation derived from the theory of shallow shells (4.27).

From the left side of the equation, the homogeneous equation is written as
4
∂2 ∂2 12 ∂ 4 w

r2 + w+ =0 (4.43)
∂x2 r2 ∂φ2 h2 ∂x4

The solution has the form of

∞
X
w= Am expM φ sin kx (4.44)
n=1,3,...

where Am and M represent eight arbitrary constants and eight roots. The constants are based

upon the longitudinal boundary conditions while the roots are based solely on the dimensions of

the shell.

The roots are computed by substituting Eq. 4.44 into Eq. 4.43

2 4
 
r2 Am expM φ sin kx −k 2 + Mr 2 + h122 k 4 Am expM φ sin kx = 0

4 (4.45)
6 4
M 2 − k 2 r2 Am expM φ sin kx + 12rh2k = 0

In substituting
 r 2
Q8 = 3(kr)4 (4.46)
h

we obtain

4
M 2 − (kr)2 Am expM φ sin kx + 4Q8 Am expM φ sin kx = 0 (4.47)

which can be rewritten as

4
M 2 − (kr)2

√ +1=0 (4.48)
2Q2
With one additional substitution,
 2
kr
γ= (4.49)
Q

Eq. 4.48 becomes

M2 √
4
√
− γ = −1 2 (4.50)
Q2
√
We are now left with the imaginary term 4 −1 that is rewritten as

√
4
√
−1 = ±i (4.51)
 127
√
where i = −1. A complex number is written as

z = x + iy = Reiθ (4.52)
p
When x = 0, y = 1, and r = x2 + y 2 = 1 the complex number becomes

i = eiθ (4.53)

where, from Euler’s equation,

eiθ = cos θ + i sin θ (4.54)

and θ = π/2 forming the complex relation of

i = eiπ/2 (4.55)

Now that we have derived the complex relation, we compute

√ 1
i = (i) 2 = eiπ/4

eiπ/4 = cos π/4 + i sin π/4 (4.56)

eiπ/4 = √1 + i √12
2

If we substitute Eq. 4.56 into Eq. 4.50, we find our eight roots

M1 = ±(α1 ± iβ1 )
(4.57)
M2 = ±(α10 ± iβ10 )

where q√
(1+γ)2 +1+(1+γ)
α1 = Q 2 = Qm1
q√
2
(1+γ) +1−(1+γ)
β1 = Q 2 = Qn1
q√ (4.58)
2
(1−γ) +1−(1−γ)
α10 = Q 2 = Qm2
q√
2
(1−γ) +1+(1−γ)
β10 = Q 2 = Qn2
We next determine the eight arbitrary constants, Am , by expanding upon Eq. 4.44.
 
(α1 +iβ1 )φ A e(α1 −iβ1 )φ
 A1 e 2 
 
0 0 0 0
 A3 e(α1 +iβ1 )φ A4 e(α1 −iβ1 )φ
 

w= 
 sin kx
 (4.59)
 A5 e−(α1 +iβ1 )φ A6 e−(α1 −iβ1 )φ 
 
 
0 0 0 0
A7 e−(α1 +iβ1 )φ A8 e−(α1 −iβ1 )φ
 128

As noted before from Euler’s equation

e±iβ1 φ = cos β1 φ ± i sin β1 φ (4.60)

allowing the first row of our matrix to be rewritten as

[(A1 + A2 ) cos β1 φ + i(A1 − A2 ) sin β1 φ] expα1 φ (4.61)

where A1 and A2 must be complex conjugate constants in order for w to be real.

A1 + A2 = 2a A1 − A2 = i2b (4.62)

and rewritten in the form of

A1 = a + ib A2 = a − ib (4.63)

We now substitute the constants back into Eq. 4.61 to simplify.

2(a cos β1 φ − b sin β1 φ) expα1 φ (4.64)

When we compare the above expression with the first term of Eq. 4.59

A1 exp(α1 +iβ1 )φ = [a cos β1 φ − b sin β1 φ + i(b cos β1 φ + a sin β1 φ)] expα1 φ (4.65)

we notice it is twice the real part, 2Re A1 expM1 φ , meaning

 
 A1 expM1 φ
+ A3 expM2 φ
w = 2Re   sin kx (4.66)

A5 exp −M 1 φ + A7 exp −M2 φ

Thus, the partial derivatives of w is easily attained and applied to Eqs. 4.16, 4.17, 4.18, 4.19 and

4.20 to obtain the stress resultants and couples in terms of Am . As an example, we will view the

term Mφ .

Eh ∂ w3 2
Mφ = − 12r 2 ∂φ2
 
A1 M12 eM1 φ + A3 M22 eM2 φ
Eh3
Mφ = − 12r 2 ∗ 2R   sin kx
 
A5 M12 e−M1 φ + A7 M22 e−M2 φ
 129

The Am terms are rewritten in the form of the eight arbitrary constants as in Eq. 4.63

A1 = a + ib A3 = c + id
(4.67)
A5 = e + if A7 = g + ih

and Mφ is rewritten as
 
 (a + ib)M12 (cos β1 φ
+ i sin β1 φ)eα1 φ 
 
0
Eh3  (c + id)M22 (cos β10 φ + i sin β10 φ)eα1 φ
 

Mφ = − 2 R 
  sin kx (4.68)
6r  (e + if )M 2 (cos β1 φ − i sin β1 φ)e−α1 φ


 1 
 
0
(g + ih)M22 (cos β10 φ − i sin β10 φ)e−α1 φ

After expansion, substitution of Mn and dropping the imaginary terms, we attain the resulting

stress couple in terms of eight unknowns.

 
h[a(1 + γ) − b] cos β φ − [a + b(1 + γ)] sin β φi e α1 φ
 1 1 
 
0
0 φ − [c + d(γ − 1)] sin β 0 φi eα1 φ 
3 h[c(γ − 1) − d] cos β

Eh 1 1
Mφ = − 2 Q2 
 
 sin kx (4.69)
6r 
 h[e(1 + γ)−] cos β1 φ + [e + f (1 + γ)] sin β1 φi e−α1 φ 

 
 
h[g(γ − 1) − h] cos β10 φ − [g + h(γ − 1)] sin β10 φi e−α1 φ

In the case of a symmetrical shell, the unknowns are reduced to four, given a = e, b = f ,

c = g, and d = h. Next we replace the exponential terms with their hyperbolic counterpart

expαφ = cosh αφ + sinh αφ

(4.70)
exp−αφ = cosh αφ + sinh αφ

to obtain the stress resultant and couple in its final form.

 
Eh3 2  [a(1 + γ) − b] cos β1 φ cosh α1 φ − [a + b(1 + γ)] sin β1 φ sinh α1 φ
 

Mφ = − 2 Q sinkx
3r  [c(γ − 1) − d] cos β 0 φ cosh α0 φ − [c + d(γ − 1)] sin β 0 φ sinh α0 φ
 

1 1 1 1
(4.71)

This process is repeated to solve all eight functions, Mφ , Mx , Mxφ , Qx , Qφ , Nφ , Nx , and

Nxφ .
 130

The next step is to solve for the edge effects due to line loads. The effects can be obtained

by setting three boundary conditions equal to zero and the fourth equal to one. We then have four

equation and four unknowns which are solved simultaneously for the four arbitrary constants a, b, c

and d. For the unsymmetrical cases, we solve eight simultaneous equations for eight unknowns. In

order to solve for the displacements, the newly solved arbitrary constants are substituted directly

into u, v, and w.

4.3.5 Edge Beams

The purpose of the edge beam is to stiffen the shell edge and, along with the shell, carry

flexural stresses. Two types of edge beams will be reviewed in the analysis of the shell, vertical and

horizontal edge beams, seen in Figs. 4.4 and 4.5. Vertical beams see extensive use in long shells

where longitudinal bending controls the design. Whereas horizontal beams are used primarily in

short shells where transverse arching controls.

ANALYSI$ OP CIRCULAR CYLINDRICAL SHELLS 183

shown here
ich indicate

points out
rrors which
ples may be
, the errors
alues given

hallow-shell
ntermediate Figure 4.4: Shell with Vertical Edge Beams, (Billington, 1965)
land or the
g shells can

ct
mally there I
beams (Fig.
where the
ntal beams
 out
hich (a) 131
y be
rors
ven

hell
iate
the
can

ct
here I
Fig.
the
ams
tion

the
dge
hell
ver-
(d~)
son-
ntal
9a). Figure 4.5: Shell with Horizontal Edge Beams, (Billington, 1965)
aint
rain
The analysis for both beams follows the typical four steps:

Primary System : The surface loads are supported by a shell containing free edges. Also at the

free edges, edge members consisting of simple beams carrying their own dead weight are

present.

Errors : Two errors are presented by the primary system of the shell and edge members. First,
S +D B .
the difference in the vertical deflections of the shell free edge and the edge beam, D10 10

Next, the difference in the longitudinal stresses in the shell edge and the top of the edge
S + fB.
beam, f20 20

Corrections : A vertical force, Vb , and shearing force, Sb are applied at the edge beam as seen

in Fig. 4.6. Vb acts upward on the shell and downward on the beam reducing the vertical
 132

deflection and longitudinal stresses in the shell. It is assumed positive when acting down-

ward on the shell and upward on the beam. Sb acts inward on the shell and outward on the

beam reducing the vertical deflection and longitudinal stresses in the shell. It is assumed

positive when acting outward on the shell and inward on the beam.

Compatibility : Two compatibility equations are applied. Vertical displacement is restored by

Eq. 4.72
S B S B S B



X1 D11 + D11 + X2 D12 + D12 + D10 + D10 =0 (4.72)

Longitudinal stresses are restored by Eq. 4.73

S B S B S B



X1 f21 + f21 + X2 f22 + f22 + f20 + f20 =0 (4.73)

184 THIN SHELL CONCRETE STRUCTURES

. 11'%
v.b siny

Figure 4.6: Corrective Line Loads Applied at Edge Members, (Billington, 1965)
The analysis is divided into the usual four steps:

4.3.5.1
Primary System. The shell, supporting surfac
Vertical Edge Beams
edges; and the edge member as a simple beam carr
Let us first review the vertical edge beam. For free edge analysis, discussed previously,
weight.
corrective line loads T and S are applied so that the free edge forces equal zero. The purpose
L
Errors. £ rin1ary-system errors are : (1) the diffe
L

deflections of the shell edge and the edge beam, D 108 +

of the vertical edge beam is to resist vertical movement and longitudinal edge strain (N /Eh). If
x

difference in the longitudinal stresses in the shell edge

edge beam, ! 20 8 + /2 08 •
 133

the beam is slender, it is assumed that negligible resistance to rotation and horizontal translation

is offered. This means that the edge forces are no longer zero and unknown edge forces of Vb and

Sb must be applied. The determination of these values depends on the properties of the edge beam

and its capacity to resist vertical deflection and longitudinal edge strain.

The values of the shell displacements and stresses were discussed previously in Sec. 4.3 via

membrane theory and th theory of shallow shells. The values of edge beam displacements and

stresses are computed by the ordinary flexural theory as follows.

4.3.5.2 Ordinary Flexural Theory

Since the shell line loads are represented by Fourier series, the same representation must

be required for the edge beam loading in order to achieve compatibility with the shell over the

longitudinal length’s entirety. The ordinary flexural theory utilizes the following relationships:

4
d y
EI dx4 = wx = loading
3
d y
EI dx3 = −Sx = shear
2
d y
EI dx2 = −Mx = moment (4.74)
dy R L Mx
dx = − 0 EI dx + f1 (x) = θ = slope
RLR x RL
y = − 0 M EI dx + 0 f1 (x)dx + f2 (x) = deflection

The shear, moment, slope, and deflection are all solved by a known loading case and basic inte-

gration. Since the load on the beam is assumed as Vb sin nπx

L the loading equation takes the form

of

∂4δ Vb nπx
= sin (4.75)
∂x4 EI L

Integrating the previous, we derive the moment equation

2
∂2δ

Vb L nπx
M= 2
=− sin (4.76)
∂x EI nπ L
 134

and the displacement

 4
Vb L nπx
δ= sin (4.77)
EI nπ L

The longitudinal stress at the top of the beam are computed as

M
ft = (4.78)
Zt

With a known shearing force, Sb cos nπx

L , a resultant thrust at the edge of the beam is produced

at any distance x from the beam support.

Z x
nπx L nπx
T =− Sb cos dx = −Sb sin (4.79)
0 L nπ L

A bending moment T e and a normal load T result from the thrust. The vertical deflection is

computed from Eq. 4.74, with f1 (x) = f2 (x) = 0 and the longitudinal stress at the top of the beam

is recomputed as
T Te
ft = + (4.80)
A Zt

4.3.5.3 Matrix Analysis

For a rectangular cross section, the previously discussed equations can be written in matrix

form. When I = bd3 /12, Z = bd2 /6 and e = d/2, the equations are summarized as seen below.

For Vb = 1:
B = L4 12
D11 Ebd3 π 4 n4
sin kx
(4.81)
B L2 6
f21 = − bd 2 π 2 n2 sin kx

For Sb = 1:
3
B =− L
D12 6
sin kx
Ebd2 π 3 n3
(4.82)
B = L 4
f22 bd πn sin kx

The beam displacement and stress due to full loads on the beam alone can be found from

Eqs. 4.81 and 4.82. Utilizing Fourier series representation for the vertical uniform load
∞
4 X 1
(wB+L )x = − wB+L sin kx (4.83)
π n
n=1,3,5...
 135

we derive
4
B = −4w
D10 L 0.12319
sin kx
π B+L Ebd3 n4
(4.84)
2
B = 4w
f20 L 0.60793
sin kx
π B+L bd2 n2

4.3.5.4 Horizontal Edge Beams

The process for analyzing a horizontal edge beam is nearly identical to that of the vertical

beam. Hb replaces the Vb values in Eq. 4.74 and the matrix equations. The d and b terms of Eqs.

4.81 and 4.82 are reversed representing the beam width and depth. Similar to the vertical beam,

it is assumed to have negligible rotational and transverse stiffness.

The primary difference is found in the addition of a vertical load VB+L and a bending moment

MB+L at the shell edge.

VB+L = wB+L d
(4.85)
2
MB+L = wB+L d2

In addition, the horizontal edge beam is assumed to carry its own load as a cantilever of span

d, and not a beam of length L.

4.3.6 Prestressing

The purpose of prestressing within the edge beam is to allow a reduction in the size of the

beam. We will once again follow our four step analysis procedure and is as follows:

Primary System : The primary system consists of two separate items. The first is the shell

carrying no load with free edges. The second is the edge beam analyzed as a simple beam

subject to the prestressing load.

Errors : Two errors are presented by the primary systems of the shell and edge members. First,

the difference in the vertical deflections of the shell free edge and the beam. Second, the

difference in the longitudinal stresses in the shell edge and the top of the edge beam. For

both cases, zero deflection will occur at the shell edge since the shell structure is unloaded.
 136

Corrections : A vertical force, Vb , and shearing force, Sb are applied at the edge beam as seen

in Sec. 4.3.5. Vb acts upward on the shell and downward on the beam reducing the

vertical deflection and longitudinal stresses in the shell. It is assumed positive when acting

downward on the shell and upward on the beam. Sb acts inward on the shell and outward

on the beam reducing the vertical deflection and longitudinal stresses in the shell. It is

assumed positive when acting outward on the shell and inward on the beam.

Compatibility : The compatibility equations are the same as those given for Sec. 4.3.5 with the

exception of the error terms derived in Eq. 4.84.

The new error terms are derived from Eqs. 4.81 and 4.82. If the edge beam is considered as a

simple beam, a parabolic reinforcing profile causes a vertical upward deflection and a longitudinal

stress at the top of the edge beam.

B =−
RLR MF
RL
D10 0 EI dxdx + 0 f1 (x)dx + f2 (x)
(4.86)
B = F MF
f20 A − Zt

where MF is the bending moment due to eccentricity of the tendon profile.

The deflections and stresses caused by this bending moment are computed throughout the

member except at the supports were the bending moment is assumed zero. The equivalent uniform

load is then used to compute the deflections and stresses.

8F ec
we = (4.87)
L2

The loading is once again represented by the Fourier series:

∞
4 X 1
(we )x = we sin kx (4.88)
π n
n=1,3,5...

If the edge beam is of a rectangular cross section, Eq. 4.86 is rewritten in the form of

B = 4 8F ec L4 0.12319
D10 π L2 Ebd3 n4
sin kx
(4.89)
B = 4 F 4 8F ec L2
f20 π bd sin kx − π L2 bd2 (0.60793) sin kx
 137

where the first term F/A uses on the first term of Fourier series.

Note that the values of VbF and SbF can by determined be inputting the values of Eq. 4.89 in

the compatibility equations (4.72) and (4.73).

4.4 Conclusion

In this chapter, we focused on the analysis of the shell structure, first through membrane

theory, and second through the theory of shallow shells. We concluded by viewing the analysis of

the edge beams and the required reinforcement.

 Chapter 5

Circular Cylindrical Shell Design Tool

5.1 Introduction

Chapter five discusses the development of a design tool to aid in future cylindrical shell

design along with a brief preliminary design. The code utilizes the theory of shallow shells in the

determination of the stresses and displacements and is presented in the appendix of the thesis

(A). A discussion of the code and results is presented followed by a reinforcement section and

the equations for required steel within the shell and beam. An example structure is presented for

validation of the code.

5.1.1 Preliminary Design Example

The structure that will be utilized for this example is a warehouse complex located in Penn-

sylvania. For purposes of this design, the structure will used to validate the results since the

structure has been previously studied using the theory of shallow shells. We will design a single

barrel from the warehouse assuming restraint from the longitudinal edge beam.
 ness of 3 in. is chosen for practical considerations. A minimum of three
layers of reinforcement is normally used: top and bottom wire fabric with
main bars in between. The cover over the fabric should be at least

139

fig. 6-4 (Courtesy of Roberts & Schaefer Co., Inc.)

j
Figure 5.1: Pennsylvania Warehouse Complex, (Billington, 1965)

208 THIN SHELL CONCRETE STRUCTURES

--1 whic
20' the b

esse
Cross section
Whe
orou

Pl
I
200'
t:: -- ~
_,_

I
-

l/
II
v
(ii] C-J
I
I
l
I
I
l
I
6-3

Figu
cove

L
'-- :.= = .=- .::: ·-·- u (•J ~J [•J
I/
I
I sylv
I
I of 24
I/ I
I
I
free
~- ~-- - -a------{~---~-- -" --- tudin
the a
Plan view of typical unit

the o
orien
mov
Figure 5.2: Pennsylvania Warehouse Roof Layout, (Billington, 1965) ness
layer
main
The structure of interest can be seen in Fig. 5.1. The warehouse roof consists of 15 barrels

1'3"
t
3•
I
3• 4·0· I I5'8'I 1· 4•0·
 I

L
'-- :.= = .=- .::: ·-·- u (•J ~J [•J
I
I/ I
I
I
I/ I 140
I
I
covering 40,000 ft2 (see Fig. 5.2). The complex~-- - -a------{~---~-- -" ---
~-
totals 16 separate units (240 barrels)covering

640,000 ft2 . Plan view of typical unit

1'3"
t
3•
I
3• 4·0· I I5'8'I 1· 4•0·
7'5'

20'8'

---+k----- 66'2' --------<---

Section A·A

fig. 6-3
Figure 5.3: Single Barrel used for Design, (Billington, 1965)

We will consider the column height and spacing fixed for the design and given in Fig. 5.3.

The barrels are oriented in the direction of the long span with a shell thickness, h, of 3 in. The

shell will contain a minimum of three layers of reinforcement. Wire fabric will be used top and
3
bottom with main bars running in between. The cover over the fabric will be no less than 8 in.
 141

1 3
assuming 2 in. thick fabric. The bars will be no greater than 4 in. in diameter.

The curvature of the barrel will be initially assumed based upon the ease of construction

and material costs. If the shell is classified as semicircular, the roof will carry low stresses but

contain a surface area of 1.5 time greater than the actual covered surface increasing material costs

greatly. Also, the concrete would be very difficult to place near the springing where the slope is

nearly vertical. The desirable slope should be no greater than 45deg. However, with the ease of

construction comes the increase in stresses. For this design, we will use a slope of φk = 45 deg with

a radius of r = 26.6 ft.

The loads considered for the design are as follows:

Shell Dead Load 40

Roofing and Mechanical Equipment 10

Live Load (snow) 30

Total: 80

Even though the snow load often requires further thought due to drifting, for simplicity, we

will consider all three loads to be uniformly distributed over the surface of the shell. Note that all

predefined variables are selected as initial values and may be modified as required to benefit the

analysis.

5.2 Matlab

The matlab program presented in the appendix A is utilized to design a reinforced concrete

circular cylindrical shell. The program allows the user to design a simply supported shell with

longitudinal edge beams. The code is based on the theory of shallow shells along with the analysis

methods discussed in the previous chapter. Following the method laid out in Sec. 3.4.4, the program

first defines the principle system by computing the moment, Mφ , shear, Nφx , and the radial and

longitudinal forces, Nφ and Nx .

The program is relatively simple to use and understand. The user will input a few values

based on given dimensions, properties and loads for the shell of interest. Once the values are the
 142

selected the program can be ran. By the default, the moment, shear, and radial and longitudinal

forces are computed for the edge of the shell at midspan. This can be easily adjusted by paging

to the final section of code entitled, Shell Forces. Here the user can select a new angle φ (distance

from longitudinal edge, rad) and new length x (distance from transverse edge, ft), depending upon

the specific location of interest. Other locations of interest would include the crown along with the

end and quarter span points.

First, the design of the shell will be discussed below.

5.2.1 Cylindrical Shell Roof

The preliminary design of the shell structure is the first task that must be accomplished in

the design of the barrel vault. The matlab program follows the design process laid out by the

theory of shallow shells. The initial inputs for the program were determined in Sec. 5.1.1 and will

be the basis for our initial sizing of the structure.

The program requires the inputs of basic shell dimensions and the loading. Dimension inputs

include the shell length, radius, height, and the slope at the springing. The external and self weight

loads must also be considered. The program analyzes a uniformly distributed load over the surface

of the shell, similar to self weight. Points loads are not considered and outside the capabilities of

the code.

Once the basic dimension and load parameters have been input, the next step is to define

the shell constants, k, γ and Q.

n∗π
k= L
q
rk h
γ= √ (5.1)
4
3r r
√8 r√
Q= 34 rk
h
where n = 1 due to the close approximation of a uniform load from the first equation of fourier

series.

We will next define the roots and powers from Eq. 4.58. This will be necessary for the

definition of our trigonometric and hyperbolic multiples, required for determining our arbitrary
 143

constants. The determination of the multiples is dependent on the location φ within the shell. The

φ will be measured from the crown with negative taken as the clockwise direction of rotation.

The final step is to solve the series of simultaneous equations consisting of four equations

and four unknowns. The equations for consideration will be Mφ , as seen in Eq. 4.71, Nφ , Q0φ , and

Nφx . The derivation of the latter three variables will follow the same process as Mφ seen in Sec.

4.3.4. Now that we have four equations and four unknowns, the final requirement is the shell edge

conditions. In order to solve the simultaneous equations, we will set the longitudinal edge force Nφ

equal to the corrective line load TL .

4
TL = − pr cos(φk − φ) sin kx
π

and the shear force Nφx equal to the line load SL .

4 2
SL = − pr × sin(φk − φ) cos kx
π rk

The remaining forces will be set to zero and the simultaneous equations are solved for a, b, c and

d. The newly solved arbitrary constants are next substituted back into the equations of interest to

solve for the stress resultants, stress couples, and displacements.

As proof for validation, the results compiled by the matlab program match those presented

in Table 6-5 (Billington, 1965). The program allows for the adjustment of φ and x to locate the

moment, shear, and axial forces throughout the entire shell. These results, however, are only

preliminary since they do not take into consideration the edge beam the shell is supported by. The

next part of the design will consider the edge beam.

5.2.2 Edge Beam

Now that we have established the primary system of the shell roof and the results, we can

analyze the edge beams. It is important that all internal forces and displacements are know to size

the edge beam. Once again for proof of validation, we will use the size chosen for the design of the

Pennsylvania warehouse. The warehouse beam design can be seen below in Fig. 5.4.
 144

The program requires the inputs of basic shell dimensions, edge beam dimensions, edge beam

properties, steel properties and the loading. In addition, since the program is yet optimized, the

user must input the longitudinal stress resultant, Nx , determined by the Shell code. Also the

constants from ASCE Manual 31 Tables 2A and 2B must be inputted.

Shell dimension inputs include the shell length, radius, height, and the slope at the springing.

The external and self weight loads must also be considered. The program analyzes a uniformly

distributed load over the surface of the shell, similar to self weight. Points loads are not considered

and outside the capabilities of the code. Edge beam dimension and property inputs include the

beam width and depth along with the density of the concrete in use. The final property input is

the steel yield stress.

We will follow the analysis discussed previously in Sec. 4.3.5. Since the edge beam is con-

sidered as a simple beam separate from the shell, the displacements and internal forces of the edge

beam will be computed separately. However, in order to join the two systems, an equation of

compatibility, consisting of two simultaneous equations, must be solved to determine the correction

forces Vb and Sb . Similar to the shell corrective line loads, TL and SL , the edge beam must also

attribute corrective values to the forces computed on the shell. After applying the correction forces,

the final stress resultants and couples are determined

 Primary System. In this case we shall take the final results of the
analysis in Sec. 6d as the primary system. It may be reemphasized that
a primary system need not be statically determinate; it is only necessary
that all its internal forces and displacements be known. We now have
145

...
, , _ _ _ 37.61 - ---H

Figure 5.4: Pennsylvania Warehouse Roof Edge Beam Dimensions, (Billington, 1965)

For simplicity, the compatibility equation can be written in matrix form

    
 B11 + S11 B12 + S12
 Vb 
    F B11 + F S11 

= (5.2)

 
B21 + S21 B22 + S22  S   F B21 + F S21
   

b

The values for B:,: have been previously derived in 4.3.5. The values of S:,: , however, require

the ASCE Manual no. 31 (Whitney, 1952). The required coefficients can be taken directly out of

the tables within for the following equations:

L4
S11 = r3 hE
C sin(kx)
L4
S12 = r3 hE
C sin(kx)
(5.3)
L2
S21 = r2 h
C sin(kx)
L2
S22 = r2 h
C sin(kx)

where C is the constant selected from Tables 2A and 2B (Whitney, 1952). The final values

required are F B and F S. The values of F B have also been computed previously in Sec. 4.3.5. F S

can be computed by the shell values calculated by the Shell Design Matlab code. F B will require

the displacements v and w along with the longitudinal stress resultant Nx .

 146

F S11 = −v sin(φk − φ) + w cos(φk − φ)

(5.4)
Nx
F S21 = h

With all variables known, the values of Vb and Sb can be determined. The final internal shell

values are computed by the summation of stress resultant or couple determined by the Shell Design

code along with C ∗ V b and C ∗ Sb where once again C is a constant taken from Table 2A (Whitney,

1952).

5.2.2.1 Reinforcement

The final step of the preliminary design is to add reinforcement to the slab and beam. Even

though concrete can resist tensile stresses up to about 0.1 of its compressive strength, steel is

required to resist the tensile forces. The computation of reinforcement requires the finalized stress

resultants and couples calculated by the Edge Beam code.

The first step is to determine the principal stresses:

r  2
Nx +Nφ 2 + Nx −Nφ
N1 = 2 + Nxφ 2
r  2 (5.5)
Nx +Nφ 2 Nx −Nφ
N2 = 2 − Nxφ + 2

and the plane on which the first principal acts

−2Nxφ
tan 2θ = (5.6)
Nx − Nφ

where θ is measured counter clockwise from the face on which Tx acts. The second principle stress

will act at a right angle to the first.

The shell reinforcement will first be considered. The required area of steel, As , can be

determined by the general formulation

N1
As = (5.7)
fs

For the presented design example at φ = 0, N1 = 82.988 kip/ft and fs = 20 ksi (depends on type of

steel in use), requiring an As = 4.149 in2 /f t. The steel requirement can be computed throughout
 147

the entire shell by altering φ and solving for N1 from the newly computed stress resultants and

couples.

The final shell requirement will be the bending steel. Bending reinforcement is required over

the center half of the shell as this is where the maximum moment and thrust occur. In computing

the bending reinforcement, the thrust Nφ can be neglected

Mφ × 12
As = (5.8)
fs × 7/8 × ds

where ds is the depth of steel.

The edge beam reinforcement will now be considered. At the midspan of the the beam, all

tension is concentrated in the edge beam. Therefore the total tension is

 
N1 + 8.32
T = − 8.32 × d (5.9)
2

and the required steel can be computed as

T
As = (5.10)
fs

The minimum steel requirement must be met throughout the shell and edge beam

amin1 = 0.35 × area of concrete (tensile zone)

(5.11)
amin2 = 0.18 × area of concrete (shell)

5.3 Discussion

Even though the purpose of this analysis was to develop a preliminary design using only

hand derived equation, it was found that it is not entirely possible. In developing the Shell Design

code, the theory of shallow shells was used in its entirety allowing for a smooth design code based

solely on hand derived equations. The difficulty, however, came when designing the edge beam.

The design requires variables for compatibility to join the shell and beam. The edge beam values

were easily computed utilizing the ordinary theory of flexure. However, the compatibility constants

required for the slab are not so easily computed. In multiple sources of literature, (Billington,
 148

1965), and (Whitney, 1952), the text refers to the tables found in Whitney (1952) with no answer

as to how the constants within those tables have been derived.

Without full knowledge of the source of the constants, the edge beam code can not be

fully automated with equations, but rather requires inputs from the tables. With the given table

constants, the program does function sufficiently and determines the finalized stress resultants and

couples. These values were not only validated by the Billington (1965) publication but also the

shell described in example 3 of Whitney (1952). The validation by two different shells proves the

code works and is accurate for design.

The current code requires about three minutes to enter the inputs for analysis and run.

However, the code would be classified as a beta version and requires cleaning up. I believe by

combining the two files and adding an separate input notebook, the program can be completely

automated allowing the user to find any value within the shell in under a minute. Even the best

structural programs today cannot compute a design in that short of time, mainly due to the time

requirement of setting up the model itself.

5.4 Conclusion

In this chapter we applied the theory of shallow shells to develop a design tool to aid in future

cylindrical shell analysis. Utilizing the design tool, a brief preliminary design was conducted. The

design consisted of two parts, the shell itself and the edge beam with reinforcement.
 Chapter 6

Conclusions and Recommendations for Future Studies

6.1 Conclusions

In conclusion, I find that hand calculations have been and will continue to be a useful tool in

the design of arches and shells. Methodologies seen throughout history have engaged the designers

in basic hand calculations which have been determined as accurate design principles. Each designer

used what was given previously to further the practice and enable growth within the sector of arch

and shell design.

Through an extensive literature review of historical analysis and design methodologies, an ap-

preciation for hand calculations and the work to derive and ensure accuracy was developed. Within

the basis of masonry arch design, principles such as basic geometric design, wedge theory, line of

thrust and the ultimate load theorem were discovered and researched to gain basic knowledge on

the development of these methods. Furthermore, designers were able to implement these principles

for the design of vaults and domes as seen in the case study of St. Peter’s Basilica by Poleni.

Only with this basic understanding of the development of these theories can a designer

continue onward toward the advanced derivations of shell theories. Similar to the development of

the masonry arch theories, equilibrium, compatibility, and stress-strain relations were considered

to step through the derivations of a beam and plate in flexure, and ultimately the shell. This

process allowed for the development of a differential equation relating transverse displacement to

the applied load. In each stage of the derivation, the designer is able to build upon the first, just

as the designers of old, always looking to improve upon the current principle.
 150

Next, a vigorous derivation was conducted to develop membrane and shallow shell theories,

which are both vital for the design of shells. Both theories utilized the previously derived shell

equation, through simplifications specific to circular cylindrical shells. First, the methodology of

membrane theory was followed to develop the stress resultant and couples within the shell, along

with the displacements.

Building upon previous design principles, the theory of shallow shells expands upon the

theory of membranes by no longer neglecting bending but rather conducting an elaborate derivation

involving an eighth order differential equation. Solving the homogeneous and particular solutions

for the arbitrary constants allow the development of a solution more accurate than membrane

theory.

Even though the process of deriving and solving the equations is tedious, once the equations

are available, they can be of great use to a skilled programmer in developing working code similar

to what I have developed. Not only are the equations easily inputted into a code, they output

highly desirable results in a short amount of time. Solely by modifying the inputs, the user can

determine the moments, thrusts, and shears throughout the shell. Hand derived equations based

on the theory of shallow shells allows for accurate solutions in developing the design of a circular

cylindrical shell. Given the time required to clean up and optimize the program, the code can allow

for fast solutions without the time required to set up the shell in a structural design program.

This ultimately proves that hand calculations can still be useful for today’s engineered world.

Rather than running to buy the newest, most expensive structural analysis tool on the market, it

can be just as useful, and efficient, to develop your own software that allows the designers to see

the basic equations and codes that are utilized in the design.

6.2 Recommendations for Future Work

In conducting the research and development of this thesis, many future work opportunities

have come to mind. The first opportunity consists of expanding upon the type of shells and

conducting a similar analyses as presented within this thesis. Shell such as hemispherical domes
 151

the hyperbolic parabola are of most interest.

In addition, analyzing additional shell structures presents an opportunity to further an un-

derstanding of the limitations of the membrane theory and draw a larger push for the theory of

shallow shells. The cylindrical shell case presented within this thesis does not allow for a clear

distinction between the two theories since many of the assumptions of membrane theory are met

within a cylindrical shell. The comparison of the theory of shallow shells and membrane theory is

more easily displayed within complex structures like the hyperbolic parabola.

Third,in keeping with the topic of cylindrical shells, instead of solely viewing a simply sup-

ported shell as in this example, the design of a multiple barrel shell would be of interest. Once

again, the theory of shallow shells would be utilized in the preliminary design of the structure.

Conducting an analysis of multiple barrels will require great attention to detail in the regions of

joined barrels and boundary conditions.

Similarly, a shell analysis of the transverse support would be important. Not all shells can

be designed as a simply supported shell and require transverse ribs. Researching the design of such

supports would be important to the design of cylindrical shells.

However, prior to optimizing the code, an understanding must be gained as to where the coef-

ficients of Whitney (1952) originate. One must fully understand the derivation of these coefficients

through the respective free body diagrams and corresponding equations. Though the equations

and coefficients of Table 1 were easily derived, Table 2 presented difficulty in understanding the

origination and derivation of both the equations and coefficients.

An additional topic of future work would be the development of the cylindrical shell program

itself. As stated previously, the code does require some optimization to allow for more efficient use

than a structural analysis tool. Code may be added or subtracted to allow for a single program,

rather than two, along with an inputs page.

Understanding the underlying mathematical model of a shell and encapsulating the governing

equations in a numerical code (as presented within thesis), could ultimately prove to be a more

appropriate tool for engineers than the use of the finite element analysis without proper under-
 152

standing of shell behavior. In optimizing the presented code, it may hopefully someday be found

useful to designers interested in the lost topic of concrete and masonry shell design.
 Bibliography

Benvenuto, E. (1991). An introduction to the history of structural mechanics; Part II: Vaulted
structures and elastic systems. Springler-Verlag.

Billington, D. (1965). Thin Shell Concrete Structures. McGraw-Hill, Inc., Princeton University.

Coulomb, C. (1773). Essai sur une application de maximis & minimis à quelques problèmes de
statique relatifs à l’architecture. mémoires de l’académie Royale des Sciences, 7:343–382.

Couplet, C. (1731). De la poussée des voûtes. Mémoires de l’académie Royale des Sciences.

Couplet, C. (1732). Seconde partie de l’examen des voûtes.

mémoires de l’académie Royale des Sciences.

de la Hire, P. (1679). Nouveaux éléments des sections coniques, les lieux géometriques. Chez Andre
Prallard, Paris.

de la Hire, P. (1712). Sur la construction des voûtes dans les édifices.

mémoires de l’académie Royale des Sciences, 69.

Derand, F. (1743). L’architecture des voûtes, ou l’art des traits, et coupes de voûtes. Chez Andre
Cailleau, Paris.

Heyman, J. (1997). The Stone Skeleton: Structural Engineering of Masonry Architecture. Cam-
bridge University Press.

Heyman, J. (1998). Structural analysis; A historical approach. Cambridge University Press.

Huerta, S. (2005). Essays in the history of the theory of structures: In honour of Jacques Heyman.
Instituto Juan de Herrera, Madrid.

Krivoshapko, S. N. and Hyeng, C.and Mamieva, I. (2014). Chronology of erection of the earliest
reinforced concrete shells. International Journal of Research and Reviews in Applied Sciences,
18(2).

Kurrer, K. (2012). The History of the Theory of Structures: From Arch Analysis to Computational
Mechanics. John Wiley & Sons.

López, G. (2006). Poleni’s Manuscripts about the Dome of Saint Peter’s.

Proceedings of the Second International Congress on Construction History, 2.
 154

Moseley, H. (1839). On the theory of the arch. In Hann, J., editor, The theory , practice and
architecture of bridges.

Navier, C. (1826). Resumé des leçons données à l’école des ponts et chaussées sur l’application de la mécanique à
Dunot, Paris.

Peerdeman, B. (2008). Analysis of thin concrete shells revisited: Opportunities due to innovations
in materials and analysis methods. Master’s thesis.

Poleni, G. (1748). Memorie Istoriche della Gran Cupola del Tempio Vaticano, e de danni di essa
e de ristoramenti loro, divise in libri cinque. Stamperia del Seminario, Padova.

Prusinski, K. (2015). Analysis of complex structures. Master’s thesis, Department of Civil, Envi-
ronmental and Architectural Engineering, University of Colorado at Boulder.

Salimbeni, L. (1787). Degli archi e delle volte. Ramanzini, Verona.

Whitney, C. (1952). Design of Cylindrical Shell Roofs. American Society of Civil Engineers, no.
31 edition.

Winkler, E. (1880). Die Lage der Stützlinie im Gewölbe. Deutsche Bauzeitung, 13:127–130.
 Appendix A

Design Code

The following code has been developed via matlab for the preliminary design of the circular

cylindrical shell example. The code is discussed in detail within Sec. 5.2.

The shell design code (part one) consists of three sections, the shell design (master copy),

A.1, the shell coefficient function, A.1.1, and the shell force function, A.1.2. The edge beam design

code (part two) A.2 consists of only one piece of code computing the final stresses in the beam and

shell along with the required reinforcement.

A.1 Shell Design - Part One

The following code is developed for the design of the shell portion of the simply supported

cylindrical shell. The code follows the design procedure outlined by the theory of shallow shells and

is discussed in Sec. 5.2.1. The purpose of the design tool is to determine the internal forces and

displacements within the shell.The shell design code consists of the basic inputs and formulations

the the shell dimensioning. The shell coefficient function produces the simultaneous equations

required to solve for the arbitrary constants. With known constant values, the shell forces function

determines the internal forces and displacement of the shell. The code is presented as follows:
 
1 close all ; fclose all ; clear all ; clc ;

2 %%I n p u t s

3 %D i m e n s i o n s and Loads

4 L = 6 6 . 5 ; %f t

5 r = 2 6 . 6 ; %f t

6 h = 0 . 2 5 ; %f t

7 p h i k =45∗ p i / 1 8 0 ; %r a d

8 p h i =0; %r a d ( measured from e d g e )

 156

9 p h i t=phi phi k ;

10 q = . 0 8 0 ; %k s f

11 Phi=p h i k phi ;

12 d i m l o a d . Length=L ;

13 d i m l o a d . R a d i u s=r ;

14 d i m l o a d . H e i g h t=h ;

15 d i m l o a d . P h i k=p h i k ;

16 d i m l o a d . Phi=p h i ;

17 d i m l o a d . Pd=q ;

18

19 %S h e l l Constants

20 n =1;

21 k= n∗ p i /L ;

22 gamma=r ∗k / 3 ^ ( 1 / 4 ) ∗ s q r t ( h/ r ) ;

23 Q= 3 ^ ( 1 / 8 ) ∗ ( r /h ) ^ ( 1 / 4 ) ∗ s q r t ( r ∗k ) ;

24 x=L / 2 ;

25 E=4.32 e 5 ;%k s f

26 D=E∗h ^ 3 / 1 2 ;

27

28 S h e l l . k=k ;

29 S h e l l . gamma=gamma ;

30 S h e l l .Q=Q;

31 S h e l l . x=x ;

32 S h e l l . E=E ;

33 S h e l l .D=D;

34

35 %%Roots

36 %from Eq . 4 . 5 4

37 m1=s q r t ( ( s q r t ((1+gamma) ^2+1)+(1+gamma) ) / 2 ) ;

38 n1=s q r t ( ( s q r t ((1+gamma) ^2+1) ( 1 + gamma) ) / 2 ) ;

39 m2=s q r t ( ( s q r t ( ( 1 gamma) ^2+1) ( 1 gamma) ) / 2 ) ;

40 n2=s q r t ( ( s q r t ( ( 1 gamma) ^2+1) + ( 1 gamma) ) / 2 ) ;

41 mn=[m1 , n1 , m2 , n2 ] ;

42 %r o o t p o w e r s Eq . 4.55

43 a l p h a 1=Q∗m1 ;

44 b e t a 1=Q∗ n1 ;

45 a l p h a 1 p=Q∗m2 ;

46 b e t a 1 p=Q∗ n2 ;

47 r p =[ a l p h a 1 , b e t a 1 , a l p h a 1 p , b e t a 1 p ] ;

48

49 Roots . M1=m1 ;

50 Roots . N1=n1 ;

51 Roots . M2=m2 ;

52 Roots . N2=n2 ;

53 Roots . Alpha1=a l p h a 1 ;

54 Roots . Alpha2=a l p h a 1 p ;

55 Roots . Beta1=b e t a 1 ;

56 Roots . Beta2=b e t a 1 p ;

57

58 %%S h e l l Edge C o n d i t i o n s ( F o u r i e r Series )

59 TL=4/ p i ∗q∗ r ∗ c o s ( p h i k p h i ) ∗ s i n ( k∗x ) ;

 157

60 SL=4/ p i ∗q∗ r ∗ 2 / ( r ∗k ) ∗ s i n ( p h i k p h i ) ∗ c o s ( k∗x ) ;

61 M phi =0;

62 Np phi = TL ;

63 Qp phi =0;

64 Np xphi = SL ;

65

66 EdgeLoad . M phi=M phi ;

67 EdgeLoad . N p h i=Np phi ;

68 EdgeLoad . Qp phi=Qp phi ;

69 EdgeLoad . N x p h i=Np xphi ;

70

71 f =[ M phi ; Np phi ; Qp phi ; Np xphi ] ;

72

73 %%S h e l l Coefficients

74 %( 1 ) Mphi ( 2 )Mx ( 3 ) Qx ( 4 ) Nphi ( 5 ) Nx ( 6 ) u ( 7 )w

75 %( 8 ) Qphi ( 9 ) Nxphi ( 1 0 ) v ( 1 1 ) Mxphi ( 1 2 ) Qphip

76

77 %M phi

78 SC=1;

79 F1= s h e l l c o e f f i c i e n t s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , x ) ;

80

81 %N p h i

82 SC=4;

83 F2= s h e l l c o e f f i c i e n t s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , x ) ;

84

85 %Qphip

86 SC=12;

87 F3= s h e l l c o e f f i c i e n t s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , x ) ;

88

89 %Nxphi

90 SC=9;

91 F4= s h e l l c o e f f i c i e n t s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , x ) ;

92 F=[F1 ; F2 ; F3 ; F4 ] ;

93

94 S h e l l C o e f . M phi=F1 ;

95 S h e l l C o e f . N p h i=F2 ;

96 S h e l l C o e f . Qp phi=F3 ;

97 S h e l l C o e f . N x p h i=F4 ;

98

99 c o n s t a n t s = F ^ 1 ∗ f ;

100

101 %S h e l l Forces

102 p h i =0; %r a d ( measured from e d g e )

103 p h i t=phi phi k ;

104 x=L / 2 ;

105 Np x = 4 / p i ∗2∗ q / ( r ∗k ^ 2 ) ∗ c o s ( p h i k p h i ) ∗ s i n ( k∗x ) ;

106 Np phi = 4 / p i ∗q∗ r ∗ c o s ( p h i k p h i ) ∗ s i n ( k∗x ) ;

107 Np xphi = 4 / p i ∗q∗ r ∗ 2 / ( r ∗k ) ∗ s i n ( p h i k p h i ) ∗ c o s ( k∗x ) ;

108 up=2∗q / (E∗h∗ r ∗k ^ 3 ) ∗ c o s ( p h i k p h i ) ∗ c o s ( k∗x ) ;

109 vp = 2 ∗ q / (E∗h∗ r ^2∗ k ^ 4 ) ∗(1+2∗ k ^2∗ r ^ 2 ) ∗ s i n ( p h i k p h i ) ∗ s i n ( k∗x ) ;

110 wp= 2 ∗ q / (E∗h∗ r ^2∗ k ^ 4 ) ∗(1+2∗ k ^2∗ r ^2+ r ^4∗ k ^ 4 / 2 ) ∗ c o s ( p h i k p h i ) ∗ s i n ( k∗x ) ;

 158

111 %M phi

112 SC=1;

113 M phi= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x ) ;

114

115 %M x

116 SC=2;

117 M x= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x ) ;

118

119 %M xphi

120 SC=11;

121 M xphi= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x ) ;

122

123 %Q phi

124 SC=8;

125 Q phi= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x ) ;

126

127 %Q x

128 SC=3;

129 Q x= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x ) ;

130

131 %N p h i

132 SC=4;

133 N p h i= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+Np phi ;

134

135 %N x

136 SC=5;

137 N x= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+Np x ;

138

139 %N x p h i

140 SC=9;

141 N x p h i= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+Np xphi ;

142

143 %u

144 SC=6;

145 u= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+up ;

146

147 %v

148 SC=10;

149 v= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+vp ;

150

151 %w

152 SC=7;

153 w= s h e l l f o r c e s ( SC , k , r , p h i t , rp , mn, Q, gamma , h , c o n s t a n t s , x )+wp ;

154

155 DV= v∗ s i n ( p h i k p h i )+w∗ c o s ( p h i k phi ) ;

156 DH=v∗ c o s ( p h i k p h i )+w∗ s i n ( p h i k phi ) ;

157

158 F o r c e s . M phi=M phi ;

159 F o r c e s . N p h i=N p h i ;

160 F o r c e s . N x=N x ;

161 F o r c e s . N x p h i=N x p h i ;
 159

162 D i s p l a c e m e n t . u=u ;

163 D i s p l a c e m e n t . v=v ;

164 D i s p l a c e m e n t . w=w ;

165 D i s p l a c e m e n t . up=up ;

166 D i s p l a c e m e n t . vp=vp ;

167 D i s p l a c e m e n t . wp=wp ;
 

A.1.1 Shell Coefficients

The shell coefficient function produces the simultaneous equations required to solve for the

arbitrary constants. The methodology for deriving coefficients 1 - 11 have been previously discussed

in Sec.4.3.4 and seen by the example equation Mφ (4.71.

 
1 %%S h e l l Coefficients

2 function S h e l l= s h e l l c o e f f i c i e n t s ( SC , k , r, phit , rp , mn, Q, gamma , h , x )

4 a l p h a 1=r p ( 1 ) ;

5 b e t a 1=r p ( 2 ) ;

6 a l p h a 1 p=r p ( 3 ) ;

7 b e t a 1 p=r p ( 4 ) ;

8 m1 = mn( 1 ) ;

9 n1 = mn( 2 ) ;

10 m2 = mn( 3 ) ;

11 n2 = mn( 4 ) ;

12

13 %T r i g o n o m e t r i c and H y p e r b o l i c Multiples a t Edge

14 T1=c o s ( b e t a 1 ∗ p h i t ) ∗ c o s h ( a l p h a 1 ∗ p h i t ) ;

15 T2=s i n ( b e t a 1 ∗ p h i t ) ∗ s i n h ( a l p h a 1 ∗ p h i t ) ;

16 T3=c o s ( b e t a 1 p ∗ p h i t ) ∗ c o s h ( a l p h a 1 p ∗ p h i t ) ;

17 T4=s i n ( b e t a 1 p ∗ p h i t ) ∗ s i n h ( a l p h a 1 p ∗ p h i t ) ;

18 T5=c o s ( b e t a 1 ∗ p h i t ) ∗ s i n h ( a l p h a 1 ∗ p h i t ) ;

19 T6=s i n ( b e t a 1 ∗ p h i t ) ∗ c o s h ( a l p h a 1 ∗ p h i t ) ;

20 T7=c o s ( b e t a 1 p ∗ p h i t ) ∗ s i n h ( a l p h a 1 p ∗ p h i t ) ;

21 T8=s i n ( b e t a 1 p ∗ p h i t ) ∗ c o s h ( a l p h a 1 p ∗ p h i t ) ;

22

23 %C o e f f i c i e n t s

24 E=4.32 e 5 ;

25 D=E∗h ^ 3 / 1 2 ;

26 %( 1 ) M phi

27 R( 1 ) = 2 ∗D/ r ^2∗ s i n ( k∗x ) ;

28 B1 ( 1 )=Q^2∗(1+gamma) ;

29 B2 ( 1 )=Q^ 2 ;

30 B3 ( 1 )=Q^ 2 ∗ (gamma 1 ) ;

31 B4 ( 1 )=Q^ 2 ;

32

33 %( 2 ) M x

34 R( 2 ) =2∗D∗k ^2∗ s i n ( k∗x ) ;

35 B1 ( 2 ) =1;
 160

36 B2 ( 2 ) =0;

37 B3 ( 2 ) =1;

38 B4 ( 2 ) =0;

39

40 %( 3 ) Q x

41 R( 3 ) = 2 ∗D∗k ^ 3 / (gamma) ∗ c o s ( k∗x ) ;

42 B1 ( 3 ) =1;

43 B2 ( 3 ) =1;

44 B3 ( 3 ) = 1 ;

45 B4 ( 3 ) =1;

46

47 %( 4 ) N p h i

48 R( 4 ) =4∗D∗ r ∗k ^4/gamma^2∗ s i n ( k∗x ) ;

49 B1 ( 4 ) =0;

50 B2 ( 4 ) =1;

51 B3 ( 4 ) =0;

52 B4 ( 4 ) = 1 ;

53

54 %( 5 ) N x

55 R( 5 ) = 4 ∗D∗ r ∗k ^4/gamma^3∗ s i n ( k∗x ) ;

56 B1 ( 5 ) = 1 ;

57 B2 ( 5 )=1+gamma ;

58 B3 ( 5 ) =1;

59 B4 ( 5 ) = 1 gamma ;

60

61 %( 6 ) u

62 R( 6 ) =4∗D∗ r ∗k ^ 3 / ( h∗E∗gamma^ 3 ) ∗ c o s ( k∗x ) ;

63 B1 ( 6 ) = 1 ;

64 B2 ( 6 )=1+gamma ;

65 B3 ( 6 ) =1;

66 B4 ( 6 ) = 1 gamma ;

67

68 %( 7 ) w

69 R( 7 ) =2∗ s i n ( k∗x ) ;

70 B1 ( 7 ) =1;

71 B2 ( 7 ) =0;

72 B3 ( 7 ) =1;

73 B4 ( 7 ) =0;

74

75 %( 8 ) Q phi

76 R( 8 ) = 2 ∗D∗k ^ 3 / ( s q r t (gamma) ) ^3∗ s i n ( k∗x ) ;

77 B1 ( 8 )=m1 n1 ;

78 B2 ( 8 )=m1+n1 ;

79 B3 ( 8 ) = ( m2+n2 ) ;

80 B4 ( 8 )=m2 n2 ;

81

82 %( 9 ) N x p h i

83 R( 9 ) =4∗D∗ r ∗k ^ 4 / ( s q r t (gamma) ) ^5∗ c o s ( k∗x ) ;

84 B1 ( 9 ) = n1 ;

85 B2 ( 9 )=m1 ;

86 B3 ( 9 )=n2 ;
 161

87 B4 ( 9 ) = m2 ;

88

89 %( 1 0 ) v

90 R( 1 0 ) =4∗D∗ r ∗k ^ 3 / (E∗h ∗ ( s q r t (gamma) ) ^ 7 ) ∗ s i n ( k∗x ) ;

91 B1 ( 1 0 )=m1+n1 ∗ ( 1 gamma) ;

92 B2 ( 1 0 )=n1 m1 ∗ ( 1 gamma) ;

93 B3 ( 1 0 ) = m2+n2 ∗(1+gamma) ;

94 B4 ( 1 0 ) = n2 m2∗(1+gamma) ;

95

96 %( 1 1 ) M xphi

97 R( 1 1 ) =2∗D∗k / r ∗ c o s ( k∗x ) ;

98 B1 ( 1 1 )=a l p h a 1 ;

99 B2 ( 1 1 )=b e t a 1 ;

100 B3 ( 1 1 )=a l p h a 1 p ;

101 B4 ( 1 1 )=b e t a 1 p ;

102

103 %Q phip

104 R( 1 2 ) = 2 ∗D∗k ^ 3 / ( s q r t (gamma) ) ^3∗ s i n ( k∗x ) ;

105 B1 ( 1 2 )=m1 ∗ ( 1 gamma) n1 ;

106 B2 ( 1 2 )=m1+n1 ∗ ( 1 gamma) ;

107 B3 ( 1 2 ) = m2∗(1+gamma) n2 ;

108 B4 ( 1 2 )=m2 n2 ∗(1+gamma) ;

109 %V a r i a b l e s

110 syms

111 i f SC<8

112 A=(B1 (SC) ∗T1 B2 (SC) ∗T2 ) ;

113 B= B2 (SC) ∗T1 B1 (SC) ∗T2 ;

114 C=B3 (SC) ∗T3 B4 (SC) ∗T4 ;

115 D= B4 (SC) ∗T3 B3 (SC) ∗T4 ;

116 e l s e

117 A=B1 (SC) ∗T5 B2 (SC) ∗T6 ;

118 B= B2 (SC) ∗T5 B1 (SC) ∗T6 ;

119 C=B3 (SC) ∗T7 B4 (SC) ∗T8 ;

120 D= B4 (SC) ∗T7 B3 (SC) ∗T8 ;

121 end

122 S h e l l =[A, B , C, D] ∗ 2 ∗R(SC) ;

123

124 end
 

A.1.2 Shell Forces

The shell force function follows the procedure of the shell coefficient function with the ex-

ception of known A, B, C and D constants. With known constant values, the shell forces function

determines the internal forces and displacement of the shell utilizes the same equations developed

in Sec.4.3.4 and seen above in A.1.1.

 
1 %%S h e l l Forces
 162

2 function F= s h e l l f o r c e s ( SC , k , r, phit , rp , mn, Q, gamma , h , constants , x)

4 a=c o n s t a n t s ( 1 ) ;

5 b=c o n s t a n t s ( 2 ) ;

6 c=c o n s t a n t s ( 3 ) ;

7 d=c o n s t a n t s ( 4 ) ;

8 a l p h a 1=r p ( 1 ) ;

9 b e t a 1=r p ( 2 ) ;

10 a l p h a 1 p=r p ( 3 ) ;

11 b e t a 1 p=r p ( 4 ) ;

12 m1 = mn( 1 ) ;

13 n1 = mn( 2 ) ;

14 m2 = mn( 3 ) ;

15 n2 = mn( 4 ) ;

16

17 %T r i g o n o m e t r i c and H y p e r b o l i c Multiples a t Edge

18 T1=c o s ( b e t a 1 ∗ p h i t ) ∗ c o s h ( a l p h a 1 ∗ p h i t ) ;

19 T2=s i n ( b e t a 1 ∗ p h i t ) ∗ s i n h ( a l p h a 1 ∗ p h i t ) ;

20 T3=c o s ( b e t a 1 p ∗ p h i t ) ∗ c o s h ( a l p h a 1 p ∗ p h i t ) ;

21 T4=s i n ( b e t a 1 p ∗ p h i t ) ∗ s i n h ( a l p h a 1 p ∗ p h i t ) ;

22 T5=c o s ( b e t a 1 ∗ p h i t ) ∗ s i n h ( a l p h a 1 ∗ p h i t ) ;

23 T6=s i n ( b e t a 1 ∗ p h i t ) ∗ c o s h ( a l p h a 1 ∗ p h i t ) ;

24 T7=c o s ( b e t a 1 p ∗ p h i t ) ∗ s i n h ( a l p h a 1 p ∗ p h i t ) ;

25 T8=s i n ( b e t a 1 p ∗ p h i t ) ∗ c o s h ( a l p h a 1 p ∗ p h i t ) ;

26

27 %C o e f f i c i e n t s

28 E=4.32 e 5 ;

29 D=E∗h ^ 3 / 1 2 ;

30

31 %( 1 ) M phi

32 R( 1 ) = 2 ∗D/ r ^2∗ s i n ( k∗x ) ;

33 B1 ( 1 )=Q^2∗(1+gamma) ;

34 B2 ( 1 )=Q^ 2 ;

35 B3 ( 1 )=Q^ 2 ∗ (gamma 1 ) ;

36 B4 ( 1 )=Q^ 2 ;

37

38 %( 2 ) M x

39 R( 2 ) =2∗D∗k ^2∗ s i n ( k∗x ) ;

40 B1 ( 2 ) =1;

41 B2 ( 2 ) =0;

42 B3 ( 2 ) =1;

43 B4 ( 2 ) =0;

44

45 %( 3 ) Q x

46 R( 3 ) = 2 ∗D∗k ^ 3 / (gamma) ∗ c o s ( k∗x ) ;

47 B1 ( 3 ) =1;

48 B2 ( 3 ) =1;

49 B3 ( 3 ) = 1 ;

50 B4 ( 3 ) =1;

51

52 %( 4 ) N p h i
 163

53 R( 4 ) =4∗D∗ r ∗k ^4/gamma^2∗ s i n ( k∗x ) ;

54 B1 ( 4 ) =0;

55 B2 ( 4 ) =1;

56 B3 ( 4 ) =0;

57 B4 ( 4 ) = 1 ;

58

59 %( 5 ) N x

60 R( 5 ) = 4 ∗D∗ r ∗k ^4/gamma^3∗ s i n ( k∗x ) ;

61 B1 ( 5 ) = 1 ;

62 B2 ( 5 )=1+gamma ;

63 B3 ( 5 ) =1;

64 B4 ( 5 ) = 1 gamma ;

65

66 %( 6 ) u

67 R( 6 ) =4∗D∗ r ∗k ^ 3 / ( h∗E∗gamma^ 3 ) ∗ c o s ( k∗x ) ;

68 B1 ( 6 ) = 1 ;

69 B2 ( 6 )=1+gamma ;

70 B3 ( 6 ) =1;

71 B4 ( 6 ) = 1 gamma ;

72

73 %( 7 ) w

74 R( 7 ) =2∗ s i n ( k∗x ) ;

75 B1 ( 7 ) =1;

76 B2 ( 7 ) =0;

77 B3 ( 7 ) =1;

78 B4 ( 7 ) =0;

79

80 %( 8 ) Q phi

81 R( 8 ) = 2 ∗D∗k ^ 3 / ( s q r t (gamma) ) ^3∗ s i n ( k∗x ) ;

82 B1 ( 8 )=m1 n1 ;

83 B2 ( 8 )=m1+n1 ;

84 B3 ( 8 ) = ( m2+n2 ) ;

85 B4 ( 8 )=m2 n2 ;

86

87 %( 9 ) N x p h i

88 R( 9 ) =4∗D∗ r ∗k ^ 4 / ( s q r t (gamma) ) ^5∗ c o s ( k∗x ) ;

89 B1 ( 9 ) = n1 ;

90 B2 ( 9 )=m1 ;

91 B3 ( 9 )=n2 ;

92 B4 ( 9 ) = m2 ;

93

94 %( 1 0 ) v

95 R( 1 0 ) =4∗D∗ r ∗k ^ 3 / (E∗h ∗ ( s q r t (gamma) ) ^ 7 ) ∗ s i n ( k∗x ) ;

96 B1 ( 1 0 )=m1+n1 ∗ ( 1 gamma) ;

97 B2 ( 1 0 )=n1 m1 ∗ ( 1 gamma) ;

98 B3 ( 1 0 ) = m2+n2 ∗(1+gamma) ;

99 B4 ( 1 0 ) = n2 m2∗(1+gamma) ;

100

101 %( 1 1 ) M xphi

102 R( 1 1 ) =2∗D∗k / r ∗ c o s ( k∗x ) ;

103 B1 ( 1 1 )=a l p h a 1 ;
 164

104 B2 ( 1 1 )=b e t a 1 ;

105 B3 ( 1 1 )=a l p h a 1 p ;

106 B4 ( 1 1 )=b e t a 1 p ;

107

108 %V a r i a b l e s

109 i f SC<8

110 A=a ∗B1 (SC) ∗T1 a ∗B2 (SC) ∗T2 ;

111 B= b∗B2 (SC) ∗T1 b∗B1 (SC) ∗T2 ;

112 C=c ∗B3 (SC) ∗T3 c ∗B4 (SC) ∗T4 ;

113 D= d∗B4 (SC) ∗T3 d∗B3 (SC) ∗T4 ;

114 e l s e

115 A=a ∗B1 (SC) ∗T5 a ∗B2 (SC) ∗T6 ;

116 B= b∗B2 (SC) ∗T5 b∗B1 (SC) ∗T6 ;

117 C=c ∗B3 (SC) ∗T7 c ∗B4 (SC) ∗T8 ;

118 D= d∗B4 (SC) ∗T7 d∗B3 (SC) ∗T8 ;

119 end

120 F=2∗R(SC) ∗ (A+B+C+D) ;

121

122 end
 

A.2 Edge Beam - Part Two

Part two of the code for shell design is presented below and discussed in Sec. 5.2.2. The edge

beam code is required to configure a compatibility relation between the preliminary values of the

shell developed by the above shell design code, and the values a an edge beam subjected to the

ordinary flexural theory. The development of the compatibility relation allows for the computation

of the final internal forces of the simply supported shell. The edge beam code also computed the

required area of steel for the beam and shell. The code is presented as follows:
 
1 %Edge Beam

2 close all ; fclose all ; clear all ; clc ;

3 %%I n p u t s

4 b =1; %f t

5 d = 2 . 2 5 ; %f t

6 L = 6 6 . 5 ; %f t

7 r = 2 6 . 6 ; %f t

8 h = 0 . 2 5 ; %f t

9 p h i k =45∗ p i / 1 8 0 ; %r a d

10 p h i =0∗ p i / 1 8 0 ; %r a d ( measured from e d g e )

11 q = . 0 8 0 ; %k s f

12 r h o = . 1 5 0 ; %k c f

13 w=b∗d∗ r h o ; %k l f

14 n =1;

15 k= n∗ p i /L ;

16 x=L / 2 ;
 165

17 E=4.32 e 5 ;%k s f

18

19 %%N x

20 N x = 9 9 . 0 3 ; %k i p / f t ( computed i n Shell Design )

21

22 %ASCE Manual 31 v a l u e s from T a b l e 2B

23 C1 = 9 6 . 1 0 ; %( V e r t i c a l Edge Load Delta v )

24 C2 = 6 4 . 6 9 ; %( H o r i z o n t a l Edge Load Delta v )

25 C3 = 2 . 5 0 0 ; %( S h e a r Edge Load Delta v )

26 %ASCE Manual 31 v a l u e s from T a b l e 2A

27 C4 = 1 9 . 6 4 ; %( V e r t i c a l Edge Load T x)

28 C5 = 1 . 3 8 8 ; %( S h e a r Edge Load T x)

29

30 %C o r r e c t i o n Values

31 Np phi = 4 / p i ∗q∗ r ∗ c o s ( p h i k p h i ) ∗ s i n ( k∗x ) ;

32 Np xphi = 4 / p i ∗q∗ r ∗ 2 / ( r ∗k ) ∗ s i n ( p h i k phi ) ∗ cos (0) ;

33 VL= Np phi ∗ s i n ( p h i k ) ;

34 HL= Np phi ∗ c o s ( p h i k ) ;

35 SL = Np xphi ;

36

37 %S h e l l

38 NS=L ^ 4 / ( r ^3∗ h∗E) ;

39 FS ( 1 , 1 )=L ^ 4 / ( r ^3∗ h∗E) ∗ ( q∗ r ∗ ( ( ( 2 ∗ r ) / ( p i ∗L ) ) ^2+2/ p i ^4+( r /L ) ^ 4 ∗ ( c o s ( p h i k p h i ) ) ^ 2 )+C1∗VL+C2∗HL+C3∗SL ) ∗ s i n ( k∗x

);

40 FS ( 2 , 1 )=N x /h ;

41 S ( 1 , 1 )=NS∗C1∗ s i n ( k∗x ) ;

42 S ( 1 , 2 )=NS∗C3∗ s i n ( k∗x ) ;

43 S ( 2 , 1 )=NS∗E∗ r /L^2∗C4∗ s i n ( k∗x ) ;

44 S ( 2 , 2 )=NS∗E∗ r /L^2∗C5∗ s i n ( k∗x ) ;

45

46 %Beam

47 NB=L ^ 4 / ( d ^3∗ b∗E) ;

48 FB ( 1 , 1 )=NB∗ 4 / p i ∗w∗ 1 2 / ( p i ∗n ) ^4∗ s i n ( k∗x ) ;

49 FB ( 2 , 1 )=NB∗E∗d/L^2∗4/ p i ∗w∗ 6 / ( p i ∗n ) ^2∗ s i n ( k∗x ) ;

50 B ( 1 , 1 )=NB∗ 1 2 / ( p i ∗n ) ^4∗ s i n ( k∗x ) ;

51 B ( 1 , 2 ) = NB∗d/L ∗ 6 / ( p i ∗n ) ^3∗ s i n ( k∗x ) ;

52 B ( 2 , 1 ) = NB∗E∗d/L ^ 2 ∗ 6 / ( p i ∗n ) ^2∗ s i n ( k∗x ) ;

53 B ( 2 , 2 )=L / ( b∗d ) ∗ 4 / ( p i ∗n ) ∗ s i n ( k∗x ) ;

54

55 %C o m p a t i b i l i t y

56 BS=S+B ;

57 F=FS+FB ;

58 X=BS \ F ;

59 Vb=X( 1 , 1 ) ;

60 SB=X( 2 , 1 ) ;

61

62 %%F i n a l Internal Forces

63 %N x

64 N x=N x+(Vb∗C4 ) +(SB∗C5 ) ;

65

66 %N x p h i
 166

67 N x p h i =0;

68 C4=0;

69 C5 = 0 . 4 ;

70 N x p h i=N x p h i+Vb∗C4+SB∗C5 ;

71

72 %N p h i

73 N p h i =0;

74 C4 = 0 . 7 0 7 ;

75 C5=0;

76 N p h i=N p h i+Vb∗C4+SB∗C5 ;

77

78 %M phi

79 M phi =0;

80 C4=0;

81 C5=0;

82 N x p h i=N x p h i+Vb∗C4+SB∗C5 ;

83

84 F o r c e s . N x=N x ;

85 F o r c e s . N x p h i=N x p h i ;

86 F o r c e s . N p h i=N p h i ;

87 F o r c e s . M phi=M phi ;

88

89 %%R e i n f o r c e m e n t

90 f s =20; %k s i

91 d s =2; %i n

92 %S h e l l Reinforcement

93 N1=(N x+N p h i ) /2+ s q r t ( N x p h i ^2+(( N x N p h i ) / 2 ) ^ 2 ) ;

94 N2=(N x+N p h i ) / 2 s q r t ( N x p h i ^2+(( N x N p h i ) / 2 ) ^ 2 ) ;

95

96 A s s h e l l=N1/ f s ;

97 As bend=M phi ∗ 1 2 / ( f s ∗7/8∗ d s ) ;

98

99 %Beam R e i n f o r c e m e n t

100 T=((N1+ 8 . 3 2 ) / 2 8 . 3 2 ) ∗d ;

101 As beam=T/ f s ;

102

103 %Min R e i n f o r c e m e n t

104 Amin1 =0.0035∗ h ∗ 1 2 ∗ 1 2 ;

105 Amin2 =0.0018∗ h ∗ 1 2 ∗ 1 2 ;