The Effect of Geometry and Topology on the

Mechanics of Grid Shells

by

Samar Rula Malek

B.S. Civil Engineering
Johns Hopkins University, 2002

S.M. Civil and Environmental Engineering

ARH
VS
Massachusetts Institute of Technology, 2007

Submitted to the Department of Civil and Environmental Engineering

in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy in the Field of Structures and Materials

at the Massachusetts Institute of Technology

June 2012

@ 2012 Massachusetts Institute of Technology. All rights reserved.

Signature of Author:
Departefient of Civil an nvironmental Engineering
a~o i A A /T
\" , 1 1')I

Certified by:
f John A. Ochsetfdorf
Associate Professor of Building Technology and Civil and Environmental Engineering
This Co-,qiinPrviqnr

Certified by:
Tomasz w ierzOicki
ofessor of Applied Mechanics
hesis Co- ervisor
Certified by:
0 Id9O M. Reis
Assistant Professor of Civil and Environmental Engineering and Mechanical Engineering
. JThesis Reader
Accepted by:
fdepf a
CharDearm C.
Chair, Departmenital Committee for Gr atfStudents
 The Effect of Geometry and Topology on the Mechanics of Grid Shells
by
Samar Rula Malek
Submitted to the Department of Civil and Environmental Engineering
on May 1, 2012 in Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in the Field of Structures and Materials

Abstract
The use of grid shell structures in architecture and structural engineering has risen in
the past decade, yet fundamental research on the mechanics of such structures is lacking.
Grid shells are long span structures comprised of a lattice of single layer members forming
a curved surface. Grid shells can be made of a wide range of materials from steel to wood.
They have potential to be used in readapting existing spaces or in new aesthetically
pleasing structures. By studying their mechanics, engineers can be more effective at the
schematic phase of design so that the potential of grid shells can be maximized.

This research conducts a parametric study that varies the topology and topogra-
phy of grid shells. The parametric space is framed around real-world design constraints
including the grid spacing, panel shape, span-to-height ratio and the use of double
curvature.

In this thesis, the buckling capacity is evaluated using finite element analysis for
two typical grid shell geometries: the spherical cap and the corrugated vault. First,
a spherical cap is considered for which an analytical solution exists and therefore the
accuracy of the numerical procedure is validated. Simple closed-form solutions are derived
using the concept of the equivalent continuum and compared to the numerical models.
Then, the parametric study of the spherical cap is performed including variations of the
grid spacing, the span to height ratio and the panel shape (triangles and quadrilaterals).

Having determined the efficiency of the computational tool the study is extended
to the barrel vault. Here the new features of the analysis are the use of double curvature
by introducing corrugation along the edge and the crown. By understanding the
fundamental mechanical behavior of grid shells, design guidelines aimed to maximize
their capacity and efficiency and intended to facilitate the discussion between architect
and engineer are proposed.

Thesis Co-Supervisor: John A. Ochsendorf

Title: Associate Professor of Building Technology and Civil and Environmental Engineering

Thesis Co-Supervisor: Tomasz Wierzbicki

Title: Professor of Applied Mechanics

Thesis Reader: Pedro M. Reis

Title: Assistant Professor of Civil and Environmental Engineering and Mechanical Engineering
Acknowledgments

I am most grateful to the guidance I have received from my co-supervisors, Professors

Wierzbicki and Ochsendorf. Thank you John for setting the example on how to pursue
research passions, how to empower students and how kindness can be an incredible moti-
vating force. I cannot thank you enough for giving me the generous opportunity to teach
your class and for extending your network of amazing scholars, engineers and architects.
Thank you Professor Wierzbicki for teaching me mechanics! For allowing me to be
your teaching assistant and your officemate. More importantly thank you for the countless
hours of discussion and advice on this dissertation, mechanics, academia, ... and classical
music. Thank you Professor Pedro Reis for your sincere enthusiasm, encouragement,
advice and for serving on my committee.
A heartfelt thank you to my undergraduate mentor Professor Roger Ghanem who
introduced me to research starting my freshman year of college and whose guidance opened
the door to MIT. Thank you for your generosity with all the research and academic
opportunities.
Thank you to my sources of funding and support and people who have influenced
my professional development: The Martin Family Fellowship for Sustainability; Professor
John Fernandez & the Building Technology Program; Kris Kipp & Department of Civil
and Environmental Engineering; Duncan Kincaid & CRON; the Department of Mechan-
ical Engineering; ODGE & MSRP; the GSC Travel Grant; Kim Bernard, Professor Linn
Hobbs & GECD.
Thank you to my students - from JHU to Teach for America to CMHS to TutoringPlus
to MIT, and to my friends and colleagues who have supported me along the way: Tia
Letras, Bryna Chen, Mira Chokshi, Dr. David Quinn, Marcia, Diana Chien, Caitlin
Mueller, Rory Clune, Dr. Jenn Zessin, and Dr. Elham Sahraei.
A special thank you to Dr. Alejandra Menchaca-Brandan and Monica Orta for being
amazing, smart, funny people. Thank you Alex for the years of so many good memories
and for always being there. Thank you Monica for introducing me to MSRP - the best
part of my MIT experience - and for being my 'rock'.
Lastly, thank you to my family. To mom and dad, for leaving their country and
sacrificing their families to give their kids a better opportunity. To my sister Alia for
being the first to set the example to follow. To my brother Hussam for giving me the
childhood memories to reminisce on. To my brother-in-law Nilesh for being the '4th'
sibling to read my essays:, but more importantly for being a great dad to my niece. To
my sister Rana whose love and unconditional support I could not live without, but most
especially for bringing Manal into this world. To the joy of my life and my reason to
persevere, Manal (and my future second niece) - I love you!

5
Table of Contents

List of Figures 8

List of Tables 12
1 Introduction 14
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1.1 Definition of Grid Shell. . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1.2 Advantages of Grid Shells . . . . . . . . . . . . . . . . . . . . . . . 15
1.1.3 Unresolved Questions . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2.1 Current Design of Grid Shells . . . . . . . . . . . . . . . . . . . . . 21
1.2.2 Thesis Goals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2.3 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.3 Outline of Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2 Literature Review 25
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 Structural Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.2.1 Design Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2.2 Design Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Structural Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.1 Numerical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.2 Analytical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Methodology 34
3.1 Procedure for Parametric Study . . . . . . . . . . . . . . . . . . . . . . . . 34
3.1.1 Parameters and Test Geometries . . . . . . . . . . . . . . . . . . . 35
3.1.2 Numerical and Analytical Methods . . . . . . . . . . . . . . . . . . 35
3.2 Computational Model Verification . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 2D Arch Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2 Grillage Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

6
 Table of Contents

3.3 Summary .......... .................................... 47

4 Spherical Cap Grid Shell Study 48

4.1 Introduction ................................... 48
4.2 Problem Definition ............................... 49
4.2.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2.2 Numerical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2.3 Analytical Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3 R esults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3.1 Model Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3.2 Collapse Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.3 Numerical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.4 Analytical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.5 Numerical and Analytical Results . . . . . . . . . . . . . . . . . . . 59
4.4 D iscussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.5 Sum m ary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5 Corrugated Vault Study 64

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.2 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2.1 G eom etry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.2.4 Analytical Models. . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.5 Numerical Models. . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3 R esults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3.1 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.2 Continuous Shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3.3 G rid Shell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.4 D iscussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.5 Sum mary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6 Conclusions 94
6.1 Summary of Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.2 Future Work. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . 95

A MATLAB Code 97
A. 1 Spherical Cap - Quadrilateral Grid . . . . . . . . . . . . . . . . . . . . . . 97
A.2 Spherical Cap - Triangular Grid . . . . . . . . . . . . . . . . . . . . . . . . 105
A.3 Corrugated Vault - Continuous Shell . . . . . . . . . . . . . . . . . . . . . 112
A.4 Corrugated Vault - Quadrilateral Grid . . . . . . . . . . . . . . . . . . . . 116
A.5 Corrugated Vault - Triangular Grid . . . . . . . . . . . . . . . . . . . . . . 121

7
Table of Contents

B Equivalent Continuum 126

Bibliography 128

8
List of Figures

1.1 The Great Court grid shell at the British Museum. . . . . . . . . . . . . . . . 15

1.2 Exam ples of grid shells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3 Quadrilateral grid at the National Portrait Gallery (left); triangular grid at
the Great Court (right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4 Coarse grid at the National Portrait Gallery (left); dense grid at the Mannheim
M ultihalle (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Shallow shell at the Odeon in Munich (left); steeper shell at the Museum for
Hamburg History (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6 The Japan Pavilion (left); the National Portrait Gallery (right). . . . . . . . . 21
1.7 The current design process of grid shells (two figures reproduced from (Harris
et al., 2008)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.8 Proposed contributions of this dissertation to the design process of grid shells. 23

2.1 The British M useum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.2 External view of the Kogod Courtyard grid shell at the National Portrait Gallery. 27
2.3 External view of the Neckarsulm Swimming Arena. . . . . . . . . . . . . . . . 27
2.4 Equivalent continuum example defined by an equivalent volume. . . . . . . . . 31

3.1 Values and illustration of the varied parameters. . . . . . . . . . . . . . . . . . 36

3.2 Examples of the spherical cap and the corrugated barrel vault as grid shells. 36
3.3 Examples of the continuous shell and the grid shell finite element model for
the spherical cap and the barrel vault. . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Examples of the inputted undeformed finite element model and the ADINA
output of the first buckling mode (magnified at a maximum factor of 4500). 38
3.5 2D geometry of the pinned-pinned arch. . . . . . . . . . . . . . . . . . . . . . 40
3.6 The three competing failure modes (buckling, snap-through, yielding) of an
arch as a function of the span-to-height ratio. . . . . . . . . . . . . . . . . . . 40
3.7 Monotonic convergence of the strain energy for a simply supported beam with
a distributed load. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.8 The strain energy convergence of a pinned-pinned arch with a normally dis-
tributed load. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.9 Comparison of the arch deflection under uniform distributed loading between
a coarse and a dense mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

9
List of Figures

3.10 Force displacement plot as a function of increasing mesh density. . . . . . . . . 42

3.11 Comparison of the theoretical buckling load (Equation 3.1c) to the numerical
linearized buckling analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.12 Comparison of linear and nonlinear buckling analysis for two scaled magnitudes
(#) of imperfection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.13 Comparison of the buckling load for the circular (C) and parabolic (P) arch
with a normally (N) and vertically (V) distributed load. . . . . . . . . . . . . 45
3.14 Comparison of the theoretical buckling load for a distributed load and the
numerical buckling load for point loading. . . . . . . . . . . . . . . . . . . . . 45
3.15 Finite element representation of rigid ends. . . . . . . . . . . . . . . . . . . . . 46
3.16 G rillage case study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.1 Spherical cap grid shell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.2 A hierarchal tree illustrating the number of models for the spherical cap study. 50
4.3 Illustration of the six combinations of the panel shape and spacing. . . . . . . 51
4.4 Example of the spherical cap grid shell finite element model with k= h 6 . . 52
4.5 The quadrilateral and triangular topology and their repeating units (the cru-
ciform and the hexagon). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.6 Comparison of the numerical and the theoretical buckling load of a continuous
spherical cap using axisymmetric shell elements. . . . . . . . . . . . . . . . . . 55
4.7 Error in the numerical buckling load to the analytical when the loading is
applied as a point load and a pressure load. . . . . . . . . . . . . . . . . . . . 55
4.8 Comparison of the collapse load to the linearized buckling load for a spherical
cap of = 20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.9 The buckling capacity of the spherical cap as a function of the grid spacing for
the quadrilateral grid (left) and the triangular grid (right). . . . . . . . . . . . 57
4.10 First buckling mode of spherical cap with h = 20 and a spacing of s = 3in:
quadrilateral topology (left), and triangular topology (right) (magnified by a
factor of 5000). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.11 Structural efficiency of the spherical cap grid shell (Equation 4.2). . . . . . . . 58
4.12 Comparison of the buckling load (Equation 4.1) for the four equivalent con-
tinuum definitions for a quadrilateral grid with k=20 . . . . . . . . . . . . . . 59
4.13 Comparison of analytical and numerical results for the spherical cap grid shell
study. The rows from top to bottom represent the topologies quadrilateral and
triangular, respectively. The columns from left to right represent decreasing
span-to-height ratios. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.14 Design guidelines for a spherical cap grid shell. Columns from left to right
represent decreasing span-to-height ratio. . . . . . . . . . . . . . . . . . . . . . 62

5.1 Examples of grid shells with corrugation at the edge and the crown. . . . . . . 64
5.2 The four groups of parameters: corrugation location, aspect ratio, span-to-
height ratio, and grid spacing (left to right). . . . . . . . . . . . . . . . . . . . 65
5.3 Straight edge barrel vault geometry and boundary conditions . . . . . . . . . . 66

10
 List of Figures

5.4 The hierarchal tree illustrating the number of corrugated grid shell models. . . 67
5.5 Corrugated barrel vault geometry. . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.6 Illustration of the eight combinations of n and j for the corrugated edge. . 69
5.7 Geometric representation of the corrugated edge as a function of y.
. . .... 69
5.8 Plane strain condition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.9 Corrugated shell cross-section. . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.10 The effect of a varied moment of inertia on the numerical factor m for the
buckling load......... ..................................... 73
5.11 The three grid shell densities used in the corrugated vault study (10x10, 20x20,
30x30)........... ....................................... 74
5.12 Dimensions of the grid spacing and the overlay of the repeating unit of the
grid shell cruciform and the square continuous shell. . . . . . . . . . . . . . . . 75
5.13 Plane strain validation for the continuous shell. . . . . . . . . . . . . . . . . . 76
5.14 The buckling mode for the continuous shells magnified by a factor of 400. . . . 76
5.15 Plane strain validation for the grid shell. . . . . . . . . . . . . . . . . . . . . . 77
5.16 The first buckling mode of the straight vault for each of the three grid shell
densities (magnified by a factor of 1000). . . . . . . . . . . . . . . . . . . . . . 77
5.17 Gain in capacity from the straight barrel vault by corrugating the edge of the
continuous shell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.18 Gain in capacity from the straight barrel vault by corrugating the crown for
the continuous shell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.19 First buckling mode for the corrugated edge case with increasing corrugation
from top to bottom (magnified by a factor of 500). . . . . . . . . . . . . . . . 79
5.20 Illustration of the effective arclength aeff. . . . . . . . . . . . . . . . . . . . . 80
5.21 A matrix of plots highlighting the gain in capacity from the straight barrel vault
by using in-phase corrugation for the continuous shell. The rows represent the
increase in corrugation amplitude by O and the columns represent the number
of w aves n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.22 Comparison of the numerical results of the corrugated crown and the in-phase
to the analytical (Equation 5.8) for the continuous shell. . . . . . . . . . . . . 83
5.23 Comparison of the gain in capacity in terms of the percent volume increase
from the straight vault for the continuous shell. . . . . . . . . . . . . . . . . . 85
5.24 Comparison of the collapse load and the buckling load for a normally dis-
tributed load (N) and a vertically distributed load (V) for the in-phase con-
tinuous shell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.25 Gain in capacity by corrugating the edge of the 30x30 grid shell. . . . . . . . . 86
5.26 The effect of varying the grid density for the corrugated edge grid shell. . . . . 88
5.27 Gain in capacity by corrugating the crown of the 30x30 grid shell. . . . . . . . 89
5.28 Gain in capacity by in-phase corrugation of the 30x30 grid shell. . . . . . . . . 90
5.29 Comparison of the gain in capacity in terms of percent volume increase from
straight vault for the grid shell for all three corrugation locations (edge, crown
and in-phase). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

11
List of Figures

5.30 Set of design guidelines to be used in the early stages of design of a corrugated
grid shell vault. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

12
List of Tables

1.1 Some grid shells built in the past 30 years. . . . . . . . . . . . . . . . . . . . . 16

1.2 Examples showing the range of the size and materials of grid shells. . . . . . . 18

2.1 Summary of previous research on the equivalent continuum. . . . . . . . . . . 32

3.1 Comparison of the analytical and numerical tip deflection w, using both the
beam and plate models for the grillage. . . . . . . . . . . . . . . . . . . . . . . 47

4.1 The equivalent thickness (teq) relations for the quadrilateral and the triangle
using the four equivalent definitions. . . . . . . . . . . . . . . . . . . . . . . . 54

5.1 Equations for calculating oaef given a span-to-height ratio and an aspect ratio
of corrugation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Maximum value of -- to increase the capacity of thte corrugated edge by at
least 50% . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.3 Maximum values of -L
ag ands. y to increase the capacity by 50% for the 30x30
grid shell .. .. .. .. .. . . .. . .. . .. . . .. . .. . . .. . .. . . .. .. 89

13
Chapter 1

Introduction

The objective of this work is to evaluate the mechanical performance of grid shells and to
provide design guidelines to facilitate the discussion between the architect and the engineer
during schematic design. A parametric study is conducted by varying the topology and
topography and calculating the change in the load-bearing capacity using both numerical
and analytical methods. The parameters are studied using two key grid shell geometries:
the spherical cap and the barrel vault.

1.1 Motivation
In the year 2000 the Great Court grid shell at the British Museum (Figure 1.1) opened
and quickly became one of the most photographed, and visited public spaces in London.
While it was not the first grid shell, its popularity and success propelled grid shells into
both the public and the professional architectural and engineering conscience. Since then
there has been a rise in the number of grid shells built. Table 1.1 lists grid shells that
have built since the year 1975.

1.1.1 Definition of Grid Shell

The term grid shell is defined more recently as "a structure with the shape and strength
of a double curvature shell, but made of a grid instead of a solid surface. These structures
can cross large spans with very few material. They can be made of any kind of material -
steel, aluminum, wood or even cardboard tubes" (Douthe et al., 2006). The terms "lattice
shell" and "reticulated shell" have also been used to describe grid shells but more by the

14
 1.1. Motivation

Figure 1.1 - The Great Court grid shell at the British Museum.

academic community and not practicing engineers. There is a debate if a grid shell is
defined by its structural action, or by its construction process. For example, grid shells
made from wood are formed by laying the lattice flat and then either pushing or dropping
the shell into place as was done in the Mannheim Multihalle (Happold and Liddell, 1975),
and the Weald and Downland Museum (Wells, 2001), respectively. For this thesis, a grid
shell is defined to be a long span structure comprised of a network of members creating
the single layer "grid" that forms the curved surface "shell".

1.1.2 Advantages of Grid Shells

Grid shells are an efficient means of spanning space. They have been used to cover both
existing spaces such as the Cabot Circus (Figure 1.2a) and new spaces like the Savill
Garden Visitor Center (Figure 1.2b). They have been used as stand-alone structures as
seen in the Weald and Downland Museum (Figure 1.2c).
The benefits of using a grid shell compared to either the conventional slab and frame
system or the continuous shell are both aesthetical and structural. Grid shells create
dramatic spaces by pulling the eye to heights higher than the building top, and by articu-
lating the space with its discretized topology. They create beautiful spaces because they

15
Chapter 1. Introduction

Table 1.1 - Some grid shells built in the past 30 years.

Year Structure Location

1975 Mannheim Multihalle Mannheim, Germany
1989 Museum for Hamburg History Hamburg, Germany
1989 Swimming Arena Neckarsulm Neckarsulm, Germany
1994 Meeting Hall Flemish Council Brussels, Belgium
1996 Zoo Berlin Hippo House Berlin, Germany
1998 DZ-Bank Berlin, Germany
2000 Great Court London, United Kingdom
2000 Japan Pavilion Hanover Expo Hanover, Germany
2001 Porticullis House London, United Kingdom
2002 Weald and Downland Museum West Sussex, United Kingdom
2002 German Historical Museum Berlin, Germany
2005 Fiera Milano Milan, Italy
2006 Berlin Main Train Station Berlin, Germany
2007 Kogod Courtyard Washington DC, USA
2007 Savill Gardens Egham, UK
2007 Odeon Courtyard Munich, Germany
2007 Golden Terraces Warsaw, Poland
2008 Cabot Circus Roof Bristol, United Kingdom
2009 Palacio de Comunicaciones Madrid, Spain
2011 Centre Pompidou-Metz Metz, France
2011 National Maritime Museum Amsterdam, Netherlands

16
 1.1. Motivation

(a) Cabot Circus (b) Savill Gardens

(c) Weald and Downland Museum (d) Japan Pavilion

Figure 1.2 - Examples of grid shells.

are light and airy due to their efficient use of material, single layer members, and open-
ings. Their fairly simple construction of members and nodes create shell-like structures
without the arduous process of form work and pouring.
Grid shells also differ from conventional frame systems and continuous shells in that
they can create more sustainable design by lowering embodied energy and by reducing
operating energy. In 2008, the UK adopted the Climate Change Act which calls for an
80% reduction in C02 emission by 2050 (Parliament, 2008). Further, all new public
buildings are to have zero carbon emissions by 2018. However, 60% of existing buildings
in the UK will exist in 2050 and they will represent 45% of the total floor space (Delay
et al., 2009). Thus effort to reduced carbon emissions is by both the refurbishments of
existing buildings and the design of new ones. Further, lighting, heating and ventilation
were found to be the predominant consumers of energy in UK buildings.

17
Chapter 1. Introduction

Grid shells facilitate sustainable development in three major ways. (i) They enable
adaptive reuse of historic buildings, allowing open spaces to become enclosed, offering
shelter, commercial, or recreational possibilities. (ii) They save structural material needs
by over 50% compared to conventional framed structures and can be constructed of a
variety of materials with lower embodied energy such as wood or cardboard tubes (Ref,
2007). (iii) They also admit greater natural light, harnessing the natural elements to
create heating and cooling environments that are comfortable (Pople, 2002a).
Grid shells also have an unexplored potential for use as temporary buildings. For ex-
ample they can be used at the Olympics, fulfilling the need for temporary large structures.
Grid shells can also be used for disaster relief situations or for exhibition spaces.
Examples of grid shells that demonstrate their vast range of material, shape and
function are the Great Court in London (Figure 1.1), the Weald and Downland in West
Sussex (Figure 1.2c), and the Japanese Pavilion from the Hanover Exposition (Figure
1.2d). Their range of material include steel, wood and cardboard and Table 1.2 lists the
range of panel shape, footprint, function and cladding seen in grid shells.

Table 1.2 - Examples showing the range of the size and materials of grid shells.

[The Great Court Weald and Downland Japan Pavilion

member material steel wood cardboard
panel shape triangle triangle quadrilateral
panel material glass wood & glazing paper membrane
shell footprint 320ft x 230ft 164ft x 50ft 236ft x 114ft
function roof building temporary structure

1.1.3 Unresolved Questions

Grid shells are highly complex structures as are their mechanics. For example, there is
no clear understanding if a quadrilateral grid is better than a triangular. Figure 1.3 are
magnified views of the National Portrait Gallery and the Great Court grid shells. In
the National Portrait Gallery the members become increasingly larger as they approach
the support. Whereas in the Great Court, the members are slender and not noticeably
varying in size. Are grid shells bending or membrane dominated and what is the trade-off
between using the quadrilateral grid versus the triangular?

18
 1.1. Motivation

Figure 1.3 - Quadrilateral grid at the National Portrait Gallery (left); triangular grid at the
Great Court (right).

The effect of topology is also dependent on the spacing. The benefit of shell structures
is the additional stiffness due to curvature. Figure 1.4 juxtaposes the aerial view of the
National Portrait Gallery on the left and the Mannheim Multihalle on the right. For the
coarse grid on the left, the surface is more faceted, whereas on the right, the denser grid
is a smoother surface. Thus, how dense must the grid be to exhibit geometric stiffness?
And how does grid spacing affect the capacity?
Because geometric stiffness results from curvature, how does the span-to-height ratio
affect the load bearing capacity? In Figure 1.5 the shallower shell at the Odeon in Munich
is compared to the steeper shell on the right at the Museum of Hamburg History. So which
structure is stronger? How does the span-to-height ratio affect the load bearing capacity
and the weight required?

19
Chapter 1. Introduction

Figure 1.4 - Coarse grid at the National Portrait Gallery (left); dense grid at the Mannheim
Multihalle (right).

Figure 1.5 - Shallow shell at the Odeon in Munich (left); steeper shell at the Museum for
Hamburg History (right).

In addition to considering the curvature along the cross section, curvature can also be
introduced along the length. In Figure 1.6, the Japan Pavilion on the left has corrugation
at the crown and the edge, but at the National Portrait Gallery the corrugation is only
at the crown and flattens at the perimeter. Where is the best location for corrugation to
maximize the vault strength and how much corrugation is needed?

20
 1.2. Problem Statement

Figure 1.6 - The Japan Pavilion (left); the National Portrait Gallery (right).

1.2 Problem Statement

1.2.1 Current Design of Grid Shells

Grid shells have been designed on a case by case basis and have not been studied as a
structural type. In the schematic phase of design, the key players are the architect and the
client where they start the design from the architect's sketch; here the global geometry
is defined purely on the aesthetics. Figure 1.7 illustrates the progression of design from
schematic design (SD) to design development (DD). Currently, the engineers come in
the design development phase to refine the details of the previously selected shape. The
engineer at this stage has no power to make any major changes to a proposed structure.
It is therefore imperative that the questions regarding the main aspects of grid shell
geometry posed in the previous section be addressed during early design stages. In order
to do that there is a need to build an intuition and to develop simple expressions that
can allow engineers to do back of the envelope calculations. Thus, the improved design
process of grid shells will include the engineers as a key player in SD and the use of "back
of the envelope" calculations in SD.

1.2.2 Thesis Goals

There is a need to evaluate the mechanical performance of grid shells during schematic
design and to provide design guidelines to facilitate discussion between architect and

21
Chapter 1. Introduction

FEstructural analysis

Figure 1.7 - The current design process of grid shells (two figures reproduced from (Harris
et al., 2008)).

engineer (Figure 1.8). Therefore the goals of this dissertation are:

" to develop a methodology to calculate the change in load-bearing capacity of grid

shells due to their curvature and topology using numerical and analytical methods;

* to provide a set of design guidelines that can contribute to maximize the mechanical
performance of grid shells during early design stages.

The thesis goals can be summarized into the following three questions.

1. How do the curvature and topology affect the load-bearing capacity of grid shells?

2. How to best describe the mechanics of these structures?

3. What are the design rules of thumb to maximize grid shell performance?

1.2.3 Approach
The approach used in this thesis is to conduct a parametric study varying the topology
and topography and to calculate the change in load-bearing capacity using numerical
and analytical methods. Normally intuition is built from theory, experiments and case

22
 1.3. Outline of Chapters

Schematic Design
Design Development:

This dissertation Previous research

Mechanics Design
Guidelines

Figure 1.8 - Proposed contributions of this dissertation to the design process of grid shells.

studies. Because these structures are complex and few have been built, identifying the
structural trends must rely on computational models. A parametric study is conducted on
two structural shapes: the spherical cap and the barrel vault. The parameters perturbed
are identified in the following Chapter as those that often drive the design process and
affect both the engineering and the aesthetics of the grid shell.

1.3 Outline of Chapters

In Chapter 2 published research on grid shells from the communities of structural design
and structural mechanics is presented. The literature review provides the context of this
dissertation and establishes the need for this original research. It also identifies the design
parameters and constraints that have the most impact on the global shape and topology
that will be used in the following chapters.
In Chapter 3, the methodology for achieving the thesis goals is presented. All com-
putational methods and modeling assumptions are also explained and verified using the
commercial finite element package ADINA (Engineering, 2010).
The first of the parametric studies, the spherical cap grid shell, is presented in Chapter
4. Both analytical and numerical methods are used to determine the load-bearing capacity
of the spherical cap grid shell. New equivalent continuum techniques are derived and
compared. The coupling of the numerical and analytical work provides an understanding

23
Chapter 1. Introduction

of the mechanics of the structures. Simple analytical methods to estimate the capacity of
a grid shell are also derived and presented.
In Chapter 5 the parametric study is extended to a barrel vault. The role of curvature
on the load-bearing capacity is investigated by corrugating the vault. In both studies a
continuous shell is first modeled so to validate the methods used. Chapter 6 concludes
with the original contributions of this dissertation.

24
Chapter 2

Literature Review

2.1 Introduction
The objectives of this chapter are to present the key contributions of previous research
done on grid shells and to provide the context for the contributions herein. There are
two communities that directly and indirectly work on grid shells: the structural design
community and the structural mechanics community. The structural design community
consists of practicing engineers and consultants who provide firsthand experience of the
design process of existing grid shells. The structural mechanics community are academics
who focus more on the nonlinear analysis techniques of shells like imperfection sensitivity.
First the work of the structural design community will be presented, followed by the
mechanics community and conclude with a summary of the remaining open problems
that this dissertation solves.

2.2 Structural Design

The structural design community provides both the overview of the design process and the
detailed analysis of existing grid shells. The publications on the overall design process are
authored by the structural engineers of the grid shells and describe specifically the design
constraints, challenges and solutions. The more detailed papers on the mathematical
progression of the grid geometry are then provided by the consultants who are tasked
with writing the specific codes that map out the grid.

25
Chapter 2. Literature Review

2.2.1 Design Constraints

Based on the literature on built grid shells, there are three main design constraints that
govern their global shape and topology: the shell height, the grid spacing, and the panel
shape. In both the design of the Great Court and the Dutch Maritime Museum, the
height of the shell was governed by the constraint that the shell was not to be seen from
the ground (Figure 2.1a) (Anderson, 2000; Adriaenssens et al., 2010). The Great Court's
shell height was further constrained because it could not be taller than the dome of the
Reading Room and obstruct the view of the dome in the London skyline (Figure 2.1b).
The second and third constraints are the panel shape and subsequently its size. The
topology of grid shells has differed from quadrilaterals to triangles to irregularly shaped
polygons. In the case of the Great Court, the selection of the triangular grid came from
the constraint that the glazing must produce a smooth flat surface (Williams, 2001).
Whereas for the Kogod Courtyard grid shell at the National Portrait Gallery, the use
of the quadrilateral grid and double curvature created a jagged external surface (Figure
2.2). In one built grid shell, the Neckarsulm Swimming Arena, the cladding was made
of curved quadrilateral glazing so to make the surface smooth (Figure 2.3) (Schlaich and
Schober, 1996).
In some cases the panel size has been determined by the panel shape and in other cases
by the required strength of the grid shell. For the Great Court, the average triangular
panel is seven feet because the glass triangular panels could not be manufactured at a

(a) Entrance (b) Skyline view

Figure 2.1 - The British Museum.

26
 2.2. Structural Design

Figure 2.2 External view of the Kogod Courtyard grid shell at the National Portrait Gallery.

Figure 2.3 - External view of the Neckarsulm Swimming Arena.

27
Chapter 2. Literature Review

larger size (Anderson, 2000). In the case of the Weald and Downland grid shell, the grid
spacing was first varied from 0.5m to 1m. However, the im spacing could not provide
the required buckling resistance so the design was reverted back to 0.5m spacing (Harris
et al., 2003).
The shell height, panel size and shape are the design constraints identified here that
impact the aesthetics and engineering of the structure. There are other design constraints
that are also described in these papers but are not within the scope of this thesis.
For example, another constraint is that the grid shell cannot impose any loads on
the existing structures they touch (Adriaenssens et al., 2010; Hart, 2001). All the forces
must be transferred to vertically supporting members. In the Dutch Maritime Museum
all the loads are transferred to the courtyard corners (Keuning, 2011) and in the National
Portrait Gallery the grid shell rests on a series of visible columns (Ouroussoff, 2007). The
restriction to not impose loads on the existing building is also described in the design
of the Portcullis House (Thornton, 2000) and the Norwich Union buildings (Stansfield,
2007).

2.2.2 Design Analysis

Grid shells are currently designed on a case by case basis. Computer codes are specifically
written to define the geometry of each grid shell and to perform the structural analysis.
The main method employed for mapping a grid to a surface is dynamic relaxation (Day,
1965). This method modifies the shape by monitoring the kinetic energy as the model is
forced to move.
As mentioned earlier, the starting point from which the engineer is involved is after
the global shape and topology have been established. For example, in the Great Court
(Williams, 2001) the process of generating the final geometry began from the input of a
triangular grid shell with the maximum shell height being reached between the perimeter
and the Reading Room dome. A similar process was done on both the Savill Gardens
and the Weald and Downland Museum (Harris et al., 2008, 2003). After Dr. Williams
performed the initial form-finding analysis on the initial geometry, the edited shape was
imported into a structural analysis program for the buckling analysis.
Because the global shape is established before the engineer is involved with the design,
the structural design community has also looked at other ways to map the grid onto a
surface by using also dynamic explicit finite element simulations (Bouhaya et al., 2010)

28
 2.3. Structural Mechanics

or by applying fractal geometry (Vyzantiadou et al., 2007).

The geometry and structural analysis of grid shells is tedious and complex and not
every design option can be investigated. Over 3000 lines of code were written specifically
for the Great Court's geometry definition and structural analysis, and it took over 5000
cycles for the model to converge (Williams, 2001). There is a need for more simplified
tools that can inform the engineer and architect during the Schematic Design phase.

2.3 Structural Mechanics

In the structural mechanics community it is well established that buckling is the dominant
failure mode for reticulated shells (Gioncu, 1985). Research on grid shells has subsequently
focused mainly on the factors influencing the buckling load like coupled instabilities, the
rigidity of joints (Lpez et al., 2007), and the effect of imperfections (Balut and Gioncu,
2000; Buchert, 1965; Calladine, 1995; Hanaor, 1995; Singer and Abramovich, 1995; Sumec
and Sumec, 1990). A more detailed overview of the state of the art on reticulated shells
has been presented (Gioncu, 1995).

2.3.1 Numerical Methods

For example, one paper studied the effects of imperfection on the stability of barrel vaults
and domes (Bulenda and Knippers, 2001). The first challenge was to define the imperfect
shape. Both the static deflected shape and the first buckling mode were considered with
various magnitudes. The varied parameters were the span-to-height ratio, the boundary
conditions from hinged to fixed and the loading from symmetric to asymmetric.
While the span-to-height ratio was varied, the range only considered steep shells. As
shown earlier in the many roof examples of grid shells, the shells are shallow. They also
did not consider the effect of other design parameters mentioned in Section 2.2.1. In
addition, they studied the effects on basic barrel vaults when in reality grid shells are
built as corrugated vaults. In fact, until this dissertation, there has not been research on
the mechanics of corrugated vault grid shells though that is what is actually built.
The main contributions of the structural mechanics community are most useful in the
Design Development phase of grid shells when the nonlinear structural analysis needs to
be completed. The numerical studies did not investigate the effects of the parameters
(shell height, panel size and shape) most influential in the design process. There has yet

29
Chapter 2. Literature Review

to be an extensive parametric study that considers those parameters and also chooses
realistic ranges of their values. Lastly, the majority of the research is of a numerical
nature and does not provide an analytical explanation into the behavior of grid shells.

2.3.2 Analytical Methods

In Chapter 3, the analytical derivation will employ the equivalent continuum technique.
The concept of equivalent thickness has been around since the 60's (Wright, 1965; Forman
and Hutchinson, 1970). Previous shell mechanics research focused on assessing the buck-
ling capacity of grid shells using an equivalent continuum idea and discrete methods. But
the validation and exploration of those methods were limited to the technology of their
time. The researchers also did not compare multiple equivalent continuum definitions.
First the concept of equivalent thickness is illustrated with a simple example. Given
a volume of material V, the volume can be "smeared" into the shape of a spherical cap
as either a thin continuous shell or into a network of bars shown in Figure 2.4.
For example, in a grid shell with a quadrilateral topology the repeating unit is a
cruciform as shown in Figure 2.4. The cross sectional area A =tt 2 is kept constant and
the network of bars is equally spaced in both directions at a value of s. The equivalent
thickness for a continuous shell and a grid shell to have the same volume is derived
in Equation 2.1. The volumes are approximately equal because there are two negligible
sources of error: the partial cruciform units at the perimeter are neglected and the volume
where the members intersect is counted twice.

Volumecontinuou Volumegridshell

teq 82 stlt2
2t 1t 2
teq ~ t (2.1)
S

The previous example considers an equivalence defined by volume only. However, other
equivalences have been and can be established. The three previously researched meth-
ods for equivalence include equivalent stiffness, equivalent split rigidity, and orthotropic
equivalence (Gioncu, 1995).
One method of equivalence is defined by an equivalent depth and an equivalent Young's
Modulus (Wright, 1965; Chriss and Wright, 1978). Two cases of a spherical shell, dis-
cretized with equilateral triangles are considered: one where the members have the same

30
 2.3. Structural Mechanics

S S
t

Figure 2.4 - Equivalent continuum example defined by an equivalent volume.

cross-section and another where the cross-sections are different. The methodology in-
cludes developing a relationship between the shell and bar forces, the elastic properties
and the buckling criteria. After establishing an equivalent thickness, the buckling load
is calculated. The proposed method was then compared to the known existing collapse
loads of domes such as the Bucharest dome (Wright, 1965).
Another equivalence definition used two different thicknesses for each of the rigidities
(axial and bending) (Forman and Hutchinson, 1970). The authors tested their method on
a spherical cap with an equilateral triangular grid. The equivalent continuum results were
compared to that of a discrete model and it was found that the accuracy had a 3-6% error
depending on the slenderness of the members. For more slender members the equivalent
continuum became more inaccurate. The authors called for a need of further theoretical
and experimental comparisons.
Lastly, equivalence has been defined using orthotropic equivalence (Kollar and Du-
lacska, 1984). Here the shell buckling equation is derived for an orthotropic continuum
where the orthotropic property mimics the placement of ribs in the x and y direction.
While this method was proposed it was not validated on any case studies.
Table 2.1 summarizes the methods and their respective researchers. In all the works
mentioned only one technique in each case was considered; the researchers did not compare
their methods with each other nor tested their methods with different topologies or as a
function of either the grid spacing or the shell height as is done in this dissertation.

31
Chapter 2. Literature Review

Table 2.1 - Summary of previous research on the equivalent continuum.

Method Approach Authors ] Time Frame

Equivalent uses an equivalent depth (Wright, 1965; Sumec 1970-1990
Stiffness and modulus of elastic- and Sumec, 1990)
ity
Split Rigidity defines two thicknesses (Buchert, 1965; Forman 1970
defined by extensional and Hutchinson, 1970)
and flexural deforma-
tions
Orthotropic derives the differential (Gioncu, 1985; Kollar 1975-1985
equations and rigidities and Dulacska, 1984)

2.4 Summary
This chapter has provided the context of this dissertation by identifying the previous re-
search done on grid shells from two communities: the structural design and the structural
mechanics communities. This chapter reviewed what research been done, the topics of
interest and has highlighted the areas in which contributions need to be made. This litera-
ture review allows the following conclusions to be drawn, and motivates the corresponding
contributions of this dissertation:

" The main design constraints that govern the global shape of the grid shell are the
shell height, the panel size and the panel shape.

" Having established buckling as the main failure mode, research within the struc-
tural mechanics community focuses on nonlinear effects like imperfections. The
implications of these effects is used in the Design Development phase when the final
structural analysis is performed and not in the Schematic Design.

" While academic research on grid shells has considered the effect of some parameters
on the stability of the structure, the parameters are not framed within the real
design constraints nor described in a language useful for practicing architects and
engineers. An extensive parametric study is needed to study the effect of real

32
 2.4. Summary

design parameters on the load-bearing capacity of grid shells and to help build the
engineer's intuition on the mechanical behavior of grid shells.

* The analysis of built grid shells is complicated and requires computer codes written
specifically to the geometry of the shell. Simpler analytical methods are needed for
back of the envelope calculations used by practicing structural engineers.

33
Chapter 3

Methodology

The methodology used to answer the research questions posed in Chapter 1 and stated
again below will be presented in this chapter.

1. How do the curvature and topology affect the load-bearing capacity of grid shells?

2. How to best describe the mechanics of these structures?

3. What are the design rules of thumb to maximize grid shell performance?

The thesis goals are to conduct a parametric study varying the topology and topography
and to calculate the change in the load-bearing capacity using numerical and analytical
methods.

3.1 Procedure for Parametric Study

In order to achieve the thesis goals, an outline of the steps required are listed below and
explained in the following subsections.

1. Identify the parameters influencing the design as defined by the real design con-
straints.

2. Define two geometries based on typical grid shell footprints.

3. Build a finite element model for each permutation for both the continuous shell and
the grid shell.

34
 3.1. Procedure for Parametric Study

4. Use the commercial finite element software ADINA to run a linearized buckling
analysis and post-process the results.

5. Derive new analytical expressions to calculate the buckling load and compare with
the numerical results.

3.1.1 Parameters and Test Geometries

The first step is to identify the parameters that drive the global shape in the design
process. The parameters are the grid spacing, the panel shape, the span-to-height ratio
and the use of double curvature by corrugation. All the parameters were identified and
explained in Sections 1.1.3 and 2.2.1. In Figure 3.1 the parameters and their ranges are
listed and illustrated.
The grid spacing is varied between 30in, 60in, and 120in. The two shapes considered
are the quadrilateral grid and the triangular grid. The three span-to-height ratios con-
sidered are : =6,11,20 representing fairly shallow shells. The use of double curvature is
implemented at either the edge, the crown, or simultaneously at the crown and edge.
The effect of changing the parameters is tested on two key geometries - the spherical
cap and the barrel vault and the results are presented in Chapters 4 and 5 respectively.
These geometries were selected because they are often used as grid shells (Figure 3.2).
The geometry, member properties, loading and boundary conditions of the spherical cap
and the barrel vault will be defined later in their respective chapters.

3.1.2 Numerical and Analytical Methods

The thesis goals are achieved using both numerical and analytical methods. Engineering
problems cannot always be solved with a closed-form derivation. Thus the mathematical
formulation is often solved using numerical (discrete) methods. A powerful method used
in engineering is the finite element method where engineering bodies are represented
as discrete elements that satisfy equilibrium, compatibility, and constitutive relations. In
this thesis, complex, doubly-curved structures are studied numerically using finite element
modeling (FEM).
A finite element model is built for each permutation of the parameters for both a
continuous shell and a grid shell. A continuous shell was modeled first to validate the
computational modeling techniques and to help build the understanding of the mechanics

35
Chapter 3. Methodology

Parameter: Value:

grid spacing 30 in - 120in

panel shape quadrilaterals

triangles

span-to-height = 20, 11, 6

double curvature

Figure 3.1 Values and illustration of the varied parameters.

Spherical Cap Barrel Vault

Figure 3.2 - Examples of the spherical cap and the corrugated barrel vault as grid shells.

36
 3.2. Computational Model Verification

before complicating them further as grid shells (Figure 3.3).

The finite element models are generated using a program written by the author and
provided in Appendix A. The user inputs the parameter values defined earlier and the
geometry of the structure, and outputs the finite element model input file to be used in
ADINA. The linearized buckling analysis is then run and the buckling load and mode are
calculated. An example of the undeformed finite element model and the post-processed
buckled shape are shown in Figure 3.4. The linearized buckling analysis does not account
for prebuckling rotations. The linear analysis is chosen over the nonlinear analysis because
it is computationally faster and similar to the predicted collapse load as will be verified
in later sections. Lastly, new analytical expressions are derived to calculate the buckling
load and compared with numerical results.

3.2 Computational Model Verification

In order to proceed with more complicated problems, the computational modeling tech-
niques are validated by studying the failure modes of a 2D arch and the bending of a
grillage. In establishing the accuracy of the models, the following technical questions are
answered:

Spherical Cap Barrel Vault

AD
NN
A

Continuous shell

DD
N
A N

Grid shell

Figure 3.3 - Examples of the continuous shell and the grid shell finite element model for the
spherical cap and the barrel vault.

37
Chapter 3. Methodology

Spherical Cap Barrel Vault

I D
N I
A N
A
Undeformed

model
p
First bucklingA

Figure 3.4 - Examples of the inputted undeformed finite element model and the ADINA output
of the first buckling mode (magnified at a maximum factor of 4500).

* What are the failure modes of an arch and the corresponding critical loads?

* What closed form solutions exist and what are their limitations?

* How fine of a mesh is needed to accurately predict the failure load?

* Is there a difference in capacity between a circular and parabolic arch or between a

normally distributed load and a vertically distributed load?

* What are the limitations of the linear eigenvalue solution for determining the critical
buckling load?

* Can the distributed load be represented by an equivalent point load?

* For the grillage model, how should the joints be modeled?

Options for validating research questions include experimental, analytical and numer-
ical techniques. For grid shells, because they are large structures and few have been built,
full scale experimental studies were not a valid option. Where appropriate, analytical
solutions were employed and derived. However, the bulk of this research uses numeri-
cal models. Even though computational methods have advanced, it is still necessary to
validate the modeling assumptions and techniques using benchmark studies.

38
 3.2. Computational Model Verification

3.2.1 2D Arch Study

A 2D arch is a good prototype structure on which various methods and approximations
can be tested. All the complications resulting from the interplay between bending and
membrane action are present in the arch. Therefore attention is focused first on the
analysis of the arch comparing the analytical to the numerical solution.
Consider a shallow, circular arch of radius R, height h, span L and angle a shown
in Figure 3.5. The arch is pinned-pinned supported with a uniformly distributed normal
load q. There are three competing failure modes derived by Timoshenko (Timoshenko,
1961): yielding, snap-through and buckling (Equations 3.1),where E is Young's Modulus
of Elasticity, I is the moment of inertia, A is the cross sectional area, o- is the yield stress.

o-yA
qcryielding = (3.1a)
2)2 + ()2

384EIh 4(1 - 4)32

qcrlIsnap-through 34 4 I 1 ++ ( 27 41AR2 )3(3.l1b) (
5L
AR2

2
EI {r
qcrI uckling = l - 1 (3.1c)

Failure by yielding is self-explanatory and less likely to happen, whereas the other two
failure modes require further explanation and are most likely to happen. In the case of
the shallow arch, the theory assumes that the arch is inextensionl, no axial deformation
is allowed and the buckled shape has an inflection point at the center of the arch. The
failure mode of snap through is for an arch of small curvature (essentially flat). In this
case, axial strain must be considered.
In order to determine the failure mode of the arch, first the variable that governs
the transition from one failure mode to another is established. Though there are four
geometrical parameters of the arch(R,a,h,L) two are sufficient to uniquely define the
arch and in this study the span-to-height ratio A becomes the independent variable. By
plotting the failure load versus the span-to-height ratio (Figure 3.6), the dominant failure
mode in a given range of the independent variable can be found.
As h A-± oc the arch becomes flat. It can be seen that the range of L/h for which
the dominant failure mode is buckling is L/h<135. The practical range of interest in this
dissertation is L/h<20, thus the arch fails first by buckling. While the above is a simple

39
Chapter 3. Methodology

analysis of a 2D arch, the lessons learned can be expanded to the failure modes of a 3D
shell. Previous research described in Chapter 2 has also concluded that global buckling
is the governing failure mode of shells.

Strain Energy Convergence

Having established the analytical solution to the failure of an arch, next the numerical
model is discussed. In verifying the accuracy of the FEM both the strain energy and
mesh convergence are checked. The limits of Timoshenko's small angle assumption is
numerically checked by calculating the difference in the buckling load between a circular
and a parabolic arch and between a normally and vertically distributed load. The current
studies use a steel arch of a span-to-height ratio of jh = 8 and a square cross-section of
5inx5in.
The sufficiency of the finite element mesh is evaluated by plotting the rate of conver-
gence of the strain energy to mesh refinement measured by - where 1
h is the element
length. The strain energy in both the continuum and the discrete model is given in Equa-
tions 3.2 and 3.3, respectively. Typically, the finite element model exhibits monotonic
convergence where the error decreases as the mesh is refined and converges from below to

Ac

Y ---- Buckling
- Snap-through
- Yielding
0.4
h
range considered
inthis dissertation
.3
a-
R
0
-J

U 0.1

(4 . 1 ~

0 10 20 30 40 50 60 70 80 90 100
L
Span-to-height ratio, L/h
Figure 3.5 - 2D geometry of the pinned- Figure 3.6 - The three competing failure
pinned arch. modes (buckling, snap-through, yielding) of
an arch as a function of the span-to-height ra-
tio.

40
 3.2. Computational Model Verification

the exact strain energy.

Eh = 2j rh~hdV (3.2)

Eh = -UTKU (3.3)
2 - -
To illustrate monotonic convergence consider the strain energy plot of a simply sup-
ported beam with a distributed load Figure 3.7). As the mesh is refined (increasing on the
x-axis), the strain energy quickly converges to the reference energy Erej. The reference
energy is either calculated by the exact solution, if available, or by the FEM of a highly
refined mesh. Because the plot shows that both a coarse mesh and a fine mesh converge
to the reference strain energy, then the coarse mesh is sufficient and the computational
expense of a finer mesh can be avoided.
For monotonic convergence the strain energy for coarser meshes cannot be greater
than Eref as a consequence of the FEM being "stiffer" than the continuum. Interestingly,
in performing the convergence study for the arch, a curved structure, the opposite is
witnessed; the strain energy converges from above.
In Figure 3.8 the strain energy convergence is plotted for an arch with a normally
distributed load. The reference mesh density was 1200 elements. For curved structures, a
coarser mesh is the equivalent to having a faceted curve where depending on the density,
the additional stiffness due to curvature is not captured. The coarser mesh will displace
more than the finer and, as a result, have a higher strain energy and converge from
above. The results from the strain energy convergence study reinforce that a continuous
curve can be represented as a faceted curve as is done in real structures. But how many
faceted elements are needed to accurately exhibit the geometric stiffness resulting from
curvature? The more coarse the mesh the less the structure exhibits geometric stiffness.
Figure 3.9 plots the deformed arch for both the coarse and the fine mesh. The less stiff
structure (the coarser mesh) displaces more for the same load. This is further seen in the
force-displacement plot in Figure 3.10 where the displacement at the middle of the arch
is measured as the mesh is refined, further explaining why the strain energy in Figure 3.8
converges from above.

Collapse Analysis

The conventional approach to determine buckling failure is to perform an eigenvalue anal-

ysis. There are two limitations to this approach. One is that buckling of shallow shells

41
Chapter 3. Methodology

6.8643

LU
.6.8643 .= 2.6

Cn 2.55
C 6.8643
u-i 2.5

C 2.45
'C&
4-06.8643 2.45

2.4

2.35

2.3 - - --- -- - -- - - - -
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
element length-1 , 1/1 h element length-1 , 1/lh
Figure 3.7 - Monotonic convergence of the Figure 3.8 - The strain energy convergence
strain energy for a simply supported beam of a pinned-pinned arch with a normally dis-
with a distributed load. tributed load.

z 0.045
LY
C 0.04
Undeformed FEM Deformed Arch
C
Q)
E 0.035
%

IA

- 0.03

0
C
~0.025
4-J
C
U
0.02

Number of elements
Figure 3.9 - Comparison of the arch de- Figure 3.10 - Force displacement plot as a
flection under uniform distributed loading be- function of increasing mesh density.
tween a coarse and a dense mesh.

42
 3.2. Computational Model Verification

is preceded by a nonlinear prebuckling phase; the eigenvalue analysis disregards the non-
linearity of the prebuckling phase. Secondly, the eigenvalue analysis cannot incorporate
structural imperfections. At the same time the equilibrium approach predicts the growth
of the initial imperfection and the magnitude of the ultimate load followed by a postbuck-
ling weakening phase. The differences between the equilibrium and eigenvalue approach
are studied on the 2D arch.
The theoretical buckling solution is compared with the linearized buckling analysis as
a function of the span to height ratio ; (Figure 3.11). Though Timoshenko's derivation is
based on shallow arches (small angle approximation), the buckling loads for steep arches
are calculated as well. For steep arches Timoshenko's theory has an error of less than
10% compared to the numerical solutions.
Because a FEM is stiffer than the continuum, the numerical frequencies are greater
than the theoretical. A continuum has an infinite number of natural frequencies because
every particle can move in infinite directions; whereas a FEM has only N natural frequen-
cies corresponding to the N degrees of freedom in the mesh. While the predicted FEM
frequencies are greater than the theoretical in Figure 3.11, within the range of shallow
arches number h <
-
20, the error is insignificant at a maximum of 3%.
The collapse load is next calculating using a Load-Displacement Constraint (LDC)
method to solve for the non-equilibrium path of the collapse and post-collapse responses
(Bathe, 1996). This method requires specifying the amount of geometrical imperfections.
As a common practice, the shape of imperfection used was the first buckling mode. Two
scales of the imperfect shape were considered: # = 0.05 and 1. The comparison between
the analytical, and both the linear eigenvalue and the nonlinear analysis is shown in Figure
3.12. There is a qualitative agreement between all three solutions. In particular, the linear
eigenvalue estimate is as accurate as both the analytical and the rigorous non-linear pre-
buckling analysis. It can be concluded that for practical applications the accuracy of
either the analytical or linear numerical solution is sufficient.

Small Angle Approximation

The small angle approximation of the Timoshenko closed-form solution is checked by com-
paring a circular arch to a parabolic arch and a normally distributed load to a vertically
distributed load. For shallow arches (' >6) there is little geometric difference between a
circle and a parabola as well as little difference in their buckling capacity (Figure 3.13).
This figure also compares the buckling load between the normal and vertical distributed

43
Chapter 3. Methodology

0.04i. -U--- Numerical-Linear

_0.035

0.03

o 0.02 0

C
0.02
M
I
u.u 11~g

0.01
0 2 4 6 8 10 12 14 16 18 20
Span-to-height ratio, L/h Span-to-height ratio, L/h
Figure 3.11 - Comparison of the theoretical Figure 3.12 - Comparison of linear and non-
buckling load (Equation 3.1c) to the numeri- linear buckling analysis for two scaled magni-
cal linearized buckling analysis. tudes (#) of imperfection.

load. For both the circular geometry with either a normally distributed or vertical loading
and for the parabolic geometry with either the normally distributed or vertical loading,
there is little difference.

Load Equivalence

In preparation for modeling grid shell structures, it is important to confirm that the
analytical solution that assumes a pressure load can still be used for a grid shell when the
loading becomes point loads at the joints. The calculation of an equivalent pressure is
provided in Equation 3.4 where P is the point load in kips, N is the number of nodes, and
S is the arclength in inches. In Figure 3.14, the theoretical buckling load for a uniformly
distributed load is compared to the numerical prediction where the loading is applied as
point loads at the joint. The distributed load is accurately represented with point loads
at a mesh density of 50 elements for an error of 9%.

qcr equivalent P xN (3.4)

44
 3.2. Computational Model Verification

0.045 - - -. 01
...... Theory
0.09 Theory
---. -...
0.04- Numerical-C, N 0.09
I
Theri
Numerical-Linear
-0 - Numerical-C, V 0.08
C 0.035 - -Numerical-P, N 007
-A--Numerical-P, V
0.03 0.06
- 0.05
o 0.025 0
-J -J0.04.

C C
0.02 0.03 ............... ............. ........

S0.02
0.015
0.01

0 2 4 6 01 12 1416 18 20 00 100 200 300 400 500

Span-to-height ratio, L/h Number of Elements
Figure 3.13 - Comparison of the buckling Figure 3.14 - Comparison of the theoretical
load for the circular (C) and parabolic (P) buckling load for a distributed load and the
arch with a normally (N) and vertically (V) numerical buckling load for point loading.
distributed load.

3.2.2 Grillage Study

In addition to the finite element modeling concerns investigated earlier on the 2D arch, a
grillage is also studied to clarify how to model rigid connections so to transfer moments.
Rigid ends are modeled within ADINA by identifying the location of the connection based
on- the element number, and inputting the rigid connection's length and stiffness. The
physical model and the finite element representation of a rigid joint is illustrated in Figure
3.15. The rigid end option assumes the the rigid ends will never undergo plasticity and
that the element is formulated as one single element. The rigid end lengths used in all the
models of this dissertation are equal to the thickness of the members which is a typical
length used for welded connections.
A simple bending case study is done to verify the modeling of the joints. A 1200inx1200in
steel cantilever plate with a tip load is modeled as an assemblage of beams forming a
quadrilateral grid (Figure 3.16). Two analytical models are used to calculate the de-
flection. The first assumes the grid to deflect as one cantilevered beam by making the
connections rigid (Figure 3.16a). The second assumes the assemblage becomes a wide
plate with a thickness of 0.8in defined by having an equivalent moment of inertia to the
beam model (Figure 3.16b). The beam has a point load of 1kip and a square cross section

45
Chapter 3. Methodology

Physical Model Finite Element Representation

I-S node
Iend length

Figure 3.15 - Finite element representation of rigid ends.

of 25in 2 . The plate has a distributed line load of 0.025k/in 2 with a thickness of 0.8in.
Using the deflection of a cantilevered beam with a tip load (Equation 3.5), the analytical
deflection is calculated and compared to the numerical results for both the beam and
plate models in Table 3.1. There is good agreement which verifies that the plate/shell is
an assemblage of beams with moment connections at the joints.

PL 3
wo = 3 (3.5)
3EI

A A
D PRESCRIBED PRESCRIBED
FORCE LINELOAD x ',
N
1.0 k I
N
A A 0.02500 k/in !

(a) Cantilevered grillage (b) Cantilevered plate

Figure 3.16 - Grillage case study.

46
 3.3. Summary

Table 3.1 - Comparison of the analytical and numerical tip deflection w, using both the beam
and plate models for the grillage.

Beam Plate
Analytical [in] 381 11,460
Numerical [in] 300 11,035

3.3 Summary
This chapter outlined the methodology for calculating the change in load-bearing capacity
of grid shells due to their curvature and topology using numerical and analytical methods.
In preparation for the grid shell parametric studies a mathematical model for which an
analytical exact solution is known, the 2D arch, was first studied and compared using
numerical methods. This builds confidence and validates the methods that will be used
as the models have more complicated features introduced to them. The computational
model verification studies confirmed that:

" the geometric stiffness of an arch is accurately represented by a faceted arch with a
mesh density of 50 the arclength.

" a linearized buckling analysis is sufficient for predicting the collapse load.

* moment connections can be accurately modeled using rigid ends in ADINA.

In the next chapter, the computational techniques verified here will be employed in the
parametric study of a spherical cap grid shell where the effect of curvature and topology
on the load-bearing capacity will be studied and quantified.

47
Chapter 4

Spherical Cap Grid Shell Study

4.1 Introduction
There is a vast body of literature touching upon the failure mode and ultimate carrying
capacity of grid shells. However there is a lack of simplified tools that would assist
in early design of such structures for maximizing their strength and/or minimizing their
weight and in facilitating the design discussion between architects and engineers. Previous
works reviewed in Chapter 2 solve only specific boundary value problems and thus fail to
contribute to the understanding of the underlying physics.
Specifically, the mechanics of lattice structures are controlled by the transition between
pure membrane response in the prebuckling domain to the combined bending/membrane
response at the point of buckling and in the post-buckling range. This chapter identifies
groups of parameters that are responsible for a given lattice structure reaching its critical
load and studies their influence using both analytical and numerical techniques. In this
connection the following global questions from Chapter 1 are revisited for spherical caps:

" How do the curvature and topology affect the load-bearing capacity of grid shells?

" How to best describe the mechanics of these structures?

* What are the design rules of thumb to maximize grid shell performance?

More specifically, here it is asked:

* What is the effect of topology, span-to-height ratio and grid spacing on the load-
bearing capacity?

48
 4.2. Problem Definition

* Can a closed-form solution be derived that accounts for the varied parameters?

4.2 Problem Definition

In order to understand how grid shells behave and to find a fast and simple analytical
estimation on the buckling capacity, a case study on a spherical cap grid shell is performed.
A spherical cap shape is a natural choice because it is a simple geometry from which more
complicated shapes can grow and because it has been seen in existing grid shells (as shown
earlier) such as the Swimming Arena in Neckarsulm, Germany (Holgate, 1997).
The buckling load is calculated analytically and numerically for a pinned supported
spherical cap under uniform pressure (Figure 4.1a). The members have a solid square
cross-section, are made from steel and have rigid connections at the joints. The load is
applied as vertical point loads at the joints (Figure 4.1b). The parametric case study is
framed around the parameters that influence most the architectural design (aesthetics)
and the engineering design (capacity). The analytical and numerical buckling loads are
then analyzed and compared.

z
Y

S
L S
(a) Spherical cap geometry (b) Member cross-section

Figure 4.1 - Spherical cap grid shell.

49
Chapter 4. Spherical Cap Grid Shell Study

4.2.1 Parameters
The three selected parameters that frame the spherical cap study are the panel shape,
the panel spacing, and the span-to-height ratio. These parameters are selected based on
the main design constraints explained in Chapter 2. The panel shapes considered are the
quadrilateral and the triangle with a spacing of s=30, 60, 120 in. The span-to-height
ratios are =20, 11, 6. The member's square cross-sectional area is A = tixt 2 and is
kept constant in this study at 25in 2 . Every permutation of the parameters, schematically
shown by the hierarchal tree of Figure 4.2, is considered, resulting in a total of 30 grid
shell models. Figure 4.3 shows more clearly all the combinations of the panel shapes and
spacing.

4.2.2 Numerical Approach

The finite element models for every permutation is created by a Matlab program written
by the author and provided in Appendix A. The grid shell models for the triangular
topology are partially built using a Matlab mesh generator code (Persson and Strang,
2004). The mesh generator uses a force-based method and Delaunay triangulation to
create an equally spaced grid. All 30 models are imported into ADINA where a linearized
buckling analysis is used to calculate the buckling load and mode.
Each member is modeled as an ADINA beam element (a 2-node Hermitian beam

Figure 4.2 - A hierarchal tree illustrating the number of models for the spherical cap study.

50
 4.2. Problem Definition

Grid Density
1Ox10 20x20 30x30

Quadrilateral

-5F

0~
0

Triangular4

Figure 4.3 - Illustration of the six combinations of the panel shape and spacing.

with 6 degrees of freedom at each node 3 translations and 3 rotations). Each member
has a mesh density of five. The joints are modeled as rigid connections as explained
in Section 3.2.2. Lastly, the perimeter members are pinned, and the loading is applied
vertically to the joints. One of the 30 models is shown in Figure 4.4, highlighting the
mesh density, boundary conditions and the loading.
In addition to the grid shell finite element models, a continuous shell is also created
in order to validate both analytical and numerical techniques. Two types of continuous
shells are defined: one employs the axisymmetric beam, and the other uses shell elements.
The models and their results will be described further in the Model Verification section.

4.2.3 Analytical Approach

One way to converge fast to an optimized solution is to make better use of existing
analytical solutions for axisymmetric shells that have uniform thickness. Therefore, the
key issue in using the analytical solution, for the purpose of establishing a practical design
tool, is a proper definition of the equivalent thickness describing the transition from
the discrete grid shell to the continuous shell. The difficulty in defining the equivalent
thickness is in understanding the interplay between the in-plane membrane response and
the out-of-plane bending response because the bending and axial stiffness include shell

51
Chapter 4. Spherical Cap Grid Shell Study

A mm"C14
NI
A

Figure 4.4 - Example of the spherical cap grid shell finite element model with L=6.

thickness raised to a different power. The issue is to decide what is more necessary to
have: sufficient cross-sectional area to withstand the membrane forces or enough moment
of inertia to withstand the bending?
Consider the spherical shallow shell with pin supports and a uniformly distributed
pressure shown earlier in Figure 4.la. The closed form solution of the buckling load by
Timoshenko, given in Equation 4.1, is obtained assuming an axisymmetric buckling mode
(Timoshenko, 1961). The radius of the spherical cap is defined by R and V is Poisson's
ratio. In the present application of this equation, teq is a thickness of an equivalent uniform
shell understood in a certain sense to be equivalent to the grid shell.

qc,= 2 (4.1)
V/3(1 -v2) R

In this research multiple definitions of the equivalent continuum are compared which
not only facilitate finding simpler approximate solutions, but also fills the knowledge
gap in understanding how these structures are behaving; specifically, their transition
from bending to membrane action. Developing this understanding helps to better design
structures in terms of weight efficiency.
In addition, the concepts reviewed in Chapter 2 and reviewed in Chapter 2 are ex-
tended here by validating various equivalent thickness methods with discrete models
(Wright, 1965; Forman and Hutchinson, 1970; Kollar and Dulacska, 1984). While one

52
 4.2. Problem Definition

paper uses discrete analysis to evaluate their equivalent continuum buckling prediction,
it was limited to the tools available in 1970 (Forman and Hutchinson, 1970). The au-
thors acknowledged that their method was to be further validated by experiments. Today
computational software facilitates completing this much needed parametric study.
Furthermore, the results in this thesis are uniquely presented in a form applicable for
design. This is done by including the parameters that influence the design in the simplest
form so as to be accessible to both architects and engineers. The previous works also did
not validate the equivalent continuum technique in terms of the design parameters like
topology. It can be concluded that there is a lack of well validated methods for a rapid
estimation of the failure of grid shells needed for preliminary design.
There is no unique method to calculate the equivalent thickness based on the topology
of the grid shell, the spacing and the member cross-section. Figure 4.5 highlights the
geometry of the repeating grid shell unit for a quadrilateral and triangular topology.
Equivalent thickness could be determined on the bases of (i) area equivalence, (ii) moment
of inertia equivalence, (iii) volume equivalence or (iv) a combination of area and moment
of inertia. The area, moment of inertia and volume equivalence are self-explanatory. In
the case of the concept of area and moment of inertia, the Timoshenko solution is valid;
however, the derivation of the final expression is more complicated. It requires rederiving
the spherical cap buckling load while keeping the thickness due to the axial rigidity and to
the bending rigidity separate. The derivation is provided in Appendix B. The summary
of the equivalence thicknesses derived in this dissertation using the above four methods
is listed in Table 4.1. Note similar expressions can be derived for different cross-section
geometries.

Figure 4.5 - The quadrilateral and triangular topology and their repeating units (the cruciform
and the hexagon).

53
Chapter 4. Spherical Cap Grid Shell Study

Table 4.1 - The equivalent thickness (teq) relations for the quadrilateral and the triangle using
the four equivalent definitions.

Definition of Equivalence Quadrilateral Topology JTriangular Topology

Area (Acq) St

Moment of Inertia (Ieq) (iq

Area & Moment of Inertia (Aeq + Ieq) t ti t S

Volume (Vq) 2tit2 stit2

4.3 Results

4.3.1 Model Verification

Before proceeding with the grid shell FEM, the computational model and the range of
analytical validity is verified first by modeling a continuous shell. The analytical model is
Timoshenko's buckling equation for a shallow spherical cap (Equation 4.1). Two numerical
models are considered: one using axisymmetric isobeam elements for < 20, and the
other using shell elements for only = 20. The latter model was used primarily to check
the load equivalence method described in Section 3.2.1. Both models have a thickness
t = 3in.
For the range of shallowness established in Chapter 3 ( < 20), the numerical results
have a maximum of 2% error at = 15 (Figure 4.6). In Chapter 3, the sources of
error between the finite element solution and the analytical were already discussed. Here,
another source of error is that the first buckling mode is assumed to be axisymmetric
in the analytical solution; however, the axisymmetric shell element predicts a symmetric
buckling mode. Because the corresponding buckling load is only 0.3% error from the
analytical this error is negligible. In the model that uses shell elements the first buckling
mode is axisymmetric and has with a 0.3% error for the buckling load.
In Section 3.2.1 the load equivalence method is checked where the buckling load is
calculated by applying point loads instead of a distributed. Here, that study is repeated
for the spherical cap. In Figure 4.7, the buckling load for the continuous shell using
distributed loading and point loading is compared. For an inverse element length of 0.025
the point load predicts the theoretical buckling load with a 5% error.

54
 4.3. Results

1 100

Theory Pressure load

Numerical Point load
0.850

Ix 0.6
(U0 0
0 7

0.4 mo

-50

0.2

01 -------- -100 0 0.01 0.02 0.03 0.04 0.05

0 5 10 15 20
Span to height ratio, Uh Element length-', lh 1 [1/in]

Figure 4.6 Comparison of the numerical Figure 4.7 - Error in the numerical buck-
and the theoretical buckling load of a contin- ling load to the analytical when the loading is
uous spherical cap using axisymmetric shell applied as a point load and a pressure load.
elements.

4.3.2 Collapse Analysis

As explained in Section 3.2.1 the linearized buckling analysis does not account for prebuck-
ling rotations. In this section the critical load is calculated using the nonlinear collapse
analysis and is compared to the linearized buckling analysis for a spherical cap of L = 20.
The concern is if the linearized buckling load prediction overestimates the collapse or if
the grid shell's collapse load is negligible because the deformed shape gains more stiffness.
In this study the collapse analysis is induced by an imperfection in the z-direction at
the center equal to 0.lin or 6i = j. In Figure 4.8, the pressure versus displacement curve is
plotted. The collapse load is 15% less than the predicted linearized buckling load. Because
the error can vary as seen in Section 3.2.1, a linearized buckling analysis is still used in
this case study because it is computationally faster to both set up and run. Also by doing
a collapse analysis other complications arise, for example like what and where should the
imperfection be and how large? Previous research has looked at imperfection sensitivity
for shell buckling analysis (Singer and Abramovich, 1995; Calladine, 1995; Blachut and
Galletly, 1995; Bernard et al., 1999). After using the trends concluded in this thesis to
guide the schematic design of a grid shell, the research on imperfection studies can be
used in the robust analysis needed for the final design.

55
Chapter 4. Spherical Cap Grid Shell Study

- - Collapse
0.025

0 2
0.015
M
0.01
0-

0.005

0 5 10 15 20
Center node displacement, 6i [in]
h0
Figure 4.8 - Comparison of the collapse load to the linearized buckling load for a spherical cap
of ;= 20.

4.3.3 Numerical Results

Recall that the objective is to study the change in buckling capacity as a function of the
shell height, the panel shape and spacing and the shell height. As an example of the FEM
output of the mode shape, the buckling mode for both the quadrilateral and triangular
grid shell are shown in Figure 4.10.
The change in the buckling load for both the quadrilateral and the triangular grids is
plotted as a function of the grid spacing in Figure 4.9. The three span-to-height ratios
heights are distinguished by their markers. For both the quadrilateral and triangular
grids, as the spacing increases the capacity decreases and all the models converge to the
same capacity regardless of their height.
As expected, the steeper the shell the higher the capacity because the buckling load
is proportional to y. The triangular grid is stronger than the quadrilateral especially as
the shell gets steeper. However to properly account for the trade-off between the triangle
and the quadrilateral, the volume of material is accounted for next.
A metric of performance is defined as the structural efficiency 7 to measure how much
a structure can carry of its own weight q,,(force per unit area) (Equation 4.2). In Figure
4.11 the structural efficiency of each structure is plotted. Note that for all models the
structural efficiency is always greater than one; the grid shell spherical cap can always

56
 4.3. Results

Quadrilateral Topology Triangular Topology

0.3
e
0.25
b
0.2

-'0.15

0.1

0.05
L..-
MEI
0 50 100 50 100
grid spacing, s [in] grid spacing, s [in]

Figure 4.9 - The buckling capacity of the spherical cap as a function of the grid spacing for
the quadrilateral grid (left) and the triangular grid (right).

Figure 4.10 - First buckling mode of spherical cap with h = 20 and a spacing of s = 30in:
quadrilateral topology (left), and triangular topology (right) (magnified by a factor of 5000).

57
Chapter 4. Spherical Cap Grid Shell Study

carry more than its weight.

r7- (4.2)
qsw

Figure 4.11 also shows that for shallow caps a quadrilateral grid is as efficient as the
triangle. However, as the shell becomes steeper, the triangle grid is up to two times more
efficient than the quadrilateral. The plot indicates that different design configurations
can have the same structural efficiency. For example, for the structure to carry 50 times
its own weight, there are four different grid shell geometries. Two options include either a
shallow, quadrilateral, dense grid or a steep, triangular, coarse grid. Therefore a designer
can consider more options when trying to achieve an efficient shell structure.
Quadrilateral Topology Triangular Topology
40 400
-+L/h=20
3 350 -e-L/h=11
30 1-4-L/h=6
5 300c: 3 '- 300
2. 25

u2 w20
10 150
100 100

5
so 100 50 100
grid spacing, s [in] grid spacing, s [in]
Figure 4.11 - Structural efficiency of the spherical cap grid shell (Equation 4.2).

4.3.4 Analytical Results

The analytical solution using Equation 4.1 for a quadrilateral grid of = 20 is plotted
in Figure 4.12. The four lines represent each of the four different equivalent thickness
definitions (area, moment of inertia, area and moment of inertia, volume) in Table 4.1.
The buckling load calculated from the four equivalent continuum methods are compared
as a function of the grid spacing s.
The figure shows that the upper and lower bound of the buckling load correspond to
the equivalent moment of inertia approach and the equivalent area approach, respectively.

58
 4.3. Results

It is expected for the extremes to indicate whether a structure is bending dominated or

membrane dominated. In between the bounds are the volume equivalent and the area and
bending equivalence. As shown more clearly in the figure, the area and bending equivalent
is not the average of the area equivalence and the bending equivalence.

4.3.5 Numerical and Analytical Results

Figure 4.13 plots both the numerical and analytical solutions for all three :h ratios and for
the two topologies. The matrix of plots is organized from left to right by decreasing span-
to-height ratio, and from top to bottom as the quadrilateral and the triangular topology.
As explained earlier, the analytical solution for the area and for the moment of inertia
equivalence will always give, respectively, the lower and upper bound solution. Here all
the numerical results fall within those bounds. Placing the results within the context
of those bounds, the shallow shells are seen to be more membrane dominated while the
steeper shells are more bending dominated. This is clearly seen on the figure where the
numerical solution is closer to either the moment of inertia or area equivalence depending
on the height.
The plots also show that for quadrilaterals the equivalent volume is best and for

Quadrilateral, Lh=20
0.07
eq
0.06 ....... Aeq+1I
eq
S0.05

S0.04
o n s
-' 0.03

0.02

0.01

C
0 50 100
grid spacing, s [in]
Figure 4.12 - Comparison of the buckling load (Equation 4.1) for the four equivalent continuum
definitions for a quadrilateral grid with L=20.

59
Chapter 4. Spherical Cap Grid Shell Study

triangles the best approach goes from area to area/bending to bending as the shell becomes
more steep. It is expected that the shallower shells trend toward membrane action,
requiring more axial rigidity because the axial force is greater the more shallow the shell.
The denser grid has a stronger buckling capacity which is expected because a more dense
grid converges to a continuous shell. It should also be noted that although the buckling
mode shape of the theoretical solution is axisymmetric and that the buckling mode in
the numerical solution for the triangle is symmetric, the corresponding buckling loads
are quite close to each other. This leads to the conclusion that the simple axisymmetric
Timoshenko equation, modified to account for an equivalent thickness, can be used to
estimate of the buckling load of grid shells. For quadrilaterals, the upper bound is best
defined by using the equivalent volume. For the triangle the volume equivalence is used
up to a spacing of 40in. For spacing less than 40in the combined area and moment of
inertia equivalence provides the best upper bound.

4.4 Discussion
The opportunity for architects and engineers to create weight efficient elegant structures is
in the preliminary stages of design. Here a framework is presented for a fast assessment of
grid shell performance in terms of the main design constraints (grid spacing, shell height
and span). The results of the parametric study performed on those variables are combined
into a set of design guidelines presented in Figure 4.14.
The guidelines are organized into three categories based on the three span-to-height
ratios. For the more shallow shells (the left column), it is more effective to decrease the
grid spacing than to change topology. Whereas for the steeper shells (the right column)
it is best to change the topology. In between those two extremes, the effect of changing
the grid spacing or the topology is similar and one can do either. The design guidelines
provide an immediate answer on how to improve the buckling capacity of the spherical
cap grid shell.
The notion of the equivalent thickness is not uniquely defined. The best equivalent
continuum approach varies between topology and curvature. There are significant differ-
ences in buckling prediction between the four equivalent continuum approaches. While a
simple bending and area equivalence give an upper and lower bound, the spread is very
wide. This further illustrates that grid shells respond both in membrane and bending
action. At the same time the volume equivalence and the combined area/bending equiva-

60
 Quadrilateral, L/h=20 Quadrilateral, Uh=1 1 Quadrilateral, L/h=6
0.07 ''~
0.71
Numerical
0.061 0.6} -- leq
0.2
...eqleq
0.05 0.5 -- V
eq
0.04' 0.15
-'
. 0.4 i-
re
-j 0.031 -A
a

0.02t

0.01
-. . .
o

m 0.05
0.1

.
'--
0.3

0.2

0.1
I %-

.
..................

0 0'0
0 50 100 00 50 100 50 100 -

grid spacing, s [in] grid spacing, s [in] grid spacing, s in]

Triangular, Lh=20 Triangular, Lh=1 1 Triangular, Lh=6
0 .1 0.5 0.5
Numerical
--- I
eq
0.0 82 0 .4 I 0.4 .....A +l
6
A,
eqe
--Veq
0.0 0 .3 0.3 -- A
0 a 0
a 0.04 1
0 .2
-j0
4M0 .2
S
32

m 0.0 S0..1 . .... . - -. 01

2n
n -- - --.

0 50 100 ''0 50 100 "0 50 100

grid spacing, s [n]l grid spacing, s [n)l grid spacing, s in]

Figure 4.13 - Comparison of analytical and numerical results for the spherical cap grid shell study. The rows from top to
bottom represent the topologies quadrilateral and triangular, respectively. The columns from left to right represent decreasing
span-to-height ratios. 0
0l
Chapter 4. Spherical Cap Grid Shell Study

Decrease grid Decrease Select a Select a

spacing grid triangular triangular
(ma),x of 40in)) spacing g rid L grid

Figure 4.14 - Design guidelines for a spherical cap grid shell. Columns from left to right
represent decreasing span-to-height ratio.

lence leads to much closer estimates when compared to the exact numerical solution. The
accuracy of the volume equivalence is better for the quadrilateral grid while the combined
area/bending equivalence gives better results for triangular meshes.

4.5 Summary
This chapter developed a framework for a quick and accurate assessment of grid shell struc-
tural performance by using available closed form solutions of continuous shells. Though
the analysis in this chapter is restricted to the spherical cap, the salient features can be
extended to more general types of grid shells. By expanding the use of the assessment of
simple shell and grid structures to complex ones, an efficient approximation is provided
which can immediately impact the design.
The central point of the method is the concept of an equivalent uniform shell where
the equivalence is based on the bending or axial rigidity, or both, or volume. This chapter
derived and validated multiple equivalent thickness techniques. The area and moment of
inertia equivalence establish the upper and lower bounds of the critical failure load.
The vast amount of data compiled and analyzed not only better explains the mechanics
of grid shells but also lays the groundwork for providing the following rules of thumb at
the start of a grid shell design.

9 The effect of topology lessens as the shell become shallower. The more shallow the
shell, the less the benefit in using a triangular topology over a quadrilateral.

62
 4.5. Summary

" Steeper shells have a higher capacity than shallow shells.

" Shallower shells require a denser grid to increase the capacity.

" An equivalent continuum defined by area equivalence provides a conservative esti-

mate for the buckling load of a grid shell that can be used in the preliminary stages
of design.

" For the quadrilateral grid the upper bound for the buckling load is defined by the
equivalent volume. For the triangular grid with spacing greater than 40in, the
upper bound is defined by the equivalent volume; for spacing less than 40in the
upper bound is the combined area and moment of inertia equivalence.

" Different design configurations of the span-to-height ratio, the spacing and the topol-
ogy can have the same structural efficiency.

63
Chapter 5

Corrugated Vault Study

5.1 Introduction
Most grid shell structures have a rectangular footprint leading to the use of a barrel vault
shape. Some geometries have gone so far as to add corrugation to the surface as seen in
the Japan Pavilion, the Weald and Downland Museum (Figure 5.1). In this chapter, the
role of curvature on a barrel vault is studied by gradually introducing double curvature
into the geometry.
In Chapter 4, the spherical cap grid shell study provided us with a rapid estimation
of grid shell capacity by using the equivalent continuum. In that chapter, the objective
was to understand the mechanics while quantifying how the most influencing real design

(a) Weald and Downland (b) Japan Pavilion

Figure 5.1 - Examples of grid shells with corrugation at the edge and the crown.

64
 5.2. Problem Definition

constraints affect the capacity and efficiency of the structure. In this chapter, the focus
is on the role of double curvature for efficiency and capacity. Here the following key
questions are asked:

" Is corrugation an effective and efficient means to increase the load-bearing capacity
of a barrel vault?

* Can a closed-form solution be derived that accounts for the varied parameters?

To answer these questions, the change in buckling capacity of three corrugated barrel
vaults is studied. The varied parameters include the location of corrugation (edge, crown
or both), the aspect ratio of corrugation, the span-to-height, and the grid spacing (Figure
5.2).

5.2 Problem Definition

5.2.1 Geometry
The dimensions of the straight edge barrel vault are a span Li of 100ft, and a longitudinal
length of 2L 2 of 162ft (Figure 5.3a). The barrel vault has a a parabolic cross section defined

Figure 5.2 - The four groups of parameters: corrugation location, aspect ratio, span-to-height
ratio, and grid spacing (left to right).

65
Chapter 5. Corrugated Vault Study

by Equation 5.1. The span Li and the length 2L 2 are kept constant in all models at 100ft
and 162ft respectively. These length values were chosen based on typical values of existing
grid shells (Cadji, 2001; Stungo, 2001; Pople, 2002b,a, 2001). The vault is loaded with a
uniform pressure and its boundary conditions include no lateral and normal translation
along the length of the edges and no longitudinal displacement and rotation about the x
axis (Figure 5.3b).

Z = h1-(5.1a)

xo = (5.1b)
xO72

5.2.2 Parameters
The four groups of parameters defining the parametric space are:

" the corrugation location: at either the edge, the crown or both the edge and the
crown (in-phase)

" the aspect ratio of the corrugation: 8 values total

" the span-to-height ratio: L=6, 11, 20

" the grid spacing: 10x1O, 20x20, 30x30

As in the case of the spherical cap study, every permutation of the parameters was con-
sidered, overall there were a total of 390 models. This is illustrated by the hierarchal tree
in Figure 5.4.

z u=0 X

x Y=0 zLY
LIXY __ __ _

U =0 U =0
u =0 u =0

07ft u =0
16 2 ft 6,1r/I0Y=0
(a) Dimensions (b) Boundary conditions

Figure 5.3 - Straight edge barrel vault geometry and boundary conditions

66
 5.2. Problem Definition

Figure 5.4 - The hierarchal tree illustrating the number of corrugated grid shell models.

The undulations at the edge and crown are defined by a cosine function and the
surface is described by Equation 5.2. The two variables defining the undulations are
the number of waves n along the length L2 and the wave amplitudes ai and a2 . The
corrugation amplitude at the edge (ai) is defined as a percentage #1~ of the span; the vault
is corrugated at the edge and tapers to a fiat crown. The corrugation amplitude at the
crown (a 2 ) is defined as a percentage #32 of the height; the corrugation is maximum at
the crown and tapers to a straight edge at the boundary (Figure 5.5 and Equations 5.3
and 5.4). In the third case where the edge and the crown are corrugated ai f a2 ; thus,
the corrugation varies along the arclength. The amplitudes are in-phase and this case of
corrugation at both the edge and the crown will be referred to as the in-phase case.

2nyry
x=zo+aicos

z =(h+ a 2 acosj (0-((5.2)

1

ai)(5.3) = (
100 2

67
Chapter 5. Corrugated Vault Study

a2 = h (5.4)
100

The two values of #1 and #2

are 5% and 10% and the number of waves n per half
length are 0.5,1,1.5,2. The total number of combinations of n and fi is eight and Figure
5.6 illustrates all the combinations. To simplify and minimize the number of parameters,
#j, and n are consumed into Aa which is defined as the aspect ratio of the corrugation
1

where A = As y oo there is no corrugation and the geometry is flat. Multiple

-+

combinations of n, and #3 can yield the same -L. In Figure 5.7 the change in corrugation
in terms of ) for the corrugated edge is illustrated.

5.2.3 Methodology
To tackle this complicated problem, the study is split into two sections. The first is to
study the role of curvature on the continuous shell and to validate the modeling techniques
used. The second part is to repeat the analysis for grid shells. The procedural steps for
both the continuous and grid shell are outlined below, beginning with the definition of the
geometry and parameters and concluding with the analysis of the numerical models. In all
of these studies the numerical results help identify the structural trends while contributing
to the understanding of the mechanics behind grid shells.

Procedural Steps

Continuous Shell:

1. Present the theoretical buckling load of a barrel vault with straight edges.

82

2L2 5L

Figure 5.5 - Corrugated barrel vault geometry.

68
 5.2. Problem Definition

n=2 n=1.5 n=1 n=0.5

=10%

Figure 5.6 - IIlustration of the eight combinations of n and 3j for the corrugated edge.

11 11 11 11
I ----------h.
I A/a

Figure 5.7 Geometric representation of the corrugated edge as a function of a,L.

69
Chapter 5. Corrugated Vault Study

2. Identify parameters describing the double curvature.

3. Construct the continuous shell finite element model of the basic barrel vault.

4. Run the buckling analysis and compare to the analytical solution.

5. Run the collapse analysis and compare to the linearized buckling load.

6. Construct the 100 finite element models in the parametric study.

7. Run the buckling analysis.

8. Plot and analyze results.

Grid Shell:

1. Establish the parameters and ranges that define the grid shell curvature and topol-
ogy.

2. Define the dimensions based on an equivalent volume between the continuous and
the grid shell.

3. Validate the straight edge grid shell with the analytical solution.

4. Proceed with the parametric study constructing 300 grid shell models.

5. Plot and analyze the results.

5.2.4 Analytical Models

In preparation for processing the numerical results, the mathematical representation of
the variables affecting the capacity are discussed. No closed form solutions have been
derived for the corrugated barrel vault. Here an attempt is made to build an analytical
model starting from the plane strain analysis of the straight edge barrel vault.
A plane strain analysis is typically used for structures long in one direction such as
a dam. The structure can be simplified to a slice of unit thickness because the strain
components Eyy, 7zy, and Tyz are zero with the axes defined in Figure 5.8. In the case
of the barrel vault, the unit slice is a parabolic arch, as the one studied extensively in
Chapter 3.

70
 5.2. Problem Definition

Because the barrel vault span-to-height ratios are relatively shallow, the unit slice is
expected to fail by buckling. Thus the arch buckling equation presented in Chapter 3 is
modified to a force per unit area in Equation 5.5. The lateral strain is accounted for using
Poisson' s ratio. /2

qcrlbarreivault =R3 (,2 i b( (5.5)

While Equation 5.5 is accurate for a basic barrel vault, the corrugated vault requires
more modification. By corrugating the vault the cross-section becomes curved (Figure
5.9) and the moment of inertia is different from the flat section. Thus Equation 5.5 is
modified to Equation 5.6 where the moment of inertia for the corrugated section Ieff is
derived in Equation 5.7. As the amplitude decreases to zero (ai -4 0), the geometry of the
cross-section becomes rectangular and Ieff = Irectangle. Although the actual corrugation
is a cosine function, here the moment of inertia is approximated by using a parabolic
function. The ratio of the buckling load of the corrugated vault with Ieff to the basic
cault can be simplified to Equation 5.8.

qcrkbarreivait - (b(1 v2 (5.6)

Z X
X Y

dx

1 unit

4N

Figure 5.8 - Plane strain condition. Figure 5.9 - Corrugated shell cross-section.

71
Chapter 5. Corrugated Vault Study

'eff JY2dA
0

(h(1 - tdx
0
3
2 2t 2( t
5 z 3 2
lim Ieff Irectangle (5.7)
aj -40

qcrlstraight A

(= (5.8)

As mentioned earlier the moment of inertia varies along the arclength. Timoshenko
derived the expression in Equation 5.9 that accounts for a larger moment of inertia at the
crown than the edge. The variables I and I1 are the moment of inertias at the crown and
the edge respectively, and m is a numerical factor depending on a, the angle of openness
and the ratio I. The numerical factor m is plotted in Figure 5.10. As the moment of
inertia goes to zero at the edge, the capacity decreases. In other words, if the arch became
thinner at the supports the arch is weakened and the capacity decreases.

I 0I1'- 1I- )0

qcr = m2R (5.9)

5.2.5 Numerical Models

The mechanics of the corrugated vaults are first studied using continuous shell models of
1 inch thickness. This validates the analytical approach explained earlier as well as the
computational modeling techniques.

72
 5.2. Problem Definition

10

a=0

....... a=30

Z 2 - 2
- a =600
. .- a=90 0

0
0 0.2 0.4 0.6 0.8 1

/1//0

Figure 5.10 - The effect of a varied moment of inertia on the numerical factor m for the
buckling load.

Continuous Shell

The FEM of the continuous shell is built using a program by the author (Appendix A)
that requires the input of the geometric parameters and outputs a point cloud defining the
surface. The surface grid is then imported into ADINA where the surface is created by
connecting the control points with a quadrilateral polygonal mesh (Engineering, 2010).
A less dense point cloud will produce a less smooth surface, and the surface becomes
approximately equal to that of Equation 5.2.
Within the ADINA interface, the surface is meshed using shell elements. As shown
earlier in Figure 5.3b, the boundary conditions include no lateral and normal translation
along the length of the edges and no longitudinal displacement and rotation about the x
axis. Because shell elements are prone to boundary layers due to incompatible boundary
conditions, the selection of proper boundary conditions was carefully considered (Chapelle
and Bathe, 2011). Finally, a surface pressure is applied, and a linearized buckling analysis
is used to calculate the buckling load as was done in Chapters 3 and 4.

73
Chapter 5. Corrugated Vault Study

Grid Shell

In the grid shell studies three grid densities are considered: 10x10, 20x20, and 30x30
(Figure 5.11). The grid shells are defined to be equivalent in volume to its continuous
shell geometric counterpart. The width and longitudinal spacing between the units are
denoted as sx and sy respectively. The member dimensions are shown in Figure 5.12.
To have an equivalent volume to the continuous and because the grid shell members
are solid squares, the cross-section dimensions are calculated using Equation 5.10 for the
quadrilateral topology. The calculated member thicknesses for an equivalent volume are
also provided in Figure 5.11; when the grid becomes finer the member thickness decreases.
As in the case of the spherical cap, sources of error in calculating the equivalent thickness
include: counting for the volume at the member intersection twice, disregarding the half
units at the boundaries, and using a planar projection for calculating the spacing.

tllIquad t2lIquad c (5.10)

sx + sy

1Ox10 grid 20x20 grid 30x30 grid

Figure 5.11 - The three grid shell densities used in the corrugated vault study (10x1O, 20x20,
30x30).

74
 5.3. Results

x
zL

i i i i 6 i 6

SY

S,

t
s

'S,

Figure 5.12 - Dimensions of the grid spacing and the overlay of the repeating unit of the grid
shell cruciform and the square continuous shell.

The finite element models of the grid shells are created using a Matlab program written
by the author and provided in Appendix A. The program defines the FEM by redefining
the nodes and element connectivity of the continuous shell to that of the joints and
members of the grid shell. The model is then imported into ADINA, and a linearized
buckling analysis is used to calculate the buckling load.

5.3 Results
The results are presented in the following subsections:

* Model Validation: the straight edge numerical models of both the continuous and
the grid shell are compared to the plane strain analysis.

" Continuous Shell: the gain in capacity by using corrugation is measured and the
results are compared to the analytical model.

" Grid Shell: the role of curvature, spacing and topology is analyzed using the nu-
merical results.

75
Chapter 5. Corrugated Vault Study

5.3.1 Model Validation

First the continuous shell of the basic barrel vault is validated. The buckling load predicted
by the numerical models is plotted against the analytical plane strain solution (Figure
5.13). The numerical results coincide with the analytical. Also, the first buckling mode
(magnified by a factor of 400 in Figure 5.14) is axisymmetric as is the assumed shape in
the analytical derivation.
The grid shell model is also first validated by plotting the analytical plane strain to
the numerical for all three grid densities as a function of the shell heights (Figure 5.'15).
The buckling modes for each grid density are plotted in Figure 5.16 with a magnification
factor of 1000. As in the case for the analytical and the continuous shell, the first mode
is axisymmetric.

5.3.2 Continuous Shell

The continuous shell results are presented separately by the corrugation locations. To
quantify the gains in adding curvature the buckling load of the corrugated barrel vaults
is normalized to the buckling load of the basic barrel vault. After the presentation of
the numerical results of each corrugation location, the analytical models are discussed.
Lastly, all three cases are compared in terms of the cost in volume. In all three cases

x 10.4
A MODE1 MODEMAG 400
D
- - Theory
Numerical
N XiY
A

0 20 40 60 80 100 120 140 160 180 200 220

Shell height, h [in]

Figure 5.13 - Plane strain validation for the Figure 5.14 - The buckling mode for the con-
continuous shell. tinuous shells magnified by a factor of 400.

76
 5.3. Results

0.01 A MODE1 MODEMAG 1000

D
-.0 1- Theory I
+ 10x10 N X! Y
0.008 - 20x20 A
A 30x30

t 0.006 1Ox10 20x20 30x30

0
0.004 -

Co

0.002

0 - -
0 50 100 150 200
Shell height, h [in]

Figure 5.15 - Plane strain validation for the Figure 5.16 - The first buckling mode of the
grid shell. straight vault for each of the three grid shell
densities (magnified by a factor of 1000).

there are significant gains by introducing double curvature.

Corrugated Edge: The gain in capacity of the corrugated edge is plotted as a function of
the aspect ratio --
al
in Figure 5.17. First consider the two extremes of the x-axis. As gL
ai
approaches infinity the corrugation is non-existent and it becomes a straight edge vault.
Thus, the gain in capacity (normalized by the straight edge capacity) converges to unity.
At the other extreme when - approaches zero, the corrugation is "tight" and the gain in
capacity ranges from 13 to 35 times the capacity of the straight edge as the span-to-height
ratio varies.
In between the two extremes, there is a clear inverse power trend in the data. To
understand this trend, in Figure 5.19 the buckling modes(magnified by a factor of 500)
are plotted. From top to bottom the aspect ratio of corrugation decreases. All the vaults
exhibit the axisymmetric mode as seen in the basic barrel vault. However, it is evident
that as the amount of corrugation increases the vault develops a shorter effective length.
In essence, the vault maintains the same radius of curvature but the angle of openness a,
becomes smaller and smaller as the vault decreases its aspect ratio.
The relationship between the effective angle aeff can be calculated from Equation 5.11
using the numerical data and was plotted as a function of the aspect ratio y.
al
A curve

77
Chapter 5. Corrugated Vault Study

35 A - 50
L/h=20 [ Lh=20
30 . L/h=11 .L/h=11
A L/h=6 40 L/h=6
25
M
30
201
W LM A

b 1 A COb20
CC-

10 A 0 A

5 AsA

0 0A*- - -
0 5 10 15 0 50 100 150
A/a1 A/a2
Figure 5.17 Gain in capacity from the Figure 5.18 - Gain in capacity from the
straight barrel vault by corrugating the edge straight barrel vault by corrugating the crown
of the continuous shell. for the continuous shell.

was fit to the data for each of the span-to-height ratios and the fitted equations are given
in Table 5.1. These equations can be used for calculated the buckling load of a corrugated
edge shell. If the span-to-height ratio is known as well as the desired aspect ratio, the
effective angle can be calculated and plugged back into Equation 5.12 to calculate the
buckling load.
aeff (5.11)
rqcr (12R3(12)
Et3 0+

(b1
qcrle=ge= E 3 (_e2
eg R ka2
-
7 .b(1 V2)/
(5.12)

Corrugated Crown: For the corrugated crown, the gain in capacity is plotted as a function
of y, the aspect ratio of the corrugation at the crown (Figure 5.18). As approaches in-
finity, the geometry approaches a straight barrel vault and the gain in capacity approaches
unity. As a2'\ approaches zero, the gain in capacity ranges from 20 to 50 times the capacity
of the straight vault as the span-to-height ratio varies from 20 to 6 respectively.

In-Phase: The final case considers the vault corrugated at both the crown and the edge.

78
 5.3. Results

z
A MODE 1 MODE MAG 500.0
D x

N
A
n=0.5, P =5%

n=0.5, 10%

n=1, 1=5%

n=1, 1 =10%

.....
.....
n=1.5, 1p=5%

n=1.5, 1=10%

n=2 1=5%

1-1...
...
...
....
...
.....

n=2 1=10%

Figure 5.19 - First buckling mode for the corrugated edge case with increasing corrugation
from top to bottom (magnified by a factor of 500).

79
Chapter 5. Corrugated Vault Study

z
Y

eff
...

Figure 5.20 - Illustration of the effective arclength aeff.

80
 5.3. Results

Table 5.1 - Equations for calculating aegg given a span-to-height ratio and an aspect ratio of
corrugation.

Oaeff

20 -0.002 l +0.23( ) 2.24

1
61 -0. 005
-+al0 A '+0. 41( + 3.08

6 -0.006( )')+0.60(-L)+ 4.1

The gain in capacity is plotted in Figure 5.21 and is measured as a function of the ratio
of corrugation amplitudes at the edge and the crown ('). As -' - 00, the corrugated
edge is approached and the gain in capacity decreases.
In the matrix plot of Figure 5.21, the rows represent the wave amplitude Oi where row
one is # = 10% and the second is O3 = 5%. The columns represent the numer of waves
with column 1 at n = 2 to column 4 with n = 0.5. The plots geometrically match up with
Figure 5.6. The presentation of the plots in a matrix shows that having a higher number
of waves n is more beneficial than a higher wave amplitude #4. In other words, it is more
beneficial to have a shorter wavelength than to increase the amplitude.
Unlike the corrugated edge case, the buckling modes for the corrugated crown and
in-phase all differ. However the mechanics of the structure can still be understood. The
weakest point of the straight vault is the flat crown where the point of inflection occurs.
If the cross-section at the crown is changed, the difference in capacity can be explained
by the increase in the moment of inertia because qc, oc I.
In Figure 5.22 the gain in capacity is calculated analytically using Equation 5.8. Math-
ematically the behavior is changed by 'eff. The plot shows good agreement especially
for the in-phase condition. This is because the analytical derivation assumes that the
cross-section geometry is the same at the crown and the edge when in fact they vary.
To reduce the error, the moment of inertia should be re-derived to take into account its
variation along the arclength.

Case Comparison: Having validated that adding double curvature greatly strengthens
the barrel vault, the best location to add curvature is measured as a function of the per-
cent increase in volume from the basic barrel vault. In Figure 5.23 all three cases are

81
 e+

n=2 n=1.5 n=1 n=0.5

160 18 16:C 16 e+
* Uh=20
140 A 140, 140- 140- Uh=11
A Uh=6
120 120, 120- 120 e+1
100 100, 100 100 M,

10% )80 80 A 80 80 e+
560 80 60 60
40 40 40 A 40
20 201 20 U 20 A

0.1
) 5 10 5 0 5 10 13 0 5 10 1: 0 5 10 15
al/a 2 a1/a 2 a 1/a 2 a,/a2

1212
160 160 16C 160
00 * Uh=20
tND 140 140, 140- 140 a Uh=11
A Uh=6
120, 120, 120- 120
100 100,
=5% 100. 100
80 80 80 80
a-
S 60 A 60 60 60
U
40 40 A 40 40
20 20 20 A U 20
A
0 C0 C
0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15
a1/a 2 a1/a 2 al/a 2 81/a2

Figure 5.21 - A matrix of plots highlighting the gain in capacity from the straight barrel vault by using in-phase corrugation
for the continuous shell. The rows represent the increase in corrugation amplitude by #i and the columns represent the number
of waves n.
 L/h=20 LJh=1 1 L/h=6
100 - - - - - , - - -- - - -- - - -- - - - 100 - - -, - - - - - - - - - - - - -.- - - - - L- - -Analytical
A Crown
80, 80 . 80 . In-Phase

60 60 60 -
U) -
40 401 br 40:
00 Cr) \%
A

20 A 20 A - - - 20 A
A
A A
A 1U-

0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

A/a2 A/a2 A/a2

Figure 5.22 Comparison of the numerical results of the corrugated crown and the in-phase to the analytical (Equation 5.8)
for the continuous shell.

U,

U'
Chapter 5. Corrugated Vault Study

plotted as a function of the percent increase in volume for all three span-to-height ratios.
For the corrugated edge, for a 1% increase in volume, the capacity is increased by a factor
of 40. For the corrugated crown, a 1% increase in volume yields a gain in capacity of
a factor of 50. However, the best option is the in-phase corrugation. For the same 1%
increase in volume, the in-phase case yields a gain in capacity by a factor of 80.

Collapse Analysis

Before proceeding to the grid shell studies, the linearized buckling analysis is compared
to the collapse analysis. The collapse analysis accounts for the prebuckling rotations and
nonlinear behavior.
The collapse analysis was performed for an in-phase geometry. Both a normally dis-
tributed load and a vertically distributed load was considered and the results are plotted
in Figure 5.24. First, both the normal load and vertical load are comparable to each
other. Second, the collapse load is approximately three times higher than the buckling
load; thus, the buckling load is a conservative estimate and appropriate to use for this
parametric study.

5.3.3 Grid Shell

Having gained insight into the behavior of the corrugated continuous shell, the grid shell
results are presented next. The same plots used in the continuous shell study are repeated.

Corrugated Edge: First consider only the 30x30 grid shell. For the corrugated edge,
the buckling load is normalized by the buckling load of the basic (cylindrical) 30x30 grid
shell barrel vault as a function of the aspect ratio yA(Figure 5.25a).
al
The three lines
represent the three different span-to-height ratios. Corrugating the edge can increase the
capacity up to 5 times that of the straight edge barrel vault. The amount of increase in
capacity is proportional to (y2 as was seen in the continuous shell.
In Figure 5.25b the buckling load is plotted as a function of the shell height. The
multiple lines represent the different values of the aspect ratio from 2 to 16. The rela-
tionship between the buckling load and the height is linear for a given edge aspect ratio
A and steeper shells are stronger. While increasing the corrugation in the grid shell
al does
benefit the gain in capacity, the amount of gain is not of the same magnitude as seen in
the continuous shell.

84
 Corrugated Edge Corrugated Crown In-phase
100 100 100

80 80 80-

-c
60 60 60 A
0) 2
C, C,
00
40 40 40 A A

20 AL"E AdF 20 AA 20 IUh=20

=Lh=1 1
AUh=6
0"0 0.5 1 1.5 '0 0.5 1 1.5 '0 0.5 1 1.5
% Volume increase % Volume increase % Volume increase

Figure 5.23 - Comparison of the gain in capacity in terms of the percent volume increase from the straight vault for the
continuous shell.

CA
Chapter 5. Corrugated Vault Study

-- - Collapse - Normal
8 -C -Collapse -Vertical
7 Buckling

5 . . .\ .

3
-I
2

0
0 2 4 6 8 10 12 14 16 18
Time step
Figure 5.24 - Comparison of the collapse load and the buckling load for a normally distributed
load (N) and a vertically distributed load (V) for the in-phase continuous shell.

5- x 10 -3
5
L/h=201
-e--U-h=1 X/a
-4- L/h11 -. 2.025/
4- -- L/h=6 -* 2.7
4.05
=. / A
=5.4
- =8.1 /-*
3
3 -=16.2

-J
M)
ni£

0
0 5 10 15
0 50 100 150 200
A/a h [in]
(a) Effect of the corrugation aspect ratio (b) Effect of the span-to-height ratio

Figure 5.25 - Gain in capacity by corrugating the edge of the 30x30 grid shell.

86
 5.3. Results

Next consider the effect of three grid densities (1Ox1O, 20x20, 30x30) on the capacity
for the corrugated edge grid shell. In Figure 5.26 the gain in capacity is plotted for all
three grid densities as a function of the aspect ratio. Depending on the grid density, the
gain in capacity can reach a maximum of 5 for the densest grid and tallest shell. Table
5.2 lists the maximum allowable corrugation 1- to increase the capacity by at least 50%.
The denser the grid, the less corrugation is needed to increase the capacity by at least
50%. The effect of grid density is more prominent the steeper shell. Also, the steeper the
shell the less corrugation is needed to increase the capacity from a straight vault. Because
corrugating the 1Ox1O grid does not improve the capacity significantly from its straight
edge counter part and because for the same amount of volume the 20x20 and 30x30 grid
have the same increase in capacity, only a 30x30 grid will be considered in the remaining
results.

Corrugated Crown: The same trends seen in the corrugated crown for the continuous
shell are exhibited in the grid shell (Figure 5.27). The effect of crown corrugation is only
significant if the crown's aspect ratio is 1- < 15. Unlike the corrugated edge, the gain
in capacity is less pronounced as the shell height changes; the data points are not easily
separated by their height. In total the gain in capacity can quadruple by corrugating the
crown.

In-Phase: In Figure 5.28, the grid shell results for the in-phase case are subplotted in
a matrix. The rows represent from top to bottom Oi=10% and .i-= 5%. The columns
from left to right represent the number of waves n=2, and 1 respectively. As in the case
of the continuous shell, it is more beneficial to increase the number of waves n than to
increase the amplitude #3 to increase the gain in capacity; a shorter wavelength is more
desirable.

Table 5.2 - Maximum value of - to increase the capacity of thte corrugated edge by at least
50%.

Grid Density -=20 -=11 -=6

10x10 not possible 2.7 4
20x20 2.7 4 5
30x30 4 4.5 8

87
 0

Cn

C,'
L/h = 20 1/h = 11 L/h = 6
5 5..
.10x10grid
- 2Ox20grid
4, 4 4 , .30x30grid

3 03
CO
C1
00 2
00
1s
-Ii
1 I a,.
*0 I

0 5 10 15 0 5 10 15 "0 5 10 15
A/a, A/a, A/a,

Figure 5.26 - The effect of varying the grid density for the corrugated edge grid shell.
 5.3. Results

*i3

010 20 40 60 80
A/a2

Figure 5.27 - Gain in capacity by corrugating the crown of the 30x30 grid shell.

Case Comparison: Table 5.3 lists the amount of corrugation needed at either the edge
or the crown to increase the capacity by 50% for each height. For the corrugated crown,
the same increase in capacity can be achieved for a looser corrugation than that at the
edge. In Figure 5.29, the three corrugation cases are plotted against each other in terms of
the percent increase in volume. For the corrugated edge and the in-phase, the maximum
percent increase in volume is 3% but yield a gain in capacity up to five and eight times,
respectively. For a less than 1% increase in volume by corrugating the crown, the capacity
is quadrupled. As evident in the plots, there is a quicker return on using a corrugated
crown.

Table 5.3 - Maximum values of g and y to increase the capacity by 50% for the 30x30 grid
shell.

Corrugation Case 1:=20 !=11 L=6

Edge _-< 4 4.5 8
Crown 1- < 21 24 19

89
Chapter 5. Corrugated Vault Study

n=2 n=1
10 10-

9 9

8 A 8

7 7

=1 0% 6 6

op 5 5

u4 4
CL
3 3

2 2

0 0-
0 2 4 6 8 10 12 0 2 4 6 8 10 12
al /a2 al /a2
10 10

9 9
8 8

.) 7 7

6 6
0.=5% -U5 5

04 4

3 A 3

2 2 A

0 u
0 2 4 6 8 10 12 0 2 4 6 8 10 12
al /a 2 al /a2

Figure 5.28 - Gain in capacity by in-phase corrugation of the 30x30 grid shell.

90
 Edge Corrugation - Grid Crown Corrugation - Grid In-Phase Corrugation - Grid
8
7

6
LU 5
65 4- A a,)
Co
a. 3A br
I'

1
n
0 1 2 3 4 5 '0 1 2 3 4 5 ''0 1 2 3 4 5
% Volume increase from Straight % Volume increase from Straight % Volume increase from Straight

Figure 5.29 - Comparison of the gain in capacity in terms of percent volume increase from straight vault for the grid shell for
all three corrugation locations (edge, crown and in-phase).

CA'

......
.....
...
Chapter 5. Corrugated Vault Study

5.4 Discussion
Corrugated grid shells have many advantages over their continuous counterpart. The
discretization and topology provides an articulated, dramatic space. Their openings allow
for more design options to be used in terms of lighting, and ventilation. The corrugation
provides not only an interesting aesthetic but has a dramatic effect on the capacity,
increasing it by a maximum factor of 8 for an increase in volume less than 5%.
The corrugated vault's behavior can be built from considering a plane strain analysis.
Physically, corrugation at different locations create different mechanical behavior. For
the corrugated edge, the increase corrugation creates an effective length and shortens the
arclength that is more susceptible to buckling. For the corrugated crown, the behav-
ior is that of adding a stiffened longitudinal beam to a vault. And lastly the in-phase
case changes the cross-section geometry and hence the moment of inertia, increasing the
bending stiffness.
In all cases decreasing the span-to-height ratio (becoming steeper) improves the ca-
pacity of the vault. However, the amount of improvement in the corrugated crown case is
independent of the shell height. The aspect ratio between the edge and the crown differ
greatly. To increase the capacity by at least 50%, the corrugated crown can have a looser
corrugation by a factor of six than the edge.
Finally, the grid spacing needs to be dense enough to pick up the extra stiffness of
curvature, but there is a limiting return. Making the grid denser eventually approaches a
continuous shell and the case for designing a grid shell over a continuous lessens. However,
as shown in the following design guidelines, increasing the grid density has a small impact
when compared to the other factors.
The conclusions made in this study are summarized in a set of design guidelines shown
in Figure 5.30 to be used in the early stage of design. The guidelines are organized by the
three corrugation locations: the edge, the crown, and at both edge and crown (in-phase).
In order to increase the buckling capacity, if the designer is considering edge corrugation,
the most effective change would be to increase the height of the shell, and in turn creating
a shorter effective arclength.

92
 5.5. Summary

Edge Corrugation Crown Corrugation In-phase

Decrease
Increase Increase number
crown aspect
height of waves
LratioAI

Icrease
Decrease edge
aspect ratio
amplitude

Figure 5.30 - Set of design guidelines to be used in the early stages of design of a corrugated
grid shell vault.

5.5 Summary
In this chapter the benefits of corrugating a vault, and varying the grid spacing and
span to height ratio are investigated and quantified. Corrugating a vault is an efficient
and effective way to increase the load-bearing capacity of a barrel vault. Adding double
curvature to the barrel vault has significant gains in capacity up to 8 times for less than
a 3% increase in volume. Closed-form solutions for calculating the buckling load were
also derived to account for the varied grid shell parameters. A plane strain analysis of a
barrel vault was modified to account for an effective arclength when corrugating the edge,
and an effective moment of inertia when corrugating the crown. Lastly a set of design
guidelines were developed for use in the early stage of design of corrugated grid shells.

93
Chapter 6

Conclusions

6.1 Summary of Contributions

The goals of this body of work were:
" to calculate the change in load-bearing capacity of grid shells due to their curvature
and topology using numerical and analytical methods,

" to provide a set of design guidelines that can contribute to maximize the mechanical
performance of grid shells during early design stages.

The goals were achieved by conducting a parametric study varying the topology and
topography of grid shells and calculating the change in load-bearing capacity using both
numerical and analytical methods. The main contributions are as follows:
* ParametricStudy: The roles of curvature, and topology on the load-bearing capacity
of grid shells were assessed using a parametric study. The parametric study was
framed within the context of real design constraints and contained over 400 finite
element models. The generation of the spherical cap and corrugated vault finite
element models were programmed and inputted into ADINA. The buckling capacity
of the finite element models for both the grid shells and the continuous shells was
calculated. It was found that:

- Corrugation is an effective and efficient means to strengthen the barrel vault.

- Corrugation on a grid shell increases capacity up to 8 times for less than a 3%
increase in volume. Corrugation on a continuous shell increases the capacity
up to 80 times for a 1.5% volume increase.

94
 6.2. Future Work

- Triangular grid are better for steeper shells (spherical cap).

- Shallower shells require a denser grid to increase the capacity.

- Different design configurations of the span-to-height ratio, the spacing and the
topology can have the same structural efficiency.

" Mechanics of Grid Shells: To describe the mechanics of grid shells, closed form
solutions were derived for calculating the buckling load of grid shells. Multiple
equivalent continuum approaches were defined and validated and helped distinguish
grid shell behavior from being membrane dominated to bending dominated. An
equivalent continuum defined by area equivalence is a conservative estimate for the
buckling load and can be used the preliminary stages of design.

The first study on the role of double curvature on barrel vaults for both continuous
shells and grid shells was completed and the mechanical behavior of the corrugated
barrel vault was validated using a plane strain analysis. It was found that corrugat-
ing the edge produces an effective arclength, while corrugating the crown increases
the moment of inertia.

" Design Guidelines: This dissertation developed rules of thumb to be used in the
initial design for assessing the behavior of grid shells as a function of their shape,
topology, curvature and grid spacing. The parameters of interest were pulled from
the literature on the design process of existing grid shells. Simple analytical meth-
ods for calculating the buckling load of both spherical cap and corrugated vault
grid shells were derived. The guidelines are classified in terms of the main design
constraints driving the design process.

6.2 Future Work

The exploration of the mechanics of grid shells have herein contributed towards a better
understanding of these complicated structures, but also motivates numerous areas of
continued research:

* Analytical Modeling: With a better understanding of the collapse analysis and the
analytical models developed herein, new analytical models for other surfaces can be
developed.

95
Chapter 6. Conclusions

" Software Tools: Having established the significant relationships between the design
parameters, the results are to be implemented into a software tool to allow designers
and engineers to play with varying the parameters while simultaneously observing
their effects on the capacity and volume of material.

" Joint Design: Often the cost of grid shells has been affected by the difficulties of
the joint design. Future work will account for the trade offs between having more
members at the nodes (increasing grid density and capacity) with the complications
that arise from designing a joint shared by more members.

96
Appendix A

MATLAB Code

A.1 Spherical Cap - Quadrilateral Grid

1 %The following code creates the finite element ADINA input file a .
spherical cap grid shell with a QUADRILATERAL grid
2 clear all
3 delete linecheckl5.in
4
[headline,geometryData]=hdrload( 'geometryData.txt');
Rvector=geometryData(:,5);
hvector=geometryData(:,3);
centervector=geometryData(:,2);
[dl d2]=size(Rvector);
for (k=l:dl)
%Mesh Density
uvector=[10;20;30;40;50;60];
[ul u2]=size(uvector);
for (1=1 :ul)
u=uvector (l,1);
v=uvector (l,1);

%% INPUTS

%Global Geometry
L=1200;

h=hvector(k,l); %arc height

R=Rvector(k,l); %radius of sphere

%center of sphere
Xo=O;
Yo=O;
Zo=centervector(k,1);

97
Appendix A. MATLAB Code

31
32 %Cross section
33 tl=5; %width of rectangular cross section
34 t2=5; %height of rectangular cross section
35
36 dx=(L)/u; %same as spacing
37 dy=(L)/v;
38
39 ndiv=floor((dx/(2*tl)));%the mesh density per line
40 %% CALCULATING Pcr THEORETICAL
41 %p-crTheoretical
42 %% CREATING THE MESH GRID
43 [X,Y] = meshgrid(-L/2:dx:L/2, -L/2:dy:L/2);
44
45 %to remove point outside of circle
46 %radius that describes the boundary circle on the plane ...
(R-h+- Zo)
47 CBR=X.^2+Y.^2;
48 index=find(CBR>(L/2)^2);
49 X(index) = [;
50 Y(index) =
51 X=X';
52 Y=Y';
53 %plot(X,Y)
54
55 %to add points that are with the xi, or yi coordinate but sit ..
on the circle
56 YcirclePos=((L/2)^2-X.^2).^.5;
57 XcirclePos=((L/2)^2-Y.^2).^.5;
58 YcircleNeg=-((L/2)^2-X.^2).^.5;
59 XcircleNeg=-((L/2)^2-Y.^2).^.5;
60
61 %check plot
62 %plot(XcirclePos,Y, 'bo',XcircleNeg,Y, 'm*')
63 %plot(X,YcirclePos,'ks', X,YcircleNeg,'g*')
64
65 %add the cirlce boundary points to the mesh grid so we can ...
find the z
66 %values at those point
67 Xnew=vertcat(X,XcirclePos, XcircleNeg,X,X);
68 Ynew=vertcat(Y,Y,Y,YcirclePos, YcircleNeg);
69
70 %plot(Xnew,Ynew,'k.');
71 points=[Xnew Ynew];
72 %points=[X Y);%un comment the line above when i add back the
circle points
73 %plot(points(:,l),points(:,2),'go');
74 pointsFinal=unique (sortrows (points),'rows');
75 %plot(pointsFinal(:,1),pointsFinal(:,2),'go');
76
77 %% CALCULATING THE Z VALUE

98
 A.1. Spherical Cap - Quadrilateral Grid

78 X=pointsFinal (:, 1);

79 Y=pointsFinal(:,2);
80
81 %need to change the - zo to + if zo is negative
82 if(Zo<0)
83 Z=(R^2-(X-Xo).^2-(Y-Yo).^2).^0.5+Zo;
84 else
85 Z=(R^2-(X-Xo).^2-(Y-Yo).^2).^0.5-Zo;
86 end
87
88 %plot3(X,Y,Z,'ko')
89 pointID=[1:size(X,1)]';
90 PointBLOCK=[pointID X Y Z zeros(size(X))];
91 % if(Zo<0)
92 % axis([-600 600 -600 600 R-h+Zo R+Zo]);
93 % else
94 % axis([-600 600 -600 600 R-h-Zo R-Zo]);
95 % end
96 PointBLOCKOriginal=PointBLOCK;
97
98
99 % ADD THE AUXILIARY NODES %
100 % %
101 auxPoints=[PointBLOCK(:,1)+5000,PointBLOCK(:,2:3),...
102 PointBLOCK(:,4)-1,PointBLOCK(:,5)];
103 %append the auxiliary point coordiates to the point block
104 PointBLOCK=[PointBLOCK; auxPoints];
105
106 %% GET THE LINE CONNECTIVITY
107 Xvector=[-L/2:dx:L/2] ';
108 Yvector=[-L/2:dy:L/2] ';
109 LineBlockFinal=[];
110 for(i=l:size(Xvector,l))
11 indicesl=find(X==Xvector(i, 1));
112 if (size ( (indicesl) , 1)>1)
113 LineBlock=[indicesl indicesl+l];
114 LineBlock (end, : )=[] ;
115 LineBlockFinal=vertcat(LineBlockFinal,LineBlock);
116 else
117 end
118 end
119 for(j=l:size(Yvector,l))
120 indices2=find(Y==Yvector(j, 1));
121 if(size((indices2),l)>1)
122 indicesA=indices2;
123 indicesA (end, ) = ];
124
125 indicesB=indices2;
126 indicesB (1,:)=[];
127
128 LineBlock=[indicesA indicesB];

99
 Appendix A. MATLAB Code

129 LineBlockFinal=vertcat(LineBlockFinal,LineBlock);
130 else
131 end
132 end
133 %% ISOLATE THE BOUNDARY NODES
134 CBR=X. ^2+Y. ^2;
135 tolerance=le-7;
136 %index3=find(CBR==(L/2)^2);
137 index4=find((l-tolerance)*(L/2)^2 CBR);% (l+tolerance)*(L/2)^2);
138 boundaryPoints=[index4];
139 %SORT BOUNDARY POINTS BASED ON THETA
140 boundaryPointsCoords=[index4 zeros (size(index4))];
141 %more complicated calculation of theta
142 for(i=1:size(index4,1))
143 if (X (index4 (i, 1) )>0)
144 boundaryPointsCoords (i,2)= ...
145 atan(Y(index4(i,1))/X(index4(i,l)));
146 end
147
148 if((X(index4(i,1))<0)&& Y(index4(i,l))>0 )
149 boundaryPointsCoords (i,2)= ..
150 atan (Y (index4 (i,1)) /X (index4 (i,1))) +pi;
151 end
152
153 if((X(index4(i,1))<0)&& Y(index4(i,1))<0)
154 boundaryPointsCoords (i,2)= ..
155 atan (Y(index4 (i,1)) /X (index4 (i,1)) )-pi;
156 end
157
158 if((X(index4(i,1))==0)&& Y(index4(i,l))>0)
159 boundaryPointsCoords (i,2)=pi/2;
160 end
161
162 if((X(index4(i,1))==0)&& Y(index4(i,1))<0)
163 boundaryPointsCoords (i,2)=-pi/2;
164 end
165
166 if((X(index4(i,1))==0)&& Y(index4(i,1))==0)
167 boundaryPointsCoords (i,2)=0;
168 end
169
170 end
171
172 sortedboundaryPoints=sortrows (boundaryPointsCoords,2);
173 %get P1 and P2 for boundary lines
174 Pl=sortedboundaryPoints (:, 1);
175 P2=Pl;
176 P2= P2([2:end 1]);
177 boundaryLines=[Pl P2];
178
179 %add boundary lines to line block

100
 A.1. Spherical Cap - Quadrilateral Grid

180 LineBlockFinal=vertcat(LineBlockFinal,boundaryLines);
181 %
182 %% WRITE THE POINT BLOCK
183 fid=fopen('linecheckl5.in','at');
184 fprintf(fid, '*\nMASTER ANALYSIS=STATIC MODEX=EXECUTE ...
TSTART=0.00000000000000 IDOF=O,\nOVALIZAT=NONE ...
FLUIDPOT=AUTOMATIC CYCLICPA=1 IPOSIT=STOP,\nREACTION=YES ...
INITIALS=NO FSINTERA=NO IRINT=DEFAULT ...
CMASS=YES,\nSHELLNDO=AUTOMATIC AUTOMATI=OFF ...
SOLVER=SPARSE,\nCONTACT-=CONSTRAINT-FUNCTION ...
TRELEASE=0.00000000000000,\nRESTART-=NO FRACTURE=NO ...
LOAD-CAS=NO LOAD-PEN=NO ...
SINGULAR=YES,\nSTIFFNES=0.000100000000000000 ...
MAP-OUTP=NONE MAP-FORM=NO,\nNODAL-DE='''' POROUS-C=NO
ADAPTIVE=O ZOOM-LAB=1 AXIS-CYC=O,\nPERIODIC=NO ...
VECTOR-S=GEOMETRY EPSI-FIR=NO ...
STABILIZ=NO,\nSTABFACT=1.OOOOOOOOOOOOOOE-10 ...
RESULTS=PORTHOLE FEFCORR=NO,\nBOLTSTEP=1 EXTEND-S=YES ...
CONVERT-=NO DEGEN=YES TMC-MODE=NO,\nENSIGHT-=NO');
185 fprintf(fid, '\n*\nCOORDINATES POINT SYSTEM=O\n@CLEAR');
186 fprintf(fid, '\n%d %d %d %d %d',PointBLOCK');
187 %% WRITE THE LINE BLOCK
188 LineID=[1:size(LineBlockFinal,1) ]';
189 LineBiockFinal2=[LineID LineBlockFinal];
190
191 fprintf(fid,'\n*\nLINE STRAIGHT NAME=%d P1=%d ...
P2=%d',LineBlockFinal2');
192 fclose(fid);
193
194 %% DEFINE BOUNDARY CONDITIONS
195 %GET POINT IDS OF POINTS THAT NEED TO BE PINNED
196 %Define pinned condition
197 fid=fopen('linecheckl5.in','at');
198 fprintf(fid,'\n*\n FIXITY ...
NAME=PINNED\n@CLEAR\n''X-TRANSLATION'...
199 '\n''Y-TRANSLATION''\n''Z-TRANSLATION''\n''OVALIZATION''');
200
201 % APPLY BOUNDARY CONDITIONS TO POINTS
202 fprintf(fid,'\n@\n*\n FIXBOUNDARY POINTS FIXITY=PINNED\n@CLEAR');
203 fprintf(fid,'\n%d ''PINNED''', boundaryPoints);
204 fprintf(fid, '\n@');
205
206 %APPLY BOUNDARY CONDITIONS TO LINES
207 fprintf(fid,'\n@\n*\n FIXBOUNDARY LINES FIXITY=PINNED\n@CLEAR');
208 fprintf(fid,'\n%d ''PINNED''', ...
LineBlockFinal2((end+1-size (boundaryLines)):end,1));
209 fprintf (fid, '\n@');
210 fclose(fid);
211
212 %% CALCULATE LINE LENGTH

101
 Appendix A. MATLAB Code

213 %this is used to calculate if the rigid end length is greater

than the
214 %line length. here we separate the lines that will have a ...
rigid end
215 %length and those that will not.
216
217 indexPl=LineBlockFinal2 (:,2);
218 indexP2=LineBlockFinal2 (:,3);
219
220 length=sqrt((PointBLOCKOriginal(indexPl,2)-...
221 PointBLOCKOriginal(indexP2,2)).^2+...
222 (PointBLOCKOriginal(indexPl,3)-...
223 PointBLOCKOriginal(indexP2,3)).^2+...
224 (PointBLOCKOriginal(indexPl,4)-...
225 PointBLOCKOriginal(indexP2,4)).^2);
226 length((end+l-size (boundaryLines)):end,l)=O;
227 %set boundary line lengths to zero so that it won't come up ...
in rigid end
228 %search
229 indexLength=find((length/ndiv)>tl);
230
231 indexNOLength=LineID;
232 indexNOLength(indexLength)=[];
233
234 %% WRITE REST OF INPUT FILE
235 %%% WRITE MATERIAL BLOCK%
236 fid=fopen('linecheckl5.in','at');
237 %below is for elastic material
238 %fprintf(fid, '\n*\nMATERIAL ELASTIC NAME=1 E=29000 NU=0.3, ...
DESNITY=2.83e-4 ALPHA=O MDESCRIP=''STEEL\' '');
239
240 %below is for bilinear material (so that we can enter a ...
sigma-y yield
241 %stress)
242 fprintf(fid,'\n*\nMATERIAL PLASTIC-BILINEAR NAME=l ...
HARDENIN=ISOTROPIC,\nE=29000.0000000000 ...
NU=0.300000000000000 ...
YIELD=50.0000000000000,\nET=50.00000000000000 ...
EPA=0.00000000000000 ...
STRAINRA=O,\nDENSITY=0.000283000000000000 ...
ALPHA=0.00000000000000,\nTREF=0.00000000000000 ...
DEPENDEN=NO ...
TRANSITI=0.000100000000000000,\nEP-STRAI=0.00000000000000 .
BCURVE=O ...
BVALUE=0.00000000000000,\nXM-INF=0.00000000000000 ...
XMO=0.00000000000000,\nETA=0.00000000000000 ...
MDESCRIP=''STEEL''');
243 fclose(fid);
244 %% WRITE CROSS SECTION BLOCK
245 fid=fopen('linecheckl5.in','at');

102
 A.1. Spherical Cap - Quadrilateral Grid

246 fprintf(fid,'\n@\n*\nCROSS-SECTIO RECTANGULAR NAME=1 ...

WIDTH=%d,\n HEIGHT=%d SC=O TC=0,\n TORFAC=1 ...
SSHEARF=0.8333,\n TSHEARF=0 ISHEAR=NO SQUARE=NO',tl,t2);
247 %% DEFINE ELEMENT TYPE
248 fprintf(fid,'\n*\nEGROUP BEAM NAME=1 SUBTYPE=THREE-D ...
DISPLACE=DEFAULT MATERIAL=1 RINT=5,\n SINT=DEFAULT ...
TINT=DEFAULT RESULTS=STRESSES INITIALS=NONE,\n ...
CMASS=DEFAULT RIGIDEND=INFINITE MOMENT-C=NO RIGIDITY=1,\n ...
MULTIPLY=1000000.00000000 RUPTURE=ADINA OPTION=NONE,\n ...
BOLT-TOL=0.00000000000000 DESCRIPT=''NONE'' SECTION=1,\n ...
PRINT=DEFAULT SAVE=DEFAULT TBIRTH=0.00000000000000,\n ...
TDEATH=0.00000000000000 SPOINT=4 ...
BOLTFORC=0.00000000000000,\n BOLTNCUR=0 TMC-MATE=1 ...
BOLT-NUM=0 BOLT-LOA=0.00000000000000,\n WARP=NO');
249 fclose(fid);
250 %% APPLY LOADS
251 % Apply Point loads to Points on Edge and in middle of ...
interesecting
252 LoadedPoints=PointBLOCKOriginal(:,1);
253 LoadedPoints([boundaryPoints])=[];
254 [dl d2]=size(LoadedPoints);
255 forceLabel=(1:1:dl)';
256 fid=fopen('linecheckl5.in','at');
257 fprintf(fid,'\n*\n LOAD FORCE NAME=1 MAGNITUD= 1 ...
FX=0.00000000000000,\nFY=0.00000000000000 ...
FZ=-1.00000000000000');
258 fprintf(fid,'\n*\n APPLY-LOAD BODY=O\n@CLEAR');
259 fprintf(fid, '\n%d ''FORCE'' 1 ''POINT'' %d 0 1 0 0 -1 0 0 0 ...
''NO'',\n 0 1 0 ''MID''',[forceLabel(:,1) ...
LoadedPoints(:,1)]');
260 fclose(fid);
261 noLoadedPoints=dl;
262 %% MESH DENSITY
263 %insert here Line Labels
264 fid=fopen('linecheckl5.in','at');
265 fprintf(fid,'\n*\nSUBDIVIDE LINE NAME=1 MODE=DIVISIONS ...
NDIV=%d RATIO=1.00000000000000,\n PROGRESS=GEOMETRIC ...
CBIAS=NO\n@CLEAR',ndiv);
266 fprintf(fid,'\n%d ', LineBlockFinal2(:,1));
267 fprintf(fid,'\n@');
268 fclose(fid);
269 %% CREATE MESH
270 fid=fopen('linecheckl5.in','at');
271 auxPointonLine=[LineBlockFinal2(:,2)+5000 LineBlockFinal2(:,1)];
272 %print to file the auxiliary point
273 %fprintf('\nauxpoint %d line %d',auxPointonLine')
274 %insert here line labels
275 fprintf(fid, '\n*\nGLINE NODES=2 AUXPOINT=%d NCOINCID=ENDS ...
NCENDS=12,\n NCTOLERA=1.OOOOOOOOOOOOOOE-05 SUBSTRUC=0 ...
GROUP=1 MIDNODES=CURVED\n @CLEAR \n%d\n@',auxPointonLine');

103
 Appendix A. MATLAB Code

276 %fprintf('\nauxpoint %d line %d', [auxPointonLine(:,2) ...

lineLabel]')
277 fclose(fid);
278 fid=fopen('linecheckl5.in','at');
279 fprintf(fid,'\n*\n');
280 fclose(fid);
281 %% ELEMENT DATA
282 tlVector(1:size(LineBlockFinal2),1)=tl;
283 %to include rigid end lengths
284 fid=fopen('linecheckl5.in','at');
285 fprintf(fid,'EDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@CLEAR\n@\n');
286 fprintf(fid,'EDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@CHAROW 1
101 101');
287 %fprintf(fid, '\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' ...
0.00000000000000 0.00000000000000 0,\nO.00000000000000 %d ...
%d\n', [LineBlockFinal2(:,1) t1Vector t1Vector]');
288 %put rigid ends on elements that are long enough
289 firstElement=(ndiv) .*LineBlockFinal2 (indexLength, 1)-(ndiv-1);
290 lastElement=(ndiv).*LineBlockFinal2(indexLength,1);
291
292 noOfElement=(1:ndiv*size (LineBlockFinal2,1))';
293 noOfElement([firstElement,lastElement])=[];
294
295 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' ...
0.00000000000000 0.00000000000000 0,\nO.00000000000000 %d ...
0', [firstElement tlVector(indexLength,1)]');
296 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' ...
0.00000000000000 0.00000000000000 0,\nO.00000000000000 0 ...
%d', [lastElement tlVector(indexLength,1)]');
297 %make NO rigid ends on elements whose length are not long enough
298 fprintf(fid, '\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' ...
0.00000000000000 0.00000000000000 0,\nO.00000000000000 0 ...
0',noOfElement);
299
300 fprintf(fid,'\n@ENDMODIFY\n*\nEDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@ENDMODIFY');
301 fclose(fid);
302 %% COPY INPUT FILE TO FILE WITH THE HEIGHT AND MESH DENSITY ...
IDENTIFIED
303 copyfile('linecheckl5.in', ['inputFiles/Height.' num2str(h) ...
'- num2str(u) 'x' num2str(u) '.in'])
304 delete linecheckl5.in
305 end
306 end

104
 A.2. Spherical Cap - Triangular Grid

A.2 Spherical Cap - Triangular Grid

1 %The following code creates the finite element input file for a spherical
2 %cap grid shell with a TRIANGULAR grid
3 clear all
4 %% INPUTS
s %Global Geometry
6 L=1200;
7 h=200;%hvector(k,l); %arc height
8 R=1000%Rvector(k,l); %radius of sphere
9 %center of sphere
10 Xo=0;
11 Yo=0;
12 Zo=-200;%centervector(k,1);
13 %Cross section
14 tl=5; %width of rectangular cross section
15 t2=5; %height of rectangular cross section
16 %mesh density
17 u=50;
18 v=50;
19 dx=(L)/u;
20 dy=(L)/v;
21
22 ndiv=floor((dx/(2*tl)));%the mesh density per line
23 %% MESH GENERATOR
24 figure
25 fd=inline('sqrt(sum(p.^2,2))-600','p');
26 [p,t]=distmesh2d(fd,@huniform,dx, [-600,-600;600,600], []);
27 % p contains the nodal points, need to insert the nodal label, and ...
put in a
28 % z coordinate of 0, and a coordinate system of 0;
29 %% edit info from mesh generator
30 % t contains teh element connectivity, but need to insert the element ...
label
31 % [dl d2)= size(t)
32 % elemid=[1:1:dl]';
33 % elemconn=[elemid, t(:,l), t(:,2), t(:,3)];
34
35 % Get POINT coordinates
36 [dl d2]=size(p);
37 pointid=[l:1:dl]';
38 X=p(:,1);
39 =p(:,2);
40 Z=zeros(dl,1);
41 coordsys=zeros(dl,1);
42 pointBlockl=[pointid X Y Z coordsys];
43 %% CALCULATING THE Z VALUE
44 %need to change the - zo to + if zo is negative
45 if(Zo<0)

105
Appendix A. MATLAB Code

46 Z=(R^2-(X-Xo).^2-(Y-Yo).^2).^0.5+Zo;
47 else
48 Z=(R^2-(X-Xo).^2-(Y-Yo).^2).^O.5-Zo;
49 end
50
si pointBlockl(:,4)=Z;
52 pointBlock=pointBlockl;
53 PointBLOCKOriginal=pointBlock;
54 %% GET THE ELEMENT CONNECTIVITY
55 elemConnectivity=t;
56 [dl d2]=size(elemConnectivity);
57 for(i=l:dl)
58 j=3*i-2;
59 lineBlock(j,1:2)=[elemConnectivity(i,l) elemConnectivity(i,2)];
60 lineBlock(j+1,1:2)=[elemConnectivity(i,2) elemConnectivity(i,3)];
61 lineBlock(j+2,1:2)=[elemConnectivity(i,3) elemConnectivity(i,l)];
62 end
63 %clean up the line block to get rid of repeating lines
64 lineBlock(:,:);
65 lineblockorig=lineBlock;
66 [dl d2]=size(lineBlock);
67 for(i=2:dl)
68 for(j=i+l:dl)
69 pairl=[lineBlock(i,l) lineBlock(i,2)];
70 pair2=[lineBlock(j,2) lineBlock(j,l)];
71
72 if(pair2==pairl)
73 lineBlock(j,:)=[];
74
75 [dl d2]=size(lineBlock);
76 break
77 else
78 end
79 end
80 end
81 lineLabel=(1:1:dl)';
82 lineBLOCKFINAL=[lineLabel, lineBlock];
83
84 %%
85 % START TO WRITE INPUT BLOCKS FOR ADINA FILE
86 %
87 %% INSERTING COMMON INFO INTO THE ADINA INPUT FILE
88 % WRITE THE POINT BLOCK
89 auxPoints=[pointBlock(:,1)+5000,pointBlock(:,2:3),pointBlock(:,4)-1,
90 pointBlock(:,5)]
91 pointBlockFINAL=[pointBlock; auxPoints]
92 %
93 %% WRITE THE POINT BLOCK
94 fid=fopen('linecheckl5.in','at');
95 fprintf (fid, '*\nMASTER ANALYSIS=STATIC MODEX=EXECUTE ...
TSTART=0.00000000000000 IDOF=O,\nOVALIZAT=NONE FLUIDPOT=AUTOMATIC

106
 A.2. Spherical Cap - Triangular Grid

CYCLICPA=1 IPOSIT=STOP,\nREACTION=YES INITIALS=NO FSINTERA=NO ...

IRINT=DEFAULT CMASS=YES,\nSHELLNDO=AUTOMATIC AUTOMATI=OFF ...
SOLVER=SPARSE,\nCONTACT-=CONSTRAINT-FUNCTION ...
TRELEASE=0.00000000000000,\nRESTART-=NO FRACTURE=NO LOAD-CAS=NO ...
LOAD-PEN=NO SINGULAR=YES,\nSTIFFNES=0.000100000000000000 ...
MAP-OUTP=NONE MAP-FORM=NO,\nNODAL-DE='''' POROUS-C=NO ADAPTIVE=O ...
ZOOM-LAB=1 AXIS-CYC=O,\nPERIODIC=NO VECTOR-S=GEOMETRY EPSI-FIR=NO ...
STABILIZ=NO,\nSTABFACT=1.OOOOOOOOOOOOOOE-10 RESULTS=PORTHOLE ...
FEFCORR=NO,\nBOLTSTEP=l EXTEND-S=YES CONVERT-=NO DEGEN=YES ...
TMC-MODE=NO,\nENSIGHT-=NO');
96 fprintf(fid, '\n*\nCOORDINATES POINT SYSTEM=0\n@CLEAR');
97 fprintf(fid, '\n%d %d %d %d %d',pointBlockFINAL');
98 fclose(fid);
99 %% WRITE THE LINE BLOCK
1o fid=fopen('linecheckl5.in','at');
101 fprintf(fid, '\n*\nLINE STRAIGHT NAME=%d Pl=%d P2=%d',lineBLOCKFINAL');
102 fclose(fid);
103 %define the auxiliary point for each line, this will be used later
104 auxPointonLine=[lineBLOCKFINAL(:,1),lineBLOCKFINAL(:,2)+50001
105 %% Find boundary lines and points
106 indexl=find((599<Z) & (Z 601));
107 boundarypoints=pointBlockFINAL(index1,1);
108 %% WRITE BOUNDARY BLOCK
109 %GET POINT IDS OF POINTS THAT NEED TO BE PINNED
11o %Define pinned condition
inl fid=fopen('linecheckl5.in','at');
112 fprintf(fid,'\n*\n FIXITY NAME=PINNED\n@CLEAR\n''X-TRANSLATION'...
113 '\n''Y-TRANSLATION''\n''Z-TRANSLATION''\n''OVALIZATION''');
114 %APPLY BOUNDARY CONDITIONS TO POINTS
115 fprintf(fid,'\n@\n*\n FIXBOUNDARY POINTS FIXITY=PINNED\n@CLEAR');
116 fprintf(fid,'\n%d ''PINNED''', boundarypoints);
117 fprintf(fid, '\n@');
118 fclose(fid);
119 %% WRITE REST OF ADINA FILE
120 %
121 %%% WRITE MATERIAL BLOCK%
122 fid=fopen('linecheckl5.in','at');
123 %below is for elastic material
124 fprintf(fid,'\n*\nMATERIAL ELASTIC NAME=1 E=29000 NU=0.3, ...
DESNITY=2.83e-4 ALPHA=0 MDESCRIP=''STEEL\''');
125 %below is for bilinear material (so that we can enter a sigma-y yield
126 %stress)
127 %fprintf(fid, '\n*\nMATERIAL PLASTIC-BILINEAR NAME=1 ...
HARDENIN=ISOTROPIC,\nE=29000.0000000000 NU=0.300000000000000 ...
YIELD=50.0000000000000,\nET=50.00000000000000 ...
EPA=0.00000000000000 STRAINRA=O,\nDENSITY=0.000283000000000000 ...
ALPHA=0.00000000000000,\nTREF=0.00000000000000 DEPENDEN=NO ...
TRANSITI=0.000100000000000000,\nEP-STRAI=0.00000000000000 ...
BCURVE=O BVALUE=0.00000000000000,\nXM-INF=0.00000000000000 ...
XMO=o.oooooo00000000,\nETA=0.00000000000000 MDESCRIP=''STEEL''');
128 fclose(fid);

107
Appendix A. MATLAB Code

129 %% WRITE CROSS SECTION BLOCK

130 fid=fopen('linecheckl5.in','at');
131 fprintf(fid,'\n@\n*\nCROSS-SECTIO RECTANGULAR NAME=1 WIDTH=%d,\n ...
HEIGHT=%d SC=0 TC=0,\n TORFAC=1 SSHEARF=0.8333,\n TSHEARF=0 ...
ISHEAR=NO SQUARE=NO',tl,t2);
132 %% DEFINE ELEMENT TYPE
133 fprintf(fid, '\n*\nEGROUP BEAM NAME=1 SUBTYPE=THREE-D DISPLACE=DEFAULT ...
MATERIAL=1 RINT=5,\n SINT=DEFAULT TINT=DEFAULT RESULTS=STRESSES ...
INITIALS=NONE,\n CMASS=DEFAULT RIGIDEND=INFINITE MOMENT-C=NO ...
RIGIDITY=1,\n MULTIPLY=1000000.00000000 RUPTURE=ADINA ...
OPTION=NONE,\n BOLT-TOL=0.00000000000000 DESCRIPT=''NONE''
SECTION=1,\n PRINT=DEFAULT SAVE=DEFAULT ...
TBIRTH=0.00000000000000,\n TDEATH=0.00000000000000 SPOINT=4 ...
BOLTFORC=0.00000000000000,\n BOLTNCUR=0 TMC-MATE=1 BOLT-NUM=0 ...
BOLT-LOA=0.00000000000000,\n WARP=NO');
134 fclose(fid);
135 %% APPLY LOADS
136 % Apply Point loads to Points on Edge and in middle of interesecting
137 LoadedPoints=pointBlock(:,1);
138 LoadedPoints([boundarypoints])=[];
139 [dl d2]=size(LoadedPoints);
140 forceLabel=(1:1:dl)';
141 fid=fopen('linecheckl5.in','at');
142 fprintf(fid,'\n*\n LOAD FORCE NAME=1 MAGNITUD= 1 ...
FX=0.00000000000000,\nFY=0.00000000000000 FZ=-1.00000000000000');
143 fprintf(fid,'\n*\n APPLY-LOAD BODY=O\n@CLEAR');
144 fprintf(fid, '\n%d ''FORCE'' 1 ''POINT'' %d 0 1 0 0 -1 0 0 0 ...
''NO'',\n 0 0 1 0 ''MID''', [forceLabel(:,1) LoadedPoints(:,l)]');
145 fclose(fid);
146 noLoadedPoints=dl;
147 %% MESH DENSITY
148 %insert here Line Labels
149 fid=fopen('linecheckl5.in','at');
iso fprintf(fid,'\n*\nSUBDIVIDE LINE NAME=1 MODE=DIVISIONS NDIV=%d ...
RATIO=1.00000000000000,\n PROGRESS=GEOMETRIC CBIAS=NO\n@CLEAR',ndiv);
151 fprintf(fid,'\n%d ', lineBLOCKFINAL(:,1));
152 fprintf(fid, '\n@');
153 fclose(fid);
154 %% CREATE MESH
155 fid=fopen('linecheckl5.in','at');
156 auxPointonLine=[lineBLOCKFINAL(:,2)+5000 lineBLOCKFINAL(:,l)];
157 %print to file the auxiliary point
158 %fprintf('\nauxpoint %d line %d',auxPointonLine')
159 %insert here line labels
160 fprintf(fid,'\n*\nGLINE NODES=2 AUXPOINT=%d NCOINCID=ENDS ...
NCENDS=12,\n NCTOLERA=1.00000000000000E-05 SUBSTRUC=0 GROUP=1 ...
MIDNODES=CURVED\n @CLEAR \n%d\n@',auxPointonLine');
161 %fprintf('\nauxpoint %d line %d', [auxPointonLine(:,2) lineLabel]')
162 fclose(fid);
163 fid=fopen('linecheckl5.in','at');
164 fprintf(fid,'\n*\n');

108
 A.2. Spherical Cap - Triangular Grid

165 fclose(fid);
166 %% isolate boundary, copied this from quad grid
167 %ISOLATE THE BOUNDARY NODES
168 CBR=X.^2+Y.^2;
169 tolerance=le-7;
170 %index3=find(CBR==(L/2)^2);
171 index4=find((1-tolerance)*(L/2)^2<CBR);% (1+tolerance)*(L/2)^2);
172 boundaryPoints=[index4];
173 %SORT BOUNDARY POINTS BASED ON THETA
174 boundaryPointsCoords=[index4 zeros(size(index4))];
175 %more complicated calculation of theta
176 for(i=l:size(index4,1))
177 if (X (index4 (i, 1) )>0)
178 boundaryPointsCoords(i,2)=atan(Y(index4(i,1))/X(index4(i,1)));
179 end
180
181 if((X(index4(i,l))<0)&& Y(index4(i,l))>0)
182 boundaryPointsCoords(i,2)=atan(Y(index4 (i, 1)) /X (index4 (i, 1))) +pi;
183 end
184
185 if((X(index4(i,1))<0)&& Y(index4(i,1))<0)
186 boundaryPointsCoords(i,2)=atan(Y(index4 (i, 1)) /X(index4 (i, 1)) )-pi;
187 end
188
189 if((X(index4(i,l))==0)&& Y(index4(i,l))>0)
190 boundaryPointsCoords(i,2)=pi/2;
191 end
192
193 if((X(index4(i,l))==0)&& Y(index4(i,l))<0)
194 boundaryPointsCoords(i,2)=-pi/2;
195 end
196
197 if((X(index4(i,l))==0)&& Y(index4(i,l))==O)
198 boundaryPointsCoords(i,2)=0;
199 end
200
201 end
202
203 sortedboundaryPoints=sortrows(boundaryPointsCoords,2);
204 %get P1 and P2 for boundary lines
205 Pl=sortedboundaryPoints(:,l);
206 P2=Pl;
207 P2= P2([2:end 1]);
208 boundaryLines=[Pl P2];
209 %- LineBlockFinal=vertcat(LineBlockFinal,boundaryLines);
210
211 %% CALCULATE LINE LENGTH
212 %this is used to calculate if the rigid end length is greater than the
213 %line length. here we separate the lines that will have a rigid end
214 %length and those that will not.
215

109
Appendix A. MATLAB Code

216 indexPl=lineBLOCKFINAL(:,2);
217 indexP2=lineBLOCKFINAL(:,3);
218
219 length=sqrt((PointBLOCKOriginal(indexP1,2)-...
220 PointBLOCKOriginal(indexP2,2)).^2+...
221 (PointBLOCKOriginal(indexP1,3)-...
222 PointBLOCKOriginal(indexP2,3)).^2+...
223 (PointBLOCKOriginal(indexP1,4)-PointBLOCKOriginal(indexP2,4)).^2);
224 length((end+1-size(boundaryLines)):end,l)=0;
225 %set boundary line lengths to zero so that it won't come up in rigid end
226 %search
227 indexLength=find((length/ndiv)>tl);
228
229 LineID=lineLabel; %this was edited for the triangle case
230 indexNOLength=LineID;
231 indexNOLength(indexLength)=[];
232 %% ELEMENT DATA
233 tlVector(1:size(lineBLOCKFINAL),l)=tl;
234 %to include rigid end lengths
235 fid=fopen('linecheckl5.in','at');
236 fprintf(fid,'EDATA SUBSTRUC=0 GROUP=1 UNDEFINE=IGNORE\n@CLEAR\n@\n');
237 fprintf(fid,'EDATA SUBSTRUC=O GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@CHAROW 1 101 ...
101');
238 %fprintf(fid, '\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 .
0.00000000000000 0,\nO.00000000000000 %d %d\n', .
[LineBlockFinal2(:,l) tlVector t1Vector]');
239 %put rigid ends on elements that are long enough
240 firstElement=(ndiv).*lineBLOCKFINAL(indexLength,1)-(ndiv-1);
241 lastElement=(ndiv).*lineBLOCKFINAL(indexLength,1);
242
243 noOfElement=(1:ndiv*size(lineBLOCKFINAL,1))';
244 noOfElement([firstElement,lastElement])=[];
245
246 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 %d 0', [firstElement ...
t1Vector(indexLength,1)]');
247 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 %d', [lastElement ...
t1Vector(indexLength,1)]');
248 %make NO rigid ends on elements whose length are not long enough
249 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 O',noOfElement);
250 %
251 fprintf(fid,'\n@ENDMODIFY\n*\nEDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@ENDMODIFY');
252 fclose(fid);
253 %
254 %% COPY INPUT FILE TO FILE WITH THE HEIGHT AND MESH DENSITY IDENTIFIED
255 % copyfile('linecheckl5.in', ['inputFiles/Height.I num2str(h) '-' ...
num2str(u) 'x' num2str(u) '.in'])

110
 A.2. Spherical Cap - Triangular Grid

256 % delete linecheckl5.in

111
Appendix A. MATLAB Code

A.3 Corrugated Vault - Continuous Shell

1 %the following file is used to create the input ADINA data (points, ...
surface
2 %grids etc) that is then used to create the CONTINUOUS Barrel Vault ...
shell.
3 %user inputs the geometrical parameters that define the corrugated vault
4 clear all
5 hold on
6
7 % INPUTS - INCHES, KIPS
8 L1=1200; %L1 is the full width
9 L2=972; %L2 is HALF the length
10 h=109; %h is the height of the arch
11
12 %INSERT as a percent of the length and width used in the amplitude, I ...
dividt
13 %by 100 later.
14 %to be OUT OF PHASE put one of the beta's as negative
15 betal=0 %PERCENT
16 beta2=10%PERCENT
17
18 %al is the amplitude of the edge curve (so at z==0)
19 al=(betal/100)*(Ll/2);
20 %a2 is the amplitude of the top undulation (so at z==h)
21 a2=(beta2/100)*(h);
22
23 %n is the number of waves on half the length, so in total there are 2*n
24 %waves
25 n=1.5
26 %steps refers to the remaking of the mesh grid, and how many steps we ...
take
27 %in the x direction
28 %THIS needs to be MORE FINE to get the curvature
29 steps=40;
30 %THIS MAKES THE NUMBER OF ROWS DIFFERENT (Sl)
31
32 %redefining the parameters
33 xo=Ll/2;%half the width
34 if n==0;
35 A=0;
36 else
37 A=al;
38 end
39
40 l=L2;
41 %dx=Ll/50;
42 dx=(xo+A)/50;
43 if(n<3)

112
 A.3. Corrugated Vault - Continuous Shell

44 dy=L2/20; %dy doesn't need to be as fine as dx, %NOT TRUE FOR ...
WHEN n=5;
45 else
46 dy=L2/50
end

[Xl,Yl]=meshgrid(O :dx:xo+A,-l:dy:1);
XX=Xl(:);
YY=Yl(:);
plot (XX,YY,'y.');

%maybe need to add boundary points

xb=(xo+A*cos(2*n*pi*YY/l));
plot(xb,YY,'k.');
XX=vertcat(XX,xb);
YY=vertcat(YY,YY);
plot(XX,YY,'gs');

points2=[XX YY];
points3=unique(poi nts2,'rows');

XX=points3(:,1);
YY=points3(:,2);
plot(XX,YY, 'b.');

dd=(xo+A*cos(2*n*pi*YY/l))-XX;

XXX=XX;
YYY=YY;
indices=find(dd<O);
XXX(indices)=[];
YYY (indices) = [I ;
plot(XXX,YYY,'cd');

XFINAL=vertcat(XXX,-l*XXX);
YFINAL=vertcat(YYY,YYY);
plot(XFINAL,YFINAL,'k+');

%clear X Y xlin ylin Input inputy inputsub

Yvalues=unique(YFINAL);
[yl y2]=size(Yvalues)

X=[];
Y=[];
Input=[];

for(i=l:yl)
indices=find(YFINAL==Yvalues(i,1));
xlinl=linspace(O,max(XFINAL(ind.ices)), (steps))';
xlin2=sortrows(vertcat(-l*xlinl,xlinl));

113
 Appendix A. MATLAB Code

94 ind=find(xlin2==0);
95 xlin2(ind(1,1))=[]
96 xlin=xlin2;
97
98 X=vertcat(X,xlin);
99 [sl s2]=size(xlin)
100
101 ylin(1:sl)=Yvalues(i,1);
102 Y=vertcat(Y,ylin');
103
104 inputy(1:sl)=i;
105 inputsub=[1:1:sl; inputy]';
106 Input=vertcat(Input,inputsub);
107 end
108 Z=(h+a2.*cos(2*n*pi.*Y./l)).*[l-((X.^2)./(xo+A.*cos(2*n*pi.*Y./l)).^2)];
109
11o POINTS=[X Y Z];
ill %reorder
112 POINTS2=sortrows(POINTS,2);
113 plot3(POINTS2(:,l), POINTS2(:,2), POINTS2(:,3),'m*');
114 X=POINTS2(:,1);
115 Y=POINTS2(:,2);
116 Z=POINTS2(:,3);
117
118 [zl z2]=size(Z);
119 label=[1:1:zl]';%
120 %points=[label p zeros(dl,1)]
121 points=[label X Y Z zeros(zl,l)];
122 points2=[X Y Z];
123 InputFinal= [Input(:,2) Input(:,l) label];
124
125 %steps
126 noRows = size(Yvalues) %becomes 3 of rows
127 noCols=sl %becomes # of cols
128

129 polylinePointl=(noCols+l)/2;
130 polylinePoint2=polylinePointl+(noCols)*(noRows-1)
131 A=noCols
132 al
133 a2
134 dlmwrite('PointsToimport.txt', points,'delimiter', '\t');
135 dlmwrite('SurfaceGridToimport.txt', InputFinal,'delimiter', '\t');
136
137 %% using MyCrust to plot the delauney triangulation
138 % Author:Giaccari Luigi
139 % Last Update: 28/01/2009
140 % Created: 15/4/2008
141 % MYCRUST - makes a surface out of nonuniform grid data
142 % Input:
143 % p is a Nx3 array containing the 3D set of points
144 % Output:

114
 A.3. Corrugated Vault - Continuous Shell

145 % t are points id contained in triangles nx3 array

146 addpath('D:\Smalek\Desktop\RESEARCH SPRING ...
2011\MATLAB\MyCrustOpenO7O9O9');
147 figure
148 [t]=MyCrustOpen(points2);
149 trisurf(t,points2(:,l),points2(:,2),points2(:,3),'facecolor', 'c',...
150 'edgecolor','b');%plot della superficie;
151 shading interp;
152 colormap(gray);
153 figure
154 trimesh(t,points2(:,1),points2(:,2),points2(:,3),'facecolor', 'c',...
155 'edgecolor','b');%plot della superficie;

115
Appendix A. MATLAB Code

A.4 Corrugated Vault - Quadrilateral Grid

1 clear all
2 % the following code takes the element connectivity of QUAD SHELL ...
ELEMENTS
3 % and gives them a line id # and the points/nodes that connect them
4 %rewrite data into line #, point 1, point 2 of ADINA Block portion
5
6 % INPUT FILES
7 %read in data
8 %file name "archQUADelements.txt" - is the exported data from adina ...
that contains
9 %the element connectivity
10 %file name "archQUADnodes.txt" - is teh exported data from adina that
11 %contains the nodal 'points and their coordinates
12
13 % OUTPUT FILES
14 %this program writes all the adina commands into one file called
is %"adinaINPUTQUAD. in"
16
17 % INPUT PARAMETERS TO EDIT
18 % NDIV = which is the number of elements to mesh each beam member into
19 ndiv=3;
20 %ALSO - delete the adinaINPUTQUAD.in file when rerunning this.
otherwise it
21 %will append the new info to the old input file.
22
23 %Cross section
24 t1=4.98; %width of rectangular cross section
25 t2=4.98; %height of rectangular cross section
26 %% INSERTING COMMON INFO INTO THE ADINA INPUT FILE
27 %specifically CMASS=Yes, allows the option to print the mass and ...
volume of
28 %the model
29 fid=fopen('adinaINPUTQUAD.in','at');
30 fprintf(fid, '*\nMASTER ANALYSIS=STATIC MODEX=EXECUTE ...
TSTART=0.00000000000000 IDOF=O,\nOVALIZAT=NONE FLUIDPOT=AUTOMATIC ...
CYCLICPA=l IPOSIT=STOP,\nREACTION=YES INITIALS=NO FSINTERA=NO ...
IRINT=DEFAULT CMASS=YES, \nSHELLNDO=AUTOMATIC AUTOMATI=OFF ...
SOLVER=SPARSE, \nCONTACT-=CONSTRAINT-FUNCTION ...
TRELEASE=0.00000000000000,\nRESTART-=NO FRACTURE=NO LOAD-CAS=NO ...
LOAD-PEN=NO SINGULAR=YES,\nSTIFFNES=0.000100000000000000 ...
MAP-OUTP=NONE MAP-FORM=NO, \nNODAL-DE='''' POROUS-C=NO ADAPTIVE=O ...
ZOOM-LAB=l AXIS-CYC=O,\nPERIODIC=NO VECTOR-S=GEOMETRY EPSI-FIR=NO ...
STABILIZ=NO,\nSTABFACT=1.OOOOOOOOOOOOOOE-10 RESULTS=PORTHOLE ...
FEFCORR=NO,\nBOLTSTEP=l EXTEND-S=YES CONVERT-=NO DEGEN=YES ...
TMC-MODE=NO, \nENSIGHT-=NO');
31 %% POINT BLOCK
32 pointBlockl=dlmread('archQUADnodes.txt');

116
 A.4. Corrugated Vault - Quadrilateral Grid

33 %define the auxiliary points

34 auxPoints=[pointBlockl(:,l)+1000,pointBlockl(:,2:3),pointBlockl(:,4)-1,..
35 pointBlockl(:,5)];
36 pointBlock=[pointBlockl; auxPoints];
37 fprintf(fid, '\n*\nCOORDINATES POINT SYSTEM=O\n@CLEAR');
38 fprintf(fid, '\n%d %d %d %d %d',pointBlock');
39 fclose(fid);
40 %% LINE BLOCK%
41 elemConnectivity=dlmread('archQUADelements.txt');
42 [dl, d2]=size(elemConnectivity);
43 elemConnectivity=elemConnectivity(l:dl,2:5);
44
45 for(i=l:dl)
46 j=4*i-3;
47 lineBlock(j,1:2)=[elemConnectivity(i,l) elemConnectivity(i,2)];
48 lineBlock(j+1,1:2)=[elemConnectivity(i,2) elemConnectivity(i,3)];
49 lineBlock(j+2,1:2)=[elemConnectivity(i,3) elemConnectivity(i,4)];
50 lineBlock(j+3,1:2)=[elemConnectivity(i,4) elemConnectivity(i,l)];
si end
52
53 %clean up the line block to get rid of repeating lines
54 lineBlock(:,:);
55 [dl d2]=size(lineBlock);
56 for(i=l:dl)
57 for(j=i:dl)
58 pairl=[lineBlock(i,l) lineBlock(i,2)];
59 pair2=[lineBlock(j,2) lineBlock(j,l)];
60 if(pair2==pairl);
61 lineBlock(j,:)=[];
62 [dl d2]=size(lineBlock);
63 break
64 end
65 end
66 end
67 lineLabel=(1:1:dl)';
68 lineBLOCKFINAL=[lineLabel, lineBlock];
69 fid=fopen('adinaINPUTQUAD.in','at');
70 fprintf(fid, '\n*\nLINE STRAIGHT NAME=%d Pl=%d P2=%d',lineBLOCKFINAL');
71 fclose(fid);
72 %define the auxiliary point for each line
73 auxPointonLine=[lineBLOCKFINAL(:,l),lineBLOCKFINAL(:,2)+1000];
74 %% WRITE MATERIAL BLOCK%
75 fid=fopen('adinaINPUTQUAD.in','at');
76 %below is for elastic material
77 fprintf(fid,'\n*\nMATERIAL ELASTIC NAME=1 E=29000 NU=0.3, ...
DESNITY=2.83e-4 ALPHA=O MDESCRIP=''STEEL\''');
78
79 %below is for bilinear material (so that we can enter a sigma-y yield
80 %stress)
81 %fprintf(fid,'\n*\nMATERIAL PLASTIC-BILINEAR NAME=l ...
HARDENIN=ISOTROPIC,\nE=29000.0000000000 NU=0.300000000000000 ...

117
Appendix A. MATLAB Code

YIELD=50.0000000000000,\nET=0.00000000000000 EPA=0.00000000000000 ...

STRAINRA=O,\nDENSITY=0.000283000000000000 ...
ALPHA=0.00000000000000,\nTREF=0.00000000000000 DEPENDEN=NO ...
TRANSITI=0.000100000000000000,\nEP-STRAI=0.00000000000000 ...
BCURVE=O BVALUE=0.00000000000000,\nXM-INF=0.00000000000000 ...
XMO=0.00000000000000,\nETA=0.00000000000000 MDESCRIP=''STEEL''');
82
83 %% WRITE CROSS SECTION BLOCK
84 fprintf(fid, '\n@\n*\nCROSS-SECTIO RECTANGULAR NAME=1 WIDTH=%d,\n ...
HEIGHT=%d SC=O TC=O,\n TORFAC=l SSHEARF=O,\n TSHEARF=O ISHEAR=NO ...
SQUARE=YES',tl,t2);
85 %DEFINE ELEMENT TYPE
86 % HERE INSERT IF THE JOINTS ARE RIGID OR NOT "RIGIDEND=INFINITE" OR ...
"RIGIDEND=NONE"
87 fprintf(fid, '\n*\nEGROUP BEAM NAME=1 SUBTYPE=THREE-D DISPLACE=DEFAULT
MATERIAL=1 RINT=5,\n SINT=DEFAULT TINT=DEFAULT RESULTS=STRESSES ...
INITIALS=NONE,\n CMASS=DEFAULT RIGIDEND=INFINITE MOMENT-C=NO ...
RIGIDITY=1,\n MULTIPLY=1000000.00000000 RUPTURE=ADINA ...
OPTION=NONE,\n BOLT-TOL=0.00000000000000 DESCRIPT=''NONE''
SECTION=1,\n PRINT=DEFAULT SAVE=DEFAULT ...
TBIRTH=0.00000000000000,\n TDEATH=0.00000000000000 SPOINT=4 ...
BOLTFORC=0.00000000000000,\n BOLTNCUR=O TMC-MATE=1 BOLT-NUM=O ...
BOLT-LOA=0.00000000000000,\n WARP=NO');
88 %% MESH DENSITY
89 %insert here Line Labels
90 %fprintf(fid, '\n*\nSUBDIVIDE LINE NAME=103 MODE=DIVISIONS NDIV=1 ...
RATIO=1.00000000000000,\n PROGRESS=GEOMETRIC CBIAS=NO\n@CLEAR\n');
91 fprintf(fid,'\n*\nSUBDIVIDE LINE NAME=103 MODE=DIVISIONS NDIV=%d ...
RATIO=1.00000000000000,\n PROGRESS=GEOMETRIC ...
CBIAS=NO\n@CLEAR\n',ndiv);
92 fprintf(fid,'\n %d ', lineLabel);
93 fprintf(fid, '\n@');
94
95 %% CREATE MESH
96 %insert here line labels
97 fprintf(fid,'\n*\nGLINE NODES=2 AUXPOINT=%d NCOINCID=ENDS ...
NCENDS=12,\n NCTOLERA=1.OOOOOOOOOOOOOOE-05 SUBSTRUC=O GROUP=1 ...
MIDNODES=CURVED\n @CLEAR \n%d \n@ ', [auxPointonLine(:,2) ...
lineLabel]');
98 %fprintf('\nauxpoint %d line %d', [auxPointonLine(:,2) lineLabel]')
99 fclose(fid);
100
1o %% CALCULATE LINE LENGTH
102 %this is used to calculate if the rigid end length is greater than the
103 %line length. here we separate the lines that will have a rigid end
104 %length and those that will not.
105
106 indexPl=lineBLOCKFINAL(:,2);
107 indexP2=lineBLOCKFINAL(:,3);
108
109 length=sqrt((pointBlockl(indexPl,2)-pointBlockl(indexP2,2)).^2+...

118
 A.4. Corrugated Vault - Quadrilateral Grid

110 (pointBlockl(indexPl,3)-pointBlockl(indexP2,3)).^2+...
il (pointBlockl(indexPl,4)-pointBlockl(indexP2,4)).^2);
112 %length((end+l-size(boundaryLines)):end,l)=0;
113 %set boundary line lengths to zero so that it won't come up in rigid end
114 %search
115 indexLength=find( (length/ndiv) >tl);
116
117 indexNOLength=lineLabel;
118 indexNOLength(indexLength)=[];
119 %% ELEMENT DATA
120 tlVector(1:size(lineBLOCKFINAL),1)=tl;
121 %to include rigid end lengths
122 fid=fopen('adinaINPUTQUAD.in','at');
123 fprintf(fid,'EDATA SUBSTRUC=0 GROUP=1 UNDEFINE=IGNORE\n@CLEAR\n@\n');
124 fprintf(fid,'EDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@CHAROW 1 101 ...
101');
125 %fprintf(fid, '\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 %d %d\n', .
[LineBlockFinal2(:,1) tlVector t1Vector]');
126 %put rigid ends on elements that are long enough
127 %place holder
128 firstElement=(ndiv).*lineBLOCKFINAL(indexLength,1)-(ndiv-1);
129 lastElement=(ndiv).*lineBLOCKFINAL(indexLength,1);
130
131 noOfElement=(l:ndiv*size(lineBLOCKFINAL,1))';
132 noOfElement([firstElement,lastElement])=[];
133
134 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 %d 0', [firstElement ...
tlVector(indexLength,1)');
l35 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 %d', [lastElement ...
tlVector (indexLength, 1) ]1');
136 %make NO rigid ends on elements whose length are not long enough
137 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 0',noOfElement);
138 %
139 fprintf(fid,'\n@ENDMODIFY\n*\nEDATA SUBSTRUC=0 GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@ENDMODIFY');
140 fclose(fid);
141 %% LOADS
142 %Apply Point loads to Points on Edge and in middle of interesecting
143 [dl d2]=size(pointBlockl);
144 forceLabel=(1:1:dl)';
145 fid=fopen('adinaINPUTQUAD.in','at');
146 fprintf(fid,'\n*\n LOAD FORCE NAME=1 MAGNITUD= 1 ...
FX=0.00000000000000,\nFY=0.00000000000000 FZ=-1.00000000000000');
147 fprintf(fid,'\n*\n APPLY-LOAD BODY=O\n@CLEAR');
148 fprintf(fid, '\n%d ''FORCE'' 1 ''POINT'' %d 0 1 0 0 -1 0 0 0 ...
''NO'',\n 0 0 1 0 ''MID''',[forceLabel(:,1) pointBlockl(:,1)]');

119
Appendix A. MATLAB Code

149 fclose(fid);
150 %% collect boundary points
151 boundaryPointsOPENINGS=[find(pointBlockl(:,3)>971.9); ...
find(pointBlockl(:,3) -971.9)];
152 boundaryPointsBOUNDARIES=[find(pointBlockl(:,4)==O)];
153 %% DEFINE BOUNDARY CONDITIONS
154 %GET POINT IDS OF POINTS THAT NEED TO BE PINNED
155 %Define pinned condition
156 fid=fopen('adinaINPUTQUAD.in','at');
157 fprintf(fid,'\n*\n FIXITY NAME=OPENINGS\n@CLEAR\n''X-ROTATION''\n'...
158 'Y-TRANSLATION''\n''OVALIZATION''');
159 fprintf(fid,'\n*\n FIXITY NAME=BOUNDARIES\n@CLEAR\n''X-TRANSLATION'...
160 '\n''Z-TRANSLATION''\n''OVALIZATION''');
161
162 %Degree(s) of freedom to be fixed. {X-TRANSLATION/Y-TRANSLATION/
163 % Z-TRANSLATION/X-ROTATION/Y-ROTATION/Z-ROTATION/OVALIZATION/
164 % FLUID-POTENTIAL/PORE-FLUID-PRESSURE/BEAM-WARP}
165 % APPLY BOUNDARY CONDITIONS TO POINTS
166 fprintf(fid,'\n@\n*\n FIXBOUNDARY POINTS FIXITY=ALL\n@CLEAR');
167 fprintf(fid,'\n%d ''OPENINGS''', boundaryPointsOPENINGS);
168 fprintf(fid,'\n%d ''BOUNDARIES''', boundaryPointsBOUNDARIES);
169 fprintf(fid, '\n@')
170
171 % %APPLY BOUNDARY CONDITIONS TO LINES
172 % fprintf(fid, '\n@\n*\n FIXBOUNDARY LINES ...
FIXITY=PINNED\n@CLEAR');
173 % fprintf(fid, '\n%d ''PINNED''',
LineBlockFinal2 ( (end+1-size(boundaryLines)) :end, 1));
174 % fprintf(fid, '\n@');
175 fclose(fid);

120
 A.5. Corrugated Vault - Triangular Grid

A.5 Corrugated Vault - Triangular Grid

1 clear all
2 % the following code takes the element connectivity of Triangular 2D ...
ELEMENTS
3 % and gives them a line id # and the points/nodes that connect them
4 %rewrite data into line #, point 1, point 2 of ADINA Block portion
5
6 % INPUT FILES
7 %read in data
8 %file name "arch4elements.txt" - is the exported data from adina that ...
contains
9 %the element connectivity
10 %file name "arch4nodes.txt" - is teh exported data from adina that
11 %contains the nodal points and their coordinates
12
13 % OUTPUT FILES
14 %this program writes all the adina commands into one file called
15 %"adinaINPUTQUAD.in"
16
17 % INPUT PARAMETERS TO EDIT
18 % NDIV = which is the number of elements to mesh each beam member into
19 ndiv=4;
20 %ALSO - delete the adinaINPUTQUAD.in file when rerunning this. ...
otherwise it
21 %will append the new info to the old input file.
22
23 %Cross section
24 tl=4.1; %width of rectangular cross section
25 t2=4.1; %height of rectangular cross section
26 %read in data
27 %rewrite data into line #, point 1, point 2 of ADINA Block portion
28 %% INSERTING COMMON INFO INTO THE ADINA INPUT FILE
29 %specifically CMASS=Yes, allows the option to print the mass and ...
volume of
30 %the model
31 fid=fopen('adinaINPUT.in','at');
32 fprintf(fid, '*\nMASTER ANALYSIS=STATIC MODEX=EXECUTE ...
TSTART=0.00000000000000 IDOF=O,\nOVALIZAT=NONE FLUIDPOT=AUTOMATIC ...
CYCLICPA=1 IPOSIT=STOP,\nREACTION=YES INITIALS=NO FSINTERA=NO ...
IRINT=DEFAULT CMASS=YES,\nSHELLNDO=AUTOMATIC AUTOMATI=OFF ...
SOLVER=SPARSE,\nCONTACT-=CONSTRAINT-FUNCTION ...
TRELEASE=0.00000000000000,\nRESTART-=NO FRACTURE=NO LOAD-CAS=NO ...
LOAD-PEN=NO SINGULAR=YES,\nSTIFFNES=0.000100000000000000 ...
MAP-OUTP=NONE MAP-FORM=N0,\nNODAL-DE='''' POROUS-C=NO ADAPTIVE=0 ...
ZOOM-LAB=1 AXIS-CYC=0,\nPERIODIC=NO VECTOR-S=GEOMETRY EPSI-FIR=NO ...
STABILIZ=NO,\nSTABFACT=1.OOOOOOOOOOOOOOE-10 RESULTS=PORTHOLE ...
FEFCORR=NO,\nBOLTSTEP=l EXTEND-S=YES CONVERT-=NO DEGEN=YES ...
TMC-MODE=NO,\nENSIGHT-=NO');

121
Appendix A. MATLAB Code

33 %% POINT BLOCK
34 pointBlockl=dlmread('arch4nodes.txt');
35 [api ap2]=size(pointBlockl);
36 %if lines get messed up, revisit the next line where i add 10,000 to the
37 %point id for creating the auxillary nodes. if the geometry has more than
38 %10,000 points then the numbering gets messed up. the 10,000 is the line
39 %below must also match the line further down ...
%auxPointonLine=[lineBLOCKFINAL(:,1),lineBLOCKFINAL(:,2)+10000];
40
41 auxPoints=[pointBlockl(:,1)+apl,pointBlockl(:,2:3) ,pointBlockl (:,4)-1,...
42 pointBlockl(:,5)];
43 pointBlock=[pointBlockl; auxPoints];
44 fprintf(fid, '\n*\nCOORDINATES POINT SYSTEM=0\n@CLEAR');
45 fprintf(fid, '\n%d %d %d %d %d',pointBlock');
46 fclose(fid);
47 %% LINE BLOCK%
48 elemConnectivity=dlmread('arch4elements.txt')
49 [dl, d2]=size(elemConnectivity)
5o elemConnectivity=elemConnectivity(l:dl,2:4)
51
52 for(i=l:dl)
53 j=3*i-2;
54 lineBlock(j,1:2)=[elemConnectivity(i,l) elemConnectivity(i,2)];
55 lineBlock(j+1,1:2)=[elemConnectivity(i,2) elemConnectivity(i,3)];
56 lineBlock(j+2,1:2)=[elemConnectivity(i,3) elemConnectivity(i,l)];
57 end
58
59 %clean up the line block to get rid of repeating lines
60 lineBlock(:,:);
61 [dl d2]=size(lineBlock);
62 for(i=l:dl)
63 for(j=i:dl)
64 pairl=[lineBlock(i,l) lineBlock(i,2)];
65 pair2=[lineBlock(j,2) lineBlock(j,l)];
66 if(pair2==pairl);
67 lineBlock(j,:)=[];
68 [dl d2]=size(lineBlock);
69 break
70 end
71 end
72 end
73 lineLabel=(1:1:dl)'
74 lineBLOCKFINAL=[lineLabel, lineBlock]
75 fid=fopen('adinaINPUT.in','at');
76 fprintf(fid, '\n*\nLINE STRAIGHT NAME=%d Pl=%d P2=%d',lineBLOCKFINAL');
77 fclose(fid);
78 %define the auxiliary point for each line
79 auxPointonLine=[lineBLOCKFINAL(:,1),lineBLOCKFINAL(:,2)+apl]
80
81 %% WRITE MATERIAL BLOCK%
82 fid=fopen('adinaINPUT.in','at');

122
 A.5. Corrugated Vault - Triangular Grid

83 %below is for elastic material

84 fprintf(fid,'\n*\nMATERIAL ELASTIC NAME=1 E=29000 NU=0.3,
DESNITY=2.83e-4 ALPHA=O MDESCRIP=''STEEL\''');
85
86 %below is for bilinear material (so that we can enter a sigma-y yield
87 %stress)
88 %fprintf(fid,'\n*\nMATERIAL PLASTIC-BILINEAR NAME=l ...
HARDENIN=ISOTROPIC,\nE=29000.0000000000 NU=0.300000000000000 ...
YIELD=50.0000000000000,\nET=0.00000000000000 EPA=0.00000000000000 ...
STRAINRA=O,\nDENSITY=0.000283000000000000 ...
ALPHA=0.00000000000000,\nTREF=0.00000000000000 DEPENDEN=NO ...
TRANSITI=0.000100000000000000,\nEP-STRAI=0.00000000000000 ...
BCURVE=O BVALUE=0.00000000000000,\nXM-INF=0.00000000000000 ...
XMO=0.00000000000000,\nETA=0.00000000000000 MDESCRIP=''STEEL''');
89 %% WRITE CROSS SECTION BLOCK
90 fprintf(fid,'\n@\n*\nCROSS-SECTIO RECTANGULAR NAME=1 WIDTH=%d,\n ...
HEIGHT=%d SC=O TC=O,\n TORFAC=l SSHEARF=O,\n TSHEARF=O ISHEAR=NO ...
SQUARE=YES',tl,t2);
91 %DEFINE ELEMENT TYPE
92 % HERE INSERT IF THE JOINTS ARE RIGID OR NOT "RIGIDEND=INFINITE" OR ...
"RIGIDEND=NONE"
93 fprintf(fid,'\n*\nEGROUP BEAM NAME=1 SUBTYPE=THREE-D DISPLACE=DEFAULT ...
MATERIAL=l RINT=5,\n SINT=DEFAULT TINT=DEFAULT RESULTS=STRESSES ...
INITIALS=NONE,\n CMASS=DEFAULT RIGIDEND=INFINITE MOMENT-C=NO ...
RIGIDITY=1,\n MULTIPLY=1000000.00000000 RUPTURE=ADINA ...
OPTION=NONE,\n BOLT-TOL=0.00000000000000 DESCRIPT=''NONE'' ...
SECTION=1,\n PRINT=DEFAULT SAVE=DEFAULT ...
TBIRTH=0.00000000000000,\n TDEATH=0.00000000000000 SPOINT=4 ...
BOLTFORC=0.00000000000000,\n BOLTNCUR=O TMC-MATE=1 BOLT-NUM=O ...
BOLT-LOA=0.00000000000000,\n WARP=NO');
94 %% MESH DENSITY
95 %insert here Line Labels
96 fprintf(fid,'\n*\nSUBDIVIDE LINE NAME=103 MODE=DIVISIONS NDIV=%d ...
RATIO=1.00000000000000,\n PROGRESS=GEOMETRIC ...
CBIAS=NO\n@CLEAR\n',ndiv);
97 fprintf(fid,'\n %d ', lineLabel);
98 fprintf(fid,'\n@');
99 %% CREATE MESH
100 %insert here line labels
101 fprintf(fid,'\n*\nGLINE NODES=2 AUXPOINT=%d NCOINCID=ENDS ...
NCENDS=12,\n NCTOLERA=1.OOOOOOOOOOOOOE-05 SUBSTRUC=O GROUP=1 ...
MIDNODES=CURVED\n @CLEAR \n%d \n@ ', [auxPointonLine(:,2) ...
lineLabel]');
102 %fprintf('\nauxpoint %d line %d', [auxPointonLine(:,2) lineLabel]')
103 fclose(fid);
104 %% CALCULATE LINE LENGTH
105 %this is used to calculate if the rigid end length is greater than the
106 %line length. here we separate the lines that will have a rigid end
107 %length and those that will not.
108
109 indexPl=lineBLOCKFINAL(:,2);

123
 Appendix A. MATLAB Code

110 indexP2=lineBLOCKFINAL(:,3);
111
112 length=sqrt((pointBlockl(indexPl,2)-pointBlockl(indexP2,2)).^2+...
113 (pointBlockl(indexP1,3)-pointBlockl(indexP2,3)).^2+...
114 (pointBlockl(indexP1,4)-pointBlockl(indexP2,4)).^2);
115 %length((end+l-size (boundaryLines)):end,l)=0;
116 %set boundary line lengths to zero so that it won't come up in rigid end
117 %search
118 indexLength=find((length/ndiv) >tl);
119
120 indexNOLength=lineLabel;
121 indexNOLength(indexLength)=[];
122 %% ELEMENT DATA
123 tlVector(l:size(lineBLOCKFINAL),l)=tl;
124 %to include rigid end lengths
125 fid=fopen('adinaINPUT.in','at');
126 fprintf(fid,'EDATA SUBSTRUC=O GROUP=1 UNDEFINE=IGNORE\n@CLEAR\n@\n');
127 fprintf(fid,'EDATA SUBSTRUC=O GROUP=1 ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@CHAROW 1 101 ...
101');
128 %fprintf(fid, '\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\n0.00000000000000 %d %d\n', .
[LineBlockFinal2(:,l) tlVector tlVector]');
129 %put rigid ends on elements that are long enough
130 %place holder
131 firstElement=(ndiv).*lineBLOCKFINAL(indexLength,l)-(ndiv-1);
132 lastElement=(ndiv).*lineBLOCKFINAL(indexLength,1);
133
134 noOfElement=(l:ndiv*size(lineBLOCKFINAL,1))';
135 noOfElement([firstElement,lastElement])=[];
136
137 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 %d O', [firstElement ...
t1Vector(indexLength,1)]');
138 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 %d', [lastElement ...
tlVector(indexLength,1)]');
139 %make NO rigid ends on elements whose length are not long enough
140 fprintf(fid,'\n%d 1 1 0 ''DEFAULT'' ''DEFAULT'' 0.00000000000000 ...
0.00000000000000 0,\nO.00000000000000 0 0',noOfElement);
141 %
142 fprintf(fid,'\n@ENDMODIFY\n*\nEDATA SUBSTRUC=0 GROUP=l ...
UNDEFINE=IGNORE\n@STARTMODIFY\n@ENDMODIFY');
143 fclose(fid);
144 %% Apply Point loads to Points on Edge and in middle of interesecting
145 [dl d2]=size(pointBlock);
146 forceLabel=(1:1:dl)'
147 fid=fopen('adinaINPUT.in','at');
148 fprintf(fid,'\n*\n LOAD FORCE NAME=l MAGNITUD= 1 ...
FX=0.00000000000000,\nFY=0.00000000000000 FZ=-1.00000000000000');
149 fprintf(fid,'\n*\n APPLY-LOAD BODY=O\n@CLEAR')

124
 A.5. Corrugated Vault - Triangular Grid

15o fprintf(fid, '\n%d ''FORCE'' 1 ''POINT'' %d 0 1 0 0 -1 0 0 0 ...

''NO'',\n 0 0 1 0 ''MID''',[forceLabel(:,1) pointBlock(:,1)]');
151 fclose(fid);
152 %%
153 %GET POINT IDS OF POINTS THAT NEED TO BE PINNED
154 boundaryPointsOPENINGS=[find(pointBlockl(:,3)>971.9); ...
find (pointBlockl (:,3):<-971.9)];
155 boundaryPointsBOUNDARIES=[find(pointBlockl(:,4)==0)];
156 %% DEFINE BOUNDARY CONDITIONS
157 %GET POINT IDS OF POINTS THAT NEED TO BE PINNED
158 %Define pinned condition
159 fid=fopen('adinaINPUT.in','at');
160 fprintf(fid,'\n*\n FIXITY NAME=OPENINGS\n@CLEAR\n''X-ROTATION''\n'...
161 'Y-TRANSLATION''\n''OVALIZATION''');
162 fprintf(fid,'\n*\n FIXITY NAME=BOUNDARIES\n@CLEAR\n''X-TRANSLATION'...
163 '\n''Z-TRANSLATION''\n''OVALIZATION''');
164
165 %Degree(s) of freedom to be fixed. {X-TRANSLATION/Y-TRANSLATION/
166 % Z-TRANSLATION/X-ROTATION/Y-ROTATION/Z-ROTATION/OVALIZATION/
167 % FLUID-POTENTIAL/PORE-FLUID-PRESSURE/BEAM-WARP}
168 % APPLY BOUNDARY CONDITIONS TO POINTS
169 fprintf(fid,'\n@\n*\n FIXBOUNDARY POINTS FIXITY=ALL\n@CLEAR');
170 fprintf(fid,'\n%d ''OPENINGS''', boundaryPointsOPENINGS);
171 fprintf(fid,'\n%d ''BOUNDARIES''', boundaryPointsBOUNDARIES);
172 fprintf(fid,'\n@');
173 fclose(fid);

125
Appendix B

Equivalent Continuum

1. Reduce the equation of equilibrium and compatibility into a couple set of two ho-
mogeneous equations in w and f.

DVw + 1V4f + -peRV2W = 0 (B.la)

V4f
EhaV 2 W = (B.1b)
R
where f must satisfy the compatibility condition and ha is the thickness from to
axial rigidity.

2. Use the solutions of w and f derived by (Hutchinson, 1967) and solve for B 1

w = cos (i cos (B.2a)

f = B 1 cos cos (K
R)
(B.2b)

where B 1 is a constant and n, and iY are wavelength parameters.

3. Substitute Equations B.2a into Equations B.la.

B1 EhaR (B.3)
+ K
K2

4. Substitute B 1 and the bending rigidity D into Equation B.1a and solve for pe.

Eh3
(B.4a)
12(1 -V2)

126
 2E hb (1 +v) + ha
pe (B.4b)
eR 12R2(i _ V2) K2 + K2

where hb is the thickness from to the bending rigidity.

5. Fin the minimum of 2 + K2 to solve for the buckling load.

e+ = 0 (B.5a)

22 2R 2 hap 2E Vh3ha
xi + (B.5b)
Y h3 R2
3(1 - v2)

127
Bibliography

(2007). Ripple effect: Glenn howells architects in windsor great park. Architecture today,
(175):72-78. 18

Adriaenssens, S., Ney, L., Bodarwe, E., and Williams, C. (2010). Dutch maritime museum:
Form-finding of an irregular faceted skeletal shell - part b. p. 1356-1366. 26, 28

Anderson, R. (2000). The Great Court at the British Museum. British Museum Press,
London. 26, 28

Balut, N. and Gioncu. V. (2000). The influence of geometrical tolerances on the behaviour
of space structures. International Journal of Space Structures, 15(3):189. 29

Bathe, K.-J. (1996). Finite element procedures. Prentice Hall, Englewood Cliffs, N.J. 43

Bernard, E. S., Coleman, R., and Bridge, R. Q. (1999). Measurement and assessment of
geometric imperfections in thin-walled panels. Thin- Walled Structures, 33(2):103-126.
55

Blachut, J. and Galletly, G. D. (1995). Buckling strength of imperfect steel hemispheres.

Thin-Walled Structures, 23(14):1-20. 55

Bouhaya, L., Baverel, 0., and Caron, J.-F. (2010). Mapping two-way continuous elastic
grid on an imposed surface: Application to grid shells. p. 989-998. 28

Buchert, K. (1965). Buckling of framed domes. AISC Engineering Journal, 2(4):120. 29,
32

Bulenda, T. and Knippers, J. (2001). Stability of grid shells. Computers & Structures,
79(12):1161-1174. 29

Cadji, M. (2001). Wood work: Weald and downland timber gridshell. RIBA journal,
108(2):78-79. RW: Cadji, Miriam. 66

Calladine, C. R. (1995). Understanding imperfection-sensitivity in the buckling of thin-

walled shells. Thin-Walled Structures, 23(14):215-235. 29, 55

128
 Bibliography

Chapelle, D. and Bathe, K.-J. (2011). The finite element analysis of shells. Springer,
Berlin ; Heidelberg, 2 edition. 73

Chriss, S. and Wright, D. (1978). Analysis of square reticular pattern for hypars. Jour-
nal of Boston Society of Civil Engineers Section, American Society of Civil engineers,
64(4):149. 30

Day, A. (1965). An introduction to dynamic relaxation. The Engineer, 219. 28

Delay, T., Famer, S., and Jennings, T. (2009). Executive summary: Building the future,
today. Technical Report CTC766, The Carbon Trust. 17

Douthe, C., Baverel, 0., and Caron, J. F. (2006). Form-finding of a grid shell in composite
materials. Journal of the International Association for Shell and Spatial Structures,
47(150):53. 14

Engineering, I. A. (2010). A UI Command Reference Manual, volume 1. Watertown, MA.

23, 73

Forman, S. E. and Hutchinson, J. W. (1970). Buckling of reticulated shell structures.

International Journal of Solids and Structures, 6(7):909-932. 30, 31, 32, 52, 53

Gioncu, V. (1985). Instability problems in space structures. International Journal of

Space Structures, 1(3):169-183. 29, 32

Gioncu, V. (1995). Buckling of reticulated shells. state-of-the-art. International Journal

of Space Structures, 10(1):1-1. 29, 30

Hanaor, A. (1995). Design and behaviour of reticulated spatial structural systems. Inter-
national Journal of Space Structures, 10(3):139-149. Compilation and indexing terms,
Copyright 2010 Elsevier Inc.; Ml: Compendex; undefined; undefined; undefined; unde-
fined; undefined. 29

Happold, E. and Liddell, W. I. (1975). Timber lattice roof for the mannheim bundes-
gartenschau. The Structural Engineer, 53(3). 15

Harris, R., Haskins, S., and Roynon, J. (2008). The savill garden gridshell: design and
construction. The Structural Engineer, 86(17):27-34. 9, 22, 28

Harris, R., Romer, J., Kelly, 0., and Johnson, S. (2003). Design and construction of the
downland gridshell. Building Research & Information, 31(6):427-454. M3: Article. 28

Hart, S. (2001). A brilliant shell game at the british museum. Architectural Record,
189(3):149-154. PD: Illustration. 28

Holgate, A. (1997). The art of structural engineering: the work of Jrg Schlaich and his
team. Edition Axel Menges, Stuttgart. 49

129
Bibliography

Hutchinson, J. (1967). Imperfection sensitivity of externally pressurized spherical shells.

Journal of Applied Mechanics, 34:49. 126

Keuning, D. (2011). Sander and ney cover two historic buildings with glazed roofs. 28

Kollar, L. and Dulacska, E. (1984). Buckling of shells for engineers. Akademai Kiado
Wiley, Budapest; Chichester [W. Sussex]; New York. 31, 32, 52

Lpez, A., Puente, I., and Serna, M. A. (2007). Numerical model and experimental tests
on single-layer latticed domes with semi-rigid joints. Computers & Structures, 85(7-
8):360-374. 29

Ouroussoff, N. (2007). A delicate glass roof with links to the past. 28

Parliament, U. K. (2008). The climate change act. 17

Persson, P. 0. and Strang, G. (2004). A simple mesh generator in matlab. SIAM Review,
46(2):329-345. 50

Pople, N. (2001). Worth the weight: Weald and downland gridshell. RIBA journal,
108(9):80-81. RW: Pople, Nicolas. 66

Pople, N. (2002a). Off the grid: Weald and downland gridshell. RIBA journal, 109(5):36-
44. RW: Pople, Nicolas. 18, 66

Pople, N. (2002b). The other gridshell: Goethean science centre, pishwanton scotland.
RIBA journal, 109(8):54-55. 66

Schlaich, J. and Schober, H. (1996). Glass-covered grid-shells. Structural Engineering

Internatinoal,6(2):88. 26

Singer, J. and Abramovich, H. (1995). The development of shell imperfection measurement

techniques. Thin-Walled Structures, 23(14):379-398. 29, 55

Stansfield, K. (2007). Gridshell covers courtyard formed by historic buildings. 28

Stungo, N. (2001). Timber talk: Weald and downland timber gridshell. RIBA journal,
108(7):60-61. RW: Stungo, Naomi. 66

Sumec, J. and Sumec, J. m. d. a. s. (1990). Regular lattice plates and shells. Elsevier,
Amsterdam ; New York. Jozef Sumec.; Translation of: Regulrne mriezkov dosky a
skrupiny.; Bibliography: Includes bibliographical references and indexes. 29, 32

Thornton, J. A. (2000). The new parliamentary building - portcullis house. The Structural
Engineer, 78(18):17. 28

Timoshenko, S. (1961). Theory of elastic stability. McGraw-Hill, New York, 2d edition.

39, 52

130
 Bibliography

Vyzantiadou, M. A., Avdelas, A. V., and Zafiropoulos, S. (2007). The application of

fractal geometry to the design of grid or reticulated shell structures. Computer-Aided
Design, 39(1):51-59. 29

Wells, M. (2001). Taking shape: on site at weald and downland. RIBA journal, 108(4):75-
76. RW: Wells, Matthew. 15

Williams, C. J. K. (2001). The analytic and numerical definition of the geometry of the
British Museum Great Court Roof, pages 434-440. Mathematics & design 2001. Deakin
University, Geelong, Victoria, Australia. Proceedings of the the Third International
Conference on Mathematics & Design, MD-2001. Deakin University, Geelong, Australia,
July 3-5, 2001. 26, 28, 29

Wright, D. T. (1965). Membrane forces and buckling in reticulated shells. American

Society of Civil Engineers Proceedings, Journal of the Structural Division, 91:173-201.
30, 31, 32, 52

131