You are on page 1of 148

Structural Design of Glass Geometries

By use of the AE-methodology

Main Report
Jouke Lutgendorf
Technical University Delft
Civil Engineering

Master thesis - Apr 2014


cover:
Maximillian’s shell, Los Angeles

time frame:
April 14, 2014

author:
Jouke Lutgendorf
Student number: 1363344
Jouke.lutgendorf@gmail.com

committee:
prof.ir. Rob Nijsse ABT bv / Delft University of Technology
prof.dr.ir. Fred Veer Delft University of Technology
prof.ir. Sander Pasterkamp Delft University of Technology
ir. Mark Feijen I-Saac
ir. Diana de Krom ABT bv
”I suppose it is tempting, if the only tool you have is a hammer, to treat everything
as if it were a nail.”

-Abraham Maslow, The Psychology of Science (1966)-


This master’s thesis is the last step to complete my Civil Engineering degree at the
University of Technology in Delft, the Netherlands. During the last eleven months I
worked with a lot of pleasure on this thesis investigation and therefore I’m proud to
present the results in the form of this report. This thesis could not be realized without
the contribution of many others, some of whom I would like to mention in particular.

First of all, I would like to thank my graduation committee: Rob Nijsse, Fred Veer,
Sander Pasterkamp, Mark Feijen and Diana de Krom. My special gratitude goes out
to Mark Feijen, who supported me throughout the whole graduation process with the
most valuable information and his creative thoughts.

Secondly, I express my sincere gratitude to my family and friends for their support,
feedback and editorial comments. Especially my sister, Dorien Lutgendorf has a big
influence on the final result. Her critical eye, support and intelligence is very much
appreciated. I owe particular thanks to my mother, Winy Laan. Not only because of
her moral support but also for borrowing me her computer (sorry if i did any irre-
versible damage to it). Last but not least I would like to thank my girlfriend Johanna
Scholtyssek for her care, guidance and advice at all times.

Without the support of the people mentioned in this paragraph and all the others,
this report would have been much different. Thank you!

If you have any questions regarding the material in this report or you just want to
share your motivation or passion, feel free to contact me at any time.

Jouke Lutgendorf

Many thanks to:


Winy Laan - Peter Lutgendorf -Johanna Scholtyssek - Dorien Lutgendorf - Marlies
Lutgendorf - Francien Lutgendorf - Sophie Lutgendorf - Jose Lutgendorf - Wouter van der
Linde - Pieterjan Nijhuis - Rinke Feenstra - Thomas Nijssen
S U M M A RY

The aim of this thesis is to investigate the use and effects of advanced computational
power in the structural design process of glass structures. This investigation resulted in
the finding of the AE-model and AE-method which will be both extensively explained
in the upcoming chapters. Below only a brief summary is given of the research ques-
tion, method, contents and results of this thesis.

At present computational power used within the design process mainly influences
the speed at which architectural concepts can be generated and visualised. This re-
sults in a growing demand for engineers whom can respond as fast as architectural
concepts can develop. This causes the following research question to arise:

Is it possible to find a methodology that fully utilizes the available computational power and
can this improve the quality and speed of the engineering advise giving in the design phase of
glass structures?

The subject of the literature study is divided into two topics: ’Glass’ and ’Geometry’
. The first topic ’Glass’ covers the knowledge about glass as a (construction) material.
The ’Geometry’ part covers the computational description of geometries starting with
the description of points followed by curves and surfaces. From August 2013 to March
2014, time was spent on investigating the main research question. As an answer to the
research question the AE-model and AE-method are proposed. The main part of this
thesis report therefore includes the description of both, AE-model and AE-method.

The AE-model, which is the abbreviation of Architect and Engineering model, is


created by chaining the latest parametric computation software. The AE-model is
designed to give rapid insight in the structural behaviour of architectural glass de-
signs and gives feedback about stiffness, strength, stability and costs. In this manner
the feasibility of a concept can be tested and, if necessary, improved towards a more
balanced, affordable and feasible design. To avoid the inefficient use of the model
the AE-method is introduced. The AE-method sums up the non-technical know-how
around the AE-model. The AE-method proposes a workflow for the AE-model and
addresses the impact which the model has on the building industry.

The AE-methodology is tested during this theses on two virtual projects and on three
projects ’ in real life.’ The virtual projects concern a dome and an ellipsoid. The
projects in real life are: ’the new Fifth Avenue Apple cube’ in New York, ’The Globe’

7
in Kuwait and ’The Canakkale antenna tower’ in Turkey. The last project, in which the
AE-model was used for the structural design of a non-glass structure, was part of an
international competition and won the first price.

The AE-model and AE-method reveal that the research question can be answered
positively: It is possible to improve the speed and quality of the engineering advise
given in the design process of glass structures. Although the quality of the engineering
advise given throughout the whole construction process might not be influenced, the
latest computational power can improve the speed at which engineering advise can
be given. As a consequence a new work flow need to be introduced. The model and
method together can bring the quality of engineering advise given during the design
phase quickly to a high level.

8
CONTENTS

i introduction 15
1 introduction 17
1.1 Research question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.3 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.4 Outline of the report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

ii lessons from literature 21


2 literature study - glass 23
2.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 History of glass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Glass the material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Engineering in glass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 literature study - geometry 31
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4 Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

iii the ae-model 35


4 general description 37
4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Description of AE-model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Model input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.4 Model output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.5 Concept and structure of the AE-model . . . . . . . . . . . . . . . . . . . 40
4.6 Used software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5 ae-model elements 45
5.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.2 Module 1: Reading architectural geometries . . . . . . . . . . . . . . . . 46
5.3 Module 2A: Panelization method; approaching . . . . . . . . . . . . . . . 51
5.4 Module 2B: Panelization method: meridian . . . . . . . . . . . . . . . . . 61
5.5 Module 3A: FE-Quad meshing: line support . . . . . . . . . . . . . . . . 67
5.6 Module 3B: FE-Tri meshing: line support . . . . . . . . . . . . . . . . . . 79
5.7 Module 3C: Windload on spheres . . . . . . . . . . . . . . . . . . . . . . 87
5.8 Module 4A: Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

9
Contents

5.9 Module 4B: Analysis cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 98


5.10 Module 4C: Load cases and load combinations . . . . . . . . . . . . . . . 100
5.11 Module 4D: Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

iv the ae-method 105


6 general description 107
6.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.2 Introducing the AE-method . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3 Purpose and place of AE-method in the construction process . . . . . . 109
6.4 The workflow of the AE-method . . . . . . . . . . . . . . . . . . . . . . . 111
6.5 Benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.6 Field of application and boundaries . . . . . . . . . . . . . . . . . . . . . 116
7 using the ae-method: some examples 119
7.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.2 The Dome . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.3 The Ellipsoid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.4 The Globe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.5 The New Apple Cube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.6 The Canakkale Antenna Tower . . . . . . . . . . . . . . . . . . . . . . . . 130

v conclusion and recommendations 133


8 conclusion 135
8.1 Technical conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.2 Process-oriented conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 136
9 recommendation 137

vi appendix 141
a categorisation of surfaces 143
b screenshot ae-model: ellipsoid 145

10
LIST OF FIGURES

Figure 1 Calculation by isolating the structural members . . . . . . . . . 18


Figure 2 Complex designs, hard to distinguish structural layers . . . . . 18
Figure 3 Overview today’s glassproduction . . . . . . . . . . . . . . . . . 24
Figure 4 Stress-strain diagram of various materials . . . . . . . . . . . . . 25
Figure 5 Material properties of float glass . . . . . . . . . . . . . . . . . . 27
Figure 6 Pre-Tensioning by heat (left) and chemicals (right) . . . . . . . . 27
Figure 7 Laminated glass (left) and glass structural components (right) . 28
Figure 8 Left and right coordinate system . . . . . . . . . . . . . . . . . . 32
Figure 9 Parametric curve representation [3D] . . . . . . . . . . . . . . . . 32
Figure 10 Parametric paraboloid surface representation [3D] . . . . . . . . 33
Figure 11 Schematization of the AE-method and model . . . . . . . . . . . 37
Figure 12 Extern and intern architectural 3D models . . . . . . . . . . . . 39
Figure 13 Chain of modules in AE-Model . . . . . . . . . . . . . . . . . . . 40
Figure 14 Overview of modules made . . . . . . . . . . . . . . . . . . . . . 41
Figure 15 Used modelling software . . . . . . . . . . . . . . . . . . . . . . . 43
Figure 16 Summary module 1 . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Figure 17 Reading cross . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Figure 18 Positive directions . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Figure 19 Organisation of output . . . . . . . . . . . . . . . . . . . . . . . . 50
Figure 20 Summary module 2A . . . . . . . . . . . . . . . . . . . . . . . . . 51
Figure 21 Left: Architectural surface, Right: Panellized approximation . . 52
Figure 22 Left: Architectural surface, Right: Panellized approximation . . 53
Figure 23 Guiding panellization by rotating the reading cross . . . . . . . 54
Figure 24 Panelization of double and single curvature surfaces . . . . . . 55
Figure 25 Step by step explanation of algorithm [2D] . . . . . . . . . . . . 56
Figure 26 Step by step explanation of algorithm [3D] . . . . . . . . . . . . 57
Figure 27 Approximation overview . . . . . . . . . . . . . . . . . . . . . . . 58
Figure 28 Geometries with curvature . . . . . . . . . . . . . . . . . . . . . . 59
Figure 29 Summary module 2B . . . . . . . . . . . . . . . . . . . . . . . . . 61
Figure 30 Step by step explanation of Module 2B . . . . . . . . . . . . . . 63
Figure 31 Example of output . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 32 Example of counting and culling . . . . . . . . . . . . . . . . . . 66
Figure 33 Summary module 3A . . . . . . . . . . . . . . . . . . . . . . . . . 67
Figure 34 Input (left) and output (right) of module 3A . . . . . . . . . . . 69
Figure 35 Connection variation (1 of 2) . . . . . . . . . . . . . . . . . . . . 70
Figure 36 Connection variation (2 of 2) . . . . . . . . . . . . . . . . . . . . 71

11
List of Figures

Figure 37 Rough steps in scripting . . . . . . . . . . . . . . . . . . . . . . . 72


Figure 38 Densifying FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Figure 39 Adjusting width of zone 2 . . . . . . . . . . . . . . . . . . . . . . 73
Figure 40 Constraints and two property zones . . . . . . . . . . . . . . . . 74
Figure 41 Exposing zone 2 to various loadcases . . . . . . . . . . . . . . . 75
Figure 42 Result test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Figure 43 Top: Linear FE, Bottom: Quadratic FE . . . . . . . . . . . . . . . 78
Figure 44 Hourglass failure mode . . . . . . . . . . . . . . . . . . . . . . . 78
Figure 45 Summary module 3B . . . . . . . . . . . . . . . . . . . . . . . . . 79
Figure 46 Input (left) and output (right) of module 3B . . . . . . . . . . . . 80
Figure 47 Sketch of script connecting problem . . . . . . . . . . . . . . . . 82
Figure 48 Script explanation part 1 . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 49 Script explanation part 2 . . . . . . . . . . . . . . . . . . . . . . . 85
Figure 50 Summary module 3C . . . . . . . . . . . . . . . . . . . . . . . . . 87
Figure 51 Explanation of windforce formula . . . . . . . . . . . . . . . . . 90
Figure 52 Example of Cpe calculation . . . . . . . . . . . . . . . . . . . . . 91
Figure 53 Example of output, sphere topview . . . . . . . . . . . . . . . . 92
Figure 54 Explanation method, part 1 . . . . . . . . . . . . . . . . . . . . . 94
Figure 55 Explanation method, part 2 . . . . . . . . . . . . . . . . . . . . . 95
Figure 56 Summary module 4A . . . . . . . . . . . . . . . . . . . . . . . . . 96
Figure 57 Explanation of module 4A . . . . . . . . . . . . . . . . . . . . . . 97
Figure 58 Summary module 4B . . . . . . . . . . . . . . . . . . . . . . . . . 98
Figure 59 Explaination of module 4B . . . . . . . . . . . . . . . . . . . . . . 99
Figure 60 Summary module 4C . . . . . . . . . . . . . . . . . . . . . . . . . 100
Figure 61 Explaination of module 4C . . . . . . . . . . . . . . . . . . . . . 101
Figure 62 Explaination module 4C . . . . . . . . . . . . . . . . . . . . . . . 102
Figure 63 Summary module 4D . . . . . . . . . . . . . . . . . . . . . . . . . 103
Figure 64 Explain module 4D . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Figure 65 AE-method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Figure 66 Place of AE-method in the construction process . . . . . . . . . 110
Figure 67 Classical design sequence . . . . . . . . . . . . . . . . . . . . . . 112
Figure 68 Proposed design sequence . . . . . . . . . . . . . . . . . . . . . . 113
Figure 69 Field of application examples . . . . . . . . . . . . . . . . . . . . 118
Figure 70 Starting sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Figure 71 Overview of ’The dome’ . . . . . . . . . . . . . . . . . . . . . . . 122
Figure 72 Output AE-method . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Figure 73 Top: Geometry, Middle:Snow Load, Bottom:Wind Load . . . . . 125
Figure 74 Overview project ’The Globe’ . . . . . . . . . . . . . . . . . . . . 127
Figure 75 Top: New Apple Cube, Bottom; FE Geometry . . . . . . . . . . 129
Figure 76 Model geometric variables . . . . . . . . . . . . . . . . . . . . . . 131
Figure 77 Top: Architectural impression, Bottom: Output AE-model . . . 132

12
List of Figures

Figure 78 2 CV electric scheme . . . . . . . . . . . . . . . . . . . . . . . . . 138

13
Part I

INTRODUCTION
1
INTRODUCTION

The building industry is an unique industry. It distinguishes itself from other in-
dustries for instance by its size of organisation, its unique products and amount of
stakeholders. Another characteristic is its strong dependancy on technological inno-
vation. History shows multiple examples of technological discoveries that changed
and revolutionized the building industry. Although the effects of these technological
innovations are not always perceived as positive they are clearly noticeable even today.
New building codes, new 3D drawing programs, new construction methods and even
totally new building typologies are examples of change which is directly or indirectly
pushed by technological innovation.

A moment in history where technology accelerated exceptionally was during the eight-
ies and nineties. In this time period the computer was invented. The computer rapidly
became accepted as a working tool, initially as a replacement for the typewriter. In
the 30 years since people spent time programming knowledge of for example math
and physics in it and tries to increase its processing speed. Not long ago the computer
reached a level were it contained enough of both, knowledge and calculation speed to
be of interest for advanced engineering calculations

At present, the influence of computational power is entering the construction indus-


try and is gradually changing the construction process. Not only from a technical
point of view but also socially in terms of role and task division. Within the field
of architecture, where innovation is a selling point, new computational technologies
have no problems being accepted and are smoothly embedded. Computers used in
the design process increases the amount of architectural freedom and the speed in
which complex architectural concepts can be generated. As a consequence engineers
face more challenging designs which causes the structural calculation to become more
complex as well. Whereas traditional ways of structural calculation (fig 1) is based on
calculating isolated structural members the new computational driven designs have a
more integrated load bearing system. This makes structural calculation by isolating
the structural members inadequate or even impossible (fig 2).

17
introduction

Figure 1.: Calculation by isolating the structural members

Figure 2.: Complex designs, hard to distinguish structural layers

With architectural designs pushing boundaries, there is a growing need to commu-


nicate meaningfully between both islands of information. With vanishing borders
between architecture and engineering both parties need to be fully aware of the re-
strictions in either domain. These complex designs, as shown in figure 2, require a
close partnership between architect and engineer, exchanging information and knowl-
edge along the way. In order to fulfil this demand, engineers need to respond to the
increasing computational complexity and the speed of concept generation by finding
new means and methods of calculation. The matching of old calculations methods
based on isolating structural members with these complex integrated designs seems
not adequate and causes a question to arise: Is it possible to find a methodology that
fully utilizes the computational power which is present in the world and can this im-

18
1.1 research question

prove the quality of the engineering advise giving in the design phase? What impact
would such a model or method have on the work flow of engineers in the design
phase? Can the design process be improved?

The research field of this thesis is narrowed to the specialisation of glass structures.
This has various reasons. The use of glass has evolved from purely decorative or ar-
chitectural to structural, encouraging glass technologies to develop. Glass is gaining
popularity in architecture causing engineers to face more and more challenging glass
designs. Glass is often used in highly expressive designs and various kinds of glass
free-forms already grace people with their presence. Due to the glass’s mechanical
properties and its critical breaking behaviour the engineering part in the development
of glass designs is essential. The constructing of glass structures demands for a good
cooperation between the architect and engineer from an early stage in the design. This
specialistic area is therefore very suitable for this thesis investigation and can really
benefit from a more detailed level of engineering advise in the design phase.

1.1 research question

Is it possible to find a design methodology that fully utilizes the available compu-
tational power and can this improve the quality of the engineering advise giving in
the design phase of glass structures?

1.2 objective

The research question formulated above leads to the following main objective: The
objective of this thesis is to find a methodology which can improve the quality of the
engineering advise giving during the design phase of glass structures.

The secondary objective is to investigate the impact and the consequences such a
model or method will have on the work flow of engineers and architects.

1.3 approach

To be able to reach the objectives stated in the previous paragraph an adequate ap-
proach was chosen. The approach is divided into 6 steps. The steps are shown on the
next page in chronological order. To get a general understanding of glass and the com-
putational modelling of geometries the literature study is divided into both subjects:
glass and geometry. After the literature study, in step 2, model concepts have been
generated (what can be implemented, what not? will it work?). In step 3 the model

19
introduction

is verified and refined by testing the model on various test cases. In step 4 the final
model is made, implementing some or all model concepts from step 2 and the lessons
learned from step 3. In step 5 the consequences of the designed model is investigated.
The last step, step 6, included the finishing of the report and the presentation.

Step 1: Understanding glass and geometry

Step 2: Investigate model concepts

Step 3: Refine and verify chosen model concepts with test cases.

Step 4: Make the final model.

Step 5: Investigate the impact on the work flow of engineers and architects.

Step 6: Make conclusions, recommendations and documentation.

1.4 outline of the report

This report is divided into six parts being ’Introduction’, ’Literature’, ’The AE-model’,
’The AE-method’, ’Conclusion and recommendations’ and ’Appendix’. These parts are
related to the steps which are described in the previous section ’Approach’.

The first part: ’Introduction’ is the part which you are reading at the moment. The
literature part will start right after and shows the result of step 1 as described in the
previous section. It includes a brief summary of the literature study done. The sum-
mary will focus on the lessons learned from literature and the specific aspects which
are used to develop the model and method. The complete literature study is gathered
in a separate book. After the literature part the report continuous with the explana-
tion of the AE-model in part III. This is the result of step 2 and 4 as described in the
’Approach’. The AE-method, part IV, will explain why the model should come with a
method. This is the result of step 5. The AE-method part will end with five examples
of the whole proposed procedure of using model and method and shows end-results
and outputs. In the final part, part V, the conclusions and recommendations are ad-
dressed. Focussing on the impact of the proposed model and the recommendations
for further development.

20
Part II

L E S S O N S F R O M L I T E R AT U R E
2
L I T E R AT U R E S T U D Y - G L A S S

2.1 abstract

This chapter includes a brief summary of the literature study about glass.

2.2 history of glass

Since the discovery of glass around 1500 BC, glass has evolved from purely decora-
tive to structural [Phil and Seff, 1999]. In the first years after its discovery the use of
glass changed from small glass covered clay-objects for decorative means to a more
practical and widely use of glass in the Roman Empire. The development in the years
after were pushed by the strive to purify the glass and to create bigger panes which
could cover window openings[Merchant, 1998]. Steps forward where most of the time
caused by technological improvements to the production process of glass; melting fur-
naces which could reach higher temperatures or an improved composition of the glass
batch.

In times of the industrial revolution (around 1800) the glass industry improved dras-
tically. Various mechanical machinery was developed to modify molten glass in flat
panes. The mechanical systems rapidly followed one another [GBO, 2012]. Examples
are the Fourcault system, the Pittsburgh-system and the Libbey-Owens-system. Finally
in 1959 sir Alastair Pilkington invented the ’Float process’ which is still used today.

Today’s glass productions shows a big variety in glass composition, modification tech-
niques and methods to assemble glass. An overview can be found in figure 3. A full
explanation of each of the points shown in the figure can be found in the Literature
study.

23
literature study - glass

Figure 3.: Overview today’s glassproduction

2.3 glass the material

2.3.1 Definition

The nature of glass causes problems for definition. This has two main reasons. The
first concerns the question whether glass is a supercooled liquid or not. Supercooling
or in other words undercooling of a liquid is a process whereby a liquid cools down
without forming a solid. Glass never becomes a true solid nor liquid, in contradiction
with for instance water when frozen. Secondly glass exist in many different forms
including different batch ingredients, making it even harder to find a fully covering
definition. The chemical nature of glass is, in contradiction with most other materials,
not suitable to define glass.
What can truly be said about glass, regardless any specific type, covers only some
information about the structure of the material (actually about the lacking of any) and
the temperature at which it will start to to find it’s semi-solid state:

”Glass can be called an amorphous solid completely lacking in long range, periodic
atomic structure, and exhibiting a region of glass transformation behaviour” [Shelby,
2005]

2.3.2 Physical properties

Glass has the ability to refract, reflect, and transmit light without scattering it. This
optical property makes glass truly unique and plays an essential role in the creation
of comfort in buildings. The amount of refraction, transmittance and reflection can be

24
2.3 glass the material

altered by modification methods until the desired property is obtained. This playing
field of possibilities makes glass an interesting material. Glass can be considered as
a durable material. Due to its chemical components and structure glass has a high
resistance to:

• Water percolation
• Corrosion
• Salt water
• Strong acids
• Organic solvents
• Ultra-violet radiation

2.3.3 Mechanical properties

Failure behaviour
Only looking at the physical and optical properties there might be thought that glass is
the ultimate construction material. Sadly enough this is not the case. The stress-strain
diagram of glass is shown in figure 4 and shows the stress strain relation of glass
together with five other materials. The figure shows that glass has no plasticity phase
which means that it only deforms elastically before showing explosive failure. This
explosive failure is unfortunate when using glass for structural purposes. A positive
aspect of materials which do show pre-breaking behaviour is that they warn passers-
by before initial collapse.

Figure 4.: Stress-strain diagram of various materials

25
literature study - glass

Strength
The strength of glass depends on a variety of factors. Most of them are related to the
breaking behaviour of glass or in other words the lack of plasticity. Although glass
seems like a perfect homogeneous material it is never perfect. On the surface of the
glass and especially on the edges micro cracks are present due to glass’s fabrication
process (sawing, cutting, drilling, edge and surface grinding). Whenever a glass pane
is loaded by wind for instance, stress can get locally higher than the covalent bond
between the atoms around those micro cracks and the crack will start to widen. The
atoms around the micro crack lack the capability to regroup or replace themselves
because they cannot deform plastically. This results in an even higher stress level and
an unstoppable cracking behaviour [F.A.Veer, 2007]. Factors which imply cracks in
glass can therefore be seen as strength reducing factors. All the factors which affect
the glass strength are summarized below:

• Surface condition and edge quality


• Load duration
• Environmental condition, especially humidity
• Stress distribution on the surface
• Size of stressed area
• Damage of glass surface, flaws and cracks

The Eurocode prescribes various formulas to calculate the strength of glass. Within
this formula the factors listed above are taken into account and considered as strength
reducing. They reduce the initial tensile (float glass) bending strength of fgk =45
[N/mm2 ]

Overview material properties


To get a basic feeling for other mechanical properties related to glass an overview table
is made shown in figure 5. The table is made for Soda-silica based unmodified glass.

2.3.4 Improving mechanical properties

There are various possibilities to improve the properties of glass. Some are focussed
on reducing the chance on micro cracks by optimizing the production process on
perfection while others do really improve the mechanical properties and thus the
strength value which is used for structural calculations. The latter is the case with
pre-tensioning. The three main methods to pre-tension glass are:

• Type 1: Heat strengthened glass [Heat treatment]


• Type 2: Toughened / Tempered glass [Heat treatment]
• Type 3: Chemically pre tensioned glass [Chemical treatment]

26
2.3 glass the material

Figure 5.: Material properties of float glass

Thermally pre tensioning (Type 1 and 2) will cause pretension in the glass by cool-
ing down the outer face of the glass pane more rapidly than the inside. The inside
glass, which has still a high temperature, wants to expand but is restricted by the
cooled down outer layer. This will cause the hotter inside glass to get stuck and there-
fore build up tension. This residue of stress inside the glass causes a different breaking
pattern than non-pretensioned glass and is therefore clearly recognisable when broken.
Chemical pre-tensioning (Type 3) uses the chemical exchange between small ion’s in
the glass and the bigger ion’s in a chemical bath. This change in volume in the glass
surfaces causes an increasing pressure on the glass surface.

The methods will lead the following characteristic bending tensile strength: Type 1:
70 [N/mm2 ], Type 2: 120 [N/mm2 ], Type 3: 150 [N/mm2 ]. Standard float glass has a
tensile bending strength of 45 [N/mm2 ].

Figure 6.: Pre-Tensioning by heat (left) and chemicals (right)

27
literature study - glass

2.4 engineering in glass

Whenever glass is used as a load bearing component it requires special treatment. To


make sure glass structures warn before collapse the missing plasticity phase in the
material needs to be simulated in an alternative way. This can be achieved by com-
bining glass with materials which do show plastic deformation behaviour. Although
there are more methods on the market, the most common method is sandwiching an
adhesive substance between multiple glass panes.
1. Laminated glass with resin interlayer:
Glass panes can be glued together by a resin made from for instance epoxy,
acrylic or polyester. The resin cures by chemical reaction or UV light. The advan-
tages of resins are the fact that for the lamination process a large autoclave is not
necessary allowing a larger pane size to be laminated. The main disadvantage,
when compared with foil interlayers, is that a lower load bearing capacity is left
whenever one of the glass panes break.’
2. Laminated glass with foil interlayer:
Glass can also be laminated with foils. These foils can be made from PVB
(polyvinylbutyral), EVA (ethyleenvinylacetaat), SGP (ionoplast) or PU (poly- ure-
thane). The main advantage of lamination with foils is that they have a higher
residual stress when one of the glass panes is broken. The shattered pieces of
glass stick to the foil and therefore residual catenary action is left in the assembly,
preventing sharp pieces of glass from falling down at the same time.

Figure 7.: Laminated glass (left) and glass structural components (right)

2.4.1 Engineering glass structures

Because the structural use of glass is relatively new the structural design codes do not
cover this subject sufficiently yet. At the same time, since glass can break, engineering

28
2.4 engineering in glass

in glass can quickly become complex. The lack of any norm description about ’sec-
ondary’ issues such as serviceability and robustness makes glass engineering depend-
ing on knowledge and experience[F.P.Bos, 2009]. The paper ’Glass structures- learning
from experts’ written by D. Honfi and M. Overend [Honfi and Overend, 2013], includes
a summary of interviews done with fourteen experts in the glass engineering industry.
These experts indicate that the aspects ’Consequences’, ’Robustness’ and ’Serviceabil-
ity’ are insufficiently covered by the norms. These three aspects are described below.

Consequences
When designing in glass the consequences of glass failure should always be kept in
mind. The consequences of failure are higher when more people are present in the
building. Therefore the primary distinction should be made between public and pri-
vate structures. Extra safety can be achieved by using more glass panes in the lami-
nated assembly than strictly necessary. In the design process distinction can be made
between primary, secondary and tertiary load bearing elements and the amount of
lamination panels need to be chosen accordingly. Since consequences can be fatal
taking into account al exposures is important. The additional non-natural exposures
mentioned during the interviews are: blast, impact, vandalism, damage due to main-
tenance, thermal stresses and NiS breakage.

Robustness
the attendants of the interviews mentioned ’robustness’ being the most important fac-
tor in glass design Acccording to the PHD thesis of F.P. Boss [F.P.Bos, 2009] a robust
structure can be achieved at three different levels.

1. Robustness at material level;


The robustness at the material level can be achieved by choosing the right type
of glass. Pre stressed glass for instance can be more vulnerable for scratches and
nickel sulphide inclusion while it has a better resistance to exposures such as
general loading, thermal stresses and surface impact.

2. Robustness at component level;


Choosing the right glass lamination assembly can also help achieving robust
structures. Laminated units composed of solely pre-stressed panels should be
avoided. These units do not have any residual stress left when broken. For
this reason at least one of the laminated plies need to be made of normal float
glass or chemically pre-stressed glass which breaks in large pieces when broken.
The catenary action left in the interlayer after breakage can prevent progressive
collapse.
Detailing is of great importance. A bad choice of connection type might lead to
stress concentrations and eventually to failure.

29
literature study - glass

3. Robustness at system level;


Whenever a structural component fails or needs to be replaced an alternative
load path should be provided. This means that in the design phase the remov-
ing of a complete glass panel should be taken into account. Secondly, statically
indeterminate constructions should be avoided as much as possible because the
underestimation of stiffness in indeterminate glass structures might lead to fail-
ure. An unsolved difficulty mentioned during the interviews are the uncertain-
ties about which load cases should be used when calculating secondary load
paths and the post breakage behaviour of the structure.

Serviceability
Serviceability requirements for glass structures are vague and leave room for discus-
sion. The requirements are insufficiently described in the norms. However especially
with floors, stairs and pedestrian bridges they can be governing. With beam, fins can-
tilevers and columns strength criterion are usually governing.

30
3
L I T E R AT U R E S T U D Y - G E O M E T R Y

3.1 abstract

After the computer was invented the use of it spread quickly through various indus-
tries. In design industries such as the car industry there came a need to draw smooth
curves and surfaces on computers in order to speed up the design process. Computer
aided design soon became the standard. Nowadays computer aided design is enter-
ing the engineering world enabling complex designs to be visualised and calculated.
This chapter is a brief summary of the literature study about geometry and will give a
short introduction about the background knowledge used to describe the geometries
and structures within this thesis.

3.2 points

To define a point in 3D space we need three values: Two values are necessary to deter-
mine the exact horizontal location and the third value determines the vertical location.
These values are fixed by use of a coordinate system described by three perpendicular
orientated axis named: x,y and z.

In this document the right handed coordinate system is used. The thumb corresponds
with the x-axis, the index finger to the y-axis and the middle finger to the z-axis. See
figure 8.

3.3 curves

Curves can be described in multiple mathematical ways. These include the implicit,
explicit and parametric description [Pottman, 2007]. When comparing these options
of describing curves there can be concluded that the ’parametric mathematical repre-
sentation of curves’ is most suitable for curve programming and manipulations by use
of computers. This because the parametric description has next to the fact of having
only one input variable (t) the benefit of having a trajectory.

31
literature study - geometry

Figure 8.: Left and right coordinate system

Figure 9.: Parametric curve representation [3D]

Curves drawn on computers use this knowledge on the background. Now the right
method to describe curves is known, the question how to input curves on computers
remains still unsolved. For the input of curves on computers another invention was
necessary. Most curves on computers can be inserted by ’control points’ which can
be placed by clicking on the mouse button. The computer afterwards interpolates be-
tween this points a curve. This interpolation between the control points can be done
with various algorithms.

Nowadays, the most commonly used curve algorithm is the NURBS curve. This
method introduces next to the control points a specific weight to each control point
allowing the curve to be pulled towards or pushed away from the control point. In
this way it becomes possible to draw smooth free-form curves but also mathematical
curves such as circles.

32
3.4 surfaces

3.4 surfaces

The theory about curves can be extended to surfaces. An example of the parametric
description of a surface can be found in figure 10 [WolframMathWorld, 2014]. The
polynomial functions transfer the 2D domain into 3D space. In this manner various
surfaces can be described.

Figure 10.: Parametric paraboloid surface representation [3D]

3.4.1 Categorisation of surfaces

With the above described description of surfaces an unlimited amount of surfaces can
be described and made. The surfaces can be categorises in various ways all having
their own benefits and characteristics. Below two ways to categorise surfaces are de-
scribed. The first categorisation divides all the surfaces in three groups:

• Mathematical surfaces. These are surfaces which can be described by math func-
tions. Such as spheres, ellipsoids and cylinders.

• Structural surfaces. Structural surfaces are shaped by the influence of a specific


load case. This load case can be for instance gravity, wind loads or pneumatic
pressure/suction.

• Freeform surfaces. Freeform surfaces do not fit in one of the above described
groups and have a random shape.

Although the above described description can be useful to understand the nature
of the surface encountered, it can also be perceived as to rough. Within each group
a lot of subgroups can be found and there are also surfaces which fit in multiple
groups. For the computational description of surfaces another categorisation can be
useful focussing on the method to describe them on computers. Since in the modern

33
literature study - geometry

construction process all structures are drawn on computers this categorisation can be
more useful. This computational categorisation can be found in appendix A.

34
Part III

THE AE-MODEL
4
GENERAL DESCRIPTION

4.1 abstract

This part ’ AE-model’ contains two chapters. The first chapter includes the general
description of the AE-model. This chapter contains general information about the AE-
model such as the description of the models input and output. It will give answers on
questions like: ”What does the model contain?” and ”what kind of software is used.”
Chapter two will go more into detail on the contents of the model itself. In this chapter
each module is explained separately in high detail.

4.2 description of ae-model

The AE-model is an abbreviation of ’Architect & Engineering model’. As the name


suggest it represents a model where the architect and engineer can work with together
in order to evaluate architectural concepts in an engineering sense. The AE-model is
designed to give rapid insight in the structural behaviour of architectural designs and
gives feedback about stiffness, strength, stability and costs. The AE-model is build by
uses of parametric programming which gives the modeller the opportunity to change
parameters until the desired end-result is obtained. In this manner the feasibility of a
concept can be tested and, if necessary, improved towards a more balanced, affordable
and feasible design.

Figure 11.: Schematization of the AE-method and model

37
general description

What: Structural design model for structural glass geometries, having no, one
or two main curvature directions. The model focusses on the
dimensioning of structural components, cost estimation and structural
safety within the playing field of a given architectural concept.
Why: To support the decision making process in the preliminary design phase.
Who: The model needs to be handled by a modelling-team and is
meant to shape the wishes and desires of the leading architect
and engineer.

4.3 model input

The possible methods of implementing architectural geometries in the AE-model can


be divided in two options. Both options are described below:

1. The first option reads out an architectural shape by loading an already existing
3D model. This model is externally made by the Architect. Option 1 is visualised in
the top part of figure 12. To ensure a good implementation of the architectural 3D
model, best is to make the architectural model in Rhinoceros. Loading 3D models out
of other software programs as for instance sketch-up is possible, but this is not tested
during this graduation.

2. The second method implies the programming of the shape manually from scratch
inside the AE-model. This can be necessary when only an architectural sketch is avail-
able at the start. Depending on the complexity of the architectural model, method
2 can take more time. But a big advantage of method 2 is the fact that, when pro-
grammed, the architectural shape is parametric and can be varied during the design
process. In other words, when the geometry is programmed manually it gives the
modeller the possibility to change the geometry when desired. Since the calculations
are parametric as well, they will follow the adapted geometry automatically.

38
4.4 model output

Figure 12.: Extern and intern architectural 3D models

4.4 model output

The AE-model has two outputs. The two outputs are described below:
1. The first output sends the data generated inside the AE-model to the calculation
software. The calculation software runs as a ’background solver’. This means it solely
digests the data which the AE-model sends. The calculation program then executes
the structural calculation and returns the calculated output back to the AE-model.
The AE-model takes over the ’traditional’ actions which are done inside a calculation
program such as determining load cases, analysis cases and defining material proper-
ties. The calculation program only executes the structural calculation.

2. The second type of data output is towards a spreadsheet program (for instance
Microsoft Excel). All the data which is added to the architectural geometry in the
AE-model is saved and exported towards a standardized spreadsheet template. The
spreadsheet contains all the known information about the geometry such as structure
height, glass area, steel section length, steel weight etc.This information can be used
for post-processing, such as for instance cost calculation.

An overview of the software used can be found in section 4.6: ’Used software’

39
general description

4.5 concept and structure of the ae-model

The toolbox itself is structured and is build in a modular way. Figure 13 shows a
schematization of the contents of the toolbox. Within the AE-model different modules
are attached to form a chain.

Figure 13.: Chain of modules in AE-Model

The modules are made in such a way that the output of module 1 is the input of
module 2 and so on. The coupling of different modules will lead to a different end
result. The modules are in fact parts of script. An overview of all the modules written
during this thesis can be found in figure 14. All modules are individually described
in chapter 5. Figure 14 shows the division of the modules into four categories. The
first category concerns the reading of architectural geometries. The second category
’Panelization’ describes various way to cover architectural geometries with flat panels.
The third category ’ Pre-engineering’ prepares all the data known so far for calculation
purposes. Finally category four contains modules in which engineering properties are
defined.

When specific modules are picked from the list, they can be attached to create a chain.
Once set, variables inside the chain of modules can be varied to tune the outcome.
Furthermore the geometry at the beginning can be altered. When altered, the changed
geometry in module 1 will automatically run through all the modules in the chain and
finally, in the last module, will be calculated in its new gesture.

40
4.5 concept and structure of the ae-model

Figure 14.: Overview of modules made

41
general description

4.6 used software

The programs used for the AE-model are summarized on the page on the right in
figure 15. The fact that the AE-methodology uses the well known programs Rhino
and Oasys GSA, makes the output easy to check for people which are not familiar
with the AE-methodology. Secondly, the ’after-engineering’ can be done within Oasys
GSA without the need of scripting knowledge. But, worthwhile noticing, the technol-
ogy (software) used within this thesis is limited to the technology within reach of the
author of this report. Other software combinations where, due to missing licensing,
impossible to investigate. Possibly they can replace one of the software programs used,
still maintaining the AE-methodology as described in this thesis.
The software programs Rhinoceros, Grasshopper and Excel are not made for engi-
neering purposes (yet). They only function as input or output to the program Oasys
GSA. Oasys GSA contains the necessary math functions to calculate structures. Oasys
Limited, which is responsible for the development of the Oasys GSA software is es-
tablished in 1976. In the past years Oasys GSA has proven to be a stable platform to
calculate (complex) structures. A big benefit of the described combination of software
in figure 15 is that the programs Rhinoceros and Oasys GSA have a good visual output.
This helps making the AE-model suitable for communicative means.

42
4.6 used software

Figure 15.: Used modelling software

43
5
AE-MODEL ELEMENTS

5.1 abstract

This chapter describes all the modules which are included in the AE-model during
this Thesis. Each module is explained using the same set of section headings. The
section headings are: Function, Description, Input, Method, Output and Possibilities
for further development. At the beginning of each section a summary is given in the
form of a figure. This figure sums up the most important information about each
module. Between the standardized way of explanation background information is
given. These can be found in the grey background boxes.

45
ae-model elements

5.2 module 1: reading architectural geometries

Figure 16.: Summary module 1

46
5.2 module 1: reading architectural geometries

5.2.1 Function

The name of this module is: ’reading architectural geometries’. As the name implies,
this module ’reads’ an architectural geometry and gathers all necessary information
about the architectural shape, such as boundary surfaces, points and curves. This
information can be used as an input for the following modules.

5.2.2 Description

This module is the first in a chain of modules leading to a finished concept and can be
seen as a pre-processing module. The module pre-processes the architectural shape in
such a way that it collects data about boundary surfaces, curves and points.
The AE-model is meant to be used during the design phase of a project. In the design
phase there are still great uncertainties around the architectural geometry and there-
fore the shape might vary in time. For this reason the module is scripted in such a
way that it can cope with these alteration. The module is suitable for shape categories
1,2,3,4 and 7. (as described in the literature study section 6.4) The information gained
in this module is organized, named and stored as output (see subsection 5.3.5).

5.2.3 Input

Generally speaking this module has two inputs: A surface component and a brep
component.

Components
A surface component represents a list of surfaces. Inside the modelling program called
Rhinoceros the surfaces are described by use of the parametric description. For
this reason they all have an area and they all have U,V domains as described in the
literature study section 6.2. The known information about each surface (Area and U,V
domains) is stored within its component.

Brep components represents Breps. Breps are different from surfaces since they
can be a shape consisting of multiple and/or trimmed surfaces. A cube for instance
can be represented in 1 brep component filled with one brep but also in a surface
component filled with 6 seperate surfaces.

The brep and surface component necessary for the input for this module can be ’filled’
with the corresponding surfaces and/or breps in two ways:

1. The first method reads out the architectural shape by loading an already existing
Architectural 3D-model made by the architect. To ensure a good implementation of

47
ae-model elements

the architectural 3D model in this module the architectural model needs to be made
in Rhinoceros. Depending on whether the architectural shape represents a brep or a
surface the architectural shape can be loaded into either the brep component and/or
the surface component. Within the program Grasshopper the ’loading’ of a compo-
nent is called ’setting’ . The setting of components is part of the standard procedures
in Grasshopper and takes no more than two mouse clicks.

2. The second method implies the programming of the shape manually from scratch.
This can be necessary when only a architectural sketch is available at the start. De-
pending on the complexity of the architectural model, method 2 can take more time.
But a big advantage of method 2 is the fact that, when programmed, the architectural
shape is parametric and can be varied during the design process. In other words,
when the geometry is programmed manually it gives the modeller the possibility to
change the geometry when desired. When altered, the changed geometry in module
1 will automatically run through all the modules in the chain and finally, in the last
module, will be calculated in its new gesture.

5.2.4 Method

The basic idea behind this module is the introduction of a ’reading cross’. This reading
cross is shown in figure 17 in red. As shown in the picture the reading cross includes
the surfaces named: Seg 1, Seg 2, Seg 3, Seg 4, Ax, Ax1, Ay and Ay1. If the reading
cross is placed over an architectural shape, intersection curves and points can be found.
The intersection curves are named Crv X, Crv X1, Crv y, Crv Y1, Crv 1, Crv 2, Crv
3 and Crv 4. At the same time the reading cross splits the architectural geometry in
four different surfaces: Srf 1, Srf 2, Srf 3 and Srf4. All the contour points, lines and
surfaces found are named according figure 17 and stored as output. Note that also the
trajectory of the curves are stored. More information about this can be found in the
grey background box below. Rotating the reading cross around the Z-axis will lead to
different boundary curves, surfaces and points. Why this can be useful will become
clear in description of the next modules.

Curve trajectories
Since the curves inside the program Rhinoceros are modelled by use of the parametric
description all the curves have a starting point and an end. More information can
be found in the literature study section 5.3. The trajectory of the curve is important
information and is used within other modules in the chain. The positive trajectory
of the curves are indicated by the black arrows in figure and follow the right handed
coordinate system with an anti-clockwise positive rotation direction

48
5.2 module 1: reading architectural geometries

Figure 17.: Reading cross

Figure 18.: Positive directions

49
ae-model elements

5.2.5 Output

As described previously the output contains all the information found about the archi-
tectural geometry which is put in module 1. All the boundary surfaces, curves and
points will be stored. The layout of the output in the script can be found in figure 19..
The output is organised in such a manner that it represents the places of the curves
and surfaces. In this way it is easy to recall each individual curve when necessary.

Figure 19.: Organisation of output

50
5.3 module 2a: panelization method; approaching

5.3 module 2a: panelization method; approaching

Figure 20.: Summary module 2A

51
ae-model elements

5.3.1 Function

This module is able to panellize the architectural surface which is given as input to
the AE-model. The architectural surface can be approximated by use of this module
with quad and/or tri shaped surfaces. These surfaces represent glass and are here-
after called panes. The module can be used for surface in categories 1,2,3,4 and 7 as
indicated in the literature study section 6.4 ’ Categorisation’.

5.3.2 Description

This module is the result of a search for a panellization method which can panellize
various kind of surfaces within a short amount of time. Although there are already
some panellization tools on the market most of them focus on a specific type of shape,
and don’t allow the loading of several shapes. The second problem is that most meth-
ods focusses on gridshells for which the division of the surface or the projection of
nodes onto the surface, is sufficient to find a working beam layout.
In the case of this thesis the glass is the leading factor. This leads to different
requirements which need to be fulfilled: Is it possible to find a method which can
slightly deviate from the architectural surface in order to keep the panels flat? How
can we approximate the architectural shape with the least amount of variation in glass
panes? A major step forward towards the solution came in the form of the MSc thesis
of M.H. Toussaint [M.H.Toussaint, 2007]. Although his thesis covers a different subject
(a gridshell tool) it was an inspiration for this module and leaded to new ideas.

Figure 21.: Left: Architectural surface, Right: Panellized approximation

52
5.3 module 2a: panelization method; approaching

Figure 22.: Left: Architectural surface, Right: Panellized approximation

53
ae-model elements

This module, module 2A, uses the architectural shape as an under layer. This under-
layer, can be seen as a mould on which the glass surface is draped. This draping of
glass is done by use of an algorithm which is described in section 5.3.4. Because the
mould and the algorithm are kept separate the algorithm becomes independent from
the architectural surface. This way of scripting increases the modules field of applica-
tion and makes the module suitable for surface categories 1, 2, 3, 4 and 7 as described
in the literature study section 6.3 ’ Surface categorisation’. A second important aspect
of this module is that it can ’propose’ node location which are slightly above or under
the architectural surface. This allows an unique way of approximation and helps to
keep the panels flat. These surfaces can have no, single or double curvature.
Because the red ’reading cross’ (shown in red in figure 24) guides the algorithm. The
rotating of the reading cross will lead to a different orientation of the glass surfaces.
In this way the orientation of the glass panes can be chosen and adapted to the wishes
of architect and engineer.

Figure 23.: Guiding panellization by rotating the reading cross

5.3.3 Input

The input of this module is the output of Module 1. All the point, curves and surfaces
found in module 1 are used within this module to approach a architectural geometry
with flat (glass)surfaces.

5.3.4 Method

Generally speaking, the whole procedure of approximating architectural surfaces within


this module can be split in two parts. The first part lays down the nodes on (or close
too) the architectural surface by use of an algorithm. The second part connects the
nodes to form quads or tri- shaped glass panes. Both parts will be explained below.

54
5.3 module 2a: panelization method; approaching

Part 1
In order to find nodes on, or close too, the architectural surface this module uses an
algorithm between the intersection curves x & y, x1 & y1, x1 & y and y1 & x in com-
bination with one of the 4 parts of the architectural surface (srf1, srf2, srf3, srf4). To
cover the whole architectural surface the algorithm needs to run four times; for each
quarter once. Two examples of the algorithm running in one quarter along the bound-

Figure 24.: Panelization of double and single curvature surfaces

ary curves x & y are shown in figure 24. The algorithm itself is explained in figure
25 and 26 on the next pages. Figure 25 explains step by step the algorithm by use of
handsketches in 2D. The steps shown in figure 26 are identical, only now the algorithm
is shown in 3D. These 3D pictures are screenshots from the program Rhinoceros.

55
Figure 25.: Step by step explanation of algorithm [2D]

56
Figure 26.: Step by step explanation of algorithm [3D]

57
ae-model elements

Part 2
When part 1 is finished there are basically nodes only. At this point information is
known about the amount of nodes, their locations and numbering. The choice for
quad shaped panes or tri-shaped panes can still be made. This depends on the way
the nodes are connected. Since the numbering of the nodes will be identical regardless
the architectural shape, the connecting procedure can easily be scripted uniformly.
Not all surfaces are suitable to be approximated with quad surfaces due to geometrical
contraints. More information about this topic can be found in the grey background
information box below. In figure 27 ’Approximation overview’ an overview is given

Figure 27.: Approximation overview

of the nodes which are found and there location in comparison with the architectural
surface. The nodes which are indicated with a green dot are exactly on the architec-
tural surface. The nodes indicated with a red dot can be slightly under or above the
surface.

58
5.3 module 2a: panelization method; approaching

Geometrical constraints
Whether architectural surfaces are suitable to be discretized into quad shaped elements
depends greatly on the amount of curvature in the architectural surface. Curvature
is the amount of curve in a surface, or how much an objects geometry deviates from
being flat. Examples of some curved geometries can be found in figure 28. The geome-
tries are coloured corresponding the amount of curvature they have. A panel in the
green zone (zone A) has only to cope with one curvature direction, just like the green
arch shaped surface shown on the right side of the figure. Zone B has two curvature
directions. For this reason zone A can be approximated with a flat quad by the above
explained algorithm but a surface in the red zone B cannot. This because the quad
surface simply cannot follow the curved surface without bending the pane.
The yellow zone in between is the transition zone. This zone indicates that there is
no distinct border between the red and green zone. The fact that this yellow zone is
a blurred line indicates the underlying question: When do we call a surface single
curved (green) or double curved (red)? The same problem occurs when looking at the
flatness of the quads: when do we call a surface flat? These transition zones between
flat or curved panels and single-curved and double-curved surfaces also effects the al-
gorithm. In contradiction with this, the software programs used define this boundaries
exact, creating a fixed border between flat an curved. Sometimes these boundaries can
be manually adjusted.
The algorithm above starts with the assumption that the architectural geometry has
single curvature and can be approximated with quads. Whenever the algorithm is at
the green zone no problem occurs. When the red zone is reached the algorithm will
give an message telling that the quad pane is ’warped out of plane’. This means the
panel cannot be made flat and crossed the ’software’s defined border of being flat’.
Solving this issue can be done by locally splitting the quad surface into Tri-elements.
This transforming of quads to tri shaped elements can be done within Rhinoceros as
well in Oasys GSA. The second option involves the adjustment of the border definition:
In other words reconsidering the definition of ’What is flat?’. More information about
curvature can be found in the literature study section 6.4

Figure 28.: Geometries with curvature

59
ae-model elements

5.3.5 Output

The output of this panellization method are the quad or tri shaped surfaces. These
surfaces have special properties:

• All quad pane edges have equal length (L1)

• The panes indicated in black in figure 24 are always flat.

• Whether the white indicated panes are flat depends on the architectural shape
(double/single curvature)

5.3.6 Possibilities for further development

The following list indicates several ideas an possibilities to improve the above ex-
plained module.

• Find smart ways to cut an trim the glass panes surrounding the ground face.

• Make the side pane length L1 variable per algorithmic step.

• Play with the definition of ’What is flat?’ as described in the grey background
box. New glass production techniques (as described in the Literature study sec-
tion 2.4 ’ Modifying techniques’) allow the bending of glass panes to a certain
extend. If we can determine, quantify and control the amount of curvature in
the quads (so play with the ’tolerance level of flatness’) it should be possible to
approximate (strongly) double curved shapes with quads only by using panels
which are slightly (cold) bended or twisted.

• Investigate the rotation of the reading cross in combination with the layout of the
panels. Is there a way to combine the information about the curvature level and
the pane layout?. The edges of the panes can easily be turned into beam sections.
These beam sections follow the edge of the pane. Changing the orientation of
the pane will change the layout of the underlying beams. Finding the best beam
layout/pane orientation when optimizing stress levels can be interesting.

• What if the reading cross has more cutting planes? In that case more (local) infor-
mation will be known about local curvatures. Can the algorithm be improved?

60
5.4 module 2b: panelization method: meridian

5.4 module 2b: panelization method: meridian

Figure 29.: Summary module 2B

61
ae-model elements

5.4.1 Function

The function of module 2B is to panellize surfaces, in quad or tri elements, by use of


meridians. This will lead to another pane layout in comparison with module 2A.

5.4.2 Description

This module was made to panellize an architectural common shape: rotational surfaces
like spheres and cylinders. For this category of surfaces it can be aesthetically attractive
to have horizontal lines of glass pane edges. This demands for a different way of
panellization in comparison with module 2A. The panellization method in module
2B is driven by meridians. The meridians divide the rotational surface in horizontal
segments, hereafter referred to as ’Rings’. Each ring is build up out of an x amount
of trapezium shaped quad elements. For each ring this trapezium quad pane is scaled
and therefore unique. The meridians guide the panellization method as shown in
section 5.4.4.

5.4.3 Input

The input of this module is the information about boundary points, curves and sur-
faces found in module 1.

5.4.4 Method

The working of Module 2B can be split up in the same way as done with module 2A.
Part 1 will focus on the finding of the node locations on the architectural surface. This
time all the node locations are exactly on the architectural surface. Part 2 includes the
forming of surfaces between the nodes found. Both parts are explained below.

Part 1
The algorithm explained in this part needs to run four times to cover the total archi-
tectural surface. One time for each quarter. The algorithm for one quarter (segment 1)
is explained hereafter. The locating of the nodes on the architectural surface, surface 1
in this case, is guided by crv x and crv y. The way the nodes are found is explained in
figure 30. This is a step by step explanation which is guided by a series of pictures.

62
5.4 module 2b: panelization method: meridian

Figure 30.: Step by step explanation of Module 2B

63
ae-model elements

Part 2
When the nodes are found the connecting of the nodes to form the glass panes is the
hard part. To find a script that can form quads and tri elements for any number of ring
division can be difficult. This because of the fact that all the data lists in the module
are depending on the horizontal and vertical ring count and therefore the amount of
nodes vary. In order to do so the script needs to be written towards these variables. A
brief explanation of the counting problems this brings along is illustrated in the grey
background box and the corresponding figure 32.

Counting and Culling


The nodes found in part 1 are stored in a list. This list has a varying length, depending
on the amount of horizontal and vertical rings chosen. To create a quad surface its
corner nodes (A, B ,C and D) need to be picked out of the node list. The difficulty
occurs when the amount of horizontal rings is adjusted and therefore the amount of
nodes in the list. This causes the amount and number of the nodes to change. A
uniform way of scripting therefore needs to be independent from the amount of nodes
in the list. This can be achieved by making all the list in the module (for instance the
node list which is plugged into A or B) depending on the horizontal and vertical ring
count.
The way to script this is to start with the complete list of nodes. After that remove
the nodes which are not part of the surface you want to make. The removing of items
from a list is called ’Culling’. For example, if you were to cull a collection of marbles
such that only blue marbles are chosen, the cull would be the set of marbles that are
not blue. In this module the culling pattern is again depending on the horizontal and
vertical ring count. Clarifying will be figure 32.

5.4.5 Output

The output of this module is a list of quad and/or tri shaped surfaces. In the following
modules these surfaces will be prepared for calculations.

5.4.6 Possibilities for further development

• Make the height and the distance between the meridians variable

64
5.4 module 2b: panelization method: meridian

Figure 31.: Example of output

65
ae-model elements

Figure 32.: Example of counting and culling

66
5.5 module 3a: fe-quad meshing: line support

5.5 module 3a: fe-quad meshing: line support

Figure 33.: Summary module 3A

67
ae-model elements

5.5.1 Function

This module prepares the quad surfaces for calculation purposes. It defines two finite
element zones for each surface which makes it possible to model line like connections
with various characteristics.

5.5.2 Description

The AE-model is used in the preliminary design process. In the design phase the lay-
out of the glass panes and the connection type might not be fixed yet. The choice for
for instance a point like support or a line like support can still be made. This choice
is bound by architectural and structural constraints. Architectural constraints will be
most of the time leaded by aesthetics, which is important but hard to measure. On the
other hand, structural constraints are more bound by numbers (max stress level, max
size, max thickness’s) and therefore a more pragmatic approach can be useful.

This module proposes a method for the modelling of the glass in combination with
various kinds of line like connections at the edges of the pane. This will help to find
the structural constraints of the connection type in combination with the total glass
geometry. Is this connection type suitable for the geometry? As with all (structural)
models there is a playing field between ’what is modelled?’ and ’what is reality?’ To
model something as precise as reality requires a lot of time. Unfortunate time is one
of the biggest constrains in the building industry.

For the reasons stated above, the drive behind this module was to find a method
which is fast and, more important, one which can approximate the real connections
precise enough to investigate its structural suitability. When used properly the struc-
tural engineer can advise by use of this module for a specific connection type or can
divide the group of connection types into: ’not possible’, and ’possible’.

68
5.5 module 3a: fe-quad meshing: line support

Figure 34.: Input (left) and output (right) of module 3A

In figure 34 you can see the input and output of the module. A Quad surface (of
any shape) comes in and runs through the module. Finally a glass pane split in two
finite element zones comes out. Zone 1 represents the glass. Zone 2 is an edge zone.
This zone (zone 2) can be adjusted to model several connection types. The material
properties for each zone will be addressed in module 4A. What can be varied within
this module is the amount of finite elements of both zones and the thickness of zone
2, see figure 38 and figure 39. Above that the layout and number of the finite elements
in zone 2 can be adjusted. This variables gives the modeller the possibility to model
various kinds of connections between the glass panes. Some examples of connections
are shown in figure 35 and figure 36. The yellow color indicates a soft connection type :
like rubber. The red indicates a more stiff connection, for instance steel. When looking
at figure 35 there can be seen that connection 1 shows the simplest connection. Two
glass plates glued together. Connection 2 is the same as connection 1 only now the
rubber is supported by a profile. Because zone 2 is split up in two identical finite ele-
ment parts the structural profile can be connected exactly in the middle of the rubber.
The placing of a structural profile at an edge is simple and van be considered when the
glass geometry itself is not capable of stabilizing and carrying the load. Connection
3,4 and 5 show variations of clamping connections.

The critical part of this way of modelling is to give zone 1 and zone 2 the right material
properties in order to give realistic calculation results. A test can be done to find the
connection’s so called ’equivalent material property’. When the connection’s equiva-
lent material property is found the connection will show the same behaviour in the
model compared with the actual connection in reality. This is elaborated in subsection
5.5.5 ’Testing and Calibration’.

69
ae-model elements

70
Figure 35.: Connection variation (1 of 2)
5.5 module 3a: fe-quad meshing: line support

Figure 36.: Connection variation (2 of 2)

5.5.3 Input

The input of this module is a list of quad surfaces.

5.5.4 Method

At the start of this module the surfaces which represent glass panes are already de-
fined. Each surface is described with a U,V system and a normal direction indicating

71
ae-model elements

the top an bottom face of the pane. Regardless the shape of the quad a quad surfaces
will have: four corner points, four edges, a domain made out of u and v coordinates
and a top and bottom face. This information is shape independent and can therefore
be used to modify the pane into the different finite element zones. The whole proce-
dure is systematized in figure 37. More information about the building of the script
itself will be shown in the next module 3B. This module has the same function as this
module only now for tri-shaped panes.

Figure 37.: Rough steps in scripting

72
5.5 module 3a: fe-quad meshing: line support

Figure 38.: Densifying FE mesh

Figure 39.: Adjusting width of zone 2

5.5.5 Testing and Calibration

The physical and structural behaviour of the modelled connection needs to correspond
as much as possible with the real structural behaviour of the connection. It will require
a lot of time to model the connection in detail. A high level of detail does not fit the
phase in which the model is used and for this reasonable simplifications are made
to speed up the calculation. Nonetheless, the modelled connection need to behave
as similar as possible to the connection in reality. To obtain this identical behaviour
calibration of the modelled connection can be necessary.

The combination of a simplified model of the connection together with the equiva-
lent material properties can be calibrated in such a way that it represent the actual
connection. This calibration can be done by the use of two tests. The first test is in
real life and includes real glass and real connection materials. The second test has the
same test setup but is virtual. After testing both setups, the outcomes (deflection e.g)
can be compared. See figure 40 for the test setup.

73
ae-model elements

Figure 40.: Constraints and two property zones

When the virtual test outcome differs from the outcome of the real setup, the mate-
rial property of the virtual edge zone (zone 2) needs to be adjusted. This can be done
until the similarities of both results are within an acceptable level. At this moment the
”equivalent material property’ of the modelled connection will be found. Once known,
the setting can be used for every geometry.

This thesis only includes the theoretical description of this test. To ensure the sound-
ness of this approach all the influence factors on the end result need to be investigated.
However, the focus of this master thesis is laid , because of time constraints, on the
model itself and therefore the testing of the connection in real life is left out of the
scope of this thesis. But to illustrate the basic idea an example is given of the virtual
test:

The tests are necessary to find the equivalent material properties of the virtual con-
nection. These property values include the Youngs modulus, poissons ratio, density,
temperature modulus, shear modulus, yield stress value and the ultimate stress value.
To find the correct settings for these values the virtual connection need to be exposed
to shear loads, bending moments, tension and pressure loads. Examples of load cases
of this virtual test can be found in figure 41.Although the actual calibrating is left out
of the scoop of this thesis some results are shown in figure 42. These indicate that the
effects of zone 2 are clearly visible. And that the equivalent material properties can be
set to model the actual connection

74
5.5 module 3a: fe-quad meshing: line support

Figure 41.: Exposing zone 2 to various loadcases

75
ae-model elements

Figure 42.: Result test setup

76
5.5 module 3a: fe-quad meshing: line support

5.5.6 Output

The output of this module is a list of finite elements. More about the choice of FE
within Oasys GSA can be read in the grey background box below.

Choosing the correct FE type


Within GSA there are four types of finite elements. First order elements include the
’Quad 4’ and ’Tri 3’ elements. Second order (or parabolic) elements are Quad 8 and
Tri 6 elements. For each element different settings such as interpolation methods are
possible. Which finite elements are most suitable for the engineering task depends on
the type of structural calculation and the characteristic of the structure.

The AE-model imports (via a DXF file) automatically the finite elements into Oasys
GSA. These form Quad 4 and Tri 3 elements at start. The elements can be switched to
Quad 8 and Tri 6 within GSA by two mouse clicks.

Quad 4 and Tri 3


The quad 4 and tri 3 first order elements give a linear approximation of displacements
over their domain and are therefore less accurate then quad 8 and tri 6 elements, see
top figure 43. Nonetheless, since they have less approximation nodes, are faster in
calculation. Care should be taken when these first order elements are used. These
elements can show inaccurate stress, strain and deflection solutions due to problems
such as shear locking, volume locking and hourglass locking. The latter, hourglass
locking, is solved since Oasys GSA implemented the MITC formulation [Bath and
Iosilevich, 1999]. The MITC formulation surpasses the trapezoid failure modes which
causes the hourglass locking see figure 44

By use of the MITC formulation for the Tri 3 and quad 4 elements the out-of-
plane performance of linear elements improves. In some cases the tri 3 and quad 4
can therefore be a practical alternative to higher order quadratic elements. [Lee and
Bathe, 2005].

Quad 8 and Tri 6


The quad 8 (and tri 6) elements provide a quadratic approximation of displacement
over their element domain and therefore provide the best accuracy when compared
with the linear approximation, see bottom figure 43. But because more nodes
are necessary for the calculation using this element can significantly influence the
calculation speed and file size.

77
ae-model elements

Figure 43.: Top: Linear FE, Bottom: Quadratic FE

Figure 44.: Hourglass failure mode

5.5.7 Possibilities for further development

• Elaborate more on the test setup and testing

• Find a similar method for point connections focussing on densifying the finite
element grid locally

• Find ways to implement sub-modelling techniques: When the first calculation


run indicates specific critical areas, these areas need to be further analysed. This
can be done by ’taking out’ the critical area and executing a more detailed anal-
ysis,

78
5.6 module 3b: fe-tri meshing: line support

5.6 module 3b: fe-tri meshing: line support

Figure 45.: Summary module 3B

79
ae-model elements

5.6.1 Function

This module prepares Tri surfaces for calculation purposes. It defines two finite ele-
ment zones for each surface which makes it possible to model line like connections
with various characteristics.

5.6.2 Description

In comparison with the last module the difference between both is the shape which
comes in as an input. In the case of this module, module 3B, these are Tri- shaped
elements (surfaces made by use of 3 corner nodes) instead of Quad shaped elements.
The input and output of the module are shown in figure 46. This figure indicates again
two finite element zones called zone 1 and zone 2.

Figure 46.: Input (left) and output (right) of module 3B

The function and output is the same as the previous module and the description of
this module is therefore kept brief. The focus of this description is laid on the expla-
nation of the scripting. Not only of this specific module but also more in general.

80
5.6 module 3b: fe-tri meshing: line support

Scripting
A computer script is nothing more than a list of commands that are executed by a
software program. A script can be compared with a cooking recipe. Following the
recipe step by step leads to a perfect diner. The ingredients are the input, the diner
the output and the recipe is the script.

In contradiction with the steps in a recipe the steps in a computer script are all
digital. It’s a very functional way to automate a series of steps. These steps can
replace actions which are normally done by hand. The benefit of (calculation) scripts
are that you can use them over and over again. This in contradiction with calculations
which are written on paper.

It is generally thought that structures and their calculations are unique and one
of a kind. This is true regarding the final structure itself but the way to get there is not.
Within the process of translating an architectural model to a structural model there
are a lot of steps which are similar for every project. Every structural model requires
for instance the determining of materials, load cases, nodes, beams and surfaces.

Scripting these repetitive parts within the procedure of translating an architec-


tural model to a calculation model makes therefore sense. There is only one constrain.
The script needs to be written with one variable in mind: the geometry of the
structure. This can be achieved by treating the geometry as the input to the general
steps. Or in other words: the geometry is the input of the script.

The difficulty in writing this script is the uniform purpose of the script. With
’uniform purpose’ is meant: ’ making it suitable for all geometries’. Creating a script
which can full fill this requirement can be difficult. What happens when the layout of
the glass surfaces change? How do all the modules connect? Next to the problem of
having a lack of scripting knowledge at the start of this thesis, this leaded to problems
which can roughly be divided into two categories:

1. Problems concerning the uniform scripting of the modules.


2. Connecting the modules.

An example of a problem out of category 2 is illustrated in the sketch shown in figure


47. This example is a good illustration of similar kinds of ’attaching’ problems which
occurred during the process of scripting.

81
ae-model elements

Figure 47.: Sketch of script connecting problem

5.6.3 Input

The input of this module is a list of tri shaped surfaces. These are surfaces which have
3 corner nodes.

5.6.4 Method

At the start of this module the surfaces which represent glass panes are already de-
fined. So the area, length of the edges and the top an bottom face of the pane are
known. This module generates a workable finite element lay-out for each surface.

In comparison with the previous module the scripting is more difficult. This because
the counting is more complex with triangular split surfaces. The approach is nonethe-
less similar. First finding nodes and afterwards connecting them to form surfaces.

The forming of the surfaces starts with the complete list of nodes. After that the nodes
which are not part of the surface you want to make need to be removed. The removing
of items from a list is called ’Culling’. In this module the culling pattern is depending
on the V-count. The V-count determines the amount of finite elements present in the

82
5.6 module 3b: fe-tri meshing: line support

glass face. A V-count of 4 corresponds with the edges of zone 1 being split in 4 equal
parts. In total this gives 16 finite element triangles within zone 1. A description of the
script concerning the forming of the finite elements in zone 1 is shown in figure 48
and 49. What can be seen in the figure is that the amount of nodes in each list as well
as the culling function (Fcull) is only depending on the variable ’v-count’.

5.6.5 Output

The output of this module is a list of finite elements.

5.6.6 Possibilities for further development

• Elaborate more on the test setup and testing.

• Find a similar method for point connections focussing on densifying the finite
element grid locally

83
ae-model elements

Figure 48.: Script explanation part 1

84
5.6 module 3b: fe-tri meshing: line support

Figure 49.: Script explanation part 2

85
5.7 module 3c: windload on spheres

5.7 module 3c: windload on spheres

Figure 50.: Summary module 3C

87
ae-model elements

5.7.1 Function

Determining of the windload on sphere like shapes.

5.7.2 Description

Wind is a complex phenomena. Loads caused by wind are hard to predict and vary
in altitude over the geometries surface. Because of its complexity it is never sure how
the structure is affected by the wind before the structure is actually build. To cope
with all these uncertainties wind loads are overestimated to guarantee safety. In most
cases, due to time constraints, the load is even more overestimated to simplify the
calculation.
The overestimation of windloads is unfavourable while it can lead to inefficient struc-
tures. In contradiction with gravity the windload is an asymmetric load case. This
can lead to high bending moments and torsion. For this reason overestimating the
wind load will have a big impact on the material usage: the structure will need more
material in order to keep the deflections within limits.

Especially in glass structures where the desire is most of the time to minimize the
presence of the structure, optimizing the windload calculation can give substantial re-
wards.

Generally speaking, Wind loads are calculated according NEN-EN 1991-1-4+A1+C2


and are prescribed for a limited variation of geometries. If the architectural geometry
follows exactly one of the geometries described in the standards the wind loads can
be determined exact. In all the other cases the wind load will be approximated by
interpolation from one of the standardized geometries. The standardized geometries
are shown in figure 51 in box 2: Cpe. This module focusses on the last geometry in the
figure: the sphere. But as said before it is allowed to slightly deviate from the shape
and therefore the module can also be used to calculate more geometries. For instance
ellipses.

In the eurocode windloads are determined by use of the following formula:

Windload = CsCd ∗ (Cpe + Cpi ) ∗ qp ∗ Are f (1)

CsCd Buildingfactor, takes into account dynamic effects of the structure.


This factor depends greatly on the height of the building and its
eigenfrequency
Cpe Shapefactors, this is a factor which accounts the maximum values
of suction and pressure. Factor depends on the shape of the structure.
’How does the wind ’grabs’ the surface?

88
5.7 module 3c: windload on spheres

Cpi0 Overpressure, factor which represents the maximum overpressure value.


Depends on the amounts of holes in the surface.
Cpiu Underpressure, factor which represents the maximum underpressure
value. Depends on the amounts of holes in the surface.
Aref Surface area where the wind blows against
qp Windpressure, at a specific reference height. How hard is the
wind blowing?

The quantities of the factors in formula 1 vary per structure. Wind loads can be either
suction or pressure. The exact location of the suction and pressure zones are difficult
to predict and depend greatly on the shape of the structural geometry. The factor Cpe
determines the zones on the architectural surface in which pressure and or suction
occurs.
The Cpe factor can be either a Cpe1 factor or a Cpe10 factor. Cpe1 is used when
the windload formula is used for local strength calculation. Like local glass surface
deflections. The number 1 in the subscript comes from 1 m2 and refers to an influence
area of the same size. The Cpe10 factor is used for the overall structural stability
calculation and takes into account the combination of suction and pressure on the
total construction. The 10 comes from 10 m2 . For spheres there are no Cpe1 factors
defined in the Eurocode.Cpe10 factors are described. For the purpose of this module
this is fine, because the influence of the wind on the total structure is to be investigated.
On the next two pages the windload calculation is elaborated further. The references in
figure 51 and figure 52 ’refer to the NEN-EN 1991-1-4+A1+C2. Figure 1 explains and
figure 52 gives an example of the Cpe values concerning a sphere and a ellips. This
gives more understanding in general and a feeling for the way wind grips spheres and
ellipses..

89
ae-model elements

Figure 51.: Explanation of windforce formula

90
5.7 module 3c: windload on spheres

Figure 52.: Example of Cpe calculation

91
ae-model elements

5.7.3 Input

The input of this module is the information of boundary surfaces, curves and points
gained in module 1. Also the geometry, size and location of the quad and tri shaped
glass panels are used to calculate the wind loads.

5.7.4 Output

The output is the windload on each finite element surface in KPa. An example of the
output can be found in figure

Figure 53.: Example of output, sphere topview

5.7.5 Method

To determine the windload on a structure the factors described in formula 1 need to be


determined. The factors, Cscd, Cpi and Qp are rather easy to determine. This because
for each structure they only need to be determined once. For the factors Cpe and Aref
this is not the case. Aref is the area the wind is blowing against. In this case these
are glass plates. All the glass plates are already defined and therefore the area can
easily be recalled. What is left is the factor Cpe. This factor is more complicated to
determine and is independent from all the digital information we have so far.

92
5.7 module 3c: windload on spheres

To find the Cpe value on each location this module uses a so called ’Cpe10 -surface’ This
Cpe10 -surface is placed over the original architectural surface. The normal distance
between a point on the architectural surface and a point on the Cpe10 -surface, is the
Cpe10 value on that exact location. To determine the windload on each specific finite
element this distance need to be found for each individual finite element. This whole
procedure is explained by use of a cartoon in figure 55.

93
Figure 54.: Explanation method, part 1

94
Figure 55.: Explanation method, part 2

95
ae-model elements

5.8 module 4a: materials

Figure 56.: Summary module 4A

96
5.8 module 4a: materials

5.8.1 Description

The upcoming four modules cover the engineering input of the AE-model. All the
points, curves and surfaces need to be changed into connections, beams, and surfaces
taking into account the corresponding material properties and applied loads. The soft-
ware’s equivalents are called structural nodes, structural beams and finite elements.
The applying of load cases and materials are normally done within a calculation pro-
gram itself. The upcoming four module take over these traditional actions.
In this module, module 4A, materials and their corresponding properties can be de-
fined. These defined materials can be addressed to Finite elements and/or structural
beams. An example of the material property ’ Glass’ set for finite elements is shown
in figure 57. After everything is set correctly, the module can send this information to
the calculation program Oasys GSA.

This module, and also module 4B, 4C and 4D, uses the excellent Geometry Gym plugin
written by Jon Mirtschin [Mirtschin, 2014].

Figure 57.: Explanation of module 4A

97
ae-model elements

5.9 module 4b: analysis cases

Figure 58.: Summary module 4B

98
5.9 module 4b: analysis cases

5.9.1 Description

This module gives the modeller the opportunity to chose a calculation solver depend-
ing on the way he wants the calculation to be made. There can for instance be cho-
sen for a linear, non-linear or form finding solver. The different ways of calculation
(solvers) can be run after each other. For this reasons stages are defined. Clarifying
will be figure 59. Within each solver load cases (LC) can be added. Resulting in that
specific load case being calculated according the solver chosen. Load cases(LC) and
load combinations (LCM) are explained in module 4C.

Figure 59.: Explaination of module 4B

99
ae-model elements

5.10 module 4c: load cases and load combinations

Figure 60.: Summary module 4C

100
5.10 module 4c: load cases and load combinations

5.10.1 Description

On the nodes, curves and surfaces which represent connections, beams and plate/shells.
loads can be applied. This can be done by addressing a type of load case to a (set of)
structural components. In this manner point loads and distributed loads can be ap-
plied. To illustrate this procedure the definition of a snowload on a glass surface is
highlighted in figure 61. This load case considers a distributed face load on a finite
element. This finite element is placed under the input indicated with the letter ’ E’ in
figure 61.

Figure 61.: Explaination of module 4C

101
ae-model elements

After the load cases are determined they run through a specific solver, as described
in module 4B. At this point for instance a load case considering snow loads, ”LC1:Snow”,
is placed in a linear solver and defined as analysis case 1 ”A1:Snow load”. The dead
loads are defined on a similar way and named ”A2:Dead load”.
To combine analysis cases and to add safety factors to them load combination (LCM)
can be made. For instance Load combination 1 can be for the Ultimate Limit State:
LCM 1: ULS[Dead + Snow]. See figure 62 for an explanation.

Figure 62.: Explaination module 4C

102
5.11 module 4d: constraints

5.11 module 4d: constraints

Figure 63.: Summary module 4D

103
ae-model elements

5.11.1 Description

This last module, module 4D, helps to ensure stability. In this module supports can be
defined. Out of a list of all structural nodes the the supports can be picked. When cho-
sen supported need to be fixed in a certain direction. Support group 1 is highlighted
in figure 64. This figure speaks for itself.

Figure 64.: Explain module 4D

104
Part IV

THE AE-METHOD
6
GENERAL DESCRIPTION

6.1 abstract

In the previous part, ’part III: AE-model’, the AE-model is explained. This part, part
IV, explains the AE-method which is required for the proper used of the model. This
part is split in two chapters. The first chapter includes the general description of the
AE-method. To help orientating through this chapter questions are formulated below.
The location where the answer can be found is placed on the right side.

The next chapter will include several examples in which the AE-model en method
are used.

Figure 65.: AE-method

Question concerning the AE-method: Answer:


• Why does the model comes along with a method? Section 6.2
• Which place does it take in the construction process? Section 6.3
• How can it be used? Section 2.3
• What is the starting point of usage? Section 6.4.1
• What is the (desired) end-result? Section 6.4.3
• What are the benefits? Section 6.5
• What is the field of application? Section 6.6
• What are the boundaries? Section ??

107
general description

6.2 introducing the ae-method

The great danger of computer models is that they have the tendency to become black
boxes. Clarifying the do’s and don’ts of the model by use of a guideline can help to
minimize this risk. The guideline which belongs to the AE-model, is introduced in this
chapter, and is called AE-method. The AE-method will describe the human interaction
of the model with the building industry and the models field of application.
Why is the AE-method so important? When a modeller starts working with the AE-
model without being completely familiar with the model’s capabilities and restrictions
the (mis)using of the model can result into failure. The main reason being the fact that
insufficient knowledge of the model will lead to an incorrect input. This on its turn
will result inevitably in an incorrect output. To prevent the creation of a black box it
is useful knowing what causes a black box in general:

Causes of Black Box:


- Underlying model methodology not visual and or not accessible.
- Bad ordering of model, no clear structure.
- No information given about boundaries and restrictions of the model.
- Insufficient knowledge of user.

Possible consequence:
- Using model for a wrong purpose.
- Using model at a wrong time.
- Giving wrong input.

Risk:
- Incorrect and inaccurate output which is perceived as being correct.

While avoiding the causes listed above will do certainly good, a big part of the solu-
tion lies within the overall perception of engineering models at the start of modelling.
In general, an engineering model should not be considered as a model that generates
unique solutions itself but as a tool to sculpt human knowledge. As a logic consequence,
user-friendliness becomes of great importance, together with the transparency of the
model and the readability of the outcome. There is no such thing as a magical box
that thinks for itself nor does it make sense trying to make a model that pretends it
does. Making technical models therefore should not focus only on the digital part but
should acknowledge human interaction as well.
To avoid the misusing of the model the upcoming paragraphs will sum up the non-
technical know-how around the AE-model. The information includes the AE-model’s
purpose, boundaries, input, output and other information concerning the use of the
model.

108
6.3 purpose and place of ae-method in the construction process

6.3 purpose and place of ae-method in the construction process

The main benefits of computers is that it can remember and process knowledge fast. It
can take away work and safe time. A process which takes a lot of time within engineer-
ing is the iteration process. This is a process in which a concept is iterated towards
a safe and efficient design by optimizing structural variables such as cross sections.
During the iteration process engineers gain insight and feel for the important param-
eters of the structure. A logical place to implement a computer model is therefore
within this time consuming iteration process: Speeding up can lead to more iteration
steps within a short amount of time and therefore more feeling for the structure. The
iteration process is highly related with optimization and both are explained more ex-
tensively hereafter.

To ensure safe buildings engineers need to propose structural load bearing systems
which meet with the strength, stability and stiffness requirements described in the
norms. When the norms are not met in the initial design, the engineer has to change di-
mensions, materials and the structural layout until all the structural values are within
its limits. This process of trial and error can take up a lot of time when done by hand.
An engineering computer model can take over part of the work and can automate the
iteration process. This automated iteration can converge, by changing structural vari-
ables such as cross sections or material properties, towards a design which has optimal
stress levels, deflections or spans. Computational optimizing in this sense of the word
takes over the iteration process which would take a very long time when done by hand.

As with all optimization, optimizing something has a certain reference and is influ-
enced by perception. For person 1 the optimal solution can be completely different
than for person 2. Fixing the reference from which the optimal solution is determined
is therefore important. This reference is hereafter referred to as ’ boundaries’. The op-
timal design lies in between these boundaries. Optimization models within the work-
ing field of engineering can work very well because the boundaries between which
the optimal solution is found can easily be determined. These can be a maximum
and minimum stress level or a minimal and maximum amount of deflection. The op-
timal solution perceived from an engineering view will never exceed these boundaries.

The difficulty occurs when architectural boundaries are added to the optimization
process. Architects have the goal to optimize aesthetics which is most of the time not
bound by numbers nor can it be determined exact. Above that it can sometimes vary
from day to day. This can influence the engineering optimization. An architect might
for instance change the amount of columns in a building for aesthetic reasons and
therefore influence the structural system and the structural optimization. Mixing up
architectural boundaries and engineering boundaries is therefore not sensible.

109
general description

The solution, as proposed in this thesis, lies within the splitting of the optimization
process as shown in figure 66. After being split the optimalisation processes need to
be placed in time behind each other. Figure 66 illustrates the explanation which is
given previously by use of a small cartoon.

Figure 66.: Place of AE-method in the construction process

Optimalisation 1, as shown in the figure, corresponds with the AE-method and fo-
cusses on the finding and quantifying of the architectural and engineering boundaries
in the design phase. During the process concepts will be winnowed down, main-
taining the concepts which are architecturally- and engineering-wise in balance. The
purpose of the AE-model and AE-method can therefore be formulated:

By use of the AE-model and method concepts can be put into perspective during the design
phase architecturally and also structurally.

110
6.4 the workflow of the ae-method

After that, whenever the design is finalized, optimalisation 2 can start in the engineer-
ing phase. This will optimize the design even further and can take care of maximum
efficiency within defined and clear boundaries.

6.4 the workflow of the ae-method

The introduction of the AE-method in the design phase influences the traditional work-
flow during a design phase. In this paragraph the traditional design phase is compared
with the proposed ’modern’ design phase. This modern design phase is in fact the
workflow belonging to the AE-method. By use of this workflow the AE-model can be
fully utilized.

On the following pages the classical design phase and modern design phase are com-
pared. Figure 67 shows the classical design phase. Figure 68 will show the modern
design phase: The AE-method. In both figures, the letter ’A’ stands for ’ Architect’ and
the letter ’E’ stands for ’Engineer’. The difference between both classic and modern is
summarized below. In the modern design phase,

- the speed in which engineering calculations can be done increases.


- the amount of people involved in the engineering calculations decreases.
- the speed at which design engineering related questions appear increases.

Nonetheless, it is worthwhile noticing that the intended goal of the traditional and
modern design phase does not change. Still the intended goal is to have a balanced
and efficient design structural and architectural wise in the end of the phase. The
design related questions, as shown in figure 67, will stay therefore the same for both
the modern and classical method. The biggest difference is the speed at which the
questions can be answered: the modern design phase can do this faster.

For the reasons stated above the modern method, figure 68, proposes three distinct
steps. Step 1, step 2 and step 3 will be explained individually in section 6.4.1, 6.4.2
and section 6.4.3 respectively.

111
Figure 67.: Classical design sequence

112
Figure 68.: Proposed design sequence

113
general description

Before step 1 (see figure 68), can take place a design team has to be formed. The
forming of an actual team is essential since the AE-model demands for a close way
of working together. Furthermore the team needs to be aware of the process which is
coming.

6.4.1 Step 1: Think Talk and Define

As a basis for step 1 an architectural sketch can satisfy. In other words, when informa-
tion is known about the basic geometry step 1 of the AE-method can be started. During
step 1, this sketch, or architectural image, can be discussed with one main question in
mind: ’How can we develop the architectural image into the most beautiful, efficient
and above all balanced design?’ This means taking into account the architectural con-
cept but also structural logic. It is therefore important for this step that architect and
engineer work on the same level without one party trying to lead the other. At the
end of step 1, when an agreement is found about the way of approach and the scoop
of the model, the agreements can be formulated and fixed in a fundamentals report.

The fundamentals report report will formulate the expectations, boundaries and cal-
culation principles of the AE-model. The report can be a guideline for the rest of the
process and forms a document which includes all agreements. If at a later moment
in the design process discussions arise this fundamentals report can help solving the
issues.

6.4.2 Step 2: Modelling

This stage includes the modelling of the structure. All the details about the existing
modules which are build during this thesis project are described in part III of this
thesis.

6.4.3 Step 3: Generate concepts and make a choice

After the model is set in step 2, the generation of various concepts can start in step 3.
With the AE-model being made to change as much variables as possible the result is a
model which can generate an unlimited amount of concepts. With all this possibilities
and variables it can be hard to converge towards solutions and an unstoppable flow of
output needs to be prevented. Structuring and ordering of the output can prevent this.
For the reasons stated above the output of the AE-method will be given in the form of
’concept-booklets’. These booklets are useful in various ways.

First of all they will give visual feedback about each concept in a structured manner.

114
6.4 the workflow of the ae-method

This makes it easy to compare the generated concepts and to put them in perspective.
The booklets contain bundled information, preventing a unnecessary and inefficient
flow of lose results being send and forth between architect and engineer. Secondly it
can be used to create a ’milestone’ moment. Each booklet can be a milestone in the
design phase creating a moment of rest to discuss the outcome and to decide how to
go on. Finally, whenever the design process takes a wrong path the booklets can be
used to get back to a concept where things were still all right.

The booklets includes the concept specific input as well as the output. The standard-
ized booklets are scripted in LaTeX allowing a very fast, almost automatic, generation
of the concept booklets. The AE-model is coupled with Excel allowing information
such as building height, glass area, section length and material properties to be read
out of the model. This information can then be used for the booklets. Several exam-
ple booklets can be found in the next chapter where the AE-model and method are
demonstrated with multiple example projects. A brief overview of the output which
can be found in each concept booklet is given below:
Output structural analysis:


Materialisation of design

Dimensioning of structural glass and/or support structure

Strength check

Deflection check

Stability check

× Fatique calculations
× Connection detailling

Output cost estimation:


Material cost, Glass

Material cost, Secondary structure

Engineering costs

× Constructing costs
× Maintenance costs
× Permits

Note: Cost estimations are left out of the scoop of this thesis.

115
general description

6.5 benefits

When executed properly the benefits of the AE-method are numerous and the most
important are summarized below:

• Engineer can propose (economically) attractive improvements for complicated


double curved geometries
• Increased speed of structural analysis allowing dialogue between architect and
engineer
• Structural analysis model can be reused.
• The model can give rapidly insight in the structural behaviour of architectural
designs. Engineer can better point out what the consequences are of any
changes in the design.

6.6 field of application and boundaries

The field of application of this method is broad. As long as it fits in the scoop of
the modules described in part III of this thesis it can be modelled very fast. When
the desired shape or architectural task cannot be made by chaining up the existing
modules as described in part III, it needs to be modelled by hand. This requires time
but only needs to be done once, it can lead to a new module. The amount of time
this takes depends greatly on the skill of the modeller. For this reason at the start of
the design phase decided should be whether the shape is worth parametrizing worth
investing the time of scripting. Generally speaking the following kind of projects are
therefore suitable:

• Complex geometries, which are still rather undefined at start. These can benefit
greatly from the adaptive parametrization which the model has to offer

• Geometries where glass needs to be activated or is used to stabilize the structure.

• Geometries where the load bearing path is influenced by the shape such as shells
and domes.

On the following page shows some projects which fit in one of the above described
categories.
The AE-model and AE-method is solely meant for dimensioning and estimation cal-
culations. With the design of the AE-model a trade-of is made between speed and the
accuracy of the structural calculation. The AE-method can not been seen as a stand-
alone calculation. In order to speed up and simplify the calculations approximations
are made. Depending on the complexity of the geometry these calculation will include
deviation from reality. Uncertainties mainly caused by unknown wind loads (not cov-
ered in the norms), ground characteristic specifications, the lack of fatigue calculations,

116
6.6 field of application and boundaries

buckling calculations and connection detailing. These aspects demand for a secondary
more detailed calculation and can be added to the calculation at a later stage.

117
Figure 69.: Field of application examples

118
7
USING THE AE-METHOD: SOME EXAMPLES

7.1 abstract

This chapter includes five projects in which the AE-model and AE-method is used. The
projects are listed below in the grey box and visualised in the corresponding picture
below the box. The first project, ’The Dome’, will be most extensively explained. After
this, the other projects are explained more briefly since they have exactly the same
approach but only a different composition and number of modules. For each project
concept booklets are made. These can be found separately with this thesis report.

1. ’The Dome’ , Virtual project Section 7.2, pg.120


2. ’The Ellipsoid’, Virtual project Section 7.3, pg.124
3. ’The Globe’, Kuwait, Kuwait Section 7.4, pg.126
4. ’The New Apple Cube’, New York, USA Section 7.5, pg.128
5. ’The Canakkale Antenna Tower’, Canakkale, Turkey Section 7.6, pg.130

119
using the ae-method: some examples

7.2 the dome

Project: The Dome (Virtual project)


Geometry: Hemisphere, scripted within AE-model
Structure Concept 0: Glass + Rubber + Steel
Concept 1: Glass + Rubber
Concept 2: Glass + Rubber
Modules used: Module 1A: ’Reading architectural geometries’
Module 2B: ’Panelization method: meridians’
Module 3A: ’FE-Quad meshing, line support’
Module 3B: ’FE-Tris meshing, line support’
Module 3C: ’Windload on spheres’
Module 3D: ’Snowloads on surfaces’
Module 4A: ’Materials’
Module 4B: ’Analysis cases’
Module 4C: ’Load cases and combinations’
Module 4D: ’Constraints’

The first geometry in this series of test projects is a Hemisphere, or in other words
half a sphere. This shape was chosen because of various reasons. First of all it is
a common shape in architecture and the shape is already build in glass a couple of
times (see the literature study section 4.4). This allows the outcomes of the model
to be compared with their real-life duplicates which can help to validate the model.
Secondly the shape is the most difficult (double curved) shape for which the wind
loads are described in the Eurocode. The proper description of the wind loads by use
of module 3C (asymmetric loadcase) can help to optimize the dimensions of the glass
and structure since wind load is a governing and unfavourable asymmetric load case.
This project is described hereafter according to the three steps in the proposed design
phase as shown in fig 68 on page 113.

Preface
Before starting with the AE-method it is important to make sure that the AE-team is
aware of the process they are going to be part of and have a general understanding of
the process. This seems obvious but is crucial since the design process will be concise
and to-the-point.

Step 1: Think, Talk & Define


During step 1 the boundaries of architect and engineer need to be discovered, clarified
and defined as good as possible. As a start a basic architectural sketch can satisfy as

120
7.2 the dome

shown in figure 70. What is the architectural image the architect is seeking for? Try to
find out what each other desires, wishes and constraints are. Think for instance about
desired dimensions, load bearing systems, penalization, shape variables and materials.
This makes clear what is expected from the model and indicates how the AE-model
should be used. In other words: What should be the variable parameters? At the end
of step 1 the decisions and ’rules’ which are agreed upon can be formulated into the
’Fundamentals booklet’ . This booklet should describe what is expected from the AE-
model. How accurate will it be? What will be included in the calculation? What are
the assumptions and simplifications made? An example of this fundamentals report
for this project ’The Dome’ can be found separately attached to this report.

Step 2: The AE-model


Within this step the right modules should be chosen and chained. The script can be
found in the bottom of figure 71. In this script nine modules are coupled. These mod-
ules are described in part III of this thesis. At the top of figure 71 a brief overview is
given. In appendix B the script can be found on A3 size.

Figure 70.: Starting sketch

121
Figure 71.: Overview of ’The dome’

122
7.2 the dome

Step 3: Generate concepts and make a choice


Once the AE-model is set the variables in the model can be altered to get an unlimited
amount of concepts. To organize and structure this flow of concepts, concept booklets
can be made. For this dome project three concepts are described in concept booklets
see fig 72. Concept 0 includes a half dome consisting of 48 panes and has a secondary
steel frame. Concept 1 consist of 96 panes and has no secondary frame. Concept
2 includes again 96 panes but with free corner nodes, The booklets help to channel
down the concepts and chose which path to follow.

Figure 72.: Output AE-method

123
using the ae-method: some examples

7.3 the ellipsoid

Project: The Ellipsoid (Virtual project)


Geometry: Half an ellipsoid, scripted within AE-model
Structure Concept 0: Glass + Rubber
Modules used: Module 1A: ’Reading architectural geometries’
Module 2B: ’Panelization method: meridians’
Module 3A: ’FE-Quad meshing, line support’
Module 3B: ’FE-Tris meshing, line support’
Module 3C: ’Windload on spheres’
Module 3D: ’Snowloads on surfaces’
Module 4A: ’Materials’
Module 4B: ’Analysis cases’
Module 4C: ’Load cases and combinations’
Module 4D: ’Constraints’

This project is described hereafter according to the three steps in the proposed design
phase as shown in fig 68 on page 113.

Step 1: Think, Talk & Define


This geometry is chosen for its irregular floor plan. This is mainly done to test the
outcomes of the windloads on an irregular shape. The Eurocode allows a slight devia-
tion from the prescribed norms. Therefore it is allowed to calculate the windloads on
half a ellipsoid by using the norm for hemispheres. For the ellipsoid described in this
project the wind is placed on the most governing direction, the ’x’ direction.

Step 2: The AE-model


Within the AE-model nine modules are chained.The modules are summarized in the
grey box on the top of this page.

Step 3: Generate concepts and make a choice


The output and results are again described in a separate concept booklet. The dimen-
sions of the geometry is shown in figure 73. Although only one concept (booklet) is
made, as with all the other projects,variables such as pane number, height, width etc
can be adjusted to generate other concepts in a short amount of time.

124
7.3 the ellipsoid

Figure 73.: Top: Geometry, Middle:Snow Load, Bottom:Wind Load

125
using the ae-method: some examples

7.4 the globe

Project: The Globe


Location: Kuwait, Kuwait
Function: Art, World clock
Geometry: Buckminster-fuller dome, geometry externally made by architect
Structure Concept 0: Steel + Glass
Client: BDM (Boondock)
Architect: CUBE Architecten
Engineer: I-Saac & BRS
Completion: 2014
Modules used: Module 3C: ’Windload on spheres’
Module 3D: ’Snowloads on surfaces’
Module 4A: ’Materials’
Module 4B: ’Analysis cases’
Module 4C: ’Load cases and combinations’
Module 4D: ’Constraints’

This project is described hereafter according to the three steps in the proposed design
phase as shown in fig 68 on page 113.

Step 1: Think, Talk & Define


For this project the engineering task for the AE-model was to put the wind loads on
the nodes of the steelframe of this buckminster-fuller dome. The geometry was made
by the architect and could be loaded into the AE-model as described in section 4.3 on
page 38. The windload is placed on the glass panes which are supported on the steel
frame. Note that the each glass pane consist of only one finite element because they
only function in this calculation for selfweight and the spreading of the windload onto
the steel frame.

Step 2: The AE-model


Within the AE-model five modules are chained. The modules are summarized in the
grey box on the top of this page.

Step 3: Generate concepts and make a choice


To explain the calculation done one concept booklet is made.

126
7.4 the globe

Figure 74.: Overview project ’The Globe’

127
using the ae-method: some examples

7.5 the new apple cube

Project: The New Apple Cube


Location: New York, USA
Function: Entrance of Apple store
Geometry: Cube, scripted within AE-model
Structure Concept 0: Glass + Rubber
Client: Apple Computer Inc.
Architect: Bohlin Cywinski Jackson (BCJ)
Engineer: Eckersley O’Callaghan
Completion: 2011
Modules used: Module 3A: ’FE-Quad meshing, line support’
Module 4A: ’Materials’
Module 4B: ’Analysis cases’
Module 4C: ’Load cases and combinations’
Module 4D: ’Constraints’

This project is described hereafter according to the three steps in the proposed design
phase as shown in fig 68 on page 113.

Step 1: Think, Talk & Define


To test the field of application and to check the outcome of the AE-model the gradua-
tion committee requested for a more regular geometry to be tested. The Apple cube
is a regular shape without any curvature. The geometry consist of fifteen cover panes
and four glass portal fins. The cover panes are shown in blue in figure 75 and the
fins in red. Between the fins and the cover panes and between each cover pane itself a
rubber sealing is modelled (purple colour) by use of module 3A.

Step 2: The AE-model


Within the AE-model five modules are chained.The modules are summarized in the
grey box on the top of this page. Due to the variables implemented in the scripting of
the geometry the cube can vary in height/width. This also counts for the width of the
fins.

Step 3: Generate concepts and make a choice


To explain the calculations done one concept booklet is made. The concept is made
from glass and rubber only.

128
7.5 the new apple cube

Figure 75.: Top: New Apple Cube, Bottom; FE Geometry

129
using the ae-method: some examples

7.6 the canakkale antenna tower

Project: The Canakkale Antenna Tower (Non-glass structure)


Location: Canakkale, Turkey
Function: Antenna + Public watch-tower
Geometry: Free-from, scipted within AE-model
Structure Concept 0: Steel structure
Client: Municipality of Canakkale
Architect: IND + Powerhouse company
Engineer: ABT
Completion: 2015
Modules used: Module 4A: ’Materials’
Module 4B: ’Analysis cases’
Module 4C: ’Load cases and combinations’
Module 4D: ’Constraints’

This project is described hereafter according to the three steps in the proposed design
phase as shown in fig 68 on page 113.

Step 1: Think, Talk & Define


The Canakkale Antenna tower project is a side step in this graduation thesis because
it describes a non-glass structure. Nevertheless it is included in this thesis report to
indicate that the use of the method has a bigger field of application than glass struc-
tures. This antenna tower project was part of an international competition done in
cooperation with IND and Powerhouse company. At the end of this thesis project it
became clear that the tower has won the first price and is scheduled to be completed
in 2015. [ArchDaily, 2014].

The area where this tower is going to be build is a highly seismic area. Therefore
engineering report for the competition needed to include, besides the static calcula-
tions, seismic calculations as well. Within step 1 of this project, so before modelling
took place, in cooperation with the architect the scoop of the model was determined
including the desired geometric variables. These are shown in figure 76

Step 2: The AE-model


Within the AE-model four modules are chained.The modules are summarized in the
grey box on the top of this page. Because the geometry is a line-model and does not
include any panels which require a finite element division the challenge in the process

130
7.6 the canakkale antenna tower

Figure 76.: Model geometric variables

of modelling of this tower lies within the scripting of the twisted legs of the tower.

Step 3: Generate concepts and make a choice


To explain the calculations done one concept booklet is made. Although more con-
cepts were made by use of the AE-model only the last version is included in this thesis
report.

131
using the ae-method: some examples

Figure 77.: Top: Architectural impression, Bottom: Output AE-model

132
Part V

C O N C L U S I O N A N D R E C O M M E N D AT I O N S
8
CONCLUSION

This thesis which is done from Jun 2013 till april 2014, is an investigation concerning
the following research question:

Is it possible to find a design methodology that fully utilizes the available computational power
and can this improve the quality of the engineering advise giving in the design phase of glass
structures?

This thesis investigation resulted in an AE-model and AE-method which are both
described in part III and part IV of this main report. To validate the AE-model and
AE-method, they were tested during this thesis on some virtual projects as well as
three projects in real life (see chapter 7). The conclusions which can be drawn from
the building, testing and usage of the AE-model and AE-method are described below
and are divided into two categories. The first category describes all the technical con-
clusions covering the technical part of the AE-model. The second category describes
the process-oriented conclusions which can be drawn from the AE-method.

8.1 technical conclusions

• With regard to the technical side of the research question, the question can be
answered positive. The AE-method and AE-model shows that computational
power has progressed to a level where it can be of great interest for engineers in
the design of glass structures.

• The AE-method increases the speed in which engineering advise can be given.
This leads to the availability of engineering advise with a higher level of de-
tail during the design phase of glass structures. This is mostly advantageous
in complex architectural designs where structure and architecture are strongly
integrated.

• The AE-model, which is build by use of parametric modelling, shows that it


can be successfully used for the panellization of architectural shapes (Module
2A & 2B) the FE-calculation of Tri and quad shaped elements (Module 3A & 3B)

135
conclusion

the placing of complex load cases on structures (Module 3C) and can be used
as a parametric engineering tool (Module 4A untill 4C). All the modules are
explained in part III of this thesis report.

• The test projects show that these modules can be successfully chained in order
to speed up the complete calculation of glass structures.

• The Antenna tower project shows that the AE-methodology can also be used for
non-glas structures.(chapter 7, section 7.6).

• The AE-model is depending on four program licenses, sometimes given by one


man companies. This indicates vulnerability.

8.2 process-oriented conclusions

During the usage and development of the AE-model and AE-method it became clear
that next to having technical conclusions there are a lot, if not more, non-technical
conclusions which can be drawn.

• The speed at which engineering related questions appear will increase. More
questions need to be answered at the beginning of the design process. This will
cause the ’traditional’ workflow to change (chapter 6, section 6.4).

• The main consequence of the proposed work flow will be that architectural and
engineering boundaries need to be known before the modelling takes place.

• The new work flow requires a closer cooperation between architect and engineer
compared with the traditional design process. The AE-model becomes therefore
just as much a cooperation tool as a calculation tool.

• The standardized concept booklets support this cooperation greatly. The book-
lets include the concept specific input as well as output and give in this manner
rapid insight in the structural behaviour of architectural designs and gives feed-
back about stiffness, strength and stability.

• Using the booklets for communicative means can lead quickly to the understand-
ing of each others expectations, constraints and desires.

136
9
R E C O M M E N D AT I O N

During this thesis not every aspect could be analysed and some assumptions have been
made. Furthermore, several aspects have been encountered which require further op-
timisation. The recommendations concerning the modules included in the AE-model
are addressed per module in the last section of each chapter in part III. All the other
recommendations are described below:

• Investigate the uncertainties in the model: unknown wind loads on complex


free-forms, the lack of fatigue calculations, buckling calculations and connection
detailing. Try to understand their influence. Hereafter, try to find a way to take
them into account in the design process without decreasing the speed of the
model.

• Investigate other software applications to diminish the existing short comings.


such as CFD (computational fluid dynamics) solvers for wind calculations (espe-
cially turbulence).

• Rewrite part of the Grasshopper script in a programming language such as


Phython in order to increase the processing speed and decrease the size.

• Test and validate the ’equivalent material properties’ And optimize the connec-
tion approach as described in module 3A & 3B.

• Improve the clarity and readability of the AE-model script. Try finding a way
in which the structure of the model becomes clear on paper like a car electric
scheme (fig 78). This can help communicating the working of the AE-model.

• Find ways to work together in a single AE-model script.

• Investigate the coupling of geometry data within the AE-model with a cost esti-
mation (Excel) sheet. In this way per concept(booklet) an indicative price can be
given. This enables the economical judgement of the concepts.

• Implement structurally optimization tools. Which can advise a new geometry


( such as new node locations) in order to improve the structural efficiency and
material usage.

137
recommendation

• Implement sub-modelling techniques for the detailed analysis of crucial details.

Figure 78.: 2 CV electric scheme

138
BIBLIOGRAPHY

ArchDaily. Winner revealed, 2014. URL http://www.archdaily.com/481309/


winner-for-the-canakkale-antenna-tower-competition-revealed/.

Klaus-Jurgen Bath and Alexander Iosilevich. An evaluation of the mitc shell elements.
Elsivier, 75:30, 1999.

F.A.Veer. The strength of glass, a nontransparent value. F.A.Veer, 2007.

F.P.Bos. Safety Concepts in Structural Glass Engineering. PhD thesis, TU Delft, 2009.

GBO. Glas door de eeuwen heen, 2012. URL http://www.glasbrancheorganisatie.


nl/glas-techniek/glas-door-de-eeuwen-heen/.

D. Honfi and M. Overend. Glass structures - learning from experts. In Mid-term


Conference on Structural Glass, 2013.

Phill-Seung Lee and Klaus-Jurgen Bathe. Development of mitc isotropic triangular


shell finite elements. Elsevier, 82:18, 2005.

Ian James Merchant. English Medieval Glass-making Technology. PhD thesis, University
of Sheffield, 1998.

M.H.Toussaint. A design tool for timber gridshells. Master’s thesis, Technische Uni-
versiteit Delft, 2007.

Jon Mirtschin. Geometry gym, rhinoceros plugin, 2014. URL http://geometrygym.


blogspot.nl/search/label/GSA.

Phil and Nancy Seff. Phoenicians credited with discovering glass, 1999. URL http:
//www.glasslinks.com/newsinfo/phoenician.htm.

Helmutt Pottman. Architectural Geometry. Bentley Institute Press, 2007.

J.E. Shelby. Introduction to glass science and technology. The royal society of Chemestry,
2005.

WolframMathWorld. Geometry, 2014. URL http://mathworld.wolfram.com/topics/


Geometry.html.

139
Part VI

APPENDIX
A
C AT E G O R I S AT I O N O F S U R FA C E S

143
Surface group Example shapes Example structures
1. Flat surfaces

shape

shape
Glass facade
concept

2. Rotational surfaces
One-sheet rotational Two-sheet rotational

shape variable rotational paraboloid


cylinder
(scaling) rotational hyperboloid
cone St Mary axe
prolate rot. ellips ( Norman Foster)
shape torus
1-sheet hyperbo- oblate rot. ellips Torre Agbar
sphere ( Jean Nouvel)

3. Translational surfaces

shape
ellips circle
shape
Japanese Pavillion
( Frei Otto)
hyperbolic paraboloid eliptical paraboloid translational paraboloid

4. Ruled surfaces

shape variable

shape variable
(scaling)

New state gallery (James Stirling) Ysios Winery ( Santiago Caletrava)

4a. HP surfaces (double

shape
or
shape

Concrete shells
( Felix Candela)
HP surfaces can be both : double ruled and translational surface

4b. Developable surfaces

shape

shape

Guggenheim
D-forms (by Tony Wills) ( Frank Gehry)
Surface group Example shapes Example structures
5. Helical surfaces

shape constant

shape constant
Spiral stair
( Unknown )

6. Pipe surfaces

shape constant

shape constant Scaffolding


( Unknown )

7. Free-from surfaces
Free-form surfaces are build from:

Admirant
( Massimiliano Fuksas )
Bezier curves B-spline curves NURBS curves Bezier surfaces B-spline surfaces NURBS surfaces
B
SCREENSHOT AE-MODEL: ELLIPSOID

145

You might also like