University of Southern California Dissertations and Theses

Performative shading design: parametric based measurement of shading system configuration effectiveness and trends

(USC Thesis Other)

Performative shading design: parametric based measurement of shading system configuration effectiveness and trends

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

PERFORMATIVE SHADING DESIGN:

PARAMETRIC BASED MEASUREMENT OF SHADING SYSTEM CONFIGURATION
EFFECTIVENESS AND TRENDS

by

Tyler James Tucker

______________________________________________________________________________

A Thesis Presented to the
FACULTY OF THE USC SCHOOL OF ARCHITECTURE
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF BUILDING SCIENCE

December 2014

Copyright 2014 Tyler James Tucker

ACKNOWLEDGEMENTS

Chuck Khuen from Weather Analytics for providing pinpoint weather data for a near-downtown Los
Angeles weather station.
Tammy Jow, David Martin, and AC Martin for providing necessary drawings and models of their
developing high rise project. For meeting with me and providing feedback on initial ideas and methods to
developing this process.
Jeffrey Vaglio and Enclos Corp. for allowing me to use time and resources from our office to complete
my research.
Jeff Landreth for going beyond your studio teaching responsibilities and helping me tackle the most
difficult technical hurdles in my research. For providing data parsing and sensitivity analysis resources
and guiding my visual representation of data.
To my thesis committee team: Douglas Noble (chair), Kyle Konis (2nd), Karen Kensek (3rd), and Jeffrey
Vaglio (4th). You cultivated my initial research ideas, guided and narrowed my interests, provided
invaluable feedback, and most importantly, patiently pushed me along – for this, I thank you.

ii
TABLE OF CONTENTS

ACKNOWLEDGEMENTS .......................................................................................................................... ii
LIST OF FIGURES ..................................................................................................................................... vi
LIST OF TABLES ....................................................................................................................................... ix
ABSTRACT .................................................................................................................................................. x
Chapter One - Introduction ........................................................................................................................... 1
1.1 Problem ......................................................................................................................................... 1
1.2 Design Space ................................................................................................................................. 2
1.3 Hypothesis..................................................................................................................................... 3
1.4 Proposed Solution ......................................................................................................................... 3
1.4.1 Boundaries ............................................................................................................................ 3
1.4.2 Process .................................................................................................................................. 4
1.5 Objective ....................................................................................................................................... 5
1.6 Chapter Structure .......................................................................................................................... 5
Chapter Two - Background ........................................................................................................................... 7
2.1 Daylight......................................................................................................................................... 7
2.1.1 Solar Radiation ...................................................................................................................... 7
2.1.2 Illuminance-based Metrics .................................................................................................. 10
2.1.3 Luminance-based Metrics ................................................................................................... 12
2.1.4 Shading ............................................................................................................................... 14
2.2 Analysis Software ....................................................................................................................... 17
2.2.1 Radiance .............................................................................................................................. 18
2.2.2 DAYSIM ............................................................................................................................. 19
2.2.3 Rhinoceros 3D and Grasshopper ......................................................................................... 19
2.2.4 DIVA ................................................................................................................................... 20
2.3 Optimization ............................................................................................................................... 21
2.3.1 Galapagos ............................................................................................................................ 22
2.4 Sensitivity Analysis .................................................................................................................... 22
Chapter Three – Methodology .................................................................................................................... 24
3.1 Reference Building ..................................................................................................................... 24
3.2 Software Platform ....................................................................................................................... 25
3.3 Parametric Definition .................................................................................................................. 26
3.4 Inputs........................................................................................................................................... 29
3.4.1 Geometry ............................................................................................................................. 29
3.4.2 Vectors and Sensors ............................................................................................................ 33
iii
3.4.3 Weather Data....................................................................................................................... 34
3.5 Parameters ................................................................................................................................... 35
3.6 Measurements ............................................................................................................................. 37
3.6.1 Solar Radiation .................................................................................................................... 39
3.6.2 Useful Daylight Illuminances ............................................................................................. 40
3.6.3 Periodic Glare Evaluation ................................................................................................... 43
3.7 Optimization ............................................................................................................................... 45
3.8 Data ............................................................................................................................................. 49
3.9 Re-instantiation ........................................................................................................................... 52
Chapter Four - Results ................................................................................................................................ 54
4.1 Process Results ............................................................................................................................ 54
4.2 Baseline – Open Office ............................................................................................................... 54
4.3 Hotel ............................................................................................................................................ 57
4.4 Closed Office .............................................................................................................................. 60
4.5 East .............................................................................................................................................. 63
4.6 Winter ......................................................................................................................................... 65
4.7 Summer ....................................................................................................................................... 67
4.8 Double UDI ................................................................................................................................. 70
4.9 Triple UDI ................................................................................................................................... 72
4.10 Double DGPs .......................................................................................................................... 74
4.11 Triple DGPs ............................................................................................................................ 77
Chapter Five - Discussion ........................................................................................................................... 80
5.1 Comparisons ............................................................................................................................... 80
5.2 Typologies ................................................................................................................................... 80
5.3 Orientations ................................................................................................................................. 83
5.4 Timeframes ................................................................................................................................. 85
5.5 Weighting Factors – UDI ............................................................................................................ 87
5.6 Weighting Factors – DGPs ......................................................................................................... 89
5.7 Process Discussion ...................................................................................................................... 91
Chapter Six - Conclusions .......................................................................................................................... 93
6.1 Process ........................................................................................................................................ 93
6.2 Future Work ................................................................................................................................ 93
6.2.1 Form-finding Geometry ...................................................................................................... 93
6.2.2 Material Characteristics ...................................................................................................... 94
6.2.3 Improved Sample Size ........................................................................................................ 94
6.2.4 Thermal integration ............................................................................................................. 95
iv
6.2.5 True Multi-Objective Solving ............................................................................................. 95
6.3 Process Conclusions .................................................................................................................... 95
BIBLIOGRAPHY ....................................................................................................................................... 97
APPENDIX: Simulation Data ................................................................................................................... 101
APPENDIX: User Guide .......................................................................................................................... 111

v
LIST OF FIGURES

Figure 1. Transmission, Reflection, Absorption, and Emission of Solar Radiation through Glass .............. 8
Figure 2. Illuminance and Luminance Measurements ................................................................................ 11
Figure 3. Exterior Shades, Interior Blinds, Frits, and Glass Types/Coatings (Enclos Corp 2014) ............. 15
Figure 4. Coating on Surface 3 for Hot Climates, Coating on Surface 2 for Cold Climates ...................... 16
Figure 5. Renderings and the Rhino Model of the Wilshire Grand Tower ................................................. 24
Figure 6. Overall Grasshopper Parametric Definition ................................................................................ 27
Figure 7. Counterclockwise: Hotel, Open Office, Closed Office, and East Facing Open Office ............... 30
Figure 8. Hotel Room Materials - Glass, Ceiling, Walls, and Floor ........................................................... 32
Figure 9. Parameters and Sensors in the Hotel Room ................................................................................. 34
Figure 10. Workflow Diagram .................................................................................................................... 38
Figure 11. DIVA Daylight Component - Solar Irradiation Settings ........................................................... 39
Figure 12. DIVA Daylight Component - Climate Based Settings .............................................................. 41
Figure 13. DIVA Daylight Component - Climate Based Outputs .............................................................. 42
Figure 14. Galapagos Component Settings ................................................................................................. 48
Figure 15. Galapagos Solving ..................................................................................................................... 49
Figure 16. Recorded and Exported Data into Excel .................................................................................... 50
Figure 17. Three Sets of Duplicate Results in Excel .................................................................................. 50
Figure 18. Linear Regression Significance Indicators for Parameters ........................................................ 51
Figure 19. Eight scatter plots with polynomial trend lines for each parameter .......................................... 52
Figure 20. Re-instantiation of Simulation #703's Configuration Back to Rhino ........................................ 53
Figure 21. (Open Office, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations ............ 55
Figure 22. Shading System Visualization and Configuration Information for #711 and #1131 ................. 56
Figure 23. (Open Office, South, Annual, 1:1:1 Weighting) Significant Parameters .................................. 56
Figure 24. (Open Office, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length ........ 57
Figure 25. (Hotel, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations ....................... 58
Figure 26. Shading System Visualization and Configuration Information for #970 and #900................... 59
Figure 27. (Hotel, South, Annual, 1:1:1 Weighting) Significant Parameters ............................................. 59
Figure 28. (Hotel, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length ................... 60
Figure 29. (Closed Office, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations ......... 61
Figure 30. Shading System Visualization and Configuration Information for #309 and #24..................... 61
Figure 31. (Closed Office, South, Annual, 1:1:1 Weighting) Significant Parameters ................................ 62
Figure 32. (Closed Office, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length ..... 62
Figure 33. (Open Office, East, Annual, 1:1:1 Weighting) Top Two Performing Configurations .............. 63
Figure 34. Shading System Visualization and Configuration Information for #1199 and #852 ................. 64
vi
Figure 35. (Open Office, East, Annual, 1:1:1 Weighting) Significant Parameters ..................................... 64
Figure 36. (Open Office, East, Annual, 1:1:1 Weighting) Louver Length and Louver Angle ................... 65
Figure 37. (Open Office, South, Winter, 1:1:1 Weighting) Top Two Performing Configurations ............ 66
Figure 38. Shading System Visualization and Configuration Information for #1012 and #1215 ............... 66
Figure 39. (Open Office, South, Winter, 1:1:1 Weighting) Significant Parameters ................................... 67
Figure 40. (Open Office, South, Winter, 1:1:1 Weighting) Number of Fins and Louver Length .............. 67
Figure 41. (Open Office, South, Summer, 1:1:1 Weighting) Top Two Performing Configurations .......... 68
Figure 42. Shading System Visualization and Configuration Information for #110 and #1....................... 69
Figure 43. (Open Office, South, Summer, 1:1:1 Weighting) Significant Parameters ................................. 69
Figure 44. (Open Office, South, Summer, 1:1:1 Weighting) Number of Fins and Horizontal Offset ....... 70
Figure 45. (Open Office, South, Annual, 1:2:1 Weighting) Top Two Performing Configurations ............ 71
Figure 46. Shading System Visualization and Configuration Information for #79 and #721..................... 71
Figure 47. (Open Office, South, Annual, 1:2:1 Weighting) Significant Parameters .................................. 72
Figure 48. (Open Office, South, Annual, 1:2:1 Weighting) Louver Length and Number of Louvers ........ 72
Figure 49. (Open Office, South, Annual, 1:3:1 Weighting) Top Two Performing Configurations ............ 73
Figure 50. Shading System Visualization and Configuration Information for #926 and #1011 ................. 73
Figure 51. (Open Office, South, Annual, 1:3:1 Weighting) Significant Parameters .................................. 74
Figure 52. (Open Office, South, Annual, 1:3:1 Weighting) Louver Length and Louver Angle ................. 74
Figure 53. (Open Office, South, Annual, 1:1:2 Weighting) Top Two Performing Configurations ............ 75
Figure 54. Shading System Visualization and Configuration Information for #57 and #61....................... 75
Figure 55. (Open Office, South, Annual, 1:1:2 Weighting) Significant Parameters .................................. 76
Figure 56. (Open Office, South, Annual, 1:1:2 Weighting) Louver Angle and Horizontal Offset ............. 76
Figure 57. (Open Office, South, Annual, 1:1:3 Weighting) Top Two Performing Configurations ............ 77
Figure 58. Shading System Visualization and Configuration Information for #17 and #43....................... 78
Figure 59. (Open Office, South, Annual, 1:1:3 Weighting) Significant Parameters .................................. 78
Figure 60. (Open Office, South, Annual, 1:1:3 Weighting) Louver Angle and Horizontal Offset ............. 79
Figure 61. Open Office, Hotel, and Closed Office Configurations ............................................................ 81
Figure 62. (Hotel/Open Office/Closed Office, South, Annual, 1:1:1 Weighting) Number of Louvers ...... 82
Figure 63. (Hotel/Open Office/Closed Office, South, Annual, 1:1:1 Weighting) Louver Length ............. 82
Figure 64. South and East Open Office Configurations ............................................................................. 83
Figure 65. (Open Office, South/East, Annual, 1:1:1 Weighting) Louver Length ....................................... 84
Figure 66. (Open Office, South/East, Annual, 1:1:1 Weighting) System H Offset .................................... 84
Figure 67. Annual, Winter, and Summer Open Office Configurations. ..................................................... 85
Figure 68. (Open Office, South, Annual/Winter/Summer, 1:1:1 Weighting) Number of Fins ................... 86
Figure 69. (Open Office, South, Annual/Winter/Summer, 1:1:1 Weighting) Louver Length .................... 86
Figure 70. Equal Weighted, Double Weighted, and Tripled Weighted UDI Configurations ..................... 87
vii
Figure 71. (Open Office, South, Annual, 1:1/2/3:1 Weighting) Louver Length......................................... 88
Figure 72. (Open Office, South, Annual, 1:1/2/3:1 Weighting) Number of Louvers ................................. 89
Figure 73. Equal Weighted, Double Weighted, and Triple Weighted DPGs Configurations ..................... 89
Figure 74. (Open Office, South, Annual, 1:1:1/2/3 Weighting) Louver Angle .......................................... 90
Figure 75. (Open Office, South, Annual, 1:1:1/2/3 Weighting) Horizontal Offset .................................... 91
Figure 76. Open Office Parameters ........................................................................................................... 101
Figure 77. Hotel Parameters ..................................................................................................................... 102
Figure 78. Closed Office Parameters ........................................................................................................ 103
Figure 79. Open Office, East Parameters .................................................................................................. 104
Figure 80. Open Office, Winter Parameters ............................................................................................. 105
Figure 81. Open Office, Summer Parameters ........................................................................................... 106
Figure 82. Open Office, 2x UDI Parameters ............................................................................................ 107
Figure 83. Open Office, 3x UDI Parameters ............................................................................................ 108
Figure 84. Open Office, 2x DGPs Parameters .......................................................................................... 109
Figure 85. Open Office, 3x DGPs Parameters .......................................................................................... 110
Figure 86. Context Geometry and Reference Building ............................................................................. 111
Figure 87. Reference Building and Outlined Rooms for Analysis ........................................................... 111
Figure 88. Open Office with Orientation and Measurement Sensors ....................................................... 112
Figure 89. Parametric Shading System Applied to Open Office .............................................................. 112
Figure 90. Simulation Data Exported to Excel Template ......................................................................... 113
Figure 91. Simulation Data Exported to Excel Template, Scatter Plots with Trend Lines ....................... 113
Figure 92. Re-instantiated Configuration in Rhino ................................................................................... 114

viii
LIST OF TABLES

Table 1. Default Material Properties ........................................................................................................... 32
Table 2. Parameter Values Range ............................................................................................................... 35
Table 3. Glazing Type Parameter Properties .............................................................................................. 36
Table 4. Daylight Glare Probability Brackets ............................................................................................. 44

ix
ABSTRACT

A tension is developing between the goal of more energy efficient buildings and the growing complexity
of architecture. Conceptual design tools are needed that generate, visualize, and optimize building
performance. Highly glazed enclosures offer excellent views and allow large amounts of natural light in
the interior. Yet even the design of shading devices must take into account dynamic solar conditions,
direct solar radiation, daylight, glare, the geometry of the building, site, glazing properties, and other
factors.
By providing designers with an automated method to generate shading systems, visualize the performance
and daylighting quality of each unique configuration, and analyze the variable trends, better informed
and more effective choices regarding façade configuration can be achieved.
It is difficult to predict the quality of daylight or solar control of conceptually designed systems with so
many variables affecting them – local climate, geometry, orientation, materials, glazing properties, etc.
The implementation of shading systems has historically involved creating a series of designs and
validating their performance through individual simulations. Therefore, a design process that optimizes
the conceptual design of shading systems, but also portrays how each variable is influencing the overall
performance, can better guide and inform designers.
A workflow has been developed that incorporates multiple design variables to be simulated
simultaneously, resulting in comprehensive results displaying both raw data to validate decisions as well
as optimized façade configurations to guide designers. The proposed optimization and visualization
process focuses on three separate metrics: solar radiation, useful daylight illuminance, and glare
probability. Rhinoceros 3D and Grasshopper are the parametric and modeling software used as the main
process interface, with DIVA included as the climate-based and daylighting simulation engine.
Galapagos functions as the evolutionary solving engine used to generate and optimize different variations
of shades as well as glazing properties. The results are optimized, parametrically driven shading solutions
x
with their solar performance and daylighting quality visually displayed. Sensitivity analysis among
design variables and spatial occupancy follows the optimization results, giving designers a full suite of
information to inform their designs.
xi
CHAPTER ONE - INTRODUCTION

1.1 Problem
The current building stock accounts for 48.7% of the total energy consumption in the United States
(Mazria 2011). The building enclosure is a large contributor to the overall energy consumption of a
building. The building envelope has been shown to contribute up to 50-60% of the total heat gain in
buildings, resulting in excessive HVAC use to balance the indoor temperature (Rashwan, Farag and
Moustafa 2014) (Mwasha, Williams and Iwaro 2011). The role of the façade is to maintain the comfort
and security of the interior against the outside environment. Regulation of temperature, protection from
sound, excess sunlight, pollution, and other environmental factors are all encompassed in the role of the
building envelope. In addition to maintaining thermal and visual comfort of the interior, the façade also
provides a visual link to the environment. Through these responsibilities, the façade directly influences
the energy consumption of a building with heating, ventilation systems, and air conditioning and artificial
lighting.
In order to meet the challenges of energy saving initiatives and codes such as Architecture 2030,
California Title 24, and ASHRAE 90.1, the building façade can and should be carefully designed to
minimize energy consumption (Sadineni, Madala and Boehm 2011). However, the effectiveness of a
façade system is not entirely based on minimizing energy usage; solar and visual considerations such as
limiting glare and the preservation of views to the exterior are also important. The performance of a
façade system can therefore be defined as a combination of thermal and visual comfort, which in turn
means energy savings as a result of reduced heating and cooling loads. Solar considerations are the
performance metrics: covering solar radiation passing through the façade, regulating and maintaining
natural light into the interior, and minimizing discomfort glare.
Highly glazed enclosures allow large amounts of natural light to enter a space, reducing the need for
artificial lighting and saving energy. However, too much light may introduce excessive heat gains and the
1
possibility of uncomfortable glare. While a reduction on glazing area may limit heat gains and reduce
glare, it may also increase the use of artificial lighting due to lack of natural light. One method of
moderating solar influences (solar radiation, daylighting, and glare), while still preserving views to the
outside, is the integration of an external shading system into the façade. Simple shading devices such as
overhangs, fins, and louvers can be designed to perform effectively in different climates, orientations, and
timeframes, providing excellent solar control.
1.2 Design Space
The choices made during the conceptual design stage of a building directly affect its performance. Many
of these decisions build the foundation for future design decisions to be made. Therefore, it is critical that
accurate and informed design moves are conducted in this stage of design (Granadeiro, et al. 2013).
Conceptual design can be considered during the pre-design or schematic design phase of a project, where
the form of the building is still being massaged into place and glass sizes and placements are still being
iterated. The design and implementation of a shading system while the building is going through
conceptual design is a responsive process based on iterating designs and receiving feedback – both
performance and design based.
Current means of developing designs and design alternatives of shading systems focus on generating
small, manageable working sets of designs and basing decisions off of those original concepts (Akin
2001). These designs are analyzed, reviewed, and adjusted accordingly, but are limited by the small
sample size of design alternatives. Furthermore, analysis is generally performed individually on each
option, therefore extra options require more time and resources to narrow down options and search for the
most effective system.
Increasing design iterations and alternatives can provide a higher likelihood of finding effectively
performing designs (Aish and Woodbury 2005). Parametric modeling incorporates defined characteristics
into a variable, allowing changes to be made to specific portions of a model quickly and in a controlled
2
environment. The parameterization of a conceptual design can lead to a generative exploration of design
alternatives. Once design intentions have been digitized via parameters, the parametric model can be
continually adjusted to produce alternative designs that follow the constraints of the project as well as the
direction and intents of the designer.
1.3 Hypothesis
By providing designers with an automated method to generate shading systems, visualize the performance
and daylighting quality of each unique configuration, and analyze the variable trends, better informed
and more effective choices regarding façade configuration can be achieved.
1.4 Proposed Solution
A workflow has been developed that incorporates multiple design variables to be simulated
simultaneously, resulting in comprehensive results displaying both raw data to validate decisions as well
as optimized façade configurations to guide designers. The proposed optimization and visualization
process focuses on three separate metrics: solar radiation, useful daylight illuminance, and glare
probability. Rhinoceros 3D and Grasshopper are the parametric and modeling software used as the main
process interface, with DIVA included as the climate-based and daylighting simulation engine.
Galapagos functions as the evolutionary solving engine used to generate and optimize different variations
of shades as well as glazing properties. The results are optimized, parametrically driven shading solutions
with their solar performance and daylighting quality visually displayed. Sensitivity analysis among
design variables and spatial occupancy follows the optimization results, giving designers a full suite of
information.
1.4.1 Boundaries
The focus of this research is based on providing a process for designers that is able to parameterize a
shading system and find an optimal and effective configuration, while also revealing how each parameter
contributed to the overall performance of the system. In particular, this study defines the performance
metrics of this process as: solar radiation, useful daylight illuminance, and daylight glare probability.
3
While many parameters are considered, only those related to daylighting and solar controls were utilized.
Thermal properties, construction materials, and HVAC systems were not considered in this study.
1.4.2 Process
The process begins with integrating a project model into Rhino and Grasshopper. Gathering up and
modeling the surrounding building context, climate data, and specifying which parts of the building are to
be analyzed. To provide accurate simulations each room needs to be fully modeled with proper materials
applied. Those materials can be included as variables down the line when the parameterization of the
model is defined. Next the application of the shading system needs to be parameterized to define the
design intent and scope of what is being tested. For example, what range of depth is to be tested on the
louvers of a shading system? A series of sensors are placed in the rooms to measure the three considered
metrics. One measures illuminance at the task surface placed at desk height in the space. Another
measuring simplified daylight glare probability is placed at eye level looking towards the exterior. Lastly,
a grid of sensors all facing the exterior are located just inside the exterior face of the room measuring the
amount of solar radiation that enters the space. These measurements combined with occupancy schedules
of the simulated spaces will form the performance criterion of shading system configuration. To apply an
evolutionary solving algorithm to rapidly iterate shading system configurations the three measurements
are combined into a single condensed fitness value which ranks the configurations and produces new
designs based on the previous high performing ones.
Once a sufficient number of simulations have been completed to find relative convergence in the
parameter values, the data is streamed into a pre-defined Excel template file. The configurations are
sorted and color coded by their fitness value – highlighting the top performing configurations. This data
also performs another role, it enables the user to re-instantiate the parameter values from any
configuration back into the Rhino model to visualize the shading system. Additionally, a simple linear
regression analysis is performed on each parameter compared to the fitness values, finding which
parameters have a strong correlation and ultimately a strong significance to the outcome. In the same
4
template file all the parameter values derived from the simulations are displayed on scatter plots with
polynomial trend lines indicating how changes in parameter values correspond with fitness values.
Seeing the shading system configurations combined with the graphical analysis and regression statistical
significance provides the designer with a suite of information about what works the best and why.
The design solution space, software platform, and parametric definition define the scope and
methodology of this research. The simulations sequence, data retrieval, sorting, and significance are
carried out at the back end forming the principle information in which guides and informs the user.
1.5 Objective
The goal of this research is to provide architects and designers a process that will inform and guide design
decisions regarding the development of shading systems for building facades. Traditionally the
environmental analysis and simulations of shading system configurations would be conducted in specific
software and often with the assistance of outside consultants. This process seeks to streamline and
simplify the workflow so that architects and designers can complete the entire design cycle, without
having to introduce other disciplines or additional software.
To achieve the objective of this research the process needs to be easy to integrate into existing designing
practices. Integrating a project is done using existing software used to model and design projects. To
facilitate the decision making process the results are easy to navigate and understand. The results show
contextually optimum solutions for the given inputs based on observed metrics, timeframes, and variable
weighting. Visualization that reflects both the individual and overall analysis from simulations is
documented in graphical and numerical form with color coding and even re-instantiated back into the 3d
model it came from.
1.6 Chapter Structure
The outlined chapters define the steps taken to conduct and document this research: introduction,
background research, methodology, results and discussion, and conclusions and future research.
5
Chapter two discusses the nature of sunlight, each of the three performance metrics (solar radiation,
useful daylight illuminance, and glare probability), and the tools and methods used simulate, measure, and
extract meaningful information from them.
Chapter three introduces the study building and the entire evaluation process of shading system
configurations from inputs to results to visualization. It begins by defining each step in the process and
detailing out how each input is fulfilled, parameter created and flexed, measurements set and conducted,
optimization is carried out, data is sorted and exported, and schemes can be color-coded and visualized.
Chapter four examines each of the ten scenarios and analyzes the data and optimal configurations. Each
scenario includes the raw data, significant parameters, re-instantiated top configurations, and scatter plots
of select parameters measured against the fitness value.
Chapter five focuses on the comparisons of the scenarios and show how optimal parameter ranges for
building typologies, timeframes, orientations, and variable weighting can be applied to shading system
and building design.
Chapter six gathers conclusions about the optimal shading system parameters and configurations for the
study building and how the process performed overall. It discusses the effectiveness and limitations of
the process and leads into future improvements and applications. It also discusses future work that can be
conducted to improve the process to include complex form-finding geometry, material properties,
expanded testing scenarios, thermal metrics, and more advanced multi-objective solving processes.
6
CHAPTER TWO - BACKGROUND

2.1 Daylight
Daylight can be described as the combination of all direct and indirect light sources occurring during the
daytime. It is made up of sunlight and diffuse sky radiation. Sunlight is composed of three types of
electromagnetic radiation – visible light, infrared, and ultraviolet. Visible light produces the light we can
see with the naked eye as well as the diffuse and ambient light that is absorbed and reflected off objects.
Infrared radiation produces radiant heat when absorbed and re-radiated by objects. Diffuse sky radiation
is created by the scattering of sunlight by molecules in the atmosphere or water vapor as in clouds. These
all combine to create the light we use to see things during the day and also the warmth felt by sunlight.
Measurements of the components of direct and indirect sunlight are divided into three categories: the
radiant heat energy of direct sunlight’s solar irradiation on surfaces, the intensity of direct and indirect
light occurring on a surface, and the amount of direct or indirect light either reflecting off a surface or
entering the eye directly.
2.1.1 Solar Radiation
Solar radiation is measured by the amount of solar energy received on the surface of an object. When
solar radiation hits an object the infrared radiation is absorbed, reflected, and re-radiated to nearby
objects. However, when solar radiation comes into contact with something transparent like glass, some
amount of solar radiation is transmitted through the glass and some absorbed (Figure 1). The absorbed
radiation is re-radiated by the glass to nearby objects both inside and outside of the space. Because part
of the radiation is re-radiated back into the space, there is a net gain compared to the direct transmission
of radiation in the space, called solar gain. The transparency and makeup of the glass affect the amount
of visible and infrared light transmitted.
7

Figure 1. Transmission, Reflection, Absorption, and Emission of Solar Radiation through Glass
Solar gains present a benefit in colder climates by trapping heat energy inside an enclosed space, but in
hotter climates solar gains must be avoided to prevent overheating. In either case, solar gains are directly
linked to energy consumption of a building due to the regulation of the temperature inside. If a space is
too cold from lack of sunlight, the furnace is used to compensate. If a space is too hot from overexposure,
the air conditioning is used to cool the space. Some means of managing solar gains are done through
shading systems, frits, coatings, and the glass products themselves. Shading systems obstruct solar
radiation before it is even able to reach the glass surface and transmit and re-radiate heat energy. Frits can
limit the amount of solar radiation that is transmitted into the space. Coatings can both inhibit the amount
of solar radiation transmitted to a space, but also selectively mitigate infrared radiation while maintaining
high visual transmittance. In the United States, windows and glass systems have ratings to reflect their
ability to transmit radiation and light through them. The Solar Heat Gain Coefficient (SHGC) reflects a
value from 0 to 1 that translates to the amount of radiation transmitted through the glass and the amount
of energy that is absorbed and re-radiated.
Total Incident
Solar Radiation
R
Absorption
Glass
Direct
Transmission
Secondary
Emission
Secondary
Emission
SHGC
8
The measurement of solar radiation on a surface over a period of time is called insolation or solar
irradiation. Solar irradiation can be physically measured using a pyranometer or a pyrheliometer. A
pyranometer measures the amount of irradiance on a surface with a hemispherical view. Solar radiation is
absorbed on a black thermopile sensor with a glass dome protecting it. The absorbed heat energy is
converted to an electrical signal and recorded. A pyrheliometer follows the sun directly and absorbs a
sunbeam into its black thermopile sensor. The resulting electrical signal can to calculate the amount of
watts per square meter.
Simulating and measuring solar radiation in 3d model space can be done with a variety of methods. One
method involves creating a grid of points acting as virtual pyranometers which can measure hourly solar
irradiance. The measurements can then be combined to form an aggregate solar irradiance measurement
of a surface. While this method works well for select periods of time, to cover an extended period of
time, for example one year, would be very computationally taxing.
The GenCumulativeSky method is an evolved methodology of calculating annual solar irradiation very
quickly with little sacrifice to accuracy (Robinson and Stone 2004). The new GenCumulativeSky module
is based on a combination of existing modules inside Radiance, a commonly used daylight simulating
software (Radiance n.d.). The GenSky and GenDaylit modules in Radiance are able to mathematically
produce continuous sky luminance distributions so that when a ray traced from a source intersects with
the sky vault (a mathematically generated construct of the sky, including direct and indirect sunlight) a
luminance value is given. With these modules it is possible to create continuous sky luminance models,
but difficult to provide cumulative continuous models. It is however possible to create a cumulative non-
continuous discretized version of a sky vault. This involves dividing up the sky vault into 145 tiles
following the Tregenza sky model, finding the centroid of each tile, applying a continuous luminance
distribution model at each centroid to find a measurement, and aggregating those measurements over a
period of time (Tregenza 1987). With the cumulative sky model in place, when a ray is traced backwards
to the source it will find cumulative radiation values in the discretized sky vault, rather than generating
9
each point one at a time as they intersect. This methodology is six times faster, with only two times
increase in root mean square error, producing an accurate yet computationally light simulation (Robinson
and Stone 2004).
2.1.2 Illuminance-based Metrics
Direct and indirect sunlight transmitting through windows and openings in rooms can adequately light
them without the need for artificial lighting. A properly daylit space can therefore save energy by
hedging its lighting needs on readily available sunlight. How adequately and how deep daylight reaches
into space, governs its effectiveness. A room that is only partially daylit will still need to rely on electric
lights to maintain an adequate brightness for its occupants to perform tasks. The levels of illuminance
required for different tasks are specified by the Illuminating Engineering Society of North America
(IESNA). The IESNA releases a lighting handbook that dictates minimum levels of illuminance for a
variety of tasks as well as required levels for safely illuminating paths of egress (Illuminating Engineering
Society of North America 2000). These minimum levels are based upon many generic attributes of the
space, such as whether the task is being conducted on a flat horizontal surface.
Illuminance is measured by the amount of light hitting a surface, measured in lux or footcandles (Figure
2). Illuminance takes into consideration properties of materials which the light is hitting, including the
reflectivity and color absorption of the surface. A light meter is used to determine the illuminance. Using
the recorded illuminance and the IESNA guidelines, one can verify if a space is properly lit for the
appropriate task. The same idea translates to digital simulations; a virtual directional light meter is used
to measure the illuminance on a surface.
10

Figure 2. Illuminance and Luminance Measurements
Daylight Factor (DF) once was one of the most commonly used methods of expressing how much
daylight was entering a space. It is found by simulating a building in a CIE overcast sky model and
dividing the illuminance values measured inside the space by the outside overall ambient illuminance.
This generates a percentage that represents the fraction of daylight that is entering the space. Because the
CIE overcast sky is used, the only contributing variable is the design of the building; orientation does not
factor in. For example, identical buildings measured for their daylight factor in a south facing room in
Arizona versus a north facing room in Alaska will have the same daylight factor. This is done to maintain
continuity of measurements across all projects; all daylight factor values can easily be understood from
project to project regardless of location or orientation. Due to its limitations in excluding orientation and
climate, a push for new, more accurate daylight measurements were introduced in the early 2000s.
Daylight Autonomy (DA) was created to measure a percentage of the hours that a threshold of
illuminance was met in the allotted timeframe. Performing this evaluation requires constant
measurements at hourly intervals; much more sophisticated that measuring daylight factor. The
measurement of daylight autonomy was aided with the introduction of software like DAYSIM that
measures illuminance on the hour, each day, over any period of time (Daysim 2013). However, even this
method is not without flaws. By including all hours of the day, the daylight autonomy would always be
11
on the low side since half of the hours were measured during the night. Christoph Reinhart and his team
added an addendum to the way daylight autonomy was measured by only including occupied hours
(Reinhart and Walkenhorst 2001). The new daylight autonomy calculation now relied on using an
occupancy schedule in conjunction with hourly measurements. This method is able to produce accurate
reflections of how well a space is being naturally lit during the hours when it is actually in use.
Based on the daylight autonomy model, a series of other hourly daylight measurements have sprung up.
One is called Useful Daylight Illuminances (UDI). UDI follows the same idea of daylight autonomy,
measuring the illuminance at each hour and matching that with a corresponding occupancy schedule.
However, rather than just checking for a simple threshold of illuminance being met, UDI includes a
double threshold for checking the illuminance values: a minimum amount of light for a task and a
maximum before it becomes debilitating (Nabil and Mardaljevic 2005) (Nabil and Mardaljevic 2006).
Depending on the task or purpose of the space to be simulated, the minimum and maximum threshold
values can be adjusted. UDI enables the daylight measurement to be very precise in predicting if the
current space is getting an appropriate amount of natural daylight for a task.
2.1.3 Luminance-based Metrics
Where illuminance measures the amount of light hitting a surface, luminance measures the amount of
light leaving a surface in a direction. Luminance indicates the perceived amount of light that is reflected
or directly coming from a light source. This perception-based metric therefore is well suited to gauge if
there is an oversupply of light in a room, causing glare. Glare can be described as when there is a light
source in the field of vision of the eye that is brighter than what is being focused on, causing difficulty
seeing. Glare is divided into two categories: disability glare and discomfort glare. Disability glare is
where there is such an overabundance of light that vision is impaired trying to discern among objects.
Discomfort glare involves no visual impairment, but rather an annoyance or tiring of the eyes.
12
Measurement of glare is conducted through luminance distributions of what an occupant would see if they
were performing a task in a space. Many glare indices have been developed over time based on human
studies and correlating that data into equations and evaluations. The indices attempt to evaluate a space
as if a human occupant was there describing their comfort levels. Disability glare can be easily measured
by identifying a considerably bright light source in the luminance distribution of the scene. Measuring
discomfort glare is more difficult, involving re-creating human satisfaction and dissatisfaction with
luminance measurements.
One example of a recently validated glare index is Daylight Glare Probability (DGP). Developed in 2006,
DGP was derived from physical and human testing with varying participants, orientations, and room
configurations (Wienold and Christoffersen 2005) (Wienold and Christoffersen 2006). Important to DGP
is the use of physical data from an actual glare source. Many previous glare indices utilized artificial
glare sources for their development. Two identical tests were conducted in separate countries at the same
time measuring participant responses to a series of tasks in identical rooms with a CCD camera and
illuminance sensors placed around the room. One sensor at eye level to mimic a person sitting and two
other horizontal sensors to ensure the illuminance was the same in both rooms. A tool developed to
evaluate luminance contrast hot spots in the CCD camera imagery was used to mimic the participants’
responses to the room. From the physically recorded measurements and human responses to the
environment, a formula was developed that correlated both data sets in approximating perceived levels of
glare in each space.
DGP relies on evaluating luminance distributions of a scene at given times and orientations to produce a
glare probability index. Even using tools developed to detect luminance contrast in Radiance generated
renderings is time consuming and not feasible for dynamic daylight measurements. Wienold developed a
simplified version of his original DGP evaluation index for use in faster simulations (Wienold 2007).
Simplified DGP (DGPs) follows the principals of DGP except instead of using the luminance of the
scene, it focuses only on the eye level illuminance in the scene. DGP is based on evaluating a
13
combination of the eye level illuminance and the incoming luminous source, angle, and position. Where:
𝐸𝐸 𝑣𝑣 is the vertical illuminance at eye level, 𝐿𝐿 𝑠𝑠 is the luminance of the source, 𝜔𝜔 𝑠𝑠 is the solid angle of the
source, and 𝑃𝑃 is the Guth position index (Wienold and Christoffersen 2005) (Wienold and Christoffersen
2006).
𝐷𝐷 𝐷𝐷 𝑃𝑃 = 5.87 × 10
− 5
× 𝐸𝐸 𝑉𝑉 + 9.18 × 10
2
× log �1 + �
𝐿𝐿 𝑠𝑠 , 𝑖𝑖 2
× 𝜔𝜔 𝑠𝑠 , 𝑖𝑖 𝐸𝐸 𝑣𝑣 1. 8 7
× 𝑃𝑃 𝑖𝑖 2
𝑖𝑖 � + 0.16
Essentially this equation is broken up into two pieces – the first involves eye level illuminance and the
second calculates luminance components from a picture or rendering of the scene. If no direct glare
source is present or obscured by a shading device the second piece of the equation can be replaced with a
properly correlated constant. A new formula based on the exclusion of a direct glare source and only
relying on eye level illuminance values is as follows, where: 𝐸𝐸 𝑣𝑣 is equal to the measured eye level
illuminance during an occupied hour (Wienold 2009):
𝐷𝐷 𝐷𝐷 𝑃𝑃 𝐷𝐷 = 6.22 × 10
− 5
× 𝐸𝐸 𝑣𝑣 + 0.184
The new DGPs method can calculate a glare probability index using a single vertical eye level
illuminance sensor, drastically speeding up the simulation and allowing more frequent measurements to
be recorded.
2.1.4 Shading
Moderating sunlight in a space can be done in a variety of ways – shading devices, frits, blinds, coatings,
etc. (Figure 3). All of these work to control the amount of direct light and solar energy entering into a
room.
14

Figure 3. Exterior Shades, Interior Blinds, Frits, and Glass Types/Coatings (Enclos Corp 2014)
Exterior shading devices obstruct the path of sunlight before it even hits the glass and transmits into the
space. Because much of the glass is largely protected by the shades no solar radiation is absorbed and re-
radiated back into the room, preventing solar gains as well as limiting direct sunlight transmittance.
Shading devices can be specifically tuned for different climates and orientations to allow sunlight into
spaces at certain desirable times. For example, angles louvers can impede the highly angled sun rays
during the summer while allowing the lower angles rays of the winter, where solar gains may be
desirable. However, exterior shades are obtrusive, require additional structure and supports, and may
impede views where other less intrusive remediation may not.
Interior blinds can provide similar benefits of exterior shades by blocking direct sunlight and solar
radiation from hitting interior surfaces. Since interior blinds are behind the glass, solar radiation will have
already been absorbed into the glazing surface and will contribute to solar gains. However, blinds
provide a dynamic user controlled solar management strategy. During the summer interior shades can be
kept down often to reduce sunlight penetration and during the winter can be kept up more often than not
to introduce extra warmth where needed.
15
Frit patterns can provide an obstruction of direct light on the glass surface itself. A ceramic dot pattern is
applied and baked into the glass surface to prevent sunlight from penetrating into the space. The pattern
density can be changed along the surface to increase where more light may come through and decrease
where views are desirable. Although the frit pattern prevents direct light into a space, solar radiation is
nonetheless absorbed and re-radiated.
Rather than obstructing sunlight, certain glass products and coatings can manipulate the way direct light
and radiation are absorbed and reflected. Low emissivity coatings can selectively reflect infrared
radiation while allowing reasonable levels of visual light transmittance. In climates where more solar
gains are a boon, the low-e coating can be moved from surface three of an insulated glass unit (IGU) to
surface two (Figure 4). Light passes through the first layer of glass and is partially reflected from the
second layer to the selectively reflective second surface of the IGU which reflects radiation back into the
glass, trapping more radiation. Glass coatings are one of the easiest solar mitigation solutions because
they can be added to the glass itself and provide considerable solar management for little effort.
However, because the entire glass is coated, not much configuration or optimization can be done.

Figure 4. Coating on Surface 3 for Hot Climates, Coating on Surface 2 for Cold Climates
Glass
Exterior
1 2,3 4
Interior
Glass
Exterior
1 2,3 4
Interior
16
2.2 Analysis Software
There is a wide range of tools available to evaluate and analyze the effects of daylight on building spaces.
Combinations of different programs work together, each with their own specific roles, to assemble a
cohesive simulation package. Several different lighting simulation engines are available and each can be
interchanged with varying user interfaces to link scene geometry, materials, and the environmental
conditions into the engine. In many cases, certain program packages have advantages over the others
when dealing with specific evaluations in terms of accuracy, time, and geometrical complexity.
Evaluating the quality of daylight for this research (measured in a balance of three metrics: solar
radiation, illuminance, and glare) was determined to be most appropriately measured by using a
combination of DAYSIM and Radiance as the lighting simulation engines. DAYSIM is a Radiance-based
simulation engine that can measure the incremental levels of daylight in spaces hourly, daily, monthly,
and annually (Daysim 2013). More specifically, DAYSIM is able to produce climate-based daylighting
metrics such as useful daylight illuminances and daylight autonomy. Radiance, a backwards ray tracing
software, can accurately simulate the effects of natural and artificial light very efficiently (Radiance n.d.).
Additionally, Radiance includes many custom modules which extend its functionality to perform
calculations such as annual irradiation images using a cumulative sky, providing an effective annual solar
radiation metric.
In addition to the lighting simulation engines, 3D modeling software was needed to generate the scene
geometry and environment on which the simulations would be based. Furthermore, a modeling package
that was already used by architects and designers would provide a better transition and integration of their
projects into the workflow. Rhinoceros 3D, a popular surface modeling program, was chosen as the 3D
modeling package (Robert McNeel & Associates 2014). One particularly important aspect of using
Rhino was the abundance of plug-ins available. Just as modules enhance Radiance’s capabilities, plug-ins
for Rhino can add automation and scripting functionality. One such plug-in is Grasshopper, a visual
scripting interface specifically developed to use Rhino (Robert McNeel & Associates 2014). Grasshopper
17
enables parametrically driven changes to be made iteratively and precisely to geometry, environment, and
many other settings.
The link between the simulation engines and the modeling software was another plug-in for Rhino called
DIVA (Solemma LLC 2014). DIVA operates in both Rhino and Grasshopper as a linking interface
between the parametric modeling software and the robust lighting simulation engines. DIVA uses
geometry in Rhino and Grasshopper to setup a scene while calling upon protocols in DAYSIM and
Radiance to simulate the space. The results are then available as visualized data in the model in Rhino or
as recorded data in Grasshopper.
2.2.1 Radiance
Radiance consists of a package of programs that work together to create a comprehensive stand-alone
lighting simulation software. The lighting simulation engine is what Radiance is most known for and is
why Radiance was chosen as one of the two simulation engines. The lighting simulation engine uses a
hybrid approach of Monte Carlo and deterministic ray tracing to accurately compute radiance values
which are used to form a photorealistic image (Ward 1994). This method starts at a point in the scene
(e.g. the viewpoint) and traces rays of light back to the emitting sources. In comparison to other ray
tracing methodologies, the backwards ray tracing is very efficient because it disregards light sources that
ultimately do not influence the starting measurement point. Scene materials, light sources, and physical
geometry are the inputs required for a simulation.
Many robust software packages are based around Radiance due to the novel efficiency afforded by the
backwards ray tracing methodology. Hourly calculations are normally very quick, but when required
over the course of one year they become quite time consuming and computationally taxing. Using
Radiance’s robust simulation engine these calculations can be done in a fraction of the time of prior
software. One climate-based software that takes advantage of this is called DAYSIM.
18
2.2.2 DAYSIM
DAYSIM is a Radiance based daylighting simulation software that can calculate a variety of annual
daylighting based metrics. DAYSIM combines the power of Radiance’s backwards ray tracing engine
and a daylight coefficient approach to generate time based measurements of luminance, illuminance, and
irradiances at virtual sensors (Daysim 2013). These measurements can be used to form detailed climate-
based hourly or sub-hourly measurements like daylight autonomy and useful daylight illuminances.
In addition to daylighting analysis, DAYSIM is able to generate annual profiles for different calculations
such as predicted artificial light usage, dynamic shading system positions, and daylight glare probability.
Through an internal module called Lightswitch, DAYSIM can use annual illuminance profiles and
occupancy schedules to predict occupancy usage of artificial lights. This information predicts the amount
of energy usage from artificial lighting in a space and can provide that data to external energy modeling
software. It can accurately model several dynamic shading systems such as venetian blinds, roller shades,
and even electrochromic glazing. Annual illuminance profiles for each shading system are generated and
joined with the predicted artificial light usage occupancy behavior to check what state the shade will be in
during different times of the year. DAYSIM also utilizes DGP to predict discomfort glare and can be
used generate annual glare probability profiles which inform the dynamic shading system state.
2.2.3 Rhinoceros 3D and Grasshopper
Rhinoceros 3D is a popular NURBS (non-uniform rational b-spline) modeling software (Robert McNeel
& Associates 2014). Originally developed by Robert McNeel and Associates for the purposes of boat
design, the smooth surface modeling capabilities soon became of great interest to architects, engineers,
and designers. Existing as a mathematical freeform modeling software as opposed to a polygonal mesh
modeling program, Rhino can quickly and accurately model any types of objects and surfaces. In
addition to its modeling capabilities, Rhino has excellent functionality to import and convert other 3D
modeling file types, making it universally usable. Rhino also includes a scripting language based on
Visual Basic called Rhinoscript which enables users to enhance the base software with extra tools and
19
functionality called plug-ins. One visual scripting plug-in, developed by David Rutten at Robert McNeel
and Associates, is called Grasshopper and greatly extends Rhino’s capabilities.
Grasshopper is a visual programming language developed specifically for Rhinoceros in 2007 (Robert
McNeel & Associates 2014). Rather than rely on traditional programming syntax to generate code,
Grasshopper uses components as the most basic building blocks to carry out commands. Components use
the most basic Rhino and Rhinoscript commands to perform actions and can be joined together by
dragging a connection between each component to each other in the canvas space. Routines are built in
Grasshopper by stringing together components to form a generative algorithm. Components can create
and manipulate geometry, operate on lists of data, and work with a variety of information synced into
Grasshopper. To expand beyond its original component list, it also includes the ability to custom create
components using native programming languages such as C#, Visual Basic, and Python.
Custom plug-ins and components in Rhino and Grasshopper can provide external resources and enhanced
functionality to the software package. Two plug-ins/components of specific interest are DIVA and
Galapagos.
2.2.4 DIVA
DIVA is a plug-in for Rhino and Grasshopper originally developed by the Graduate School of Design at
Harvard University (Solemma LLC 2014). It functions as an interface between Rhino and Grasshopper
and environmental simulation software – EnergyPlus, Radiance, and DAYSIM. DIVA has extensive
capabilities for advanced daylighting simulation and analysis. Leveraging the power of DAYSIM and
Radiance, DIVA can produce radiation maps, photorealistic renderings, climate-based metrics, and glare
analysis. These simulations can be conducted on geometry existing in Rhino or parametrically linked or
created geometry in Grasshopper. DIVA provides simulation functionality inside 3D modeling software
that previously would need to be handled by external analysis programs. The ability to link DIVA to
Grasshopper allows parametric simulations to be run. Design iterations and model parameter changes can
20
be instantly done and re-simulated, providing quick feedback for designers. This cohesive package of
design and 3D modeling software and integrated simulations is one of the reasons DIVA has created such
a large following.
2.3 Optimization
A typical optimization problem consists of finding an input within a defined domain that minimizes or
maximizes the computed output of a function. Applying that same concept to the field of design involves
finding a set of design considerations which minimizes or maximizes the performance of a design;
performance criteria being defined as what is most important to the project. A simple optimization
problem focuses on providing the best available single outcome from inputs; however gravitating towards
one output is not always the case. Where multiple objectives are considered tradeoffs between outputs
are required to find a working solution. An infinite number of solutions can be produced to satisfy
equally valued multiple objectives, where tradeoffs between conflicting outputs occur continuously.
Subjective grading of outputs can lead the multi-objective optimization problem in a specific direction.
Solving optimization problems can be done using several different methods – algorithms, iteration, and
heuristics. Algorithms work through a finite number of steps to find a solution where possible. They are
tailored and applied to solve specific optimization problems. An iterative method involves generating a
sequence of improving approximate solutions to the problem until a termination criteria or relative
convergence is reached. In contract with a direct method which exhaustively solves for all inputs in the
domain, the iterative method begins with a guess and generates approximate solutions from that
attempting to reach a satisfactory solution. Heuristics are techniques of solving problems quickly where
traditional methods may be too slow or not find an approximate satisfactory solution. The objective of a
heuristic is to find a solution within a reasonable amount of time; this solution may not be the best
possible, but at the time of completion is a “good” approximate solution to the optimization problem.
21
Genetic algorithms (GA) are one type of search heuristic which follows the process of natural selection to
find approximate solutions to a problem. They are under the greater umbrella of evolutionary solving
algorithms. Genetic algorithms work by creating an initial population of inputs. The initial population is
generally a randomized mix of inputs to create a robust design space that the solving starts with. A
proportion of the initial population is selected based on a pre-determined fitness function which ranks
how each scheme is performing. The selected configurations are then mutated and bred with each other
to form a second generation population. Two selected schemes are mixed together to breed a child which
takes input traits from its parents in hopes of generating an even better configuration. Some portion of the
selected population not bred together is randomly mutated, changing the input clusters attempting to
create a more optimal scheme. This process is stopped when a termination condition is met, whether it be
a fixed number of generations has been reached, a solution meets a minimum fitness criteria, relative
convergence has been found where successive iterations no longer achieve substantial change in fitness of
the populations, or a decision maker subjectively ends it.
2.3.1 Galapagos
Galapagos is a native solving component inside Grasshopper that provides optimization support (Robert
McNeel & Associates 2014). It offers a single objective simulated annealing (SA) or genetic algorithmic
(GA) approach for solving optimization problems. Galapagos can be linked to multiple inputs and sliders
inside Grasshopper to iterate and control geometry and data as part of an optimization exercise.
Furthermore, Galapagos stores data and allows the user to visualize the solving process as it searches for a
solution through its SA or GA methods. This data can be retrieved to understand the solution solving
process and visualize iterative changes made to geometry linked to Galapagos.
2.4 Sensitivity Analysis
Sensitivity analysis is the study of how the uncertainties of inputs in a mathematical or simulated model
affect the uncertainty of outputs in that same model. Essentially, how sensitive measured outcomes are to
22
each parameter or variable. Sensitivity analysis can be applied to many scenarios to determine how
inputs and outputs are related to each other, and by how much.
Given the scenario of providing a design process to guide and influence designers on developing shading
system configurations, it is important to relate how each parameter that is configuring the system affects
the overall performance of the system.
There are many different methods to conduct sensitivity analysis, such as simple one-at-a-time methods,
scatter plots, and regression analysis or more complex variance-based methods.
One of the simplest types of sensitivity analysis is the one-at-a-time method. By changing the range on a
single input variable while keeping the rest of the inputs at a constant baseline, the output can be
monitored as the single variable is varying. Since each variable is varied while the rest are at a constant
baseline, the effects on the outcome can easily be compared to each other. One flaw with the one-at-a-
time method is it doesn’t take into consideration the simultaneous adjustment of multiple inputs and their
interactions, leaving out potentially important connections between input variables.
Scatter plots can indicate the relationship between inputs and outputs by comparing them against each
other on separate axes. Each input variable is compared to the output by plotting the data set on opposing
axes. A visual representation of the correlation between the input and the output can be seen through the
scatter plot. Additionally, trend lines, using regression analysis can be overlaid on the plots, measuring
the level of correlation between the inputs and outputs. Simple linear regression models can be used to
indicate the level of significance that input variables have on the outcome.
Variance-based methods work by quantifying the uncertainty of the inputs and outputs as probability
distributions. The output variance is then decomposed and the contributing variances can be traced to
particular inputs. The sensitivity of input variables to the outputs can be defined as the amount of output
variance caused by the input variables. Variance-based methods, while computationally taxing, allow for
full exploration of the interactions of inputs.
23
CHAPTER THREE – METHODOLOGY

3.1 Reference Building

Figure 5. Renderings and the Rhino Model of the Wilshire Grand Tower
In order to maintain a realistic scenario for testing, a developing building design was chosen to serve as
the reference building. More importantly, a project with a high amount of surface area exposed to the sun
and different programmatic spaces housed inside the building was especially desirable (e.g. a skyscraper).
AC Martin, a well-known Los Angeles architecture firm, graciously allowed their developing Wilshire
Grand high rise project to be used. At the time the firm was already looking for outside consulting for the
design of the sun shading systems on the southwestern façade.
The Wilshire Grand tower encompasses a city block in downtown Los Angeles that once was the site of
the former Wilshire Grand hotel. It is bordered by Figueroa Street, 7th Street, Francisco Street, and
Wilshire Boulevard. Situated in the heart of downtown Los Angeles, the Wilshire Grand tower is
surrounded on all sides by tall buildings, which throughout the year restrict light access and shade
Wilshire Grand’s expansive façade. The 70 story tower is oriented southwest and features a tapering and
curving southern façade. The north façade is flat, with the east and west sides tapering as the reach the
top. Wilshire Grand tops out at 1100 feet, making it the tallest building on the western United States. A
24
three story podium houses hotel amenities and retail spaces, while the lower portion of the tower consists
of both creative (open plan) and standard office layouts (closed offices). The upper two thirds of the
building is used for hotel rooms with the top including a sky lobby, restaurant, and pool area. The
Wilshire Grand design featured as much floor to ceiling glass as allowed by code throughout the tower.
Office spaces and hotel rooms would have a combination of translucent and clear panels in order to reach
the allowable glazing area.
A shading system consisting of metal louvers and fins attached to the tower’s unitized curtain wall was
being developed by the architect. The system was mainly designed for the curving southwestern façade
to prevent excessive glare and solar radiation, but was also to be used partially on the east and west sides.
The number, placement, angle, and depth of the louvers and fins were under development; these
parameters and context would serve as the foundation for the following research.
3.2 Software Platform
Rhinoceros 4.0 SR9 with the Grasshopper 0.9.0014 plug-in was used as the main software platform. Two
external and one internal plug-ins for Grasshopper were used as the simulating engine, optimization
solver, and data exporter. DIVA 2.0.1.0 with component version 2.0.0.6 functioned as the simulation
engine, calling upon DAYSIM and Radiance to perform the climate-based illuminance and radiation
calculations. Galapagos was used as the single-objective optimization solver, using its built-in genetic
algorithm. Lunchbox 0.35, a multi-functional plug-in developed by Nathan Miller, enabled Grasshopper
to collect and export data into Excel sheets as well as read back Excel data for re-instantiation (Miller
2014).
Rhino and Grasshopper were used as the main software platform because of their flexible visual
programming interface, parametric capabilities, and ease of use. One of the most appealing aspects of
Grasshopper is its shallow learning curve; everything is graphically laid out as opposed to typing lines of
code, making for a much easier and faster user experience. As a visual programming interface,
25
Grasshopper allows users with little to no coding experience to compile complex sequences. A completed
Grasshopper sequence of components is known as a definition. Grasshopper has all the capabilities of
more traditional code-based software with the inclusion of Visual Basic, C#, or Python components built
inside. Most importantly, because of the ease of use, Rhino and Grasshopper have found their way into
most design schools. In turn, recent graduates and young professionals have begun integrating Rhino and
Grasshopper into the professional environment. The prevalence in education of the software as well as
professional use makes Rhino and Grasshopper an excellent choice versus other software of similar
capabilities.
3.3 Parametric Definition
The process definition built in Grasshopper consists of six different pieces (Figure 6):
• Inputs: geometry, materials, directional vectors, sensor placement, and climate data
• Parameters: shading system configuration parameters
• Measurements: daylight performance measurements (solar radiation, useful daylight
illuminances, simplified daylight glare probability)
• Optimization: evolutionary solver and metric weighting factors
• Data: collecting, sorting, and exporting completed simulation data
• Re-instantiation: visualizing and displaying simulation results
26

Figure 6. Overall Grasshopper Parametric Definition
27
The results of the six pieces are a compiled data set of shading system configurations, the corresponding
performance measurements, and a visual representation of each system setup. 10 scenarios were tested
using the process, comparing different typologies, timeframes, orientations, and weighting factors during
the optimization evaluation. Each comparison was done to examine how the optimum shading system
parameters would change as the simulation conditions changed. The sensitivity of each parameter would
correspondingly change as the conditions for the simulations were altered; highlighting how each
parameter reacts to particular changes.
Through the architects’ design intentions and the solar exposure of the southwestern façade, it was
deemed to be the main area of focus for the simulations. All testing was conducted with the room facing
south, unless otherwise noted. All simulations were conducted over the course of a year (January 1st to
December 31st); concentrating on testing the effectiveness of the static shading system year round rather
than on any particular dates or timeframes. Furthermore, all performance metrics were valued as equal by
default, with the ability to change the significance of each metric as needed.
The preliminary/baseline scenario analyzed the open office space, facing south and throughout an entire
year. The hotel and closed office were analyzed under the same conditions to compare the typologies
against each other. The comparison of the three spaces examines how the shading system configures to
adapt to the different sizes and occupied times of each room.
The baseline scenario was then compared against two different timeframes – one month during the winter
(December 21st to January 21st) and one month during the summer (June 21st to July 21st). Narrowing
the simulation time examines how the shading system configuration might be optimized for those
particular times of the year in contrast to an annual simulation.
To examine the effect orientation has on the shading system and its corresponding parameters an east
facing office were simulated. An identical size and sensor layout open plan office was used to simulate
the eastern facing office. Both scenarios were conducted over the entire year.
28
Lastly, the baseline was compared against two sets of scenarios with varying performance valuing
(weighting) during the optimization evaluation sequence. The baseline simulation equally valued each
performance metric (solar radiation, UDI, DGPs). A set of simulations, changing the significance of UDI
by a factor of 2 and 3, respectively, were conducted to investigate the shading system configuration and
parameter response by altering which performance metrics were deemed more important. The second
comparison altered the value of the modified DGPs metric by factors of 2 and 3 also.
Each part of the process definition could be changed or disabled by the user. The entire process was kept
open source to allow for customization of any inputs, parameters, equations, or weighting factors.
3.4 Inputs
The user inputs for the definition consist of the selecting the rooms that are to be analyzed, the
surrounding context buildings, and appropriate climate data. Localized axes are derived from the room
geometries to address directionality inside the Grasshopper interface. Sensors for measuring solar
radiation, daylighting, and glare use the local axes to find their positions inside the rooms. All the inputs
are designed to easily integrate any project for analysis, requiring only the minimal user setup before
getting started.
3.4.1 Geometry
The full model of the reference building and surrounding blocks, provided by the architect, was used as
the basis of the inputs. Typical floor plans, as indicated by the architect, were added at the appropriate
floor levels into the model. Although the whole building model would not be used for all the simulations,
it was still valuable to have visual context for where each room would be located in the building and the
overall shape of the tower’s façade.
Three rooms were digitally modeled from the typical plans inserted into the building – a hotel room on
the 35
th
floor and both an open plan office and standard enclosed office on the 15th floor (Figure 7). The
hotel was modeled as a 14’-4” by 28’-6” room with 9’-8” ceilings and a small notch in the back,
29
indicating the enclosed bathroom area. The open office was modeled as a large 30’ by 30’ foot
unobstructed space with 10’ ceilings. The office area extends all the way to the service core of the
building, 30 feet away from the glass façade. An additional open office was modeled on the eastern
façade for the multiple orientation comparison scenarios. The closed office was modeled around the
perimeter of the glass façade as a small 10’-3” by 15’ room with 10’ ceilings. All rooms were located in
the center of their respective floors.

Figure 7. Counterclockwise: Hotel, Open Office, Closed Office, and East Facing Open Office
The hotel room was chosen to represent a medium sized space where it would be occupied during the
mornings, evenings, and throughout the entire week (8am to 6pm, seven days a week). In addition to the
specific occupancy times of the hotel room were the unique solar concerns. Glare avoidance and
30
daylighting reaching the back of the room was less important compared to blocking solar radiation and
preserving views outside the windows.
The open office is the largest space of the three; an open plan layout allows light to reach to the edge of
the space. The open office is occupied according to a standard work day, during weekdays (9am to 5pm,
with a 1 hour lunch break from 12pm to 1pm). Important performance criteria for the open office were to
maintain appropriate horizontal task levels of illuminance and focus on natural daylight.
The closed office represents a standard perimeter office, a small room enclosed on all sides, which would
be occupied according to a standard work day, during weekdays (9am to 5pm, with a 1 hour lunch break
from 12pm to 1pm). Given the proximity of a person’s work area to the window and the small volume of
space, the minimization of glare and solar radiation are of high concern whereas proper daylight
illumination is already achieved because of the shorter distance to the window.
Each room was broken down into components, based on their respective materials – ceilings, floors,
walls, and glass (Figure 8). In addition to the individual room components, the surrounding context
buildings in a three block radius were accounted for, along with the ground surface. Each of these
components were assigned a material that carefully matches the actual reflectivity of the material. In
simulating an average setting within the rooms, the surrounding buildings, and the street, a set of baseline
materials was used based on daylight simulation setup guidelines from daylight simulation researcher
Christoph Reinhart and other sources (Reinhart 2011).
31
Table 1. Default Material Properties
Geometrical Component Visual Properties
Interior floor 20% diffuse reflectance
Interior wall 50% diffuse reflectance
Interior ceiling 80% diffuse reflectance
Single glazing 90% direct visual

Exterior building surfaces 44% diffuse reflectance
Exterior ground 20% diffuse reflectance

Radiance interprets materials based on a text file it calls upon during the simulations. The text file
dictates the optical properties of materials used in the scene. Plastics, metals, mirrors, glass, etc. are
described by varying optical properties. For example, plastic has five variables which describe: its red
green blue reflectance, fraction of specularity, and roughness value. Glass has three variables describing
the RGB transmissivity at normal incidence. The ability to create custom materials with specific diffuse
reflectance or visual transmittances is easily done by editing the material.rad text document which DIVA
reads. For the purposes of this research, the aforementioned baseline materials were constructed for use
by Radiance as well as 10 different types of glass with varying transmissivity.

Figure 8. Hotel Room Materials - Glass, Ceiling, Walls, and Floor
32
3.4.2 Vectors and Sensors
In addition to the material selection of the geometrical components, each room needed to include a set of
three vectors that indicate directionality within the space (Figure 9). Three lines were drawn to represent
each direction, or in the context of the analyzed space – horizontal, vertical, and towards interior
directions. These vectors were unique to each room and needed to be drawn up for each space. With
these vectors in place, Grasshopper knew which direction was up and down, where to place daylight and
irradiation sensors, and how to manipulate the shading system parameters.
Two illuminance sensors and one grid of irradiation sensors were placed throughout each room. The
illuminance sensors were divided between one measuring horizontal task illuminance and the other
measuring eye level vertical illuminance. The horizontal sensor was placed at two thirds the room
distance towards the back and 30 inches off the ground, simulating a standard table or desk height. The
position of the horizontal sensor was based on attempting to achieve desirable illuminance levels deeper
into the space. The vertical illuminance sensor was placed at one third of the room, closer to the front and
nearest to the glass. The sensor faced the glass at a height of 48 inches off the ground, picking up light
that would potentially be entering the eyes of a seated person at their desk. Lastly, solar radiation was
measured via a grid of sensors that were strategically placed one inch beyond the glass (towards the
interior). The array of 81 sensors picked the solar radiation that had immediately passed through the glass
before diffusing deeper into the space. The number and position of all the illuminance and irradiation
sensors can be changed by the user, the default settings were made to emulate a baseline scenario.
33

Figure 9. Parameters and Sensors in the Hotel Room
3.4.3 Weather Data
Beyond the physical inputs required to get the process started was the location-based climate data for the
simulations. Weather data, also called typical meteorological year (TMY), is a collation of recorded
meteorological phenomena for a specific location gathered over the course of many years. It presents the
average weather over the course of one year averaged from decades of weather. TMY includes
temperature, precipitation, solar radiation values, and other weather-related data recorded in hourly
increments. This information was used to generate average environmental conditions during simulations.
The referenced building was located in downtown Los Angeles. Weather data was acquired from a
nearby weather station for the central Los Angeles area, just south of downtown. While DIVA normally
allows the inclusion of non-default climate data into its simulations, the central Los Angeles weather data
was unable to be used. The DIVA engine would lock up after the proper weather file was selected and
would stop responding. Ultimately, the standard Los Angeles International Airport weather station data
was used, a standard file included with DIVA.
34
3.5 Parameters
Nine parameters were developed to control the shading system configuration. The parameters were
designed to include a large range of variation and be as flexible as possible to avoid minimizing the
exploration space of the optimization routine. The parameters focus on glazing properties, the number
and physical properties of louvers and fins, and the position of the entire system from the building face.
Table 2. Parameter Values Range
Parameter Min. Value Max. Value Step
Glazing Type 0 10 1
Number of Louvers 1 10 1
Louver Length 1” 24” 1”
Louver Angle -45° 45° 3°
Number of Fins 2 10 1
Fin Length 1” 24” 1”
Fin Angle -45° 45° 3°
System H Offset 0” 24” 2”
System V Offset -48” 48” 6”

The glazing type parameter controls a series of 11 different types of glass with varying visual
transmittance levels; starting at 21% visual transmittance and moving to 91% with a 7% step. However,
for Radiance to understand and interpret the optical properties of each glass makeup, the visual
transmittance needs to be converted to transmissivity. The particular equation for converting visual
transmittance to transmissivity was extracted from (Radiance n.d.). Where: 𝑇𝑇 = transmissivity and 𝑉𝑉 𝑡𝑡 =
visual transmittance.
𝑇𝑇 =
� 0.8402528435 + 0.0072522239 𝑉𝑉 𝑡𝑡 2
− 0.9166530661
0.0036261119
𝑉𝑉 𝑡𝑡
35
With the values converted, they could be added into the material.rad file for Radiance and DIVA to
interpret all the materials in the Grasshopper definition.
Table 3. Glazing Type Parameter Properties
Glazing Type Visual

Transmissivity
0 21% 22.91%
1 28% 30.54%
2 35% 38.17%
3 42% 45.8%
4 49% 54.43%
5 56% 61.05%
6 63% 68.67%
7 70% 76.28%
8 77% 83.89%
9 84% 91.5%
10 91% 99.1%

The original intention of the glass type parameter was to allow the optimization process to have the ability
to control the amount of light transmitting through the glass in conjunction with the shading system. It
was thought that a non-intuitive configuration between glass type and shading system parameters might
come about. Unfortunately, the glazing type variable proved to be a dominate trait and the optimization
would always pick type 10, the glazing type with the highest levels of visual transmittance. In addition to
the dominance recorded by the glazing type parameter, it was discovered that Radiance, and therefore
GenCumulativeSky does not take into consideration any physical optical properties of glass when it
performs simulations. Only transmissivity of the glass was utilized during the simulations. Thus, despite
the glass parameter changing the glass types, the amount of radiation passing through the glass would
never change. Between the trait dominance and dysfunction with solar radiation measurements this
parameter was ultimately fixed at 91% visual transmittance and would not influence the outcomes beyond
that.
36
Three parameters control the number, length, and angle of the louvers on the shading system. Similarly to
the louver parameters, the number, length, and angle of fins were provided as three more parameters. A
wide range of parameter values were allowed in the physical parameters to allow the optimization
sequence to adequately explore the design space. Like any other piece of the 3D model being simulated,
the shading system needed material properties. The shading system (louvers and fins) has a diffuse
reflectance of 0.52 and emulates a matte silver metal type material.
Two parameters controlled vertical and horizontal system offsets from the face of the building. The
horizontal offset parameter could move the entire shading system off the building façade 24 inches in 2
inch increments. While the vertical offset parameter could shift the shading system up and down 48
inches in each direction in 6 inch intervals. Both parameters exist to give the shading system range of
motion to allow additional ambient or direct light into the space as needed.
3.6 Measurements
There were two phases in which the simulations are conducted – a preliminary run that simulated and
stored data for the optimization equation performed later and the main run, which ran continuously and
recorded all data for exporting (Figure 10). The preliminary set attempted to build a maximum scenario
based on the given inputs and parameters; for example introducing the maximum amount of radiation,
highest percentage of UDI, and highest DGPs. To achieve the maximum scenario, the shading system
was turned off allowing light and radiation to enter the space and hit the sensors without any obstructions.
The preliminary run was only done once at the beginning of each simulation sequence. The main run
included all scene geometry with materials applied, contextual surrounding geometry, and shading device
geometry. The main simulation was connected to Galapagos, which continuously ran simulations as it
attempted to find an optimum solution.
37

Figure 10. Workflow Diagram
Within the preliminary and main phase runs two different types of simulations were run; one measuring
solar radiation and the other measuring illuminance – with a horizontal sensor at the work plane and a
vertical sensor eye level. Each simulation component required four things – (1) geometrical components
of the room to be measured with the proper DIVA materials applied, (2) the location of each analysis
node, (3) the vector direction of each analysis node, and (4) a switch was used to control the simulation
from running. The switch is used to control the different phases of simulations, as well as disabling
simulations from running as changes are made to the geometry, parameters, analysis nodes, etc.
・ Project Information
・ Project Geometry
・Sensor Setup
・Parameters
Occupancy Schedules
Metric Weighting
Pre-Simulation
Main Simulation
Fitness Equation
DIVA
MODEL
INPUTS
OUTPUTS
RESULTS
Data/Graphs: Excel
38
The DIVA daylight component was able to call upon the Radiance backwards ray tracing software as well
as coded modules inside of it and DAYSIM to perform climate based simulations. The daylight
component included changing the simulation timeframe, specific measurement parameters, and Radiance
scene parameters.
3.6.1 Solar Radiation
Radiation was measured using the GenCumulativeSky module inside Radiance which created a
continuous cumulative sky radiance distribution. Radiance then used this cumulative sky in a backwards
ray-tracing simulation to calculate the radiation at each sensor. The DIVA daylight component included
the ability to change the run period for the simulation from a series of built-in ones (annual, extreme
weeks, and extreme days) to custom, where the range could be manually inputted (Figure 11). The three
sets of conducted scenarios were – annual (Jan 1 – Dec 31), summer (Jun 21 – Jul 31), and winter (Dec 21
– Jan 21). Annual was already a preset in DIVA while the summer and winter simulations were setup by
manually changing the start and end dates.

Figure 11. DIVA Daylight Component - Solar Irradiation Settings
39
A grid of analysis nodes placed just slightly past the exterior glass façade (1 inch) was used. The sensor
array measured the amount of radiation that got past the external shading system and entered and heat up
the room. The irradiation at each node was averaged together to produce a single number that was
representative of the radiation let through by the shading system.
3.6.2 Useful Daylight Illuminances
Illuminance was measured using the climate based DAYSIM to generate hourly measurements over the
course of a year. Between the two illuminance sensors, the horizontal sensor placed at task height was
used to calculate UDI. The illuminance data collected at the horizontal sensor was compared to the
selected occupancy schedule and the occupied hours of data were pulled out. DIVA includes climate
based simulations as one of their standard options in the daylight component. Inside the climate based
selection are parameters for selecting an occupancy schedule, setting a minimum illuminance threshold,
and electric lighting controls and parameters (Figure 12). The occupancy schedules used in the DIVA
daylight component can be selected from a dropdown menu of pre-made choices that come with standard
with DIVA. Additional custom occupancy schedules can be created by editing the comma separated
value occupancy schedule files.
40

Figure 12. DIVA Daylight Component - Climate Based Settings
Each occupancy schedule file was made up of a short description of the overall schedule (e.g. weekdays 9
to 5, one hour lunch break, daylight savings time lasts from 2nd Sunday of March to the first Sunday of
November) and a list of 8760 hours starting from January 1st at 12:30 am to December 31st at 11:30 pm,
with January 1st considered a Monday. Following each hour is either a 0 or 1, representing if the hour is
occupied or unoccupied.
It was important to match the run period of the UDI simulations with the radiation simulations. While the
radiation DIVA component has the option to adjust the run period, the daylight autonomy setting does not
and instead requires a matching occupancy schedule. Given the three simulations being conducted, three
occupancy schedules for annual, summer, and winter were necessary. Additionally, each typology
required a separate set of occupancy schedules. The hotel was set to be occupied 7 days a week from 8
am to 6 pm with no lunch break and daylight savings set from the 2nd Sunday of March to the first
Sunday of November. The open and closed offices were set to be occupied 5 days a week from 8 am to 5
41
pm with a one hour lunch break and daylight savings set from the 2nd Sunday of March to the first
Sunday of November.
The output of the DIVA daylight component set to climate based simulations yielded seven different
options: daylight autonomy, continuous daylight autonomy, daylight availability, UDI from 100 to 2000
lux, UDI less than 100 lux, UDI of greater than 2000 lux, and the lighting load schedule file path (Figure
13). While DIVA includes a large variety of fixed outputs, the ability to change the minimum and
maximum thresholds for UDI was not present. Additionally, all output values were rounded to the
nearest whole number, which is problematic when sending these values to be evaluated through the
optimization sequence. Completely different shading system parameter arrangements could have yielded
similar UDI values, but when rounded they became the same, despite one configuration being superior to
the other. To overcome these limitations in the native climate-based outputs, a custom UDI sequence was
developed.

Figure 13. DIVA Daylight Component - Climate Based Outputs
42
The front-end of the original illuminance simulations worked fine; illuminance at each both of the
horizontal and vertical sensors is measured at each hour of the year. The data was stored among the
DIVA system files in a comma separated value file. Grasshopper was able to link both the recorded
illuminance measurements and any of the occupancy schedules. Similar to how the DIVA daylight
component functions, the illuminance measurements are evaluated based on the selected occupancy
schedule, with the occupied measurements being pulled out. Minimum and maximum thresholds are able
to be set for the UDI calculation through the use of conditional statements. Where: ℎ is equal to the
number of occupied hours, 𝜌𝜌 is equal to the evaluated illuminance measurements, 𝛼𝛼 is equal to the
measured horizontal illuminance during an occupied hour, and 𝑈𝑈 is the percentage of occupied hours
where the measured illuminance meets the minimum and maximum thresholds (UDI).
𝑈𝑈 =
∑ 𝜌𝜌 ℎ
𝑖𝑖 = 0
ℎ

𝜌𝜌 𝑖𝑖 = �
1, 𝑖𝑖𝑖𝑖 300 ≤ 𝛼𝛼 𝑖𝑖 < 2000
0, 𝑒𝑒 𝑒𝑒 𝐷𝐷 𝑒𝑒
Given the minimum and maximum illuminance thresholds, the occupied hourly illuminance
measurements are evaluated. The summation of the evaluated illuminance data divided by the total
number of occupied hours yields the percentage of the simulated timeframe that daylight alone is able to
usefully light the task surface. Additionally, the measured UDI percentage is in the appropriate
significant figures now.
3.6.3 Periodic Glare Evaluation
Glare was evaluated using the same climate-based illuminance measurements used for useful daylight
illuminances. The vertical eye level sensor illuminance data was compared to the corresponding
occupancy schedule and the occupied hourly measurements were pulled out. These illuminance numbers
were then evaluated by using DGPs. The following equation calculates DGPs, where: 𝐸𝐸 𝑣𝑣 is equal to the
measured eye level illuminance during an occupied hour.
43
𝐷𝐷 𝐷𝐷 𝑃𝑃 𝐷𝐷 = 6.22 × 10
− 5
× 𝐸𝐸 𝑣𝑣 + 0.184
The DGPs calculation yields a unit-less value that represents the probability of dissatisfaction based on
glare within the given space at the snapshot of measurement (Wienold 2007).
Due to the methodology of the optimization performed later, each performance metric (radiation, daylight
autonomy, and glare) could only be represented only as a single number. Therefore, it was required to
interpolate the large sample of results into a single number with a second modified equation. Wienold
introduced several numerical brackets as a way of interpreting what the DGP values meant (Wienold
2009).
Table 4. Daylight Glare Probability Brackets
DGP Brackets
≤ 0.35 imperceptible
≤ 0.40 perceptible
≤ 0.45 disturbing
> 0.45 intolerable

To define up to a years’ worth of DGPs values, while still maintaining a means of quantifying the quality
of glare of each reading, a modified series of evaluations in conjunction with the original DGPs was used.
Where: ℎ is equal to the number of occupied hours, 𝑦𝑦 is equal to the evaluated DGPs measurements, and
𝐷𝐷 is representative of the periodic quality of glare measured in the space.
𝐷𝐷 =
∑ 𝑦𝑦 𝑖𝑖 ℎ
𝑖𝑖 = 0
ℎ

𝑦𝑦 𝑖𝑖 = �
0, 𝑖𝑖𝑖𝑖 𝐷𝐷 𝐷𝐷𝑃𝑃𝐷𝐷 𝑖𝑖 ≤ 0.35
1, 𝑖𝑖𝑖𝑖 𝐷𝐷 𝐷𝐷𝑃𝑃𝐷𝐷 𝑖𝑖 > 0.45
( 𝐷𝐷 𝐷𝐷 𝑃𝑃 𝐷𝐷 𝑖𝑖 − 35)/0.1, 𝑖𝑖𝑖𝑖 0.35 < 𝐷𝐷 𝐷𝐷 𝑃𝑃 𝐷𝐷 𝑖𝑖 ≤ 0.45

44
This evaluation takes into consideration the quality of glare between imperceptible and intolerable,
assigning it a normalized value between 0 and 1. For example, if:
𝐷𝐷 𝐷𝐷 𝑃𝑃 𝐷𝐷 𝑖𝑖 = 0.4
𝑦𝑦 𝑖𝑖 = 0.5
𝐷𝐷 𝐷𝐷𝑃𝑃𝐷𝐷 𝑖𝑖 = 0.38
𝑦𝑦 𝑖𝑖 = 0.3
The DGPs measurements were summed and divided by the number of measured hours to produce a
number that represents the periodic quality of glare in the measured space.
3.7 Optimization
Once the preliminary simulation and an initial main simulation had been completed successfully, the
process was calibrated and ready to begin optimizing different configurations. Galapagos functioned as
the solving component. All parameters were plugged into the Galapagos to be flexed during the solving
process. One of Galapagos’ limitations is its ability to only solve one objective at a time. The objective
is achieved by minimizing or maximizing a single fitness number. While native Galapagos is limited to
single objective optimizations, multiple objectives can be solved for, but requires a separate precursor
process; an equation to condense multiple objectives into one.
The condensed fitness equation can expand the number of objectives beyond just one by introducing
tradeoffs between each objective. Where each objective is ranked against each other, competing to find
an optimal solution that satisfies each metric equally. The fitness equation places a weighting coefficient
on each metric to induce tradeoffs between the metrics. If solar radiation is deemed of higher importance
to the design compared to glare and illuminance, the tradeoffs between the three metrics will reflect a
direction towards supporting the solar radiation metric. These weighting factors can help to guide the
simulation results towards the designer’s intent or planned use of the space.
45
The condensed fitness objective can be defined as a function of the sum of the each objective
(measurement). Each metric is either positive or negative based on if the goal is to minimize or maximize
that particular metric. In this case, a combination of minimizing solar radiation (negative), maximizing
UDI (positive), and minimizing glare (negative) is applied to the new pseudo multi-objective fitness
equation. Additionally, to ensure each metric is equally measured against each other, the measurements
going into the equation are normalized to a 0 to 1 scale. The maximum conditions measured and stored
from the preliminary run are fed into the normalizing process.
The equation follows, where:
𝐹𝐹 = the combined fitness objective
𝑈𝑈 𝑤𝑤 = the weighting factor for the UDI measurement (default is 1)
𝑈𝑈 𝑖𝑖 = the UDI measurement
𝑈𝑈 𝑚𝑚 𝑖𝑖 𝑚𝑚 = the minimum range for UDI measurements within the given scene
𝑈𝑈 𝑚𝑚 𝑚𝑚𝑚𝑚
= the maximum range for UDI measurements within the given scene, taken from the
preliminary simulation
𝑅𝑅 𝑤𝑤 = the weighting factor for the radiation measurement (default is 1)
𝑅𝑅 𝑖𝑖 = the radiation measurement
𝑅𝑅 𝑚𝑚 𝑖𝑖 𝑚𝑚 = the minimum range for radiation measurements within the given scene
𝑅𝑅 𝑚𝑚 𝑚𝑚𝑚𝑚
= the maximum range for radiation measurements within the given scene, taken from the
preliminary simulation
𝐷𝐷 𝑤𝑤 = the weighting factor for the periodic glare measurement (default is 1)
𝐷𝐷 𝑖𝑖 = the glare measurement
46
𝐷𝐷 𝑚𝑚 𝑖𝑖 𝑚𝑚 = the minimum range for glare measurements within the given scene
𝐷𝐷 𝑚𝑚𝑚𝑚𝑚𝑚
= the maximum range for glare measurements within the given scene, taken from the
preliminary simulation
𝐹𝐹 𝑖𝑖 = ( 𝑈𝑈 𝑤𝑤 ∗ (( 𝑈𝑈 𝑖𝑖 − 𝑈𝑈 𝑚𝑚 𝑖𝑖 𝑚𝑚 )/ 𝑈𝑈 𝑚𝑚 𝑚𝑚𝑚𝑚
)) − ( 𝑅𝑅 𝑤𝑤 ∗ (( 𝑅𝑅 𝑖𝑖 − 𝑅𝑅 𝑚𝑚 𝑖𝑖 𝑚𝑚 )/ 𝑅𝑅 𝑚𝑚 𝑚𝑚𝑚𝑚
)) − ( 𝐷𝐷 𝑤𝑤 ∗ (( 𝐷𝐷 𝑖𝑖 − 𝐷𝐷 𝑚𝑚 𝑖𝑖 𝑚𝑚 )/ 𝐷𝐷 𝑚𝑚𝑚𝑚𝑚𝑚 ))
Once each measurement is weighted, normalized, and added together they form a singular fitness
objective that can be connected to Galapagos.
Galapagos includes a simulated annealing and genetic algorithm as its solvers; the genetic algorithm was
chosen for the optimization processing. The genetic algorithm provides a flexible and iterative solving
process that can tackle many parameters at once without strain. The customization of genetic algorithms
is apparent in Galapagos with settings available to set the initial population, following generations,
percentage of mutations, breeding, and termination constraints (Figure 14). The fitness number can be set
to either search for the minimum or maximum fitness value. A runtime termination constraint can be
enabled to stop the solver after an inputted period of time (although this setting did not work in my
experience). Each batch of simulations was run through Galapagos for approximately 25 generations, or
1300 simulations total (24 generations of 50 plus the initial population of 100). A reasonable minimum
number of simulations was required to achieve a good sample size to find global optimums, as well as
decipher parameter trends within the data (Figure 15).
47

Figure 14. Galapagos Component Settings
48

Figure 15. Galapagos Solving
3.8 Data
During the course of the optimization process, all data was continuously recorded before being exported
to Excel for analysis and graphical representation (Figure 16). Each parameter setting, all three daylight
metrics, and the overall fitness number were recorded. Once the optimization process had completed an
appropriate number of generations or relative convergence was reached, the solver ended and the data
stopped being recorded.
49

Figure 16. Recorded and Exported Data into Excel
Due to limitations with the recording speed and calculations inside Grasshopper, partial duplicate records
of some of the simulation runs occurred throughout the sequence. Hiccups in the software caused a
partial record of a simulation to be stored with a second record immediately following with the same
parameters but different metrics and fitness values (Figure 17). Grasshopper stored the values of the
parameters immediately, but the calculations of the metrics and fitness value had not finished yet, and
therefore pieced together the current parameter settings with the previous runs’ metrics and fitness value.
The second record was correct, with the parameter settings and calculated metrics and fitness values
matching. These hiccups created false records with information that was not entirely correct, skewing the
results.

Figure 17. Three Sets of Duplicate Results in Excel
50
A short definition was written inside Grasshopper to take the recorded data and sort out all the duplicates
and remove the first ones (the pieced together parameters with the previous runs’ metrics and fitness
values) and only keep the correct second ones.
Once the data had been cleaned up it was exported to a pre-formatted Excel template. The template had
color coding for the fitness values to indicate the more desirable shading system configurations.
Additionally, the Excel template included a LINEST function (calculates the statistics for a line by using
the “least squares” method to calculate a straight line that best fits the data) that performs linear
regression on the parameters (x-axis) versus the fitness value (y-axis) to indicate which parameters are
considered statistically significant (Figure 18). If the calculated P-value for a parameter is below 0.01 it
indicated that particular parameter was highly unlikely to have affected a fitness change by coincidence
and was therefore considered significant to the fitness outcome value. Preset conditional formatting
existed where the P-values of the LINEST function were stored. A statistically significant value (<0.01)
would be displayed in green text with a green cell background and insignificant values (>0.01) would
have no highlighting.

Figure 18. Linear Regression Significance Indicators for Parameters
A second sheet in the Excel template workbook had eight scatter plots prepared to display each of the
parameters plotted against the fitness value (Figure 19). The scatter plots automatically filled themselves
in as the data was streamed into the template document.
51

Figure 19. Eight scatter plots with polynomial trend lines for each parameter
3.9 Re-instantiation
A post-analysis ability of the process is to re-instantiate data back into the definition and visually display
different shading system configurations (Figure 20). This was done by re-reading the Excel templates
back into Grasshopper using the same component which wrote the data into Excel originally. The user
could then choose which configuration they want to see on the 3D model and it would parse the Excel
document and configure the shading system parameter values accordingly.
Beyond just showing the geometry created from those specific parameter values, the re-instantiation
sequence color codes the shading system based on its fitness value. Dark green being the highest
performing of the batch and yellow to red as underperforming. The parameter values, solar
measurements, fitness value, and run number are also displayed next to the model to provide all the
relevant information back to the designer in one comprehensive visual.
52

Figure 20. Re-instantiation of Simulation #703's Configuration Back to Rhino

53
CHAPTER FOUR - RESULTS

4.1 Process Results
As outlined in the methodology, 10 unique scenarios were simulated, optimized, and recorded for review.
• Typologies: hotel, open office, and closed office
• Orientations: south and east
• Timeframes: annual, winter, and summer
• Weighting: 1x, 2x, 3x weighting of UDI and 1x, 2x, 3x weighting of DGPs
These scenarios resulted in over 10,000 simulations being conducted; testing the different shading system
configurations’ effectiveness against measured amounts of solar radiation, daylighting, and glare
probability. Due to the sheer number of parameters and scenarios analyzed, only two of the nine
parameters are discussed for each case, with full documentation of each scenario located in APPENDIX:
Simulation Data. The two chosen parameters for each case represent two parameters which are both
shown to be statistically significant and present interesting findings.
4.2 Baseline – Open Office
The baseline scenario to be analyzed was the open office space, facing south and simulated over the
course of an entire year. The open office simulation was conducted with all performance metrics equal in
value to each other during the optimization sequence. After completing 1217 simulations, approximately
23 generations through Galapagos, a substantial range of configurations were documented. All the data is
recorded and streamed into an Excel document template that is used to review each configuration as raw
numerical data and graphically through scatter plots. All nine parameters are recorded along with the
mean irradiance values, useful daylight illuminances, simplified daylight glare probability, derived fitness
value, and the simulation run number. The fitness vales shown in column M (Fitness) indicate the best
configurations for the range of parameters (Figure 21).
54

Figure 21. (Open Office, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations
Given the fitness values above, simulations #711 and #1131 rank as the top two most effective shading
system configuration based on the combined performance of mean irradiance, useful daylight
illuminances, and simplified daylight glare probability. These two configurations can be compared to
examine the differences between highly performing configurations and their corresponding parameter
values. Minor differences exist between #711 and #1131, but most parameters are towards the same end
of each parameter range.
Graphic representation each of the different shading configurations makes it easier to observe the distinct
differences between each configuration. The 3D models help to quickly convey minor changes between
each configuration set. Both effective configurations include a high number of louvers with medium
depth and a strong negative angle. They have an average amount of fins with a short depth and strong
negative angle. Both systems are offset in both directions a few inches off the face of the building (Figure
22).
55

Figure 22. Shading System Visualization and Configuration Information for #711 and #1131
Using the LINEST array function in Excel, a simple regression analysis was conducted with each
parameter compared to the fitness, indicating which parameters are most significant to providing change
in the fitness value. In this baseline configuration almost all the parameters excluding the Fin Length and
Vertical Offset parameters were discovered to be significant facilitators of fitness value change (Figure
23). As a result of these early observations, two particular parameters, the Number of Louvers and
Louver Length were chosen to have their influence further analyzed through polynomial trend lines over
scatter plots.

Figure 23. (Open Office, South, Annual, 1:1:1 Weighting) Significant Parameters
The Number of Louvers parameter follows a fairly steep convex trend line (Figure 24). A convex trend
line indicates that somewhere in the middle of the parameter range is the best configuration, and in this
case the curve reaches its peak at eight louvers. Before and after eight louvers the trend line slopes
downwards, signifying a loss of fitness and a less optimal parameter choice.
56
The Louver Length parameter has a larger convex trend line that further demonstrating how the fitness
values react as the length of the louver is changed (Figure 24). A short louver depth has a poor fitness
value, but too deep of louver is also undesired. A range between 12 to 16 inches for the louver length
achieves the highest fitness values.
Both parameters exhibit trends that hint towards the usefulness of this process. Given the range for these
parameters, too low or too high values will produce suboptimal fitness values, however given a number
somewhere in the middle – in which this process helps to narrow down and locate – is the shown to be the
most effective configuration. This narrowing process is what removes the ambiguity from designing
shading systems, the optimally performing parameter range can be identified and locked down.

Figure 24. (Open Office, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length
4.3 Hotel
The hotel room was analyzed over the course of a year, facing south, and with equal weighting on all
three metrics (solar radiation, UDI, and DGPs). 1570 simulations were analyzed, finding #970 and #900
as the top two performing configurations (Figure 25). These two followed mostly similar parameter value
concentrations yet substantial differences in their measured solar radiation and illuminance values.
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0 1 2 3 4 5 6 7 8 9 10 11
N of Louvers
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
57
Configuration #970 has higher average solar radiation and subsequently higher percentage of time with
UDI than #900 which has both lower radiation and UDI, yet both achieve a similar combined fitness
score, indicating their parameter values are in and around the right spot for optimal configuration.

Figure 25. (Hotel, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations
The most significant differences between the two systems exist in their horizontal and vertical offsets
from the face of the building. #970 has no vertical offset and 10 inches of horizontal offset, explaining
why it received more solar radiation and more time in the useful range of illuminance on the task surface
compared to #900 which had 4 inches less horizontal offset (Figure 26). Minor changes in number of
louvers (7 to 8), louver length (13 to 11), and fin length (6 to 5) and angle (-30 to -36) affect the outcome
also, but in a very minor way. Many of the differences balance each other out, for example one less
louver, but the increase of louver depth may achieve the same result as a higher number of shorter
louvers.
58

Figure 26. Shading System Visualization and Configuration Information for #970 and #900
The Number of Louvers and Louver Length parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 27).
The Number of Louvers parameter trend line follows a convex curvature, leveling at towards the end
range of the parameter at a value of 10 (Figure 28). Both top performing configurations gravitate towards
a higher number of louvers, 7 and 8, validating their high fitness values.
The Louver Length parameter features a bell curve trend line with higher fitness values towards the
middle of the parameter range, 12 to 13, with fitness values trailing off as you move away in either
direction (Figure 28). Again #970 and #900 corroborate this trend with their louver length parameters at
13 and 11 inches deep.

Figure 27. (Hotel, South, Annual, 1:1:1 Weighting) Significant Parameters
59

Figure 28. (Hotel, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length
4.4 Closed Office
The closed office was analyzed over the course of a year, facing south, and with all metrics equally
valued. 1372 simulations were conducted within the closed office. Two optimal configurations were
found early in the process. The second highest performing configuration was recorded on the 24th
simulation, and later an even better performing configuration was documented on the 309th simulation
(Figure 29). Due to the initial set of simulations (the first 100 simulations) having already found an
excellent configuration, the subsequent simulations would all narrowly focus on similar parameter
concentrations. For example, the top ten configurations all have almost the same values for all parameters
except the Louver Angle and Fin Angle. The process was able to lock down the other six parameters as
being at their most effective value and had begun to try and pin down the optimal values for the louver
and fin angles.
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 1 2 3 4 5 6 7 8 9 10 11
N of Louvers
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
60

Figure 29. (Closed Office, South, Annual, 1:1:1 Weighting) Top Two Performing Configurations
Due to the aforementioned determination of one of the better performing configurations in the initial
population and following population and mutations based on that one configuration, the differences
between #309 and #24 and even the next 8 configurations are minimal. All 10 fall within a couple steps
of parameter values each of other. The main differences are in the fin angles which range from -15 to 9
degrees (steps of 3) (Figure 30).

Figure 30. Shading System Visualization and Configuration Information for #309 and #24
The Number of Louvers and Louver Length parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 31).
The trends of both the Number of Louvers and Louver Length parameters are a slight departure from the
previously reviewed open office and hotel spaces. The Number of Louvers parameter trend line covers a
61
full convex bell curve, indicating optimal fitness performance with 5 to 6 louvers and tapering
performance as the number of louvers increase or decrease (Figure 32).
The Louver Length parameter range doesn’t cover the entirety of the convex trend line, but shows enough
to indicate the leveling out, optimal range, of the curve towards the end range of the parameter, around 22
to 24 inches deep (Figure 32). The range of parameter values for the top 10 configurations falls between
17 to 21 inches, creeping towards that optimally high depth.

Figure 31. (Closed Office, South, Annual, 1:1:1 Weighting) Significant Parameters

Figure 32. (Closed Office, South, Annual, 1:1:1 Weighting) Number of Louvers and Louver Length
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
0 1 2 3 4 5 6 7 8 9 10 11
N of Louvers
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
62
4.5 East
Two directions were analyzed over the course of this study to accurately portray the two types of
direction conditions – south and east/west (west was excluded in favor of an eastern simulation being
satisfactory in covering the lateral directions).
The open office space was recreated, rotated to face the eastern direction and analyzed annually with all
metrics equally valued. Two optimal configurations out of 1209 were recorded at simulations #1199 and
#852 (Figure 33). Both configurations were documented towards the end of the simulation session, along
with the other three in the top five, indicating relative convergence of optimal configurations.

Figure 33. (Open Office, East, Annual, 1:1:1 Weighting) Top Two Performing Configurations
The two configurations are very closely related, with a difference of number of louvers and fins balanced
with an increase or decrease in the louver or fin depth. #1199 has less and shallower fins compared to
#852 allowing for a higher percentage of occupied hours in the year that the task surface illuminance
threshold is met, but also introduces a higher chance of discomforting glare (Figure 34). Given the
current equally weighted fitness equation, the greater amount of time within the UDI threshold is more
desirable than the increase in potential glare within the space; edging out #852 to the second spot. This
particular simulation brings attention to the importance of the weighting of each metric according to the
usage of the space and intended occupants. #852 may be a more desirable configuration (or even another
configuration not already considered in this study) depending on whether potential glare is a bigger
detriment versus a more properly illuminated room.
63

Figure 34. Shading System Visualization and Configuration Information for #1199 and #852
The Louver Length and Louver Angle parameters were found to be both statistically significant and were
further analyzed via trend lines and scatter plot graphs (Figure 35).
The Louver Length parameter range almost covers a full polynomial trend line’s bell curve range, with it
leveling out at 17 inches (Figure 36). The top 10 parameters all fall within 15-17 inches, reaching toward
that optimal parameter value at the apex of the trend line.
The Louver Angle parameter follows a continuously rising convex curve, with no leveling out of the trend
line within the tested parameter range (Figure 36). Despite no visible peak in the curve, the trend line
direction points towards more negative louver angles producing higher fitness values. The best
performing configurations are all concentrated in the high negative angle range.

Figure 35. (Open Office, East, Annual, 1:1:1 Weighting) Significant Parameters
64

Figure 36. (Open Office, East, Annual, 1:1:1 Weighting) Louver Length and Louver Angle
4.6 Winter
The open office was simulated under two seasonal conditions to obtain more specific shading
configurations tuned for seasonal temperatures, daytime lengths, and sun angles. Additionally, the fitness
formula was changed to reflect the desire for more solar radiation in the winter months, as opposed to the
annual formula which tries to minimize solar radiation.
The open office was simulated as facing south, over a one month period from December 21st to January
21st, and with all metrics valued equally. 1284 simulations were run, with the two most effective being
documented reasonably late into the process. Simulations #1012 and #1215 were the top two performing
configurations (Figure 37).
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
-48 -42 -36 -30 -24 -18 -12-6 0 6 12 18 24 30 36 42 48
Louver Angle
65

Figure 37. (Open Office, South, Winter, 1:1:1 Weighting) Top Two Performing Configurations
#1012 is shown to generally decrease light into the space via an increase in the number (7 compared to 6)
and depth (17 compared to 12 inches) of the louvers versus #1215 (Figure 38). This achieved a lowered
probability of glare, in which the change in glare probability outweighs the decrease in UDI. Both
configurations included the minimum amount of fins in the parameter range, attempting to minimize the
amount of sunlight lost while maintaining a low glare environment.

Figure 38. Shading System Visualization and Configuration Information for #1012 and #1215
The Number of Fins and Louver Length parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 39).
66
The concave trend line covering the Number of Fins parameter plot indicate the minimum number of fins
led to the highest performing configurations (Figure 40). Fewer fins would allow more low angles
sunlight to pass through the louvers while still minimizing glare and direct sunlight.
The Louver Length parameter follows a convex trend line with a peak at 14-15 inches for the depth of the
louver system (Figure 40).

Figure 39. (Open Office, South, Winter, 1:1:1 Weighting) Significant Parameters

Figure 40. (Open Office, South, Winter, 1:1:1 Weighting) Number of Fins and Louver Length
4.7 Summer
Next the open office was analyzed facing south, during the summer (June 21st to July 21st), and with all
metrics equally valued. After 1276 simulations, the top two configurations were documented as #110 and
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
1 2 3 4 5 6 7 8 9 10 11
N of Fins
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
67
#1, indicating the initial generations in the optimization process were able to find highly effective
configurations and narrow down the subsequent configurations in the following generations (Figure 41).

Figure 41. (Open Office, South, Summer, 1:1:1 Weighting) Top Two Performing Configurations
The very first configuration created indicated a very high fitness value and immediately narrowed down
the parameter values. Even after 1276 simulations, the top 4 configurations were recorded within the first
162 simulations conducted. The top performing configuration performs considerably better than the
second best and as such the parameter values are quite different. The louver length difference is 4 inches
and the number of fins, angles, lengths and vertical offset vary widely (though the Fin Angle, Fin Length,
and Vertical Offset parameters were deemed statistically insignificant towards the fitness value)(Figure
42). Due to the high summer sun angles, the increase or decrease of the vertical offset parameter did not
affect the space much.
68

Figure 42. Shading System Visualization and Configuration Information for #110 and #1
The Number of Fins and Horizontal Offset parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 43).
The Number of Fins and Horizontal Offset parameters both show shallow trend line curves (Figure 44).
The remaining 7 parameters also have shallow curves due to the early find of an exceptionally performing
configuration. The Number of Fins shows a shallow convex curve with a peak at 5 fins.
The Horizontal Offset parameter shows a trend towards having a low value or no horizontal offset at all
providing the best fitness values (Figure 44). Having a shading system which pushed away from the
space would allow the high angled summer sun to enter the space unobstructed, introducing undesirable
glare and solar radiation.

Figure 43. (Open Office, South, Summer, 1:1:1 Weighting) Significant Parameters
69

Figure 44. (Open Office, South, Summer, 1:1:1 Weighting) Number of Fins and Horizontal Offset
4.8 Double UDI
The following four scenarios consider the baseline open office scenario (south facing and annually) with
one specific metric differently weighted in each simulation set. This was done to observe how the
weighting coefficients affected the optimal parameter choices in the shading system configurations.
This scenario weights UDI twice as important in the fitness equation compared to solar radiation and
glare metrics. Over the course of 1213 simulations the two performing configurations were #79 and #721
(Figure 45). However, inspection of #79’s parameter values and even re-instantiated model show a
substantial departure in measurements and parameters compared to the other top performing
configurations, yet still performing the best. Given this information it can be concluded that #79 was
somehow erroneously placed at the top (a data recording hiccup is discussed in Chapter 3.8; it appears the
DGPs measurement is erroneous, causing a seemingly poor configuration to be much ranked higher than
intended). Correcting for this the top two performing configurations are #721 and #172.
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
1 2 3 4 5 6 7 8 9 10 11
N of Fins
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
System H Offset
70

Figure 45. (Open Office, South, Annual, 1:2:1 Weighting) Top Two Performing Configurations
A comparison of the top 10 configurations shows a pretty stable parameter value concentration the louver
angle and length and number of fins. Other parameter values seem to jump around a bit, and can be
explained by their lack of statistical significance. The three parameters with were deemed significant all
clustered around each other – number of fins ranged from 3 to 5, louver angle from -45 to -33, and louver
length from 15 to 19 inches (Figure 46). Given the range of each of these parameters, the change was
very small.

Figure 46. Shading System Visualization and Configuration Information for #79 and #721
The Louver Length and Louver Angle parameters were found to be both statistically significant and were
further analyzed via trend lines and scatter plot graphs (Figure 47).
71
The Louver Length parameter follows a full convex trend line with a peak achieved at around 16 inches in
depth (Figure 48). The Louver Angle parameters has a shallow almost straight trend line pointing
towards more negative angles producing higher fitness configurations (Figure 48).

Figure 47. (Open Office, South, Annual, 1:2:1 Weighting) Significant Parameters

Figure 48. (Open Office, South, Annual, 1:2:1 Weighting) Louver Length and Number of Louvers
4.9 Triple UDI
The triple weighted UDI scenario follows the same baseline open office simulation setup but with UDI
weighted three times as much as UDI and DGPs in the fitness equation. 1056 simulations were
completed with this scenario. The top two performing configurations were #926 and #1011 (Figure 49).
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
-48 -42 -36 -30 -24 -18 -12-6 0 6 12 18 24 30 36 42 48
Louver Angle
72

Figure 49. (Open Office, South, Annual, 1:3:1 Weighting) Top Two Performing Configurations
Both top performing configurations were documented towards the end of the simulation cycle and
therefore have achieved relative convergence of their parameter values. Small variations, one step value
or less, occur between the 5 statistically significant parameters – the number, angle, and length of louvers,
fin length, and system horizontal offset (Figure 50).

Figure 50. Shading System Visualization and Configuration Information for #926 and #1011
The Louver Length and Louver Angle parameters were found to be both statistically significant and were
further analyzed via trend lines and scatter plot graphs (Figure 51).
The Louver Length parameter displays a full convex trend line with a peak being achieved at 14 inches
for optimal fitness values (Figure 52). The Louver Angle parameter has a near linear trend line pointing
73
towards more negative angles achieving better fitness values. Both of these trends are observed in the top
two configurations (Figure 52).

Figure 51. (Open Office, South, Annual, 1:3:1 Weighting) Significant Parameters

Figure 52. (Open Office, South, Annual, 1:3:1 Weighting) Louver Length and Louver Angle
4.10 Double DGPs
This scenario uses the baseline open office, facing south, over the course of the year, with DGPs weighted
twice as much as UDI and solar radiation. 1209 simulations were conducted using this scenario. Both
top performing configurations were recorded in the same initial population, narrowing down the
parameter values to optimal ranges very early (Figure 53). However, none of the other configurations
were able to exceed those two in the initial population. Thus, their parameter values are quite varied for
each performing as well as they do.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 2 4 6 8 10 12 14 16 18 20 22 24
Louver Length
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
-48 -42 -36 -30 -24 -18 -12 -6 0 6 12 18 24 30 36 42 48
Louver Angle
74

Figure 53. (Open Office, South, Annual, 1:1:2 Weighting) Top Two Performing Configurations
#57 and #61 achieve similarly high fitness values with inconsistent parameter values, and accordingly
with quite different measured results (Figure 54). In comparison with #61, #57 has less solar radiation
(good), less UDI (bad), and significantly less DGPs (good). Since DGPs is weighted twice as important
as the other metrics #57’s measured performance is considered better.

Figure 54. Shading System Visualization and Configuration Information for #57 and #61
The Louver Angle and Horizontal Offset parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 55).
The Louver Angle parameter features a partial convex trend line curve with no peak in the given
parameter range (Figure 56). Given this trend line it can be inferred more negative louver angles produce
75
higher fitness values. The top 10 configurations range from -24 to -39, keeping in line with this trend
direction.
The System H Offset parameter indicates a shallow convex trend line which indicates a low to zero
horizontal offset will generate optimally performing configurations (Figure 56). However, the top two
configurations utilize reasonably high horizontal offsets and have achieved the best results. The reason
for these two outliers is the result of an initial population generating good configurations while the rest of
the simulations generated configurations based on continuously narrowing parameter ranges. The nine
next best configurations however corroborate with the given trend line, offsetting between 0 and 6 inches
away from the space to achieve optimal results.

Figure 55. (Open Office, South, Annual, 1:1:2 Weighting) Significant Parameters

Figure 56. (Open Office, South, Annual, 1:1:2 Weighting) Louver Angle and Horizontal Offset
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
-48 -42 -36 -30 -24 -18 -12-6 0 6 12 18 24 30 36 42 48
Louver Angle
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
System H Offset
76
4.11 Triple DGPs
The last scenario considered was a three times DGPs weighted baseline open office setting simulated
annually and facing south. 1223 simulations were run, with the top two configurations coming up as #17
and #43 (Figure 57). Within the top 10 configurations were 3 from the initial population generated
randomly at the start of the evolutionary solving startup. Similar to the twice weighted DGPs scenario,
the top two performers were generated in the initial batch. The parameter ranges have very little
similarity and the measurements are generally balanced with the exception of one having less glare
probability, the weighted metric, and thus performing better than the other.

Figure 57. (Open Office, South, Annual, 1:1:3 Weighting) Top Two Performing Configurations
Attempting to drastically reduce the potential for glare in the space, #17 includes a high number of
louvers which are negatively angled to protect against any direct sunlight entering the space (Figure 58).
It achieves a mediocre amount of time with a properly lit task surface, but most importantly has a
significantly reduced chance of glare to the occupant. The different between the first and second
configuration is relatively significant (-0.14574 compared to -0.31554 and -0.33191 behind it). Thus, #43
is not quite competitive with #17. The lack of louvers and almost flat louvers suggest it would achieve
poorer solar radiation and glare control – which it does.
77

Figure 58. Shading System Visualization and Configuration Information for #17 and #43
The Louver Angle and Horizontal Offset parameters were found to be both statistically significant and
were further analyzed via trend lines and scatter plot graphs (Figure 59).
The Louver Angle parameter achieves a slight convex trend line curve, with an apparent peak towards the
more negative louver angles, protecting the interior space from direct sunlight and increasing fitness
values (Figure 60). The System H Offset parameter forms a very slight and almost linear trend line
indicating little to no horizontal would produce the best performing configurations (Figure 60).

Figure 59. (Open Office, South, Annual, 1:1:3 Weighting) Significant Parameters
78

Figure 60. (Open Office, South, Annual, 1:1:3 Weighting) Louver Angle and Horizontal Offset
-2
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
-48 -42 -36 -30 -24 -18 -12-6 0 6 12 18 24 30 36 42 48
Louver Angle
-2
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
System H Offset
79
CHAPTER FIVE - DISCUSSION

5.1 Comparisons
The results of the 10 different scenarios showcases the trends and statistically significant parameters in
each instance. Individually each instance represents unique configurations to optimally fit for the size of
the room, time of year, or direction. When the significant parameters of each set of simulations are
compared together, insight into how the optimal configurations were achieved were discovered and why
each parameter is tuned to that particular scenario. The following five grouping of the 10 original
scenarios compares them against each other based on their typologies, orientations, timeframes, and
weighting factors.
5.2 Typologies
The three different types of spaces being analyzed were brought together and compared, focusing on
parameters that were all deemed significant across each room. The Number of Louvers and Louver
Length parameters were considered significant between the open office, hotel, and closed office spaces.
A comparison of three spaces reveals that the two larger rooms, open office and hotel, follow similar
trends while the smaller, closed office space, does not (Figure 61). The number of louvers in the open
office and hotel seem to peak at around 8 or 9, with the closed office finding its optimal louver number at
6 (Figure 62). Comparing the louver length parameter, the open office and hotel are again in agreement,
both capping out at 13 to 15 inches with the closed office instead peaking at the max range of the Louver
Length parameter of 23 inches (Figure 63).
80

Figure 61. Open Office, Hotel, and Closed Office Configurations
These trends are explained by the proximity of the window and shallow room depth in the closed office
compared to the considerably larger and deeper open office and hotel rooms. Deeper louvers, but fewer
of them, allow the closed office to minimize glare and solar radiation while still maintaining high levels
of natural daylight into the space. A decrease in louver depth with a corresponding increase in the
number of louvers performed more appropriately for the larger two rooms.
This examination of parameter trends between different spaces in matching simulation settings can help to
inform designers working on a variety of room sizes and purposes. For example, if a new medium sized
space (between the size of the closed office and open office) is added to the project programming, it’s
number of louvers and the length of those louvers can be determined to range between 5 to 9 with lengths
from 13 to 23 inches, narrowing down the design parameters immediately.
81

Figure 62. (Hotel/Open Office/Closed Office, South, Annual, 1:1:1 Weighting) Number of Louvers

Figure 63. (Hotel/Open Office/Closed Office, South, Annual, 1:1:1 Weighting) Louver Length
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
0 1 2 3 4 5 6 7 8 9 10 11
Fitness
N of Louvers
Poly. (Hotel) Poly. (Open Office) Poly. (Closed Office)
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
0 2 4 6 8 10 12 14 16 18 20 22 24
Fitness
Louver Length
Poly. (Hotel) Poly. (Open Office) Poly. (Closed Office)
82
5.3 Orientations
The south and east facing open office spaces were compared with each to analyze the difference of
parameter trends based on orientation. The Louver Length and Horizontal Offset parameters were
deemed statistically significant across both scenarios.
The Louver Length parameter of the south facing open office peaks at a shorter depth compared to the
eastern facing version. Additionally, the south office is optimally configured with a few inches of
horizontal offset, while the east office excels with little to no offset (Figure 64).

Figure 64. South and East Open Office Configurations
These trends can be explained by the morning light as the low angled sun passes from the east to the
south, penetrating into the eastern open office. A tighter shading system with no horizontal offset and
longer louvers ensures less direct sunlight can enter the space creating unwanted glare and solar radiation
(Figure 65) (Figure 66). As the sun travels to the south open office, the louvers are shorter an even a
slight horizontal offset is introduced to bring in desirable ambient light. The high sun angle is easily
blocked by shortened louvers and cannot deeply penetrate the space from a small horizontal offset.
83

Figure 65. (Open Office, South/East, Annual, 1:1:1 Weighting) Louver Length

Figure 66. (Open Office, South/East, Annual, 1:1:1 Weighting) System H Offset
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24
Fitness
Louver Length
Poly. (Open Office - South) Poly. (Open Office - East)
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
Fitness
System H Offset
Poly. (Open Office - South) Poly. (Open Office - East)
84
5.4 Timeframes
Analysis of the seasonal extremes shows how the configuration of a shading system parameters change
across the year. Parameters such as the number of fins vary greatly across the three compared
timeframes, while the louver length parameter stayed consistent throughout (Figure 67).
The Number of Fins parameter peaks at 5 to 6 for both the summer month and the annual scenarios. In
the winter month, the number of fins is trending towards having the minimum value of two (Figure 68).
This is contrasted with all three measured periods matching their trend line flattening out at a louver depth
of 14 inches (Figure 69). The winter and annual trend lines even follow each other’s curvature fairly
closely.

Figure 67. Annual, Winter, and Summer Open Office Configurations.
The fewer numbers of fins in the winter month can be explained by the increase in desirability of solar
radiation in cooler times of the day where the sun is rising and setting. Less fins mean more low angled
sun can penetrate in the room and provide warmth, where the excess direct sun is still obstructed by the
angled louvers. In the summer an average amount of fins is necessary to block out excessive sunlight and
prevent overheating. Over the course of the year the optimal number of fins match the summer scenario
due to the relatively warm climate the building is tested in, as opposed to a more balanced environment.
85

Figure 68. (Open Office, South, Annual/Winter/Summer, 1:1:1 Weighting) Number of Fins

Figure 69. (Open Office, South, Annual/Winter/Summer, 1:1:1 Weighting) Louver Length
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
1 2 3 4 5 6 7 8 9 10 11
Fitness
N of Fins
Poly. (Summer) Poly. (Annual) Poly. (Winter)
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24
Fitness
Louver Length
Poly. (Summer) Poly. (Annual) Poly. (Winter)
86
5.5 Weighting Factors – UDI
At the heart of the performance evaluation of each configuration is the fitness equation. At default each
metric is equally weighted with each other, meaning an equivalent change in glare reduction equally
offsets an increase in solar radiation gains. By changing the weighting of each metric to each other, the
optimal configuration will be skewed towards supporting the metric that is worth more in comparison to
the others. This allows customization of design intent corresponding with space occupancy purpose when
running this work flow.
This comparison follows three different weightings of the useful daylight illuminances metric, altering the
weighting in the fitness equation by no change, twice as important, and three times as important. The
following analysis covers and exhibits the change in optimal configurations as the weighting factor is
applied (Figure 70).

Figure 70. Equal Weighted, Double Weighted, and Tripled Weighted UDI Configurations
At equal weighting the Louver Length parameter peaks at 14 inches and the number of louvers at 8. As
the weighting is increased for UDI, the louvers increase in depth to 16 inches and the number of louvers
decrease to 6. Finally at three times the weighting factor for UDI the louver depth decreases back to the
original 14 inches as the number of louvers stays the same at 6 (Figure 71).
As the important of UDI is increased the optimal configuration trends shift. The increase of louver depth
to compensate for the decrease is total number of louvers is observed in the doubled important of the
metric as the parameter values seek to find a new balance. At three times the weighting the parameters
87
decrease the louver depth again while keeping the same lowered number of louvers – opening up the
space to even more potential daylight (Figure 72).

Figure 71. (Open Office, South, Annual, 1:1/2/3:1 Weighting) Louver Length
-1.5
-1
-0.5
0
0.5
1
1.5
0 2 4 6 8 10 12 14 16 18 20 22 24
Fitness
Louver Length
Poly. (2:1 UDI) Poly. (1:1 UDI) Poly. (3:1 UDI)
88

Figure 72. (Open Office, South, Annual, 1:1/2/3:1 Weighting) Number of Louvers
5.6 Weighting Factors – DGPs
Similar to the weighting of UDI, a comparison of three different weightings for DGPs was documented.
The baseline open office scenario with all DGPs weighted equally with the other metrics produced an
average configuration meant to protect against direct solar radiation and glare while maximizing ambient
light for illumination of a task surface (Figure 73).

Figure 73. Equal Weighted, Double Weighted, and Triple Weighted DPGs Configurations
As the weighting and importance was increased for DGPs, the original 6 to 8 inches of horizontal offset
began to scale back. The double and triple weighted DPGs configurations trended towards little to no
-1.5
-1
-0.5
0
0.5
1
1.5
0 1 2 3 4 5 6 7 8 9 10 11
Fitness
N of Louvers
Poly. (2:1 UDI) Poly. (1:1 UDI) Poly. (3:1 UDI)
89
horizontal offset, maintaining less chances of any direct sunlight and therefore lower glare probability
into the space (Figure 74). Since the original baseline configuration was performing fine with the Louver
Angle parameter, not much change was made across the double and triple weighting of DGPs for that
parameter (Figure 75).

Figure 74. (Open Office, South, Annual, 1:1:1/2/3 Weighting) Louver Angle
-2
-1.5
-1
-0.5
0
0.5
-48 -42 -36 -30 -24 -18 -12 -6 0 6 12 18 24 30 36 42 48
Fitness
Louver Angle
Poly. (2:1 DGPs) Poly. (1:1 DGPs) Poly. (3:1 DGPs)
90

Figure 75. (Open Office, South, Annual, 1:1:1/2/3 Weighting) Horizontal Offset
5.7 Process Discussion
Individual and comparative assessment of the optimized data sets produced from using this process can
provide the beginnings to shading system designs. Guidelines for different sizes and uses of spaces,
orientations, focused timeframes, and weighted importance of different metrics can be extracted from the
optimized configuration data sets.
The baseline open office scenario produced an optimal shading system configuration based on the Los
Angeles environment. A high number of average depth louvers combined with fins and offset both
vertically and horizontally away from the building face was the optimal configuration achieved. Shallow-
angled louvers at 30 degrees blocks direct sunlight will keeping line of sight and ambient daylight
unobstructed.
Comparing that baseline open office space configuration to other types of spaces with varying sizes and
occupancies results in quick guidelines for that analyzed building. Deeper but less louvers work better
-2
-1.5
-1
-0.5
0
0.5
-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
Fitness
System H Offset
Poly. (2:1 DGPs) Poly. (1:1 DGPs) Poly. (3:1 DGPs)
91
with smaller spaces, providing good line of sight and direct sunlight protection for a space with less
window area. On the other side, an increase of shorter louvers works better for larger spaces such as the
open office and hotel rooms. The increase in window area keeps the ambient light appropriate for task
illuminance with the increase in total number of louvers.
When considering the orientation into the design of shading systems the optimal configurations point to
more closed and dense systems for the east and west directions with more open and lighter designs for
south facing spaces. The deeper louvers provide excellent sun protection on the southern direction,
giving the shading system more room to allow ambient light in. On the east or west side the low angled
morning and evening sun are able to penetrate past the negatively angled louvers better and should be
designed deeper, steeper angled, and with more fins than the south.
Although the default scenario is based in the warm climate of Los Angeles, it is still important when
designing a shading system, to ensure enough warmth will be achieved in the winter. A reduction of fins
may be one method of achieving those solar gains in the winter. Maintaining similar louver lengths as in
the summer and year round while removing fins allows the lower angled winter sun to penetrate into the
space when it is still a cooler temperature outside in the morning or becoming cooler in the evening.
Weighting metrics to guide the shading system design into focusing on specific design needs can
highlight particular parameters which are significant to optimal shading configurations. When
maintaining task illuminance in the space is the priority, the number and depth of the louvers in the
optimal configurations recedes as the importance of allowing more light in at the cost of increased glare
probability and solar radiation increases. The number of louvers and louver length parameters are shown
to be significant parameters affecting the performance of the space in that given scenario and should be a
priority design decision to maintain proper illuminance in the room.
92
CHAPTER SIX - CONCLUSIONS

6.1 Process
A design process that can help guide designers into making informed and effective decisions regarding
shading systems can help to improve energy efficiency and occupant comfort. The described design
process was used to construct parametric shading systems and optimize the system based on three solar-
based metrics. It allowed the user to manipulate select pieces of the workflow, including which metrics
are possibly more important to the design than the others and guide the optimization towards catering to
those measurements.
The process produces two sets of information. First, a list of optimized shading system configurations
based on the user inputs, design constraints, and metric weighting. Second, graphical representations of
how each parameter influenced the overall performance of the system. The results highlight the impact of
each parameter and potentially which variables of the parametric shading system should be focused on for
design alternatives.
6.2 Future Work
The workflow was able to tackle multiple parameters at once using a pseudo multi-objective evolutionary
solving process; producing optimized shading configurations based on input geometry, climate, and
orientation. Future work could improve the workflow through the use of form-finding geometry,
additional material characteristics, improving the sample size, expanding the scope to include thermal
measurements, and true multi-objective algorithms.
6.2.1 Form-finding Geometry
The current workflow includes the manipulation of louvers and fins to achieve variance in the shading
system. An improvement on static geometry varied by stock parameters would be an introduction of
form-found geometry, modified by the same evolutionary solving process used to flex the parameters of
93
the original louvers and fins. A point cloud could be setup at the face of the analyzed space with a series
of curves running through that point grid to create a new shade. As iterations of shades are created they
are simulated and analyzed for their performance with each iteration refining its shape into something
better and better.
Form-found geometry can more appropriately shade or allow light into the space based on the prescribed
intent of the space and any particular metric weighting. It can be customized to forms that perfectly
follow the sun’s path and building form.
6.2.2 Material Characteristics
Although the current workflow includes standard materials for the walls, floors, ceiling, shading devices,
and glass, more materials could be introduced and could be parameterized to begin to influence the entire
process. Additional material choices in the shading system especially could for example enables it to
reflect light deeper in the space with highly reflective materials. Though testing for glass types was
discussed in Chapter 3.5, it could be refined to be a less dominant trait. The introduction of more material
choices in the workflow would enable it to simulate a more realistic scenario and provide more accurate
results to the user.
6.2.3 Improved Sample Size
Only one reference building with a few types of rooms in a consistent climate was used in this case study.
An expansion of building types and climates would help to validate the workflows’ optimal configuration
generation as well as parameter trends. The baseline climate used was Los Angeles, which is typically
hotter year round, thus predicting optimal shading configurations to be more protective of direct sunlight.
A cooler climate, with considerably colder winters could be integrated to highlight the differences of what
an optimal shading configuration would look like where more sunlight penetrating into a space may be
more desirable for more the year.
94
6.2.4 Thermal integration
As mentioned in Chapter 1.4.1, daylighting and solar control metrics are the focus to guide the
performance criteria of the shading systems. Thermal considerations are important, but were left for
future work. It would be beneficial to consider all environmental impacts as performance criteria. DIVA
for Rhino includes the ability to perform single-zone thermal modeling using EnergyPlus as the
simulation engine. Already built-in to the workflow, the DIVA engine could consider indoor
temperature, heating and cooling loads, and advanced material properties to more accurately shape and
configure shading systems.
6.2.5 True Multi-Objective Solving
The pseudo multi-objective solving process currently in place utilizes a normalized combined fitness
equation to take into consideration multiple objectives being simulated at once and comparing to each
other. This was one of the best methods for comparing multiple objectives against each other.
A new plug-in for Grasshopper called Octopus is able to tackle multiple objectives using the same
evolutionary solving algorithms used in this workflow to generate a range of optimized trade-off
solutions. Based on SPEA-2 (Strength Pareto Evolutionary Algorithm), Octopus can accurately analyze
multiple objectives (varying metrics) at the same time and find tradeoffs between these goals; this results
in a truer optimal configuration (Vierlinger 2014).
6.3 Process Conclusions
Ultimately this process is a tool for designers in the early stage of design. It can be easily adapted to any
project considering a shading system. Basic Grasshopper knowledge is needed to parameterize the façade
intent, but once implemented the process can cycle through those and simulate those design options with
ease. This workflow is intended to aid designers working on shading systems at the conceptual stage,
guiding them with narrowed down parameter values, significant parameters, and even in which direction
parameters are trending towards.
95
Beyond just generating optimized results for a basis of design, this workflow analyzes the simulations
used to create those optimal configurations and can break down how each component is contributing to
the overall performance of the system. Narrowing down which parameters are most important in a design
can signal where to focus design options and alternatives, drastically reducing time focusing on
insignificant variables. Additionally, with the tendencies of each parameter outlined in scatter plots with
trend lines – predicting where the optimal values occur – it reinforces intuitive design decisions made by
supporting them with hard data and visual cues.
This novel approach to introducing simplified multi-variable simulations to designers is a step in the right
direction of providing more integrated and holistic approaches to crucial daylighting and energy
simulations at the earliest stages of design. Tools like this workflow can lead to comprehensive software
packages where the design model can stay from inception to construction – with the design continuously
informed through simulations along the way.
96
BIBLIOGRAPHY

Aish, Robert, and Robert Woodbury. 2005. "Multi-level Interaction in Parametric Design." Smart
Graphics (Spring Berlin / Heidelberg) 3638: 924.
Akin, Omer. 2001. "Variants in Design Cognition." In Design Knowing and Learning: Cognition in
Desgin Education, 105-124. Elsevier Ltd.
Daysim. 2013. Daysim. Accessed November 1, 2014. http://daysim.ning.com/.
Eisenhower, Bryan, Zheng O'Neill, Vladimir A. Fonoborev, and Igor Mezic. 2011. "Uncertainty and
Sensitivity Decomposition of Building Energy Models." Journal of Building Performance
Simulation.
Enclos Corp. 2014. Enclos. Accessed November 1, 2014. http://enclos.com/.
Gagne, Jaime M. L., and Marilyne Andersen. 2010. "Multi-objective facade optimization for daylighting
design using a genetic algorithm." Proceedings of the 4th National Conference of IBPSA-USA.
New York.
Granadeiro, Vasco, Joao R. Correia, Vitor M.S. Leal, and Jose P. Duarte. 2013. "Envelope-related energy
demand: A design indicator of energy performance for residental buildings in early design
stages." Energy and Buildings 215-223.
Illuminating Engineering Society of North America. 2000. The IESNA Lighting Handbook, Ninth Edition.
Edited by Mark S. Rea. New York.
Kleindienst, Sian A, and Marilyne Andersen. 2009. "The Adaptation of Daylight Glare Probability to
Dyanmic Metrics in Computational Setting." Lux Europa 2009. Istanbul.
Mazria, Edward. 2011. Architecture 2030. Accessed August 13, 2013. http://architecture2030.org/.
97
Miller, Nathan. 2014. LunchBox for Grasshopper. Accessed November 1, 2014.
http://wiki.theprovingground.org/plugin-tpg.
Mwasha, Abraham, Rupert G. Williams, and Joseph Iwaro. 2011. "Modeling the performance of
residential building envelope: The role of sustainable energy performance indicators." Energy
and Buildings 2108-2117.
Nabil, A, and J Mardaljevic. 2005. "Useful daylight illuminance: a new paradigm for assessing daylight in
buildings." Lighting Research and Technology 37 (1): 41-59.
Nabil, Azza, and John Mardaljevic. 2006. "Useful daylight illuminances: A replacement for daylight
factors." Energy and Buildings 905-913.
Radiance. n.d. Radiance Materials. Accessed November 1, 2014.
http://radsite.lbl.gov/radiance/refer/ray.html#Materials.
—. n.d. Radiance WWW Server. Accessed November 1, 2014. http://radsite.lbl.gov/radiance/.
Rashwan, Ahmed, Osama Farag, and Wael Seddik Moustafa. 2014. "Energy performance analysis of
integrating building envelopes with nanomaterials." International Journal of Sustainable Built
Environment.
Reinhart, Christoph F. 2011. "Daylight performance predictions." In Building Performance Simulation for
Design and Operation, 235-271. Routledge.
Reinhart, Christoph F, and Oliver Walkenhorst. 2001. "Validation of dynamic RADIANCE-based
daylight simulations for a test office with external blinds." Energy and Buildings 683-697.
Reinhart, Christoph F, John Mardajevic, and Zack Rogers. 2006. "Dynamic Daylight Performance
Metrics for Sustainable Building Design." Leukos 3 (1): 7-31.
98
Robert McNeel & Associates. 2014. Grasshopper - algorithmic modeling for Rhino. Accessed November
1, 2014. http://www.grasshopper3d.com/.
—. 2014. Rhinoceros. Accessed November 1, 2014. http://www.rhino3d.com/.
Robinson, Darren, and Andrew Stone. 2004. "Irradiation modelling mode simple: the cumulative sky
approach and its applications." PLEA 2004 - The 21st Conference on Passive and Low Energy
Architecture. Eindhoven.
Sadineni, Suresh B., Srikanth Madala, and Robert F. Boehm. 2011. "Passive buildings energy savings: A
review of building envelope components." Renewable and Sustainable Energy Reviews 3617-
3631.
Solemma LLC. 2014. DIVA for Rhino. Accessed November 1, 2014. http://diva4rhino.com/.
Tregenza, P. 1987. "Subdivision of the sky hemisphere for luminance measurements." Lighting Research
and Technology 19: 13-14.
Vierlinger, Robert. 2014. Octopus | Grasshopper. Accessed November 1, 2014.
http://www.food4rhino.com/project/octopus?ufh.
Ward, Gregory J. 1994. "The RADIANCE Lighting Simulation and Rendering System." SIGGRAPH.
Orlando.
Wienold, Jan. 2009. "Dynamic Daylight Glare Evaluation." Eleventh International IBPSA Conference.
Glasgow. 944-951.
—. 2007. "Dynamic Simulation of Blind Control Strategies for Visual Comfort and Energy Balance
Analysis." Building Simulation 2007. 1197-1204.
99
Wienold, Jan, and Jens Christoffersen. 2006. "Evaluation methods and development of a new glare
prediction model for daylight environments with the use of CCD cameras." Energy and Buildings
743-757.
—. 2005. "Towards a new daylight glare rating." Lux Europa. Berlin.
Zelenay, Krystyna, Mark Perepelitza, and David Lehrer. 2011. "High-Performance Facades: Design
Strategies and Applications in North America and Northern Europe." Center for the Built
Environment.

100
APPENDIX: SIMULATION DATA
Figure 76. Open Office Parameters
101

Figure 77. Hotel Parameters
102

Figure 78. Closed Office Parameters
103

Figure 79. Open Office, East Parameters
104

Figure 80. Open Office, Winter Parameters
105

Figure 81. Open Office, Summer Parameters
106

Figure 82. Open Office, 2x UDI Parameters
107

Figure 83. Open Office, 3x UDI Parameters
108

Figure 84. Open Office, 2x DGPs Parameters
109

Figure 85. Open Office, 3x DGPs Parameters
110
APPENDIX: USER GUIDE

• Reference context geometry and building to be analyzed (Figure 86)

Figure 86. Context Geometry and Reference Building
• Pick out specific rooms and spaces in the building to be analyzed (Figure 87)

Figure 87. Reference Building and Outlined Rooms for Analysis
111
• Apply materials to the room, locate sensors, and orient the model in Grasshopper (Figure 88)

Figure 88. Open Office with Orientation and Measurement Sensors
• Create parametric model in Grasshopper representing shading system design intent (Figure 89)

Figure 89. Parametric Shading System Applied to Open Office
112
• Apply weighting coefficients to fitness formula and run simulations, results exported to Excel
template (Figure 90) (Figure 91)

Figure 90. Simulation Data Exported to Excel Template

Figure 91. Simulation Data Exported to Excel Template, Scatter Plots with Trend Lines
113
• Read back Excel data into Rhino to visualize results (Figure 92)

Figure 92. Re-instantiated Configuration in Rhino
114

Abstract (if available)

Abstract A tension is developing between the goal of more energy efficient buildings and the growing complexity of architecture. Conceptual design tools are needed that generate, visualize, and optimize building performance. Highly glazed enclosures offer excellent views and allow large amounts of natural light in the interior. Yet even the design of shading devices must take into account dynamic solar conditions, direct solar radiation, daylight, glare, the geometry of the building, site, glazing properties, and other factors. ❧ By providing designers with an automated method to generate shading systems, visualize the performance and daylighting quality of each unique configuration, and analyze the variable trends, better informed and more effective choices regarding façade configuration can be achieved. ❧ It is difficult to predict the quality of daylight or solar control of conceptually designed systems with so many variables affecting them—local climate, geometry, orientation, materials, glazing properties, etc. The implementation of shading systems has historically involved creating a series of designs and validating their performance through individual simulations. Therefore, a design process that optimizes the conceptual design of shading systems, but also portrays how each variable is influencing the overall performance, can better guide and inform designers. ❧ A workflow has been developed that incorporates multiple design variables to be simulated simultaneously, resulting in comprehensive results displaying both raw data to validate decisions as well as optimized façade configurations to guide designers. The proposed optimization and visualization process focuses on three separate metrics: solar radiation, useful daylight illuminance, and glare probability. Rhinoceros 3D and Grasshopper are the parametric and modeling software used as the main process interface, with DIVA included as the climate-based and daylighting simulation engine. Galapagos functions as the evolutionary solving engine used to generate and optimize different variations of shades as well as glazing properties. The results are optimized, parametrically driven shading solutions with their solar performance and daylighting quality visually displayed. Sensitivity analysis among design variables and spatial occupancy follows the optimization results, giving designers a full suite of information to inform their designs.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Effective light shelf and form finding: development of a light shelf design assistant tool using parametric methods

PDF

The intelligent control strategy of kinetic façades for daylight and energy performance: evaluating the daylight effect of adaptive systems based on parametric workflow

PDF

Adaptive façade controls: A methodology based on occupant visual comfort preferences and cluster analysis

PDF

Environmentally responsive buildings: multi-objective optimization workflow for daylight and thermal quality

PDF

Dynamic shading and glazing technologies: improve energy, visual, and thermal performance

PDF

Double skin façades performance: effects on daylight and visual comfort in office spaces

PDF

Evaluation and development of solar control screens: using daylight simulation to improve the performance of facade solar control screens

PDF

Evaluation of daylighting circadian effects: Integrating non-visual effects of lighting in the evaluation of daylighting designs

PDF

Multi-domain assessment of a kinetic facade: determining the control strategy of a kinetic façade using BIM based on energy performance, daylighting, and occupants’ preferences; Multi-domain asse...

PDF

Kinetic facades as environmental control systems: using kinetic facades to increase energy efficiency and building performance in office buildings

PDF

Changing space and sound: parametric design and variable acoustics

PDF

Aerophysics of double-skin facades: simulation-based determination of pressure coefficients for multi-story double-skin facades

PDF

ctrl+z: exploring the effects of undoing retrofits to pre-war buildings in Los Angeles

PDF

Daylight and health: exploring the relationship between established daylighting metrics for green building compliance and new metrics for human health

PDF

Daylight prediction: an evaluation of daylighting simulation software for four cases

PDF

CFD visualization: a case study for using a building information modeling with virtual reality

PDF

Structural design tool for performative building elements: a semi-automated Grasshopper plugin for design decision support of complex trusses

PDF

Development of circadian-effective toplighting strategies: using multiple daylighting performance goals for dementia care communities

PDF

Tilted glazing: angle-dependence of direct solar heat gain and form-refining of complex facades

PDF

Comparing visual comfort metrics for fourteen spaces using simulation-based luminance mapping

Asset Metadata

Creator Tucker, Tyler James (author)

Core Title Performative shading design: parametric based measurement of shading system configuration effectiveness and trends

School School of Architecture

Degree Master of Building Science

Degree Program Building Science

Publication Date 11/11/2014

Defense Date 11/03/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag daylighting,DIVA,grasshopper,multi-objective,OAI-PMH Harvest,parametric,shading

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Noble, Douglas (committee chair), Kensek, Karen M. (committee member), Konis, Kyle (committee member), Vaglio, Jeffrey (committee member)

Creator Email tylerjtucker@gmail.com,tylertuc@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-515614

Unique identifier UC11298291

Identifier etd-TuckerTyle-3076.pdf (filename),usctheses-c3-515614 (legacy record id)

Legacy Identifier etd-TuckerTyle-3076.pdf

Dmrecord 515614

Document Type Thesis

Format application/pdf (imt)

Rights Tucker, Tyler James

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA