University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ADAPTIVE FAÇADE CONTROLS A Methodology Based on Occupant Visual Comfort Preferences and Cluster Analysis by KUSHNAV ROY A Thesis Presented to the SCHOOL OF ARCHITECTURE UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF BUILDING SCIENCE AUGUST 2018 2 COMMITTEE CHAIR: Douglas E. Noble, FAIA, Ph.D. Associate Professor School of Architecture University of Southern California dnoble@usc.edu (213)740-4589 COMMITTEE MEMBER #2: Karen M. Kensek, LEED AP BD+C Associate Professor of the Practice of Architecture School of Architecture University of Southern California kensek@usc.edu (213)740-2081 COMMITTEE MEMBER #3: Kyle Konis, AIA, Ph.D. Assistant Professor School of Architecture University of Southern California kkonis@usc.edu 3 DEDICATION This work is dedicated to my parents who have always and unconditionally been by my side, to the vast sea of inspiration out there that has propelled me to successfully finish the work that I have started, to chance and time that struck the right balance, to the world and all well-wishers, and to me for truly believing in myself, mostly at times when no one else would. 4 ACKNOWLEDGEMENTS This work would not have been possible without the help of several individuals who went beyond their way to provide insight into improving my work. I would like to thank my Chair, Prof. Douglas Noble, and my committee members, Prof. Karen Kensek and Prof. Kyle Konis. Prof. Noble’s critical analysis of the work helped to expand the research extents and to understand the implications and assumptions of the work. Prof. Kensek’s role was pivotal to the growth, guidance, and realization of the work. Her contributions were essential in shaping key ideas through discussion in the initial stages of all the phases of the research. Prof. Konis’s suggestions were important in rethinking several aspects of the research in a different light and were particularly key in analysis and proposition of future work. I would like to thank Prof. Marc Schiller, Prof. Joon Ho Choi, and Prof. Jae Yong Suk, for resolving several queries regarding daylighting and glare metrics, and in clearing early concepts. For that I am grateful. I would also like to thank Sebanti Banjerjee and BuroHappold Engineering in allowing me access to drawings and documentation of their office space in Los Angeles. And lastly, I would like to thank my friends in the Master of Building Science program, for their continuous support. 5 TABLE OF CONTENTS DEDICATION .............................................................................................................................................................. 3 ACKNOWLEDGEMENTS ........................................................................................................................................... 4 LIST OF FIGURES ....................................................................................................................................................... 8 LIST OF TABLES ....................................................................................................................................................... 11 LIST OF EQUATIONS ............................................................................................................................................... 13 ABSTRACT, HYPOTHESIS, & RESEARCH OBJECTIVES ................................................................................... 14 Chapter 1 – Introduction ................................................................................................................................................ 1 1.1 Why Building Envelopes? .............................................................................................................................. 1 1.2 Adaptive Building Envelopes ......................................................................................................................... 2 1.3 COST Action TU1403 .................................................................................................................................... 3 1.4 Characterization of Adaptive Envelopes ......................................................................................................... 3 1.5 Kinetic Facades ............................................................................................................................................... 3 1.6 Adaptive Façade Controls ............................................................................................................................... 4 1.7 Problem and Proposed Solution ...................................................................................................................... 5 1.8 Summary ......................................................................................................................................................... 6 Chapter 2 – Background ............................................................................................................................................... 7 2.1 Daylighting ..................................................................................................................................................... 7 2.2 Effective Daylighting ...................................................................................................................................... 7 2.3 Barriers to Effective Daylighting .................................................................................................................... 8 2.4 Planning for Effective Daylighting ................................................................................................................. 8 2.5 Daylight Availability ....................................................................................................................................... 8 2.6 Daylighting Design ......................................................................................................................................... 9 2.7 Daylighting Design Strategies ......................................................................................................................... 9 2.7.1 Direct Solar Control ............................................................................................................................... 9 2.7.2 Uniformity ........................................................................................................................................... 10 2.7.3 Illuminance Targets ............................................................................................................................. 11 2.7.4 Energy Targets ..................................................................................................................................... 11 2.7.5 Usability & Architectural Integration .................................................................................................. 11 2.8 Visual Comfort & Performance ...................................................................................................................... 11 2.8.1 Factors for accessing visual comfort indices ....................................................................................... 12 2.8.2 Illuminance .......................................................................................................................................... 13 2.8.3 Uniformity ........................................................................................................................................... 14 2.8.4 Luminance & Contrast ......................................................................................................................... 15 2.8.5 Glare .................................................................................................................................................... 15 2.8.6 Visual Amenity .................................................................................................................................... 15 2.9 Daylight Calculation Parameters ................................................................................................................... 15 2.10 Daylight Metrics & Calculation Methods ..................................................................................................... 16 2.10.1 Illuminance Metrics ........................................................................................................................... 16 2.10.2 Lighting Uniformity Metrics .............................................................................................................. 18 2.10.3 Glare Metrics ..................................................................................................................................... 18 2.10.4 Luminance Contrast Ratio ................................................................................................................. 18 2.10.5 Occupant Preference as Metric .......................................................................................................... 19 2.11 Daylighting System Control Methods ........................................................................................................... 19 2.11.1 Model-Based Control Strategies ........................................................................................................ 19 2.11.2 Control of Shading and Redirecting Systems .................................................................................... 21 2.12 Daylighting Systems ..................................................................................................................................... 23 6 2.12.1 Light Shelves ..................................................................................................................................... 23 2.12.2 Illuminance Preference ...................................................................................................................... 26 2.13 User Preference-Based Ranking ................................................................................................................... 26 2.14 Summary ....................................................................................................................................................... 26 Chapter 3 – Methodology ............................................................................................................................................ 27 3.1 Methodology Framework .............................................................................................................................. 27 3.2 Test Space ..................................................................................................................................................... 27 3.2.1 Location ............................................................................................................................................... 27 3.2.2 Climate Data ........................................................................................................................................ 28 3.2.3 Reference Building .............................................................................................................................. 28 3.2.4 Physical Context .................................................................................................................................. 28 3.3 Test Space Selection ..................................................................................................................................... 28 3.4 Model Simplifications ................................................................................................................................... 30 3.5 Software Platform ......................................................................................................................................... 31 3.6 Control Logic ................................................................................................................................................ 33 3.7 Parametric Definition .................................................................................................................................... 33 3.7.1 Workflow Modules - Grasshopper ...................................................................................................... 33 3.7.2 Workflow Modules - Excel ................................................................................................................. 34 3.7.3 Input ..................................................................................................................................................... 34 3.7.4 Daylighting System ............................................................................................................................. 37 3.7.5 Occupant Data ..................................................................................................................................... 39 3.7.6 Simulation ............................................................................................................................................ 40 3.7.7 Excel Write .......................................................................................................................................... 40 3.7.8 Calculation ........................................................................................................................................... 40 3.7.9 Data Analysis ....................................................................................................................................... 47 3.7.10 Visualization ........................................................................................................................................ 48 3.8 Assumptions .................................................................................................................................................. 49 3.9 Scope ............................................................................................................................................................. 50 3.10 Methodology Steps ....................................................................................................................................... 51 3.11 Summary ....................................................................................................................................................... 51 Chapter 4 – Results ...................................................................................................................................................... 52 4.1 Occupant Preference Cases ............................................................................................................................. 52 4.1.1 Case 1: Equal Preferences and Importance Factors ........................................................................... 52 4.1.2 Case 2: Equal Preferences, High Importance Factor for Glare .......................................................... 52 4.1.3 Case 3a: Equal Preferences, High Importance Factor for Illuminance (500lx) .................................. 53 4.1.4 Case 3b: Equal Preferences, High Importance Factor for Illuminance (2000lx) ............................... 53 4.1.5 Case 3c: Equal Preferences, High Illuminance Importance Factor (unequal ranges) ........................ 54 4.1.6 Case 4a: Equal Preferences, High Importance Factor for Contrast Ratio (1.0) .................................. 54 4.1.7 Case 4b: Equal Preferences, High Importance Factor for Contrast Ratio (3.0) ................................. 54 4.1.8 Case 4c: Equal Preferences, High Importance Factor for Contrast Ratio (unequal ranges) ............... 55 4.1.9 Case 5: Randomly Generated Occupant Preferences and Importance Factors .................................. 55 4.1.10 Case 6: Assumed Worst Case Scenario ............................................................................................. 56 4.2 Result Framework ......................................................................................................................................... 56 4.3 June 21, 9am ................................................................................................................................................. 60 4.4 June 21, noon ................................................................................................................................................ 63 4.5 June 21, 3pm ................................................................................................................................................. 66 4.6 September 22, 9am ........................................................................................................................................ 69 4.7 September 22, noon ....................................................................................................................................... 72 4.8 September 22, 3pm ....................................................................................................................................... 75 4.9 December 22, 9am ........................................................................................................................................ 78 4.10 December 22, noon ....................................................................................................................................... 81 4.11 December 22, 3pm ........................................................................................................................................ 84 7 4.12 Summary ....................................................................................................................................................... 87 Chapter 5 – Discussion ................................................................................................................................................ 88 5.1 Discussion Framework ................................................................................................................................... 88 5.2 Assumptions ................................................................................................................................................... 89 5.3 Cross-Comparisons ......................................................................................................................................... 92 5.4 Evaluation of Overall Visual Comfort Satisfaction ........................................................................................ 93 5.5 Façade Configuration Cluster Patterns ........................................................................................................... 95 5.6 Summary ....................................................................................................................................................... 102 Chapter 6 – Conclusions ............................................................................................................................................ 103 6.1 Conclusion Framework ................................................................................................................................. 103 6.2 Tangential Findings ...................................................................................................................................... 103 6.3 Future Work .................................................................................................................................................. 104 6.3.1 Improvements to Methodology ........................................................................................................ 104 6.3.2 Research Possibilities....................................................................................................................... 104 References ................................................................................................................................................................. 106 8 LIST OF FIGURES Figure 1. Façade energy impacts in U.S. commercial building sector .......................................................................... 1 Figure 2. Schematic role of an adaptive façade ............................................................................................................. 2 Figure 3. Overview of characterization concepts for envelope adaptivity ..................................................................... 3 Figure 4. Vertical axis glazed sun louvers, The Oval Offices in Cologne ..................................................................... 4 Figure 5. Control system diagram ................................................................................................................................. 5 Figure 6. Solar shading types and classification .......................................................................................................... 10 Figure 7. Cumulative number of visual comfort indices over time ............................................................................. 12 Figure 8. Daylighting (illuminance-based) metric definitions ..................................................................................... 17 Figure 9. Glare discomfort index comparison on a nine-point glare sensation scale ................................................... 18 Figure 10. Model-based control strategies ................................................................................................................... 20 Figure 11. Flowchart of solar protection control strategy ............................................................................................ 22 Figure 12. Different blind positions between the previous and the new control method ............................................. 23 Figure 13. Semi-transparent double light shelves made of reflective glass ................................................................. 23 Figure 14. Top and bottom sections of interior and exterior light shelves with specular surfaces .............................. 23 Figure 15. Exterior semi-reflective light shelf in a downward-tilted position ............................................................. 24 Figure 16. Interior view of the test room with a semi-transparent interior light shelf ................................................. 24 Figure 17. The Variable Area Light Reflecting Assembly (VALRA) ......................................................................... 25 Figure 18. Illuminance preference setpoints ................................................................................................................ 26 Figure 19. Scope and methodology framework ........................................................................................................... 27 Figure 20. Location of 800 Wilshire, the reference building ....................................................................................... 28 Figure 21. Reference building and physical context .................................................................................................... 29 Figure 22. Floor plan of the office of BuroHappold Engineering and the test space ................................................... 29 Figure 23. The south-east corner test space extents ..................................................................................................... 30 Figure 24. Original test space and simplifications ....................................................................................................... 31 Figure 25. Proposed software workflow ...................................................................................................................... 32 Figure 26. Proposed control logic ................................................................................................................................ 33 Figure 27. Modules forming the parametric definition ................................................................................................ 34 Figure 28. Implemented parametric definition in Grasshopper ................................................................................... 34 Figure 29. Sky types .................................................................................................................................................... 35 Figure 30. Occupant viewpoint location ...................................................................................................................... 36 Figure 31. Visual display unit ...................................................................................................................................... 36 Figure 32. Proposed daylighting system ...................................................................................................................... 38 Figure 33. Excel computation and analysis ................................................................................................................. 41 Figure 34. Excel facilitator tool: Data collection sample ............................................................................................. 47 Figure 35. Automated visualization ............................................................................................................................. 48 Figure 36. A close-up view of the tog-based visualization .......................................................................................... 48 9 Figure 37. Annotated diagram explaining the visualization components .................................................................... 48 Figure 38. Randomly generated occupant preference model case (seed value=5500) ................................................ 49 Figure 39. Occupant preference and importance factor (Case 1) ................................................................................. 52 Figure 40. Equal preferences, high glare importance factor (Case 2) .......................................................................... 52 Figure 41. Equal preferences, high importance factor for illuminance (500 lx, Case 3a) ............................................ 53 Figure 42. Equal preferences, high importance factor for illuminance (2000lx, Case 3b) .......................................... 53 Figure 43. Equal preferences, high importance factor for illuminance (randomly generated, Case 3c) ...................... 54 Figure 44. Equal preferences, high importance factor for contrast ratio (Case 4a) ..................................................... 54 Figure 45. Equal preferences, high importance factor for contrast ratio (Case 4b) ..................................................... 55 Figure 46. Equal preferences, high importance factor for contrast ratio (Case 4c) ..................................................... 55 Figure 47. Randomly generated occupant preferences and importance factors (Case 5) ............................................ 56 Figure 48. Assumed worst case scenario (Case 6) ....................................................................................................... 56 Figure 49. Result framework and organization ............................................................................................................ 57 Figure 50. Visual representation of façade configuration and visual comfort condition (June 21, 9am) .................... 61 Figure 51. Cluster patterns for Case 2 ......................................................................................................................... 62 Figure 52. Cluster patterns for Case 3b ....................................................................................................................... 62 Figure 53. Visual representation of façade configuration and visual comfort condition (June 21, noon) ................... 64 Figure 54. Cluster patterns for Case 2 ......................................................................................................................... 65 Figure 55. Cluster patterns for Case 3b ....................................................................................................................... 65 Figure 56. Visual representation of façade configuration and visual comfort condition (June 21, 3pm) .................... 67 Figure 57. Cluster patterns for Case 2 ......................................................................................................................... 68 Figure 58. Cluster patterns for Case 3b ....................................................................................................................... 68 Figure 59. Visual representation of façade configuration and visual comfort condition (September 22, 9am) ........... 70 Figure 60. Cluster patterns for Case 2 ......................................................................................................................... 71 Figure 61. Cluster patterns for Case 3b ....................................................................................................................... 71 Figure 62. Visual representation of façade configuration and visual comfort condition (September 22, noon) .......... 73 Figure 63. Cluster patterns for Case 2 ......................................................................................................................... 74 Figure 64. Cluster patterns for Case 3b ....................................................................................................................... 74 Figure 65. Visual representation of façade configuration and visual comfort condition (September 22, 3pm) .......... 76 Figure 66. Cluster patterns for Case 2 ......................................................................................................................... 77 Figure 67. Cluster patterns for Case 3b ....................................................................................................................... 77 Figure 68. Visual representation of façade configuration and visual comfort condition (December 22, 9am) ........... 79 Figure 69. Cluster patterns for Case 2 ......................................................................................................................... 80 Figure 70. Cluster patterns for Case 3b ....................................................................................................................... 80 Figure 71. Visual representation of façade configuration and visual comfort condition (December 22, noon) .......... 82 Figure 72. Cluster patterns for Case 2 ......................................................................................................................... 83 Figure 73. Cluster patterns for Case 3b ....................................................................................................................... 83 10 Figure 74. Visual representation of façade configuration and visual comfort condition (December 22, 3pm) ........... 85 Figure 75. Cluster patterns for Case 2 ......................................................................................................................... 86 Figure 76. Cluster patterns for Case 3b ....................................................................................................................... 86 Figure 77. Maximum Percentage Reduction in D CL (P r ) for the nine points in time ................................................... 93 Figure 78. Median Percentage Reduction in D CL (P r ) for the nine points in time ........................................................ 93 Figure 79. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, June 21 (9am) ................... 96 Figure 80. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, June 21 (9am) ................. 96 Figure 81. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, June 21 (noon) .................. 96 Figure 82. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, June 21 (noon) ................ 97 Figure 83. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, June 21 (3pm) ................... 97 Figure 84. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, June 21 (3pm) ................. 97 Figure 85. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, September 22 (9am) ......... 98 Figure 86. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, September 22 (9am) ....... 98 Figure 87. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, September 22 (noon) ........ 99 Figure 88. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, September 22 (noon) ...... 99 Figure 89. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, September 22 (3pm) ......... 99 Figure 90. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, September 22 (3pm) ..... 100 Figure 91. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, December 22 (9am) ........ 100 Figure 92. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, December 22 (9am) ...... 100 Figure 93. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, December 22 (noon) ....... 101 Figure 94. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, December 22 (noon) ..... 101 Figure 95. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 2, December 22 (3pm) ........ 102 Figure 96. Scatterplot of Percentage Reduction & Performance Score vs Index, Case 3b, December 22 (3pm) ...... 102 11 LIST OF TABLES Table 1. Radiance simulation parameters .................................................................................................................... 37 Table 2. Optical material properties ............................................................................................................................ 37 Table 3. Example of preference modeling for three occupants ................................................................................... 40 Table 4. Control variables and control values ............................................................................................................. 43 Table 5. Occupant Preference and Importance Factor Variables, Abbreviations, Possible Ranges & Values ............ 43 Table 6. Simulation points-in-Time and Occupant Preference Cases .......................................................................... 46 Table 7. Proposed metrics, variables, and abbreviations ............................................................................................. 46 Table 8. Summary of proposed metrics, variables, and abbreviations ......................................................................... 47 Table 9. Maximum Percentage Reduction ................................................................................................................... 57 Table 10. Distribution of Percentage Reduction .......................................................................................................... 58 Table 11. Top Five Ranked Configurations by Case ................................................................................................... 58 Table 12. Summary of Nomenclature .......................................................................................................................... 59 Table 13. Summary of Results by Case (June 21, 9am) .............................................................................................. 60 Table 14. Distribution of Percentage Reduction by Case (June 21, 9am) ................................................................... 61 Table 15. Top Five Ranked Configurations by Case (June 21, 9am) .......................................................................... 62 Table 16. Summary of Results by Case (June 21, noon) ............................................................................................. 63 Table 17. Distribution of Percentage Reduction by Case (June 21, noon) .................................................................. 64 Table 18. Top Five Ranked Configurations by Case (June 21, noon) ......................................................................... 65 Table 19. Summary of Results by Case (June 21, 3pm) .............................................................................................. 66 Table 20. Distribution of Percentage Reduction by Case (June 21, 3pm) ................................................................... 67 Table 21. Top Five Ranked Configurations by Case (June 21, 3pm) .......................................................................... 68 Table 22. Summary of Results by Case (September 22, 9am) .................................................................................... 69 Table 23. Distribution of Percentage Reduction by Case (September 22, 9am) .......................................................... 70 Table 24. Top Five Ranked Configurations by Case (September 22, 9am) ................................................................. 71 Table 25. Summary of Results by Case (September 22, noon) ................................................................................... 72 Table 26. Distribution of Percentage Reduction by Case (September 22, noon) ......................................................... 73 Table 27. Top Five Ranked Configurations by Case (September 22, noon) ............................................................... 74 Table 28. Summary of Results by Case (September 22, 3pm) .................................................................................... 75 Table 29. Distribution of Percentage Reduction by Case (September 22, 3pm) .......................................................... 76 Table 30. Top Five Ranked Configurations by Case (September 22, 3pm) ................................................................ 77 Table 31. Summary of Results by Case (December 22, 9am) ..................................................................................... 78 Table 32. Distribution of Percentage Reduction by Case (December 22, 9am) .......................................................... 79 Table 33. Top Five Ranked Configurations by Case (December 22, 9am) ................................................................. 80 Table 34. Summary of Results by Case (December 22, noon) .................................................................................... 81 Table 35. Distribution of Percentage Reduction by Case (December 22, noon) ......................................................... 82 Table 36. Top Five Ranked Configurations by Case (December 22, noon) ................................................................ 83 12 Table 37. Summary of Results by Case (December 22, 3pm) ..................................................................................... 84 Table 38. Distribution of Percentage Reduction by Case (December 22, 3pm) .......................................................... 85 Table 39. Top Five Ranked Configurations by Case (December 22, 3pm) ................................................................. 86 Table 40. Example of Cross-Comparison of Maximum Percentage Reduction (P r ) Values ........................................ 87 Table 41. Cross-Comparison of Maximum Percentage Reduction (P r ) Value for all Nine Points in Time ................. 92 Table 42. Cross-Comparison of Maximum Percentage Reduction (P r ) Cases for all Nine Points in Time ................. 92 Table 43. Overall Visual Comfort Satisfaction - Relevant Summary of Results for June 21, 9am ............................. 94 Table 44. Overall Visual Comfort Satisfaction - Relevant Summary of Results for June 21, noon ............................ 94 Table 45. Overall Visual Comfort Satisfaction - Relevant Summary of Results for June 21, 3pm ............................. 94 Table 46. Overall Visual Comfort Satisfaction - Relevant Summary of Results for September 22, 9am ................... 94 Table 47. Overall Visual Comfort Satisfaction - Relevant Summary of Results for September 22, noon .................. 94 Table 48. Overall Visual Comfort Satisfaction - Relevant Summary of Results for September 22, 3pm ................... 94 Table 49. Overall Visual Comfort Satisfaction - Relevant Summary of Results for December 22, 9am .................... 94 Table 50. Overall Visual Comfort Satisfaction - Relevant Summary of Results for December 22, noon ................... 94 Table 51. Overall Visual Comfort Satisfaction - Relevant Summary of Results for December 22, 3pm .................... 94 Table 52. Façade Configuration Cluster Patterns- Relevant Summary of Results for June 21, 9am ........................... 95 Table 53. Façade Configuration Cluster Patterns- Relevant Summary of Results for June 21, noon .......................... 96 Table 54. Façade Configuration Cluster Patterns- Relevant Summary of Results for June 21, 3pm ........................... 97 Table 55. Façade Configuration Cluster Patterns- Relevant Summary of Results for September 22, 9am ................. 98 Table 56. Façade Configuration Cluster Patterns- Relevant Summary of Results for September 22, noon ................ 98 Table 57. Façade Configuration Cluster Patterns- Relevant Summary of Results for September 22, 3pm ................. 99 Table 58. Façade Configuration Cluster Patterns- Relevant Summary of Results for December 22, 9am ................ 100 Table 59. Façade Configuration Cluster Patterns- Relevant Summary of Results for December 22, noon ............... 101 Table 60. Façade Configuration Cluster Patterns- Relevant Summary of Results for December 22, 3pm................ 101 13 LIST OF EQUATIONS Equation 1. Daylight Glare Probability metric calculation .......................................................................................... 12 Equation 2. Useful Daylight Illuminance .................................................................................................................... 16 Equation 3. Frequency of Visual Comfort ................................................................................................................... 17 Equation 4. Intensity of Visual Discomfort ................................................................................................................. 17 Equation 5. Illuminance Uniformity ............................................................................................................................ 18 Equation 6. The full Daylight Glare Probability equation ........................................................................................... 20 Equation 7. Simplified glare calculation ...................................................................................................................... 21 Equation 8. Supplementary illuminance requirement from artificial lighting ............................................................. 21 Equation 9. Cumulative Dissatisfaction as sum of Individual Dissatisfaction values ................................................. 42 Equation 10. Individual Glare Dissatisfaction ............................................................................................................. 44 Equation 11. Individual Illuminance Dissatisfaction ................................................................................................... 44 Equation 12. Individual Contrast Ratio Dissatisfaction ............................................................................................... 44 Equation 13. Individual Dissatisfaction ....................................................................................................................... 44 Equation 14. Reduction in Cumulative Dissatisfaction ............................................................................................... 44 Equation 15. Percentage Reduction in Cumulative Dissatisfaction ............................................................................. 44 Equation 16. Façade Configuration Performance Score .............................................................................................. 45 Equation 17. Façade Configuration Performance Score for configuration ranking and clustering ............................. 59 14 ABSTRACT A static daylighting façade system creates varying and uncontrolled levels of daylighting and visual conditions in a space. These visual comfort conditions are evaluated based on standard requirements or averaged visual comfort preference ranges that might not reflect true individual preferences. Such a control strategy therefore fails to address stochastic phenomena, such as changing, and often competing, individual visual comfort requirements and bias, and factors such as occupant position and viewing direction in a space. An adaptive daylighting façade system can consist of multiple independently-controllable components that are driven by a control logic which determines its control states. Using an objective function and through quantification of the required features, a control logic can assist in finding one or more possible daylighting system configurations that most closely produce certain predetermined criteria, such as glare or illuminance. For a multiple-occupancy condition, a set of individual and cumulative dissatisfaction indices were proposed that represented and helped quantify how effective an adaptive façade configuration was in producing the occupant-preferred sets of visual comfort performance. In a case study, adaptive facade configurations at 15-minute intervals were ranked based on their performance, finding control states that affected the results the most, developing and improving daylighting control strategies, and providing a more personal and generally satisfactory and controlled daylit environment. The dynamic real-time control logic framework accounted for three visual comfort factors: average task surface illuminance, glare, and task surface to background luminance contrast ratio. The south-east corner of an existing open-plan office in downtown Los Angeles was modeled with electrochromic vision glazing, kinetic shading, and light shelves. Using the Grasshopper and Honeybee interfaces, daylighting simulations under climate-based sky conditions were performed for three times daily for three days of the year to evaluate the proposed framework. The framework integrated a simplified occupant preference model with a simplified glare calculation method and a Radiance-based illuminance-luminance simulation workflow, using calculation and data analysis modules. Python was used to create a normalized distribution of personal occupant visual comfort ranges based on predefined lists and to generate flat importance factors, for glare, illuminance, and contrast ratio. The variations of each measured occupant comfort value from the preferred range, weighted by their respective importance factors, was calculated to provide a net deviation value for each occupant. A brute force simulation method was used to generate all possible façade configurations, and the respective visual comfort factor deviation values were then recorded, analyzed, and compared against a baseline case. The proposed framework substantially increased the overall as well as individual percentage occupant visual satisfaction in a multiple occupancy condition in comparison with the baseline case. A substantial reduction of cumulative deviation from the baseline case was observed for certain façade configurations, thus validating that the proposed control framework can be used to optimize façade control states that improve the overall occupant satisfaction with the visual environment by making more people more satisfied with their visual task environment. Analyses of the control states of the high performing façade configurations were useful in developing daylighting control strategies and to find adaptive façade elements and states that have the most effect on the visual comfort condition at different times of the year. Hypothesis Using a real-time control algorithm for the operation of adaptive façade daylighting systems in a multiple occupancy condition, the cumulative dissatisfaction with the visual comfort condition can be reduced by as much as 30 percent, based on occupant-chosen preferences and importance factors. In addition, analysis of the calculated dissatisfaction data can be used to develop adaptive façade control strategies and to rank and cluster façade configuration options based on their performance. Research Objectives • To develop a numerical technique to consolidate multiple visual comfort factors into a single value based on preset occupant visual comfort preferences and importance factors. • To validate that adaptive facades can improve the overall as well as individual visual comfort satisfaction and by as much as 30 percent compared to a static façade baseline case. • To develop a technique to rank and cluster adaptive façade configurations, based on how well they perform in reducing overall cumulative visual comfort dissatisfaction, for use by architects and façade consultants. • To determine adaptive façade control states and elements which have a higher impact on reducing overall visual comfort dissatisfaction. • To develop control strategies that will determine the of the dynamic façade components under varying annual daylight conditions for a multi-occupancy conditions. CHAPTER 1 1. INTRODUCTION Adaptive facades have the potential to improve indoor environmental conditions and building performance to a greater extent compared to static facades. This chapter introduces the problem in greater detail and strives to establish initial ideas and proposals of a methodology for designing and controlling adaptive facades in a way that improves overall and individual visual comfort conditions in open-plan offices. 1.1. Why Building Envelopes? Buildings are estimated to be responsible for more than 32% of total global energy consumption with operational energy constituting almost one-third of global black carbon emissions (Lucon et al. 2014). In Europe, building operational energy comprises 30% of all end user energy consumption. Amongst all building systems, the building envelope plays the most important role in determining overall building performance as well as the indoor environmental conditions, behaving essentially as the foremost environmental regulator. For example, in cooling- dominated climates, a significant portion of the net solar heat gain takes place through the envelope making a building heavily reliant on the HVAC system to optimize indoor environmental quality (IEQ) and cooling energy demand. In some cases, it has been observed that the application of dynamic and integrated facades has resulted both in a substantial decrease in HVAC loads and added to the façade’s daylighting potential (Fig. 1). Static facades are only slightly better than the facades of current stock of buildings. In comparison, low-e glazed facades perform much better only to be outperformed by dynamic, highly-insulated dynamic, and integrated facades. Figure 1: Facade Energy Impacts in U.S. Commercial Building Sector (Konis and Selkowitz 2017). Data from (Apte and Arasteh 2006). The indoor environmental quality (IEQ) of a building is dependent on a multitude of indoor environmental conditions, such as indoor air temperature, mean radiant temperature, lighting, glare, acoustics, air quality, and relative humidity, as well as human conditions such as activity, physiology, and clothing, and the cumulative effect of these conditions determines the overall comfort and well-being directly perceived by an occupant (Al horr et al. 2016). Building IEQ and performance interchangeably take precedence in the design approach. The focus of improving the IEQ of a building has essentially been to enhance occupant living conditions while adhering to minimum performance standards or code compliancy, whereas performance-based design has shifted the design focus from integrating satisfactory IEQ with effective performance measures, to maximizing building energy performance. In either of the approaches, building envelope design is important. It can be concluded from research that in commercial buildings in the United States, building envelope fenestration design and decision-making accounts for the highest fraction of total end energy use (U.S. Energy Information 2013). This conclusion was supported by a finding that in office buildings, façade design and performance greatly affects artificial lighting and HVAC loads, and that the latter two cumulatively constituting the largest share of load (U.S. Energy Information 2013). To draw impactful solutions in the quest to achieving net positive and healthy buildings, it is therefore essential that building envelopes of new as well as existing buildings receive our continued attention. In terms of occupant comfort, design decisions of the envelope directly affect indoor lighting and thermal conditions which also feature as the most common areas of occupant discomfort in the workplace. Therefore, the process of decision-making for envelope design must factor in both performance as well as comfort requirements. 2 1.2. Adaptive Building Envelopes In consideration of meeting both performance as well as occupant comfort, adaptive façade systems are a more viable solution than their static counterparts. Adaptive façade systems outperform fixed façade systems by responding “intelligently” to changing environmental conditions. Their intrinsic nature of materially or kinetically morphing to changing conditions gives them an opportunity to improve or maintain relatively higher performance levels throughout their operational period. Using non-adaptive facades, Ideally, to meet multiple objectives of performance and comfort simultaneously at any given time, the façade would need to be designed to adapt and change in real-time. With the advent of curtain wall technology featuring highly glazed envelopes, the need to devise ways of maintaining indoor environmental quality and user comfort became paramount. Even the high-performing glazed facades are still low on overall insulation values when compared to opaque envelopes that feature a lower window-wall ratio. For example, a basic wood stud 2x4 construction at 16-inch o.c. has an assembly U-factor of 0.096, whereas an argon filled double glazed low-e IGU (3mm-13mm-3m) has a higher U-value of 0.30. There have been, however, benefits and shortcomings with new glazing technologies- increased daylighting but more heat inflow and potential visual disability and discomfort arising from direct and indirect glare sources (Giovannini et al. 2015). Adaptive building envelopes are designed to undergo local or global changes in material properties, form, systems, components, or behavior, responding in real-time to surrounding internal as well as environmental conditions, and occupant behavior. Unlike fixed envelope systems, these are designed to leverage the additional performance- based benefits resulting from their ability to respond in real-time to multiple environmental factors (Fig. 2). Figure 2: Schematic role of an adaptive façade (Aelenei, Aelenei, and Vieira 2016). Adaptive envelopes should not be confused with intelligent building systems. Smart or intelligent building systems feature additional integrated abilities to learn and evolve through performance and user comfort feedback loops. They might be considered as a specialized sub-classification of adaptive facades displaying a control logic that modifies itself based on accumulated user data, real-time weather information, or occupant behavior patterns. A literature review of existing long-term research found that the energy performance and long-term economic benefits of adaptive building envelope systems outperform their static, fixed counterparts (Konstantoglou and Tsangrassoulis 2016). Static facades are not necessarily unadaptable. There has been substantial research on adaptive envelopes that has brought forward a range of concepts, innovations, and emerging solutions that have the potential to fuel the design of new envelope systems. A kinetic façade is a specific category of adaptive envelopes that features kinetic, actuated components. One can also differentiate between kinetic and responsive envelopes based on their scope of intervention. As Edward Peck, AIA, of Thornton Tomasetti puts it, “I tend to use kinetics when it’s the big picture movement, and then active or dynamic when it’s components within a system” (Enclos.com, 2017). Accordingly, kinetic envelope components might be wholly restricted to translational, rotational, or motion in general (Mahmoud and Elghazi 2016). Adaptive components could extend the possibilities to options such as additive insulation, phase-change materials, modularization, multi- functionality, or vegetated facades, often guided by algorithmic control, generative design, or even biomimetic design 3 approaches (Aelenei, Aelenei, and Vieira 2016). Several adaptive façade technologies are currently being researched and developed, such as at the Façade Research Group at the Delft University of Technology (TU Delft). Research concludes that the primary objective of such adaptable envelopes remains the same: the reversible control of environmental and man-made effects such as blast impacts, noise and fires, ideally throughout the life-cycle of the building with the aim to improve the overall building performance in terms of human conditions- comfort, interaction, aesthetics, and energy-efficiency parameters (Wang, Beltrán, and Kim 2012). 1.3. COST Action TU1403 There is substantial current research and a growing scientific community supporting adaptive facades, especially supported by the European Cooperation in Science and Technology’s COST Action TU1403 Adaptive Façade Network that comprises of an international scientific community that is jointly contributing to research in this area. There are three working research groups in the organization and their roles are divided into technology development, characterization methods, and whole-life evaluation methods, as well as user integration (Nextfacades.eu, 2017). Of its multiple objectives, COST TU1403 seeks to support development of new approaches to adaptive envelope design as well as establishing new metrics for evaluating its performance. An extensive list of current research being undertaken by the organization can be found in the Adaptive Facade Network (http://tu1403.eu/?page_id=209). 1.4. Characterization of Adaptive Envelopes One step towards adopting adaptive envelopes is their systematic characterization based on multiple parameters, of which the purpose of use is the most important. An example of such a classification matrix would include characterization parameters or concepts such as purpose, responsive function, operation, technologies (materials and systems), spatial scale, visibility, and the degree of adaptability. Thermal comfort, indoor-air quality, visual performance (illuminance, glare, and view), acoustic quality, energy generation, and personal control, are the final goals or purpose of adaptive systems (Fig. 3). Figure 3: Overview of characterization concepts for envelope adaptivity (Aelenei, Aelenei, and Vieira 2016) and (Loonen et al. 2015). In most cases it has been found that adaptive technologies are integrated into the design as a visual element. The research also found that some of the important adaptive envelope technologies are dynamic exterior shading, electrochromic glazing, double-skin facades, and photovoltaics, among others (Aelenei, Aelenei, and Vieira 2016). 1.5 Kinetic Facades Kinetic facades are adaptive envelopes that have movable, actuating components that dynamically respond to changing external environmental conditions (Mahmoud and Elghazi 2016). Kinetic facades are not a new concept, but the purpose of their design has changed and evolved over time. Passively-controlled window and daylighting systems, 4 such as shutters and blinds, have been around for a long time. Depending primarily on the climate, these systems were designed to be able to be manually regulated to respond to environmental conditions. Some well-known examples of actively controlled kinetic facades were used for aesthetic and social purposes rather than environmental control. Examples of such media facades would be the BIX façade at Kuntshaus Graz designed by Realities United and Electroland’s Enteractive Façade. Others, such as the infamous Institut du Monde Arabe, designed by Jean Nouvel, having automated sensor-based external environmental controls, were simple interpretations and almost completely mechanically reliant solutions otherwise that neither focused on more than one environmental aspect, usually solar radiation control, nor tackled the complexities involved in resolving environmental and human comfort requirements. Another example is The Oval Offices in Cologne, displaying vertical axis glazed sun louvers (Fig. 4). Figure 4: Vertical axis glazed sun louvers, The Oval Offices in Cologne (arch. Sauerbruch and Hutton, 2010). Adaptable facades are not to be confused with flexible or polyvalent facades, the latter constituting ones designed to serve a multitude of functions or to take up a range of roles throughout the life of the building (Leupen 2005). An example of a flexible system is the curtain wall system. Curtain walls can be upgraded or removed depending on the change in building use in time, but in some cases, they might neither wholly respond to changing environmental conditions in shorter time periods nor show dynamic operation of components. Polyvalent facades, however, for example media facades, are ones that can be interpreted differently depending on the user. Adaptable facades lie somewhere in between polyvalent and flexible façade systems. Media facades are therefore, in a way, kinetic, but their purpose is almost entirely aesthetic or multi-sensory based on preset or live data input, but do not serve any performance-based benefits. Depending on the nature of their integration with the building, kinetic architectural components can be dynamic, deployable or embedded systems. Kinetic facades and their components are essentially embedded systems with locally defined kinetics. 1.6 Adaptive Façade Controls The responsive action of some adaptive, especially kinetic facades and technologies like switchable glazing, is inherently controlled by a control algorithm (Konstantoglou and Tsangrassoulis 2016). Often, this façade control is integrated with other building systems through an array of sensors and control devices through the building 5 management system (Fig. 5). The control algorithm can be simple or complex, but it is essentially based on several design and performance variables and seeks to find a set of better performing combinations of these parameters for control of the dynamic façade components. Figure 5: Control system diagram (Konstantoglou and Tsangrassoulis 2016) 1.7 Problem & Proposed Solution An adaptive control algorithm is usually autonomous, which means that it works independently based on performance variables, but without any substantial active intervention from users. Different users in the same space might have varying comfort requirements and studies have shown that it is a significant reason why users do not respond very well to kinetic facades. It has been found that often, users have been reported to manually override systems to meet their personal workspace comfort requirements (Konstantoglou and Tsangrassoulis 2016). Ideally, the façade not only needs to respond well to the comfort requirements of multiple occupants, but it also needs to meet standard energy performance criteria. For example, an occupant might want a higher light level at their task surface than available and would prefer the façade be more transparent to daylight penetration and outdoor views. This would have two implications: the increase in illumination levels might exceed the acceptable range of illumination levels of other adjacent occupants resulting in conflict, subsequent negotiation, and adjustment of the façade to optimize visual comfort of the affected occupants. This is at the scale of a single occupant whose visual comfort condition is being optimized based on their individual preferences and the preferences of other occupants. In the case of multiple occupants using a given daylit space, this scenario is iterated for every occupant versus the other occupants. It would then be necessary not just to control the façade based on standard criteria but according to individual preferences of all occupants optimized to reach a façade configuration that is acceptable to all occupants. This is an ideal condition when every occupant’s preference is met completely. For example, assuming only illuminance as a visual comfort factor and that there are three occupants, in a space daylit by an adaptive façade, who have preferred illuminance levels of 500lx, 650lx, and 950lx at their task surfaces, the ideal condition would be such that the façade is configured such that the occupants get the exact illuminance setpoints that they desire. However, with a limited number of façade configurations and closely seated occupants, such as in an office space, it was understood that it might not be practically feasible to provide each occupant with their exact preferred light level, especially if the preferred values differed by a lot for adjacent occupants. This statement also holds true in the case of multiple visual comfort factors. For example, the visual comfort condition in a daylit space not only depends on the incident light levels on the various surfaces, but also on its distribution, the luminance contrast ratios produced in between the foreground to the background surface, direct or indirect glare, among other factors. To be able to say that an occupant’s visual comfort condition is ideal and acceptable, all these factors would need to be individually acceptable. In the case of multiple occupants in an open floor plan space, such as an open plan office, these factors would need to be measured for each occupant with respect to their seating location, viewing orientation, and eye level. To make each occupant completely satisfied with their individual visual comfort condition, it would be necessary to adjust the façade through its various states till the requirement is met. As stated previously, it was observed that it is mostly not possible to completely meet the set of requirements, for example, the individual visual comfort preferences set by the occupants. 6 The adaptive façade comprises of certain elements that adapt and adjust based on a control logic or algorithm that sets its state based on the solar condition, time of day, and other predefined factors. Each of these façade components or elements can take up a different configuration, called a control state. For example, an adaptive façade may comprise of electrochromic glazing, which can take up multiple tint shades. These tint shades would then be called the control states of the electrochromic glazing. Depending on the façade control logic adopted, the façade would need to take up a different tint shade. For example, if there is a high level of glare in a daylit space, then the control logic may inform the façade controller that the tint of the electrochromic glazing be changed to one with a lower visible light transmittance to reduce the probability of discomfort associated with direct solar glare. The daylighting condition in a space would depend on the exact façade configuration and its constituent control states. If a facade daylighting system has fewer number of control states, then for a given point in time, there will also be fewer resulting unique daylighting conditions. If a façade is comprised of multiple adaptable elements, say 2, and each of which can take up certain number of control states, say ‘a’ and ‘b,’ then the façade itself is said to have ‘a x b’ possible unique configurations. For each such configuration, the daylighting condition would theoretically be unique. Therefore, theoretically, the higher the number of possible unique configurations a façade is designed to take, the more accurately can the daylighting condition in the space be matched to the occupant requirement. However, to make this possible, the number of possible unique façade configurations would need to be extremely large. From a practical standpoint, this is not feasible. Several practical assumptions need to be made to adopt this ideal methodology in a working system. First, the number of unique façade configurations would need to be limited and of an order which can be computed and controlled by a façade controller. The occupant visual comfort preference is not very meaningful as a single setpoint value as not only is a single setpoint hard to create by changing the façade configuration, but also because an occupant’s perception and preference is usually defined by a range of values rather than a single value. Therefore, limited number of unique façade configurations were used along with occupant preference ranges, instead of single setpoint values. As stated previously, it is not practically feasible to simultaneously reach exact visual comfort factor values for each occupant. Therefore, instead of checking for an exact match, the closeness of the produced value to the expected value is measured and minimized. Here it is assumed that for an occupant, the closer the measured or calculated value to the expected preference range midpoint, the closer the individual visual comfort condition is to complete satisfaction. In a real scenario, if the calculated value, for example, the illuminance level on a task surface, is not within the desired preference range, the visual comfort condition might be completely unsatisfactory to an occupant. However, it has been assumed that the visual satisfaction increases linearly with an increase of the closeness value. The visual comfort satisfaction of a single occupant is not representative of the conditions for all the other occupants in a multiple-occupancy condition. The proposed control logic has been designed to consider multiple visual comfort satisfaction values and to optimize the façade configuration based on this cumulative satisfaction value. At any given point in time, different façade configurations will produce a different cumulative and individual satisfaction values. Analyzing and ranking the configurations based on how well they can cumulatively and individually satisfy occupants in a given space is also useful. Furthermore, depending on this ranking, it was also possible to find which control states for the high-performing façade configurations produced the highest impact on cumulative and individual satisfaction. 1.8 Summary Chapter 1 introduced kinetic facades, highlighting its difference from the superset of adaptive facades. It also highlighted the aspect of façade control which can be optimized based on certain preset criteria. Chapter 1 discusses and focuses the research problem as a resolution of visual comfort for multiple occupants with varying comfort preferences and importance factors. To do so, certain research objectives were set and some others were developed during the research, based on calculated data. The first chapter sets the platform to discuss the idea of subjectivity in human visual comfort. Chapter 2 discusses the literature review findings pertaining to these research objectives and introduces other useful ideas. 7 CHAPTER 2 2. BACKGROUND The following sections provide a background and a compilation of literature reviews of daylighting, visual comfort, and metrics. Sections 2.1 to 2.7 explain the concept of effective daylighting, its planning, availability, and design techniques and control strategies for improving daylighting performance. Section 2.8 explains the concept of visual comfort and performance, providing a review of established visual comfort indices for illuminance, glare, and contrast ratio, as well as factors for accessing them. Section 2.9 highlights the daylight calculation parameters. Section 2.10 explains important daylight metrics and calculation methods and Section 2.11 discusses different daylighting system control methods. Section 2.12 describes the existing daylighting system, explaining the factors adopted for designing the proposed daylighting system and its components. Section 2.13 discusses various studies on user-preference-based ranking and Section 2.14 summarizes the main areas of findings that are further explored in the subsequent chapters. 2.1. Daylighting The primary advantages of using daylight in buildings are that it comes free, is renewable, and can therefore partially or completely offset electrical lighting and associated energy use during daylit hours. A study has found that daylight provides numerous desirable physiological as well as psychological benefits that makes it the perfect source of lighting, especially for the office environment (Tenner 2003). It has been observed that daylight enhances occupant visual comfort and performance in two important ways, by improving both task and spatial visibility, and by providing environmental stimulation (Boyce 1998). In addition to admitting daylight into the building, apertures in the façade usually also provide views and a connection to the outdoors, commonly categorized as visual amenity features of a daylighting system. Daylight also provides a large workable range of illuminances and excellent color discrimination and color rendering properties (IEA 2000). Daylighting, therefore, is the design and practice of controlled natural light entry into a building that aims at meeting these performative as well as physiological considerations. The benefits of daylighting are therefore evaluated two-fold: the resulting psychological and physiological benefits and the energy saving potential. Good daylighting design practices and control strategies translate the dynamic and wide range of outdoor illuminances provided by daylight into acceptable indoor lighting conditions and can thus determine and regulate the nature and extent of visual and thermal stimulation perceived by occupants through daylight and associated factors such as visual amenity. Research shows that regular visual stimulation in the workplace through good daylighting practices can reduce stress and discomfort that is increased through prolonged work in an artificially illuminated work environment (IEA 2000). The overall benefit of the organization resulting from increased occupant productivity is certainly an important incentive to improve daylighting in the occupied spaces. However, it has been found that poor daylighting is also associated with visual discomfort resulting from undesirable effects such as glare, veiling reflections, or visual stress caused by high luminance contrasts in the field of view (Konis 2009). Planning and designing for daylight delivery therefore forms an important aspect of the initial design decision- making process. 2.2. Effective Daylighting Besides being an important consideration for energy-saving, it has been observed that good daylighting practice results in improved occupant comfort, well-being, and workplace satisfaction (Konis and Selkowitz 2017). Daylighting measures can affect both indoor thermal as well as visual comfort conditions. High-performance façade design processes are known to be driven primarily by its energy-saving and performance goals, though there is absent a clear framework for designers to integrate these benefits with long-term human comfort and well-being objectives such as visual comfort, daylight sufficiency, quality of outdoor view (Konis and Selkowitz S 2017), and other non-visual factors such as circadian entrainment and alertness (Konis 2017). A successful daylight delivery system is not just an assemblage of components and design features. In order to have the daylighting system respond optimally to daily and seasonally-changing daylight incidence directions and intensities, there exists a need for a daylight-responsive control system. The challenge of effective daylighting is therefore to establish a design approach that successfully integrates façade system technologies and their controls with other building systems, operations, individual occupant behavior, and preferences, to produce a desirable and high- 8 performing visual environment. Such a design approach would need to address the problem of the stochastic nature of occupant behavior and their individual preferential and physiological needs, as well as meet defined building performance objectives. 2.3. Barriers to Effective Daylighting Though effective daylighting been established as an important step in the design of healthy and high-performing buildings, it has not always been easy to integrate it successfully into the design process. Research shows that the primary reasons for this gap are often the unavailability of low-cost high performing systems, lack of appropriate software tools in the design workflow to predict system performance, and a design process that does not successfully integrate daylight planning (Konis and Selkowitz 2017). Other barriers to effective daylighting are the lack of quantified evidence of its advantages due to aforementioned limitations in performance prediction, and limited knowledge among designers about the range of available daylighting systems and control strategies (IEA 2000). Even with the availability of technology, a slow market adoption is the result of higher associated risks, such as the adoption of new, high-performing daylighting systems. A research highlights that to overcome some of these barriers steps were taken, such as the testing of full-scale advanced daylighting system models in actual buildings as well as in scaled building models, both in natural daylight conditions and under artificial skies (Eleanor S. Lee et al. 2005). The research continues to document an example of such a daylighting study conducted in a full-scale office prototype off-site test space undertaken by the Lawrence Berkeley National Laboratory (LBNL) for designing the façade for the New York Times Headquarters Building. The results from these tests indicate that the application of such advanced daylighting systems and design strategies, such as kinetic shading and light shelves, and electrochromic glazing, can indeed improve daylighting performance in perimeter building zones when compared to conventional fenestration daylighting performance (IEA 2000). 2.4. Planning for Effective Daylighting Documentation of research has found that daylighting design forms an essential part of architectural design (IEA 2000). The study suggests that for new constructions, the planning for effective daylighting starts from the conceptual design phase of architectural design, continuing through the design development and construction phases, and right through to the commissioning and post-occupancy phase. Each phase contributes in specific ways to the overall daylighting design. The research also concluded that the process of improving daylighting in an existing structure can either incorporate a large change, such as a systems and envelope retrofit, or it could be as unobtrusive as a lighting control calibration or an interior finish upgrade. (IEA 2000) It has been found that the challenge of effective daylighting is best tackled at the initial design phases to leverage the usefulness of daylight potential at the site. The results of the study suggest that form, massing, and whole-building orientation are key variables in new design. The study also concludes that for both existing as well as new construction, an initial step in planning for daylighting is to determine the overall lighting requirements in different spaces and on task surfaces, and the degree of lighting control that is necessary (IEA 2000). 2.5. Daylight Availability The sun is the direct source of daylight. It is well documented that a substantial portion of it reaching a space is comprised of indirect and diffused light reflecting from the sky, cloud cover, ground, and from surrounding building and landscape surfaces. The study also suggests that the latitude of the site, time of day, sky condition and sunshine probability, prevailing climate conditions, building massing and orientation, envelope optical properties, and the surrounding context are primary determinants of the daylight intensity and availability in the building (IEA 2000). Surrounding context plays an important role in determining the daylight availability in a space and is usually considered as visual or daylight obstructions, as glare enhancing objects, or as visual interests for the occupant. The extent of surrounding context could encompass both surrounding buildings, vegetation, and other structures and objects which obstruct daylight entry or the quality of view. It has been observed that one square meter of daylight has sufficient lumens to effectively daylight 200 square meters of office space when uniformly distributed (Konis and Selkowitz 2017). If daylight availability in a space reaches a maximum during the late afternoon, which is when the electric load resulting from cooling and artificial lighting is incidentally also at a cumulative peak, then it can be effectively harnessed, controlled, and distributed, and this could help cut down electricity costs, especially during peak load hours. 9 2.6 Daylighting Design Daylighting design is associated with the technique of delivery of daylight in a space and how well it fulfills task and building performance needs. Given an instance of outdoor daylight intensity and availability, its distribution in a partially or completely enclosed space would depend on the optical properties of the interior surfaces and finishes, internal spatial geometry, dimensions and arrangements, visual obstructions, and on the design of the envelope. However, the overall perception and effectiveness of daylighting in a space can be evaluated based on how closely design, and occupant visual comfort and physiological criteria, are met, as well as on the extent of energy savings that it allows. Direct daylight incidence on interior task surfaces is generally avoided owing to the discomfort that it might cause to the occupants through phenomena such as glare. It has been found that daylighting entry into buildings is usually preferred to be either diffused or redirected off non-task and room surfaces that are not directly in the field of view of occupants, such as the portions of walls that are parallel to the occupant viewing direction and is above the general viewing height (IEA 2000). Daylight can enter a space by side-lighting (such as through windows, glazed curtain-walls, light shelves), or by top lighting (such as clearstories and skylights). In side-lighting, the ratio of the opening height to the depth of the space to be effectively daylit is very important. Daylighting standards specify that the primary daylit zone extends up to a depth of 15' perpendicular from the window and the width extent being the width of glazing plus 2' on either side. The standards also specify that the secondary daylit zone is either 25' or 2.5 times the height of the top of glazing, whichever is lesser (DiLaura et al. 2011). This can be used for furniture layouts so that the workstations in an office are placed in the first and second zones while the circulation area that requires less light can be placed in the farthest zone (Banerjee 2015). 2.7 Daylighting Design Strategies Besides the primary function of providing adequate daylight in a space, a window can also be used a medium for providing occupants a medium for viewing the outdoors. While selecting and designing the daylighting system, an important consideration would be, for example, to ensure that the design decisions of window positioning, size, and divisions be tuned to provide good quality, unobstructed views of the outside from occupant seating positions. The daylighting design for larger spaces having more complex lighting requirements usually call for application of composite daylighting systems serving perimeter and interior zones simultaneously. Such composite systems can be comprised of, for example, light shelves, exterior shading, and electrochromic glazing. Studies have found that a strategy often implemented in cooling dominated climates is designing for reducing solar heat gain while still meeting daylighting requirements. It was observed that strategies such as double-pane glazing, double-skin facades, and collector windows are used for such purposes, and window operability, privacy requirements, and glare concerns can affect the choice of daylighting system. The studies also observed that daylighting strategies are developed with consideration to different sky conditions that might prevail at the same location and that it needs to take into consideration four important types of daylighting conditions resulting from clear and cloudy skies, direct sunlight, and diffuse skylight (IEA 2000). Direct sunlight has a much higher intensity than diffused sunlight and naturally needs a smaller aperture to provide an equal overall incident number of lumens (Konis and Selkowitz 2017). The efficiency of the daylighting system lies in how evenly it distributes this incident light over a given area without causing glare and by meeting minimum illuminance standards (IEA 2000). Different windows or apertures on the same façade can have specific daylighting functions. Often, a window can be designed heterogeneously, such that the same window is functionally divided into two or more segments, each designed and employing a different daylighting system and strategy to solve problems such as glare or daylight penetration. For example, light shelves (see Section 2.12.1) allow for greater daylight penetration and reduce glare. However, it can reduce illuminances in spaces right beneath it, thus needing the use of an additional vision glazing component that not just provides additional illumination at peripheral spaces, but also quality outdoor views that light shelves are neither designed nor positioned to provide. 2.7.1 Direct Solar Control When the entry of daylight in a space is allowed, it can sometimes be accompanied by direct solar incidence on task surfaces, visual field, or on the display screen. The direct or indirect glare resulting from objects displaying high and 10 contrasting luminance values in the visual field of an occupant can be mitigated using direct solar control strategies such as shading devices. Direct solar control is usually one of the primary steps in the daylighting design process. Direct Sun Test Studies suggest that to know if a task surface or other occupied space needs direct solar control, a direct sun test can be performed on the surface(s) receiving direct solar incidence (Energy Design Resources 2018). The test might be conducted at the boundary solar positions, for example on the summer solstice or on either of the equinoxes, for three times daily (usually morning, noon, and afternoon). However, the selected boundary solar positions may differ depending on the design intent or on the building operation schedule. These together constitute the altitudinal and azimuthal bounds of the sun’s motion, and this gives a certain, though not complete, idea of the expected range of solar conditions. However, this is not always the case as an intermediary altitudinal or azimuthal solar position might cause an unpredictable daylighting condition, such as the reflection of light off specular surfaces and which might ultimately add to glare. The intent of the direct sun test is to find if the illuminance measured on a surface exceeds a certain threshold value and from which one can probably infer if sunlight is directly incident on the task surface and to what extents (Energy Design Resources 2018). Solar Shades Direct solar control can be achieved using solar shades. In addition to electrochromic glazing, solar shades can be used to cut down on direct solar component which might not be preferred at occupant workstations. By using exterior solar shades, electrochromic glazing can be kept at a high visible light transmittance allowing more daylight entry and yet cut-off the direct solar component. There are several types of solar shades. The classification is based on the relative position of the shade, the material, and the type (Fig. 6). Figure 6: Solar shading types and classification with some picture examples (Grynning et al. 2017) 2.7.2 Uniformity Many types of activities and space types demand a uniformly-lit environment. Uniformity refers to the overall consistency and evenness of lighting level values measured over a given surface area (Carlucci et al. 2015). The surface area for measurement can be an entire continuous floor area or it could even be a task surface. Uniformity indices are a better representation of the distribution of illuminance values over a small surface than an average illuminance value. For example, if three evenly spaced sensor points along the centerline of a 5’x2’6” task surface measure illuminance as 500lx, 700lx, and 3000lx, then the average illuminance is 1400lx. The averaged value does not provide an indication of the high illuminance measured by one sensor point, which can result from, for example, a direct or indirect solar incidence causing a high illuminance patch on the task surface. If 11 one had a similar task surface but having measured illuminance values of 1500lx, 1400lx, and 1300lx, the average illuminance would still be 1400lx. In this case as well, the averaged illuminance value does not signify the difference in the range of distribution of the illuminances on the two task surfaces. If a researcher was to use a uniformity index, for example, the average illuminance to maximum illuminance ratio, then it would provide us with calculated uniformity values of 0.47 and 0.93 for the two mentioned cases. The value of 0.47 clearly indicates that the maximum lies further from the average in the first case than in the second case, thus providing us with a metric to measure how uniformly the illuminances are spread over a surface. The closer the uniformity value is to unity, the more uniform the illuminance distribution. If illuminances and uniformity indices are measured for a small surface, such as a workstation, and for the floor surface, in a diffused-daylit condition, then the smaller surface would show a range of closely space illuminances resulting in a uniformity value close to unity. On the other hand, sensor points at either ends of a floor surface might measure very contrasting illuminance values. In this case, the measured uniformity ratio would be far from unity, thus wrongly implying non-uniformity when the distribution of illuminances on the floor surface is indeed more uniform owing to the diffused daylighting. A uniformly-lit space can help avoid visual stress, which can be a result of continuous eye adaptations to two or more adjacent over-lit and under-lit environments (Carlucci et al. 2015), or caused due to presence of glare patches, high luminance contrast ratios, and shadowed regions in the visual field (Energy Design Resources 2018). 2.7.3 Illuminance Targets Required illuminance targets for different space types and activities are specified by the Illuminating Engineering Society of North America (IESNA). The next step in daylighting design is to find these illuminance targets and to apply design measures to compensate for a lack or excess of light levels in the process of meeting these targets, then categorize targets based on functionality. An open-plan office, might have two main lighting requirements, e.g. for circulation and for visually demanding tasks. Such a categorization requires use of ambient and task lighting illuminance targets which are met separately to provide a net lighting level at any given point in the space. 2.7.4 Energy Targets Energy efficiency standards, such as Title 24 and ASHRAE 90.1, specify different maximum Lighting Power Densities (LPD) for space types such as schools and offices. For example, Title 24 2008 specifies an LPD of 1.3 watt per square foot for libraries whereas ASHRAE 90.1-2013 recommends an LPD of 1.19. Lower LPD provisions mean that more daylight and natural illumination is needed to reach target lighting levels, thus promoting the design and use of effective daylighting strategies and systems. 2.7.5 Usability & Architectural Integration Usability refers to design functionality and considerations directed towards the end-user. For example, the interior spatial layout might be oriented in a way so that the occupant using a visual display unit is oriented parallel to the window as this reduces glare in the field of view of the occupant. Architectural integration refers to how the daylighting design complements and synergizes with the other building systems, such as the envelope, HVAC systems. 2.8 Visual Comfort & Performance It has been found that visual comfort is an effect of the visual condition of a space which enhances the well-being and performance of the occupants. The research also suggests that factors such as the physiology of the eye, the light level, its distribution in the space, probability of glare, the color rendering ability of the light, and the spectral emission of the source, cumulatively define the visual comfort as perceived by an occupant (Carlucci et al. 2015). However, the perception of light and its effects on occupants might differ, which is why visual comfort is a subjective condition. It means that the visual condition in a space might be completely acceptable to one occupant but may still fall short of the ideal visual comfort conditions for another occupant. Like thermal comfort, visual comfort might also be accessed and represented on a predefined scale. Designing the lighting environment in a space can certainly be assisted by use of such a scale. However, its use does not naturally imply that all or even a single occupant might be satisfied with the visual conditions. One of the ways of getting around this problem would be to establish personalized visual comfort ranges based on individual visual preferences. Several indices (see Section 2.8.2 to Section 2.8.5) exist to measure the different facets of visual comfort in a space, such as glare or illuminance, but none exist that defines visual comfort on an overall basis (Carlucci et al. 2015). These indices, however, may be used and modified on an individual level to represent the varying sets of individual preferences. 12 Visual Comfort Indices Visual comfort in a lit environment, whether naturally, artificially, or both, can be assessed on certain defined factors- the amount of light, its uniformity or distribution, glare, and the quality of light in rendering colors (Carlucci et al. 2015). These are further dependent and accessed using measurements, such as task surface and ambient illuminance, vertical plane illuminance at the eye level, task background reflectance, scene luminance contrast ratios, and discomfort and disability glare. A study has found that 50% of the existing indices are devoted for glare assessment or prediction while 26% measure the amount of light, 21 %, the light quality, and 3% the light uniformity (Fig. 7). The research also observed that no available metric presently represents the overall visual comfort in one value (Carlucci et al. 2015). Figure 7: Cumulative number of visual comfort indices over time (Carlucci et al. 2015) This study also found that not all the indices are applicable in all test conditions. It observes that each index has been designed to tackle specific aspects and differs from the other indices based on several factors, such as the scope of assessment, the physical quantities involved, the calculation period, the light source, the acceptability criterion, and a defined threshold (Carlucci et al. 2015)”. 2.8.1 Factors for accessing visual comfort indices (Carlucci et al. 2015) The following factors for accessing visual comfort indices are described: scope of index, light source, space discretization, comfort thresholds, acceptability criterion, and comfort thresholds. Scope of Index Different indices are tailored to assess different aspects of visual comfort, such as for glare, uniformity, light quality, and lighting levels. Light Source The applicability of an index also depends on the light source. Some indices are more generally applied, such as Daylight Glare Probability (DGP), whereas some, such as Unified Glare Rating (UGR), is applicable only for measuring glare in artificially-lit environments. This is because the mathematical calculation methods for each index is based on certain assumptions that does not allow its application in other conditions. For example, DGP is given by the formula as below (Equation 1) and is only defined for vertical illuminance E v greater than 380 lx. Equation 1: Daylight Glare Probability Metric Calculation Space Discretization Some indices are applicable locally and some globally. For example, the illuminance at a point in space is a local measurement. An array of such measurements would be necessary to map the illuminances of a surface. Such indices 13 are locally applicable. On the other hand, global indices are applicable and measured for an entire space, for example daylight autonomy. These indices do not represent individual measurements but use it for a global measurement. Time Discretization Based on the time-period of measurements, indexes can be categorized as “static” and “dynamic”, as used by Reinhart (Reinhart, Mardaljevic, and Rogers 2006). Static indices are measurements taken only for a single point in time whereas dynamic indices are usually calculated over the course of a year or a longer term (Carlucci et al. 2015). The latter may cumulate a series of measurements into one aggregate value representing overall performance or may represent a series of discrete measurements. Such an index could be used to evaluate the annual performance of building under variable sky and weather condition. Dynamic indices may be further divided into time series and cumulative indices. The former produces a series of measurements taken throughout the year representing different points in time whereas the latter forms a cumulative or aggregate representation of all the measurements resulting in one single value. An example of a cumulative daylighting index is the Spatial Daylight Autonomy (sDA) which calculates the percentage of an analysis area that meets a certain minimum threshold illuminance for a given percentage of the operating hours of the year. (Carlucci et al. 2015) Comfort Thresholds Luminous thresholds help us determine if the index measurements are within acceptable visual standards. Indices such as illuminance do not have a fixed threshold as the acceptable boundary conditions are dependent on factors such as the assessed phenomenon, the nature of task, and the lighting technique. Acceptability Criterion Indices access an aspect of the lighting environment by comparing a measured value to a single or multiple reference bounding values. For example, the glare index, Daylight Glare Probability, has a minimum lower bound of 380 lx as the vertical illuminance measurement and produces a minimum of approximately 0.2, but does not have any upper bound (Carlucci et al. 2015). Illuminance-based metrics on the other hand usually have a lower as well as an upper bound. Based on whether a visual comfort index has one or two such bounds defining the acceptable quantity for visual comfort, indices may be categorized as one-tailed and two-tailed respectively. (Carlucci et al. 2015) Comfort Thresholds Research also suggest that indices can also be categorized depending on their acceptable threshold values. The study defines a threshold as a fixed or dynamic set of values defining the acceptable range of an index, for example, in terms of visual comfort (Carlucci et al. 2015). 2.8.2 Illuminance The illuminance at a given point on a surface refers to the ratio of the total luminous flux incident on an infinitesimally small surface unit area at the point, to the area (Carlucci et al. 2015). It determines the total light level at a given point incident from all sources and is independent of the direction of view, unlike luminance. However, it is dependent on the angle of incidence and the direction of incidence. The more the angle of incidence of light falling on a surface, the lesser the area component perpendicular to the direction of light and thus a lesser illuminance at the surface (DiLaura et al. 2011). It is also important to note that illuminance is not just a result of direct light but also indirect light resulting from one or more bounces off luminous surfaces in the scene. Such bounced indirect light is called the inter-reflected component. In outdoor scenes, the inter-reflected component is presumed to be negligible compared to the direct component. In indoor scenes, the inter-reflected component constitutes a higher percentage (DiLaura et al. 2011). Illuminance is measured using a chroma meter, illuminance meter, or an illuminance spectrophotometer, in units of lux (lx) or foot-candles (fc) such that 10 fc=1 lux. Illuminance criteria calculated for a given space directly determines whether it is sufficient for a given task or activity (DiLaura et al. 2011). For example, the standard uniform illuminance at the work plane recommended for a general office space, resulting from ambient and task lighting sources, is 500 lux. More detail-oriented tasks such as sculpture, drafting, or weaving, or those requiring greater speed and accuracy, need a much higher illuminance level, sometimes greater than 1000 lux. (DiLaura et al. 2011) The standard illuminance level requirement is dependent on the nature of the activity (DiLaura et al. 2011). The viewer’s age also plays a very important factor as well and a higher illuminance level is recommended to counter the disability resulting from a higher age, especially for tasks demanding close visual inspection. Even in an office, work 14 plane illuminance is usually based on lighting level requirements for paper-based tasks, but which only form a certain share of the total activity types, mostly visual and display-based. (DiLaura et al. 2011) Using an accurate physics-based lighting simulation engine such as Radiance and a simulation program such as Daysim, illumination can be calculated virtually at any three-dimensionally defined sensor location, given a direction vector normal to the plane and point at which the measurement is computed. An array of these virtual photo-sensors, usually selected on one continuous plane, could “measure” the distribution of illuminances. The selection of the plane of measurement depends on the illumination criteria and the nature of activity in the given space. For example, in an office space the horizontal work plane illuminance might be a good representation of the amount of light incident on the work surface, but is does not give an idea of the illuminance on vertical surfaces such as walls or at vertical intersections. Illuminance by itself is not the absolute measure of how well a certain object is visible in the field of view and must be accompanied by other factors such as acceptable contrast of the object with its background and its specularity. Performance Metrics for Evaluating Illuminance Potential of Daylighting Systems • The illumination level available at a given point in space differs throughout the year. Measuring the effectiveness of a daylighting system therefore needs to take into consideration this change over time. Effectiveness of daylighting systems in achieving design illuminance can be measured based on how well the systems fulfills the selected daylighting design criteria. • If it is a conventional window system, the ratio of number of hours for which design illuminance is fulfilled to the total annual daylit hours will be proportional to the daylighting effectiveness of the system. • Conventional window systems can effectively daylight a room up to a depth of 1.5 to 2 times the height of the top of the opening from the finished floor surface (IEA 2000). Light shelves redirect light to greater depths than conventional window systems. For these and other light redirecting systems, a metric for accessing its performance could be based on the number of times the system is able to make daylight reach a certain depth annually. Ambient Lighting Ambient lighting is the general baseline lighting level required in a space and usually refers to the minimum lighting level required for way finding or to tell apart larger objects that have a high luminance contrast ratio (DiLaura et al. 2011). Task Lighting Task lighting is the additional task specific and localized lighting that supplements the ambient lighting to produce a visually acceptable and satisfactory illuminance range that is required for a specific task, such as general paperwork or computer-based work in an office environment (DiLaura et al. 2011). The aim of the task lighting is therefore to compensate for the low ambient levels at task surfaces, without having to increase the ambient lighting levels. This is done to save the additional lighting energy use cost that is incurred by wastefully using high ambient lighting levels. The required illuminance levels are based on the nature of task as well as the age of the occupants. These lighting level requirements are recommended by the Illuminating Engineering Society of North America (IESNA). 2.8.3 Uniformity The eye finds it difficult to adapt to contrasting light levels in the field of view. Daylighting uniformity measures how lighting levels vary from one point to another across a plane or surface. Different metrics are used to analyze the illuminance levels across a surface. The most common method is to average all the measured illuminances over a certain area. However, an average value is not indicative of uniformity unless the area of measurement is delimited to a small surface such as, for example, an individual workspace. To get a better idea of the range of illuminances on a given surface, metrics use the maximum, minimum, and average values (Carlucci et al. 2015). The ratio in use could be of two illuminance quantities across a surface- maximum to minimum, maximum to average, and average to minimum. However, this is also not very useful unless it is measured locally. For larger contiguous spaces, this ratio does not describe local fluctuations in illuminance. Another metric, the coefficient of variation (CV), which is the unitless ratio of the standard deviation to the mean, and denoted by the equation below, measures the dispersion of data from the average. “CV is often presented as the given ratio multiplied by 100. […] the higher the CV, the greater the dispersion and the more different (higher or lower) the values are from the mean.” (Carlucci et al. 2015) 15 2.8.4 Luminance & Contrast Luminance refers to the flux of light reflected or emitted off a surface per unit area and in a given direction. Unlike illuminance, luminance is associated with the direction of view and is also dependent on surface optical characteristics such as reflectivity, or material optical characteristics such as transmissivity. A surface with poor reflectivity in a high illuminance scene will appear optically flat, dark, and having indiscernible surface features, in contrast to a surface with higher reflectivity. Such low reflecting objects may be discernible by their outlines contrasting against a surface with a different set of optical characteristics, surface shadows, or by virtue of occlusion. In the context of an indoor scene such as an office space, the perception and visual quality of the inner surfaces, such as walls, desks, floors and ceilings, are a direct result of luminance perception by the eye. A shiny floor on which sunlight is directly incident will cause a high luminance patch in the field of view of an occupant in the same plane as the incident rays, whereas it might not be as bright when seen from a different angle. Luminance is dependent on the illuminance of incident light. If the illuminance at a given point is low, the luminance measured at that point will also be low when compared to a high illuminance condition. 2.8.5 Glare Glare occurs when there is a high luminance contrast among two or more adjacent patches in the field of view (Wienold 2007; Koo, Yeo, and Kim 2010). It may be categorized as discomfort glare, disability glare, or as veiling reflections based on the source and the nature of its effect on people. Discomfort glare occurs when there is a relatively high contrast ratio in between patches in the field of view but which is not high enough to impair vision. In such cases, the objects displaying glare are visible but it is a source of continuous discomfort to the human eye. Disability glare occurs when the contrast ratio is high enough to make the object in the foreground hardly discernible. (Wienold 2007; Koo, Yeo, and Kim 2010) Glare might also be categorized as direct or indirect based on the cause and source of the sensation (Suk 2014). For example, sunlight falling on the eye would be a direct glare source whereas when reflected from a highly specular surface might be an indirect glare source to the eye when viewed. Glare can also be categorized as absolute or relative. Absolute glare occurs due to the presence of an excessive bright source, the sun or a luminaire of very high luminance that will cause discomfort regardless of the environment since the absolute luminance is too high for the photoreceptors to process. Relative glare occurs due to the contrast between the glare source and background. (Suk 2014) In evaluating glare sources, several factors such as geometry, size, location and number of surfaces in the field of view of the occupant, and age are taken into consideration. Common sources of discomfort glare are windows, surfaces of fenestration shading elements, and luminaires in the field of view. A key point to keep in mind while designing with daylighting is to avoid glare. 2.8.6 Visual Amenity Depending on the quality of the outside view, there can be compromises on the tolerance levels of glare (IEA 2000). The nature of view differs with the type of daylighting system chosen, the relationship of the exterior context in relation to the interior, and the location, orientation of field of view, and height of the occupant in relation to the interior layout. Four basic types of outside view cases might exist and which are ranked from best to worst: complete unmitigated, undistorted view, partial view, occasional view, and no view. (IEA 2000) 2.9 Daylight Calculation Parameters The following sub-sections highlight important daylight calculation parameters such as the test location, climate, chosen time step, shading element automation, design illuminance, and spatial considerations. Location and Climate The geographical location, the building orientation, siting, seasonal and climatic conditions, sky cover- can affect daylight calculation results. Time Frame This is the time interval between each set of measurements. The time frame is usually chosen so that there is no substantial solar change in that period. It is also selected based on the accuracy and fineness of measurement needed. For more fine-grained results, the time frame needs to be smaller, such as 15 minutes. This increases the annual simulation time, for example, but also generates more accurate results. The occupancy schedule can also be used to mask the annual daylight hours to give a sense of the daily time frame or the occupied hours for which the simulation 16 needs to be run. Other parameters (e.g. shading element automation, design illuminance, and spatial considerations) that affect daylight calculations are not discussed further. 2.10 Daylight Metrics & Calculation Methods The following sections describe popular daylight metrics and calculation methods for illuminance, luminance, uniformity, and glare. 2.10.1 Illuminance Metrics Section 2.10.1 explains popular illuminance metrics, classified as either static or dynamic daylight metrics. Static Daylight Metrics The following explains popular static daylight metrics for illuminance, such as the maximum to minimum illuminance ratio, and the single point in time (SPT) measurement. Another method that has been in use but not so much lately is the Daylight Factor (DF) method (Carlucci et al. 2015). Maximum to Minimum Illuminance Ratio This calculation method is used for finding the extremes of illuminances measured on a given area. It could be used to detect potential glare and unbalanced illuminances by matching it against a defined threshold value. If the ratio is very high, for example on a work plane, it may be concluded that at least one point on the plane has an illuminance that is much higher than the other illuminance values. The limitation of this method is that it does not provide any information on the illuminance distribution gradient, location, or the distribution frequency which a standard deviation might better represent. Single Point in Time (SPT) This method uses a work plane illuminance measurement for only one time annually to represent an “average” daylight condition at that time. The time selected for the measurement should be such that the measurement could guide building orientation or design. For example, if the measurement is taken at a fixed time, say, at 1500 hrs. on the summer equinox, June 21, it could be used to compare between two or more design orientations or shading mechanisms to get an idea of its effectiveness when the sun altitude is highest. Dynamic Daylight Metrics The following explains popular dynamic daylight metrics, such as daylight autonomy (DA), useful daylight illuminance (UDI), frequency of visual comfort (FVC) and intensity of visual discomfort (IVD). Daylight Autonomy (DA) This method is used with an annual simulation to know the percentage of the year a minimum work plane illuminance is met at a given point only through daylight (Carlucci et al. 2015). There are variations to the original DA method such as the incremental summing, continuous summing, and the maximum daylight autonomy method. Useful Daylight Illuminances (UDI) Not all indoor illuminances are useful. Some of these are under and some over the recommended limits. Useful Daylight Illuminance (UDI) is defined as the number of occupied hours that the horizontal illuminance at a select point on the work plane is in the useful range of 100 lux – 2000 lux (Carlucci et al. 2015). UDI is a long-term and local index (Carlucci et al. 2015). Therefore, UDI does not only provide the useful range of illuminances but also the frequency with which they occur annually. A limitation and often debated assumption in the UDI calculation is the definition of the acceptable illumination range (Equation 2). (Carlucci et al. 2015) Equation 2: Useful Daylight Illuminance (Carlucci et al. 2015) 17 Frequency of Visual Comfort (FVC) A space might receive appropriate levels of daylight only for a fraction of the total time of measurement. The FVC measures, for a long-term (usually monthly or annually) period of measurement, the percentage of time when the average of the measured illuminances in a space, resulting from daylight alone, is within a certain acceptable range bounded by a lower and an upper bound (Carlucci et al. 2015). An assumption of a satisfactory FVC would be 0.8 according to a study (Carlucci et al. 2015), which means that the illuminance at the point of measurement is within the acceptable range, say E under =150 lx, E over =750 lx, for 80% of the total time period of measurement. The Frequency of Visual Comfort is calculated using the following equation (Equation 3). (Carlucci et al. 2015) Equation 3: Frequency of Visual Comfort (FVC) (Carlucci et al. 2015) Intensity of Visual Discomfort (IVD) For an upper and lower illumination setpoint defined for visual comfort, say E over (say 750 lx) and E under (say 150 lx), this metric calculates (Equation 4) an aggregate of the difference between the spatial average illuminance and the illumination boundaries (Carlucci et al. 2015). Some common daylighting metric definitions are shown (Fig. 8). Equation 4: Intensity of Visual Discomfort (IVD) (Carlucci et al. 2015) Figure 8: Daylighting (Illuminance-based) Metric Definitions (Architectural Energy Corporation) 18 2.10.2 Lighting Uniformity Metrics Illuminance Uniformity (U o ), the lighting uniformity metric, can be calculated as either the ratio of the minimum to the average illuminance or the minimum to the maximum illuminance on a plane (Equation 5). There are certain acceptable ranges and threshold values of uniformity ratios (Carlucci et al. 2015). Some lighting standards specify a uniformity ratio of 0.8 (min/average) and 0.7 (min/max). Some others also require certain illuminance values immediately outside and in the background of the task area. For the former, a minimum value of 0.4-0.7 is to be maintained, whereas for the latter a minimum value of 0.1 is recommended. In most cases, these ratios have been defined for artificial lights or skylights. (Carlucci et al. 2015) Equation 5: Illuminance Uniformity (U o ) (Carlucci et al. 2015) 2.10.3 Glare Metrics Glare risk is usually a result of luminance contrast of objects in the field of view (Hopkinson 1972; Pierson, Wienold, and Bodart 2017). However, it can also be caused by objects of high luminance in the visual field. The visual field itself can be divided into segments depending on how far it is from the central line of vision. The portion of the visual field in front of the occupant is the visual task or the ‘central zone’. The zone immediately around the ‘central zone’ is called the ‘adjacent zone’ and is bounded by a 60 o cone, which in turn is surrounded by a non-adjacent zone, delimited by a 120 o cone. (Hopkinson 1972; Pierson, Wienold, and Bodart 2017) 2.10.4 Luminance Contrast Ratio Some studies have documented luminance contrast ratios of 1:3:10 as ideal specifications though others have actually documented the human eye being able to adapt and be comfortable with a luminance contrast ratio of even 1:40 or 1:100 (Carlucci et al. 2015). However, the following threshold contrast ratios are usually considered: • Visual Task vs Immediate Surroundings: 1:3 • Visual Task vs Near Surfaces: 1:10 • Visual Task vs Distant Surfaces: 1:20 • Visual Task vs Other Surfaces: 1:40 There are several glare metrics that are used in the industry. Some of these are Daylight Glare Index (DGI), CIE Glare Index, Visual Comfor Probability (VCP), CIE Unified Glare Rating (UGR), and Daylight Glare Probability (DGP) (Suk, Schiler, and Kensek 2017). Of these, the Daylight Glare Probability (DGP) metric has been seen to form a greater correlation with the occupant glare perception. The following table lists the definitions of the degree of glare sensation for each of the five glare metrics (Fig. 9). Figure 9: Glare discomfort index comparison on a nine-point glare sensation scale (Carlucci et al. 2015) 19 2.10.5 Occupant Preference as Metric The above-mentioned daylighting metrics and criteria represent either point-in-time measurements or an accumulation of annual measurements. However, these do not evaluate the performance of an adaptive daylighting system in real- time. For a given system which operates based on user-preferences, a sufficient metric would be to check if the illuminance provided closely matches or falls within the range of the user-specified illuminance. The other daylighting metrics do not evaluate the system consider this aspect but based on certain fixed illuminance ranges against which measured annual illuminance values are matched. To measure the performance of such a preference-driven dynamically changing or adaptive daylighting system, measurements at each time-step needs to be evaluated against current preferences and relayed back to the system through a feedback loop. 2.11 Daylighting System Control Methods The following sub-sections discuss different proposed and tested daylighting system control methods. These control methods or control logic can be understood as decision-frameworks which operate in real-time based on input, such as from outdoor solar radiation sensors. 2.11.1 Model-Based Control Strategies Environmentally responsive facades can simultaneously reduce building energy consumption, maintain high levels of indoor environmental quality, and provide visual amenity, such as thermal and visual comfort, privacy, and views to the outside. User satisfaction is key to operation of automated facades. The key problem lies in the unpredictable array of user’s personal preferences, their goals, personal motivations, behavior, and requirements that change over time. It would seem obvious to conclude that manual controlled daylighting systems would outperform automated systems in terms of occupant overriding capability as they seemingly give complete control to all users. However, it has been seen that neither do manually controlled facades perform well in terms of energy nor do they solve the problem of conflicting user preferences. Fully-automated daylighting systems are algorithmically-controlled that may or may not consider users’ personal preferences and comfort conditions. “POE research conducted by LBNL, for example, shows that provided the systems are carefully designed, commissioned, and maintained, it is possible to develop energy-saving automated control strategies at high occupant satisfaction.(Lee, Fernandes, and Coffey 2013)”. There has been a study that contradict this as well (Bakker et al. 2014). For example, it has been identified in post- occupancy evaluations, instances of occupant discomfort and frustration owing to seemingly undesired daylighting conditions resulting from automatic blind operation. It has been found that systems with more user-override controls, especially quick response control systems, are preferred by occupants in contrast to systems with limited options for user intervention. User satisfaction is partially dependent on the degree and extent of user control and interaction that an adaptive façade system allows. Therefore, what gains importance is the design of control strategies that are used in the operation of the dynamic façade components. (Bakker et al. 2014) There are several ways of designing the control mechanism. For example, the system could operate autonomously without any occupant feedback or overriding controls. Such a control mechanism is based on assumed conditions and works based on a preset algorithm that does not adapt to changing environmental or occupancy conditions and preferences. These systems may be classified as adaptive, but they do not actively respond to any live feedback or sensor data collected at the interface. Rather, their response is designed based on historic environmental data. Environmentally responsive systems are controlled by sensing real-time data, by sensors located, for example, both on the interior and exterior, as well as through occupant feedback and control. Intelligent control algorithms are needed for the operation of such systems. “Model-based control (MBC) strategies that use (detailed or low-order) models with real-time input data and robust control schemes have a great potential to reduce energy use and improve indoor environmental comfort (Xiong and Tzempelikos 2016)”. Research in this area has mostly focused on thermal and HVAC system operation (Li et al. 2015; Moroşan et al. 2010), and on indoor air quality applications (Lu, Lü, and Viljanen 2011), based on embedded optimization techniques (Li et al. 2015; Hu and Karava 2014). However, there have been few studies on shading controls based on visual comfort criteria. In one such research, an integrated shading and lighting model-based control algorithm has been developed and implemented for perimeter zones and is based on visual comfort criteria (Xiong and Tzempelikos 2016). The algorithm takes beam and diffuse 20 (sky and ground reflected) transmitted illuminance and solar radiation as inputs, which is used to generate illuminance and luminance mapping for the given space geometry using a hybrid raytracing and radiosity method (Fig. 10). In this method (Xiong and Tzempelikos 2016), among the predetermined parameters are the space geometry, occupancy information (such as seating location, layout, and viewing direction), and the optical properties of the fenestration system obtained from WINDOW (“LBNL Windows & Daylighting Software -- WINDOW” 2017). Figure 10: Model-based control strategies (Xiong and Tzempelikos 2016) The output includes the work plane illuminance, vertical (on eye) illuminance, and the direct glare probability (DGP) based on a preset calculation grid using the full DGP equation (Wienold 2007) as shown below (Equation 6). Equation 6: The full Daylight Glare Probability equation (Wienold 2007) , where L s is glare source luminance, ω s is source solid angle, P is the position index, and E v is the vertical illuminance on the eye. Every workspace is divided into sub-surfaces to enable more accurate glare calculations. The sub-surfaces (patches) which are at 78 o and higher are assumed to be outside the occupant’s field of view and are therefore neglected. Sub-surfaces within the field of view are considered for further glare analysis. The luminance values of each sub-surface is compared against a reference glare threshold, defined as 4 times of the average work plane illuminance (Wienold 2007), to indicate a glare source. The hybrid daylighting-glare model (Xiong and Tzempelikos 2016) calculates the illuminance, luminance and evaluates glare probability for 11 pre-defined, discrete shading positions at each time step of 1 minute, and changing at 10% increments from a fully open to a fully closed position. The logged simulation data is then sent to a decision- making module which selects the highest (more open) shading position among the 11 predefined positions that also satisfies the selected visual comfort criterion. Three visual comfort-based shading control criteria are considered in this study. A simple DGP-based control selects the highest shading position for which the DGP is less than or equal to 0.35 at each time step. The DGP value for this control criterion is based on the more complex DGP calculation, as shown 21 above (Equation 6). A second control criterion, based on vertical illuminance (E v ) and on a simplified DGPs equation (Equation 7), is used for all cases except when the sun is directly within the field of vision of the occupant i.e. direct light is incident on the eye. Equation 7: Simplified Glare Calculation (Wienold 2007) It is seen that DGPs is less than or equal to 0.35 for E v <=2670 lx. Adding a safety factor, we see that for a vertical illuminance (E v ) less than or equal to 2500 lx results in a DGP < 0.35. This criterion is used to select the highest shading position for each time step. The third control criterion, the work plane illuminance (E wp ) based control or the effective illuminance control, selects the highest shading position for which the work plane illuminance is lower than 2000 lx without any direct sunlight reaching the work plane. Dimmable lighting is used to provide lighting supplement at the grid points on the work planes where the illumination value falls below 500 lx, the supplementary and additional lighting can be calculated (Equation 8). Equation 8: Supplementary Illuminance requirement from Artificial Lighting (Wienold 2007) , where E wpi is the daylight work plane illuminance on a grid ‘i’ and ‘Grid#’ is the total number of grids in the daylight model (Wienold 2007). To overcome the problem of using a very short time-step of one minute, the research suggests a 15-minute time-step. However, instead of using a traditional method of “freezing” the shading controls between time-steps and which disregards potential issues of glare that may result, a more advanced control logic is developed to solve this problem. (Wienold 2007) 2.11.2 Control of Shading and Redirecting Systems Daylighting systems can be divided into two main categories depending on their function (Konstantoglou and Tsangrassoulis 2016). Based on the categorization, daylighting systems can either be shading in nature and which diffuses light, or it could be daylight redirecting systems that direct the light deeper into the building and away from the perimeter zone, such as light shelves (IEA 2000). Other than these broad categories, there are other categorizations of daylighting systems. The choice of daylighting system largely depends on the climatic conditions (IEA 2000). For example, in high latitude countries where the solar incidence is relatively lower, a light shelf daylight redirecting system would need additional shading to prevent issues of visual discomfort, even though the heat gain might be necessary to offset heating loads. In contrast, in low-latitudes, the use of light shelves is more effective as the solar incidence is generally higher and results in a slightly lower light level in the portions of the room away from the façade, but simultaneously providing substantial solar shading. In a place like Los Angeles or down south, where cooling loads are higher but solar gain is also necessary, automated light shelf systems can be effective. Four factors affect daylight harvesting in commercial buildings: the glazing type, the dynamic control nature, character of shading devices, and the degree of automation (Colaco et al. 2008). “The review concludes that the use of artificial intelligence might enhance buildings' energy efficiency and occupants' thermal and visual comfort. It therefore suggests that future research should focus on control systems optimization by improving the placement of photocells and the system connection with real time measurements (Konstantoglou and Tsangrassoulis 2016).” A study of the Oakland Federal Building by the Lawrence Berkeley National Laboratory (LBNL) examined the impact of automated blinds, controlled by a sun blocking strategy for a time step of 30s in maintaining a work plane illuminance in the range of 540-740 lx, on a range of factors. These included the building’s energy performance, work plane illuminance and outdoor view. The results of the study were then compared to the performance of a static system. It concluded that the integration of the automated blind control system with dimming lighting controls resulted in 7%- 15% and 19%-50% energy savings in cooling and lighting respectively (E.S. Lee, DiBartolomeo, and Selkowitz 1998). 22 The BELOK collaboration team, designed a motorized version of conventional blinds daylighting system such that the tilt angles of the upper and lower blinds were different (Bülow-Hübe 2007). This would allow the daylight to enter deeper into the space. The control system closed the blinds when the outdoor façade sensor read greater than 20 klx for a minute and if the value read less than 15 klx for an extended period, the blinds would be raised with a delay of 30 minutes. If the illuminance increases, the blinds would be set back to its correct slat angle (Fig. 11). (Bülow-Hübe 2007) Figure 11: Flowchart of solar protection control strategy (Bülow-Hübe 2007; Konstantoglou and Tsangrassoulis 2016) In simulating the performance of automated split-controlled blinds and defining the slat tilt angle developed an illuminance-based Slat Angle Selection (ISAS) methodology to estimate optimum slat angles for providing desired illuminance, based on Artificial Neural Networks (ANNs) using Useful Daylight Illuminance (UDI) calculated by EnergyPlus (Hu and Olbina 2011). The resultant ISAS model was tested to be 94.7% and 98.5% predictive for estimating illuminance and optimum slat angle respectively. In an integrated approach for evaluating the performance of dynamic systems, it was observed that multiple variables such as work plane illuminance, total energy flux and vertical irradiance on façade, daylight glare probability, contrast ratio on VDU-screen, and quality of outdoor view can lead to opposing shading positions. It was seen that in some cases, the selection of the best strategy is also dependent on the size of the opening (David et al. 2011). Different approaches have been studied, the primary focus areas being control strategies for adaptive and manually controlled blinds, and the selection and evaluation of performance criteria that will provide the required daylighting levels while simultaneously providing solar protection, comfort and reduction in energy consumption (Konstantoglou and Tsangrassoulis 2016). The research also concluded that users are more accepting of dynamic shading and daylighting systems if it is possible for them to override their control system. The extent of their acceptance is also dependent on the system’s simplicity and intuitiveness. “Compared to conventionally controlled blinds where blinds maintain the same tilt angle at a particular point in time, spit- controlled blinds consume up to 20% less energy. With fuzzy control strategies applied on daylighting controls, energy savings may reach up to 60%. When coupled with dimmable electric lighting systems, automated blinds may save up to 80% electric lighting energy (Konstantoglou and Tsangrassoulis 2016).” The research also found that control strategies that try to optimize both thermal as well as visual performance, such as internal temperature and solar irradiation, are usually those that address multiple performance criteria and have been observed to most effectively balance comfort and energy savings. In the initial stage of design, simplified metrics and tools are used for selection of the best daylighting and shading systems. In contrast, selection and definition of dynamic systems and their control logic is done in the commissioning phase of the project. Further observations and conclusions (Konstantoglou and Tsangrassoulis 2016) state that there is a great potential and need for development of control strategies for emergent dynamic façade systems, especially in the form of a tool which could shorten and streamline 23 the performance analysis procedure. In addition, further research on post-occupancy surveys and occupant behavioral patterns is needed to support and improve the control strategies (Konstantoglou and Tsangrassoulis 2016). It has also been found that blinds are the most used dynamic systems, with the most commonly used control strategy being “optimum slat angle” control, while work plane target illuminance based strategies make up about 20% of studies, while 60% of them showed coupling of the system with continuous dimming lighting control. An integrated approach splits the shading device controller into two parts depending on user presence (Guillemin and Morel 2001). In presence of the user, priority is given to visual comfort aspects while more priority is given to thermal comfort aspects when the user does not occupy the space. The user preferences are also used to control the level of illuminance of the artificial lighting in the room (Guillemin and Morel 2001). “Overall, in most cases, integrated lighting and daylight control outperforms all other strategies in energy and visual comfort performance (Shen, Hu, and Patel 2014).” Double-sided blind performance was studied (Oh, Lee, and Yoon 2012) and proposed automatically changing slat side based on the heating or cooling period. In one research (Koo, Yeo, and Kim 2010), the control strategy was used to create variable levels of daylighting in multiple user-defined zones by using multiple blinds and an important regulatory variable being the lower end of each blind (Fig. 12). Figure 12: Different blind positions between the previous and the new control method: (a) previous control method (b) new control method (Koo, Yeo, and Kim 2010). 2.12 Daylighting Systems The following sections review light shelves and its variations as specific examples of redirecting and shading daylighting systems. 2.12.1 Light Shelves “A light shelf is generally a horizontal or nearly horizontal baffle positioned inside and/or outside of the window facade. The light shelf can be an integral part of the facade or mounted on the building (IEA 2000).” Therefore, the production of light shelves is not standardized but rather customized to fit specific architectural and environmental conditions. Some examples of light shelves and their application can be seen below (Fig. 13 and Fig. 14). Figure 13 (left): Semi-transparent double light shelves made of reflective glass (Littlefair 1996) Figure 14 (right): Top and bottom sections of interior and exterior light shelves with specular surface, showing the path of sunlight (IEA 2000). 24 Categorization Parameters Daylighting systems are generally categorized based on the following parameters: location in façade system, façade design considerations, and performance testing conditions. Location in Façade System Depending upon the horizontal location of the light shelf in the façade in relation to the building interior, they may be classified as either internal or external light shelves, or both. An internal light shelf, in contrast to an external one, has been observed to allow lesser light into the space, even though it is redirected. It has been found to produce solar shading near the façade but with a reduced daylight factor on the work plane throughout the interior space when compared to an equally sized unshaded window (IEA 2000). In contrast, an externally placed light shelf redirects more light into the space than admitted through an equally sized unshaded window. This is because it increases probability of exposure to sky zenith, which produces the highest illuminance. Light shelves are usually most useful on the south façade in the northern hemisphere, and on the north façade in the southern hemisphere. They work best when the sun is at a higher altitude than not which is why they do not perform as well on the east and west facades. Design Considerations There are standards which guide the design, placement and orientation of conventional light shelves. There are tradeoffs that result from the process of selecting which configuration of light shelf if best suited to meet a specific climatic and hourly need. Often, the most important factors that undergo trade-off are solar shading extents, illuminance, and daylighting uniformity. In designing a light shelf, the main objectives are to block direct sun, increase daylight illuminance, uniformity, luminance gradient, and minimize solar heat gains. Most importantly, a conventional light shelf in a space dimensionally resembling an office can increase light penetration by up to 10m (IEA 2000). An interior light shelf is usually recommended to be deep as the height of the clearstory opening (IEA 2000). A greater depth increases solar shading but reduces daylighting uniformity and illuminance near the window. An external light shelf is similar to an internal light shelf in terms of its solar shading principle, but it reduces the amount of area shaded. In such a case, the solar shading is concentrated near the window but there is more illuminance and daylighting uniformity in the space. In winter, neither internal or external light shelves are very effective in blocking low altitude sunlight. Such a situation needs an additional shading device in the space between the light shelf surface and the ceiling to block the low angle solar incidence. At low latitudes, an internal light shelf might need to be deeper to block direct sunlight coming in through the clerestory at all times during the day, but this might compromise the outdoor view. A downward tilt will increase the shading but will simultaneously reduce the light bounced off the ceiling (Fig. 15 and Fig. 16). (IEA 2000) Figure 15: The exterior semi-reflective light shelf in a downward-tilted position (cite from IEA illus.); Figure 16: Interior view of the test room with a semi-transparent, interior light shelf on an overcast winter day (cite from IEA illus.); 25 In a contrasting condition when the light shelf slopes upward, the shading area near the window gets reduced but the indoor illuminance and depth of penetration increases. However, the probability of glare from direct sunlight and from the diffused sky luminance increases. “A horizontal light shelf usually provides the best compromise between shading requirements and daylight distribution (IEA 2000).” To be effective, the ceiling finish above a light shelf and close to the façade needs to be non-glossy with a relatively low specularity. An ideal finish is white diffusing or low-gloss paint. This would ensure that the reflected light close to the light shelf does not cause internal glare. The effectiveness of a light shelf is highly dependent on the nature of its reflective surface. A highly reflective surface would certainly increase illumination in the room but it can also cause reflections of dirt on the ceiling if it is not clean (Lam 1986). A perfectly diffuse surface, called a Lambertian surface, would only diffuse half of the incident light into the room, though there would also be reflections from the inner surface of the clearstory glazing to the interiors (IEA 2000). Ideally, a semi-specular finish works best. Light shelves may also be optically treated. Their surface does not necessarily need to be flat and could be segmented into a concave, convex, or ogee profile. Specular optical films can also increase efficiency of light shelves in some situations (Beltrán et al. 1997) whereas reflective prismatic films can increase the depth of light penetration into the room (Littlefair 1996). In some cases, the material of the light shelf might be semi-transparent. However, such a design strategy might work effectively in a higher latitude location. Another type of light shelf is the sun-tracking light shelf that has movable rollers mounted on a fixed light shelf system. Though it has not been installed in any building to date (IEA 2000), it has a great potential to adapt to different solar angles (Fig. 17). Figure 17: The VALRA (Variable Area Light Reflecting Assembly), (Littlefair 1996, Howard Et Al. 1986) The position of the clearstory glazing is an important consideration. Some light shelves have their interior and exterior shelves within the external glazing. Usually, the clearstory glazing is located in between the external and internal shelves. Maintenance is an important issue for light shelves, especially due to the accumulation of dust on the horizontal surfaces. However, in the case of optically treated light shelf, all the components are housed in a glass enclosure that prevents dust accumulation on the inner components. Dust, however, accumulates on the outer surfaces of the enclosure. First cost is yet another downside of installing light shelves. Performance Testing The effectiveness and performance of light shelves should be tested under a clear as well as an overcast sky and in three different zones depending upon the distance from the facade. The three zones at which the illuminance might be measured can be the window, intermediate, and the rear zone. The sensor distances from the façade differs according to the sky type and the latitude of the location. All measurements are taken on the work plane. For overcast skies and at high latitudes, the average sensor distances from the façade were 1m, 2.75m, and 6.5m. For clear skies, the average sensor distances from the façade were 2.7m, 4.5m and 8.1m (IEA 2000). Below is a list of some of the potential design and operational parameters of the light-shelf system: • Vertical tilt angle • Interior Depth of Shelf • Exterior Depth of Shelf 26 2.12.2 Illuminance Preference The lower limit of illuminance would be determined by the minimum office illumination requirement standard and the top illuminance limit would be determined by the maximum possible illuminance possible on all the task surfaces (Fig. 18). Figure 18: Illuminance Preference Setpoints 2.13 User Preference-Based Ranking Use user preferred comfort ranges and the relative personal ranking of criteria based on personal importance to select the correct sets of control data. 2.14 Summary Chapter 2 provided a background and a compilation of literature reviews conducted to support the research methodology discussed in Chapter 3. Among the aspects explained were the concept of effective daylighting, its planning, availability, and design techniques and control strategies for improving daylighting performance. Section 2.8 explained the concept of visual comfort and performance, providing a review of established visual comfort indices for illuminance, glare, and contrast ratio, as well as factors for accessing them. Section 2.9 highlighted the daylight calculation parameters. Section 2.10 explained important daylighting metrics and calculation methods and Section 2.11 discussed different daylighting system control methods. Metrics for illuminance, glare, and contrast ratio are selected in Chapter 3 based on the literature review of the existing daylighting metrics as discussed in Section 2.10. Section 2.12 described the existing daylighting system, explaining the factors adopted for designing the proposed daylighting system and its components. Section 2.13 discussed various studies on user-preference-based ranking and Section 2.14 summarized the main areas of findings which are essential to the research and furthered explored in the subsequent chapters. The findings from the literature review has been used to formulate a methodology, as explained in the Chapter 3. 27 CHAPTER 3 3 METHODOLOGY The following sections highlight the structure of this chapter, the adopted methodology, the workflow, control logic, important research assumptions and the scope of study. 3.1 Methodology Framework This section explains the structure of this chapter and the methodology framework (Fig. 19). Section 3.2 describes the reference building, the test space, its location, climatic condition, and physical context. Section 3.3 explains the selection criteria for choosing the test space within the reference building. This is followed by Section 3.4 which describes the software platform that the research workflow is based on. Section 3.5 and 3.6 lists the research assumptions and scope respectively. Section 3.7 provides an overview of the parametric definition and its primary components. Section 3.8 describes the control logic and parametric workflow adopted. Section 3.9 highlights the proposed calculation methods, visual comfort metrics, and variables. Section 3.10 describes the process of generation of occupant preferences and importance factors. Figure 19: Scope and Methodology Framework 3.2 Test Space The following segment (Sections 3.2.1 to 3.2.4) discusses the location of the test space, the other viable locations for future studies, the climatic data of the chosen test location, the reference building, and its physical context. 3.2.1 Location A single test location was chosen to evaluate the performance of the proposed control framework, although it would have also been useful to conduct the study for multiple test locations displaying contrasting climates and to draw comparisons among the results. The criteria for selection of test location comprised of five factors- the latitude, annual average dry bulb temperature, annual average relative humidity, and direct normal irradiation. Of the four potential cities- Los Angeles, Chicago, Phoenix, and Miami, representing four different climates, Los Angeles was preferred and chosen because the selected test building, as described later, is also located in Los Angeles with readily available building information and three-dimensional models. Together, the four shortlisted viable test locations were found to 28 be representative of the average and extreme climatic conditions prevalent in the United States. For future studies, the framework might even be tested for these locations and even more. Los Angeles was chosen as representative of a moderately warm and dry climate and the selected test building, 800 Wilshire, is in Downtown Los Angeles. The office of BuroHappold Engineering, on the 16 th floor of 800 Wilshire, was chosen as the location of the test space, though the exact extents of the test area has been defined in section 3.3. The BuroHappold Engineering office layout was assumed to be representative of a generic office space. 3.2.2 Climate Data For selection of the test locations, climate data from weather files provided by the United States Department of Energy were accessed, visualized, and interpreted using Climate Consultant 6.0 and Ladybug tools for the Grasshopper environment. Developed by David Rutten at Robert McNeel & Associates, Grasshopper is a node-based visual programming language and scripting environment that is designed to run with the 3D-modeling program, Rhinoceros. Ladybug is primarily a powerful weather data visualization plug-in for Grasshopper. 3.2.3 Reference Building As mentioned in section 3.2.1, the BuroHappold office in Downtown Los Angeles was chosen as the test space. The reason for selection was the ready availability of detailed drawings and a building information model, accessibility, the relative position of the office in the building, and a scope for contributing to an ongoing daylighting improvement assessment. The office is located on the top floor of 800 Wilshire, a sixteen-story 226,797sf Class A office tower built in 1972 and located at the west corner of the intersection of Wilshire Boulevard and South Flower Street in Downtown Los Angeles (Fig. 20). Figure 20: Location of 800 Wilshire, the reference building (Google Earth: October 25, 2016) 3.2.4 Physical Context Within 50 feet to the west of the 800 Wilshire lies the twenty-four-story Home Savings of America Tower. Due to its relative location, the Home Savings Tower blocks most of the west sunlight during the evening, though there is reported glare at certain times of the year. The façade of the adjoining tower displays use of precast concrete, copper and marble finish. Due to its low window to wall ratio, its facade was observed to not create substantial reflected glare in the test space and thus reflected glare from the facades of adjoining buildings could be neglected. However, being a top floor office space with relatively low obstructions to direct sunlight from neighboring buildings and practically no ground reflectance except for that coming off adjoining perimeter terraces on the north and south sides, the daylighting conditions in the space are not entirely representative of a typical office space. 3.3 Test Space Selection The south-east corner of the office was chosen as the test space. The corner is aligned, such that the south-east facade receives solar incidence from both the south as well as the east directions, and the south-west face from the south and the west. An aerial view shows the orientation and context of the reference building (Fig. 21). 29 Figure 21: Reference Building and Physical Context (Google Earth: October 25, 2016) As the proposed framework is intended to be applicable to perimeter zones of open plan offices, modeling the building corner as a test space and optimizing daylight conditions for it is a research advantage as it accounts for adjacent solar orientations and provides a unique opportunity to test this spatial condition for better understanding the contributions and synergies of each façade in aiding occupant visual comfort. (Fig. 22). Figure 22: Floor plan of the office of BuroHappold Engineering and the test space This created the need to consider and design specific adaptable systems into all the facades for solar control and made it necessary to use different control strategies for each orientation. It was required that the test space potentially be at any of the corners of the floor plan as it was assumed that this would ensure that the space is simultaneously representative of and oriented to two solar conditions- either east or west, and north or south. As the office space has 30 closed rooms in the east corner, only the other corners were taken into further consideration. The corner condition is not a necessary requirement for the functioning of the proposed framework but simply a complex test condition chosen to evaluate the effectiveness of the framework. A second consideration was that the test space should not be daylit just by the north façade. This is because it was assumed that spaces at the center of the north-east perimeter daylight zone would usually not need dynamic façade control to regulate workspace light levels as most of the incident daylight is diffused and difficult to control using methods such as redirection. However, in a typical urban context there might be severe issues of glare from reflected south light from highly glazed facades of buildings bordering the test building on the north. In this case, however, due to an absence of an adjacent tall building in the north, there is a negligible instance of reflected façade glare and thus a north-oriented test space was not considered. Occupant workspaces along the west façade were not considered as there is minimal glare resulting from west sunlight due to the presence of the Home Savings of America Tower in that direction. However, the south-west corner was considered for its corner location. Another important consideration in the selection of the south-east and south-west corners was that the workstations in these test spaces be substantially guarded from direct glare from the south terrace as much as possible as this is not a usual feature in an office. This is to eliminate potential added glare and illumination resulting from direct light bounces from the terrace surface into the interior. 3.4 Model Simplifications The existing selected test space consists of geometrical complexities which have been either left out or simplified to form the test space that was used for the simulations. The selected test space is an L-shaped configuration comprising of 26 occupant workstations that forms the south-eastern and southern perimeter spaces of the office (Fig. 23). Figure 23: The south-east corner test space extents The simplified test space was formed by bounding the corner space (Fig. 24) and consists of only 16 occupant workstations. Each workstation geometry was simplified to decrease the raytracing and simulation time. The test space workstations were formed by L-shaped volumes (Fig. 23), derived from the original geometry, and comprises of two primary perpendicularly-oriented surfaces forming the top of the workstation, and a vertical surface facing the occupant that forms the task background. Of the two top workstation surfaces, the one facing the occupant was retained for sensor allocation and the other not considered for the simulation. 31 Both the vertical as well as the horizontal surfaces were parametrized and equipped with arrays of virtual luminance and illuminance sensors respectively. The occupant workstations were numbered linearly starting from the one closest to the south-eastern corner as the first occupant workstation to the one furthest from the southern façade but closest to the eastern façade as the 16 th occupant workstation. Every workstation geometry and sensor points were numbered in this order, parametrized, and organized using nested list data structures in Grasshopper. This allowed a logical access to each geometrical element in the space (Fig. 23). The occupant seating furniture did not form a part of the simulations. These simplifications and organization resulted in the final test space (Fig. 24). Figure 24: Original test space and simplifications 3.5 Software Platform The parametric definitions were scripted and set up using Grasshopper 0.9.76 for Rhinoceros 5.0. As explained previously (Section 3.2.2), Grasshopper is a node-based visual programming language and scripting environment that is designed to run with the 3D-modeling program, Rhinoceros, and was developed by David Rutten at Robert McNeel & Associates. Rhinoceros is a computer-aided design and NURBS-based three-dimensional modeling software, also developed by Robert McNeel & Associates. The reference drawings for the project were referred from a decentralized Revit file, graciously shared by Sebanti Banerjee from BuroHappold Engineering, Los Angeles, and with necessary permissions for use. Revit is a building information modeling software developed by Autodesk and primarily used by the architecture, engineering, and building construction industries. The interior layout and furniture were either used with modeling simplifications or remodeled as the original in Rhinoceros. Efforts were taken to make the model almost completely parametric so that different model elements can be referenced or re-instantiated through the Grasshopper script definition without having to explicitly refer it through the model space, thus saving time, and aiding model and data organization. Grasshopper’s visual node-based programming enabled laying out of a complex sequence of interconnected processes and use of multiple level data structures to organize data more efficiently, just like a high-level programming language but with a relatively easier learning curve and constant visual feedback. Use of Grasshopper altered the structure and organization of elements in the model space, making it more parametrically accessible to the daylighting simulation program than it would have been in its native state in Rhino or Revit. This organization allowed easy segregation and categorization of data, such as points, vectors, surfaces, solids, architectural elements, simulation parameters, and optical materials. 32 Ladybug, a powerful weather data analysis and visualization software with additional capabilities, was used to gather necessary climatic information and feed it into Honeybee for subsequent calculations. Honeybee is a powerful building energy, lighting, comfort, and daylighting simulation plug-in for Grasshopper which was used to set up the daylighting model and materials, and to perform the daylighting simulation using capabilities of validated simulation engines such as Radiance and Daysim. Grid-based simulation recipe components for Honeybee were used to prepare the model for the daylighting simulation. Honeybee enabled the assignment of material properties, such as reflectance and transmittance, to different objects in the scene such as task surfaces, floors, and glazing. Such a set of materials, test surfaces, points, directionality vectors, and other necessary parameters, were input into Honeybee for the daylighting simulations. Honeybee was used for calculating the illuminance, luminance, and glare Daylight Glare Probability values in the scene. A possible alternative to Honeybee is DIVA, which features similar capabilities. However, Honeybee was chosen as its source code is open and can be edited to tailor-fit workflow requirements. It was also considered advisable to use the same simulation platform throughout to minimize error, and Honeybee was chosen as it provides added functionality. Thornton Tomasetti’s plug-in for Grasshopper, TT Toolbox, was used to access, read, and write to Excel worksheets. The plug-in offered near real-time data access and updating capabilities. It was used extensively and multiple times in the definition, both to read from pre-existing data sets as well as to store values calculated in Grasshopper. The Excel read/ write component in Lunchbox, a multi-functional Grasshopper plug-in developed by Nathan Miller, could also be used to access worksheets. Python 3.6 was used for generating occupant preference models by randomizing selection from predefined sets of occupant preferences and importance factors. Using Honeybee’s source code editing capabilities and Grasshopper’s Python editor, the program could be run directly inside Grasshopper and was integrated with the main workflow. A facilitator tool was developed in Excel for calculating and analyzing the resulting illuminance, luminance, and glare simulation data as produced in Grasshopper. The functionality of the Excel tool has been described in greater detail later in this chapter. The software workflow integrating the described parts adopted is as shown below (Fig. 25). Figure 25: Proposed software workflow 33 3.6 Control Logic The control logic represents the complete process of how the adaptive façade is controlled in a real-life application (Fig. 26). This framework does not represent the aspect of physical control or kinetics, but rather the logic that guides the selection of which façade state is optimum for improving the visual comfort satisfaction of all occupants at any given point in time. The control logic has been designed using Rhino, Grasshopper, Honeybee, and Excel. In a real- world application, this logic may be run using a different set of applications, and may or may not be required to be run in real-time for every time-step for which the façade configuration changes. The control logic essentially consists of two main parts. The first part virtually simulates daylighting data or collects real-time data through physical sensors. The second part of the control logic computes the individual and cumulative visual discomfort based on the simulated or measured illuminance, glare, and luminance values produced in the first part. Alternatives to using real-time simulation data would be performing calculations would either be derived from machine-learnt data and historical data, or from real-time calculations. In a real-world application, physical sensors may or may not be used to measure the illuminance, luminance, and glare values. The proposed control logic lays out the possibility of using real-time or historically simulated data in real-time calculations using a façade controlling computing resource (Fig. 26). Figure 26: Proposed control logic 3.7 Parametric Definition The proposed control logic has been implemented using a parametric definition. The purpose of the parametric definition is to represent the iterative logic and process of decision-making that is used to control the facade. In a real- world application, the proposed parametric definition would need to be adapted as per several real-world considerations as discussed in Chapter 5. Sections 3.7.3 to 3.7.10 lays out the parametric definition. 3.7.1 Workflow Modules - Grasshopper The parametric workflow in Grasshopper is divided into 7 modules, 4 within Grasshopper and 3 within Excel (Fig. 27). Each module is designed to work both for the test case as well as the baseline case which are eventually compared in the ‘Data Analysis’ module. Results from the first four modules are written to Excel to undergo operation using the last three modules. The following provides an overview of the flow of control of data among the modules (Fig. 27). 34 Figure 27: Modules forming the parametric definition The following are the list of modules in Grasshopper and Excel which are designed to work in succession. In simple terms, the Grasshopper modules are used to take in all inputs, control the daylighting system, generate the occupant data sets, and to calculate illuminance, luminance, and contrast ratio values by running simulations. The Grasshopper script is made of these modules: Input, Daylighting System, Occupant Data, and Simulation (Fig. 28). Figure 28: Implemented parametric definition in Grasshopper 3.7.2 Workflow Modules - Excel The Excel modules: Calculation, Occupant Preference, Data Analysis, and Visualization- form part of an Excel facilitator tool that are used to generate the results based on the output of the first four modules. 3.7.3 Input This comprises of the climate data, sensor count, placement, and directionality information, optical properties of materials in the scene, existing test space, workstations, and urban context geometry, low-height partitions, the interior layout, occupant eye position, and test space boundaries. Weather Data Location-based weather data is necessary information for running the simulations. The type of weather data used is the Typical Meteorological Year (TMY), which is basically an annual collation of hourly weather data sourced from a weather data bank averaged over a specific period, usually several years. TMY data is frequently used in simulations that are dependent on weather information, such as daylighting or energy simulations for buildings. Depending on which historical period and number of stations were used to collect data for forming the TMY collection, TMY’s can be categorized as TMY, TMY2, or TMY3. The TMY3 data set contains data collected from 1020 stations across the United States and certain adjoining locations, derived from two periods, 1976-2005 and 1991-2005. The TMY data seeks to represent an average estimate of weather conditions based on historic data. The weather data is stored for every hour of the year in a .epw weather file published by EnergyPlus and presents data such as dry-bulb temperature, precipitation, humidity, and direct solar radiation. As the reference building is in downtown Los Angeles, the climate data is referred from the .epw file containing data recorded at the weather station nearest downtown at 34.033 o N and -118.217 o W. 35 Climate-Based Daylight Modeling The appearance of the sky can be very dynamic and unpredictable, transitioning from partly cloudy to overcast or clear in a matter of hours and the extents of change varies with time, season, and year (Fig. 29). Figure 29: Sky Types (“Sky Types” 2018) As the reflected luminance from the sky substantially contributes to the overall daylight in a space, it is necessary to be able to predict an approximate and average sky condition for a given time. This has been achieved using mathematical models that attempt to three-dimensionally model the sky and calculate luminance distributions according to different algorithms. The International Commission on Illumination (CIE) has attempted to create such sky models, as shown below. It classifies skies as clear, intermediate, overcast, and uniform. A climate-based sky technique models the solar radiant and luminous quantities and distributions of the sky using the actual geographical, solar, and weather information extracted from meteorological data sets. Using the TMY3 weather data set for downtown Los Angeles, the ‘Honeybee Generate Climate-based Sky’ is used to create sky-luminance distributions for given times of the year, for which daylight simulations will be performed, and for a given direction representing the geographic North. Simulation Period The simulations are run for three days of the year- June 21, September 21, and December 22- and for three times of the day, at 9am, noon, and 3pm. These three days are chosen as closely representative of the annual highest, lowest, and intermediary solar altitudinal conditions for the specific latitude. As the proposed research strives to create a framework for controlling an adaptive façade, it’s proposed results would be the possible percentage improvement in how closely the given set of users’ preferences are met for the proposed case and a baseline case. Assuming all users have a fixed set of visual comfort preferences throughout the year, a change in the day of simulation might affect the percentage improvement in a positive or negative way. However, this change will not be dependent on the number of times of the year the simulation is performed. As the resources to perform multiple simulations and optimizations is limited, only a few runs are performed, with the assumption that boundary solar conditions on the selected dates and time are somewhat representative of the total range of indoor daylighting and glare conditions for the active hours of the building operation. In all, nine optimizations are performed. Each optimization run is preceded by two sets of simulations, one calculating illuminance and the other, glare. The other parameters, such as the contrast ratio are mathematically derived from the illuminance values obtained already. Existing Geometry The following are all the existing scene geometry that have been either re-referenced as it is or simplified followed by parametrization into Grasshopper for better organization. These included the test surfaces, points, and vectors; optical material properties; occupant preference and importance factors; and simulation parameters. Reference Building The office was modeled in detail for daylighting simulation in Rhino as per the actual drawings and BIM Revit model received from BuroHappold Los Angeles. The south-east corner was detailed in greater complexity. To be representative of a typical office floor, the south-side open terrace and the building façade profile along the south-face has been matched with the façade profile of the 16 th floor. The entire floor extents are 98’ x 149’9”. The test space is selected at the southern side of the building and along the longer length of the building (149’9”) but only up to a depth of 33’9” from the southern façade. The overall test space dimensions are therefore 33’9” x 149’9”. The original floor plan features more than 52 workstations. However, for the purposes of this research, occupancy is considered only in the south-east corner and therefore unnecessary low-height furniture from the west end of the space was removed to simplify the model. Likewise, floor sensor points are also located only in the test space under study. 36 A total of 16 workspaces were considered for this study. Each occupant workstation is considered as a unit consisting of sensor points, occupant viewpoints, task surfaces, and geometry. All the workstations are instantiated into the Grasshopper definition and the data organized into multi-level list such that any particular task surface, point, or element can be accessed from the list, for example by specifying the occupant workstation index location. Each workstation is enclosed within a 5’ x 5’ rectangular floor space and is comprised of a primary and a secondary task surface. The larger task surface measures 5’ x 2’6” whereas the smaller adjoining task surface measures 3’ x 1’6”. Facing workstations are separated by low 4’ high opaque partitions. Each of the larger task surfaces comprise of two adjacent visual display units facing the occupant seating furniture. Referenced Surfaces The following test surfaces: floor, primary task surface, walls, partitions, ceiling, curtainwall frame, glazing, and visual display units- were isolated from the existing geometry for purpose of use in the simulation. The glazing surfaces are stored in three lists for the spandrel, vision, and clearstory glazing surfaces. The façade enclosing and separating the test space from the exterior is made up of nine planar surfaces. Each surface is divided horizontally and vertically by transoms and mullions into sub-surfaces. The list structure for each glazing type, say the vision glazing, comprises of nine levels representing the nine façade surfaces, each level containing multiple sub-levels of vision glazing panels in that surface. These three primary lists of the three glazing types are combined and stored as sub-lists in a main list structure in Grasshopper. Storing the individual glazing surfaces in one list allows the workflow to easily access and reference any instance by simply using its list structure address. The visual display unit, and the primary and secondary task surfaces, are stored similarly using a nested list data structure. The other surfaces such as walls and ceilings are stored in non-listed lists as they do not require such a type of data organization. Test Points & Vectors For each surface, grid-based test points are defined to position virtual sensors that would measure either the horizontal or the vertical illuminance. Unitized, or a unit length, of vector was defined at each virtual sensor location to represent the direction at the test point which is normal to the test surface. By knowing this direction, the software can calculate which side of the test surface it should place the sensor points to measure illuminance. Occupant Viewpoint A virtual sensor point was located at the eye position of the occupant based on ergonomic standards and the visual display dimensions (Fig. 30 and Fig. 31). This is used to calculate the simplified DGP glare index for all those cases when the sun is not directly in the field of view of the occupant. Figure 30: Occupant Viewpoint Location (The Ergonomics of Standing And Sitting, https://cabin.sickchickchic.com/att/the- ergonomics-of-standing-and-sitting-amazing-ergonomic-standing-desk-height-6-7434856/) Figure 31: Visual Display Unit (View Sonic, https://www.viewsonic.com/us/vx2253mh- led.html?dir=asc&mode=list&order=name) Task Surface Each workstation task surface consists of two sub-surfaces, one in front of the occupant and the other to the side. Five sensor points are placed on the larger surface and along the centerline at a spacing of 1’3”. Three sensor points are 37 placed on the smaller surface and along the centerline parallel to the longer side at a spacing of 1’6”. Overall, eight sensor points are used to take task surface readings for each workstation. A nested list structure is used to store the sensor locations, containing 26 sub-lists, representing the 26 workstations, and 8 sensor point locations for each sub- list. Task Background The partition separating the facing workstations also serve as the background for the displays. The amount of contrast between the luminance of the screen to the background affects visual comfort of the occupant. Three virtual sensor points, aligned with the horizontal task surface points, are positioned along the centerline of the visual display unit and on the partition surface facing the occupant. Simulation Parameters The Radiance simulation parameters were set up as following (Table 1 & Table 2). Radiance illuminance and luminance simulation parameters were selected to produce a more accurate simulation rather than a faster and inaccurate one (Table 1). Here, ‘ab’ represents the number of ambient bounces, ‘ad’ the number of ambient divisions, ‘as’ the number of ambient super-samples, ‘ar’ the ambient resolution, and ‘_aa_’ the ambient accuracy (Jakubiec and Reinhart 2011). Table 1: Radiance Simulation Parameters Illuminance & Luminance Simulations Simulation Parameter Abbreviation Value Ambient Bounces ab 5 Ambient Divisions ad 1024 Ambient Super-samples as 16 Ambient Resolution ar 256 Ambient Accuracy aa 0.10 Optical Material Properties The optical material properties of the interior surfaces are constant except for the glazing surfaces. For the daylight simulation, the following optical properties were assigned to the geometrical elements: a diffuse reflectance of 0.20 for the floor, 0.50 for the interior walls, 0.80 for the ceiling, and 0.20 for the exterior ground (Table 2). However, as electrochromic glazing is used as a dynamic element, its transmittance is variable. The reflectance of the upper surface of the exterior solar shading was assumed to be 0.90. Table 2: Optical Material Properties Optical Material Properties Element Value Floor 0.20 (Reflectance) Ceiling 0.80 (Reflectance) Walls 0.50 (Reflectance) Exterior Ground 0.20 (Reflectance) Glazing Variable transmittance Occupant Preference and Importance Factors In a real-world application of the proposed framework, occupants themselves set their own preferences and importance factors for glare, illuminance, and contrast ratio, using a virtual interface, such as a mobile or computer application. These are then processed by the façade controller in near real-time using the proposed control logic and which is used to determine and set the best façade configuration state. Occupant preferences are discussed in greater detail in a subsequent section in this chapter. 3.7.4 Daylighting System This module consists of the parametric definitions constructing the three adaptive daylighting systems that been studied. Each case is studied separately and thus the workflow provides a mechanism to isolate one active definition by selection. A single multi-level data structure is used to store the intrinsic data for each daylighting system, such as glazing index, vertical and horizontal position, orientation, sequence, and transmissivity. This data structure serves as 38 an information model providing all necessary information relating to the daylighting system for use in the subsequent modules. 3.7.4.1 Existing Daylighting System The purpose of documenting the existing system is to generate simulation results that are very close to real measurements made in the space. The simulation results, and hence the workflow, could then be calibrated by comparing the two sets of data. The selected test building displays a central service core and perimeter open workspaces and closes offices. The office is daylit by floor to ceiling fully glazed windows with a uniform head height (HH) of 13’ 11 ¾” measuring from the finished floor finish to the top of the opening. The opening is split into three glazing areas- a 1’ 10 ½” spandrel glazing measured from the finished floor level to the centerline of the sill frame, a 10’ 10 ½” high central vision glazing measured between centerlines of the sill and the transom, and a 3’ 1 ¼” high clearstory. Only the central vision glazing has operable roller shades. 3.7.4.2 Proposed Daylighting System The intent of proposing a new daylighting system rather than using the existing system assumes that a design improvement in the workflow of effective daylighting will get us closer to the high performance and comfort values expected of the system using minimum control, strategies, and by streamlining the workflow. Therefore, the first step in the workflow proposed in this research is the design of the daylighting system. The proposed daylighting system here is to be considered as a test example displaying multiple control techniques and variables (Fig. 32). However, the proposed workflow should work with just any other side-lighting system having, possibly, a different set of control variables. A potential daylighting system is assumed to provide better daylighting in the space to improve occupant’s visual comfort besides reducing the electric lighting use. Figure 32: Proposed Daylighting System Design Considerations Several parameters and design factors were considered while designing the proposed system, while several parameters were not. Though cost is a very important factor, it was not considered. Other factors such as kinetic performance, maintenance, mechanical complexity, thermal properties of the system were also not considered. As this design proposition is exemplary of an adaptive system, it is assumed that a better performing cost-effective system will be designed as per specific needs in the future. However, geometric, material, and kinetic parameters which directly affect the indoor daylighting condition were considered as part of the optimization process. Below is a list of key factors that were considered in choosing the system and which were adapted and modified (IEA 2000). • Site Location & Context • Sky and Daylighting Conditions: The criteria for selection of the system was applicability in overcast, clear, as well as partially clear sky conditions. 39 • Daylighting Objectives: The criteria for selection were solar shading, diffuse light redirection, glare protection, and daylighting uniformity. The criteria for glare protection was that the system should be able to minimize direct • Visual Amenity: The criteria for design and selection of the daylighting system was provision of clear unobstructed or partial views of the surrounding context but not of the sky. • Operational Constraints: The criteria for selection was a system which could be tracked diurnally as well as seasonally. Preference was given to systems with reduced mechanical complexity. • System Availability & Building Integration: An essential requirement for the selection of the system was its market availability and potential for building integration. Systems in the testing stage were not considered due to limited understanding of their applicability and effects. Performance Objectives & Expectations Below is a list of the major daylighting objectives and expectations from the proposed system that were considered: • Redirection of light to spaces away from the façade and to zones with low light levels. • Provide acceptable task and ambient illumination levels. • Create locally uniform lighting conditions by keeping light levels within acceptable illuminance contrast ratios, both on horizontal planes such as work surfaces, floors and ceilings, and on the surrounding planes such as walls. • Maintain luminance contrast ratios within acceptable limits and minimize discomfort glare. • Maximize the quality of outdoor views. • Provide solar shading close to the façade. Design Scope There are also factors that were not considered. • Economic Constraints & First Costs • Interior and Exterior Aesthetic Considerations System Design & Optical Properties A composite daylighting system was designed (Fig. 3.7.4.1). The total façade opening height is 14’. The spandrel glazing has a low visible transmittance, is unshaded, and goes up to 2’ from the finished floor surface. The façade opening above the spandrel is split into two parts- a vision glazing component and a clearstory component. Electrochromic Glazing The vision glazing was designed as an electrochromic glazing having four switchable tint states, based on a typical Sage dynamic glass electrochromic glazing product. The electrochromic glazing can take up four tint states having visible light transmittances of 1%, 6%, 18%, and 60%. The tint shades for glazing panels of a given façade are considered to be equal. However, the tint shades of glazing panels of two different facades can be different. Each façade takes up any one of the tint shades: 0.01, 0.06, 0.18, and 0.60. Exterior Solar Shading The exterior solar shading is designed to be hinged at the transom level of vision glazing. It is extendable up to lengths of 36”, 60”, and 84”, and can be rotated starting from an angle of zero degree measured from the horizontal, through 30 and 60 degrees, and up to an angle of 90 degrees and parallel to the vision glazing. 3.7.5 Occupant Data This module was used to randomly generate, store, and access occupant preferences and importance factors from existing ranges. A random normal distribution function in Python was used to randomly select the preferences and importance factors from predefined lists. Preference Types Two types of preferences were used to model the user preferences- discrete values or setpoints, and equally-spaced ranges. This was done as in reality, an occupant may choose a single specific preference value, or simply a range. Also, these ranges are suggestive and have been either derived or assumed to implement the proposed methodology. Other setpoints or ranges can also be used to form specific occupant models. 40 For glare range-based preferences, the existing IES standard ranges were extrapolated for the entire set of possible values of the metric (0.20 to 0.45). Each sub-division was therefore equal to 0.05. For the illuminance-based ranges, the IES minimum perceivable light level change in an indoor environment (150 lx) was used to divide the entire range of acceptable illuminances (300 lx to 2000 lx) into 11 ranges. For the luminance contrast ratio, the IES recommended acceptable range of 1:3 to 3:1 for the primary field of view cone was used. This range was divided by 0.33 intervals to form 7 ranges. The 0.33 interval was chosen arbitrarily and can be smaller or larger depending on how fine the control needs to be. In fact, as mentioned, these preferences are exemplary and not binding. Therefore, depending on how fine or exact the daylighting control might need to be, the preference ranges may be divided into finer or larger ranges. For setpoint-based preferences, specific values are selected by the occupant which lie within the acceptable glare, illuminance, and contrast ratio ranges. The following range-based preferences were used to model the user preferences that were used for further calculation. • Glare (DGP metric): 0.20-0. 25,0. 25-0.30, 0.30-0.35, 0.35-0.40, and 0.40-0.45. • Illuminance: 300-450,450-600, 600-750, 750-900, 900-1050, 1050-1200, 1200-1350, 1350-1500, 1500-1650, 1650-1800, 1800-1950. • Contrast ratio: 0.33-0.67, 0.67-1.0, 1.0-1.33, 1.33-1.67, 1.67-2.0, 2.33-2.67, 2.67-3.0. Personal Importance Factors A set of personal importance factors and ranging from a minimum of 5 and a maximum of 90 was chosen. The step- factor for the importance factors is 5. These factors were used to specify and quantify the personal preference for any visual comfort factor. A Python code featuring random number selection was used to generate 3 importance factors, for glare, illuminance, and contrast ratio, such that the total of the importance factors was 100. Based on the personal importance factors, a personal ranking of the visual comfort criteria was derived. For example, the preferences and importance factors (IF) of three occupants, A, B, and C, can be modeled as shown below (Table 3). It should be noted that for each occupant, the total of all the three importance factors needs to be equal to 100, as per the proposed methodology. Table: 3: Example of Preference Modeling for Three Occupants Occupant Glare Preference Glare Importance Factor (IF) Illuminance Preference (lux) Illuminance Importance Factor (IF) Contrast Ratio (CR) Preference CR Importance Factor (IF) Total IF A 0.20-0.25 10 750 85 0.33-0.67 5 100 B 0.35-0.40 90 1950 5 2.67-3.00 5 100 C 0.25-0.30 30 300 20 0.33-0.67 50 100 The techniques adopted to calculate the individual and cumulative visual comfort dissatisfaction, and the performance scores of façade configurations, based on the visual comfort simulation or measurement data, and the occupant preference data, is discussed in detail in Section 3.7.8 of this chapter. The implications of these decisions and calculations are discussed in Chapter 5. 3.7.6 Simulation This module contains the portion of the workflow that performs grid-based illuminance and luminance simulations, as well as simplified glare calculation. 3.7.7 Excel Write This module contains the portion of the workflow that writes all the simulation and other raw data to the Excel file. 3.7.8 Calculation This module uses the user preferences from the ‘Occupant Data’ module and the simulation results from the ‘Simulation’ module to calculate preference-based weighted totals. In the first step, the deviation of each of the user- preference from the measured illuminance, luminance, and glare values are calculated. To do so, a logic test is used to determine if the measured values lie within the preferred ranges. If not, the differences between the measured value and the bounding values of the preference are measured. The smaller value of these represents the deviation of that measured visual comfort factor from the respective occupant’s preferred range. In the next step, an importance factor weighted squared average of the individual deviation values is calculated to represent the net deviation from preference 41 for each occupant. In the final step, the cumulative addition of the individual deviations is calculated. Using net deviation and cumulative deviation values for the 16 occupants, the median and average deviation values are calculated, including a standard deviation of the deviations from zero, which represents the ideal case of no deviation, which would mean that all occupants’ preferences have been successfully met using that particular daylight system configuration. Each facade daylighting system has three types of controls: visible light transmittance of the electrochromic glazing (4 states), exterior light shelf and shading angle (4 states) and length (3 states). As the two façades operate independently, this allows a total of 2304 (4 x 4 x 3 x 4 x 4 x 3) combined daylighting system configurations. For each configuration, a simulation was run to calculate all the above stated values and which was continuously written as rows to an Excel file using the TT Toolbox plug-in for Grasshopper. Simultaneously, to visually record the induvial measurements, configuration states, and the illuminance distribution in the space, the data was visualized in the Rhino model space using annotation tags and color. Illuminance, luminance, glare, contrast ratio, and the Useful Daylight Illuminance (UDI) values were calculated and displayed in the visualization. This entire process was performed parametrically for 2304 times using a system of logic gates, flow of control, and useful plugins for Grasshopper and stored for further analysis. The 2304 sets of data represent the measurements and calculations for all the 2304 possible daylighting system configurations for a given point in time. This process was repeated for 3 times of the day, (8am, noon, and 4pm) for three days of the year (June 21, September 22, and December 22). Each simulation was expected to take a few minutes on average to complete. Estimating from this, the time at hand, and the available computational resources, it was concluded that a total of 6 times of the year would be used as test periods and which was expected to generate exactly 21,600 sets of simulation data resulting from 2304 grid-based illuminance and luminance simulation data. A snapshot subset of the Excel calculation module has been shown (Fig. 33). Figure 33: Excel Computation and Analysis Proposed Metrics and Variables A new set of metrics have been proposed for measuring and analyzing the outcome. Section 4.1 of this chapter explains these, highlighting the list of associated terminologies, a discussion of their implications, ranges, criteria, units, and symbolic representations. These have been extensively used throughout this chapter to measure the outcome of each proposed case and this discussion holds for each such case. To avoid a lot of repetition, hese have been referred to directly without additional explanation. An additional explanation been provided to only in certain exceptions to further elucidate the content. It must be noted that these metrics are either unitless or are represented as positive or negative percentages. The cumulative dissatisfaction (D CL ) value is used to represent the sum of the occupants’ individual dissatisfactions, represented by D i , where ‘i’ represents the number denoting the position of the occupant in the space and ranges from 1 to 16 corresponding to a sequentially-numbered seating arrangement. The baseline individual and cumulative 42 dissatisfaction values, represented by D base and D CLbase respectively, measure the individual and cumulative dissatisfaction values of occupants calculated for the baseline condition. The baseline condition is a certain chosen static configuration of the façade which remains the same throughout all the cases and points in time. When this configuration is used for different points in time, it produces different measurements of glare, illuminance, and contrast ratio calculated for all the sixteen occupants. Each point in time is further analyzed using ten different occupant preference cases. For a given point in time these values of glare, illuminance, and contrast ratio remain the same, irrespective of which case is being analyzed, as these measurements are independent of the occupant preference case. For example, the glare, illuminance, and contrast ratio values calculated for all occupant workstations will be the same for a given day and time. However, for a certain configuration under study, at a given point in time, the individual and cumulative dissatisfaction values are dependent on the importance factor-weighted deviation of the calculated glare, illuminance, and contrast ratio values from their preference values, and its aggregate. Thus, this makes the individual and cumulative dissatisfaction values dependent on the occupant preference cases being analyzed, resulting in ten different baseline values of individual and cumulative dissatisfaction, each of which is then compared with their measured case for each of the 2304 configurations. As explained, for each case the measured individual and cumulative dissatisfaction conditions are compared against the baseline values of that case. The measured dissatisfaction values might be greater, lesser, or equal to the baseline values. The ideal dissatisfaction value is zero, which implies that an occupant or all the occupants cumulatively are completely satisfied, and that their preferences have all been successfully been met. However, this is almost never the case, as discussed in chapter three. The cumulative dissatisfaction is the sum of all the individual dissatisfactions (Equation 9), DCL=∑ Di Equation 9: Cumulative Dissatisfaction as sum of Individual Dissatisfaction values where D CL is the Cumulative Dissatisfaction, D i is the individual dissatisfaction, and ‘i’ is the number representing the occupant and it takes up all the values in the range of 1 to 16. A dissatisfaction value for a case, when greater than the respective baseline value indicates an increase in dissatisfaction and is not desired. On the contrary, a dissatisfaction value that is lesser than the baseline value is desired. The absolute difference between the calculated and baseline cumulative dissatisfaction values is represented as a percentage of D CL by the term Percentage Reduction in Cumulative Dissatisfaction (P r ). Similarly, the absolute difference between the calculated and baseline individual dissatisfaction values is represented as a percentage of D i by the term Percentage Reduction in Individual Dissatisfaction (p r(i) ). Therefore, a positive reduction value implies a decrease in cumulative or individual dissatisfaction, whereas a negative reduction value implies otherwise. The Mean Positive Reduction (P r(mean) ) represents the arithmetic mean of only the reduction values which are positive and not considering the values which correspond to an increase in cumulative or individual dissatisfaction. The Percentage Partially Satisfied (P ps ) represents the number of occupants, as a percentage of the total number of occupants, whose individual dissatisfaction values are lesser than the corresponding baseline individual dissatisfaction values D i(base) . For example, a P ps value of 75% would mean that 75% of the total 16 occupants (12 occupants) have had their visual comfort increased by differing extents and that the remaining four occupants have had their individual visual comfort reduced by varying extents, both in comparison with the baseline case. The proposed framework is used to test 2304 façade configurations to recommend the best performing ones, ranked based on certain criteria, as discussed later in this section. The order of testing of façade configurations is kept fixed to be make comparison and analysis more systematic. Each configuration is referred to by its index, termed the Configuration Index (C index ), and which can range from 1 to 2304. Each configuration comprised of a set of control variables having certain control states or control values, which represented physical attributes of the adaptive façade elements, as discussed in chapter three. Each configuration represented a certain state of the adaptive façade under study. The control variables and their control values, along with their abbreviations, which includes the addend SE or SW, representing the orientation, either south-east or south-west, of the façade element which is being controlled have been documented (Table 4). 43 Table 4: Control Variables and Control Values Control Variable Façade Element Notation Units Control Values Shading Angle South-East Exterior Shading A SE degree 0, -30, -60, -90 Shading Angle South-West Exterior Shading A SW degree 0, -30, -60, -90 Shading Length South-East Exterior Shading L SE inch 36, 60, 84 Shading Length South-West Exterior Shading L SW inch 36, 60, 84 Visible Light Transmittance South-East Vision Glazing VT SE <unitless> 0.01,0.06,0.18,0.60 Visible Light Transmittance South-West Vision Glazing VT SW <unitless> 0.01,0.06,0.18,0.60 As there were two independently controlled adaptive facades in this case, each configuration constituted a collection of two such sets of control variables and control values. For the purposes of notation, a façade configuration under study is represented using the following (Notation 1), Configuration <Cindex> (<ASE>, <ASW>, <LSE>, <LSW>, <VTSE>, <VTSW>) Notation 1: Format for Representation of Façade Configuration State where C index refers to the configuration index, and each variable has been explained in Table 4.1. An example of such a notation would be Configuration 481 (0, -90, 60, 36, 0.06, 0.60) representing the 481 st configuration having a zero- degree south-east façade exterior shading angle; a 90-degree south-west façade exterior shading angle; a 60-inch (5 ft.) south-east façade exterior shading length; a 36-inch (3 ft.) south-west exterior façade shading length; a 0.06 (6%) visible light transmittance for the south-east façade vision glazing; and a 0.60 (60%) visible light transmittance for the south-west façade vision glazing. A listing of the occupant preference and importance factor variables, abbreviations, possible ranges & values have been provided (Table 5). Table 5: Occupant Preference and Importance Factor Variables, Abbreviations, Possible Ranges & Values Variable Abv. Ranges / Possible Values Occupant Index (Occ. #) i 1<=i<=16 Glare Range R g(i) 0.20-0.25,0.25-0.30,0.30-0.35,0.35-0.40,0.40-0.45 Lower Bound: Glare Range R gL(i) 0.20,0.25,0.30,0.35,0.40 (n th value) Higher Bound: Glare Range R gH(i) 0.25,0.30,0.35,0.45 (n+1) th value Glare Importance Factor If g(i) (5=<If g(i) <=90) AND mod(If g(i) ,5) =0 Half-Glare Range Interval G intv = (0.05/2) or 0.025 Illuminance Range R e(i) 300-450,450-600,600-750,750-900,900-1050,1050-1200,1200- 1350,1350-1500,1500-1650,1650-1800,1800-1950 Lower Bound: Illuminance Range R eL(i) 300,450,600,750,900,1050,1200,1350,1500,1650,1800 (j th ) Higher Bound: Illuminance Range R eH(i) 450,600,750,900,1050,1200,1350,1500,1650,1800,1950 (j+1) th Illuminance Importance Factor If e(i) (5=< If e(i) <=90) AND mod(If e(i) ,5) =0 Half-Illuminance Range Interval E intv = (150/2) or 75 Contrast Ratio Range R cr(i) 0.33-0.67,0.67-1.00,1.00-1.33,1.33-1.67,1.67-2.00,2.00- 2.33,2.33-2.67,2.67-3.00 Lower Bound: Contrast Ratio Range R crL(i) 0.33,0.67,1.00,1.33,1.67,2.0,2.33,2.67 (k th value) Higher Bound: Contrast Ratio Range R crH(i) 0.67,1.00,1.33,1.67,2.00,2.33,2.67,3.00 (k+1) th value Contrast Ratio Importance Factor IF cr(i) (5=< IF cr(i) <=90) AND mod(IF cr(i) ,5) =0 Half-Contrast Ratio Range Interval C intv = (1/6) or 0.167 Total Importance Factor IF tot =100 For each occupant and the associated workstation, and for each point in time under analysis, glare, illuminance, and contrast ratio, calculations are made. These are represented by G i , E i , and C i respectively, where ‘i’ represents the number of the occupant. To summarize, the glare and illuminance values are calculated using values from the grid- based simulation calculating horizontal and vertical illuminance, whereas the luminance contrast ratio is calculated based on values produced by a grid-based simulation calculating luminance values. Deviation of each of these values 44 from their individual preferred values is calculated and then weighted using their respective importance factor to produce individual glare, illuminance, and contrast dissatisfaction values using the following formulae. For calculating individual glare dissatisfaction, Equation 10 is used where G intv = 0.025 (Table 5). Dg(i) = √({|Gi – [(RgL(i) + RgH(i))/2]| / Gintv x Ifg(i)}/300) Equation 10: Individual Glare Dissatisfaction For calculating individual illuminance dissatisfaction, Equation 11 is used, where E intv = 75 (Table 5). De(i) = √({|Ei – [(ReL(i) + ReH(i))/2]| / Eintv x Ife(i)}/300) Equation 11: Individual Illuminance Dissatisfaction For calculating individual contrast ratio dissatisfaction, Equation 12 is used where C intv = (1/6) or 0.167 (Table 5). Dcr(i) = √({|Ci – [(RcL(i) + RcH(i))/2]| / Cintv x Ifc(i)}/300) Equation 12: Individual Contrast Ratio Dissatisfaction The individual dissatisfaction is calculated using Equation 13: Di = Dg(i) + De(i) + Dcr(i) Equation 13: Individual Dissatisfaction where D i is the individual dissatisfaction, and D g(i) , D e(i) , and D cr(i) are the glare, illuminance and contrast ratio dissatisfactions calculated using Equations 10 to 12, as shown previously. D CLmean represents the mean cumulative dissatisfaction, whereas D CLmed represents the median cumulative dissatisfaction, and D CLmode represents the mode cumulative dissatisfaction. The mean and mode values are used to get an idea of the skewness of the cumulative dissatisfaction curve plotted against the configuration numbers. The median values are used to get an idea of dissatisfaction values associated with at least 50% of the configurations. The reduction (R CL ) in cumulative dissatisfaction from the baseline cumulative dissatisfaction (D base ) and the percentage reduction (P r ) are calculated using Equations 14 and 15. RCL= Dbase - DCL Equation 14: Reduction in Cumulative Dissatisfaction Pr=(RCL/DCL) x 100 Equation 15: Percentage Reduction in Cumulative Dissatisfaction A negative reduction value in cumulative dissatisfaction would mean that the calculated case performs worse than the baseline. In that case, the percentage reduction value would also be negative. Such a case is not desirable. The number of configurations for which there is a positive reduction in cumulative dissatisfaction compared to the baseline case are considered better performing configurations. This is represented by N better . The number of configurations for which there is a negative reduction in cumulative dissatisfaction, or an increase in cumulative dissatisfaction, are considered worse than the baseline configuration. This number is represented by N worse . The percentage of configurations performing better than the baseline of the total number of configurations is referred by P config . The configuration performance score is used to rank the configurations from the most acceptable to the least acceptable. The ranking also highlights the worst performing cases and the configurations which need to be avoided. The performance score and ranking can be conducted in two different ways: a direct ranking method using Percentage 45 Reduction in Cumulative Dissatisfaction (P r ) value, and a performance score method which uses the P r and P ps values together as described below. The direct ranking method uses the Percentage Reduction as a measure of façade configuration performance. However, it must be noted that this value signifies higher cumulative satisfaction levels but does not consider how many people are getting satisfied. For example, a Percentage Reduction of 30% means that the cumulative dissatisfaction of all occupants combined has been reduced by 30%. This implies that all the occupants as whole are more satisfied. However, this does not guarantee a rise in individual happiness of all the occupants. In this case, for example, 50% of people might not have experienced a rise in visual comfort satisfaction, whereas 50% might have. For the latter, the cumulative rise in satisfaction might be higher than the fall in satisfaction for the former, causing the 30% overall rise in cumulative dissatisfaction. To get a better sense of how the calculated cumulative dissatisfaction affects individual occupants, it is important to also consider the percentage partially satisfied (P ps ) value. If two configurations produce equal cumulative dissatisfaction, the one with the higher P ps value would be the better choice as it satisfies a higher percentage of people. Now, if the P ps for two cases were to be equal, it would mean that both configurations make equal number of people more satisfied but it would not indicate which one increases satisfaction more. In such a case, the one producing the higher percentage positive reduction in cumulative dissatisfaction would be the ideal choice. Therefore, to bring together these two aspects, a product of the two values, P r and P ps , would be the correct indicator of higher performance. This was represented as the Configuration Performance Score (C sc ) and calculated as explained (Equation 16). Csc = 10 x log10 (Pr / Pps) Equation 16: Façade Configuration Performance Score The Configuration Performance Score metric has a defined range, with a minimum possible value of zero and a maximum possible value of 100. The C SC is equal to zero when P ps and P r values are both zero, which is the baseline case. This is because, for the baseline case, none of the occupants are satisfied with the baseline configuration, and the percentage reduction in cumulative dissatisfaction is also zero, as by definition. The CSC is equal to 100 when both P ps and P r are equal to 100. Such a case occurs when all the occupants are partially satisfied and the calculated cumulative dissatisfaction is reduced by 100% to a value of zero, representing no dissatisfaction. This is the ideal case and which was not observed to occur for any of the cases. For such a case, when all occupants are satisfied and the entire room as a whole is also satisfied with the visual comfort conditions, the C SC value is maximum and is equal to 100. Therefore, a higher CSC value, represents a better configuration, with 0 and 100 being the worst and best scores respectively. For a given case, it is now known that the P ps value represents the percentage of occupants whose satisfaction has increased from the baseline. It is seen later in the results that it is not common to get a P ps of 100%, which means that it is not possible to make everyone more satisfied and that one or more occupants might be dissatisfied in the process. It must be noted that a certain P ps value greater than zero might also mean that one or more persons might have had their preferences completely satisfied, but then again it is seen that not everybody’s preferences are met simultaneously. An ideal case would therefore be for a P ps value of 100% and a P r value of 100%. This would mean that all the occupants have had their preferences met on basis of their assigned importance factors. Another case might be for a P ps value of zero. This means that none of the occupants’ preferences have been met. Occupant Preference Cases The proposed research methodology was virtually tested under certain assumed possibilities of occupant preferences for glare, illuminance, and contrast ratio, and their corresponding importance factors. Several user preference models were created virtually to represent real-life preference and importance factor conditions for all the 16 occupants. No occupant surveys were done. These preferences and importance factors were constructed by virtually devising a set of ten worst-case and randomly selected scenarios which the framework might be subject to it in a real-life working situation. The framework was proposed to be tested for 3 days of the year and three times during each of these days. As explained previously, the three dates (June 21, December 22, and September 22) were chosen to be closely representative of three different annual solar altitude conditions. The three times of each day were chosen as 9am, noon, and 3pm. These times were chosen to represent three different evenly-spaced solar positions during the working hours of the day. The time 3pm 46 was chosen instead of a later time as the sun is already quite low at 3pm during the winter and because it was decided to maintain the same time for the three dates to be able to compare them. These are tabulated (Table 6). Table 6: Simulation Points-in-Time and Occupant Preference Cases Day Time Cases Time Cases Time Cases Total Cases Sims. / Time 1 Sims. / Time 2 No. of Configs. 3 June 21 9am 10 noon 10 3pm 10 30 2304 2304 2304 September 22 9am 10 noon 10 3pm 10 30 2304 2304 2304 December 22 9am 10 noon 10 3pm 10 30 2304 2304 2304 Table 7 summarizes the proposed metrics and measurement variables for ready-reference in studying the results in this chapter. Please note that the notations and abbreviations in this table shall be used throughout this chapter as well as in the remaining chapters and shall hold the same meaning unless stated otherwise. Table 7: Proposed Metrics, Variables, and Abbreviations Metrics & Variables Notation Units Range Criteria Cumulative Dissatisfaction D CL unitless >=0 minimum Percentage Reduction in D CL P r % 100% Mean Positive Reduction P r(mean) % Median Positive Reduction P r(med) % Percentage Partially Satisfied P ps % 0<=P ps <=10 0 Percentage Dissatisfied P ds % Average Cumulative Dissatisfaction D avg unitless Baseline Cumulative Dissatisfaction D base Configuration Index C index unitless Mean Cumulative Dissatisfaction D CLmean unitless Median Cumulative Dissatisfaction D CLmed unitless Mode Cumulative Dissatisfaction D CLmode unitless Expected Value of a function or variable E(f(x)) Glare (occupant ‘i’) G i unitless >=0.2 Weighted Glare Dissatisfaction (occupant ‘i’) D g(i) unitless >=0 Illuminance (occupant ‘i’) E i lux Weighted Illuminance Dissatisfaction (occupant ‘i’) D e(i) unitless >=0 Contrast Ratio (occupant ‘i’) C i unitless Weighted Contrast Ratio Dissatisfaction (occupant ‘i’) D cr(i) unitless >=0 Individual Dissatisfaction D i unitless >=0 Baseline Individual Dissatisfaction D i(base) Percentage Reduction in D i p r(i) % Minimum Individual Dissatisfaction (occ. ‘i’) D min(i) unitless Maximum Individual Dissatisfaction (occ. ‘i’) D max(i) unitless Median Individual Dissatisfaction (occ. ‘i’) D med(i) unitless Number of Better Performing Configurations N better Number of Under-Performing Configurations N worse Percentage Better Configurations P config % Configuration Performance Score C sc Excel Facilitator Tool Before delving into the structure and details for the sub-cases, it is important to highlight how the results were being documented. Figure 34 below shows a partial view of how the primary data was collected and used to calculate the other values. 47 Figure 34: Excel Facilitator Tool: Data Collection Sample The Grasshopper script generated the raw and preliminary calculation data, which were the individual workstation illuminance, glare, and contrast ratio values. These were generated both for the baseline case as well as for the other 2304 configurations. This set of data was simultaneously written to an Excel file that was automated to calculate the cumulative dissatisfaction and other related values based on a toggle for selecting one of the ten-different occupant preference cases. One Excel workbook was maintained for each point in time. Each workbook comprised of ten worksheets highlighting the ten different occupant preference models, a primary worksheet that documented and calculated all the results, an ‘Analysis’ worksheet which essentially documented, organized, and tabulated the required data in a presentable format, and a ‘Charts’ worksheet which was used to automatically generate useful charts and graphics based on the data in the ‘Analysis’ worksheet. In addition to these, there were several other worksheets which were not primarily important to discuss here but which helped bridge the gap between the primary worksheets in producing the final results and graphics. For the final results, most of calculations were automated using Excel by incorporating preset formulae that brought together data from several sheets. The automation using Excel was primarily a logistic decision but also a means to store large amounts of data in a more organized format for further analysis. The Excel tool served as an independent tool and a primary facilitator capable of generating all the results discussed in this chapter by simply receiving a set of simulation values. The following (Table 8) is a list of all the data headers under which data was recorded, stored, or calculated in the primary Excel worksheet. Table 8: Summary of Proposed Metrics, Variables, and Abbreviations Primary Excel Worksheet Column Headers 1 Simulation Number 9 Maximum Individual Dissatisfaction (D max ) 2 Control Variable States- A SE , A SW , L SE , L SW , VT SE , and VT SW . 10 Average Individual Dissatisfaction (D avg ) 3 Cumulative Dissatisfaction 11 Glare (G i ) 4 Percentage Reduction in D CL (P r ) 12 Weighted Glare Dissatisfaction (D g(i) ) 5 Percentage Partially Satisfied (P ps ) 13 Illuminance (E i ) 6 Average Cumulative Dissatisfaction (D avg ) 14 Weighted Illuminance Dissatisfaction (D e(i)) 7 Individual Dissatisfaction (Di, ‘i’ in 1 to 16) 15 Contrast Ratio (C i ) 8 Minimum Individual Dissatisfaction (D min ) 16 Weighted Contrast Ratio Dissatisfaction D cr(i) For each of these nine points in time of the year, the proposed framework was then tested under the following ten different cases of occupant preferences and importance factors, as discussed later in this chapter. The results from these cases were compared with each other and cross-compared to corresponding cases for the other points in time. 3.7.9 Data Analysis This module is used to organize and interpret the resultant data from all the modules in Excel and Grasshopper and compare it with the baseline case. For each set of 2304 data sets, a baseline case was chosen, which is the case for which the façade control states are closest to zero or minimum, and for which the electrochromic glazing visible transmittance is 0.6. This case was chosen to represent a static, unresponsive façade with minimum shading. The occupant preference deviation values corresponding to this baseline case are then compared against the other 3599 cases to determine which façade configurations produce lower deviations from the set preference. Even if one 48 configuration allows a lower deviation, it would mean that if the façade was configured to that particular state, then the occupants’ preferences would be most closely met. 3.7.10 Visualization The output of the ‘Data Analysis’ module for each simulation run is visualized using an automated technique designed in Grasshopper (Fig. 35 and Fig. 36). The visualization displays the distribution of light levels across the floor surface, the exact façade configuration under consideration, the façade control states (denoted by numerical tags representing the control state values), the horizontal work plane illuminance, glare, and contrast ratio measurements, and the individual deviation values calculated for each occupant. Figure 35: Automated Visualization; Fig. 36: A close-up view of the tag-based visualization Figure 37: Annotated diagram explaining the visualization components 49 3.8 Assumptions The following is a list of assumptions on which the thesis methodology is based. These are: preset discrete occupant preference ranges, flat importance factors, linear deviation and dissatisfaction criteria, cumulative deviation, and occupant behavior. The implications these are more discussed in more detail in chapter five. Preset Discrete Occupant Preference Ranges For each visual comfort factor, the choice of preference ranges available to the occupant are preset. For example, the glare preference ranges are fixed at discrete intervals providing ranges of 0.20-0.25,0.25-0.30,0.30-0.35,0.35-0.40, and 0.40-0.45. The occupant must choose from amongst these five options only, with 0.2-0.25 being the least perceptible glare to 0.40-0.45 as fringing on glare intolerance. A similar sub-divisional approach is applied to both illuminance as well as luminance contrast ratio. In However, real measured preferences were not used although in a real-world, occupants would choose their preferences from the predefined list. Rather, the preference sets for all the sixteen occupants were modeled virtually using a Python script to selecting preferences in a normal distribution from the pre-defined lists. This generated several sets of preferences, from which only one was chosen. Seed values provided along with the random function in Python was used as a key or seed value to regenerate or reproduce the exact random normal distribution whenever needed. A seed value of 5500 was used to generate one set of occupant preferences and importance factors (Fig. 38). Figure 38: Randomly generated occupant preference model case (seed value=5500) A total of ten occupant preference cases were modeled and generated, which are described in further detail in Chapter 4. The other nine occupant preference model cases were not generated randomly but specifically designed so that the results from those cases could be used to answer specific questions and to develop further hypotheses. Implications of this assumption is discussed in Chapter 5. Flat Importance Factors To mathematically depict the importance of the any visual comfort factor for an occupant, importance factors were assigned on a scale of a 5-90, such that all the importance factors for an individual would add up to 100. As discussed previously, in a real-world scenario, the importance factors would be input by the user directly using a graphical control interface. The 5-90 range was chosen to provide occupants the highest possible flexibility of being assigned very low or very high importance factors by the random function in a manner similar to the assigning or selection of occupant preferences, as described in the previous section. A total of 10 occupant preference models were generated, one of which was generated using the random-selection technique, and the other nine were assigned, as shown in Chapter 4, Section 4.1. Linear Deviation and Dissatisfaction Criteria The deviation values calculated for each visual comfort factor for a given occupant were consolidated to represent a net deviation value. However, each of the deviation values for glare, illuminance, and contrast ratio have different units and therefore need to be remapped to make the net deviation value more meaningful. For example, for an occupant the glare deviation might be 0.10 whereas the illuminance deviation might be 400 lx. The two deviations 50 represent two different units, the glare deviation being unitless whereas the illuminance deviation is measured in lux. Assuming importance factors of 20 and 50 for these two factors, it is seen that the resulting weighted deviations are farther apart from each other and so is the net deviation value. To normalize the deviation values, each value is divided by a constant factor, assumed to be the range subdivision length. Therefore, the deviations for glare, illuminance, and contrast ratio are divided by their range subdivision value, which are 0.05, 0.33, and 150 lx respectively. This provides normalized values, but which are still relatively large. Therefore, square roots of each of these three values are used to represent singular deviation values, and which when added forms the net deviation value that represents the occupant’s visual dissatisfaction, zero being the least dissatisfied and the best possible case. Implications of this assumption are discussed in Chapter 5. Cumulative Deviation The numerical deviation of each measured comfort factor from its preferred range is therefore not by itself proportionally representative of occupant dissatisfaction. However, as explained, a linear proportional relationship of absolute deviation to dissatisfaction has been assumed. This provides us with three independent sets of absolute deviations for glare, illuminance, and contrast ratio. Each of these contribute to the cumulative deviation that causes an overall dissatisfaction that the occupant feels. It is assumed that the cumulative deviation is calculated using an addition of the individual deviations, each weighted by its importance factor. Occupant Behavior Occupant behavior is stochastic in nature. This means that any decision which is taken based on occupant behavior has a certain degree of unpredictability. In this research, the scope of applicability of occupant behavior has been limited to a simplified occupant preference model for visual comfort only. 3.9 Scope Based on the discussed assumptions, it was possible to define a research scope. The research started off with a narrower scope which was widened with discovery. The following are within the scope of this research: the daylighting system design, adaptive façade control logic design, resolution of the multiple occupancy condition, the development of a consolidated visual comfort metric denoting overall dissatisfaction, clustering of façade configurations as design options, and the proposition and implementation of a methodology to decrease the cumulative visual discomfort. Daylighting System Design The conceptual design of the daylighting system for the two adjoining facades. This includes the decision-making process for selection of the daylighting system type, the definition of the control variables, control states, and the type of adaptations and kinetic movements that the system incorporates. Adaptive Façade Control Logic Design The control framework or the complete control logic that would drive the selected adaptive façade in a real setting using real data is based on strategies and techniques developed by analyzing the research results, as discussed in Chapter 5. In a real-world operation of the adaptive façade, the control logic proposes how to integrate these control strategies with real-time visual comfort and daylighting measurements, historic and machine-learnt data. Demonstration of the complete control logic is not detailed out. Resolving the multiple occupancy condition The proposed control logic and strategies are developed for a multiple occupancy condition. The goal of application of the control logic is to propose adaptive façade control strategies for the resolution of conflicts arising from differences in real-time visual comfort requirements of multiple occupants in a space like an open-plan office. Chapter 5 discusses the complete breadth and implications of all assumptions and possible alterations and alternatives to the control logic and strategies when resolving this problem for other conditions, such as fewer occupants. Development of a Consolidated Visual Comfort Metric Glare, illuminance, and contrast ratio at the workstation are treated independently as they measure different quantities. At any given time, the visual comfort perceived by an occupant is the cumulative effect of these measurements. However, a study of existing visual comfort metrics revealed that there does not exist any metric which consolidates the effects of these independently daylighting measurements in relation to what the occupant requires at any given point in time. A visual discomfort metric has been proposed. 51 Design Option Clustering Based on their performance, façade configurations were clustered to form groups of design alternatives. This clustering study was done using two different performance indicators representing two different aspects of design selection. Cumulative Visual Discomfort The proposed control logic is aimed to reduce the cumulative visual discomfort but its recommendations are not addressed towards reducing all instances of individual visual discomfort. This is discussed in Chapter 5. The following are not within the scope: • Thermal, energy, and outdoor view calculations and considerations • In a real-world application of the proposed control logic, a user control interface would be required to record user preferences and importance factors which would be input into the façade controller. • Precise and statistically accurate user-modeling. • A real study of occupant preference using a questionnaire and a post-occupancy evaluation was not done. Occupant preferences and personalized importance factors have been modeled digitally using a simplified approach. 3.10 Methodology Steps Below is a list of the important steps in the methodology. 1. In the first step, visual comfort data was either simulated for each occupant workstation location for a particular point in time. Historic data can also be used. 2. Input of occupant preferences and importance factors, ideally through a user-interface or based on historic preference data sets using machine learning techniques. The design of the user interface was outside the research scope. However, in a real-life situation, a user-interface would be required for the occupant to select their preferences and importance factors from predefined lists. 3. Based on the simulated or measured data, and the preferred data, scaled numerical deviations multiplied by the respective occupant-set importance factors were calculated. The deviations were scaled to make the numbers smaller and more manageable. Also, deviations of the different quantities featuring different units were normalized by division. 4. The calculated scaled and normalized deviation values were consolidated for glare, illuminance, and contrast ratio, and for each occupant. 5. A sum of these individual deviation values was calculated to represent individual dissatisfaction values. A higher deviation would imply higher dissatisfaction. 6. A sum of the individual dissatisfaction was calculated to form the cumulative dissatisfaction which represents the dissatisfaction of the entire room with the respective visual comfort conditions. 7. Steps 1 to 6 were also implemented for a chosen baseline case and baseline individual and cumulative dissatisfaction values were calculated for that point in time under consideration. 8. All the simulated values were written to Excel which was used to implement steps 9 to 11 in real-time. 9. The Percentage Reduction in Cumulative Dissatisfaction (Pr) from the baseline case was calculated. 10. Using the baseline and test case values for individual dissatisfaction, the percentage of occupants who are partially happy or satisfied relative to the baseline is also calculated and represented by Percentage Partially Satisfied (P ps ). 11. For each of the 2304 possible configurations, the P r and P ps values, and the Cumulative Dissatisfaction values, were calculated and were used to denote façade performance. These values were then plotted to recognize clusters of closely-performing configurations and also to rank the configurations and control states. 12. Automated visualization of these data using Grasshopper as explained in Section 3.7.10. 3.11 Summary Chapter 3 discussed the methodology adopted and the adopted daylighting system and its design. It also laid out the control logic that will be used to control the two adjacent facades. A metric for measuring cumulative visual discomfort resulting from glare, illuminance, and contrast ratio, based on occupants’ personal preferences and importance factors, was discussed and adopted for quantifying the results. Based on the value of this metric, a method was adopted to rank façade configurations. The ranking helped to recognize clusters among façade control states, thus allowing multiple design options to provide the same amount of visual comfort. These aspects were used to quantify results in the next chapter and their implications, application, and areas of improvement are discussed in greater detail in Chapter 5. 52 CHAPTER 4 4 RESULTS The following sections discuss the results and findings based on the methodology proposed in Chapter 3. Section 4.1 describes the ten occupant preference cases that were selected for study, describing the reasons for selection of each. Section 4.2 describes the chapter framework and organization of results, whereas Section 4.3 to 4.11 describe the results for the nine points in time under study. Section 4.12 concludes and summarizes the findings. 4.1 Occupant Preference Cases The following is a list of the ten cases under which the proposed framework was tested for each point in time. 4.1.1 Case 1: Equal Preferences and Importance Factors Case 1 (Fig. 39) considers occupants to have equal preference ranges (shown in green or uncolored) and importance factors (represented in blue). A 0.20 DGP value (Rg L= Rg H ) was chosen as it is the most desirable value for glare. A discrete contrast ratio value of 1.0 (R cr ) and a single illuminance value (R e =500lx) were chosen keeping in mind that standards such as the IES suggest a 500lx value for occupant task surfaces and a contrast ratio equal to one (DiLaura et al. 2011). Equal importance factors (If g , If e , and If cr ) of 33-1/3 were chosen for each preference, totaling 100. Figure 39: Occupant Preference and Importance Factor (Case 1) 4.1.2 Case 2: Equal Preferences, High Importance Factor for Glare The second case (Fig. 40) was derived from ‘Case 1’, with similar equal preference ranges but varying importance factors for the three visual comfort parameters. To see the effect of change in importance factors for all the three visual Figure 40: Equal Preferences, High Glare Importance Factor (Case 2) 53 comfort parameters in relation to ‘Case 1’, ‘Case 2’, ‘Case 3’, and ‘Case 4’ were designed, each featuring one visual comfort parameter with the highest importance factor possible (IF=90), and two other parameters with the lowest importance factor possible (IF=5). Accordingly, ‘Case 2’ features a high glare importance factor (If g =90) and low importance factors for illuminance and contrast ratio (If e =5 and If cr =5), totaling 100. The intent of this case is to understand and analyze how the overall results change based on a high glare importance factor. 4.1.3 Case 3a: Equal Preferences, High Importance Factor for Illuminance (500lx) As explained in section 4.1.2, ‘Case 3a’ (Fig. 41) was derived from ‘Case 1’, with similar equal preference ranges but varying importance factors for the three comfort parameters. To see the effect of change in importance factors for all the three visual comfort parameters in relation to ‘Case 1’, ‘Case 2’, ‘Case 3’, and ‘Case 4’ were designed, each featuring one visual comfort parameter with the highest importance factor possible (IF=90), and two other parameters with the lowest importance factor possible (IF=5). Accordingly, ‘Case 3’ features a high illuminance importance factor (If e =90) and low importance factors for glare and contrast ratio (If g =5 and If cr =5), totaling 100. ‘Case 3’ is divided into 3 sub-cases: ‘Case 3a’, ‘Case 3b’, and ‘Case 3c’. ‘Case 3a’ displays a discrete and fixed illuminance value of 500lx (Fig. 41) for reasons as stated in section 4.1.1. The intent of the division of ‘Case 3’ into 3 sub-cases is to cross- compare them amongst each other to draw conclusions about the overall results based only on changing illuminance preference values. Accordingly, the three cases, in order, assume a standard recommended illuminance preference values of 500lx, a maximum illuminance value of 2000lx, and a random preference case (Case 3c). Figure 41: Equal Preferences, High Importance Factor for Illuminance (500lx, Case 3a) 4.1.4 Case 3b: Equal Preferences, High Importance Factor for Illuminance (2000lx) The fourth case (Fig. 42) was derived from ‘Case 3a’, with a different set of equal illuminance preference ranges Figure 42: Equal Preferences, High Importance Factor for Illuminance (2000lx, Case 3b) 54 but similar importance factors for the three comfort parameters. As explained in Section 4.1.3, to see the effect of change in illuminance preference range, ‘Case 3b’ was designed based off ‘Case 3a’. ‘Case 3b’ features a discrete illuminance preference (Re L =Re H =2000lx) but is otherwise similar to ‘Case 3a’. 4.1.5 Case 3c: Equal Preferences, High Illuminance Importance Factor (unequal ranges) The fifth case (Fig. 43) was derived from ‘Case 3a’, with a different set of equal illuminance preference ranges but similar importance factors for the three comfort parameters. As explained in Section 4.1.3, to see the effect of change in illuminance preference range, ‘Case 3c’ was designed based off ‘Case 3a’. ‘Case 3c’ features randomly generated (seed value=5500) illuminance preference ranges, instead of discrete values, but is otherwise similar to ‘Case 3a’. Figure 43: Equal Preferences, High Importance Factor for Illuminance (randomly generated, Case 5) 4.1.6 Case 4a: Equal Preferences, High Importance Factor for Contrast Ratio (1.0) Just as ‘Case 3’ was sub-divided into 3 sub-cases based on differing illuminance preference setpoints or ranges, ‘Case 4’ was also sub-divided into 3 cases based on differing contrast ratio preference setpoints and ranges. All the three cases are based off ‘Case2’ but with a high importance factor for contrast ratio (IF cr =90) and a low importance factor for glare and illuminance (If e =5 and IF g =5). ‘Case 4a’ assumes a contrast ratio of 1.0 (Fig. 44). The intent of this case is to analyze how the results change based on a high contrast ratio importance factor and a R cr setpoint of 1. Figure 44: Equal Preferences, High Importance Factor for Contrast Ratio (R cr =1.0, Case 4a) 4.1.7 Case 4b: Equal Preferences, High Importance Factor for Contrast Ratio (3.0) As explained in Section 4.1.6, ‘Case 4’ was sub-divided into 3 cases based on differing contrast ratio preference setpoints and ranges. All the three cases are based off ‘Case2’ but with a high importance factor for contrast ratio (IF cr =90) and a low importance factor for glare and illuminance (If e =5 and IF g =5). ‘Case 4b’ assumes a contrast ratio 55 of 3.0 (Fig. 45). The intent of this case is to analyze how the results change based on a high contrast ratio importance factor and a R cr setpoint of 3. Figure 45: Equal Preferences, High Importance Factor for Contrast Ratio (R cr =3.0, Case 4b) 4.1.8 Case 4c: Equal Preferences, High Importance Factor for Contrast Ratio (unequal ranges) As explained in Section 4.1.6, ‘Case 4’ was sub-divided into 3 cases based on differing contrast ratio preference setpoints and ranges. All the three cases are based off ‘Case2’ but with a high importance factor for contrast ratio (IF cr =90) and a low importance factor for glare and illuminance (If e =5 and IF g =5). ‘Case 4b’ assumes a randomly generated set of contrast ratio ranges (Fig. 46). The intent of this case is to analyze how the results change based on a high contrast ratio importance factor and unequal contrast ratio ranges. Figure 46: Equal Preferences, High Importance Factor for Contrast Ratio (random ranges, Case 4c) 4.1.9 Case 5: Randomly Generated Occupant Preferences and Importance Factors (seed=5500) ‘Case 5’ (Fig. 47) was modeled to represent a case when all the importance factors and visual comfort preference ranges are different, as for a real occupant, with randomly generated preferences and importance factors. A random function in Python was used to randomly select the preference ranges and importance factors from predefined range options. The preference for glare is maintained at 0.20 as it is the most desirable glare value as measured by the DGP metric. The illuminance and contrast ratio values are randomly selected from their bounding ranges of 300lx-1950lx, and 0.33-3.0. Importance factors for all three visual comfort factors are also randomly selected and are different and randomly distributed. However, each set of importance factors for glare, illuminance, and contrast ratio totals 100. 56 The random selection is assumed to be a more accurate representation of the occupant preferences and importance factors than ‘Case 1’, which is an ideal case. The reason for selection of this case was to see the effect of random preferences and importance factors on the overall results and configuration performances. This was also cross- compared with the other cases, ‘Case 1’ to ‘Case 6.’ Figure 47: Randomly generated occupant preferences and importance factors (seed=5500) 4.1.10 Case 6: Assumed Worst Case Scenario This case (Fig. 48) was designed with the intention of creating a worst-case scenario by requiring peripherally situated occupants to prefer lower illuminance values, while setting a higher illuminance preference for non-peripheral occupant positions. The contrast ratio for all occupants were kept the same in this case and a minimum importance factor of 33-1/3 was assigned to it. Also, the glare preference was chosen as 0.20, that being the most ideal and preferred value. An importance factor of 33-1/3 was assigned to it. The importance factor assigned to illuminance was also 33-1/3. The occupants were split into three groups based on their distance from the façade. Occupants represented by indices 1,2,3,4,8, and 9, formed the group nearest to the façade. For this group, the lowest acceptable discrete illuminance of 300lx was considered, as a low illuminance is usually difficult to achieve near the façade. A opposite strategy was chosen for occupants furthest away from the façade, with a discrete preferred illuminance of 2000lx. For occupants in between, represented by indices 5,6,7,10, and 15, a discrete preferred illuminance of 800lx was chosen. Figure 48: Assumed Worst Case Scenario (Case 6) 4.2 Result Framework This section is intended to provide an understanding of the framework for analysis of the results. Chapter 4 has subsequently been divided into nine primary result sections. These highlight the results and analysis for the nine points in time under study. Each of these nine sections is structured into a summary section followed by ten sub-sections, 57 each of which pertains to each of the ten occupant preference cases under study, as explained in Section 4.1. Each of the nine sections starts off with a description of the point in time under study, for example “June 21, 3pm,” tables summarizing the primary results and observations for all the ten occupant preference cases, followed by a tabulated representation of the cluster patterns observed and other supporting results. At the end of each section are ten sub- sections that highlight points of interest or observations, accompanied by scatterplots, showing the distribution of cumulative dissatisfaction and performance scores with configuration indices, for some of the cases. The results in this chapter follow are organized hierarchically (Fig. 49). Figure 49: Result Framework and Organization Maximum Percentage Reduction A tabular format was used to summarize the results for cases: 1, 2, 3a, 3b, 3c, 4a, 4b, 4c, 5, and 6 (Table 9). Table 9: Typical Format of Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 37.90 20.92 26.81 47.81 28.88 43.90 40.52 41.69 38.71 46.73 Max. Cumulative Dissatisfaction 157.3 203.7 73.87 97.33 84.77 176.1 172.5 174.4 155.4 177.4 Min. Cumulative Dissatisfaction 36.75 20.10 25.70 40.46 27.71 39.81 38.54 37.47 36.90 41.91 Median Cumulative Dissatisfaction 66 30 40 65 50 89 85 87 71 90 Mode Cumulative Dissatisfaction 41.00 24.00 58.00 67.00 67.00 42.00 42.00 40.00 113.0 44.00 Average Cumulative Dissatisfaction 70.44 32.44 42.73 64.30 50.38 95.13 92.05 92.98 74.48 96.79 Maximum Percentage Reduction 3.03 3.90 4.14 15.37 4.06 9.32 4.88 10.14 4.66 10.32 Minimum Percentage Reduction 0.03 0.11 0.03 0.11 0.10 0.03 0.03 0.01 0.18 0.07 Mean Positive Reduction 1.31 1.96 2.02 6.19 1.79 4.31 2.44 4.46 2.07 4.96 Median Positive Reduction 1.24 1.86 2.16 5.37 1.37 4.21 2.70 4.37 1.82 5.08 Max. Percentage Partially Satisfied 81.25 68.75 75.00 100.0 81.25 100.0 75.00 93.75 75.00 100.0 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 48 71 27 322 210 308 87 279 123 318 No. 100% Occ. Partially Satisfied 0 0 0 84 0 22 0 0 0 33 No. of Cases Occ. Dissatisfied 844 575 580 1420 1732 844 575 575 605 844 Minimum Individual Dissatisfaction 0.56 0.35 0.24 0.32 0.19 0.46 0.27 0.18 0.24 0.47 Maximum Individual Dissatisfaction 41.94 65.37 16.62 18.62 18.26 22.08 22.00 22.10 44.00 22.00 Median Individual Dissatisfaction 3.63 1.75 2.42 3.82 3.10 4.69 4.26 4.40 3.94 4.93 Average Individual Dissatisfaction 4.40 2.03 2.67 4.02 3.15 5.95 5.75 5.81 4.66 6.05 Better Performing Configurations 42 30 8 319 17 357 131 342 97 369 Under-Performing Configurations 2262 2274 2296 1985 2287 1947 2173 1962 2207 1935 Percentage Better Configurations 1.82 1.30 0.35 13.85 0.74 15.49 5.69 14.84 4.21 16.02 Configuration Index (best) 277 1429 277 2161 273 41 185 597 41 41 Configuration Performance Score 13.77 14.92 16.87 37.96 18.16 28.77 16.59 26.71 17.91 30.58 Percentage Partially Satisfied (best) 62.50 62.50 68.75 93.75 81.25 93.75 68.75 75.00 68.75 93.75 Percentage Reduction in DCL 3.03 3.56 4.14 15.37 4.06 8.83 4.00 9.51 4.66 9.97 Least Effective Configuration Index 89 1421 265 2085 233 1357 2165 941 209 1921 Performance Score 1.02 2.58 1.37 2.41 2.56 1.11 1.19 0.67 3.20 1.84 Percentage Partially Satisfied 37.50 62.50 56.25 25 68.75 37.50 50.00 37.50 56.25 50.00 Percentage Reduction in DCL 0.03 0.11 0.03 0.23 0.10 0.03 0.03 0.01 0.18 0.07 58 Besides other values, Table 4.2.1 also lists the Baseline Cumulative Dissatisfaction, Maximum Percentage Reduction, Maximum Percentage Partially Satisfied, No. of Configurations for which 100% Occupants are partially satisfied, Percentage Better Configurations, the best Configuration Index and its Configuration Performance Score, as well as the Least Effective Configuration Index. Using the table, comparisons are drawn in between these values calculated for the different cases, ‘Case 1’ to ‘Case 6’. The definition, scope, and range of these variables have been explained previously in Chapter 3, Section 3.7.8. Distribution of Percentage Reduction The next table, displaying a typical format (Table 10), is used for each point in time and describes the distribution of the number of configurations by the percentage reduction in cumulative dissatisfaction that they produce. Table 10: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 42 30 8 145 17 216 131 202 97 181 5-10% 0 0 0 113 0 141 0 139 0 182 10-15% 0 0 0 58 0 0 0 1 0 6 15-20% 0 0 0 3 0 0 0 0 0 0 20-25% 0 0 0 0 0 0 0 0 0 0 25-30% 0 0 0 0 0 0 0 0 0 0 30-35% 0 0 0 0 0 0 0 0 0 0 35-40% 0 0 0 0 0 0 0 0 0 0 40-45% 0 0 0 0 0 0 0 0 0 0 45-50% 0 0 0 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 42 30 8 319 17 357 131 342 97 369 Figure 4.2.2: Exemplary format for categorization of number of better-performing configurations into 5% reduction bands The percentage reduction in cumulative dissatisfaction is divided into twenty, five percent range bands. To know the top performing configurations, the top 5 configurations, ranked based on their performance score, as explained previously, are summarized and listed for each of the 10 cases. Ranked Configurations The top 5 configurations for each case were tabulated in a typical format (Table: 11). This was done to compare and analyze the configurations to make conclusions regarding the control states and the resultant performance of the configurations. For ease of representation, a concatenated format is used to represent the control states. Table 11: Typical Format of Top Five Ranked Configurations by Case 1 0,-30,84,84,0.01,0.6 0,-30,84,84,0.18,0.6 0,-30,84,60,0.18,0.6 0,-30,84,60,0.06,0.6 0,0,84,84,0.01,0.6 2 0,-30,84,84,0.01,0.6 0,-30,84,84,0.18,0.6 0,-60,36,84,0.01,0.6 0,-60,36,84,0.06,0.6 0,-60,36,84,0.6,0.6 3 0,0,36,84,0.06,0.6 0,0,36,84,0.01,0.6 0,0,36,84,0.18,0.6 0,0,36,84,0.6,0.6 0,-30,36,60,0.01,0.6 4 -60,-90,36,36,0.01,0.6 -90,-90,36,36,0.18,0.6 -90,-90,36,36,0.01,0.6 -90,-90,36,36,0.6,0.6 -90,-90,36,36,0.06,0.6 5 0,0,36,60,0.18,0.6 0,0,36,84,0.06,0.6 0,0,36,60,0.06,0.6 0,0,36,84,0.6,0.6 0,0,36,84,0.01,0.6 6 0,0,36,84,0.6,0.6 0,0,36,84,0.06,0.6 0,0,36,60,0.18,0.6 0,0,36,84,0.18,0.6 0,0,36,60,0.06,0.6 7 0,0,36,60,0.18,0.6 0,0,36,60,0.06,0.6 0,0,36,84,0.06,0.6 0,0,36,84,0.6,0.6 0,-30,36,36,0.18,0.6 8 0,0,36,60,0.18,0.6 0,0,36,84,0.6,0.6 0,0,36,60,0.01,0.6 0,-60,36,36,0.01,0.6 0,0,36,60,0.06,0.6 9 0,0,84,84,0.01,0.6 0,-30,84,84,0.6,0.6 0,-30,84,84,0.01,0.6 0,0,36,84,0.6,0.6 0,-30,60,84,0.6,0.6 10 0,0,36,60,0.18,0.6 0,0,36,60,0.01,0.6 0,0,36,84,0.06,0.6 0,0,36,60,0.06,0.6 0,0,36,84,0.6,0.6 59 In the format above, the six numbers, separated by commas, represent the six control states in this order: A SE , A SW , L SE , L SW , VT SE , and VT SW , where ‘A’ represents control angle, ‘L’ is the control length, ‘VT’ is the visible light transmittance, and subscripts ‘SE’ and ‘SW’ are used to represent the south-east and south-west facades. Configuration Cluster Patterns Two kinds of performance metrics values were used to evaluate the performance of the configurations (see Chapter 3, Section 3.7.8). The percentage reduction in cumulative dissatisfaction (P r ) and the performance score values for each configuration were plotted in the same sequence as the configurations. A higher percentage reduction in cumulative dissatisfaction and a higher performance score value both imply higher performance of the configuration. However, the performance measures implied by the two metrics are different. A better performing configuration represented by a higher percentage reduction in cumulative dissatisfaction value implies that it reduces the overall visual discomfort to a greater extent. However, as the metric is cumulative, it does not provide a complete understanding of how many occupants are dissatisfied and how many are satisfied, if the cumulative effect is positive. To overcome this limitation, the performance score (CS) metric was used (see Chapter 3, Section 3.7.8). The performance score (Equation 17) is the logarithm (base 10) of the ratio of the percentage partially satisfied (P ps ) to the percentage reduction in cumulative dissatisfaction (P r ). 𝑪 𝒔𝒄 = 𝟏𝟎 𝒙 𝒍𝒐𝒈 𝟏𝟎 (𝑷 𝒓 / 𝑷 𝒑𝒔 ) Equation 17: Façade Configuration Performance Score for Configuration Ranking and Clustering It can be seen from the equation (Equation 4.1) defining the performance score that as long as P ps >P r , the performance score is negative and that for P r >P ps , the performance score is positive. A negative performance score implies that more people are satisfied but each to a lesser degree. A positive value implies that less number of occupants are getting partially satisfied but to a higher degree of satisfaction. From a design standpoint, using a configuration having a negative performance score implies that more people can be made satisfied but each to a lesser degree. From a designer’s point of view, this would mean that lesser people would be very happy with their visual comfort condition. It was seen that for most of the results, this was not the case and that the performance score is usually a positive value. It also follows from the equation that if P r =P ps , the performance score would be zero. It was observed that the performance scores of two or more configurations are similar or closely spaced, forming clusters of solutions. As far as design options are considered, these configurations could be used interchangeably by the designer as similarly- performing options and could be shortlisted based on other factors such as cost, design preference, and technical constraints. As the configurations are listed in a sequence, it was possible to detect a pattern in the performance of the configurations by analyzing a plot of the performance values of the configurations against the configuration index. Following this pattern, it was possible to predict the performance of a certain configuration. These aspects of solution clusters, performance prediction, and the proposed performance metrics are discussed in more detail in the following chapter. Summary This section consolidates and compares the results from all the cases and points in time in a tabulated format. Nomenclature As explained in Chapter 3, certain nomenclatures (Table 12) have been adopted to explain the results. Table 12: Summary of Nomenclature Symbol Term Symbol Term D CL Cumulative Dissatisfaction CS Configuration Performance Score P r Percentage Reduction in D CL IF Importance Factor P ps Percentage Partially Satisfied C index Configuration Index D i Baseline Individual Dissatisfaction P config Percentage Better Configurations D base Baseline Cumulative Dissatisfaction D CLmean Mean Cumulative Dissatisfaction P r(mean) Mean Positive Reduction P r(med) Median Positive Reduction D i(base) Individual Dissatisfaction N better No. of Better Performing Configurations A SE , A SW , L SE , L SW , VT SE , and VT SW Control Variables (in order of listing) 60 4.3 June 21, 9am A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 13). The following values correspond only to calculations made for a single point-in-time, 9am on June 21. Table 13: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 37.90 20.92 26.81 47.81 28.88 43.90 40.52 41.69 38.71 46.73 Max. Cumulative Dissatisfaction 157.3 203.7 73.87 97.33 84.77 176.1 172.5 174.4 155.4 177.4 Min. Cumulative Dissatisfaction 36.75 20.10 25.70 40.46 27.71 39.81 38.54 37.47 36.90 41.91 Median Cumulative Dissatisfaction 66 30 40 65 50 89 85 87 71 90 Mode Cumulative Dissatisfaction 41.00 24.00 58.00 67.00 67.00 42.00 42.00 40.00 113.0 44.00 Average Cumulative Dissatisfaction 70.44 32.44 42.73 64.30 50.38 95.13 92.05 92.98 74.48 96.79 Maximum Percentage Reduction 3.03 3.90 4.14 15.37 4.06 9.32 4.88 10.14 4.66 10.32 Minimum Percentage Reduction 0.03 0.11 0.03 0.11 0.10 0.03 0.03 0.01 0.18 0.07 Mean Positive Reduction 1.31 1.96 2.02 6.19 1.79 4.31 2.44 4.46 2.07 4.96 Median Positive Reduction 1.24 1.86 2.16 5.37 1.37 4.21 2.70 4.37 1.82 5.08 Max. Percentage Partially Satisfied 81.25 68.75 75.00 100.0 81.25 100.0 75.00 93.75 75.00 100.0 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 48 71 27 322 210 308 87 279 123 318 No. 100% Occ. Partially Satisfied 0 0 0 84 0 22 0 0 0 33 No. of Cases Occ. Dissatisfied 844 575 580 1420 1732 844 575 575 605 844 Minimum Individual Dissatisfaction 0.56 0.35 0.24 0.32 0.19 0.46 0.27 0.18 0.24 0.47 Maximum Individual Dissatisfaction 41.94 65.37 16.62 18.62 18.26 22.08 22.00 22.10 44.00 22.00 Median Individual Dissatisfaction 3.63 1.75 2.42 3.82 3.10 4.69 4.26 4.40 3.94 4.93 Average Individual Dissatisfaction 4.40 2.03 2.67 4.02 3.15 5.95 5.75 5.81 4.66 6.05 Better Performing Configurations 42 30 8 319 17 357 131 342 97 369 Under-Performing Configurations 2262 2274 2296 1985 2287 1947 2173 1962 2207 1935 Percentage Better Configurations 1.82 1.30 0.35 13.85 0.74 15.49 5.69 14.84 4.21 16.02 Configuration Index (best) 89 1421 265 597 233 1629 2165 941 209 1921 Configuration Performance Score 31.31 27.69 32.24 28.37 28.58 30.96 32.50 34.95 24.91 28.69 Least Effective Configuration Index 273 1577 277 2221 285 5 169 5 45 1 Performance Score 12.97 11.59 12.20 7.28 12.65 9.41 10.62 8.31 11.34 8.96 Performance Range Ratio 2.41 2.39 2.64 3.90 2.26 3.29 3.06 4.21 2.20 3.20 Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 13) to be lowest for ‘Case 1’ and highest for ‘Case 3b’, varying in the range of 3.03% to 15.37%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 4.66%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially increased when all occupants preferred an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants increased significantly for variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) actually performed better at a P r of 10.32%. It was observed that a high P r value (Cases 3b, 4a, 4c, and 6) is associated with much higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3a’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-preferred occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 61 Figure 50: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 50 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 13). It is seen here that 75% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 14 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 14: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 42 30 8 145 17 216 131 202 97 181 5-10% 0 0 0 113 0 141 0 139 0 182 10-15% 0 0 0 58 0 0 0 1 0 6 15-20% 0 0 0 3 0 0 0 0 0 0 20-25% 0 0 0 0 0 0 0 0 0 0 25-30% 0 0 0 0 0 0 0 0 0 0 30-35% 0 0 0 0 0 0 0 0 0 0 35-40% 0 0 0 0 0 0 0 0 0 0 40-45% 0 0 0 0 0 0 0 0 0 0 45-50% 0 0 0 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 42 30 8 319 17 357 131 342 97 369 62 Ranked Configurations Table 15 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 15: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 0,0,60,84,0.06,0.6 0,-60,60,60,0.01,0.6 0,-30,60,60,0.01,0.6 0,-60,60,60,0.18,0.6 0,-30,60,60,0.18,0.6 2 -60,-30,84,60,0.01,0.6 -60,-60,84,60,0.01,0.6 -60,-60,84,60,0.6,0.6 -30,-60,60,84,0.18,0.6 -60,-60,84,60,0.18,0.6 3a 0,-30,84,60,0.06,0.6 0,0,84,84,0.01,0.6 0,-30,84,60,0.01,0.6 0,-30,84,60,0.18,0.6 0,-30,84,84,0.6,0.6 3b -30,0,36,60,0.18,0.6 -30,0,36,60,0.01,0.6 -30,-60,60,36,0.01,0.6 -30,-30,36,36,0.01,0.6 -30,-30,36,36,0.6,0.6 3c 0,-30,60,84,0.06,0.6 0,-30,60,84,0.01,0.6 -60,-90,60,60,0.18,0.6 0,0,84,60,0.06,0.6 -60,-90,60,60,0.01,0.6 4a -60,-90,36,84,0.01,0.6 -60,-30,60,36,0.01,0.6 -90,-30,60,36,0.01,0.6 -30,0,84,84,0.18,0.6 -90,-90,36,84,0.06,0.6 4b -90,-90,36,36,0.18,0.6 -30,-30,36,36,0.18,0.6 -30,0,60,36,0.01,0.6 0,-90,60,36,0.06,0.6 0,-90,36,36,0.6,0.6 4c -30,-60,60,60,0.01,0.6 -90,0,60,60,0.18,0.6 -90,-90,36,60,0.6,0.6 -90,-90,36,60,0.18,0.6 -30,-60,60,60,0.18,0.6 5 0,-30,60,60,0.6,0.6 -90,0,36,84,0.6,0.6 -90,0,36,84,0.01,0.6 -90,0,36,84,0.06,0.6 -90,0,36,84,0.18,0.6 6 -90,-30,60,36,0.6,0.6 -90,-30,60,36,0.06,0.6 -30,-30,60,84,0.6,0.6 -90,-30,60,36,0.01,0.6 -60,-90,36,84,0.18,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3b’, the Cumulative Dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 51 and Fig. 52). For ‘Case 2’, the P ps values were highest for C index =2,6,10…2+4n, and lowest for the C index =4,8,12…4+4n. From the D CL curve, it was observed that D CL is highest for C index =4,8, 12…4+4n, and lowest for C index =2,6,10…2+4n. Figure 51: Cluster patterns for Case 2 For ‘Case 3b’, the P ps values were highest for C index =1,5,9…,1+4n, and lowest for the C index =i+4n, where i= {2,3,4}. From the D CL curve, it was observed that D CL is highest for C index =4,8, 12…,4+4n, and lowest for C index =2,6,10,…,2+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 52: Cluster patterns for Case 3b 63 4.4 June 21, noon A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 16). The following values correspond only to calculations made for a single point-in-time, noon on June 21. Table 16: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 38.53 22.27 28.92 45.04 41.56 41.96 39.25 39.80 38.33 44.47 Max. Cumulative Dissatisfaction 118.7 50.97 62.67 85.69 179.2 173.3 169.6 171.5 124.0 174.6 Min. Cumulative Dissatisfaction 36.48 19.32 25.46 37.08 35.93 38.83 38.43 36.98 36.74 40.70 Median Cumulative Dissatisfaction 68 31 42 66 97 92 87 89 74 93 Mode Cumulative Dissatisfaction 77.00 35.00 62.00 70.00 111.0 105.0 40.00 101.0 82.00 106.0 Average Cumulative Dissatisfaction 71.38 32.95 42.97 64.37 100.6 96.66 92.63 94.29 75.46 98.28 Maximum Percentage Reduction 5.32 13.26 11.96 17.68 13.55 7.45 2.10 7.09 4.14 8.48 Minimum Percentage Reduction 0.02 0.03 0.16 0.02 0.01 0.02 0.02 0.01 0.03 0.02 Mean Positive Reduction 1.64 4.21 3.32 6.82 5.37 3.26 0.59 3.60 1.31 3.88 Median Positive Reduction 1.37 3.16 2.42 6.38 5.27 3.31 0.48 4.23 1.24 3.84 Max. Percentage Partially Satisfied 81.25 87.50 68.75 100.0 93.75 93.75 68.75 93.75 81.25 100.0 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 99 203 48 253 257 304 42 264 152 294 No. 100% Occ. Partially Satisfied 0 0 0 45 0 0 0 0 0 4 No. of Cases Occ. Dissatisfied 576 12 579 1785 1728 1152 576 576 576 1152 Minimum Individual Dissatisfaction 0.53 0.42 0.27 0.43 0.22 0.43 0.32 0.23 0.84 0.56 Maximum Individual Dissatisfaction 13.95 5.73 6.47 7.89 21.91 21.53 21.45 21.55 13.69 21.46 Median Individual Dissatisfaction 3.46 1.71 2.38 3.78 4.72 4.38 3.91 4.18 3.76 4.61 Average Individual Dissatisfaction 4.46 2.06 2.69 4.02 6.29 6.04 5.79 5.89 4.72 6.14 Better Performing Configurations 120 231 134 354 285 322 84 264 116 318 Under-Performing Configurations 2184 2073 2170 1950 2019 1982 2220 2040 2188 1986 Percentage Better Configurations 5.21 10.03 5.82 15.36 12.37 13.98 3.65 11.46 5.03 13.80 Configuration Index (best) 1481 1869 2170 2049 405 797 1177 1029 1317 1221 Configuration Performance Score 32.56 32.58 24.37 34.03 40.78 35.01 33.22 38.81 32.85 33.14 Least Effective Configuration Index 421 1005 417 2209 9 1 169 33 173 5 Performance Score 9.50 7.15 6.39 7.07 7.78 9.89 14.22 9.94 12.15 9.47 Performance Range Ratio 3.43 4.56 3.81 4.81 5.24 3.54 2.34 3.90 2.70 3.50 Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 16) to be lowest for ‘Case 4b’ and highest for ‘Case 3b’, varying in the range of 2.10% to 17.68%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 4.14%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially increased when all occupants preferred an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants increased significantly for variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) actually performed well at a P r of 8.48%. It was observed that a high P r value is associated with much higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. However, a low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3a’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-preferred occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 64 Figure 53: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 53 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 16). It is seen here that 81.25% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 17 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 17: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 117 134 104 122 137 251 84 177 116 209 5-10% 3 83 25 162 124 71 0 87 0 109 10-15% 0 14 5 57 24 0 0 0 0 0 15-20% 0 0 0 13 0 0 0 0 0 0 20-25% 0 0 0 0 0 0 0 0 0 0 25-30% 0 0 0 0 0 0 0 0 0 0 30-35% 0 0 0 0 0 0 0 0 0 0 35-40% 0 0 0 0 0 0 0 0 0 0 40-45% 0 0 0 0 0 0 0 0 0 0 45-50% 0 0 0 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 120 231 134 354 285 322 84 264 116 318 65 Ranked Configurations Table 18 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 18: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 -60,-60,36,84,0.06,0.6 -60,0,60,60,0.06,0.6 -60,-30,60,84,0.18,0.6 0,-30,84,36,0.01,0.6 -30,-30,84,60,0.18,0.6 2 -90,0,84,84,0.01,0.6 0,-30,36,84,0.01,0.18 -30,-30,60,36,0.01,0.6 -30,-90,84,84,0.01,0.6 -90,0,84,84,0.06,0.6 3a -90,-90,36,36,0.06,0.18 -60,-90,84,60,0.6,0.6 0,-90,84,60,0.06,0.6 -60,-90,84,60,0.01,0.6 -30,-30,84,36,0.18,0.6 3b -90,-60,36,84,0.6,0.6 -30,0,84,36,0.6,0.6 0,-90,84,60,0.01,0.6 0,-90,84,60,0.06,0.6 -30,0,84,36,0.06,0.6 3c 0,-60,84,60,0.18,0.6 -90,0,60,84,0.01,0.6 -30,0,60,84,0.6,0.6 -30,-30,60,36,0.06,0.6 -30,0,60,84,0.01,0.6 4a -30,-30,60,60,0.01,0.6 -30,-60,60,36,0.06,0.6 -60,-60,36,60,0.01,0.6 -30,0,84,60,0.01,0.6 -60,-60,36,60,0.18,0.6 4b -60,0,36,60,0.06,0.6 -60,-30,36,36,0.6,0.6 -90,-30,36,36,0.6,0.6 0,-90,60,36,0.18,0.6 0,-60,60,84,0.6,0.6 4c -30,-90,36,60,0.18,0.6 -90,-30,36,84,0.01,0.6 -60,-30,36,84,0.06,0.6 0,-90,84,60,0.01,0.6 0,-90,60,84,0.01,0.6 5 -60,-30,36,60,0.18,0.6 0,-30,60,36,0.18,0.6 -60,0,36,60,0.6,0.6 0,-60,36,36,0.01,0.6 0,-30,84,36,0.6,0.6 6 -60,0,60,60,0.18,0.6 -30,-30,60,60,0.6,0.6 -30,0,84,36,0.6,0.6 -60,0,60,60,0.01,0.6 -90,0,60,84,0.6,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3b’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 54 and Fig. 55). For ‘Case 2’, the P ps values were highest for C index =2,6,10…,2+4n, and lowest for the Cindex=4,8,12…4+4n. From the D CL curve, it was observed that D CL is highest for C index =4,8, 12…,4+4n, and lowest for C index =2,6,10,…,2+4n. Figure 54: Cluster patterns for Case 2 For ‘Case 3b’, the P ps values were seen to be highest for C index =1,5,9…,1+4n, and lowest for the C index =i+4n, where i={2,3,4}. From the D CL curve, it was observed that D CL is highest for C index =4,8,12…,4+4n, and lowest for C index =2,6,10,…,2+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 55: Cluster patterns for Case 3b 66 4.5 June 21, 3pm A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 19). The following values correspond only to calculations made for a single point-in-time, 3pm on June 21. Table 19: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 44.16 28.35 39.24 38.89 34.26 39.18 40.68 38.12 39.52 39.97 Max. Cumulative Dissatisfaction 174.2 185.8 83.78 107.1 92.81 151.5 157.6 157.2 169.0 152.7 Min. Cumulative Dissatisfaction 35.87 17.83 22.34 31.90 25.87 37.53 37.43 36.06 35.09 38.65 Median Cumulative Dissatisfaction 47 25 33 57 41 58 54 56 52 60 Mode Cumulative Dissatisfaction 48.00 21.00 32.00 57.00 42.00 39.00 40.00 37.00 53.00 39.00 Average Cumulative Dissatisfaction 53.83 26.08 35.78 55.05 42.13 68.30 64.83 65.96 56.17 69.73 Maximum Percentage Reduction 18.77 37.10 43.09 17.98 24.47 4.22 7.99 5.40 11.21 3.30 Minimum Percentage Reduction 0.04 0.01 0.02 0.07 0.03 0.005 0.004 0.06 0.02 0.01 Mean Positive Reduction 9.26 24.06 22.84 6.18 4.91 1.51 2.17 2.11 4.19 1.68 Median Positive Reduction 10.05 26.19 19.63 5.28 2.80 1.48 1.66 1.98 3.21 1.82 Max. Percentage Partially Satisfied 100.0 100.0 100.0 100.0 68.75 87.50 93.75 87.50 81.25 87.50 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 704 1905 1354 291 222 382 517 443 289 392 No. 100% Occ. Partially Satisfied 21 56 42 5 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 196 129 150 1749 0 654 97 111 376 575 Minimum Individual Dissatisfaction 0.34 0.39 0.15 0.37 0.17 0.22 0.29 0.13 0.64 0.30 Maximum Individual Dissatisfaction 34.93 55.79 14.57 16.01 15.14 23.34 23.26 23.26 36.62 23.43 Median Individual Dissatisfaction 3.02 1.51 2.08 3.39 2.73 3.22 2.78 2.96 2.89 3.44 Average Individual Dissatisfaction 3.36 1.63 2.24 3.44 2.63 4.27 4.05 4.12 3.51 4.36 Better Performing Configurations 866 1426 1430 309 448 491 522 546 276 458 Under-Performing Configurations 1438 878 874 1995 1856 1813 1782 1758 2028 1846 Percentage Better Configurations 37.59 61.89 62.07 13.41 19.44 21.31 22.66 23.70 11.98 19.88 Configuration Index (best) 233 793 281 645 494 1417 69 1057 1269 1653 Configuration Performance Score 29.68 37.08 33.07 28.77 31.08 40.85 46.24 29.17 34.11 36.25 Least Effective Configuration Index 910 862 1918 1917 2273 2049 2009 2049 2045 2057 Performance Score 5.57 3.43 2.75 6.07 3.61 9.05 9.35 9.66 6.69 11.55 Performance Range Ratio 5.33 10.81 12.03 4.74 8.61 4.51 4.95 3.02 5.10 3.14 Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 19) to be lowest for ‘Case 6’ and highest for ‘Case 3a’, varying in the range of 3.30% to 43.09%. The minimum P r was found to be 0.004%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 11.21%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially decreased when all occupants demanded an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants decreased for a variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) performed lowest at a P r of 3.30%. It was observed that the highest P r value are associated with high Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. However, a low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. However, no correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3a’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-preferred occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 67 Figure 56: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 56 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 19). It is seen here that 81.25% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 20 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 20: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 184 107 81 143 315 491 464 543 183 458 5-10% 243 76 149 116 47 0 58 3 81 0 10-15% 394 119 51 42 61 0 0 0 12 0 15-20% 45 24 474 8 13 0 0 0 0 0 20-25% 0 131 76 0 12 0 0 0 0 0 25-30% 0 551 85 0 0 0 0 0 0 0 30-35% 0 342 252 0 0 0 0 0 0 0 35-40% 0 76 235 0 0 0 0 0 0 0 40-45% 0 0 27 0 0 0 0 0 0 0 45-50% 0 0 0 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 866 1426 1430 309 448 491 522 546 276 458 68 Ranked Configurations Table 21 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 21: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 0,-30,60,84,0.06,0.6 -30,0,84,84,0.01,0.6 -60,-30,60,36,0.06,0.6 -30,-90,60,84,0.18,0.6 -30,-90,36,60,0.06,0.6 2 -30,-30,60,60,0.06,0.6 -30,-60,36,60,0.6,0.6 -30,-60,36,60,0.01,0.6 -30,-30,36,84,0.6,0.6 0,-30,84,84,0.18,0.6 3a 0,-30,84,84,0.06,0.6 -30,-60,36,60,0.6,0.6 0,-60,84,60,0.18,0.6 0,-60,84,60,0.01,0.6 0,-60,84,60,0.6,0.6 3b -30,0,60,60,0.18,0.6 0,-90,60,36,0.06,0.6 -90,-60,36,60,0.01,0.6 -60,0,36,36,0.06,0.6 -90,-90,36,60,0.18,0.6 3c 0,-90,60,36,0.01,0.18 -60,-30,60,36,0.18,0.18 -30,0,60,84,0.06,0.18 0,-90,60,36,0.6,0.18 -60,-30,36,60,0.18,0.18 4a -60,-30,84,60,0.06,0.6 -60,-60,84,36,0.6,0.6 -90,-90,60,36,0.01,0.6 -90,-90,60,36,0.18,0.6 -60,-90,60,84,0.01,0.6 4b 0,0,60,60,0.18,0.6 -30,0,36,36,0.06,0.6 -30,0,84,60,0.01,0.6 -30,-90,84,84,0.18,0.6 -30,-90,60,36,0.6,0.6 4c -30,-90,60,36,0.6,0.6 -30,-90,60,36,0.01,0.6 0,-90,60,36,0.01,0.6 -30,-90,60,36,0.18,0.6 0,-90,60,36,0.06,0.6 5 -60,0,84,60,0.18,0.6 -90,-30,84,60,0.6,0.6 -60,0,60,84,0.01,0.6 -60,0,60,84,0.18,0.6 0,-30,84,60,0.01,0.6 6 -60,-90,60,60,0.18,0.6 0,-90,36,36,0.18,0.6 0,-90,84,84,0.06,0.6 -30,-90,84,36,0.06,0.6 -30,-90,36,84,0.6,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 57 and Fig. 58). For ‘Case 2’, the P ps values were highest for C index =2,6,10…,2+4n, and lowest for the Cindex=4,8,12…4+4n. From the D CL curve, it was observed that D CL is highest for C index =1, i+4n, i= {4,5}, and lowest for C index =i+4n, i= {2,3}. Figure 57: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index =2,6,10…,1+4n, and lowest for C index =1,5,9,...1+4n. From the D CL curve, it was observed that D CL is highest for C index =1,i+4n, i={4,5}, and lowest for C index =2,6,10,…,2+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 58: Cluster patterns for Case 3a 69 4.6 September 22, 9am A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 22). The following values correspond only to calculations made for a single point-in-time, 9am on September 22. Table 22: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 76.63 56.97 81.85 75.97 76.23 46.46 49.02 46.20 67.46 46.43 Max. Cumulative Dissatisfaction 101.7 73.44 118.2 105.0 113.9 127.4 122.0 124.6 91.48 129.6 Min. Cumulative Dissatisfaction 34.67 17.49 21.88 36.02 28.49 36.79 35.69 34.75 34.98 38.38 Median Cumulative Dissatisfaction 57 30 40 65 48 52 47 48 54 54 Mode Cumulative Dissatisfaction 48.00 30.00 39.00 65.00 48.00 47.00 44.00 48.00 45.00 54.00 Average Cumulative Dissatisfaction 61.11 33.57 49.32 61.73 53.57 64.86 60.72 61.92 60.06 66.30 Maximum Percentage Reduction 54.75 69.30 73.26 52.59 62.63 20.82 27.18 24.78 48.15 17.32 Minimum Percentage Reduction 0.12 1.82 0.01 2.56 0.04 0.003 0.01 0.004 0.27 0.006 Mean Positive Reduction 28.11 45.20 48.01 23.65 36.44 5.79 10.21 6.31 25.47 5.39 Median Positive Reduction 33.48 47.62 52.10 18.66 37.01 3.57 9.60 4.21 24.59 2.64 Max. Percentage Partially Satisfied 100.0 100.0 100.0 75.00 81.25 81.25 81.25 75.00 87.50 75.00 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 1741 2008 1982 189 1510 137 1082 254 1385 194 No. 100% Occ. Partially Satisfied 388 1116 1114 0 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 152 47 161 0 63 30 2 0 45 49 Minimum Individual Dissatisfaction 0.53 0.41 0.28 0.37 0.21 0.43 0.21 0.11 0.76 0.56 Maximum Individual Dissatisfaction 13.95 9.61 11.98 11.69 11.89 21.68 21.60 21.70 13.69 21.77 Median Individual Dissatisfaction 3.23 1.57 2.03 3.33 2.63 3.76 3.23 3.36 3.31 3.73 Average Individual Dissatisfaction 3.82 2.10 3.08 3.86 3.35 4.05 3.80 3.87 3.75 4.14 Better Performing Configurations 1868 2138 2046 2069 2075 531 1231 843 1534 408 Under-Performing Configurations 436 166 258 235 229 1773 1073 1461 770 1896 Percentage Better Configurations 81.08 92.80 88.80 89.80 90.06 23.05 53.43 36.59 66.58 17.71 Configuration Index (best) 2045 1733 573 1585 2057 610 1245 101 2181 710 Configuration Performance Score 25.09 11.92 37.20 14.29 30.65 45.67 35.84 53.90 21.44 49.79 Least Effective Configuration Index 691 1279 1266 739 714 2113 1725 673 714 2113 Performance Score 1.98 1.59 1.35 -0.22 0.41 4.77 3.72 4.34 1.99 5.28 Performance Range Ratio 12.74 7.50 27.56 -64.9 74.76 9.57 9.63 12.42 10.77 9.43 Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 22) to be lowest for ‘Case 6’ and highest for ‘Case 3a’, varying in the range of 17.32% to 73.26%. The minimum P r was found to be 0.003%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 48.15%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially decreased when all occupants demanded an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants decreased for variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) did perform the worst at a P r of 17.32%. It was observed that the highest P r value is associated with much higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3a’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 70 Figure 59: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 59 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 22). It is seen here that 87.50% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 23 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 23: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 108 25 76 47 107 304 202 468 113 275 5-10% 319 74 13 283 11 129 469 258 39 36 10-15% 60 70 0 561 0 8 316 20 26 36 15-20% 71 14 22 177 51 89 190 3 274 61 20-25% 126 27 56 260 250 1 27 94 341 0 25-30% 113 116 152 69 246 0 27 0 110 0 30-35% 247 84 66 180 100 0 0 0 457 0 35-40% 508 233 0 119 363 0 0 0 51 0 40-45% 128 128 155 144 388 0 0 0 106 0 45-50% 123 519 280 28 249 0 0 0 17 0 50-55% 65 276 514 201 205 0 0 0 0 0 55-60% 0 290 434 0 60 0 0 0 0 0 60-65% 0 126 122 0 45 0 0 0 0 0 65-70% 0 156 96 0 0 0 0 0 0 0 70-75% 0 0 60 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 1868 2138 2046 2069 2075 531 1231 843 1534 408 71 Ranked Configurations Table 24 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 24: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 -90,-60,36,60,0.01,0.6 -90,-60,36,60,0.06,0.6 -90,-60,36,60,0.18,0.6 -90,-30,36,84,0.6,0.6 -90,-30,36,84,0.18,0.6 2 -90,0,36,36,0.18,0.6 -90,0,36,36,0.01,0.6 -90,0,36,36,0.06,0.6 -90,0,36,36,0.6,0.6 -90,0,36,60,0.6,0.6 3a 0,-90,84,84,0.01,0.6 0,-90,84,84,0.6,0.6 -90,-90,36,60,0.06,0.6 -90,-90,36,60,0.6,0.6 0,-60,84,60,0.01,0.6 3b -60,-90,36,36,0.6,0.6 -60,-90,36,36,0.01,0.6 -60,-90,36,36,0.06,0.6 -60,-90,36,36,0.18,0.6 -90,-90,84,36,0.01,0.01 3c -90,-60,36,84,0.06,0.6 -90,-60,36,84,0.01,0.6 0,-30,84,36,0.18,0.6 0,0,84,84,0.01,0.6 0,-90,84,36,0.6,0.6 4a -30,0,36,84,0.6,0.18 -30,-30,36,84,0.6,0.18 -30,0,36,36,0.01,0.18 -90,-30,36,84,0.01,0.6 -90,-30,36,36,0.06,0.18 4b -60,0,60,84,0.01,0.6 -30,-30,60,84,0.01,0.6 -90,-30,36,36,0.18,0.6 -30,-30,60,36,0.6,0.6 0,-30,84,60,0.06,0.6 4c 0,0,84,36,0.18,0.6 -90,-60,36,36,0.18,0.6 -90,-90,36,36,0.6,0.18 0,0,84,60,0.01,0.6 0,-90,84,36,0.6,0.6 5 -90,-90,36,60,0.18,0.6 -90,-30,36,60,0.18,0.6 -90,-30,36,60,0.01,0.6 -90,-30,36,60,0.6,0.6 -90,-90,36,60,0.6,0.6 6 -30,0,84,84,0.18,0.18 -30,-30,84,84,0.01,0.18 -90,-60,84,84,0.6,0.6 -30,-90,84,60,0.06,0.18 -30,-30,84,60,0.18,0.18 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 60 and Fig. 61). For ‘Case 2’, the P ps values were highest for C index =i+4n, i= {2,3}, and lowest for the C index =5,9,13…5+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5,9…,1+4n, and lowest for C index =3,7,11…,3+4n. Figure 60: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index =i+4n, i={2,3}, and lowest for C index =1,5,9,..,1+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5,9,…,1+4n, and lowest for C index =4,8,12,…,4+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 61: Cluster patterns for Case 3a 72 4.7 September 22, noon A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 25). The following values correspond only to calculations made for a single point-in-time, noon on September 22. Table 25: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 47.30 31.34 44.82 36.32 38.97 38.21 40.47 36.90 41.08 38.53 Max. Cumulative Dissatisfaction 176.1 215.0 82.49 105.9 89.85 157.5 153.2 155.1 159.5 159.0 Min. Cumulative Dissatisfaction 35.96 18.49 22.91 37.19 26.29 32.43 36.47 35.11 35.63 33.68 Median Cumulative Dissatisfaction 53 30 41 58 45 65 62 63 55 67 Mode Cumulative Dissatisfaction 67.00 30.00 42.00 80.00 66.00 90.00 82.00 85.00 70.00 91.00 Average Cumulative Dissatisfaction 63.72 30.41 40.87 60.40 47.27 82.78 78.36 79.86 65.25 84.31 Maximum Percentage Reduction 23.97 40.99 48.89 -2.40 32.53 15.13 9.86 4.85 13.26 12.59 Minimum Percentage Reduction 0.01 0.01 0.23 0.01 0.04 0.17 0.01 0.04 0.06 0.19 Mean Positive Reduction 9.95 18.65 20.91 N/A 14.62 4.13 5.27 2.32 5.98 1.91 Median Positive Reduction 9.42 22.05 23.03 N/A 14.38 0.95 6.42 2.26 6.17 0.57 Max. Percentage Partially Satisfied 87.50 100.0 100.0 68.75 68.75 56.25 75.00 62.50 62.50 56.25 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 593 1248 1251 47 748 14 562 310 210 2 No. 100% Occ. Partially Satisfied 0 346 4 0 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 69 63 66 1619 0 1060 477 486 486 1057 Minimum Individual Dissatisfaction 0.48 0.42 0.23 0.39 0.18 0.43 0.30 0.23 0.84 0.56 Maximum Individual Dissatisfaction 60.97 97.00 24.59 26.08 25.75 29.42 29.36 29.37 62.91 29.51 Median Individual Dissatisfaction 3.39 1.63 2.33 3.67 2.86 4.09 3.56 3.79 3.51 4.18 Average Individual Dissatisfaction 3.98 1.90 2.55 3.78 2.95 5.18 4.90 4.99 4.08 5.27 Better Performing Configurations 802 1623 1691 0 577 6 338 85 491 9 Under-Performing Configurations 1502 681 613 2304 1727 2298 1966 2219 1813 2295 Percentage Better Configurations 34.81 70.44 73.39 0.00 25.04 0.26 14.67 3.69 21.31 0.39 Configuration Index (best) 365 316 2189 N/A 874 373 1509 1313 2001 389 Configuration Performance Score 37.83 40.73 24.80 N/A 31.50 21.58 36.63 31.75 29.57 24.66 Least Effective Configuration Index 425 714 434 N/A 853 1 377 577 857 1 Performance Score 5.06 3.37 2.71 N/A 2.27 4.61 7.40 10.64 5.76 5.41 Performance Range Ratio N/A Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 25) to be lowest for ‘Case 3b’ and highest for ‘Case 3a’, varying in the range of -2.40% to 48.89%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 13.26%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative dissatisfaction of all occupants substantially increased when all occupants demanded an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the P r value of all occupants decreased significantly for a variable contrast ratio. However, the cumulative satisfaction was generally higher for Case 4b. In addition, the assumed worst-case scenario (Case 6) actually performed better at a P r of 12.59%. It was observed that a high P r value (Cases 3b, 4a, 4c, and 6) is associated with higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially- satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3a’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 73 Figure 62: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 62 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 25). It is seen here that 62.50% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 26 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 26: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 204 489 29 0 95 4 123 85 145 8 5-10% 233 36 571 0 85 1 215 0 320 0 10-15% 199 89 49 0 126 0 0 0 26 1 15-20% 98 120 106 0 131 1 0 0 0 0 20-25% 68 265 172 0 44 0 0 0 0 0 25-30% 0 391 275 0 75 0 0 0 0 0 30-35% 0 88 282 0 21 0 0 0 0 0 35-40% 0 126 119 0 0 0 0 0 0 0 40-45% 0 19 66 0 0 0 0 0 0 0 45-50% 0 0 22 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 802 1623 1691 0 577 6 338 85 491 9 74 Ranked Configurations Table 27 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 27: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 0,-60,60,60,0.01,0.6 -60,-30,36,84,0.06,0.18 -30,-60,60,36,0.18,0.18 -90,0,84,60,0.18,0.6 -30,0,84,36,0.01,0.18 2 0,-60,36,60,0.06,0.01 0,0,36,60,0.06,0.01 0,-30,36,84,0.06,0.01 0,-30,60,36,0.18,0.01 0,-60,36,36,0.6,0.01 3a -90,-90,36,60,0.01,0.6 -90,-60,36,60,0.01,0.6 -90,0,36,84,0.06,0.6 -90,0,36,84,0.01,0.6 -90,-90,36,60,0.6,0.6 3b N/A N/A N/A N/A N/A 3c -30,-60,36,36,0.06,0.18 0,-90,60,60,0.18,0.18 -60,-30,36,36,0.18,0.18 -30,-60,36,36,0.01,0.18 -60,-30,36,36,0.06,0.18 4a 0,-60,60,84,0.18,0.6 0,-60,60,84,0.01,0.6 0,-60,60,84,0.6,0.6 0,-60,60,84,0.06,0.6 0,-60,84,84,0.06,0.6 4b -60,-60,60,60,0.18,0.6 -60,-30,60,84,0.6,0.6 -90,-30,60,84,0.18,0.6 -90,-60,36,84,0.18,0.6 -90,-30,36,36,0.01,0.6 4c -60,-30,36,60,0.6,0.6 -30,-30,36,60,0.18,0.6 0,-90,60,84,0.18,0.6 -30,-30,60,36,0.06,0.6 0,-60,36,84,0.18,0.6 5 -90,-30,84,84,0.6,0.6 -90,-90,84,60,0.6,0.6 -90,-90,84,60,0.06,0.6 -90,-60,84,60,0.6,0.6 -60,-90,84,36,0.6,0.6 6 0,-60,60,60,0.01,0.6 -60,-30,36,84,0.06,0.18 -30,-60,60,36,0.18,0.18 -90,0,84,60,0.18,0.6 -30,0,84,36,0.01,0.18 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 63 and Fig. 64). For ‘Case 2’, the P ps values were highest for C index =i+4n, i= {2,3}, and lowest for the C index =1,5,9,…,1+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5,9…,1+4n, and lowest for C index =3,7,11…,3+4n. Figure 63: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index = i+4n, i={2,3}, and lowest for C index =5,9,..,5+4n. From the D CL curve, it was observed that D CL is highest for C index =5,9,13,…,5+4n, and lowest for C index =2,6,10,…,2+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 64: Cluster patterns for Case 3a 75 4.8 September 22, 3pm A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 28). The following values correspond only to calculations made for a single point-in-time, 3pm on September 22. Table 28: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 61.94 41.70 63.18 64.80 60.94 44.88 47.13 43.96 58.83 45.40 Max. Cumulative Dissatisfaction 333.1 464.7 162.7 171.3 161.6 177.9 183.7 179.0 334.2 179.0 Min. Cumulative Dissatisfaction 35.76 18.64 22.41 35.45 29.91 38.83 39.91 36.99 39.64 38.78 Median Cumulative Dissatisfaction 53 29 39 57 43 60 61 60 59 63 Mode Cumulative Dissatisfaction 51.00 36.00 45.00 71.00 42.00 49.00 47.00 46.00 59.00 49.00 Average Cumulative Dissatisfaction 60.35 30.80 40.95 58.49 46.94 74.16 72.07 72.68 63.42 75.80 Maximum Percentage Reduction 42.26 55.31 64.53 45.29 50.92 13.48 15.33 15.86 32.61 14.59 Minimum Percentage Reduction 0.12 0.08 0.19 0.04 1.01 0.05 0.004 0.02 0.01 0.06 Mean Positive Reduction 19.42 29.82 38.89 17.91 25.00 8.51 5.37 4.83 11.88 6.99 Median Positive Reduction 18.48 32.18 39.49 17.30 29.29 12.38 4.78 2.56 11.96 4.58 Max. Percentage Partially Satisfied 81.25 100.0 87.50 100.0 75.00 68.75 81.25 75.00 75.00 81.25 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 763 1532 1906 234 752 351 502 368 326 334 No. 100% Occ. Partially Satisfied 0 22 0 3 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 0 0 0 0 0 575 0 0 0 575 Minimum Individual Dissatisfaction 0.56 0.38 0.24 0.29 0.09 0.56 0.37 0.10 0.17 0.51 Maximum Individual Dissatisfaction 194.2 311.0 82.70 82.39 82.60 75.93 77.10 76.71 199.4 75.94 Median Individual Dissatisfaction 3.46 1.69 2.22 3.69 2.85 4.26 3.90 3.95 3.66 4.37 Average Individual Dissatisfaction 3.77 1.92 2.56 3.66 2.93 4.64 4.51 4.54 3.96 4.74 Better Performing Configurations 1519 2128 2135 1613 2175 60 395 167 1203 83 Under-Performing Configurations 785 176 169 691 129 2244 1909 2137 1101 2221 Percentage Better Configurations 65.93 92.36 92.66 70.01 94.40 2.60 17.14 7.25 52.21 3.60 Configuration Index (best) 973 1633 1057 1173 1593 2045 1305 1337 1379 745 Configuration Performance Score 26.64 27.59 21.98 29.61 16.29 29.86 43.21 34.75 36.14 29.67 Least Effective Configuration Index 95 269 225 3 186 233 185 237 135 233 Performance Score 0.88 0.92 0.16 -0.33 0.43 4.50 4.74 4.24 1.44 4.53 Performance Range Ratio Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 28) to be lowest for ‘Case 4a’ and highest for ‘Case 3a’, varying in the range of 13.48% to 64.53%. The minimum P r was found to be 0.004%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 32.61%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially increased when all occupants demanded an illuminance setpoint of 500lx, in contrast to when they all preferred a variable or an illuminance of 2000lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the P r value increased for an ideal contrast ratio (R CR =1) than for a variable contrast ratio or a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) actually performed poorly at a P r of 14.59%. It was observed that the highest P r value is associated with high Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3b’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 76 Figure 65: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 65 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 28). It is seen here that 75% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 29 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 29: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 33 48 41 47 71 24 215 119 401 45 5-10% 57 101 31 275 406 0 122 12 112 2 10-15% 280 449 26 281 174 36 56 31 239 36 15-20% 517 59 20 409 82 0 2 5 197 0 20-25% 310 166 57 296 140 0 0 0 146 0 25-30% 259 135 459 170 317 0 0 0 72 0 30-35% 27 227 240 95 539 0 0 0 36 0 35-40% 4 185 202 25 208 0 0 0 0 0 40-45% 32 441 204 14 200 0 0 0 0 0 45-50% 0 245 343 1 34 0 0 0 0 0 50-55% 0 69 323 0 4 0 0 0 0 0 55-60% 0 3 153 0 0 0 0 0 0 0 60-65% 0 0 36 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 1519 2128 2135 1613 2175 60 395 167 1203 83 77 Ranked Configurations Table 30 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 30: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 -60,-90,60,36,0.6,0.6 -60,-90,60,36,0.06,0.6 -90,0,60,60,0.18,0.6 -60,-90,60,36,0.18,0.6 -60,-90,60,36,0.01,0.6 2 -60,-90,60,36,0.6,0.6 -60,-90,60,36,0.06,0.6 -90,0,60,60,0.18,0.6 -60,-90,60,36,0.18,0.6 -60,-90,60,36,0.01,0.6 3a -30,-90,60,36,0.6,0.6 0,-60,84,36,0.06,0.6 0,-60,84,36,0.01,0.6 -90,-30,60,36,0.18,0.6 -90,-30,60,36,0.6,0.6 3b -60,0,36,60,0.18,0.6 -30,0,36,60,0.01,0.6 -60,0,36,60,0.01,0.6 -30,0,36,60,0.18,0.6 -60,-90,60,36,0.01,0.6 3c -60,-90,36,36,0.06,0.6 0,-60,36,36,0.01,0.6 0,-60,36,36,0.06,0.6 -30,-90,36,36,0.01,0.6 -30,-90,60,36,0.06,0.6 4a -90,-60,36,60,0.01,0.6 -30,-30,36,60,0.01,0.6 -90,-30,36,84,0.18,0.6 -90,-30,60,60,0.18,0.6 0,0,60,84,0.06,0.18 4b -60,-30,36,36,0.06,0.6 -60,0,60,36,0.06,0.6 0,-90,36,36,0.06,0.6 -60,-60,36,36,0.06,0.6 -30,-60,36,36,0.01,0.6 4c -60,-30,36,84,0.06,0.6 0,-60,84,60,0.06,0.6 -30,-60,36,60,0.06,0.6 -30,-30,84,60,0.01,0.6 0,-60,60,60,0.18,0.6 5 -60,-30,60,84,0.6,0.06 -90,-90,36,36,0.01,0.6 -30,-30,84,60,0.18,0.06 -30,-30,36,60,0.6,0.06 -90,-30,84,36,0.01,0.6 6 -30,-30,36,60,0.06,0.6 -30,-30,36,60,0.18,0.6 -60,0,84,84,0.06,0.6 -90,-30,60,60,0.6,0.6 -90,-30,60,60,0.18,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 66 and Fig. 67). For ‘Case 2’, P ps values were generally highest for C index =i+4n, i={2,3}, and lowest for C index =1,5,9,…,1+4n. From the D CL curve, it was observed that it is highest for C index =1,5,9…,1+4n, and generally lowest for C index =3,7,11…,3+4n. Figure 66: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index =2,6,…2+4n, and lowest for C index =1,5,9,..,1+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5,9,13,…,1+4n, and lowest for C index =3,7,11,…,3+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 67: Cluster patterns for Case 3a 78 4.9 December 22, 9am A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 31). The following values correspond only to calculations made for a single point-in-time, 9am on December 22. Table 31: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 37.77 20.91 24.91 48.82 28.28 45.50 40.68 42.75 38.75 48.50 Max. Cumulative Dissatisfaction 123.3 52.75 64.54 87.50 73.41 180.8 177.4 179.2 128.4 182.1 Min. Cumulative Dissatisfaction 38.43 19.95 26.29 38.04 27.43 42.04 39.20 39.33 37.39 43.71 Median Cumulative Dissatisfaction 72 32 43 67 53 98 92 95 78 99 Mode Cumulative Dissatisfaction 80.00 36.00 63.00 86.00 72.00 110.0 102.0 107.0 86.00 111.0 Average Cumulative Dissatisfaction 74.39 33.80 43.96 65.43 51.38 102.1 97.48 99.90 78.57 103.7 Maximum Percentage Reduction -1.75 4.58 -5.55 22.09 3.01 7.60 3.62 7.99 3.50 9.88 Minimum Percentage Reduction 0.00 0.04 0.01 0.08 0.43 0.05 0.01 0.01 0.04 0.04 Mean Positive Reduction N/A 2.03 N/A 11.14 1.69 3.87 1.39 2.95 1.11 5.27 Median Positive Reduction N/A 1.97 N/A 11.13 1.43 3.93 1.44 2.89 1.01 5.50 Max. Percentage Partially Satisfied 56.25 75.00 50.00 100.0 68.75 87.50 62.50 81.25 68.75 93.75 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 1 60 0 424 227 297 196 256 163 352 No. 100% Occ. Partially Satisfied 0 0 0 188 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 1167 591 625 1728 1728 1662 613 580 1242 1497 Minimum Individual Dissatisfaction 0.64 0.43 0.28 0.42 0.17 0.44 0.22 0.11 0.82 0.56 Maximum Individual Dissatisfaction 13.92 5.71 6.45 7.88 6.99 21.49 21.41 21.51 13.66 21.41 Median Individual Dissatisfaction 3.86 1.81 2.54 3.92 3.17 5.02 4.62 4.71 4.09 5.26 Average Individual Dissatisfaction 4.65 2.11 2.75 4.09 3.21 6.38 6.09 6.25 4.91 6.48 Better Performing Configurations 0 75 0 503 5 312 137 253 83 344 Under-Performing Configurations 2304 2229 2304 1801 2299 1992 2167 2051 2221 1960 Percentage Better Configurations 0.00 3.26 0.00 21.83 0.22 13.54 5.95 10.98 3.60 14.93 Configuration Index (best) N/A 681 N/A 857 369 1065 1185 417 461 1081 Configuration Performance Score N/A 30.51 N/A 27.75 21.58 29.55 38.81 46.31 31.89 31.83 Least Effective Configuration Index N/A 997 N/A 2093 373 1 73 1 373 1 Performance Score N/A 11.02 N/A 5.85 13.59 9.94 11.91 9.35 12.51 8.81 Performance Range Ratio N/A N/A Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 31) to be lowest for ‘Case 3a’ and highest for ‘Case 3b’, varying in the range of -5.55% to 22.09%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 3.50%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially increased when all occupants demanded an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants increased significantly for variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) actually performed better at a P r of 9.88%. It was observed that a high P r value is associated with high Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3b’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 79 Figure 68: Visual representation of façade configuration and visual comfort condition As explained in section 4.2, Figure 68 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 31). It is seen here that 68.75% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 32 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 32: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 0 75 0 88 5 217 137 224 83 149 5-10% 0 0 0 122 0 95 0 29 0 195 10-15% 0 0 0 134 0 0 0 0 0 0 15-20% 0 0 0 131 0 0 0 0 0 0 20-25% 0 0 0 28 0 0 0 0 0 0 25-30% 0 0 0 0 0 0 0 0 0 0 30-35% 0 0 0 0 0 0 0 0 0 0 35-40% 0 0 0 0 0 0 0 0 0 0 40-45% 0 0 0 0 0 0 0 0 0 0 45-50% 0 0 0 0 0 0 0 0 0 0 50-55% 0 0 0 0 0 0 0 0 0 0 55-60% 0 0 0 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 0 75 0 503 5 312 137 253 83 344 80 Ranked Configurations Table 33 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 33: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 N/A N/A N/A N/A N/A 2 -30,0,84,36,0.06,0.6 -60,-60,84,60,0.18,0.6 -60,-60,84,84,0.06,0.6 -30,-60,84,36,0.06,0.6 -30,0,84,36,0.01,0.6 3a N/A N/A N/A N/A N/A 3b -30,-30,84,84,0.06,0.6 -30,-30,84,84,0.18,0.6 -30,-30,84,84,0.01,0.6 -30,-90,84,84,0.18,0.6 -30,-30,84,60,0.6,0.6 3c 0,-60,60,84,0.6,0.6 0,-60,84,84,0.6,0.6 0,-60,60,84,0.06,0.6 0,-60,60,84,0.01,0.6 0,-60,60,84,0.18,0.6 4a -30,-90,60,36,0.06,0.6 -30,-30,60,60,0.01,0.6 -60,-90,36,84,0.18,0.6 -30,-30,60,60,0.6,0.6 -60,-90,36,84,0.01,0.6 4b -60,0,36,84,0.6,0.6 -90,-30,36,36,0.01,0.6 -90,0,36,60,0.18,0.6 -90,0,36,60,0.01,0.6 -30,-30,36,60,0.01,0.6 4c 0,-60,84,84,0.6,0.6 -30,-30,60,36,0.6,0.6 0,-90,84,60,0.18,0.6 -60,-30,36,84,0.6,0.6 -90,-90,36,36,0.18,0.6 5 0,-90,36,60,0.01,0.6 0,-30,84,84,0.06,0.6 -30,0,36,36,0.06,0.6 0,-60,60,36,0.6,0.6 0,-60,36,60,0.06,0.6 6 -30,-90,60,60,0.06,0.6 -30,-60,60,60,0.06,0.6 -90,0,60,60,0.18,0.6 -90,0,60,84,0.01,0.6 -90,0,60,84,0.06,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3b’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 69 and Fig. 70). For ‘Case 2’, P ps values were generally highest for C index =2,6,10,…,2+4n, and lowest for C index =4,8,12,…,4+4n. From the D CL curve, it was observed that it is highest for C index =4,8,12…,4+4n, and generally lowest for C index =2,6,10…,2+4n. Figure 69: Cluster patterns for Case 2 For ‘Case 3b’, the P ps values were seen to be highest for C index =2,6,…2+4n, and lowest for C index =4,8,12,..,4+4n. From the D CL curve, it was observed that D CL is highest for C index =4,8,12,…,4+4n, and lowest for C index =2,6,10,…,2+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 70: Cluster patterns for Case 3b 81 4.10 December 22, noon A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 34). The following values correspond only to calculations made for a single point-in-time, noon on December 22. Table 34: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 57.32 39.36 59.47 53.87 55.70 39.75 42.53 38.76 52.01 39.56 Max. Cumulative Dissatisfaction 246.3 330.1 112.2 129.2 116.3 153.1 163.2 159.3 233.3 154.8 Min. Cumulative Dissatisfaction 38.34 18.65 26.43 45.60 28.46 38.02 41.30 35.95 39.53 38.92 Median Cumulative Dissatisfaction 54 28 40 56 44 56 52 53 54 59 Mode Cumulative Dissatisfaction 78.00 24.00 28.00 69.00 54.00 44.00 45.00 43.00 81.00 114.0 Average Cumulative Dissatisfaction 58.82 30.01 42.68 58.50 47.78 69.52 66.73 67.20 60.51 70.99 Maximum Percentage Reduction 33.12 52.62 55.55 15.35 48.90 4.37 2.88 7.24 24.00 1.63 Minimum Percentage Reduction 0.13 0.13 0.32 0.03 0.23 0.32 0.02 0.03 0.02 0.01 Mean Positive Reduction 17.83 29.01 35.71 7.89 21.87 1.78 0.96 4.25 9.27 0.63 Median Positive Reduction 19.15 34.80 41.55 8.52 26.58 1.89 0.84 4.69 8.20 0.52 Max. Percentage Partially Satisfied 93.75 100.0 100.0 75.00 75.00 56.25 75.00 68.75 68.75 56.25 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 1110 1860 2021 220 1372 4 696 86 410 5 No. 100% Occ. Partially Satisfied 0 622 312 0 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 24 28 15 0 0 575 0 0 0 575 Minimum Individual Dissatisfaction 0.67 0.39 0.32 0.41 0.19 0.57 0.34 0.19 0.22 0.63 Maximum Individual Dissatisfaction 71.90 115.8 29.75 27.69 29.33 28.64 29.67 29.25 73.97 29.09 Median Individual Dissatisfaction 3.33 1.69 2.28 3.40 2.83 3.63 3.20 3.32 3.44 3.64 Average Individual Dissatisfaction 3.68 1.88 2.67 3.66 2.99 4.35 4.17 4.20 3.78 4.44 Better Performing Configurations 1357 2009 1975 737 1785 42 36 45 980 29 Under-Performing Configurations 947 295 329 1567 519 2262 2268 2259 1324 2275 Percentage Better Configurations 58.90 87.20 85.72 31.99 77.47 1.82 1.56 1.95 42.53 1.26 Configuration Index (best) 1605 1909 1461 2141 1349 701 973 949 2170 377 Configuration Performance Score 25.80 26.01 22.38 31.28 23.78 21.40 34.49 31.63 33.29 41.38 Least Effective Configuration Index 483 1175 1183 997 982 997 997 1001 854 1001 Performance Score 3.01 1.94 1.84 3.88 1.48 10.01 12.91 8.39 3.96 14.28 Performance Range Ratio Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 34) to be lowest for ‘Case 6’ and highest for ‘Case 3a’, varying in the range of 1.63% to 55.55%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 24%. The low P r value for ‘Case 1’ implies that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially decreased when all occupants demanded an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the cumulative satisfaction of all occupants increased significantly for variable or an ideal contrast ratio (R CR =1) than for a contrast ratio setpoint of 3. In addition, the assumed worst-case scenario (Case 6) actually performed the least at a P r of 1.63%. It was observed that the highest P r value is associated with high Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3b’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 82 Figure 71: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 71 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 34). It is seen here that 68.75% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 35 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 35: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 79 255 38 264 440 440 42 28 259 29 5-10% 234 37 34 152 70 70 0 17 381 0 10-15% 130 190 65 315 62 62 0 0 101 0 15-20% 275 330 229 6 37 37 0 0 204 0 20-25% 315 33 133 0 178 178 0 0 35 0 25-30% 293 26 314 0 347 347 0 0 0 0 30-35% 31 140 37 0 465 465 0 0 0 0 35-40% 0 409 80 0 144 144 0 0 0 0 40-45% 0 188 443 0 20 20 0 0 0 0 45-50% 0 215 271 0 22 22 0 0 0 0 50-55% 0 186 328 0 0 0 0 0 0 0 55-60% 0 0 3 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 1357 2009 1975 737 1785 1785 42 45 980 29 83 Ranked Configurations Table 36 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 36: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 -60,-90,36,60,0.18,0.6 -60,-90,36,60,0.6,0.6 -60,-90,36,60,0.06,0.6 -60,-60,36,84,0.06,0.6 -90,-90,60,84,0.18,0.6 2 -90,-30,36,84,0.18,0.6 -90,-30,36,84,0.06,0.6 -90,-30,36,84,0.6,0.6 -30,0,84,36,0.01,0.6 -30,0,84,36,0.06,0.6 3a -60,-60,36,60,0.18,0.6 -60,-60,36,60,0.06,0.6 -90,0,60,36,0.6,0.6 -60,-60,36,60,0.01,0.6 -90,0,60,36,0.01,0.6 3b -90,-60,84,60,0.01,0.6 0,-90,60,36,0.06,0.06 0,-30,36,36,0.01,0.06 -90,-60,84,60,0.6,0.6 -90,-90,60,36,0.06,0.6 3c -60,-30,60,36,0.18,0.6 -90,0,60,60,0.06,0.6 -90,0,60,36,0.06,0.6 -30,-90,84,84,0.06,0.01 -60,-30,84,84,0.01,0.01 4a -30,0,84,60,0.01,0.6 -30,-90,84,36,0.18,0.6 -30,-90,84,36,0.6,0.6 -30,0,84,84,0.01,0.6 -30,0,84,84,0.06,0.6 4b -30,-60,84,36,0.01,0.6 -30,0,84,84,0.18,0.6 -30,-90,84,36,0.18,0.6 -90,0,84,36,0.6,0.6 -30,-30,84,36,0.18,0.6 4c -30,-60,60,84,0.18,0.6 -30,-60,60,84,0.01,0.6 -30,0,84,60,0.06,0.6 -30,0,84,60,0.01,0.6 -30,0,84,60,0.18,0.6 5 -90,-90,36,36,0.06,0.18 -60,0,36,60,0.18,0.06 -60,-30,60,36,0.6,0.6 -90,-90,60,36,0.06,0.18 -90,-90,60,36,0.18,0.18 6 0,-60,60,84,0.06,0.6 -30,-30,84,84,0.18,0.6 -30,-60,84,84,0.01,0.6 -30,0,84,60,0.01,0.6 0,-60,84,84,0.18,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 72 and Fig. 73). For ‘Case 2’, P ps values were generally highest for C index =i+4n, i= {2,3} and lowest for C index =1,5,9,…,1+4n. From the D CL curve, it was observed that it is highest for C index =1,5,9…,1+4n, and generally lowest for C index =3,7,11…,3+4n. Figure 72: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index =i+4n, i={2,3} and lowest for C index =5,9,13,..,5+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5, 9,…,1+4n, and lowest for C index =3,7,11,…,3+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 73: Cluster patterns for Case 3a 84 4.11 December 22, 3pm A summary of all the primary calculations has been provided for the ten occupant preference cases (Table 37). The following values correspond only to calculations made for a single point-in-time, 3pm on December 22. Table 37: Summary of Results by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 55.88 36.91 53.33 52.10 51.49 44.86 41.71 41.73 51.26 44.44 Max. Cumulative Dissatisfaction 408.7 579.2 182.0 194.4 184.7 226.9 238.5 233.4 383.7 228.4 Min. Cumulative Dissatisfaction 40.42 20.52 23.44 38.33 31.33 40.38 37.63 37.93 38.30 41.62 Median Cumulative Dissatisfaction 52 27 36 57 43 62 63 62 55 65 Mode Cumulative Dissatisfaction 45.00 23.00 47.00 73.00 57.00 45.00 42.00 40.00 49.00 45.00 Average Cumulative Dissatisfaction 60.84 30.69 40.09 58.21 46.49 76.32 74.10 74.60 63.98 77.89 Maximum Percentage Reduction 27.68 44.41 56.04 26.43 39.15 9.99 9.80 9.11 25.28 6.36 Minimum Percentage Reduction 0.04 0.02 0.01 0.03 0.77 0.03 0.03 0.06 0.01 0.05 Mean Positive Reduction 13.35 27.28 28.05 10.11 20.32 2.60 3.72 3.79 6.59 2.02 Median Positive Reduction 14.04 31.28 33.21 8.53 20.41 1.79 3.14 3.54 4.88 1.45 Max. Percentage Partially Satisfied 100.0 100.0 100.0 81.25 62.50 81.25 75.00 81.25 68.75 81.25 Min. Percentage Partially Satisfied 0 0 0 0 0 0 0 0 0 0 No. >50% Occ. Partially Satisfied 675 1798 1899 150 1272 180 98 95 143 125 No. 100% Occ. Partially Satisfied 38 254 9 0 0 0 0 0 0 0 No. of Cases Occ. Dissatisfied 153 54 94 139 0 575 575 575 139 575 Minimum Individual Dissatisfaction 0.53 0.34 0.24 0.40 0.19 0.33 0.44 0.10 0.17 0.37 Maximum Individual Dissatisfaction 189.5 309.2 74.97 74.26 74.44 74.11 75.28 74.89 194.3 74.15 Median Individual Dissatisfaction 3.52 1.71 2.28 3.69 2.82 4.18 3.73 3.96 3.56 4.32 Average Individual Dissatisfaction 3.80 1.92 2.51 3.64 2.91 4.77 4.63 4.66 4.00 4.87 Better Performing Configurations 1434 1685 2131 765 1576 256 140 278 782 186 Under-Performing Configurations 870 619 173 1539 728 2048 2164 2026 1522 2118 Percentage Better Configurations 62.24 73.13 92.49 33.20 68.40 11.11 6.08 12.07 33.94 8.07 Configuration Index (best) 301 664 493 469 1710 733 613 505 938 733 Configuration Performance Score 28.43 35.76 34.18 31.99 17.52 32.82 31.16 28.62 36.57 30.18 Least Effective Configuration Index 2129 535 1423 1166 2057 1751 1901 2113 2053 2021 Performance Score 3.15 2.38 1.60 3.30 1.16 8.38 7.59 7.90 3.31 10.17 Performance Range Ratio Maximum Percentage Reduction The Maximum Percentage Reduction in Cumulative Dissatisfaction (P r ) was found (Table 37) to be lowest for ‘Case 6’ and highest for ‘Case 3a’, varying in the range of 6.36% to 56.04%. The minimum P r was found to be 0.01%. ‘Case 5’ is a closer representation of an unexpected distribution of real occupant preferences and for this case, the P r value was calculated to be 25.28%. The lower P r values for ‘Case 1’ imply that while designing adaptive façade controls, assuming equal visual comfort preferences and importance factors for all occupants (Case 1) might not be the most effective way to control the façade to increase the visual comfort satisfaction of all the occupants. It was observed that for cases 3a, 3b, and 3c, which all assume a high occupant-set importance factor for illuminance (If e =90) with varying illuminance setpoints, the cumulative satisfaction of all occupants substantially decreased when all occupants preferred an illuminance setpoint of 2000lx, in contrast to when they all preferred a variable or a standard illuminance setpoint of 500lx. For cases 4a, 4b, and 4c, which assume a high occupant-set importance factor for contrast ratio (If cr =90), it was observed that the P r value of all occupants increased slightly for ‘Case 4a’ but in general, the percentage reduction was highest for ‘Case 4c’. In addition, the assumed worst-case scenario (Case 6) actually performed least at a P r of 6.36%. It was observed that a high P r value is associated with much higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. A low P r value does not necessarily imply a low P ps value. A significant number of configurations were found to partially-satisfy all the occupants. No correlation was found between P ps and the number of satisfying configurations. Both cases, ‘Case 2’, providing a higher degree of glare-protection, and ‘Case 3b’, providing a higher degree of illuminance, assume an ideal glare DGP value of 0.20 and a standard illuminance of 500lx. These two cases can be considered to be the more commonly-demanded occupant visual comfort preferences and are discussed in greater detail in the next chapter and in relation to ‘Cases 1’ and ‘Case 3b’. 85 Figure 74: Visual representation of façade configuration and visual comfort condition As explained in Section 4.2, Figure 74 is an automatically-generated, color-coded, visual and informative representation of the façade configuration, occupant visual comfort conditions, and the number as well as distribution of partially satisfied and unsatisfied occupants in the test space, calculated for Case 5 and displayed for the best performing configuration (Table 37). It is seen here that 68.75% of occupants and those away from the south-western façade, are more partially satisfied than those who are not. Distribution of Percentage Reduction Table 38 shows the distribution of the number of configurations performing better than the baseline according to the range of percentage reduction they produce in the cumulative dissatisfaction. Table 38: Distribution of Percentage Reduction by Case Case 1 2 3a 3b 3c 4a 4b 4c 5 6 0-5% 208 187 203 207 21 213 92 217 410 172 5-10% 271 109 99 224 87 43 48 61 239 14 10-15% 280 96 471 145 273 0 0 0 37 0 15-20% 442 51 69 69 352 0 0 0 46 0 20-25% 206 32 29 116 345 0 0 0 47 0 25-30% 27 281 51 4 413 0 0 0 3 0 30-35% 0 267 274 0 67 0 0 0 0 0 35-40% 0 523 208 0 18 0 0 0 0 0 40-45% 0 139 472 0 0 0 0 0 0 0 45-50% 0 0 165 0 0 0 0 0 0 0 50-55% 0 0 86 0 0 0 0 0 0 0 55-60% 0 0 4 0 0 0 0 0 0 0 60-65% 0 0 0 0 0 0 0 0 0 0 65-70% 0 0 0 0 0 0 0 0 0 0 70-75% 0 0 0 0 0 0 0 0 0 0 75-80% 0 0 0 0 0 0 0 0 0 0 80-85% 0 0 0 0 0 0 0 0 0 0 85-90% 0 0 0 0 0 0 0 0 0 0 90-95% 0 0 0 0 0 0 0 0 0 0 95-100% 0 0 0 0 0 0 0 0 0 0 Total 1434 1685 2131 765 1576 256 140 278 782 186 86 Ranked Configurations Table 39 shows the composition of the top five configurations for each of the ten cases and ranked by their performance scores. Cases 2, 3a, 3b, and 5 are discussed in more detail in the next chapter. Table 39: Top Five Ranked Configurations by Case Rank #1 Rank #2 Rank #3 Rank #4 Rank #5 1 0,-60,36,36,0.01,0.6 -60,-90,36,36,0.6,0.6 -90,0,84,84,0.18,0.06 -60,-90,60,36,0.06,0.06 -90,-30,36,36,0.01,0.06 2 -30,0,60,84,0.18,0.01 -30,0,36,84,0.18,0.01 0,0,60,36,0.6,0.01 -60,-30,60,36,0.01,0.01 0,0,36,84,0.01,0.01 3a 0,-90,60,36,0.01,0.6 -90,-60,36,84,0.6,0.01 -90,-60,60,84,0.18,0.01 -90,-30,36,84,0.18,0.01 -90,-30,36,60,0.6,0.01 3b 0,-90,36,84,0.18,0.6 -60,-90,36,36,0.6,0.6 -60,-90,36,36,0.18,0.6 -60,-60,36,84,0.18,0.18 -90,-90,60,36,0.01,0.18 3c -60,-90,84,60,0.01,0.18 0,-90,36,36,0.06,0.6 0,-90,36,36,0.01,0.6 0,-90,36,36,0.6,0.6 0,-90,36,36,0.18,0.6 4a -30,-30,36,36,0.01,0.6 -30,-30,84,60,0.01,0.6 0,-30,36,36,0.01,0.6 -30,-30,60,36,0.06,0.6 -60,-30,36,36,0.06,0.6 4b -30,0,36,84,0.18,0.6 -30,0,60,84,0.18,0.6 -60,-30,36,60,0.06,0.6 0,0,60,84,0.18,0.6 -60,-30,36,60,0.6,0.6 4c 0,-90,60,60,0.06,0.6 -30,0,60,60,0.18,0.6 -60,-30,84,60,0.18,0.6 -90,-30,36,84,0.01,0.6 0,-90,36,60,0.18,0.6 5 -30,-60,60,60,0.06,0.18 -60,-90,36,36,0.01,0.18 -30,-30,36,84,0.18,0.18 -90,-60,36,60,0.18,0.18 -30,-90,60,60,0.06,0.18 6 -30,-30,36,36,0.01,0.6 -60,-30,36,60,0.6,0.6 0,-30,36,36,0.01,0.6 -60,0,84,84,0.06,0.6 -90,-30,84,60,0.01,0.6 Configuration Cluster Patterns For ‘Case 2’ and ‘Case 3a’, the cumulative dissatisfaction (D CL ) and the Percentage Partially Satisfied (P ps ) values were plotted against representative (1 to 100) Configuration Indices (C index ), as shown (Fig. 75 and Fig. 76). For ‘Case 2’, P ps values were generally highest for C index =i+4n, i= {2,3} and lowest for C index =1,5,9,…,1+4n. From the D CL curve, it was observed that it is highest for C index =1,5,9…,1+4n, and generally lowest for C index =3,7,11…,3+4n. Figure 75: Cluster patterns for Case 2 For ‘Case 3a’, the P ps values were seen to be highest for C index =i+4n, i={2,3} and lowest for C index =5,9,13,..,5+4n. From the D CL curve, it was observed that D CL is highest for C index =1,5, 9,…,1+4n, and lowest for C index =3,7,11,…,3+4n. From both graphs, it was evident that the values of D CL and P ps are in opposite phases, and follow a periodicity of 4 owing to the sequence in which the configurations were generated. This feature helped to cluster configurations by their D CL and P ps values, and to recognize this pattern. These are discussed in the next chapter. Figure 76: Cluster patterns for Case 3a 87 4.12 Summary In Chapter 4, the results for each of the ten occupant preference cases were observed per point in time. However, these cases were cross-compared by case for all the points in time. A summary of the Maximum Percentage Reduction in Cumulative Dissatisfaction is provided in Chapter 5, Section 5.3. This is used to cross-compare the P r values across all the cases and points in time. This was done to understand how the P r value changes with a change in the preference case or point in time of the year. The summary table documenting all the P r values is as shown below (Table 40). Table 40: Cross-Comparison of Maximum Percentage Reduction (P r ) Value for all Nine Points in Time Point in Time 1 2 3a 3b 3c 4a 4b 4c 5 6 June 21 (9am) 3.03 3.90 4.14 15.37 4.06 9.32 4.88 10.14 4.66 10.32 June 21 (noon) 5.32 13.26 11.96 17.68 13.55 7.45 2.10 7.09 4.14 8.48 June 21 (3pm) 18.77 37.10 43.09 17.98 24.47 4.22 7.99 5.40 11.21 3.30 September 22 (9am) 54.75 69.30 73.26 52.59 62.63 20.82 27.18 24.78 48.15 17.32 September 22 (noon) 23.97 40.99 48.89 -2.40 32.53 15.13 9.86 4.85 13.26 12.59 September 22 (3pm) 42.26 55.31 64.53 45.29 50.92 13.48 15.33 15.86 32.61 14.59 December 22 (9am) -1.75 4.58 -5.55 22.09 3.01 7.60 3.62 7.99 3.50 9.88 December 22 (noon) 33.12 52.62 55.55 15.35 48.90 4.37 2.88 7.24 24.00 1.63 December 22 (3pm) 27.68 44.41 56.04 26.43 39.15 9.99 9.80 9.11 25.28 6.36 A discussion of the conclusions that can be made from this data and the implications of its findings in real-life design scenarios are discussed in Chapter 5. In addition, Chapter 5 also analyzes the data in two ways: the evaluation of overall visual comfort satisfaction and the analysis of façade configuration performance using the defined metrics in Chapter 3. Discussion of these two aspects are facilitated using a series of questions highlighted in Chapter 5, Section 5.1. A discussion of the most important assumptions and the scope is also made in Chapter 5, Section 5.2. 88 CHAPTER 5 5. DISCUSSION 5.1 Discussion Framework This chapter is divided into four parts: assumptions and expected findings, test case results and conclusions, tangential findings, and exploratory research. The first part discusses the different research assumptions, the scope, and the expected research findings. The second part, the test case results and the points of analysis, is structured into nine different sub-sections depending on the point in time under analysis. For each of the sub-sections, conclusions are made regarding observations made in Chapter 4. A summary section follows that discusses trends and patterns inferred from repeating conclusions. Each of the nine cases discuss the research framework for a given point in time. A subsequent section makes a cross- comparison of results, by occupant preference case, rather than by point in time. Certain useful conclusions are made more apparent. The third part discusses tangential findings and unanticipated results that include positive and negative outcomes, as well as outlier conditions, based on the hypothesis. Other conclusions that were not part of the initial research proposition are also discussed. This part references these additional tangential discussions to the fourth part of this chapter where they are discussed in greater detail. These are organized as sub-sections and include topics of discussion such as the potential advantages of dynamic facades in relation to their static counterparts, the effectiveness of using standard daylighting criteria in increasing overall visual comfort, advantages of split-control adaptive facades, potential uses and limitations of the proposed ‘Cumulative Dissatisfaction’ index and related indices. Possibilities for future work on aspects such as the real-world application of the methodology, cluster analysis of façade configurations based on their performance scores, as well as potential energy and cost benefits are discussed. The chapter ends with a summary of the key findings, conclusions, and new avenues of research. The results of each point-in-time have been highlighted in sections 4.3 to 4.12 in the form of tables. Some of these results were extracted in this chapter for further discussion. The framework for the discussion in this chapter is also divided and arranged in the same order as sections 4.3 to 4.12 of the previous chapter, as sections 5.3 to 5.11. Each of these sections highlight specific portions of the data from its corresponding section in chapter four. Each section, from 5.3 to 5.11, is structured using a table that extracts certain data information from questions. The following questions, derived from the research objectives, were answered for each point in time under broad headings: the evaluation of overall visual comfort satisfaction and the analysis of façade configuration performance. Evaluation of Overall Visual Comfort Satisfaction The first set of questions were posed to evaluate the first part of the hypothesis. These drew conclusions regarding the overall visual comfort of all occupants in the space using the proposed Maximum Percentage Reduction in D CL (P r ) metric. The DCL metric was used as a directly proportional measure of overall occupant satisfaction. The analysis also uses the proposed Percentage Partially Satisfied (Pps) metric to calculate the percentage of occupants who are partially satisfied and those who are not. Both these values are discussed for the ten different occupant preference cases. The ten constructed occupant preference cases determine how the adaptive façade control logic might accept user visual comfort preferences. For example, the control system might not allow a higher preference for any particular visual comfort factor. In such a case, Case 1 would be followed, as described in Chapter 4. In another condition, if occupants can fully describe their preferences, the condition might be better represented by Case 5. Another example is Case 3a, which is a preference model that is more biased towards illuminance satisfaction than glare or contrast ratio satisfaction. It was seen that these different cases produce different overall visual comfort satisfaction results, which necessitates the choice of the right occupant preference model. Ideally, occupants can have any set or combination of preferences within the defined ranges. In the absence of an occupant preference due to an absent occupant or a missing preference input, or when the system does not provide complete preference selection and needs to fix one or more preference or importance factor by default, or when the system itself suggests favorable preferences to the occupants, this comparison of occupant preference model might be useful. This problem is explored in this section using the results as supporting data. For each question, the broader implications of the findings have been discussed using the results in the previous chapter as starting points. 89 • How does the P r differ for the different cases? What does a ranking of the Pr values signify? • Which cases produce the highest and lowest P r value? What are the possible reasons for this? • Cross-comparing, which points in time feature the highest and lowest Pr values? What is the significance of this? • Which cases consistently perform better in terms of their P r value, and which cases perform unpredictably? • What is the significance of the P r values for the ideal occupant preference case (Case 1)? • What is the significance of the P r values for the random occupant preference case (Case 5)? • Comparing cases 3a, 3b, and 3c, what is the effect of changing the preferred illuminance set-point on the P r value? • Comparing cases 4a, 4b, and 4c, what is the effect of changing the contrast-ratio set-point on the P r value? • Comparing cases 2, 3a, and 4a, what is the effect of changing importance factors on the P r value? • How does the P ps value differ for the different cases? Does it have a correlation with the P r value? • How do these findings affect the proposed adaptive façade control methodology in a real-world scenario? • Can dynamic facades increase overall visual comfort in a multiple-occupancy condition? Analysis of Façade Configuration Performance An abridged version of the results and the plots of the performance scores against the configuration indices have been provided to facilitate the discussion. Certain values that are documented in Chapter 4 - Percentage Better Configurations (P config ), the periodicity, the control states of the best performing cases, and cluster patterns arising from closeness of performance scores of configurations. The following questions form the basis of the discussion in relation to the respective results documented in the previous chapter. • Which occupant preference case allows the highest P config ? • Is an occupant preference case having a higher P config value better as a design choice than one with a higher P r ? • Does the occupant preference case with the highest P r value also the one with the highest number of better performing configurations? Is there a correlation? • What is the distribution of the better-performing configurations among the 5% bands? What are the conclusions? • Which set of façade configuration states produce the highest P r value? • Which state governs the configuration performance the most? What is its relationship with its orientation? • Which state least governs the configuration performance? What is its relationship with its orientation? • What are the implications of adoption of either of the performance metrics on the designer’s choice and the performance of the configuration? • What were the cluster patterns that were observed in the data? • Can design options be largely predicted by analyzing the cluster pattern using the observed periodicity? • How are these façade configuration performance results affected by the façade orientation? 5.2 Assumptions The following is a list of the assumptions on which the thesis has been based. These have previously been introduced in Chapter 3, Section 3.8, and have been discussed in greater detail here. Preset Discrete Occupant Preference Ranges For each visual comfort factor, the choice of preference ranges available to the occupant are preset. For example, the glare preference ranges are fixed at discrete intervals providing ranges of 0.20-0.25,0.25-0.30,0.30-0.35,0.35-0.40, and 0.40-0.45. The occupant must choose from amongst these five options only, with 0.2-0.25 being the least perceptible glare to 0.40-0.45 as fringing on glare intolerance. A similar sub-divisional approach is applied to both illuminance as well as luminance contrast ratio. These sub- division ranges are based partly on the Illumination Engineering Society lighting standard recommendations, literature review findings, and assumptions, but these might not be adequate or appropriate for true modeling of occupant preference range requirements in the real world. For example, an occupant might prefer the range 0.225-0.25 much more than 0.20-0.25 range. The occupant’s preference can be wider or narrower than the preset range. For this work, equal preference divisions are used. Also, the preferred preference range might lie across two existing preference ranges. The preset visual comfort range sub-divisions in this research are equal and continuous, but a user preference might be spread over a larger or smaller area, might be a single point value or a discontinuous range, and might even lie outside standard recommendations for 90 visual comfort. In the latter case, it is even possible for an occupant preference to lie at either of the extreme ends of the comfort criteria scale, or might not have any upper or lower bounding value. The preferences and individual importance factors for multiple users even be the same and exactly equal. In these cases, the approach to modeling the user preferences would substantially change from the one. The purpose of simplified digital modeling of the user preferences and importance factors is to provide a set of values to test the framework and hypothesis. A more accurate modeling of preferences could essentially form the basis of future research. Flat Importance Factors To mathematically depict the importance of the any visual comfort factor for an occupant, importance factors were assigned on a scale of a 5-90, such that all the importance factors for an individual would add up to 100. As discussed previously, in a real-world scenario, the importance factors would be input by the user directly using a graphical control interface. The 5-90 range was chosen to provide occupants the highest possible flexibility of being assigned very low or very high importance factors by the random function. These are chosen randomly, just like occupant preferences are, for testing the hypothesis. This means that if an importance factor of 5 is chosen to depict, say the importance of an occupant’s personal preference towards resolving glare, and a 5, to depict the importance laid on contrast ratio preference, it means that the importance assigned to illuminance is automatically 90. This means that the proposed framework would try its best to produce illuminance values at the occupant’s task surface that are close to the preferred illuminance values, but will not try as hard to minimize glare or contrast ratio deviation. The deviation value is defined as the amount by which the calculated visual comfort parameter value differs from the preferred value. If the calculated value, such as illuminance, lies within an occupant’s preference range, a deviation value of zero is used to denote that the preference has been met. However, if the calculated value lies outside the preference range, the lesser difference between the bounds of the latter, and the former, is measured and is called a deviation value of illuminance, for example, for that occupant. For example, an illuminance preference range of 300- 450 lx compared against a measured value of 500 lx will give a net deviation of 50 lx (500-450), whereas a measured value of 350 lx will produce only a deviation value of zero denoting the that the preference has been met. Two occupants in the test space might have completely different set of preferences and importance factors. For example, assume one occupant has chosen an importance factor of 30 each for glare, illuminance, and contrast ratio, while another occupant chooses factors of 5,5, and 90 for the same visual comfort factors. The proposed framework, while looking for the most generally acceptable solution, would not prioritize minimizing glare deviation for the second occupant, as the associated glare importance factor is low. However, for the first occupant, the system will try to equally prioritize glare, illuminance, and contrast ratio. Having the importance factors sum up to 90 means that they can be distributed equally, thus signifying equal importance. The assigned importance factors are unweighted. This means that they are all equally important. In a more statistically accurate model, the importance factors themselves would have individual weights to denote differences in importance factor values. Another factor that is not within the scope of this research is occupants having different importance themselves. For example, in an office space, the effort to provide a certain section of users or spaces more control over their environment might be higher or lower than others. This might be because of the nature of activity performed by that section of occupants or simply because of a design requirement. In such cases, additional importance factors would need to be assigned to groups or individuals. Linear Deviation and Dissatisfaction Criteria The deviation values calculated for each visual comfort factor for a given occupant were consolidated to represent a net deviation value. However, each of the deviation values for glare, illuminance, and contrast ratio have different units and therefore need to be remapped to make the net deviation value more meaningful. For example, for an occupant the glare deviation might be 0.10 whereas the illuminance deviation might be 400 lx. The two deviations represent two different units- the glare deviation being unitless whereas the illuminance deviation is measured in lux. Assuming importance factors of 20 and 50 for these two factors, it is seen that the resulting weighted deviations are farther apart from each other and so is the net deviation value. To lessen the problem due to the range of units and to make all the values unitless, each deviation value is divided by a constant factor, assumed to be the range subdivision length. Therefore, the deviations for glare, illuminance, and 91 contrast ratio are divided by their range subdivision value: 0.05 for glare, 0.33 for contrast ratio, and 150 lx for illuminance respectively, and multiplied by their respective importance factors, divided by a common selected factor of 300. The calculated value is still quite large, which is why a square root is performed to scale down the value. This provides smaller normalized unitless values. Each represents the respective scaled deviation for glare, illuminance, and contrast ratio from the simulated values. When added together these become the net scaled deviation value that represents the occupant’s visual dissatisfaction. A scaled deviation of zero represents the least dissatisfied and is the best possible case. It is assumed the visual comfort dissatisfaction score is directly proportional and linearly related to the absolute deviation from set preferences. Therefore, a change in the deviation might not represent a proportional change in the occupant’s extent of dissatisfaction. In fact, a change in deviation might even adversely affect the occupant’s dissatisfaction. For example, assuming an occupant prefers an average task surface illuminance of 600-750 lx but is provided with 900 lx in one case and a 1000 lx in another case, then as per the assumptions of this research, the case providing a 900 lx would be more favorable to the occupant than the 1000 lx case. In a real situation, there can be several scenarios that arise out of these requirements. The occupant might not prefer 900 lx as much as a 1000 lx, in which case the calculated illuminance deviation would be higher for 900 lx than for 1000 lx. Another possibility is that the occupant might prefer values that are lower than the set preference and none which are above it. In that case, the values exceeding the set preference would not be considered as valid solutions. In yet another case, the extents of dissatisfaction resulting from a 100 lx deviation and a 200 lx deviation might not be in the ratio of one to two, as a 200 lx deviation might cause greater or less than two times the dissatisfaction resulting from a 100 lx change. Cumulative Deviation The numerical deviation of each measured comfort factor from its preferred range is therefore not by itself proportionally representative of occupant dissatisfaction. However, as explained, a linear proportional relationship of absolute deviation to dissatisfaction has been assumed. This provides three independent sets of absolute deviations for glare, illuminance, and contrast ratio. Each of these contribute to the cumulative deviation that causes an overall dissatisfaction that the occupant feels. It is assumed that the cumulative deviation is calculated using an addition of the individual deviations, each weighted by its importance factor. The proposed dissatisfaction value when computed for each occupant, and then cumulated for the entire room to give an idea of the overall dissatisfaction, results in several possible scenarios when compared with respective values of a baseline case. One possibility is the ideal case, resulting in all occupants having had their preferences met, thus producing deviation and dissatisfaction values of zero. These cases are represented in green and would mean everybody is a hundred percent satisfied. However, the chances of every person having had their preferences met simultaneously is very low, as discussed in chapter four and five. Thus, such a case, though theoretically possible, might not actually occur in the test cases or in actual operation, unless there are certain special circumstances or climatic conditions at that point in time. In another possible case, some or all the occupants might have their preferences met while other occupants can be said to be partially satisfied compared to the baseline, following the assumption that dissatisfaction is directly proportional to the calculated deviation. These are represented in yellow to indicate the presence of a wide band of people who fall in between the unsatisfied and fully satisfied groups. A third possible case might have some or all occupants who are dissatisfied with their current conditions when compared to their baseline conditions. These people are red-flagged indicating that the net effect of personal satisfaction and dissatisfaction with glare, illuminance, and contrast ratio measured at their workstation and compared to their baseline conditions, has created an overall dissatisfactory visual condition. As hinted, a red flagged occupant might not be completely dissatisfied with the visual condition but rather might be more biased in his judgement towards a more important and underperforming factor, that in his opinion affects their entire sense of satisfaction negatively. Occupant Behavior Occupant behavior is stochastic in nature. This means that any decision that is taken based on occupant behavior has a certain degree of unpredictability. The comfort experienced by an occupant in a space is not simply affected by comfort factors but also by other factors such gender, physiology, clothing level, metabolic level, personal preferences, motivations, mood, and a host of other factors that do not necessarily relate to the immediate conditions in the 92 occupant’s physical environment. In this research, the scope of applicability of occupant behavior has been limited to a simplified occupant preference model for visual comfort only. In a real-world situation, other factors such as occupancy, occupant movement within the space, physical interaction with the system, or integration with other building systems through the building management system would also be necessary. In such a case, an optimized façade configuration considering visual comfort alone considering all these assumptions might not be the best possible case when considering aspects such as thermal comfort, energy use, cost, or others. 5.3 Cross-Comparisons In Chapter 4, the results were collected per point in time. Of primary interest is the Maximum Percentage Reduction (P r ) value. In this section, the P r values are cross-compared by occupant preference case to identify important relations and conclusions. It was observed that cases 3a and 3b outperformed all the other cases over the results from the nine points in time and as marked (Table 41). Cases 3a and 3b produced significantly higher P r values. Table 41: Cross-Comparison of Maximum Percentage Reduction (P r ) Value for all Nine Points in Time Point in Time 1 2 3a 3b 3c 4a 4b 4c 5 6 June 21 (9am) 3.03 3.90 4.14 15.37 4.06 9.32 4.88 10.14 4.66 10.32 June 21 (noon) 5.32 13.26 11.96 17.68 13.55 7.45 2.10 7.09 4.14 8.48 June 21 (3pm) 18.77 37.10 43.09 17.98 24.47 4.22 7.99 5.40 11.21 3.30 September 22 (9am) 54.75 69.30 73.26 52.59 62.63 20.82 27.18 24.78 48.15 17.32 September 22 (noon) 23.97 40.99 48.89 -2.40 32.53 15.13 9.86 4.85 13.26 12.59 September 22 (3pm) 42.26 55.31 64.53 45.29 50.92 13.48 15.33 15.86 32.61 14.59 December 22 (9am) -1.75 4.58 -5.55 22.09 3.01 7.60 3.62 7.99 3.50 9.88 December 22 (noon) 33.12 52.62 55.55 15.35 48.90 4.37 2.88 7.24 24.00 1.63 December 22 (3pm) 27.68 44.41 56.04 26.43 39.15 9.99 9.80 9.11 25.28 6.36 Occupant preference cases are ranked based on decreasing order of their Maximum Percentage Reduction value (Table 42). It is observed that cases 2, 3a, and 3b are generally the top three cases that produce the highest Maximum Percentage Reduction values. Specifically, Case 3a and 3b are observed to perform much better than Case 2 in this respect. Table 42: Cross-Comparison of Maximum Percentage Reduction (P r ) Value for all Nine Points in Time Point in Time #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 June 21 (9am) 3b 6 4c 4a 4b 5 3a 3c 2 1 June 21 (noon) 3b 3c 2 3a 6 4a 4c 1 5 4b June 21 (3pm) 3a 2 3c 1 3b 5 4b 4c 4a 6 September 22 (9am) 3a 2 3c 1 3b 5 4b 4c 4a 6 September 22 (noon) 3a 2 3c 1 4a 5 6 4b 4c 3b September 22 (3pm) 3a 2 3c 3b 1 5 4c 4b 6 4a December 22 (9am) 3b 6 4c 4a 2 4b 5 3c 1 3a December 22 (noon) 3a 2 3c 1 5 3b 4c 4a 4b 6 December 22 (3pm) 3a 2 3c 1 3b 5 4a 4b 4c 6 It was also observed (Table 42) that apart from the top three ranks (usually cases 2, 3a, and 3c), the remaining cases do not follow a pattern in how well they perform based on their P r values. Case 3b was observed to feature in the top three only for two points in time. Similar observations can be made from the following plots of the Maximum and Median Percentage Reduction (P r ) in Cumulative Dissatisfaction values for each of the nine points in time. The maximum values of P r might provide an exaggerated distribution of P r values per case, which is why the median and mean P r values are also plotted to have a general idea of the distribution (Fig. 77 and Fig. 78). From these plots, it was also observed that the P r value is usually highest for September 22 (9am) and September 22 (3pm), and lowest for June 21 (9am), June 21 (noon), and December 22 (9am). Intermediary values were observed for June 21 (3pm), September 22 (noon), December 22 (noon), and December 22 (3pm). 93 Figure 77: Maximum Percentage Reduction in DCL (Pr) for the nine points in time Figure 78: Median Percentage Reduction in DCL (Pr) for the nine points in time 5.4 Evaluation of Overall Visual Comfort Satisfaction An important conclusion from the data (Table 5.4.1 to Table 5.4.9) is that using real-time occupant preference-based dynamic façade controls can outperform static facades in terms of increasing overall visual comfort. Considering occupants to have differing preferences and visual comfort factor importance factors, rather than equal ones, can make the adaptive façade control system perform better. Both the observations support the idea that an adaptive façade controlled in real-time using a constantly updating set of occupant visual comfort preferences and importance factors, can outperform static facades whose design is based on a fixed set of uniform and standard visual comfort requirements. The proposed framework is also seen to increase the ease and efficiency of the control logic in selecting viable façade configurations as multiple solutions fulfill the same visual comfort criteria, thus allowing flexibility in choice based on other design factors such as aesthetics, ease of fabrication, and cost reduction. The proposed framework can be used to find these equally well-performing façade configuration solutions so that designers can have the option to choose which ones to use. It is also observed that occupant preference models that are positively biased towards meeting illuminance requirements perform better than models biased towards meeting glare requirements. 94 Table 43: Relevant Summary of Results for June 21, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 37.90 20.92 26.81 47.81 28.88 43.90 40.52 41.69 38.71 46.73 Maximum Percentage Reduction 3.03 3.90 4.14 15.37 4.06 9.32 4.88 10.14 4.66 10.32 Max. Percentage Partially Satisfied 81.25 68.75 75.00 100.0 81.25 100.0 75.00 93.75 75.00 100.0 Table 44: Relevant Summary of Results for June 21, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 38.53 22.27 28.92 45.04 41.56 41.96 39.25 39.80 38.33 44.47 Maximum Percentage Reduction 5.32 13.26 11.96 17.68 13.55 7.45 2.10 7.09 4.14 8.48 Max. Percentage Partially Satisfied 81.25 87.50 68.75 100.0 93.75 93.75 68.75 93.75 81.25 100.0 Table 45: Relevant Summary of Results for June 21, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 44.16 28.35 39.24 38.89 34.26 39.18 40.68 38.12 39.52 39.97 Maximum Percentage Reduction 18.77 37.10 43.09 17.98 24.47 4.22 7.99 5.40 11.21 3.30 Max. Percentage Partially Satisfied 100.0 100.0 100.0 100.0 68.75 87.50 93.75 87.50 81.25 87.50 Table 46: Relevant Summary of Results for September 22, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 76.63 56.97 81.85 75.97 76.23 46.46 49.02 46.20 67.46 46.43 Maximum Percentage Reduction 54.75 69.30 73.26 52.59 62.63 20.82 27.18 24.78 48.15 17.32 Max. Percentage Partially Satisfied 100.0 100.0 100.0 75.00 81.25 81.25 81.25 75.00 87.50 75.00 Table 47: Relevant Summary of Results for September 22, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 47.30 31.34 44.82 36.32 38.97 38.21 40.47 36.90 41.08 38.53 Maximum Percentage Reduction 23.97 40.99 48.89 -2.40 32.53 15.13 9.86 4.85 13.26 12.59 Max. Percentage Partially Satisfied 87.50 100.0 100.0 68.75 68.75 56.25 75.00 62.50 62.50 56.25 Table 48: Relevant Summary of Results for September 22, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 61.94 41.70 63.18 64.80 60.94 44.88 47.13 43.96 58.83 45.40 Maximum Percentage Reduction 42.26 55.31 64.53 45.29 50.92 13.48 15.33 15.86 32.61 14.59 Max. Percentage Partially Satisfied 81.25 100.0 87.50 100.0 75.00 68.75 81.25 75.00 75.00 81.25 Table 49: Relevant Summary of Results for December 22, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 37.77 20.91 24.91 48.82 28.28 45.50 40.68 42.75 38.75 48.50 Maximum Percentage Reduction -1.75 4.58 -5.55 22.09 3.01 7.60 3.62 7.99 3.50 9.88 Max. Percentage Partially Satisfied 56.25 75.00 50.00 100.0 68.75 87.50 62.50 81.25 68.75 93.75 Table 50: Relevant Summary of Results for December 22, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 57.32 39.36 59.47 53.87 55.70 39.75 42.53 38.76 52.01 39.56 Maximum Percentage Reduction 33.12 52.62 55.55 15.35 48.90 4.37 2.88 7.24 24.00 1.63 Max. Percentage Partially Satisfied 93.75 100.0 100.0 75.00 75.00 56.25 75.00 68.75 68.75 56.25 Table 51: Relevant Summary of Results for December 22, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Baseline Cumulative Dissatisfaction 55.88 36.91 53.33 52.10 51.49 44.86 41.71 41.73 51.26 44.44 Maximum Percentage Reduction 27.68 44.41 56.04 26.43 39.15 9.99 9.80 9.11 25.28 6.36 Max. Percentage Partially Satisfied 100.0 100.0 100.0 81.25 62.50 81.25 75.00 81.25 68.75 81.25 It was observed that a high P r value (Cases 3b, 4a, 4c, and 6) is associated with much higher Maximum Percentage Partially Satisfied (P ps ) and Percentage Better Configurations values. 95 Discussion- Part A As seen in Fig 77 and Fig. 78, the P r value is highest for Cases 2, 3a, and 3b. A ranking of the P r values signify that the top three P r values are produced only by Cases 2, 3a, 3b, and 3c. Case 1, the ideal case, showed no trend or predictability in ranking, though it did not feature among the top three for any of the data sets for the nine points in time of the year. Case 5, the random case showed a similar ranking characteristic. The first conclusion from these observations for Case 1 is that treating all occupants as equal in terms of visual comfort preferences and using standard daylighting criteria (DGP=0.20, E H =500lx, Contrast ratio= 1.0) can limit the overall visual comfort performance of adaptive facades. This reinforces the need to adopt a non-uniform occupant preference model to control such facades. As Case 5 represents a purely random preference case, its low performance implies that allowing occupants to set any preference and importance factors can prevent the adaptive façade system from substantially reducing the overall visual discomfort. Cases 3a, 3b, and 3c, and Case 2 were observed to perform much better than the other cases (Fig. 77 and Fig. 78). From the high performance of cases 3a, 3b, and 3c, it can be concluded that when all occupants assign a high importance to illuminance simultaneously, it is easier to increase their overall visual comfort satisfaction level, when compared to the situation when all occupants assign a high importance to glare. It can also be observed that it is generally easier to satisfy illuminance preference ranges and setpoints that are closer to standard daylighting criteria (DGP=0.20, E H =500lx, Contrast ratio= 1.0) than it is for, say, an E H of 2000lx or even a random horizontal illuminance preference demand. A further conclusion of this observation is if the solar condition demands a high priority to illuminance and standard horizontal illuminance levels (E H =500lx). At the end of the day or under cloudy sky conditions, it might be easier to partially meet this preference demand and make the overall room occupants happier using this methodology, than it would be for a situation demanding a high-priority for glare control. The lowest P r values are produced by the preference cases that demand a high bias towards contrast ratio or the one with equal occupant preferences (Case 1). A probable reason for this might be that the adaptive façade cannot be used to regulate contrast ratio as easily as it can be used to reduce glare or illuminance. It was also seen that 9am and 3pm on September 22 consistently show the top two highest P r values among the other points in time under study and for almost all the preference cases. It might be intuitive to believe that all the points in time would behave similarly in terms of increasing overall visual comfort. However, as mentioned, it was apparent that the overall visual comfort of all the occupants could be increased to greater degree at 9am and 3pm on September 22. It was also observed that the Partially Percentage Satisfied (P ps ) value generally increased with an increase in P r . This observed correlation implies that if a greater number of occupants are partially satisfied, it could increase the overall satisfaction of the room, when compared to the case when only a few occupants have their visual comfort preferences met while others are not satisfied. Therefore, to make the entire room occupants happier, there is a need to adopt an adaptive façade control strategy that will try to maximize the number of people who are happier, even if by small amounts, than meet the preferences of fewer occupants. 5.5 Façade Configuration Cluster Patterns The proposed framework can be used to find these equally well-performing façade configuration solutions so that designers can have the option to choose which ones to use. It is also observed that occupant preference models that are positively biased towards meeting illuminance requirements perform better than models. The following results summarize the façade configuration performance indices and score. Scatterplots are used to plot the Percentage Reduction (P r ) and the Performance Score (C sc ) for all the 2304 configurations. These plots helped recognize clusters of closely-performing configurations. From a design application standpoint, each configuration of such a cluster is equally good as any other configuration in that cluster. This finding can be used to improve adaptive façade control strategy by letting the system predict the possible outcomes of certain configurations. Table 52: Relevant Summary of Results for June 21, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 1.82 1.30 0.35 13.85 0.74 15.49 5.69 14.84 4.21 16.02 Configuration Index (best) 89 1421 265 597 233 1629 2165 941 209 1921 Configuration Performance Score 31.31 27.69 32.24 28.37 28.58 30.96 32.50 34.95 24.91 28.69 96 Figure 79: Scatterplot of Percentage Reduction & Performance Score against Configuration Index, Case 2, June 21 (9am) Figure 80: Scatterplot of Percentage Reduction & Performance Score against Configuration Index, Case 3a, June 21 (9am) Table 53: Relevant Summary of Results for June 21, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 5.21 10.03 5.82 15.36 12.37 13.98 3.65 11.46 5.03 13.80 Configuration Index (best) 1481 1869 2170 2049 405 797 1177 1029 1317 1221 Configuration Performance Score 32.56 32.58 24.37 34.03 40.78 35.01 33.22 38.81 32.85 33.14 Figure 81: Scatterplot of Percentage Reduction & Performance Score against Configuration Index, June 21 (noon) 97 Figure 82: Scatterplot of Percentage Reduction & Performance Score vs. Configuration Index, Case 3b, June 21 (noon) Table 54: Relevant Summary of Results for June 21, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 37.59 61.89 62.07 13.41 19.44 21.31 22.66 23.70 11.98 19.88 Configuration Index (best) 233 793 281 645 494 1417 69 1057 1269 1653 Configuration Performance Score 29.68 37.08 33.07 28.77 31.08 40.85 46.24 29.17 34.11 36.25 Figure 83: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 84: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters 98 The following results were consolidated for September 22, 9am. Table 55: Relevant Summary of Results for September 22, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 81.08 92.80 88.80 89.80 90.06 23.05 53.43 36.59 66.58 17.71 Configuration Index (best) 2045 1733 573 1585 2057 610 1245 101 2181 710 Configuration Performance Score 25.09 11.92 37.20 14.29 30.65 45.67 35.84 53.90 21.44 49.79 Figure 85: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 86: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters The following results were consolidated for September 22, noon. Table 56: Relevant Summary of Results for September 22, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 34.81 70.44 73.39 0.00 25.04 0.26 14.67 3.69 21.31 0.39 Configuration Index (best) 365 316 2189 N/A 874 373 1509 1313 2001 389 Configuration Performance Score 37.83 40.73 24.80 N/A 31.50 21.58 36.63 31.75 29.57 24.66 99 Figure 87: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 88: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters The following results were consolidated for September 22, 3pm. Table 57: Relevant Summary of Results for September 22, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 65.93 92.36 92.66 70.01 94.40 2.60 17.14 7.25 52.21 3.60 Configuration Index (best) 973 1633 1057 1173 1593 2045 1305 1337 1379 745 Configuration Performance Score 26.64 27.59 21.98 29.61 16.29 29.86 43.21 34.75 36.14 29.67 Figure 89: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters 100 Figure 90: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters The following results were consolidated for December 22, 9am. Table 58: Relevant Summary of Results for December 22, 9am Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 0.00 3.26 0.00 21.83 0.22 13.54 5.95 10.98 3.60 14.93 Configuration Index (best) N/A 681 N/A 857 369 1065 1185 417 461 1081 Configuration Performance Score N/A 30.51 N/A 27.75 21.58 29.55 38.81 46.31 31.89 31.83 Figure 91: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 92: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters 101 The following results were consolidated for December 22, noon. Table 59: Relevant Summary of Results for December 22, noon Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 58.90 87.20 85.72 31.99 77.47 1.82 1.56 1.95 42.53 1.26 Configuration Index (best) 1605 1909 1461 2141 1349 701 973 949 2170 377 Configuration Performance Score 25.80 26.01 22.38 31.28 23.78 21.40 34.49 31.63 33.29 41.38 Figure 93: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 94: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters The following results were consolidated for December 22, 3pm. Table 60: Relevant Summary of Results for December 22, 3pm Case 1 2 3a 3b 3c 4a 4b 4c 5 6 Percentage Better Configurations 62.24 73.13 92.49 33.20 68.40 11.11 6.08 12.07 33.94 8.07 Configuration Index (best) 301 664 493 469 1710 733 613 505 938 733 Configuration Performance Score 28.43 35.76 34.18 31.99 17.52 32.82 31.16 28.62 36.57 30.18 102 Figure 95: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters Figure 96: Scatterplot of Percentage Reduction & Performance Score against Configuration Index showing clusters 5.6 Summary Chapter 5 discussed the implications of the results. It was concluded that the overall visual comfort satisfaction of occupants can be substantially increased using adaptive facades and a control logic that allows occupants to set their own comfort preferences and importance factors. It was also inferred from the data that it is easier for the control logic to arrive at a more acceptable and satisfying solution set of visual comfort values if occupants set more importance to illuminance than glare or contrast ratio. In the absence of an occupant or when the occupant does not know which set of visual comfort values are ideal at that time for their particular case, the system or control logic can suggest such illuminance-preferred solutions. It was also found that for particular times of the year, the system performs more efficiently allowing higher cumulative visual comfort satisfaction. In terms of the P ps and P r values, it was found that by increasing overall comfort, the individual comfort may or may not always show an increase. Also, it was observed that the P ps value is usually lesser than the P r value, which means that is much easier to increase the overall visual comfort, such that some people are happy and some unhappy, than it is to make everyone happier by small or large amounts. It also means that there are usually some percentage of people who are not satisfied. In terms of façade performance, it was observed that multiple façade configurations perform equally well in terms of increasing overall visual comfort. Recognition of these clusters can assist the designer in the pre-design and design development phases by suggesting alternatives which can then be evaluated based on other factors such as design intent, cost, and rationalization. 103 CHAPTER 6 6 CONCLUSION 6.1 Conclusion Framework This chapter is divided into two sections- tangential findings from the research and future work. The first section lists the tangential findings. These include positive and unanticipated results. During the research, several new ideas were formed which have led to creation of additional hypotheses. These are highlighted in Section 6.2. Each of these create additional avenues of research and possibilities of future work that are described in Section 6.3. 6.2 Tangential Findings The most important contribution of was to prove that controlled operation of adaptive facades can in fact increase the overall occupant visual comfort in a multiple occupancy space in comparison to static facades. The second most important contribution was to propose a simplified method to consolidate and quantify multiple visual comfort metrics based on an input model of occupant visual comfort preferences and importance factors. The idea for this method stemmed from the problem of sometimes contradictory singular visual comfort factors in determining the actual visual comfort condition in a space and the occupant response to it. This created the need to understand the overall effect of multiple visual comfort factors. It was assumed that this combined effect can only be evaluated based on the response it receives from individual occupants. It was also assumed that the response received from each occupant would be a difference of the visual condition that they perceive and the one that they prefer. This created the need to consolidate and quantify multiple visual comfort metrics into a single value, the Individual Dissatisfaction (D i ) using individual- specific visual comfort preference model. By its definition, this proposed metric is relative to the individual it is calculated for and dependent on their personal set of assigned preferences and importance factors. The other positive result was to combine the calculated individual dissatisfaction values to inform a quantification of the cumulative effect of the individual dissatisfaction values. Though this was implemented through a direct aggregation of values forming a Cumulative Dissatisfaction value, though in a real-case scenario the difference in occupant ranks in a given space might necessitate the use of a weighted system to calculate the overall dissatisfaction. However, this overall dissatisfaction value fails to determine how many occupants are satisfied. Recognizing the need to also quantify this and assuming that it is not possible to fully satisfy the visual comfort condition of all occupants in a space at a given time, another important metric was introduced called the Percentage Partially Satisfied (P ps ). An important contribution was also to propose the idea of ranking of adaptive façade configurations based on their performance in satisfying overall visual comfort of occupants in a space. A necessary bridging step was to quantify adaptive façade configuration performance. A method was proposed to quantify the façade configuration performance using a mathematical combination of the Maximum Percentage Reduction and Percentage Partially Satisfied values. Having the performance values of façade configurations that represent how well they can increase overall visual comfort of maximum occupants in a Multiple Occupancy Condition, it was possible and necessary to rank façade configurations. The two important reasons to do were to assist designers in façade configuration selection as well as to increase façade control efficiency by using higher performing configurations instead of low performing configurations. Façade configuration ranking also opened a new path to understand which individual façade control states contribute to high performance scores. A relation of the control state, its contribution to the overall configuration performance score, and the solar orientation in which it is used, was also recognized. The last contribution, perhaps one of the most important, was to recognize clusters in the performance values calculated for façade configurations. It was concluded that not all adaptive façade configurations perform equally well, but it was also evident that most of these configurations can be categorized and clustered under different performance ranges depending on their performance scores. Each of the configurations in these observed clusters perform equally well. This was recognized to be a method to facilitate design selection and façade operation. A benefit of the observed clustering technique is possible use as a decisive factor by designers in selection of façade configurations when multiple criteria are involved. In a contrasting scenario, two equally performing and satisfactory façade configurations maybe selected based on a third criterion, such as the quality of outdoor view or thermal performance. This can add more flexibility to the designer as well as to the building operation manager. This added flexibility also allows the user to select or prefer a configuration when presented with several equally performing configurations. This opens a yet another avenue of research which would tackle the problem of resolving multiple occupants’ preferences for façade states based on preference criteria such as the quality of outdoor view and the façade aesthetics. This would be in 104 addition to resolving the problem of overall visual comfort. Yet another application of this clustering phenomenon might be in predictive modeling wherein the system would know which states will perform equally. In doing so, the real-time need for computational resources needed to find the right configuration can be substantially diminished. Predictive modeling would also allow predictive energy analysis and this could be used to predict, for example, the lighting control and energy use that would be needed at some point of the working day. Also, based on the clustering, it would be possible to predict how certain façade configurations will behave. This would aid in predictive modeling. Designers can use this in a reverse technique to position occupants or design the interior layout in a way that allows the chosen configurations to work the best. 6.3 Future Work Future work possibilities can be of two types: it could either be an improvement over the current methodology, or it could refer to larger issues that broaden the research scope. The following sections discuss these under two separate headings. 6.3.1 Improvements to Methodology Several assumptions were made while devising the research methodology (Chapter 5.2). In future work, some of these assumptions can be replaced by proven techniques or established methods. The nature of most of these assumptions are statistical simplifications as listed below. Preset Discrete Occupant Preference Ranges For each visual comfort factor, the choice of preference ranges available to the occupant are preset. For example, the glare preference ranges are fixed at discrete intervals providing ranges of 0.20-0.25,0.25-0.30,0.30-0.35,0.35-0.40, and 0.40-0.45. The occupant must choose from amongst these five options only, with 0.2-0.25 being the least perceptible glare to 0.40-0.45 as fringing on glare intolerance. A similar sub-divisional approach is applied to both illuminance as well as luminance contrast ratio. Real measured preferences were not used although in a real-world, occupants would choose their preferences from the predefined list. Rather, the preference sets for all the sixteen occupants were modeled virtually using a Python script to selecting preferences in a normal distribution from the pre- defined lists. A future work possibility is to allow the occupant to input a discrete or range of values for multiple visual comfort factors. Flat Importance Factors An improvement over the current set of importance factors could be by using weighted factors differing for different occupants. This would allow a more statistically accurate user modeling. Linear Deviation and Dissatisfaction Criteria It was assumed that the dissatisfaction with a particular set of visual comfort factors is directly proportional to the its relative deviations of the measured values from the preferred values. In reality, this might is usually not the case as the user might just prefer a certain range of values and not accept any other values. In that case, all values outside the preferred range would be equally undesirable. This aspect can also be improved in future work using more accurate user-modeling and statistical techniques. Cumulative Dissatisfaction The main research objective has been to minimize the total visual comfort dissatisfaction of all occupants combined. However, all occupants might not be equally important. In addition, the cumulative dissatisfaction criterion does not help us regulate the individual dissatisfactions. A future work possibility would be to also allow control and regulation of individual visual comfort satisfaction. 6.3.2 Research Possibilities Section 6.3.1 discussed the future work that could stem from improvements made to the methodology. Section 6.3.2 describes the larger issues and possibilities that broaden the research scope. Split-Control Adaptive Facades The split control of facades is theoretically recognized as a way to increase the effectiveness of adaptive facades and daylighting systems. It should be able to more easily increase the overall visual comfort and to provide added daylighting control by using a much larger number of possible façade configurations. By splitting a façade into more number of independently controlled parts, the daylighting and indoor visual comfort condition can be controlled more 105 accurately. This idea can be extended to, for example, the use of pixelated-shading systems with gradient control, such that the pixels change orientation with changing solar conditions. Such a pixelated system could be used to control direct glare by interception or perform selective shading or daylighting in a space. The advantage of such a system is the extensive control it provides resulting from the divisional approach to designing adaptive facades. Another alternative is the creation of such pixelated systems using electrochromic glazing technology instead of actual mechanical shades. Such split systems with large number of controls would be difficult to control using the proposed methodology because of the huge computational requirement. This might need the use of divide and conquer or other efficient algorithms. Another important avenue of future work is the use of machine learning to solve this problem. Potential Energy-Use and Cost Benefits The effects of choice in performance-based selection of façade configuration can be affected by introduction of other assessment metrics, such as cost, energy-use, and lighting energy use. Cumulative Visual Comfort Metric The proposed cumulative visual comfort metric is simplified and makes certain assumptions, as discussed in the previous chapter. However, the idea of using user preferences to form a consolidated comfort metric can be applied to visual as well as thermal comfort metrics. A study of how accurately this new metric represents the visual comfort condition of an occupant can be studied using surveys involving real occupants. This study can also compare the effectiveness of this proposed metric against established standard daylighting metrics and criteria in determining, for example, the overall or individual visual comfort condition. Cluster Analysis Performance-based cluster analysis of façade configurations can be studied separately and in greater depth using other performance metrics. Such cluster analysis can help discover more efficient solution sets. This avenue of research could also consider performance-based ranking, optimization, and prediction of façade configurations and control states using real-time weighted occupant preferences. Location In future works featuring studies across multiple test locations with varying climates, this layout or a more typical office layout might be tested in similar physical and urban contexts. Though more test locations could be adopted, their inclusion in this research was not feasible for the limitation in time. This restricts the applicability of this work to Los Angeles and other locations featuring similar climatic conditions. A future work possibility would be to conduct the research in other climates. Control Interface Development In a real-world application of the proposed control logic, a user control interface would be required to record user preferences and importance factors which would be input into the façade controller. Visual Comfort, Thermal Comfort, and Visual Amenity Factors In future work, other factors such as the quality of outdoor view, thermal comfort, and the amount of circadian stimulus provided by a daylighting system can also be studied. Summary The proposed research methodology has been able to validate the hypothesis. While implementing the first proposed methodology, there have been several revisions and additions to the research objectives. Several new avenues of research were also discovered. The primary objective of validating that adaptive facades can in fact increase the overall visual comfort of occupants in an office space in a climate like Los Angeles, has been met. The other objectives of combining multiple visual comfort criteria into one based on occupant preference models was also met. Other objectives, which include the development of a technique for clustering and ranking of adaptive façade configurations based on how well they perform, the establishment of visual comfort-based adaptive façade performance metrics, and the resolution of visual comfort in a multi-occupancy situation were met and discussed extensively. In addition, application of this methodology in the façade design process and use a façade design option tool has also been considered and discussed. Several future work possibilities were discussed that include improvements to the existing methodology as well as new avenues of research. It is possible to increase overall visual comfort in a multiple occupancy condition using composite adaptive façade daylighting systems. 106 References Aelenei, Daniel, Laura Aelenei, and Catarina Pacheco Vieira. 2016. “Adaptive Façade: Concept, Applications, Research Questions.” In Energy Procedia, 91:269–75. doi:10.1016/j.egypro.2016.06.218. Apte, J, and Dariush Arasteh. 2006. “Window-Related Energy Consumption in the US Residential and Commercial Building Stock.” Buildings, 1–38. Bakker, L. G., E. C M Hoes-van Oeffelen, R. C G M Loonen, and J. L M Hensen. 2014. “User Satisfaction and Interaction with Automated Dynamic Facades: A Pilot Study.” Building and Environment 78: 44–52. doi:10.1016/j.buildenv.2014.04.007. Bülow-Hübe, Helena. 2007. “Solar Shading and Daylight Redirection Demonstration Project for a System of Motorised Daylight Redirecting Venetian Blinds and Light Controlled Luminaire Energy and Building Design.” http://www.lth.se/fileadmin/energi_byggnadsdesign/images/Publikationer/EBD_R_07- 15_E_report_English.pdf. Carlucci, Salvatore, Francesco Causone, Francesco De Rosa, and Lorenzo Pagliano. 2015. “A Review of Indices for Assessing Visual Comfort with a View to Their Use in Optimization Processes to Support Building Integrated Design.” Renewable and Sustainable Energy Reviews. doi:10.1016/j.rser.2015.03.062. Colaco, Sheryl G., Ciji P. Kurian, V. I. George, and Anitha M. Colaco. 2008. “Prospective Techniques of Effective Daylight Harvesting in Commercial Buildings by Employing Window Glazing, Dynamic Shading Devices and Dimming Control—a Literature Review.” Building Simulation 1 (4): 279–89. doi:10.1007/s12273-008- 8126-8. David, M., M. Donn, F. Garde, and A. Lenoir. 2011. “Assessment of the Thermal and Visual Efficiency of Solar Shades.” Building and Environment 46 (7): 1489–96. doi:10.1016/j.buildenv.2011.01.022. DiLaura, David L., Kevin W. Houser, Richard G. Mistirck, and Gary R. Steffy. 2011. “The Lighting Handbook 10th Edition Reference and Application.” In The Lighting Handbook 10th Edition Reference and Application, Section 13.3. Energy Design Resources. 2018. “Understanding Daylight Metrics.” Accessed January 21. https://energydesignresources.com/media/1702/EDR_DesignBriefs_daylightmetrics.pdf?tracked=true. Giovannini, Luigi, Valerio R.M. Lo Verso, Boris Karamata, and Marilyne Andersen. 2015. “Lighting and Energy Performance of an Adaptive Shading and Daylighting System for Arid Climates.” In Energy Procedia, 78:370–75. doi:10.1016/j.egypro.2015.11.675. Grynning, Steinar, Nicola Lolli, Solvår Wågø, and Birgit Risholt. 2017. “Solar Shading in Low Energy Office Buildings - Design Strategy and User Perception.” Journal of Daylighting 4 (1): 1–14. doi:10.15627/jd.2017.1. Guillemin, A., and N. Morel. 2001. “Innovative Lighting Controller Integrated in a Self-Adaptive Building Control System.” Energy and Buildings 33 (5): 477–87. doi:10.1016/S0378-7788(00)00100-6. Hopkinson, R. G. 1972. “Glare from Daylighting in Buildings.” Applied Ergonomics 3 (4): 206–15. doi:10.1016/0003-6870(72)90102-0. horr, Yousef Al, Mohammed Arif, Martha Katafygiotou, Ahmed Mazroei, Amit Kaushik, and Esam Elsarrag. 2016. “Impact of Indoor Environmental Quality on Occupant Well-Being and Comfort: A Review of the Literature.” International Journal of Sustainable Built Environment. doi:10.1016/j.ijsbe.2016.03.006. Hu, Jia, and Svetlana Olbina. 2011. “Illuminance-Based Slat Angle Selection Model for Automated Control of Split Blinds.” Building and Environment 46 (3): 786–96. doi:10.1016/j.buildenv.2010.10.013. Hu, Jianjun, and Panagiota Karava. 2014. “A State-Space Modeling Approach and Multi-Level Optimization Algorithm for Predictive Control of Multi-Zone Buildings with Mixed-Mode Cooling.” Building and Environment 80: 259–73. doi:10.1016/j.buildenv.2014.05.003. 107 IEA. 2000. “Daylight in Buildings.” Task 21/ ECBCS Annex 29 3 (3): 262. doi:10.1001/jama.1936.02770020010056. Jakubiec, J. Alstan, and Christoph F. Reinhart. 2011. “DIVA 2.0: Integrating Daylight and Thermal Simulations Using Rhinoceros 3D, DAYSIM and EnergyPlus.” In Proceedings of Building Simulation 2011: 12th Conference of International Building Performance Simulation Association, 2202–9. doi:10.1017/CBO9781107415324.004. Konis, Kyle. 2017. “A Novel Circadian Daylight Metric for Building Design and Evaluation.” Building and Environment 113: 22–38. doi:10.1016/j.buildenv.2016.11.025. Konis, Kyle, and Stephen Selkowitz. 2017. “The Challenge of Effective Daylighting.” In Green Energy and Technology, 1–31. doi:10.1007/978-3-319-39463-3_1. Konis, Kyle Stas. 2009. “Effective Daylighting: Evaluating Daylighting Performance in the San Francisco Federal Building from the Perspective of Building Occupants.” Spring 94 (2): 9–10. doi:10.1080/09613218.2011.556008. Konstantoglou, Maria, and Aris Tsangrassoulis. 2016. “Dynamic Operation of Daylighting and Shading Systems: A Literature Review.” Renewable and Sustainable Energy Reviews. doi:10.1016/j.rser.2015.12.246. Koo, So Young, Myoung Souk Yeo, and Kwang Woo Kim. 2010. “Automated Blind Control to Maximize the Benefits of Daylight in Buildings.” Building and Environment 45 (6): 1508–20. doi:10.1016/j.buildenv.2009.12.014. “LBNL Windows & Daylighting Software -- WINDOW.” 2017. Accessed August 27. https://windows.lbl.gov/software/window/window.html. Lee, E.S., D.L. DiBartolomeo, and S.E. Selkowitz. 1998. “Thermal and Daylighting Performance of an Automated Venetian Blind and Lighting System in a Full-Scale Private Office.” Energy and Buildings 29 (1): 47–63. doi:10.1016/S0378-7788(98)00035-8. Lee, Eleanor S., Stephen Selkowitz, Robert Clear, Mehlika Inanici, Vorapat Inkarojrit, and Judy Lai. 2005. “Daylighting the New York Times Headquarters Building.” Windows.lbl.gov. http://windows.lbl.gov/comm_perf/pdf/daylighting-nyt-final-iii.pdf. Lee, Es, Ll Fernandes, and B Coffey. 2013. “A Post-Occupancy Monitored Evaluation of the Dimmable Lighting, Automated Shading, and Underfloor Air Distribution System in The New York Times Building.” Lawrence Berkeley National Laboratory, no. January: 1–69. http://escholarship.org/uc/item/3km3d2sn.pdf. Leupen, Bernard. 2005. “Towards Time-Based Architecture.” In Time-Based Architecture. Li, Siwei, Jaewan Joe, Jianjun Hu, and Panagiota Karava. 2015. “System Identification and Model-Predictive Control of Office Buildings with Integrated Photovoltaic-Thermal Collectors, Radiant Floor Heating and Active Thermal Storage.” Solar Energy 113: 139–57. doi:10.1016/j.solener.2014.11.024. Loonen, R. C. G. M., J. M. Rico-Martinez, F. Favoino, M. Brzezicki, C. Menezo, G. La Ferla, and L. Aelenei. 2015. “Design for Façade Adaptability – Towards a Unified and Systematic Characterization.” Proceedings of the 10th Conference on Advanced Building Skins, 1284–94. Lu, Tao, Xiaoshu Lü, and Martti Viljanen. 2011. “A Novel and Dynamic Demand-Controlled Ventilation Strategy for CO2 Control and Energy Saving in Buildings.” Energy and Buildings 43 (9): 2499–2508. doi:10.1016/j.enbuild.2011.06.005. Mahmoud, Ayman Hassaan Ahmed, and Yomna Elghazi. 2016. “Parametric-Based Designs for Kinetic Facades to Optimize Daylight Performance: Comparing Rotation and Translation Kinetic Motion for Hexagonal Facade Patterns.” Solar Energy 126: 111–27. doi:10.1016/j.solener.2015.12.039. Moroşan, Petru-Daniel, Romain Bourdais, Didier Dumur, and Jean Buisson. 2010. “Building Temperature Regulation Using a Distributed Model Predictive Control.” Energy and Buildings 42 (9): 1445–52. doi:10.1016/j.enbuild.2010.03.014. 108 Oh, Myung Hwan, Kwang Ho Lee, and Jong Ho Yoon. 2012. “Automated Control Strategies of inside Slat-Type Blind Considering Visual Comfort and Building Energy Performance.” Energy and Buildings 55: 728–37. doi:10.1016/j.enbuild.2012.09.019. Pierson, Clotilde, Jan Wienold, and Magali Bodart. 2017. “Discomfort Glare Perception in Daylighting: Influencing Factors.” In Energy Procedia, 122:331–36. doi:10.1016/j.egypro.2017.07.332. Reinhart, Christoph F., John Mardaljevic, and Zack Rogers. 2006. “Dynamic Daylight Performance Metrics for Sustainable Building Design.” LEUKOS - Journal of Illuminating Engineering Society of North America 3 (1): 7–31. doi:10.1582/LEUKOS.2006.03.01.001. Shen, Eric, Jia Hu, and Maulin Patel. 2014. “Energy and Visual Comfort Analysis of Lighting and Daylight Control Strategies.” Building and Environment 78: 155–70. doi:10.1016/j.buildenv.2014.04.028. “Sky Types.” 2018. Accessed April 1. https://www.new- learn.info/packages/clear/visual/daylight/sun_sky/sky_types.html. Suk, Jae Yong, Marc Schiler, and Karen Kensek. 2017. “Investigation of Existing Discomfort Glare Indices Using Human Subject Study Data.” Building and Environment 113: 121–30. doi:10.1016/j.buildenv.2016.09.018. Tenner, Ariadne D. 2003. “A Healthy Future for Office Lighting?” Journal of Light & Visual Environment. doi:10.2150/jlve.27.172. U.S. Energy Information, a. 2013. “Annual Energy Outlook 2013.” Office of Integrated and International Energy Analysis 1: 1–244. doi:DOE/EIA-0383(2013). Wang, Jialiang, Liliana O. Beltrán, and Jonghoon Kim. 2012. “From Static to Kinetic: A Review of Acclimated Kinetic Building Envelopes.” World Renewable Energy Forum, WREF 2012, Including World Renewable Energy Congress XII and Colorado Renewable Energy Society (CRES) Annual Conferen 5: 4022–29. http://www.scopus.com/inward/record.url?eid=2-s2.0-84871549544&partnerID=tZOtx3y1. Wienold, Jan. 2007. “Dynamic Simulation of Blind Control Strategies for Visual Comfort and Energy Balance Analysis.” IBPSA 2007 - International Building Performance Simulation Association 2007, 1197–1204. doi:10.1016/S0378-7788(01)00058-5. Xiong, Jie, and Athanasios Tzempelikos. 2016. “Model-Based Shading and Lighting Controls Considering Visual Comfort and Energy Use.” Solar Energy 134: 416–28. doi:10.1016/j.solener.2016.04.026.

Abstract (if available)

Abstract A static daylighting façade system creates varying and uncontrolled levels of daylighting and visual conditions in a space. These visual comfort conditions are evaluated based on standard requirements or averaged visual comfort preference ranges that might not reflect true individual preferences. Such a control strategy, therefore, fails to address stochastic phenomena, such as changing, and often competing, individual visual comfort requirements and bias, and factors such as occupant position and viewing direction in a space. An adaptive daylighting façade system can consist of multiple independently-controllable components that are driven by a control logic which determines its control states. Using an objective function and through quantification of the required features, a control logic can assist in finding one or more possible daylighting system configurations that most closely produce certain predetermined criteria, such as glare or illuminance. For a multiple-occupancy condition, a set of individual and cumulative dissatisfaction indices were proposed that represented and helped quantify how effective an adaptive façade configuration was in producing the occupant-preferred sets of visual comfort performance. In a case study, adaptive facade configurations at 15-minute intervals were ranked based on their performance, finding control states that affected the results the most, developing and improving daylighting control strategies, and providing a more personal and generally satisfactory and controlled daylit environment. The dynamic real-time control logic framework accounted for three visual comfort factors: average task surface illuminance, glare, and task surface to background luminance contrast ratio. The south-east corner of an existing open-plan office in downtown Los Angeles was modeled with electrochromic vision glazing, kinetic shading, and light shelves. Using the Grasshopper and Honeybee interfaces, daylighting simulations under climate-based sky conditions were performed for three times daily for three days of the year to evaluate the proposed framework. The framework integrated a simplified occupant preference model with a simplified glare calculation method and a Radiance-based illuminance-luminance simulation workflow, using calculation and data analysis modules. Python was used to create a normalized distribution of personal occupant visual comfort ranges based on predefined lists and to generate flat importance factors, for glare, illuminance, and contrast ratio. The variations of each measured occupant comfort value from the preferred range, weighted by their respective importance factors, was calculated to provide a net deviation value for each occupant. A brute force simulation method was used to generate all possible façade configurations, and the respective visual comfort factor deviation values were then recorded, analyzed, and compared against a baseline case. The proposed framework substantially increased the overall as well as individual percentage occupant visual satisfaction in a multiple occupancy condition in comparison with the baseline case. A substantial reduction of cumulative deviation from the baseline case was observed for certain façade configurations, thus validating that the proposed control framework can be used to optimize façade control states that improve the overall occupant satisfaction with the visual environment by making more people more satisfied with their visual task environment. Analyses of the control states of the high performing façade configurations were useful in developing daylighting control strategies and to find adaptive façade elements and states that have the most effect on the visual comfort condition at different times of the year.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

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

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

PDF

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

PDF

Simulation-based electric lighting control algorithm: integrating daylight simulation with daylight-linked lighting control

PDF

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

PDF

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

PDF

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

PDF

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

PDF

Human–building integration: machine learning–based and occupant eye pupil size–driven lighting control as an applicable visual comfort tool in the office environment

PDF

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

PDF

Energy savings by using dynamic environmental controls in the cavity of double skin facades

PDF

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

PDF

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

PDF

Occupant-aware energy management: energy saving and comfort outcomes achievable through application of cooling setpoint adjustments

PDF

Developing environmental controls using a data-driven approach for enhancing environmental comfort and energy performance

PDF

Blind to light loss: evaluating light loss through commercial building facades as a contribution to urban light pollution

PDF

Enhancing thermal comfort: data-driven approach to control air temperature based on facial skin temperature

PDF

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

PDF

Exploring the influence of scheduling on daylighting performance evaluation in assisted living: testing two new metrics based on sDA and ASE

PDF

Multi-occupancy environmental control for smart connected communities

Asset Metadata

Creator Roy, Kushnav (author)

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

School School of Architecture

Degree Master of Building Science

Degree Program Building Science

Publication Date 07/26/2018

Defense Date 04/26/2018

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

Tag adaptive facade control,adaptive facades,cluster analysis,control states,cumulative dissatisfaction,daylighting,daylighting simulation,multiple occupancy condition,OAI-PMH Harvest,performance clustering,performance ranking,preference aggregation,user preferences,visual comfort,visual comfort preference

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Noble, Douglas (committee chair), Kensek, Karen (committee member), Konis, Kyle (committee member)

Creator Email kroy@usc.edu,kushnavroy@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-31541

Unique identifier UC11670511

Identifier etd-RoyKushnav-6515.pdf (filename),usctheses-c89-31541 (legacy record id)

Legacy Identifier etd-RoyKushnav-6515.pdf

Dmrecord 31541

Document Type Thesis

Format application/pdf (imt)

Rights Roy, Kushnav

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