Citation
Characterization of Near-Surface Turbulent Boundary Layer Flows for Wind Load Analysis and Optimization of Large Frame Structures under Wind Action

Material Information

Title:
Characterization of Near-Surface Turbulent Boundary Layer Flows for Wind Load Analysis and Optimization of Large Frame Structures under Wind Action
Creator:
Fernandez Caban, Pedro L
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (153 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Civil Engineering
Civil and Coastal Engineering
Committee Chair:
MASTERS,FORREST J
Committee Co-Chair:
CONSOLAZIO,GARY R
Committee Members:
BRIDGE,JENNIFER ANNE
HAFTKA,RAPHAEL TUVIA

Subjects

Subjects / Keywords:
buildings -- optimization -- pressures -- wind
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Civil Engineering thesis, Ph.D.

Notes

Abstract:
This dissertation presents three interrelated studies of importance to the quantification of boundary layer wind loads on civil infrastructure. The first study refutes the longstanding assumption that near surface wind velocity exhibits Gaussian behavior irrespective of the aerodynamic roughness length of the upwind fetch. Analysis of a large database comprising ground-level anemometric observations during land falling hurricanes, and a series of boundary layer wind tunnel (BLWT) experiments suggest that the longitudinal component of the wind velocity positively skews in more built-up terrain conditions during strong wind events. Quadrant analysis of the Reynolds stress revealed the strong linkage between non-Gaussian observations and sweep events (high-velocity air masses moving towards the surface). The study demonstrates and discusses possible implications of observed non-Gaussian trends on peak loads on buildings. The second study leverages a comprehensive series of experiments in a large boundary layer wind tunnel to investigate the variation of external pressure coefficients with increasing surface roughness on low-rise buildings. BLWT modeling was carried out on three rigid building models of the Wind Engineering Research Field Laboratory (WERFL) experimental building with geometric scales of 1:20, 1:30, and 1:50. A total of 33 upwind terrain configurations were explored for each model scale. BLWT pressure tests were also compared to aerodynamic experiments of a 1:100 model of the WERFL building conducted at University of Western Ontario (UWO). The findings revealed the dependency of extreme pressure coefficients with increasing surface roughness. Peak pressures near the leading edge of the roof appeared to vary linearly with increasing turbulence levels at eave height. Only slight variations in the mean pressures were observed with increasing turbulence levels. The third study addresses the optimization of large frame structures subjected to wind action. It presents a new metaheuristic technique to minimize the weight of the structure while satisfying strength and serviceability requirements. The algorithm transitions from the global examination of the design domain, to the restricted investigation around promising regions (subdomains). The latter is performed through a discrete stochastic search scheme, which conducts an exhaustive search near the global optimum during late stages of the optimization process (exploitation). ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2017.
Local:
Adviser: MASTERS,FORREST J.
Local:
Co-adviser: CONSOLAZIO,GARY R.
Statement of Responsibility:
by Pedro L Fernandez Caban.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
LD1780 2017 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

CHARACTERIZATION OF NEAR SURFACE TURBULENT BOUNDARY LAYER FLOWS FOR WIND LOAD ANALYSIS AND OPTIMIZATION OF LARGE FRAME STRUCTURES UNDER WIND ACTION By PEDRO LUIS FERN NDEZ CAB N A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2017

PAGE 2

2017 Pedro Luis Fern ndez Cabn

PAGE 3

To my parents Delia Iraida Cab n D vila and Pedro Carlos Fern ndez Torres

PAGE 4

4 ACKNOWLEDGMENTS I am deeply grateful for the unconditional suppor t and guidance of Dr. Forrest Masters, and for the trust that he had i n me throughout the development of this work. H is mentoring has proven indispensable in my growth as an academic researcher and, most importantly, as a person I would like to recognize my committee members, Dr. Jen nifer Rice, Dr. Gary Consolazio, and Dr. Raphael Haftka for their assistance and contribution to this dissertation. Dr. Luis Aponte and Ryan Catarelli also deserves my gratitude. Finally, I would like to thank and acknowledge the staff of the Powell Struct ures and Materials Laboratory for their passionate involvement and support in field measurements and wind tunnel testing.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 14 C HAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 16 Objectives and Motivation ................................ ................................ ....................... 16 Organization ................................ ................................ ................................ ........... 18 2 CHARACTERIZATION OF NEAR SURFACE HURRICANE WINDS IN BUILT UP TERRAIN ................................ ................................ ................................ .......... 20 Background ................................ ................................ ................................ ............. 20 Methodology ................................ ................................ ................................ ........... 21 Field Experiments of Land Falling Tropical Cyclones ................................ ....... 21 Data Processing ................................ ................................ ............................... 22 Conditional Analysis of the Reynolds Stress ................................ .................... 24 Boundary Layer Wind Tunnel Measurements ................................ .................. 26 Results ................................ ................................ ................................ .................... 28 Higher Order Central Moments ................................ ................................ ........ 28 Reynolds Stress Fractions ................................ ................................ ................ 29 Relation of Stress Fraction Difference and Skewness ................................ ...... 30 Discussion ................................ ................................ ................................ .............. 31 Summary ................................ ................................ ................................ ................ 33 3 WIND SIMULATION IN A LARGE BOUNDARY LAYER WIND TUNNEL ............... 48 Background ................................ ................................ ................................ ............. 48 Approach to Modeling ................................ ................................ ............................. 49 ESDU Model ................................ ................................ ................................ ..... 49 Morphologic Models for Roughness Length Estimates ................................ .... 50 Met hodology ................................ ................................ ................................ ........... 51 Wind Tunnel ................................ ................................ ................................ ..... 51 Roughness Element Grid and Vortex Generators ................................ ............ 52 Flow Measurements ................................ ................................ ......................... 52 Experimental Design ................................ ................................ ........................ 53 Resul ts and Discussion ................................ ................................ ........................... 53 Mean Velocity Profiles ................................ ................................ ...................... 53

PAGE 6

6 Turbulence intensity Profiles ................................ ................................ ............ 54 Summary ................................ ................................ ................................ ................ 55 4 SURFACE PRESSURES ON LOW RISE BUILDINGS ................................ .......... 65 Background ................................ ................................ ................................ ............. 65 Simulation of Approach Flow ................................ ................................ .................. 67 Low Rise Building Models ................................ ................................ ....................... 68 Pressure Measurements ................................ ................................ ......................... 68 Instrumentation ................................ ................................ ................................ 68 Pressure Coefficients ................................ ................................ ....................... 69 Peak Pressures ................................ ................................ ................................ 70 Experimental Design ................................ ................................ ............................... 71 Results and Discussion ................................ ................................ ........................... 72 Validation of Pressure Measurements ................................ .............................. 72 Spatial Distribution of Surface Pressures ................................ ......................... 74 Characterizing Upwind Terrain and Model Scale Comparison ......................... 76 Summary ................................ ................................ ................................ ................ 77 5 WIND LOAD DESIGN OPTIMIZATION OF STEEL FRAMES ................................ 96 Background ................................ ................................ ................................ ............. 96 Discrete Optimization of Multi Story Frame Structures ................................ ........... 99 Proposed ETE Algorithm ................................ ................................ ...................... 102 Particle Swarm Optimization (PSO) ................................ ............................... 102 ETE Algorithm ................................ ................................ ................................ 103 Linearly Varying Control Parameter ................................ ......................... 104 Discrete Stochastic (Exploitation) Scheme ................................ .............. 105 Results ................................ ................................ ................................ .................. 106 24 Story Three Bay frame ................................ ................................ .............. 106 60 Story Seven Bay Frame ................................ ................................ ............ 108 Summary ................................ ................................ ................................ .............. 111 6 CONCLUSIONS AND FUTURE WORK ................................ ............................... 126 Field Measurements of Hurricane Winds ................................ .............................. 126 Surface Pressures on Low Rise Buildings ................................ ............................ 126 Design Optimization of Large Civil Structures Subject to Wind Loads .................. 127 A P PENDIX A UWO PRESSURE TAP LAYOUT AND TESTING PARAMETERS ...................... 129 B UFL WERFL MODELS ................................ ................................ ......................... 132 C GEOMETRY OF WERFL MODELS (UFL) ................................ ............................ 136 LIST OF REFERENCES ................................ ................................ ............................. 145

PAGE 7

7 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 153

PAGE 8

8 LIST OF TABLES Table page 2 1 FCMP datasets included in this study in addition to those analyzed in Balderrama et al. (2012). ................................ ................................ .................... 35 2 2 Total number of 15 min records ( = 10 m) stratified into mean wind speed and roughness length based on ASCE 7 10 exposure categories. .................... 35 2 3 Mean roughness lengt hs stratified into mean wind speed and roughness length ( = 10 m) based on ASCE 7 10 exposure categories. ............................ 35 2 4 Quadrants of e vents contributing to the total mean Reynold stress. ................... 36 2 5 Equivalent full scale roughness lengths obtained from suburban terrain BLWT simulation. ................................ ................................ ............................... 36 2 6 Statistics of 15 min records stratified into mean wind speed and roughness length based on ASCE 7 10 exp osure categories. ................................ ............. 36 2 7 Mean stress fraction difference between sweep and ejection stratified into mean wind speed and roughness length. ................................ ........................... 37 2 8 Proposed exposure coefficient values obtained from non Gaussian peak factor model ................................ ................................ ................................ ........ 37 3 1 BLWT aerodynamic parameters. ................................ ................................ ........ 56 4 1 WERFL testing parameters ................................ ................................ ................ 78 5 1 Summary of statistics for optimum design of 24 story three bay frame (100 Runs). ................................ ................................ ................................ ............... 112 5 2 Percent difference relative to best run for 24 story three bay frame (100 Runs). ................................ ................................ ................................ ............... 112 5 3 Comparison of best designs for 24 story three bay frame. ............................... 11 3 5 4 Location of structural members in the 60 story frame. ................................ ...... 114 5 5 Summary of statistics for opti mum design of 60 story seven bay frame (Independent Runs = 15). ................................ ................................ ................. 114 5 6 Percent difference relative to best run for 60 story seven bay frame (Independent Runs = 15). ................................ ................................ ................. 114 5 7 Optimum design comparison for the 60 story planar frame. ............................. 115

PAGE 9

9 C 1 WERFL model tap coordinates (1:20) ................................ .............................. 137 C 2 Vertices of WERFL building model (1:20) ................................ ......................... 144

PAGE 10

10 LIST OF FIGURES Figure page 2 1 Plan view of the BLWT at the University of Florida. ................................ ............ 38 2 2 Spatially averaged mean velocity profiles of six independent BLWT simulations measured at location = 29.5 m ................................ ..................... 39 2 3 Longitudinal turbulence spectra of six independent BLWT simulations, measured at location = 29.5 m ................................ ................................ ........ 40 2 4 Observed mean (longitudinal) skewness values stratified into mean wind speed and terrain roughness. ................................ ................................ ............. 41 2 5 Observed mean (longitudinal) kurtosis values stratified into mean wind speed and terrain roughness. ................................ ................................ ........................ 41 2 6 Mean Reynolds stress fractions for mean wind speed values stratified into mean wind speed and terrain roughness for the field observations. ................... 42 2 7 Spatially averaged Reynolds stress fractions of six independent BLWT simulations measured at location = 29.5 m. ................................ .................... 43 2 8 Theoretical and measured stress fraction difference between sweep and ejection quadrants ................................ ................................ .............................. 44 2 9 Measured stress fraction difference versus longitu dinal skewness .................... 45 2 10 Measured stress fraction difference v ersus vertical skewness ........................... 46 2 11 Proposed non Ga ussian exposure coefficient versus elevation above ground . 47 3 1 Boundary layer wind tunnel at the University of Florida ................................ ...... 57 3 2 Set of screens and honeycomb system downwind of the fan bank .................... 57 3 3 Pitot tube mounted to the wall of the tunne l ................................ ........................ 58 3 4 Freestre am reference anemometry ................................ ................................ .... 58 3 5 Meteorological weather station located adjacent to the exit of the BLWT ........... 59 3 6 Irwin spires located downwind of the screens and honey comb .......................... 59 3 7 Longitudinal m ean velocity profiles for a narrow edge windward element orientation. ................................ ................................ ................................ .......... 60

PAGE 11

11 3 8 Longitudinal Mean velocity profiles for a wide edge windward element orientation. ................................ ................................ ................................ .......... 60 3 9 Semi l ogarithmic mean velocity profiles for the narrow edge windward element orientation. ................................ ................................ ............................ 61 3 10 Semi logarithmic mean velocity profil es for the wide edge windward element orientation. ................................ ................................ ................................ .......... 61 3 11 Longitudinal turbulence intensity profiles for a narrow edge windward element orientation. ................................ ................................ ............................ 62 3 12 Longitudinal turbulence intensity profiles for a wide edge windward element orientation. ................................ ................................ ................................ .......... 62 3 13 Lateral turbulence intensity profiles for a narrow edge windward element orientation. ................................ ................................ ................................ .......... 63 3 14 Lateral turbulence intensity profiles for a wide edge windward element orientation. ................................ ................................ ................................ .......... 63 3 15 Ver tical turbulence intensity profiles for a narrow edge windward element orientation. ................................ ................................ ................................ .......... 64 3 16 Vertical turbulence intensity profiles for a wide edge windward element orientation. ................................ ................................ ................................ .......... 64 4 1 Pressure tap layout for 1:20, 1:30, and 1:50 WERFL building models. .............. 79 4 2 Measured and fitted response amplitudes up to a frequency of 600 Hz. ............ 80 4 3 Schematic diagram of BLWT physical arrangement near the test section for pressure tests. ................................ ................................ ................................ .... 80 4 4 BLWT ...................... 81 4 5 Comparison of mean pressure co efficients for Narrow edge, = 150 mm, and 90 ................................ ................................ ................................ ....... 82 4 6 Comparison of peak pressure coefficients for Wide edge, = 80 mm, and 90 ................................ ................................ ................................ .............. 83 4 7 Pressure spectra comparison of Tap 216 for the 1:20 WERFL model. ............... 84 4 8 Distribution of mean pressure coeffic ients for the 1:20 WERFL model for Wide edge and 0 ................................ ................................ ....................... 85 4 9 Distribution of RMS pressure coeffic ients for the 1:50 WERFL model for Wide edge and 90 ................................ ................................ .............................. 86

PAGE 12

12 4 10 Distribution of peak pressure coefficients for the 1:30 WERFL model for Wide edge and 45 ................................ ................................ .............................. 87 4 11 Distribution of mean pressures along a line of taps parallel to the long building dimension for the 1:20 WERFL model (Wide edge, 0). ................. 88 4 12 Distribution of mean pressures along a line of taps parallel to the long building dimension for the 1:30 WERFL model (Wide edge, 0). ................. 88 4 13 Distribution of mean pressures along a line of taps parallel to the long building dimension for the 1:50 WERFL model (Wide edge, 0). ................. 89 4 14 Distribution of mean pressures along a line of taps perpendicular to the long building dimension for the 1:20 WERFL model (Wide edge, 90). ............... 89 4 15 Distribution of mean pressures along a line of taps perpendicular to the long building dimension for the 1:30 WERFL model (Wide edg e, 90). ............... 90 4 16 Distribution of mean pressures along a line of taps perpendicular to the long building dimension for the 1:50 WERFL model (Wide edge, 90). ............... 90 4 17 Distribution of peak pressures along a line of taps parallel to the long building dimension for the 1:20 WERFL model (Wide edge, 0). .............................. 91 4 18 Distribution of peak pressures along a line of taps parallel to the long building dimension for the 1:30 WERFL model (Wide edge, 0). .............................. 91 4 19 Distribution of peak pressures along a line of taps parallel to the long building dimension for the 1:50 WERFL model (Wide edge, 0). .............................. 92 4 20 Peak pres sures on all 266 taps for open country terrain simulation ................... 93 4 21 Peak pressures on all 266 taps for suburban terrain simulation ........................ 94 4 22 Peak pressures on all 266 taps for open country terrain simulation with narrow and wide edge windward element orientation. ................................ ........ 95 5 1 Linear variation of during the optimization process. ................................ ..... 116 5 2 Conditional subroutine for determining ................................ ........................ 116 5 3 Flowchart of ETE for discrete sizing design optimization of framed structures. 117 5 4 T opology and loading condition for the 24 story three bay planar frame. ......... 118 5 5 ETE mean optimization history for the 24 story three bay frame ..................... 119 5 6 Member sizing, inter story drift, and load capacity ratios for the 24 story three bay fr ame obtained using ETE for Case 1 ................................ .............. 120

PAGE 13

13 5 7 Member sizing, inter story drift, and load capacity ratios for the 24 story three bay fr ame obtained using ETE for Case 3 ................................ ............... 121 5 8 To pology and loading condition for the 60 story seven bay planar frame ......... 122 5 9 ETE mean optimization histo ry for the 60 story seven bay frame .................... 123 5 10 Member sizing, inter story drift, and load capacity ratios for the 60 story seve n bay frame obtained using ETE for Case 2 ................................ ............. 124 5 11 Member sizing, inter story drift, and load capacity ratios for the 60 story seve n bay frame obtained using ETE for Case 3 ................................ ............. 125 A 1 UWO pressure tap layout for 1:100 WERFL model. ................................ ......... 130 A 2 UWO test parameters for 1:100 WERFL model ................................ ................ 131 B 1 1:20 UFL WERFL model for Narrow edge, = 20 mm, and 0 ................. 133 B 2 1:20 UFL WERFL model for Narrow edge, = 80 mm, and 45 .............. 133 B 3 1:20 UFL WERFL model for Narrow edge, = 160 mm, and 90 ............ 134 B 4 1:20 UFL WERFL model for Wide edge, = 80 mm, and 0 ..................... 134 B 5 1:20 UFL WERFL model for Wide edge, = 20 mm, and 45 ................... 135 B 6 1:20 UFL WERFL model for Wide edge, = 160 mm, and 90 ................ 135

PAGE 14

14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy CHARACTERIZATION OF NEAR SURFACE TURBULENT BOUNDARY LAYER FLOWS FOR WIND LOAD ANALYSIS AND OPTIMIZATION OF LARGE FRAME STRUCTURES UNDER WIND ACTION By Pedro Luis Fern ndez Cabn August 2017 Chair: Forrest J. Masters Major: Civil Engineering This dissertation presents three interrelated studies of importance to the quantification of boundary layer wind loads on civil infrastructure The first study refutes the longstanding assumption that near surface wind velocity exhibits Gaussian behavior irrespective of the aerodynamic roughness length of the upwind fetch. A nalysis of a large data base comprising ground level anemometric observat ions during land falling hurricanes and a series of boundary layer wind tu nnel (BLWT) experiments suggest that the longitudinal component of the wind velocity positively skews in more built up terrain conditions during strong wind events. Quadrant analysis of the Reynolds stress revealed the strong linkage between non Gaussian observations and sweep events high velocity air masses moving towards the surface. The study demonstrates and discusses possible implications of observed non Gau ssian trends on peak loads on buildings The second study leverages a comprehensive series of experiments in a large boundary layer wind tunnel to investigate the variation of external pressure coefficient s with increasing surface roughness on low rise buildings. BLWT modeling was carried

PAGE 15

15 out on three rigid building models of the Wind Engineering Research Field Laboratory (WERFL) experimental buildi ng with geometric scales of 1:20, 1:30, and 1:50 A total o f 33 upwind terrain configurations were explored for each model scale. BLWT pressure tests were also compared to aerodynamic experiments of a 1:100 model of the WERFL building conducted at University of Western Ontario (UWO). The findings revealed the dependency of extreme pressure coefficients with increasing surface roughness Peak pressures near the leading edge of the roof appeared to vary linearly with increasing turbulence levels at eave height. Only slight variations in the mean pressures were ob served with increasing turbulence levels. The third study addresses the optimization of large frame structures subjected to wind action. It presents a new metaheuristic technique to minimize the weight of the structure while satisfying strength and service ability requirements. The algorithm transitions from the global examination of the design domain, to the restricted investigation around promising regions ( subdomains ) The latter is performed through a discrete stochastic search scheme, which conducts an exhaustive search near the global optimum during late stages of the optimization process (exploitation).

PAGE 16

16 CHAPTER 1 INTRODUCTION Despite recent advancements in computational wind engineering in the past few decades ( e.g., Blocken 2014; Troldborg et al., 2014 ) a ccurately modeling peak wind loads on civil structures through purely numerical techniques remains elusive Therefore, physical measurements are obtained in boundary layer wind tunnel (BLWT). The research herein seeks to advance our understanding of turbulent wind loading and the associated structural response. It addresses basic assumptions about simulated appr oach flow conditions in the BLWT, including the experimental limitations imposed by model scale, and presents a new method to optimization large frame structures under wind action. Objectives and Motivation The purpose of this dissertation was three fold. The first objective was to evaluate how the near surface (<20 m) wind field evolves as the aerodynamic roughness increases. Analysis of field and BLWT three dimensional velocity records found that the surface wind field trends to non Gaussian behavior as the upwind terrain becomes more built up. The cause is attributed to the linkage between the longitudinal skewness and downward transfers of momentum (sweep events) in strong winds. The implication is significant for commissioning of wind tunnel experimen ts. Current standards omit any reference to this phenomenon. The second objective was to examine how the magnitude and spatial distribution of aerodynamic loads on low rise buildings change with terrain, angle of attack, and model scale. Pressure coeffici ents on models of the Texas Tech University Wind Engineering Research Field Laboratory (WERFL) experimental buildi ng with geometric

PAGE 17

17 scales of 1:20, 1:30, and 1:50 were immerse d in a wide range of turbulent boundary layer flows using precise adjustment of a computer control terrain generator called the Terraformer. The series of experiments will serve as a testbed for exploring more complex upwind terrains, such as heterogeneous roughness element fields, to simulate more realistic topographic conditions and the adverse wind induced effects on low r ise buildings. A large database will be stored on the Natural Hazards Engineering Research Infrastructure ( NHERI ) Cyberinfrastructure node at the University of Texas at Austin, which can be accessed by researchers w orldwide. The third objective was to create a structural design framework suitable for sizing optimization of large scale civil structures subject to wind loads. A hybrid metaheuristic search algorithms was developed for cost optimization of steel frames s ubject to code based strength and serviceability requirements. This work adds to the knowledge base for the a utomation of the structural design process of civil structures which is expected to grow significantly in the coming decades Efforts for the development of frameworks, such as Performance Based Design (PBD), for the analysis and design of buildings subject to wind loads are ongoing (e.g., Spence and Gioffr 2012; Spence and Kareem, 2013). This PBD philosophy has been successf ully applied in the field of seismic engineering. The wind engineering community is progressively adapting PBD strategies to obtain optimum and cost effective solutions in the design of civil structures exposed to extreme wind events. However, direct trans fer of the PBD concepts from seismic to wind is not conceivable due to the complex mechanism such as with complex dimensional turbulence and vortex shedding on a structure. Although there has been significant progress in the computational modelling of thes e multifaceted wind

PAGE 18

18 phenomena the use computational fluid dynamics (CFD) and large eddy simulation (LES) their practical application to design optimization problems of civil structures is not well yet established. Consequently, researchers have relied on large databases of detailed aerodynamic and climatologic information from boundary layer wind tunnel (BLWT) tests and full scale field measurements in the design optimization of buildings (Spence and Kareem, 2013). Organization The current dissertation is organized into six chapters that present three manuscripts out of which one is accepted and the other two will soon be submitted to peer reviewed journals. Chapter 2 discusses results from a large database of field experiments to examine non Gaussian wind behavior in suburban terrains, and probable implication s to peak loads on low rise structures. This chapter was recently accepted by the Journal of Wind Engineering and Industrial Aerodynamics. Chapters 3 and 4 will later be comb ined to produce a single journal article, and will be submitted to the Journal of Wind Engineering and Industrial Aerodynamics. Chapter 3 describes the components of a large boundary layer wind tunnel (BLWT) located at the University of Florida The chapte r discusses methods applied for the validation of an automated terrain generator (i .e., the Terraformer). Chapter 4 outlines findings from a comprehensive series of aerodynamic tests on low rise structures in the BLWT to examine the effects of small pertur bations in upwind terrain conditions on peak loads.

PAGE 19

19 Chapter 5 presents a hybrid metaheuristic search algorithm for discrete sizing optimization of large steel frames subject to wind loads. This chapter will be submitted to the Journal of Computers and Stru ctures. Chapter 6 (Conclusions and Future W ork) summarizes general findings, limitations, and recommendations resulting from this dissertation.

PAGE 20

20 CHAPTER 2 CHARACTERIZATION OF NEAR SURFACE HURRICANE WINDS IN BUILT UP TERRAIN This chapter presents a nalysis of anemometric observations collected during the 1999 2016 Atlantic hurricane seasons and boundary layer wind tunnel (BLWT) measurements indicate that the near surface wind behavior deviates from Gaussian as surface roughness increases. Quadrant an alysis of the Reynolds stress clearly demonstrates the linkage between the longitudinal skewness and downward transfers of momentum (sweep events) in strong winds, which previous studies have also observed. Modern wind load provisioning and BLWT similarity requirements do not explicitly account for this effect, although the experimental configuration of typical wind tunnel development sections may correctly simulate the phenomena. Here we quantify how the observed positive skewness affects the terrain expos ure coefficient profile for ASCE 7. The results demonstrate the need to update peak factor calculations to accurately predict extreme winds acting on low rise buildings in built up terrain. Background T he dominant assumption in wind engineering is that th e surface (< 25 m) wind field exhibits Gaussian behavior over all terrain types. Results from 17 years of anemometric observations in tropical cyclones and recent boundary layer wind tunnel (BLWT) experiments indicate that the skewness ( ) of the longi tudinal velocity component increases with aerodynamic roughness length ( ) a phenomena not currently addressed by exposure conversion factors in wind load provisioning nor in standards to commission facilities for BLWT modeling, e.g. ASCE/SEI 49 12 and A WES QAM 1 2001. The cause is attributed to the prevalence of sweeps, i.e. downward and forward departures from the mean longitudinal flow, which positively

PAGE 21

21 skew the velocity. This mechanism, in part, also explains why field studies (e.g., Baldocchi and Hut chison 1987; Baldocchi and Meyers 1988; Maitani and Shaw, 1990) have observed variations in high order moments near the canopy of suburban regions. Our observations strongly corroborate with previous findings originating from field measurements (e.g., Che n, 1990; Rotach, 1995; Rotach et al., 2005; Shaw et al., 1983 ) and the BLWT (e.g., Finnigan, 2000 ; Zhu et al., 2007; Raupach, 1981; Bhm et al., 2013 ), which have documented the dominance of large coherent structures within and just above the canopy of th e turbulent boundary layer, and the correlation of these structures to the third moment of the longitudinal and vertical velocity components. The chapter extends the work of Balderrama et al. (2012), expanding its dataset to include targeted field measure ments in suburban terrain and ultrasonic anemometer measurements at five levels spanning 5 to 15 m. It also incorporates new approach flow data collected in a novel BLWT that can rapidly reconfigure the roughness element grid to achieve a user specified for a user specified geometric scale. The results suggest the standards should be conservatively updated to incorporate a linear change in from [0.46, 0.0] over [4.4, 24.6] m, holding 0.46 below this extent. Applying the non Gaussian up crossin g rate defined in Kareem and Zhou (1994) to the wind speed (pressure) conversion described in Masters et. al. (2010), we find up to a 14% increase in the pressure loading on suburban low rise structures (ASCE 7 10 2010 ) Methodology Field Experiments of L and Falling Tropical C yclones surface wind fi eld observations collected durin g 1999 2016 Atlantic tropical cyclones by the Florida C oastal Monitoring Program (FCMP ), a multi institution consortium that includes the University of Florida,

PAGE 22

22 Clemson University, Florida International University, and the Insurance Institute for Business & Home Safety. Balderrama et al. (2011 2012 ) describe th e program in detail The dat aset includes records from 86 deployments of portable weather stations in 25 named storms. The current study incorporates 13 new field experiments that deployed upgraded 1 5 m portable weather stations ( Table 2 1), with suburban terrain being the primary st udy target. The original 10 m FCMP weather stations are equipped with anemometers at 5, and 10 m. Two custom arrays of three fixed axis anemometers (RM Young Model Number 27106R) collect wind velocity observations (3D wind speed and direction) at the 5 and four blade polypropylene helicoid propellers (Model Number 08234) include a 2.7 m, 63% recovery distance constant and a damped natural wavelength of 7.4 m. A wind monitor (RM Young Model Number 0510 3V) installed at the 10 m level serves as a redundant anemometer system to monitor the horizontal component of the wind. The wind monitor 50% recovery vane delay distance is 1.3 m and it is rated for a 100 m/s gust survival. In 2010, two weather stations w ere upgraded with high resolution ultrasonic anemometers ( WindMaster Pro Model 1561 PK 020 ) installed at 5, 7.5, 10, 12.5, and 15 m above ground level. The units have a wind speed range of 0 65 m/s with a resolution of 0.01 m/s, and measure instantaneous , and wind components with a maximum sampling rate of 32 Hz. In this study, data were sampled at or resampled to 10 Hz. Data P rocessing Data were segmented into continuous non overlapping 15 min time histories. Quality control included comparing th e propeller measurements at an elevation of 10 m to the redundant wind monitor measurements at the same elevation to detect

PAGE 23

23 anomalies. Data segments with mean wind velocities below 5 m/s were removed to eliminate any effects from convection. The tower tilt correction described in Foken and Nappo ( 2008) was performed to align the anemometer coordinate system into the streamlines and towards the mean flow coordinate system. Negligible mean vertical and lateral wind components we re verified to satisfy the requ irements of an eddy covariance method of analysis. Linear trend removal methods were performed on all 15 min records to remove first order non stationarities. Shear (friction) velocities for each data segment were calculated directly from the three measure d orthogonal velocity components following Weber (1999): ( 2 1) The logarithmic mean velocity profile was then used to estimate the roughness length : ( 2 2) where is the mean wind speed at elevation constant., and is the zero plane displacement height. Measurements were obtained at 2 5 levels, preventing direct estimation of Thus the heuristic method of Jackson (1981) was applied: ( 2 3)

PAGE 24

24 where is the average roughness element height, nominally assumed to be 5 m, which is representative of locations surrounded by clusters of single family dwellings. Displacement height was assumed to be effectively zero for records with < 0.15 m. For qualit y assurance, an independent procedure was performed to calculate A bounded nonlinear function minimization was applied to satisfy the logarithmic formulation of the mean velocity profile and the modified Harris and Deaves (1980) variance model descri bed in Vickery and Skerjl (2005). Percent differences between the two procedures were in the range of 2 6% for records corresponding to suburban terrain ( 0.15 m > ). Data records were stratified by surface roughness (exposure) categories defined in ASCE 7 10 (2010), i.e. B (suburban), C (open), and D (marine). Table 2 2 lists the number of 15 min records stratified by surface roughness and the mean wind speed at 10 m. Table 2 3 lists the corresponding mean roughness length estimates, which were o bta ined from Equation 2 2 for Conditional Analysis of the Reynolds S tress Conditional (also known as quadrant) analysis of the Reynolds stress is useful to quantify the turbulent mechanisms of organized structures in the inertial sublayer and the roughness sublayer (RS), i.e. the region below the inertial sublayer affected by the coherent motions spawned by buildings, trees, etc. upwind (Lu et al., 1973). It applies conditional statistical averaging to partition the contribution of the mean Re ynold stress into four quadrants based on the sign of the (mean removed) longitudinal ( ) and vertical ( ) wind velocity components as shown in Table 2 4. Sweep and ejection events (quadrants two and four) make positive contributions to the Reynolds stress, with

PAGE 25

25 the former associated with downward motion and forward velocity greater than the mean. Quadrants one and three correspond to inward and outward inte raction events and contribute negatively to the shear stress. The contribution of the total Reynolds stress from quadrant is ( 2 4 ) where is the averaging time and is the indicator function. takes a value of one if otherwise, allows examination of large scale and sparse structures (i.e., coherent motions) by eliminating small and frequent c ontributions. The fraction of the shear stress transported is obtained from normalizing Equation 2 4 by the mean Reynolds stress ( 2 5 ) The difference in stress fraction between quadrants two and four ( 2 6 ) expresses the imbalance of sweep and ejection events (> 0 = sweep dominance). Raupach (1981) established a relation between and the third moments (skewness) of the fluctuations in the velocity components and which can be derived by a cumulant discard method ( Antonia and Atkinson, 1973 ; Nakagawa and Nezu 1977 ). For the theoretical difference in all stress contributions in the sweep and ejection quadrants ( ) can be expressed as

PAGE 26

26 ( 2 7 ) where, ( 2 8 ) ( 2 9 ) In Equations 2 8 and 2 9 is the correlation coefficient and represent the third moments ( , ). Boundary Layer Wind Tunnel M easurements FCMP field observations were compared against flow measurements in the which is a low speed open circuit tunnel with dimensions of 6 m W x 3 m H x 38 m L. Comparison of field and wind tunnel observations was conducted during a validation study of the Terraformer which is an automated roughness element grid that rapidly reconfigures the height and orientation of 11 16 roughness elements in a 62 X 18 grid to achieve desired upwind terrain conditions. The grid extends nominally 18.3 m along the length of the tunnel (Figure 2 1). Dimensions of the elements are 5 cm by 10 cm, and they are spaced 30 cm apart in a staggered pattern. Height and orientat ion can be varied from 0 160 mm and 0 360 degrees, respectively. Two orientations were applied. The narrow and wide edge cases refer to the 5 cm and 10 cm face perpendicular to the flow, respectively. Spatially averaged semi logarithmic mean velocity profi les are shown in Figures 2 2. The profiles were measured over the last row of the roughness elements closest to the test section, at location = 29.5 m ( Figure 2 1), using an automated gantry system

PAGE 27

27 instrumented with four Turbulent Flow Instrumentation Cobra pressure probes that measure , and velocity components and static pressure within a 45 acceptance cone. Response characteristics include a maximum frequency response of 2 kHz and a 2 100 m/s sensing range. Three vertical traverses were tak en across the width of the tunnel at the centerline and 500 mm off the centerline of the tunnel. Aerodynamic parameters and values (Table 2 5) were estimated from a non linear least squares fit of the log law to the mean velocity profile in the in ertial sublayer (ISL) region ( 150 900 mm), following the curve fitting method in Karimpour et al. (2012). Zero plane displacement height estimates were based on the geometric arrangement of the roughness array following MacDonald et al. (1998): ( 2 10 ) where is the plan area index defined as the ratio of the lot area to the plan area of the roughness element, and is a constant taken as 4.43 the recommended value for a staggered array. The reference mean wind velocity was measured at a height 1. 48 m Plots of BLWT longitudinal turbulence spectra at a height = 610 mm are shown in Figure 2 3 for the six corresponding mean velocity profiles. Measured data were compared with the power spectra model in ESDU 74030 and 74031 (1974), which was first derived by von Krmn (1948) for isotropic turbulence: ( 2 1 1 )

PAGE 28

28 where is the power spectral density function, is the frequency, is the longitudinal mean velocity, and is the longitudinal integral length in Bendat and Piersol (2000) Time histories were segmented into eight contiguous blocks, and a Hanning tapering window with 50% overlap was applied to suppress side lobe leakage. These plots are representative of measurements obtained between 320 860 mm, outside of the immediate influence of the individual roughness elements. In all cases in this range, similitude was achieved. Results Higher Order Central M oments Figures 2 4 and 2 5 show semi logarithmic whisker subplots of the mean longitudinal skewness ( ) and kurtosis ( ) at elevations of 5 m and 10 m for three mean wind speed ranges for data collected in the field. The whisk ers bound the 25th and 75th quantiles, and the letters B, C and D inside the whisker box indicate the exposure category. An increase in skewness with terrain roughness is evident for all wind speed ranges. Skewness values at 5 m are consistently larger tha n 10 m values for the three exposures. Records with mean wind speeds above 25 m/s in exposure B (suburban) show mean skewness values of 0.49 and 0.37 for elevations of 5 m and 10 m, respectively. Table 2 6 summarizes the statistics of all 5 m and 10 m 15 min records based on exposure category and mean wind speed. The table shows that for a given terrain exposure, the mean may be assumed to be invariant to mean wind speed No significant change in kurtosis was observed.

PAGE 29

29 Reynolds Stress F ractions In Fi gure 2 6, the average Reynolds stress fractions are shown as a function of for and exposure categories Observations at 10 m (top three subplots) display a slight decrease in ejection contributions with increasing roughness. At 0 ejection contributions in higher winds ( m/s) account for ~60, ~70, and ~82% of the total mean stress for exposures B, C, and D, respectively. Sweep events appear to contribute equally for the three exposure categories at = 10 m. The 10 m subplots in Figure 2 6 also suggest a minor reduction in the contribution of inward/outward interaction events in more built up terrain. At 5 m, a significant increase in sweep contributions is observed for exposure B (suburban) in high winds ( m/s), comprisin g over 95% of the total mean stress for Sweep contributions are still present for larger hole sizes. At mean associated with sweeps account for ~18% ( m/s), while ejection events appear to vanish for The two interaction ev ents at 5 m display greater contributions to the Reynolds stress in rougher terrain for m/s (bottom right subplot). Subplots of spatially averaged Reynolds stress fractions from BLWT simulation are shown in Figure 2 7. Two elevations (above the tun nel floor) where normalized by the six roughness length estimates presented in Table 2 5. Large sweep contributions (greater than 90%) for = 0 can be observed for element heights of 40, 60 and 100 mm, all oriented with the narrow edge perpendicular t o the flow. For = 60 mm, a height of z = 5 m will result in a roughness length of 0.36 m ( = 14) which would classify as Exposure B in ASCE 7 10. Additionally, sweep contributions are still present up to = 20 for these three element configurati ons. For = 80 mm, sweep

PAGE 30

30 contributions account for 80% of the Reynolds stress at = 0. However, ejections at H = 0 are approximately 60%, which results in sweep dominance. Subplots for the two wide edge element orientations (top two subplots) show sweep contributions around 80% for = 0. However, capturing the coherent structures in the roughness sublayer is more challenging as the element height approaches zero (i.e., flush tunnel floor) i.e., = 0 mm. Relation of Stress Fraction Difference and S kewness Figure 2 8 plots the theoretical stress fraction difference obtained from Equation 2 7 against the measured values i.e., Equation 2 6 for three wind speed ranges. Observations at both 5 m and 10 m heights show good agreement of measured with values found from Eq uation 2 7 The bulk of records associated with exposure B (suburban) favor positive values of hence suggesting swee p dominance. Figure 2 9 plots the longitudinal skewness values against values The six scatter plots exhibit a positive linear trend between and Proportionality coefficients obtained from linear regression analysis are included in the subplots. Slightly higher linear coefficient values are found at 5 m heights for all three wind speed regimes. This suggests greater sweep dominance at 5 m than over 10 m heights. Linear coefficient values for records with m/s closely match prop ortionality values of and established in Raupach (1981). However, the coefficient decreases for 25 m/s, with values of 0.33 and 0.27 for 5 m and 10 m observations, respectively. Scatter plots in Figure 2 10 show the relation of the vertical skewness and An approximate negative linear relation between and is observed. Oikawa

PAGE 31

31 and Meng (1995) detected similar trends from field observations at heights above the ground near the suburban canopy region. In addition, vertical skewness values in Figure 2 10 compare well with calculated values from Raupach et al. (1996), which ranged from 0.5 to 1.0 within a vegetated canopy layer. A robust multilinear regression was performed to the scattered data using an iteratively r e weighted least squares method with a bi square weighting function. The linear coefficients are included in the subplots. No prior work was found for comparison of regression analysis coefficient for Nevertheless, it can be deduced from Figure 2 9 that sweep events in suburban regions are predominantly associated with negatively skewed distributions of the vertical wind component in strong winds. Table 2 7 summarizes results presented in this section, which reveal the increasing trend of sweep dominance in rougher upwind terrain. Measured values exceed 20% for the three wind speed ranges in Exposure B at = 5 m. However, Exposures C and D show mean stress fraction differences in the range of 2% 12% for the same measurement height. Furthermore, negative values i.e., ejection dominance are observed for two wind speed ranges collected at 10 m in Exposure D. Discussion Effects of observed non Gaus sian trends within the roughness sublayer on suburban exposure coefficient values ( values in ASCE 7 10) are now examined. Figure 2 11 includes the proposed non Gaussian exposure coefficient profile for suburban exposure based on linearly vary ing values with height above ground level i.e., the skewness profile The proposed non Gaussian exposure coefficient is calculated from

PAGE 32

32 ( 2 12 ) in which is the 3 s gust velocity at 10 m in open terrain ( 0.02 m), is the gust velocity profiles obtained from where is the mean velocity profile (Equation 2 2). The shear velocity in Equation 2 2 is determined for a given gradient wind speed based on the relation between and described in Irwin (2006). The turbulent intensity is computed from Harris and Deaves (1980) variance model Balderrama et al. (2012) demonstrated that the non Gaussian peak factor model in Kareem and Z hao (1994, henceforth KZ94) is more accurate than the Gaussian model defined in Davenport (1964, henceforth D64). The non Gaussian peak factor from KZ94 can be expressed as where is the crossing rate, and is the duration of the record. The parameters , and are dependent on the third (skewness) and fourth (kurtosis) moments (KZ94). In the special case where and Equation 2 14 takes the form of the Gaussian D64 peak factor model. A constant kurtosis value of = 3.0 was selected in Equation 2 14 to generate values. A value of = 0.15 m was chosen for the proposed profile. This ( 2 13 ) ( 2 14)

PAGE 33

33 roughne ss length corresponds to the lower limit value specified in ASCE 7 10 (2010) for suburban terrain The (linear) skewness profile shown in Figure 2 11 (left subplot) was constructed from mean values of FCMP observations at 5 m, 7.5 m, 10 m, 12.5 m, and 15 m for suburban exposure ( = 0.15 m 0.7 m). Full scale values in Figure 2 10 are limited to 15 min records collected from 2012 2016 using ultrasonic (US) anemometry Table 2 1. Only records satisfying 10 m/s were considered. Horizontal w hiskers at the three FCMP elevations indicate the 25 th and 75 th quantiles. Below = 4.6 m, skewness values are kept constant ( = 0.46). BLWT skewness profiles from the six independent tests summarized in Table 2 5 are also shown in Figure 2 11 (lef t subplot). Most of the BLWT data show skewness values in the range of 0.20 0.5 up to 5 m full scale elevation. Above 5 m, the BLWT data exhibit a noticeable decreasing trend towards zero skewness. Table 2 8 summarizes proposed skewness values and non Gau ssian values for heights up to 24.4 m. The table shows a ~14% increase in suburban exposure coefficient values at 10 m from values provided in ASCE 7 10 (2010). Summary Analysis of anemometric observations collected during 1999 2016 Atlantic hur ricane seasons, along with BLWT data, exhibit non Gaussian wind behavior in suburban terrain exposures within the suburban roughness sublayer. Observations in suburban terrain predominantly show positively skewed distributions of the along wind component, thus increasing the probability of peak gust events. In addition, field data suggests a dependence of longitudinal skewness with height in rougher terrain. Positive

PAGE 34

34 skewness values appear to increase with decreasing elevation near the suburban canopy. BLWT studies for suburban terrain also seem to follow similar tendencies. Quadrant analysis reveals the linkage between positive values of the longitudinal skewness and downward transfers of momentum (sweep events). Most of these records are associ ated to rougher terrain conditions (suburban). However, sweep dominance appears to be less pronounced in higher winds. Positively skewed distributions measured within the roughness sublayer were incorporated into exposure coefficient calculations for subur ban terrain through a non Gaussian peak factor model. A non trivial increase in suburban exposure coefficient values was found.

PAGE 35

35 Table 2 1. FCMP datasets included in this study in addition to those analyzed in Balderrama et al. (2012). Year Tropical Cycl one Station ID Latitude Longitude Town Anemometry 2016 Matthew T1 +28.1947 80.5944 Satellite Beach, FL 3 Axis Propeller 2016 Matthew T2 +32.7141 79.9664 Charleston, SC Ultrasonic 2016 Matthew T2 +27.1889 80.2411 Stuart, FL Ultrasonic 2016 Matthew T3 +32.7902 79.9881 Charleston, SC Ultrasonic 2016 Matthew T3 +28.1937 89.6056 Satellite Beach, FL Ultrasonic 2016 Matthew T5 +27.1802 80.2295 Stuart, FL 3 Axis Propeller 2016 Hermine T2 + 29.6735 83.3746 Steinhatchee, FL Ultrasonic 2016 Hermine T3 + 29.6731 83.3798 Steinhatchee, FL Ultrasonic 2016 Hermine T5 + 29.6728 83.3703 Steinhatchee, FL 3 Axis Propeller 2014 Arthur T2 + 35.2322 75.6215 Hatteras, NC Ultrasonic 2012 Sandy T3 + 39.3208 74.5953 Linwood, NJ Ultrasonic 2012 Isaac T3 + 29.6487 90.6940 Houma, LA Ultrasonic 2012 Isaac T2 + 29.5385 89.7751 Bohemia, LA Ultrasonic Table 2 2. Total number of 15 min records ( = 10 m) stratified into mean wind speed and roughness length based on ASCE 7 10 exposure categories Terrain Type ASCE 7 Exposure Roughness Length Regimes No. of 15 min data segments WS1 WS2 WS3 Total Suburban B 0.15 m > 507 233 105 845 Open country C 0.01 m > 483 350 760 1593 Very smooth D < 0.01 m 73 129 499 701 Total 1063 712 1364 3139 Note: WS1 = 5 m/s <15 m/s, WS2 = 15 m/s <25 m/s, and WS3 = 25 m/s Table 2 3. Mean roughness lengths stratified into mean wind speed and roughness length ( = 10 m) based on ASCE 7 10 exposure categories. Terrain Type ASCE 7 Exposure Roughness Length Regimes Mean Roughness Length, (m) WS1 WS2 WS3 Suburban B 0.15 m > 0.499 0.536 0.423 Open country C 0.01 m > 0.119 0.095 0.064 Very smooth D < 0.01 m 0.005 0.004 0.004 Note: WS1 = 5 m/s < 15 m/s, WS2 = 15 m/s <25 m/s, and WS3 = 25 m/s

PAGE 36

36 Table 2 4. Quadrants of events contributing to the total mean Reynold stress. Quadrant ( ) Event 1 Outward interactions 2 Ejections 3 Inward interactions 4 Sweeps Table 2 5. Equivalent full scale roughness lengths obtained from suburban terrain BLWT simulation. BLWT Test No. Terraformer Configuration BLWT Aerodynamic Parameters Equivalent Full Scale Roughness Length Roughness Element Orientation Element Height (mm) Geometric Scale (m) 1 Wide 20 29.9 0.131 0.021 400 0.171 2 Wide 30 16.4 0.131 0.076 100 0.227 3 Narrow 40 19.7 0.131 0.017 300 0.208 4 Narrow 60 16.5 0.131 0.037 120 0.267 5 Narrow 80 14.2 0.131 0.066 70 0.371 6 Narrow 100 12.5 0.131 0.095 60 0.572 Table 2 6. Statistics of 15 min records stratified into mean wind speed and roughness length based on ASCE 7 10 exposure categories ASCE 7 10 Exposure range (m/s) (m/s) 5 m 10 m 5 m 10 m 5 m 10 m 5 m 10 m B 5 <15 8.2 9.9 0.33 0.31 0.50 0.43 3.28 3.10 15 < 25 17.8 20.8 0.31 0.30 0.55 0.47 3.28 3.05 25 28.6 32.8 0.29 0.28 0.53 0.41 3.24 2.97 C 5 <15 9.3 10.5 0.25 0.23 0.36 0.30 3.06 2.97 15 < 25 18.6 20.7 0.23 0.22 0.39 0.33 3.06 2.96 25 35.6 40.3 0.21 0.20 0.33 0.25 2.99 2.89 D 5 <15 9.5 10.8 0.18 0.16 0.24 0.17 2.97 3.00 15 < 25 18.2 20.5 0.19 0.17 0.28 0.24 2.99 2.99 25 37.7 42.6 0.17 0.15 0.21 0.17 2.93 2.90

PAGE 37

37 Table 2 7. Mean stress fraction difference between sweep and ejection stratified into mean wind speed and roughness length based on ASCE 7 10 exposure categories. ASCE 7 Exposure range (m/s) Measured Theoretical 5 m 10 m 5 m 10 m 5 m 10 m B 5 <15 0.48 0.41 0.202 0.153 0.192 0.146 15 < 25 0.49 0.41 0.229 0.161 0.222 0.155 25 0.49 0.37 0.239 0.132 0.227 0.126 C 5 <15 0.31 0.28 0.093 0.090 0.094 0.087 15 < 25 0.36 0.31 0.120 0.091 0.114 0.088 25 0.31 0.24 0.090 0.058 0.088 0.058 D 5 <15 0.24 0.17 0.004 0.012 0.004 0.003 15 < 25 0.26 0.22 0.036 0.024 0.038 0.024 25 0.21 0.17 0.021 0.005 0.026 0.000 Table 2 8. Proposed exposure coefficient values obtained from non Gaussian peak factor model Height above ground level, Peak factor, Suburban exposure coefficient, % Increase ft (m) Proposed ( ) ASCE 7 10 Exposure B 15 4.6 0.46 3.48 0.63 0.57 +9.8 20 6.1 0.43 3.44 0.70 0.62 +12.3 25 7.6 0.39 3.40 0.75 0.66 +13.5 30 9.1 0.35 3.36 0.80 0.70 +14.0 40 12.2 0.27 3.28 0.86 0.76 +13.9 50 15.2 0.20 3.21 0.91 0.81 +13.0 60 18.3 0.12 3.13 0.95 0.85 +11.7 70 21.3 0.04 3.05 0.98 0.89 +10.2 80 24.4 0.00 3.01 1.01 0.93 +9.3

PAGE 38

38 Figure 2 1. Plan view of the BLWT at the University of Florida, illustrating the two element orientations considered for this study, namely wide and narrow edge windward.

PAGE 39

39 Figure 2 2. Spatially averaged mean velocity profiles of six independent BLWT simul ations measured at location = 29.5 m.

PAGE 40

40 Figure 2 3. Longitudinal turbulence spectra of six independent BLWT simulations, measured at location = 29.5 m ( = 610 mm)

PAGE 41

41 Figure 2 4. Observed mean (longitudinal) skewness values stratified into mean wind speed and terrain roughness. Figure 2 5. Observed mean (longitudinal) kurtosis values stratified into mean wind speed and terrain roughness.

PAGE 42

42 Figure 2 6. Mean Reynolds stress fractions for mean wind speed values stratified into mean wind s peed and terrain roughness for the field observations.

PAGE 43

43 Figure 2 7. Spatially averaged Reynolds stress fractions of six independent BLWT simulations measured at location = 29.5 m.

PAGE 44

44 Figure 2 8. Theoretical and measured stress fraction difference between sweep and ejection quadrants for mean wind speed values stratified into mean wind speed and terrain roughness. Suburban (Exposure B) measurements are denoted by black circular markers.

PAGE 45

45 Figure 2 9. Measured stress fraction difference versus longitudinal skewness for mean wind speed values stratified into mean wind speed and terrain roughness. Suburban (Exposure B) measurements are denoted by black circular markers.

PAGE 46

46 Figure 2 10. Measured stress fraction difference versus vertical skewness for mean wind speed ranges: (a) 5 m/s 15 m/s, (b) 15 m/s 25 m/s, and (c) greater than 25 m/s. Suburban (Exposure B) measurements are denoted by black circular markers.

PAGE 47

47 Figure 2 11. Proposed non Gaussian exposure coefficient versus elevation above ground

PAGE 48

48 CHAPTER 3 WIND SIMULA TION IN A LARGE BOUNDARY LAYER WIND TUNNEL This chapter presents results from a series of flow measurements in a large boundary layer wind tunnel for validation of a computer controlled terrain generator and gives a detailed description of the flow conditioning and mixing devices of the BLWT instrumentation utilized for flow m easurements, and validation models for adequate simulation of the atmospheric boundary layer (ABL) for testing of building models. Background Despite significant advancements in computational modeling of boundary layer flows (e.g., CFD, LES), boundary layer wind tunnel (BLWT) testing still remains an indispensable tool for simulation of complex turbulent flows. BLWTs require adequate simulation of the atmospheric boundary layer (ABL) the lowest part of the atmosphere where the air flow is influenced by the Earth Although there is ongoing debate about what characteristics of the ABL should be match in wind tunnels, statistics of primar y interest for wind engineering applications include: (1) mean longitudinal velocity component, (2) longitudinal turbulence intensity, (3) power spectral densities, and (4) integral scales in the along, across, and vertical directions. Satisfying similarit y requirements of ABL characteristics and test models is still a topic of discussion among wind tunnel modelers. Properly scaling the mean velocity and turbulence characteristics of the ABL to match a scaled model is typically a trial and error process. T he natural scaling of the ABL in wind tunnels is in the range 1:400 to 1:600 ( Davenport, 2007 ) thus test models should be built at similar geometric scales. This is readily achievable for tall building models. However, low rise models with detailed featur es require larger geometric scales ranging (from 1:20 to 1:100) to better

PAGE 49

49 capture flow separation regions as well as wind effects on building features such as parapets and chimneys. Additio nal flow modification devices are used to grow the ABL (i.e., scale up the approach flow) and achieve better agreement between the scale of th e model and the approach flow. M odification devices such as barriers, vortex generators, and spires have been employed to stimulate the growth of turbulent boundary layers. Approach to M odeling ESDU M odel The turbulent structure of the ASL, in strong winds, can be described by the boundary layer model developed in Harris and Deaves (1981), which introduces the s surface. Near the ( 3 1 ) w here is the longitudinal wind velocity (typically mean hourly) at elevation is the frict ion (shear) velocity and is the aerodynamic roughness length. The longitudinal turbulence intensity is defined as the standard deviation of the velocity ( ) divided by the mean velocity and is expressed analytically as ( 3 2 ) where is the Coriolis parameter = in which rad/s and = latitude in degrees.

PAGE 50

50 If the turbulence intensity and mean velocity are known for a specific height and can be obtained through a bounded nonlinear function minimization that simultaneously satisfies Equations 3 1 and 3 2. This procedure excludes the zero plane displacement height, which is a n important parameter for more built up terrain (e.g., suburban). Morphologic M odels for Roughness L ength E stimates A wide range of approaches can be found in literature (e.g., Lettau, 1969; Counehan, 1971; Macdonald, 1998 ) to estimate surface roughness from the geometric arrangement of upstream obstacles Lettau (1969 ) suggested a simple expression for estimating the roughness length for a uniform array of obstacles ( 3 3 ) w here is the mean obstacle height, is the total frontal area of the obstacle, and is the total plan area of the obstacles. This method for estimating roughness length is included in the Commentary section of ASCE 7 10 (2010). Equation 3 3 only applies for roughness area densities ( i.e., ) below 20 30% (Macdonald et al., 1998). Mac donald et al. (1998) presented an improved method for the estimation of surface roughness, which accounts for the nonlinear decrease of roughness length at high roughness densities ( i.e., 20% ). Additionally, the model explicitly includes the drag co efficient ( ) of the obstacles, based on the shape and layout of the roughness array, and the zero plane displacement height ( ) This results in the following expression:

PAGE 51

51 ( 3 4 ) where is a multiplicative factor, which may be calibrated against experimental data. The displacement height ( ) in which is the plan area index defined as the ratio of the plan area of the roughness obstacles and Methodology Wind T unnel Tunnel (BLWT), a low speed open circuit tunnel with dimensions of 6 m W x 3 m H x 38 m L Figure 3 1 shows a schematic plan drawing of the tunnel. The inlet of the tu nnel houses eight vane axial fans, each driven by a 75 hp (56 kW) electric motor. Flow conditioning system consists of five sets of screens and honeycombs, located approximately 3 m downwind of the fans (Figure 3 2). The test section (1 m turntable) is loc ated 31.5 m from the fan bank. The air speed through the test section can be set from 1 to 16 m/s (measured at a reference height). The tunnel is instrumented with reference anemometry (Figure 3 3) for static pressure measurements at specific locations alo ng the streamwise direction. Reference velocity in the tunnel is taken at an elevation of 1.48 m above the floor using Pitot tubes (Figure 3 4 ) located approximately 2.5 m upwind of the center of the turntable. The ceiling of the tunnel is adjustable, whi ch allows for zero pressure gradient along the length of the test section. Data is collected from a meteorological station (Figure 3 5 ) is in close proximity to the exit of the tunnel for monitoring of local atmospheric conditions.

PAGE 52

52 Roughness Element G rid a nd Vortex G enerators Passive mixing devices in the tunnel include Irwin spires, downwind of the screens and honeycombs (Figure 3 6 ), and a computer controlled roughness array named the Terraformer, a 62 X 18 roughness element grid that rapidly reconfigures the height and orientation of 1116 roughness elements to achieve desired upwind terrain conditions. The grid extends nominally 18.3 m along the length of the tunnel Dimensions of the elements are 5 cm by 10 cm, and they are spaced 30 cm apart in a stagge red pattern. Height and orientation can be varied from 0 160 mm and 0 360 degrees, respectively. The narrow and wide edge cases refer to the 5 cm and 10 cm face perpendicular to the flow, respectively. The height and orientation of each element in the roug hness grid is controlled through LabView software. Consequently, the Terraformer can readily simulate an extensive series of homogeneous and heterogeneous terrains. Reconfiguration of all 1116 elements is typically achieved in less than 60 seconds. Flow M e asurements An automated (computer controlled) gantry system was used for three dimensional mapping of the flow. The gantry traverses four velocity probes in the three orthogonal directions (i.e., three translational degrees of freedom) of the tunnel, namel y along ( ), across ( ) and vertical ( ). The position and probe parameters (e.g., sampling rate and duration) is controlled through LabView code. Velocity sensors consist 3 hole (i.e., taps) Cobra pressure probes from Turbulent Flow Instrumentation (TF I). The probes measure , and velocity components and static pressure within a 45 acceptance cone. Response characteristics include a maximum frequency response of 2 kHz and a 2 100 m/s sensing range. Accuracy of the probes is typically between 0.5

PAGE 53

53 m/s, although this is dependent on the turbulence levels. The probes remain relatively accurate for turbulence intensities greater than 30%. Pressure data from the four taps of the probe are measured by a data acquisition (DAQ) system and processed by the TFI Device Control software to resolve the three components of the flows and static pressure measurements. Experimental Design For this experiment, 32 homogeneous terrain configurations were examined. The roughness element grid was raised (uniformly) from 10 160 mm, using 10 mm increments. Vertical and lateral traverses were taken for each element height increment for both narrow and wide edge windward element orientations. Time histories of the , and components of the approach flow were measured at the test section ( = 31.52 m).. Vertical traverses consisted of 44 measurements along the height of the tunnel from = 5 16 7 0 mm using cobra probes. Three lateral traverses were taken across the width of the tunnel at the centerline and 500 mm off the centerline of the tunnel. A triple rotation procedure described in Foken and Nappo (2008) was performed to align the probe coordinate system into the streamlines and towards the mean flow coordinate system. Measurements for each probe position were collected for 30 seconds at a sampling rate of 1250 Hz. Results and Discussion Mean Velocity P rofiles Figure s 3 7 and 3 8 show subplots of spatially averaged mean velocity profiles for a narrow and wide edge windward element orientation, respectively. Mean velocities are normalized by the reference wind velocity at a height of = 16 7 0 mm The reference wind velocity appeared to mildly increase with increasing element height A

PAGE 54

54 noticeable difference in the shape of the profiles between wide and narrow edge windward element orientations is observed for high element heights ranging from 80 to 160 mm. The profi le curves in Figure 3 8 (wide case) contain inflection points at approximately 500 mm for high element height. This is not observed in the narrow case. Furthermore, a small inner boundary layer (IBL) appears to be forming close to the tunnel. This might be a result of the rough to smooth transition between the test section and the last row of roughness elements Semi logarithmic mean velocity profiles are depicted in Figure s 3 9 and 3 10 for both element orientations. Roughness length ( ) and shear velo city ( ) values were obtained from a non linear least square fit of the logarithmic law (Equation 3 1) following Karimpour et al. (2012). The zero plane displacement height ( ) was estimated based on the morphometric models of roughness arrays found in Macdonald et al. (1998). The logarithmic fits were limited to data points in the inertial sublayer (ISL), above the wake region. Aerodynamic parameters for 16 element heights and two element orientations are presented in Table 3 1. For the same element height, larger roughness length estimates and shear velocities were obtained from the wide edge windward element orientation. Turbulence intensity P rofiles T urbulence intensity pro files of the longitudinal velocity component ( ) f or a narrow and wide edge windward element orientation, respectively are shown in Figure s 3 11 and 3 12 The profiles suggest that a greater range of turbulence inte nsity levels can be generated by orienting the roughness elements in a wide edge windward manner For an element height of 160 mm (wide), values can exceed 30% below

PAGE 55

55 500 mm. In comparison, maximum turbulence levels for the narrow case are around 22% at elevations below 200 mm. Figures 3 1 3 and 3 1 4 include profiles of the lateral turbulen ce component ( ). Below = 100 mm the wide element orientation can achieve levels ranging from 7 29%, while the narrow case can generate lateral turbulence levels between 8 23%. Vertical turbulence intensity ( ) profiles are shown in Figures 3 1 5 and 3 1 6 for the two element orientations. Maximum values for the wide edge orientation occurs at approximately 250 mm above the floor. The peak for the narrow case happens at = 100 mm Summary This chapter presents results from an extensive series of flow measurements in a large boundary layer wind tunnel (BLWT) for validation of the Terraformer, a compu ter controlled roughness element grid. Three dimensional instantaneous velocities were measured using an automated gantry, instrumented with Cobra probe sensors, for a wide range of homogeneous terrain configurations. Aerodynamic parameters, such as shear velocity and roughness length, were obtained from a non velocity profiles. Longitudinal turbulence intensity values in the range of 6 to 32 were observed at elevations below 500 mm from the tunnel floor. The wide edge windward element orientation generates higher turbulence levels.

PAGE 56

56 Table 3 1. BLWT aerodynam ic parameters. (mm) (m) (m/s) (m/s) ( = 1.67 m) Narrow Wide Narrow Wide Narrow Wide 10 0.0018 0.00291 0.883 0.966 15.290 15.508 20 0.0026 0.00564 0.944 1.094 15.415 15.594 30 0.0041 0.01075 1.014 1.246 15.465 15.708 40 0.0070 0.02250 1.135 1.480 15.536 15.712 50 0.0081 0.03649 1.172 1.678 15.573 15.730 60 0.0121 0.04832 1.275 1.825 15.486 15.794 70 0.0136 0.06327 1.315 1.990 15.616 15.980 80 0.0153 0.06885 1.348 2.039 15.628 15.966 90 0.0194 0.08751 1.433 2.237 15.703 16.069 100 0.0213 0.09081 1.466 2.250 15.686 16.126 110 0.0244 0.09406 1.522 2.290 15.778 16.252 120 0.0305 0.10140 1.619 2.349 15.868 16.271 130 0.0336 0.11345 1.663 2.446 15.804 16.298 140 0.0354 0.10582 1.692 2.327 15.858 16.250 150 0.0410 0.11514 1.769 2.438 15.838 16.322 160 0.0453 0.13393 1.832 2.587 15.933 16.279

PAGE 57

57 Figure 3 1. Boundary layer wind tunnel (BLWT) at the University of Florida (Photo courtesy of John Jernigan) Figure 3 2. Set of screens and honeycomb system downwind of the fan bank (Photo courtesy of author)

PAGE 58

58 Figure 3 3. Pitot tube mounted to the wall of the tunnel (located at the test section ) (Photo courtesy of author) Figure 3 4. Freestream reference anemometry (Pitot tubes) (Photo courtesy of author)

PAGE 59

59 Figure 3 5. Meteorological weather station located adjacent to the exit of the BLWT (Photo courtesy of author). Figure 3 6. Irwin spires located downwind of the screens and honeycomb (Photo courtesy of author)

PAGE 60

60 Figure 3 7. Longitudinal Mean velocity p rofiles for a narrow edge windward element orientation Figure 3 8. Longitudinal Mean velocity profiles for a wide edge windward element orientation

PAGE 61

61 Figure 3 9 Semi logarithmic mean velocity profiles for the narrow ed ge windward element orientation. Figure 3 10 Semi logarithmic mean velocity profiles for the wide ed ge windward element orientatio n.

PAGE 62

62 Figure 3 11 Longitudinal turbulence intensity profiles for a narrow edge windward element orientation Figure 3 1 2 Longitudinal turbulence intensity profiles for a wide edge windward element orientation

PAGE 63

63 Figure 3 1 3 Lateral turbulence intensity profiles for a narrow edge windward element orientation Figure 3 1 4 Lateral turbulence intensity profiles for a wide edge windward element ori entation

PAGE 64

64 Figure 3 1 5 Vertical turbulence intensity profiles for a narrow edge windward element orientation Figure 3 1 6 Vertical turbulence intensity profiles for a wide edge windward element orientation

PAGE 65

65 CHAPTER 4 SURFACE PRESSURES ON LOW RISE BUILDINGS This chapter presents the research design and subsequent findings from a comprehensive series of experiments in a large BLWT to invest igate the variation of pressure coefficients with increasing surface roughness on low rise buildings. Reduced geometric scale models of the Wind Engineering Research Field Laboratory ( WERFL ) experimental building ( Tieleman et al., 1996 ) were subjected to a wide range of turbulent boundary layer flows, through precise adjustment of a computer control terrai n generator called the Terraformer. The Terraformer allows rapid reconfiguration of upwind terrains ranging from marine to dense suburban exposures. Surface p ressures on low rise models were measured downwind of the Terraformer for a large series of upwind terrains, with particular attention to regions prone to flow separation Background Significant research was conducted in the 1990s to assess uncertainties associated with wind tunnel tests of low rise structures, and their accuracy in replicating full scale surface pressures. Tieleman (1992) compared BLWT simulations with full scaled pressure measurements from the Wind Engineering Research Field Laboratory (WERFL) experimental building at Texas Tech University. Th e study examines the importance of generating small scale turbulence in the incident BLWT flow, to properly simulate peak full scale suction pressures, present near the edges and roof corners. This and other studies (e.g., Tieleman and Reinhold, 1978 ; Hill ier and Cherry, 1981 ; Gartshore, 1984 ) argue d that adequate matching of the turbulence intensity takes priority over simulating

PAGE 66

66 the shape of the mean velocity profiles, given that vortex formation on roof edges and corners are primarily controlled by the p resence of the small scale turbulence in the incident flow. In addition, several authors (e.g., Stathopoulos and Surry, 1983; Meecham et al., 1991; Lin et al., 1995 ) have recommended the use of larger models to better duplicate full scale turbulence charac teristics. However, most boundary layer tunnels are incapable of accommodating large low rise models without causing excessive blockage effects (Krishna, 1995). One governing factor for assessing wind loads on low rise buildings is the effect of upwind te rrain conditions. The shape and turbulence characteristics of the approaching boundary layer flow is strongly dependent on the level of surface roughness upwind of the building. The s urface roughness is defined by the s surface such as buildings and trees, which retards the along wind flow. In rougher terrain, the longitudinal component of the mean velocity tends to slow down, while the along wind turbulence intensity increases (i.e. the coefficient of variation) Wind load provisions, such as ASCE 7 10 (2010) introduce exposure coefficient profiles (Irwin, 2006) in design wind pressure calculations to account for the variation of wind velocity with terrain and elevation above ground. In the case of low rise buildings, a single exposure coefficient is, in most cases, applied to design pressures acting on roof and walls. Therefore, it is implicitly assumed that the effect of terrain exposure on design pressures is the same for different regions on the building surface. Ho wever, studies (e.g., Tieleman et al., 1997 ; Pierre et al., 2005 ) have shown significant variation on peak pressure coefficients close to flow separation regions, such as roof edges and corners, with increasing turbulence.

PAGE 67

67 The bulk of wind tunnel studies assessing extreme surface pressures on buildings have been limited to two terrain conditions, namely open and suburban (e.g., Hunt, 1982 Pierre et al., 2005 ). In boundary layer wind tunnels, simulation of these two terrains is typically attained by manual modification of upwind terrain via mixing devices e.g., vortex generators, roughness grid, barriers until aerodynamic parameters, such as roughness length, closely match predefined values from wind load provisions. This study is the first of its kind to systematically characterize both the flow and pressure coefficient distribution on building models. Simulation of Approach Flow BLWT experiments were conducted at the University of Florida (UF) Natural Hazard Engineering Research Infrastructure (NHERI) Exp erimental Facility. The BLWT at UF is a low speed open circuit tunnel with dimensions of 6 m W x 3 m H x 38 m L ( Chapter 4 ). Simulation of terrain roughness is performed via the Terraformer, an automated roughness element grid that rapidly reconfigures the height and orientation of 1116 roughness elements in a 62 X 18 grid to achieve desired upwind terrain conditions. The grid extends nominally 18.3 m along the length of the tunnel. Dimensions of the elements are 5 cm by 10 cm, and they are spaced 30 cm apa rt in a staggered pattern. Height and orientation can be varied from 0 160 mm and 0 360 degrees, respectively. The presents results from an extensive series of boundary layer flow measurements at the test section of the tunnel.

PAGE 68

68 Low Rise Building Models Pressure tests were conducted on three rigid building models of Wind Engineering Research Field Laboratory ( WERFL ) experimental building with geometric scales of 1:2 0 1: 30 and 1: 50 Multiple model scales were tested to investigate the effects of reattachment regions. The three models are identical but scale by a constant geometric multiplier The models were instrumented with 26 6 pressure taps on the four walls and roof. The tap loc ation follows the same layout as UWO (Pierre et al., 2005) however 6 0 additional pressure taps were added on the roof of the model ( Figure 4 1 ) The roof pitch is 1:48 for all models. The model s were placed at the center of the test section. Pressure Mea surements Instrumentation Time series of differential pressures acting on each tap were measured using eight 64Px ZOC33 electronic pressure scanning modules (64Px ZOC33, Scanivalve Corp.). Each module consists of 64 channels, and are housed in a rugged stain less steel thermal control unit (TCU), along with an Ethernet remote analog to digital (A/D) module (E RAD 4000 Scanivalve Corp.). The TCU can operate in temperatures ranging from 45C up to 65C. Each ZOC33 module incorpo rates 64 individual silic on pressure sensors, cali bration valving, a high speed multiplexer (45 kHz), and an instrumentation amplifier. Clear urethane tubing (URTH 063 Scanivalve Corp.) were use d to connect each pressure tap to a corresponding channel in a pressure scanning module. The tubes have an outer diameter of 0.086 inches and inner diamet er of 0.054 inches. The length of the tube was 48 in for all taps and model scales.

PAGE 69

69 The use of tubing s ystems to connect pressure taps to the pressure scanners creates distortion of the pressure fluctuations i.e., resulting from resonance and damping effects in the tubing system (Irwin et al., 1979). The distortion depends on several factors including: (1) tube length and diameter, (2) geometry of the pressure scanner internal passageways, (3) and on the transducer internal volume. To correct for distortion effects, measured pressures were digitally filtered using tubing system transfer functions ( Figur e 4 2 ) for several tube lengths. The frequency response function for tubing of a given length is obtained by generating a random input pressure signal containing all of the desired frequencies (i.e., a pink noise signal). The pressure signal is measured at two taps, one connected to a pressure transducer through a minimal length tube (< 6 inches) and the other through a tube of a specific length for which distortion effects need to be corrected. The ratio of the FFTs of the two signals provides the frequenc y response function, which essentially contains factors to adjust the amplitude and phase of the distorted signal at specific frequencies from 0 to half of the sampling frequency, known as the Nyquist frequency. Pressure Coefficients External pressure coe fficients for all experiments were calculated as the ratio of the differential pressure and the velocity (dynamic) pressure at model eave height: ( 4 1 )

PAGE 70

70 where is the (absolute) pressure measured, is the reference (static) pressure, is the air density, and is the mean longitudinal velocity at eave height. Normalization of surface pressures to the velocity pressure at eave height is common practi ce in pressure tests of generic low rise models (e.g., Uematsu and Isyumov, 1999; Pierre et al., 2005). Static reference pressure s ( ) will be taken fro m the static port of the Pitot tube, en suring stable measurements with negligible fluctuations. Air de nsity ( ) will be estimated based on the air temperature, barometric pressures, and relative humidity measured during each test. During testing, were measured indirectly based on Pitot tube measurements. A conversion factor was used to relate th e mean wind speed at Pitot tube height to the model eave height ( Figure 4 3 ). Direct measurement of would require placing a velocity probe (i.e., Cobra Probe ) in close proximity to the model, causing flow distortion and adverse effects of pressure measurements in the model. Additionally, distortion in the flow due to blockage effects (ASCE SEI 49 12) are non existing given the large cross section of the tunnel relative to the model dimensions. Peak Pressures Extreme value analysis was applied to provide a more reliable estimate of peak surface pressures (Lieblein, 1974). A Fisher Tippett Type I also known as Gumbel extreme value distribution has proven to be a useful method for assessing peak pressures in building studies (Mayne and Cook, 1979). The

PAGE 71

71 Gumbel fit models the peak pressures well for a modest number of peak events i.e., less than 100 ( Holmes and Cochran, 2003 ) for a positive or neg ative pressure regime. The procedure for estimating peak pressures can be summarized as follows: (1) Peak (maximum or minimum) values from non overlapping segments are obtained from pressure time series. (2) Peak values are sorted in ascending orde r, and allocated a Gumbel plotting position given by where is the order. (3) The plotting parameter is transformed into a reduced variate: (4) values are plotted against and a linear regression fit is performe d. (5) Mode ( ) and shape ( ) parameters are obtained from the fit. (6) Peak estimates for a given probability of exceedance can be obtained from the Gumbel cumulative distribution function: where is the random variable i.e., peak Experimental Design Three models of the WERFL building were immerse in 33 turbulent flow fields. The approach flow was varied by changing the configuration of the roughness element grid i.e.,Terraformer upwind of th e model. Two element orientations were be considered, namely wide and narrow edge windward. Roughness elements were elevated from 0 160 mm using increments of 10 mm, thus generating 16 upwind terrain conditions for each element orientation for a total of 3 3 terrains. Three wind angles were considered for each terrain configuration and model scale, namely 0, 45, and 90 degrees. Figure 4 4 shows three wind tunnel tests from the 1: 3 0 WERFL model The figure provides a representative case of

PAGE 72

72 the three model or ientations: (1) long building dimension parallel to the wind 0, (2) cornering wind 45, and (3) short building dimension parallel to the wind 90. particular (1) terrain configurati on, (2) model scale, and (3) wind orientation. Sampling rate will be selected according to reduced frequency relation: ( 4 5 ) w here is the sampling rate, is the characteristic length, and is the reference wind velocity for model ( ) and full scale ( ), respectively. For all the tests, t he mean reference wind velocity at Pitot tube height (Figure 4 3 ) was approximately 15 m/s. The characteristic length of the model is directly proportio nal to the geometric scale of the model. The sampling period of each test must be sufficiently long to provide stable estimates of the statistics of surface pressures mean and root mean square ( RMS ). Additionally, peak pressures must provide a representative estimate of full scale intervals of approximately 3600 seconds, because these are used in combination with the statistics of hourly mean wind speeds for calculation of full scale peak pressures ( Davenport 2007 ). Table 4 1 summarizes the sel ected testing parameters for the three model scales. Results and Discussion Validation of Pressure Measurements P ressure coefficients were compared to aerodynamic tests conducted at the University of Western Ontario (UWO) Two terrains were examined on UW WERFL model tests, namely, open ( = 0.01 m) and sparse

PAGE 73

73 suburban ( = 0.087). The measured turbulence intensity at eave height for the two exposures were approximately = 14 and 22%, for open and sparse suburban, respectively. Figure 4 5 show s a comparison of mean pressure coefficients between the 1:100 UWO model and the three models tested in this study A roughness element configuration was selected to closely match the turbulence le vels of UWO (22%). Data points i n Figure s 4 5 correspond to mean pressure coefficients from 206 pressure taps. The three models scales 1:20, 1:30, and 1:50 show reasonable agreements to surface pressures from UWO. However, most of the data points are below the 45 line, indicating that mean pres sures from this study were slightly larger than the tests conducted at UWO. Figure 4 6 include s representative peak (minimum) pressure coefficients corresponding to a 78% probability of non exceedance. The data points in the figure are more scattered compa red to mean pressures. The bulk of the experiments showed a cceptable agreement between UWO and UFL WERFL data when turbulence levels at eave height were matched Illustrative power spectral densities of pressure coefficient time histories for the 1:20 WER FL model (45 degree angle of attack) are depicted in Figure 4 7 The pressure signal corresponds to T ap 216, located at the roof corner closest to the main wind direction (Figure 4 1) The signal was low pass filtered at 150 Hz. A roughness element height was selected for a narrow (left subplot) and wide (right subplot) element orientation to approximately match the longitudinal turbulence at eave height ( ). A greements is observed between UWO and UFL data.

PAGE 74

74 Spatial Distribution of Surface Pressures Contour subplots of m ean pressure s ( ) from four BLWT tests are shown in Figure 4 8 for the 1:20 WERFL model and a wind direction parallel to the long building dimension ( 0) The four subplots correspond to turbulence intensities at eave heigh t of 10.4, 18.7, 25.7 and 27.7%, all achieved from a wide edge windward element orientation. Positive mean pressures of 0.5 are observed near the centroid of the windward wall stagnation point for turbulence levels of 10.4% and 18.7% (top two subplots). H igher turbulence intensities show mean values around 0.75 (bottom subplots) Roof taps close to the windward wall show mean suction (negative) of approximately 1.2 for = 0.10 and 0.30, and 1.75 for h/H = 0.56 and 0.71. Figure 4 9 illustrates th e spatial distribution of root mean squares ( RMS ) of pressures for the 1:50 WERFL model and a wind direction perpendicular to the long building dimension ( 9 0) It is evident from the four subplots that the two roof corners facin g the main wind directi on experience the highest RMS values Maximum RMS pressures range from 1.2 for = 11.7% to 2.2 for = 27.3 % Maps of peak pressure coefficients from four representative BLWT tests are presented in Figure 4 10 for the 1:30 WERFL model and a corneri ng wind angle of 45. Peak pressure values at the roof corner closest to the main wind direction were around 6 for = 0.15, while = 1.06 produced peak pressures exceeding 14. Figures 4 11 4 12 and 4 13 show the distribution of mean pressures along a line of taps in the long building dimension for the 1:20, 1:30, and 1:50 WERFL models, respectively The wind direction is along the tap line ( 0).

PAGE 75

75 Each figure includes 16 curves, one for each element height ( ) oriented in t he wide edge windward element orientation. Turbulence intensity values at eave height ( ) for each elements height are included in the legend of the figure. For the 1:20 model (Figure 4 11 ), mean pressures vary from 0.9 to 1.3 near the leading edge of t he roof (i.e., = 0) for turbulence levels between 8.6 29.1%. Comparable ranges of mean pressures at = 0 can be observed for the 1:30 model (Figure 4 12 ). However, slightly lower values of mean pressures at = 0 were found for the 1:50 WER FL model. This is depicted in Figure 4 13 Mean pressures along a line of taps perpendicular to the ridgeline of the roof for 9 0 are shown in Figures 4 14 4 1 5 and 4 16 for the three model scales and a wide edge element orientation For the 1:20 and 1:30 models (Figures 4 14 and 4 15), mean pressures at = 0 range from 1.1 to 1.5. The 1:50 model shows slightly lower values ranging from 1.0 to 1.2 near the leading edge of the roof. Representative p rofiles of peak pressures a long a line of roof taps parallel to the long building dimension for the three WERFL models are displayed in Figures 4 17 4 1 8 and 4 19 for 0 A large spread in peak pressures near the leading edge of the roof is evident on the three figures Howev er, the spread for the 1:50 WERFL model is small in relation to the other two models. Variation in peak pressures were between 3 and 9 for the 1:20 and 1:30 models, while peak pressures only ranged from 2.5 to 6 at = 0 for the 1:50 model.

PAGE 76

76 Charac terizing Upwind Terrain and Model Scale Comparison Fi gure 4 20 includes peak pressures on all 266 pressure taps f r om the 1:30 and 1:50 WERFL (UFL) models plotted against the 1:20 model The top subplot display s pr essures from selected upwind terrain configurations matching turbulence intensities (at eave height) representative of open exposure ( ~14%). The bottom subplot includes pressures corresponding to roughness array configurations matching equivalent full scale roughness lengths for open country exposure BLWT roughness lengths were geometrically scaled based on the model scale ASCE 7 10 (2010) provides a range of roughness length s for open terrain raging from = 0.01 0.15 m. Comparable (peak) pressures are observed for the two subplo ts in Figure 4 20. In contrast, Figure 4 21 show s significant differences in peak pres sures for suburban simulation. Larger peak values ( 6 to 8) are detected for = 0, when reconfiguring the Terraformer to attain the required turbulence levels at eav e height. Nevertheless, both suburban and open simulation, show consistent agreement between the thre e model scales, when the turbulence intensity was matched at the eave of the model. Figure 4 22 provides a comparison of two simulations of open terrain achieved using a narrow (top subplot) and wide (bottom subplot) edge windward element orientation. Roughness element heights of 30 mm and 60 mm were selected for the wide and narrow subplots, respectively. The resulting turbulence intensities (at eave heig ht) for the three model scales are included. The two upwind terrains appear to generate comparable (peak) pressures for the considered wind direction.

PAGE 77

77 Summary A series of experiments was conducted in a large BLWT to investigate the effects of envelop pressures on low rise buildings with increasing surface roughness M odels of the WERFL experimental building were immersed in 33 of turbulen t boundary layer flows via precise regulation of a computer control terrain generator. BLWT pressure tests were compared to aerodynamic experiments of a 1:100 model of the WERFL building conducted at UWO. In general, good agreement was found between UWO and UFL WERFL pressure measurements when matching the turbulence levels at eave height. Furthermore, the 1:50 model E xperiments revealed the dependency of extreme (minimum) pressure coefficients with increasing surface roughness Peak pressures near the leading edge of the roof appeared to vary linearl y with increasing turbulence levels at eave height. The trend was more pronounced in the 1:20 and 1:30 models. Only slight variations in the mean pressures were observed with increasing turbulence levels

PAGE 78

78 Table 4 1 WERFL testing parameters Model Scale 1:20 1:30 1:50 Sampling period, 300 s 180 s 120 s Sampling rate, 625 Hz Wind direction, 0, 45, and 90 Mean reference wind speed (pitot), 15.3 m/s Eave height, (mm) 198 132 80 Number of pressure taps 266 Roof slope 1:48

PAGE 79

79 Figure 4 1 Pressure tap layout for 1:20, 1:30, and 1:50 WERFL building models

PAGE 80

80 Figure 4 2. Measured and fitted response amplitudes up to a frequency of 600 Hz (Courtesy of Dr. David Roueche). Figure 4 3. Schematic diagram of BLWT physical arrangement near the test section for pressure tests.

PAGE 81

81 Figure 4 (Photos courtesy of author).

PAGE 82

82 A B C Figure 4 5 Comparison of mean pressure coefficients for the A ) 1:20 B) 1:30, and C) 1:50 WERFL model s (Narrow edge, = 15 0 mm, 90 ).

PAGE 83

83 A B C Figure 4 6 Comparison of peak pressure coefficients for the A) 1:20 B) 1:30, and C) 1:50 WERFL model s ( Wide edge, = 8 0 mm, 90 ).

PAGE 84

84 Figure 4 7 Pressure spectra comparison of Tap 216 (located near the roof corner closest to the approach flow) for the 1:20 WERFL model.

PAGE 85

85 Figure 4 8 Distribution of mean pressure coefficients for the 1:20 WERFL model (Wide edge, 0 ).

PAGE 86

86 Figure 4 9 Distribution of RMS pressure coefficients for the 1: 5 0 WERFL model (Wide edge, 9 0).

PAGE 87

87 Figure 4 10 Distribution of peak pressure coefficients for the 1: 3 0 WERFL model (Wide edge, 45 ).

PAGE 88

88 Figure 4 11 Distribution of mean pressures along a line of taps ( parallel to the long building dimension ) for the (a) 1:20 WERFL model ( Wide edge, 0 ). Figure 4 1 2 Distribution of mean pressures along a line of taps (parallel to the long building dimension) for the 1: 3 0 WERFL model ( Wide edge, 0 ).

PAGE 89

89 Figure 4 1 3 Distribution of mean pressures along a line of taps (parallel to the long building dimension) for the 1: 5 0 WERFL model ( Wide edge, 0 ). Figure 4 1 4 Distribution of mean pressures along a line of taps (perpendicular to the long building dimension) for the 1:20 WERFL model ( Wide edge, 90 ).

PAGE 90

90 Figure 4 1 5 Distribution of mean pressures along a line of taps (perpendicular to the long building dimensi on) for the 1: 3 0 WERFL model ( Wide edge, 90 ). Figure 4 1 6 Distribution of mean pressures along a line of taps (perpendicular to the long building dimension) for the 1: 5 0 WERFL model ( Wide edge, 90 ).

PAGE 91

91 Figure 4 1 7 Distribution of peak pressures along a line of taps (parallel to the long building dimension) for the 1:20 WERFL model ( Wide edge, 0 ). Figure 4 1 8 Distribution of peak pressures along a line of taps (parallel to the long building dimension) for the 1: 3 0 WERFL model ( Wide edge, 0 ).

PAGE 92

92 Figure 4 1 9 Distribution of peak pressures along a line of taps (parallel to the long building dimension) for the 1: 5 0 WERFL model ( Wide edge, 0 ).

PAGE 93

93 A B Figure 4 20. Peak pressures on all 266 taps for upwind terrains matching A) turbulence intensities ( at eave height ) and B ) roughness length s satisfying open country (0.01 0.15 m)

PAGE 94

94 A B Figure 4 21. Peak pressures on all 266 taps for upwind terrains matching A ) turbulence int ensiti es (at eave height) and B ) roughness lengths satisfying suburban exposure (0.01 0.15 m).

PAGE 95

95 A B Figure 4 22. Peak pressures on all 266 taps for open country ( ~14%) terrain simulation with A ) n arrow and B ) w ide edge windward element orientation.

PAGE 96

96 CHAPTER 5 WIND LOAD DESIGN OPTIMIZATION OF STEEL FRAMES This chapter presents an improved metaheuristic optimization method for large scale frame structures that minimizes weight while satisfying strength and serviceability requirements. The proposed alg orithm first explores then exploits (ETE) the search space, adjusting the influence of the local and global best designs on the selection of new candidate solutions. A discrete (stochastic) search scheme is activated in the late stages of the optimization procedure to exploit the (local) search space near the global optimum. The proposed method is successfully applied to two benchmark large steel frame structures: (1) a three bay 24 story moment resistant frame and (2) a seven bay 60 story structure. ETE pr oduced optimum weights for the 24 story frame that outperformed recently develop metaheuristic strategies. For the 60 story frame, optimum designs from 45 independent runs produced frame weights within 2% of results found through rigorously derived optimal ity criteria methods. The paper demonstrates how the proposed stochastic (local) search strategy performs minute alterations to the best design, while only permitting the creation of new designs capable of improving the (current) global best. ETE appears t o significantly enhance the exploitation capabilities of Big Bang Big Crunch method, specifically for discrete sizing optimization of large steel frames with vast design domains. Background Automation of the structural design process of civil structures via high level optimization strategies is expected to grow significantly in the upcoming decade, particularly for structural systems comprising hundreds to upwards of thousands of structural e lements. Efficiently sizing these members to satisfy

PAGE 97

97 code based strength and serviceability requirements is a computationally demanding task, hence the need for new research to efficiently evaluation optimal solutions. Deterministic optimization algorithms such as mathematical programming e.g., sequential quadratic programming and optimality criteria (OC) have been widely held methods for solving struct ural optimization problems (Haftka, 1985; Arora, 2004; Haftka, 2012; Fenton, 1974; Feury and Geradin, 1978 ) The OC algorithm is one of the most established methods for optimizing la rge scale civil structures (Saka, 1984; Tabak and Wright, 1981; Khan, 1984; Sadek, 1992) For example, Chan and Grierson (1993) applied an efficient pseudo discrete OC procedure fo r member sizing optimization of a 60 story planar steel framework subject to multiple deflection constraints. This resizing technique has been extensively applied to optimize the stiffness of tall buildings subject to mul tiple drift requirements (Chan et a l., 2010; Zou and Chan, 2005; Huang et al., 2012; Spence and Gioffre, 2012; Spence and Kareem, 2013) In practice, member sizing optimization of frames typically involves the selection of commercially available cross sectional shapes from a list of standa rd steel sections that satisfy code based serviceability and strength requirements while reducing the cost of material by minimizing the frame weight. Despite modifications to traditional deterministic optimization algorithms, such as OC, for discretizing s izing optimization problems (Arora, 2000) their practical application has been limited due to their complexity and inefficiency when applie d to large scale structures (Saka and Geem, 2013) Furthermore, large scale problems are

PAGE 98

98 highly susceptible to be t rapped in subdomains (i.e., local optima) given the relatively vast search space. The present study addresses these issues through a hybridized metaheuristic optimization algorithm designed to solve large scale structural frame problems. Inspired from bio logical and physical processes found in nature, metaheuristic algorithms use high level population based strategies to guide stochastic search through the design domain of candidate solutions. Numerous studies (Camp et al., 2005; Li et al., 2009; Kaveh and Talatahari, 2009) have demonstrated their effectiveness in civil structure applications. For example, Li et al. (2009) created a heuristic particle swarm optimizer (HPSO) for pin connected structures by hybridizing PSO and harmon y search (HS) schemes. Cam p (2007) implemented a big bang big crunch (BB BC) algorithm for continuous and discrete optimization of space trus ses. Kaveh and Abbasgholiha (2011) applied BB BC for weight minimization of steel sway frames subject to AISC LRFD (2010) strength and servic eability requirements. However, only a few studies (e.g., Kicinger et al., 2005; Hasancebi et al., 2011; Azad and Hasancebi, 2015; Kaveh and Bolandgerami, 2017) have applied metaheuristic algorithms on large structural systems. Optimization of these system s comprising thousands of design variables require intelligent strategies capable of performing a global examination of the search space (i.e., exploration) and the restricted investigation around promising regions (i.e., exploitation) (e.g., Cuevas et al. 2014; Mitalili et al., 2014) The proposed methods is based on PSO and BB BC for discrete sizing optimization of large scale steel frames. A modification is introduced to allow a

PAGE 99

99 steady transition from exploration during early stages of the optimization process to exploitation of the feasible domain at late stages, and a discrete (stochastic) search scheme is initiated in the late stages of the optimization procedure to exploit the designs near the global optimum. The proposed algorithm is su ccessfully applied to two benchmark problems: (1) a 24 story three bay planar frame (2) and a 60 story seven bay planar frame. Results obtained by the proposed algorithm for the 24 story frame compare to recently develop metaheuristic strategies. For the 6 0 story frame, optimum designs from 45 independent runs produced frame weights within 2% of results found through rigorously derived optimality criteria methods. The paper illustrates how the proposed stochastic search strategy performs small adjustments to the global best during late stages, while only permitting the generation of new candidate designs with the potential to overtake the (current) best design. This discrete stochastic scheme appears to enhance the exploitation capabilities of BB BC (mostly influenced by a random distribution generator) for large scale discrete sizing optimization problems. Discrete Optimization of Multi Story Frame Structures Structural frame optimization selects the lightest combination of standard structural sections, res ulting in minimum material cost while satisfying code based serviceability and strength requirements. Here we analyze planar steel frames based on AISC LRFD specifications (2010). Mathematically, the optimum design of multi story steel frames can be formul ated as follows: find ( 5 1 )

PAGE 100

100 to minimize ( 5 2 ) subject to ( 5 3 ) ( 5 4 ) ( 5 1 ) where each member of set is associated with a standard steel section for member group ; is the total number of design variables (i.e., member groups). In Equation 5 2, is the weight of the frame; and are the density and cross sectional area of members in group respectively; is the length of member and is the total number of members; Equation 5 5 constrains the bounds of each design variable entry ; is the total number of standard steel sections in member group Equations 5 3 and 5 4 represent inequality constraints for strength (Equation 5 3) and serviceability (Equation 5 4) requirements based on AISC LRFD specifications (2010). In Equation 5 3, is the load capacity ratio defined as: ( 5 6 )

PAGE 101

101 where is the required axial strength (i.e., tension or compressio n), and are the required flexural strength for strong and weak axis bending, respectively ( for planar frames ; is the available axial strength; and are the available flexural strength for strong and weak axis bending, respectively; is the resistance factor for bending equal to 0.90; is the axial resista nce factor, which equals 0.90 and 0.90 for tension and compression members, respectively. S trength constraints presented herein for steel frames using AISC LRFD strength requirements can readily be reformulated for optimum design of concrete (ACI 318, 2008) an d wood framed structures (Wheat and Cramer, 2006) Equation 5 4 shows the inequality co nstraint for inter story drift requirement of multi story buildings, where can be expressed as follows: ( 5 7 ) where is the relative lateral displacement of adjacent stories; is the allowable interstory drift limit; is the total number of stories. The following penalty function approach is applied to transform the constrained optimization problem presented in Equations 5 1 through 5 5 into an unconstrained one: ( 5 8 ) where is a penalty coefficient ( typically 0.5 1.5 ), and is the penalty function defined as

PAGE 102

102 Proposed ETE Algorithm Particle Swarm Optimization (PSO) PSO is a stochastic optimization technique (Eberhart and Kennedy, 1995) inspired from the social behavior of bird flocking and fish schooling. In PSO, a population (swarm) of individuals (particles) move through the search space of candidate solutions, periodically updating t heir positions and velocities based on both their own experience (particle best) and the experience of the swarm (global best). The updating rule in the original PSO is given by: where and is the position (Equation 5 1) and velocity vectors of particle at iteration respectively; is the position of the best solution found among all candidates up to iteration (global best); is the best position found by particle up to iteration (particle best); is the inertial weight parameter; is the population size; and are called acceleration coefficients; and are uniformly distributed random numbers in the range of [0,1]. PSO has proven effective in the global investigation (i.e., exploration) of the design domain. However, the algorithm has a tendency to easily be trapped in local optima (i.e., subdomains) during late stages of the optimization process ( 5 9 ) ( 5 10 ) ( 5 11 )

PAGE 103

103 (Yang et al., 2007). Design problems with large search spaces e.g., hundreds of design varia bles are more susceptible to this phenomenon. Consequently, modifications to the original PSO scheme are abundant in literature (e.g., Li and Huang, 2007 ; Perez and Behdinana, 2007; Yang et al., 2007) This study attempts to address the limitation s of PSO by precise control of the influence of on new particles while preserving the exploratory capabilities of the original algorithm. ETE Algorithm The proposed ETE algorithm regulates the influence of at different stages of the optimization process via a single control parameter. In addition, a normal distribution operator adapted from the BB BC algorithm ( Erol and Eksin, 2006 ) is introduced to exploit promising regions within the search space at late iteration stages. The updating technique is form ulated as follows: where is a control parameter that linearly increases over a user specified number of generations to control the relative influence of and ; is a normal distribution operator from the BB BC algorithm ( Erol and Eksin, 2006 ). In this study, is defined as: in which is a random generated from a standard normal distribution; is a parameter for controlling the size of the search space; is an exponential parameter ( = 3 for this study); and are the position vectors of the ( 5 12 ) ( 5 13 )

PAGE 104

104 upper and lower bounds of each design variable, respectively. A comparable updatin g scheme to the one shown in Equation 5 12 can be found in Kaveh and Abbasgholiha (2010). Figure 5 1 presents a flowchart of the proposed ETE algorithm. Similar to most metaheuristic methods, an initial population i.e., swarm is created via a uniform random number generator bounded by and The penalized objective function is then evaluated for each candidate solution i.e., par ticle. After assessing the fitness of each particle, and are determined. New candidate solutions emerge from Equation 5 12. In subsequent iterations, is recalculated using Equation 5 14. This process is repeated until the maximum number of iterations is reached. Linear ly Varying Control Parameter During each iteration, is adjusted to progressively intensify the influence of the global best solution ( ) on the swarm, thus effecting a gradual transition from exploration t o exploitation of the search space. In Equation 5 12, is increased linearly after each iteration following where is the maximum number of iterations; is a parameter which defines the iteratio n when will transition from a linear variation to a final constant value; and are the initial and final values ( and ). The dynamic behavior of is represented graphically in Figure 5 2. At early iterations, it is desirable for new candidate solutions to be influenced by both the best solution found by the particle ( ) and swarm ( ). For example, if 0.5, ( 5 14 )

PAGE 105

105 and will equally impact the new position As the optimizati on carries on, new positions gradually become more attracted to the global best solution, searching more in the vicinity of When iteration is reached, will be mostly or completely, given 1.0 influenced by as portrayed by the exploitation phase. In this final stage, the algorithm performs a local search around i.e., Discrete Stochastic (Exploitation) Scheme In metaheuristics, design optimization of large steel frames typically requires a high num ber of iterations ( Hasanebi et al., 2011 ). From Equation 5 13, it is evident that approaches 0 as becomes large. If 0 and 1.0, the new particle will remain in the same position as the global best according to Equation 5 12 Additionally, if every entry in is equal to or greater than 0, is certain to produce a worst solution than To address this, a conditional routine (Figure 5 3) is added to the proposed algorithm if either condition is satisfied. Fir st, a new is pre allocated with a vector of zeros. Second, an entry in vector is selected randomly and assigned a negative integer between and 1. Finally, an integer between 0 and is allocated to either entry or of The r ationale behind the routine is that nearby entries in typically implies neighboring members in the physical structure. A small increase in the cross sectional area (i.e., size up) of a group of members may allow a considerable size reduction (i.e., si ze down) of a nearby member group.

PAGE 106

106 Results This section examines the efficiency and robustness of ETE for optimizing large planar steel frames. Two benchmark frame problems are considered: (1) a three bay 24 story steel frame and (2) a seven bay 60 story steel structure, both subject to strength and inter story drift requirements from AISC (2010). The ETE was applied to each benchmark problem for three cases, namely 0.25, 0.50, and 0.75. In both problems, = 80, = 0.5, 0.5, 1.0, and 0.6. For the 24 story frame, 100 independent runs were performed for each case 300 total runs. The algorithm was executed 15 times on the 60 story frame for each case, for a total of 45 runs. The maximum number iterations was set to 200 and 1000 for the 24 and 60 story frames, respectively. Structural analysis (i.e., 1 st order, elastic direct stiffness technique) and the ETE algorithm were coded in MATLAB. Runs were executed from a personal computer 8 Core Intel Xeon CPU E 5440 @ 2.83 GHz with 24 GB RAM using independent parallel processing. 24 Story Three Bay frame Figure 5 4 shows the topology, load condition, and member grouping for the subject three bay twenty four story planar frame. The frame has been studied by Kaveh and Talatahari (2010), applying an improved ant colony optimization (IACO), Togan (2012) via teaching learning based optimization (TLBO) algorithm, Safari et al. (2013) utilizing a modified multiple deme genetic algorithms (MMDGA), among others. The struct ure comprises 168 members arranged into 20 groups. Groups 1 4 and 5 20 correspond to beam and column

PAGE 107

107 members, respectively. Beam sections are selected from all 267 W Shapes and columns are chosen from a set of 37 W14 shapes. The magnitude of the loads disp layed in Figure 5 4 are = 25.628 kN, = 4.378 kN/m, = 6.362 kN/m, = 6.917 kN/m, = 5.954 kN/m. The modulus of elasticity and yield stress are = 205 GPa and Fy = 230.3 MPa, respectively. Design for strength follows LRFD AISC specificati ons (2010). Beams and columns are assumed unbraced along their length. The effective length factors of the members are computed as 0 for a sway permitted frame and the out of plane effective length factor is = 1.0. Inter story drift limit is set to /300 for all stories. Table 5 1 summarizes the statistics of final weights from 300 independent runs for the 24 story frame. The lightest design resulted in a final weight of 894.3 kN For this run, 0.25 (Case 1). The worst (heaviest) desig n produced a final weight of 986.3 kN (Case 2). A slight reduction in the variability (standard deviation) of the final weight is observed as increases. Additionally a greater number of optimum designs (runs) are closer to the bes t design for each cas e as shown in Table 5 2 This might be attributed to a more exhaustive (i.e., exploration) search with increasing the influence of is present for a longer period. For instance, runs with 0.75, 52 of 100 runs were within +1.0% of the best (case) solution ( Table 5 2), while 91 of 100 were within +5 .0% Figure 5 5 includes iteration histories of the mean weight from 100 runs for the 24 story frame. During iterations ( = 20 100), case 1 shows a faster (mean) weight improvement than Cases 2 and 3, suggesting a stronger influence of the global best ( ) for case 1 during early stages For example, at = 50, 1.0, 0.75, and 0.66 for cases 1, 2, and 3, respectively. However, at iterati ons

PAGE 108

108 150 200, mean weights of case 3 surpasses cases 1 (at ~162) and 2 (at ~165). Optimum weights and W shape sections obtained by ETE are shown in Table 5 3, and compared to other recent metaheuristic techniques. The best design attained by ETE (89 4.3 kN) is 8.4%, 1.1%, and 0.43% lighter than IACO, TLBO, and MMDGA, respectively. Figures 5 6 and 5 7 show the inter story drift, load capacity ratio, and member size distribution of the best run from Cases 1 and 3 for the 24 story frame, respectively. Member line thicknesses are proportional to the square root of the area (i.e., ). The best design o btained by ETE satisfies strength and serviceability requirements for all three cases of the 24 story frame. 60 Story Seven Bay Frame The sixty story seven bay trussed frame shown in Figure 5 8 was first optimized in Chan (1992) using the OC method. All se ven bays span 6.096 m and all stories are 3.658 m in height. Beam to column joints are rigidly connected, while K bracing members are assumed to be pinned at their ends. The eight joints at the base of the frame are modeled as fixed supports. Only wind loa ding is considered. The fram e consists of 1080 members ( Table 5 4), which are convened into 240 groups. Beams located on the same story are grouped together (i.e., 60 beam groups). The same applies for diagonal bracings. Columns are grouped over two adjac ent stories and symmetrical columns correspond to the same group. Steel sections for beam groups selected from a discrete set of 21 W24

PAGE 109

109 sections, while column and bracing groups can chose from 36 W14 sections. The total number of combinations is approximat ely 10 360 In this study, the frame is design for inter story drift and strength requirements based on LRFD AISC specifications (2010). Beams and columns are assumed to be unbraced along their length. The modulus of elasticity is = 200 GPa. A yield stre ss of Fy = 300 MPa was selected to match the material properties used in Walls and Elvin (2010). Inter story drift limit is h/400 for each of the 60 stories. Statistics of final weights from 45 independent runs 15 runs per case for the 60 story frame are s hown in Table 5 5. As is the case for the 24 story frame, the worst run correspond to 0.25 (22.835 MN). The best run resulted in a final weight of 22.448 MN. The parameter was set to 0.75 in this run. Mean (final) weights seem to decrease as incr eases. Case 1 produced the highest variability in the final weight, with a standard deviation of 0.122 MN. Case 2 ( 0.25) produced less spread in the final weight. Table 5 6 includes a comparison of ETE with other optimization techniques for the 60 stor y frame. The best run of ETE is 1.2% lighter than the OC simple round up solution, and 1.58% heavier than the pseudo round up technique (Chan, 1992). However, neither of the two OC methods presented in Chan (1992) considered strength requirements. Wall and Elvin (2010) introduced strength requirements to the 60 story frame based on SANS (2005), and achieved an optimum weight of 22.436 MN using VWO. This design is 0.05% lighter than the best run of ETE (22.445 MN), which incorporated strength limit states fr om AISC LRFD (2010) Distinctions in code specification hinders direct

PAGE 110

110 comparison of the results of VWO and ETE. Nonetheless, it is evident that ETE can achieve comparable results to numerical methods (i.e., OC and VWO) for large scale frames. Mean iterati on history curves for the 60 story frame are shown in Figure 5 9. Similar to the 24 story frame, case 1 shows a rapid reduction of the mean weight during early iterations ( = 250 600) compared to cases 2 and 3. However, the iteration curve for case 3 fal ls just below Case 2 from = 200 525, contrary to what should be expected. This could be due to the limited number of runs for the 60 story problem 15 runs Regardless, mean weights of case 3 better cases 1 and 2 at late stages of the optimization ( = 750 1000). While absent in Figure 5 5, Figure 5 9 shows abrupt slope changes (e.g., at = 500 for case 2). mark points where the algorithm first reaches 1.0. A more rapid improvement stepper slope in the mean weight is observed past these points. At this stage of the optimization process, some particles are satisfying the conditional statement of Figure 5 3, allowing recalculation of As a consequence, more particles have a chance to ove rtake Inter story drift and member size distribution, and load capacity ratios of the best run from Cases 2 and 3 for the 60 story frame are shown in Figures 5 10 and 211, respectively. As previously mentioned, member line thicknesses are scaled pro portional to It is apparent that inter story drift constraints controls the optimum design.

PAGE 111

111 Summary This chapter presents an exploration to exploitation (ETE) algorithm for discrete optimization of large frame structures. The algorithm was tested on two benchmark steel frame problems. First, ETE was applied to a 24 story planar steel frame subject to AISC LRFD strength and inter story drift requirements. The best designs from ETE outperformed recently developed meta heuristic algorithms (IACO, TLBO, a nd MMDGA). The second benchmark problem was a 60 story seven bay frame. Optimum designs obtained using ETE compared to results from CH92 and WE10. Differences in optimum weights can be ascribed to (1) the absence of strength constraints in the OC algorithm and (2) distinctions in code specifications used to satisfy strength requirements AISC LRFD (2010) and SANS (2005) This study demonstrated the capabilities of ETE to execute a comprehensive search of the vast design domain associated with large steel st ructures. Despite the random nature of the algorithm, optimum designs from ETE compared to rigorously derived deterministic methods, without the use of domain knowledge OC and VWO employ the principle of virtual work for identifying the most effective stru ctural members. Subsequent work will concentrate on significantly reducing the computational time number of structural analysis of ETE, which remains one of the main drawbacks of most metaheuristic algorithms for structural optimization.

PAGE 112

112 Table 5 1. Summary of statistics for optimum design of 24 story three bay frame (100 Runs). Case Case 1 = 0.25 Case 2 = 0. 5 Case 3 = 0. 7 5 Final Weight (k N ) Maximum ( w orst r un) 966.9 986.3 967.7 Mean 913.4 912.6 910.7 Minimum ( b est r un) 894.3 895.8 896.5 Standard Deviation (k N ) 18.0 17 .0 15 2 Coefficient of variation (%) 1.97 1.87 1.67 Table 5 2. Percent difference relative to best run for 24 story three bay frame (100 Runs). % Greater than Best (Case) Run No. of Independent R uns Case 1 = 0.25 Case 2 = 0. 5 Case 3 = 0. 7 5 + 0.1 0 1 3 + 0.5 14 22 28 + 1.0 31 42 52 + 5.0 84 89 91

PAGE 113

113 Table 5 3. Comparison of best designs for 24 story three bay frame. Group No. AISC W Shapes IACO (Kaveh et al., 2010) TLBO (Togan, 2012) MMDGA (Safari et al., 2013) Present Study Case 1 = 0.25 Case 2 = 0. 5 Case 3 = 0. 7 5 1 W30X99 W30X90 W30X90 W30X90 W30X90 W30X90 2 W16X26 W8X18 W8X15 W6X15 W6X15 W6X15 3 W18X35 W24X62 W24X55 W24X55 W24X55 W24X55 4 W14X22 W6X9 W10X15 W6X8.5 W6X8.5 W6X8.5 5 W14X145 W14X132 W14X159 W14X159 W14X15 9 W14X14 5 6 W14X132 W14X120 W14X132 W14X132 W14X12 0 W14X13 2 7 W14X120 W14X99 W14X90 W14X109 W14X10 9 W14X12 0 8 W14X109 W14X82 W14X90 W14X74 W14X74 W14X74 9 W14X48 W14X74 W14X65 W14X61 W14X61 W14X68 10 W14X48 W14X53 W14X48 W14X38 W14X43 W14X38 11 W14X34 W14X34 W14X48 W14X38 W14X34 W14X30 12 W14X30 W14X22 W14X22 W14X22 W14X22 W14X22 13 W14X159 W14X109 W14X109 W14X90 W14X90 W14X99 14 W14X120 W14X99 W14X99 W14X99 W14X10 9 W14X99 15 W14X109 W14X99 W14X99 W14X90 W14X90 W14X90 16 W14X99 W14X90 W14X74 W14X90 W14X90 W14X90 17 W14X82 W14X68 W14X68 W14X74 W14X74 W14X68 18 W14X53 W14X53 W14X53 W14X61 W14X61 W14X61 19 W14X38 W14X34 W14X26 W14X30 W14X34 W14X38 20 W14X26 W14X22 W14X22 W14X22 W14X22 W14X22 Weight (k N ) 969.25 903.88 898.19 894.32 895.83 896.54

PAGE 114

114 Table 5 4. Location of structural members in the 60 story frame. Members Location in the structure 1 60 Exterior columns (C1) 61 120 1 st interior columns (C2) 121 180 2 nd interior columns (C3) 181 240 3 rd interior columns (C4) 241 300 3 rd Interior columns (C5) 301 360 2 nd interior columns (C6) 361 420 1 st interior columns (C7) 421 480 Exterior columns (C8) 481 540 Beams bay 1 541 600 Beams bay 2 601 660 Beams bay 3 841 900 Beams bay 4 901 960 Beams bay 5 661 720 Beams bay 6 721 780 Beams bay 7 781 840 Beams bay 8 961 1020 Bracings bay 4 1021 1080 Bracings bay 5 Table 5 5. Summary of statistics for optimum design of 60 story seven bay frame (Independent Runs = 15). Case Case 1 = 0.25 Case 2 = 0.5 Case 3 = 0.75 Final Weight (k N ) Maximum (worst run) 22.835 22.699 22.713 Mean 22.625 22.581 22.571 Minimum (best run) 22.454 22.500 22.448 Standard Deviation (k N ) 0.122 0.053 0.071 Coefficient of variation (%) 0.54 0.23 0.31 Table 5 6 Percent difference relative to best run for 60 story seven bay frame (Independent Runs = 15). % Greater Than Best (Case) Run No. of Independent R uns Case 1 = 0.25 Case 2 = 0.5 Case 3 = 0.75 + 0.1 0 1 0 + 0.5 5 11 7 + 1.0 10 14 12 + 5.0 14 14 14

PAGE 115

115 Table 5 7. Optimum design comparison for the 60 story planar frame. Method Weight (kN ) % Greater (Heavier) Than Best R un Constraints OC simple round up (Chan, 1992) 22.716 +1.20 Inter story drift OC pseudo round up (Chan, 1992) 22.093 1.58 Inter story drift VWO (Wall and Elvin, 2010) 22.436 0.05 Inter story drift and strength (SANS, 2005) Present Work Worst Run 22.827 +1.70 Inter story drift and strength (AISC LRFD, 2010) Best Run 22.445 Inter story drift and strength (AISC LRFD, 2010)

PAGE 116

116 Figure 5 1. Linear variation of during the optimization process. Figure 5 2. Conditional subroutine for determining

PAGE 117

117 Figure 5 3. Flowchart of ETE for discrete sizing design optimization of framed structures.

PAGE 118

1 18 Figure 5 4. Topology and loading condition of 24 story three bay planar frame.

PAGE 119

119 Figure 5 5. ETE mean optimization history (24 story three bay frame).

PAGE 120

120 Figure 5 6. Member sizing, inter story drift, and load capacity ratios for the 24 story three bay frame obtained using ETE (Case 1, = 0.25).

PAGE 121

121 Figure 5 7. Member sizing, inter story drift, and load capacity ratios for the 24 story three bay frame obtained using ETE (Case 3, = 0.75).

PAGE 122

122 Figure 5 8. Topology and loading condition of 60 story seven bay planar frame.

PAGE 123

123 Figure 5 9. ETE mean optimization history (60 story seven bay frame).

PAGE 124

124 Figure 5 10. Member sizing, inter story drift, and load capacity ratios for the 60 story seven bay frame obtained using ETE (Case 2, = 0.5).

PAGE 125

125 Figure 5 11. Member sizing, inter story drift, and load capacity ratios for the 60 story seven bay frame obtained using ETE (Case 3, = 0.75).

PAGE 126

126 CHAPTER 6 CONCLUSIONS AND FUTURE WORK Contributions, final remarks, and recommendations for future research resulting from this dissertation a re outlined herein. The sections follow the order in which the topics were presented in the document. Field Measurements of Hurricane W inds As part of this dissertation, nine new f ield experiments (from Hurricanes Hermine and Matthew) collected during the 2016 Atlantic hurricane season were added to the FCMP database, which includes near surface wind observations from 25 named storms since 1998. Recent field deployments have targete d suburban terrain exposures to investigate the turbulent structures in the roughness sublayer (RS) during strong win d events Skewed distributions of the longitudinal velocity component were o bserved within the RS. Quadrant analysis of the Reynolds stress demonstrated the link of skewed distributions swee p dominance downward transfers of momentum. Future field experiments should continue to target suburban terrains, since most of the FCMP database consist of wind observations measured in open and marine e xposures. Furthermore, the use of 15 m weather stations, equipped with ultrasonic anemometry along the height of the tower provides valuable insight of how the statistical moments of the three velocity components (e.g., mean, variance, skewness, and kurto sis) vary with height in the RS. Surface Pressures on Low Rise B uildings Most wind tunnel studies assessing extreme surface pressures on low buildings have been limited to two terrain conditions, namely open and suburban. Simulation of these two terrains i s traditionally achieved by manual modification of the upwind terrain

PAGE 127

127 through implementation of mixing devices e.g., vortex generators, roughness grid, barriers until aerodynamic parameters, such as roughness length, closely match predefined values from wi nd load provisions. Preliminary results from a comprehensive series of experiments on low rise buildings revealed that the peak (minimum) pressure coefficients are terrain dependent Modern wind load provisioning such as ASCE 7 uses external pres sure coefficients re normalized to open exposure terrain, and do not, for example, prescribe pressure distributions on the building for terrain type. Further more t he increase was found to be more pronounced in regions of the large st negative pressure (e.g ., roof corners). R oot mean square values of pressure coefficients appear to follow a nearly linearly proportional relation with increasing surface roughness Wind tunnel experiments presented in this dissertation could serve as a testbed for exploring mor e complex upwind terrains, such as heterogeneous roughness element fields, to simulate more realistic topographic conditions and the adverse wind induced effects on low rise buildings. Further, all data will be available in the Design Safe Cyberinfrastruct ure (CI) node at the University of Texas at Austin Design Safe is a web based research platform of the Natural Hazards Engineering Research Infrastructure (NHERI) Network. The website provides computational tools needed for managing, analyzing, and unders tanding critical data for natural hazards research. Design Optimization of Large Civil Structures Subject to Wind Loads Metaheuristic search techniques have the potential to become a common tool for design practitioners to enhance the performance of civil structures subject to wind. The simple yet versatile nature of these algorithms when compared to gradient based methods (e.g., optimality criteria) makes them a viable choice for practical design

PAGE 128

128 problems, and could be incorporated to existing performance based design (PBD) frameworks for optimal design of wind exited tall buildings. Future research should focus on significantly reducing the computational time number of structural analysis of the algorithms for large problems. However, the ever growing abun dance computational resources could naturally resolve this limitation.

PAGE 129

129 APPENDIX A UWO PRESSURE TAP LAYOUT AND TESTING PARAMETERS This appendix contains pertinent information related to pressure tests conducted at the University of Western (UWO) on a 1:100 model of the WERFL experimental building. This study was part of a cooperative agreement between the National Institute of Standa rds and Technology (NIST) and Texas Tech University.

PAGE 130

130 Figure A 1. UWO pressure ta p layout for 1:100 WERFL model (Source: Ho et al., 2003 ).

PAGE 131

131 Figure A 2. UWO test parameters for 1:100 WERFL model (Source: Ho et al., 2003 )

PAGE 132

132 APPENDIX B UFL WERFL MODELS This appendix includes illustrative images of experimental configurations for pressure tests of a 1:20 model of the WERFL experimental building. The experiments

PAGE 133

133 Figure B 1. 1:20 UFL WERFL model (Narrow edge, = 2 0 mm, 0 ) (Photo courtesy of author) Figure B 2. 1:20 UFL WERFL model (Narrow edge, = 8 0 mm, 45 ) (Photo courtesy of author)

PAGE 134

134 Figure B 3. 1:20 UFL WERFL model (Narrow edge, = 16 0 mm, 90 ) (Photo courtesy of author) Figure B 4. 1:20 UFL WERFL model (Wide edge, = 8 0 mm, 0 ) (Photo courtesy of author)

PAGE 135

135 Figure B 5. 1:20 UFL WERFL model (Wide edge, = 2 0 mm, 45 ) (Photo courtesy of author) Figure B 6. 1:20 UFL WERFL model (Wide edge, = 16 0 mm, 90 ) (Photo courtesy of author)

PAGE 136

136 APPENDIX C GEOMETRY OF WERFL MODELS (UFL) Vertices and pressure tap coordinates for the 1:20 WERFL model are tabulated in t his appendix Physical coordinates for the 1:30 and 1:50 models can be obtained by applying scaling factors to tabulated values. The scaling factors for the 1:30 and 1:50 are 0.67 and 0.4, respectively.

PAGE 137

137 Table C 1. WERFL model tap c oordinates (1:20) Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 1 101 3.8 0.0 52.9 1 2 102 60.0 0.0 59.1 1 3 103 148.0 0.0 59.1 1 4 104 172.4 0.0 59.1 1 5 105 235.0 0.0 59.1 1 6 106 284.8 0.0 59.1 1 7 107 300.4 0.0 59.1 1 8 108 397.2 0.0 59.1 1 9 109 453.4 0.0 52.9 1 10 110 3.8 0.0 120.7 1 11 111 60.0 0.0 126.9 1 12 112 148.0 0.0 126.9 1 13 113 172.4 0.0 126.9 1 14 114 235.0 0.0 126.9 1 15 115 284.8 0.0 126.9 1 16 116 300.4 0.0 126.9 1 17 201 397.2 0.0 126.9 1 18 202 453.4 0.0 120.7 1 19 203 3.8 0.0 191.7 1 20 204 60.0 0.0 198.0 1 21 205 148.0 0.0 198.0 1 22 206 172.4 0.0 198.0 1 23 207 235.0 0.0 198.0 1 24 208 284.8 0.0 198.0 1 25 209 300.4 0.0 198.0 1 26 210 397.2 0.0 198.0 1 27 211 453.4 0.0 191.7 1 28 212 0.0 4.7 59.1 2 29 213 0.0 4.7 128.0 2 30 214 0.0 4.7 192.6 2 31 215 4.4 8.2 198.2 6 32 216 33.7 8.2 198.8 6 33 301 71.8 8.2 199.6 6 34 302 148.0 8.2 201.2 6 35 303 224.2 8.2 202.8 6 36 304 300.4 8.2 201.4 5 37 305 376.6 8.2 199.8 5 38 306 414.7 8.2 199.0 5 39 307 452.8 8.2 198.2 5 40 308 457.2 5.0 192.6 4

PAGE 138

138 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 41 309 457.2 3.2 128.0 4 42 310 457.2 3.2 59.1 4 43 311 0.0 80.0 59.1 2 44 312 0.0 80.0 128.0 2 45 313 0.0 88.2 192.6 2 46 314 4.4 83.5 198.2 6 47 315 33.7 83.5 198.8 6 48 316 71.8 83.5 199.6 6 49 401 148.0 83.5 201.2 6 50 402 224.2 83.5 202.8 6 51 403 300.4 83.5 201.4 5 52 404 376.6 83.5 199.8 5 53 405 414.7 83.5 199.0 5 54 406 452.8 83.5 198.2 5 55 407 457.2 91.4 192.6 4 56 408 457.2 79.1 128.0 4 57 409 457.2 79.1 59.1 4 58 410 0.0 155.3 59.1 2 59 411 0.0 155.3 128.0 2 60 412 0.0 162.5 192.6 2 61 413 4.4 158.8 198.2 6 62 414 33.7 158.8 198.8 6 63 415 71.8 158.8 199.6 6 64 416 148.0 158.8 201.2 6 65 501 224.2 158.8 202.8 6 66 502 300.4 158.8 201.4 5 67 503 376.6 158.8 199.8 5 68 504 414.7 158.8 199.0 5 69 505 452.8 158.8 198.2 5 70 506 457.2 165.8 192.6 4 71 507 457.2 154.2 128.0 4 72 508 457.2 154.2 59.1 4 73 509 0.0 234.1 59.1 2 74 510 0.0 234.1 128.0 2 75 511 0.0 234.1 192.6 2 76 512 4.4 234.1 198.2 6 77 513 33.7 234.1 198.8 6 78 514 71.8 234.1 199.6 6 79 515 148.0 234.1 201.2 6 80 516 224.2 234.1 202.8 6

PAGE 139

139 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 81 601 300.4 234.1 201.4 5 82 602 376.6 234.1 199.8 5 83 603 414.7 234.1 199.0 5 84 604 452.8 234.1 198.2 5 85 605 457.2 234.1 192.6 4 86 606 457.2 234.1 128.0 4 87 607 457.2 234.1 59.1 4 88 608 0.0 305.7 59.1 2 89 609 0.0 305.7 128.0 2 90 610 0.0 312.9 192.6 2 91 611 4.4 309.4 198.2 6 92 612 33.7 309.4 198.8 6 93 613 71.8 309.4 199.6 6 94 614 148.0 309.4 201.2 6 95 615 224.2 309.4 202.8 6 96 616 300.4 309.4 201.4 5 97 701 376.6 309.4 199.8 5 98 702 414.7 309.4 199.0 5 99 703 452.8 309.4 198.2 5 100 704 457.2 317.1 192.6 4 101 705 457.2 305.7 128.0 4 102 706 457.2 305.7 59.1 4 103 707 0.0 380.1 59.1 2 104 708 0.0 380.1 128.0 2 105 709 0.0 368.7 192.6 2 106 710 4.4 384.7 198.2 6 107 711 33.7 384.7 198.8 6 108 712 71.8 384.7 199.6 6 109 713 148.0 384.7 201.2 6 110 714 224.2 384.7 202.8 6 111 715 300.4 384.7 201.4 5 112 716 376.6 384.7 199.8 5 113 801 414.7 384.7 199.0 5 114 802 452.8 384.7 198.2 5 115 803 457.2 372.9 192.6 4 116 804 457.2 380.1 128.0 4 117 805 457.2 380.1 59.1 4 118 806 0.0 459.9 59.1 2 119 807 0.0 459.9 128.0 2 120 808 0.0 459.9 192.6 2

PAGE 140

140 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 121 809 4.4 459.9 198.2 6 122 810 33.7 459.9 198.8 6 123 811 71.8 459.9 199.6 6 124 812 148.0 459.9 201.2 6 125 813 224.2 459.9 202.8 6 126 814 300.4 459.9 201.4 5 127 815 376.6 459.9 199.8 5 128 816 414.7 459.9 199.0 5 129 901 452.8 459.9 198.2 5 130 902 457.2 459.9 192.6 4 131 903 457.2 459.9 128.0 4 132 904 457.2 459.9 59.1 4 133 905 0.0 531.6 59.1 2 134 906 0.0 531.6 128.0 2 135 907 0.0 520.0 192.6 2 136 908 4.4 535.2 198.2 6 137 909 33.7 535.2 198.8 6 138 910 71.8 535.2 199.6 6 139 911 148.0 535.2 201.2 6 140 912 224.2 535.2 202.8 6 141 913 300.4 535.2 201.4 5 142 914 376.6 535.2 199.8 5 143 915 414.7 535.2 199.0 5 144 916 452.8 535.2 198.2 5 145 1001 457.2 523.3 192.6 4 146 1002 457.2 530.5 128.0 4 147 1003 457.2 530.5 59.1 4 148 1004 0.0 606.7 59.1 2 149 1005 0.0 606.7 128.0 2 150 1006 0.0 594.4 192.6 2 151 1007 4.4 610.5 198.2 6 152 1008 33.7 610.5 198.8 6 153 1009 71.8 610.5 199.6 6 154 1010 148.0 610.5 201.2 6 155 1011 224.2 610.5 202.8 6 156 1012 300.4 610.5 201.4 5 157 1013 376.6 610.5 199.8 5 158 1014 388.6 614.0 199.5 5 159 1015 414.7 610.5 199.0 5 160 1016 452.8 610.5 198.2 5

PAGE 141

141 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 161 1101 457.2 597.6 192.6 4 162 1102 457.2 605.8 128.0 4 163 1103 457.2 605.8 59.1 4 164 1104 0.0 682.6 59.1 2 165 1105 0.0 682.6 128.0 2 166 1106 0.0 680.8 192.6 2 167 1107 4.4 677.6 198.2 6 168 1108 33.7 677.6 198.8 6 169 1109 71.8 677.6 199.6 6 170 1110 148.0 677.6 201.2 6 171 1111 224.2 677.6 202.8 6 172 1112 300.4 677.6 201.4 5 173 1113 376.6 681.1 199.8 5 174 1114 386.0 671.5 199.6 5 175 1115 414.7 677.6 199.0 5 176 1116 452.8 677.6 198.2 5 177 1201 457.2 681.1 192.6 4 178 1202 457.2 681.1 128.0 4 179 1203 457.2 681.1 59.1 4 180 1204 3.8 685.8 52.9 3 181 1205 60.0 685.8 52.9 3 182 1206 148.0 685.8 52.9 3 183 1207 172.4 685.8 52.9 3 184 1208 235.0 685.8 52.9 3 185 1209 284.8 685.8 52.9 3 186 1210 300.4 685.8 52.9 3 187 1211 397.2 685.8 52.9 3 188 1212 453.4 685.8 52.9 3 189 1213 3.8 685.8 120.7 3 190 1214 60.0 685.8 120.7 3 191 1215 148.0 685.8 120.7 3 192 1216 172.4 685.8 120.7 3 193 1301 235.0 685.8 120.7 3 194 1302 284.8 685.8 120.7 3 195 1303 300.4 685.8 120.7 3 196 1304 397.2 685.8 120.7 3 197 1305 453.4 685.8 120.7 3 198 1306 3.8 685.8 191.7 3 199 1307 60.0 685.8 191.7 3 200 1308 148.0 685.8 191.7 3

PAGE 142

142 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 201 1309 172.4 685.8 191.7 3 202 1310 235.0 685.8 191.7 3 203 1311 284.8 685.8 191.7 3 204 1312 300.4 685.8 191.7 3 205 1313 397.2 685.8 191.7 3 206 1314 453.4 685.8 191.7 3 207 1315 233.0 8.2 202.8 5 208 1316 233.0 83.5 202.8 5 209 1317 233.0 158.8 202.8 5 210 1318 233.0 234.1 202.8 5 211 1319 233.0 309.4 202.8 5 212 1320 233.0 384.7 202.8 5 213 1321 233.0 459.9 202.8 5 214 1322 233.0 535.2 202.8 5 215 1323 233.0 610.5 202.8 5 216 1324 233.0 677.6 202.8 5 217 1325 148.0 45.9 201.2 6 218 1326 148.0 121.2 201.2 6 219 1327 148.0 196.4 201.2 6 220 1328 148.0 271.7 201.2 6 221 1329 148.0 347.0 201.2 6 222 1330 148.0 422.3 201.2 6 223 1331 148.0 497.6 201.2 6 224 1332 148.0 572.9 201.2 6 225 1333 148.0 644.0 201.2 6 226 1334 300.4 45.9 201.4 5 227 1335 300.4 121.2 201.4 5 228 1336 300.4 196.4 201.4 5 229 1337 300.4 271.7 201.4 5 230 1338 300.4 347.0 201.4 5 231 1339 300.4 422.3 201.4 5 232 1340 300.4 497.6 201.4 5 233 1341 300.4 572.9 201.4 5 234 1342 300.4 644.0 201.4 5 235 1343 19.1 234.1 198.5 6 236 1344 52.7 234.1 199.2 6 237 1345 109.9 234.1 200.4 6 238 1346 186.1 234.1 202.0 6 239 1347 266.7 234.1 202.1 5 240 1348 338.5 234.1 200.6 5

PAGE 143

143 Table C 1. Continued Tap Index Tap ID X (mm) Y (mm) Z (mm) Surface 241 1349 395.6 234.1 199.4 5 242 1350 433.7 234.1 198.6 5 243 1351 266.7 459.9 202.1 5 244 1352 338.5 459.9 200.6 5 245 1353 395.6 459.9 199.4 5 246 1354 433.7 459.9 198.6 5 247 1355 19.1 459.9 198.5 6 248 1356 52.7 459.9 199.2 6 249 1357 109.9 459.9 200.4 6 250 1358 186.1 459.9 202.0 6 251 1359 4.4 45.9 198.2 6 252 1360 33.7 45.9 198.8 6 253 1361 71.8 45.9 199.6 6 254 1362 224.2 45.9 202.8 6 255 1363 233.0 45.9 202.8 5 256 1365 376.6 45.9 199.8 5 257 1366 414.7 45.9 199.0 5 258 1367 452.8 45.9 198.2 5 259 1368 4.4 644.0 198.2 6 260 1369 33.7 644.0 198.8 6 261 1370 71.8 644.0 199.6 6 262 1371 224.2 644.0 202.8 6 263 1372 233.0 644.0 202.8 5 264 1373 376.6 645.8 199.8 5 265 1374 414.7 644.0 199.0 5 266 1375 452.8 644.0 198.2 5

PAGE 144

144 Table C 2. Vertices of WERFL building model (1:20) Surface X (mm) Y (mm) Z (mm) Building Feature 1 0.0 0.0 0.0 Wall 457.2 0.0 0.0 457.2 0.0 198.1 228.6 0.0 202.9 0.0 0.0 198.1 2 0.0 0.0 0.0 Wall 0.0 0.0 198.1 0.0 685.8 198.1 0.0 685.8 0.0 3 0.0 685.8 198.1 Wall 228.6 685.8 202.9 457.2 685.8 198.1 457.2 685.8 0.0 0.0 685.8 0.0 4 0.0 685.8 198.1 Wall 228.6 685.8 202.9 457.2 685.8 198.1 457.2 685.8 0.0 0.0 685.8 0.0 5 228.6 685.8 202.9 Roof 457.2 685.8 198.1 457.2 0.0 198.1 228.6 0.0 202.9 6 0.0 685.8 198.1 Roof 228.6 685.8 202.9 228.6 0.0 202.9 0.0 0.0 198.1

PAGE 145

145 LIST OF REFERENCES ASCE 7 10 (2010). Minimum Design Loads for Buildings and Other Structures, American Society of Civil Engineers, Reston, VA. ACI Committee, American Concrete Institute, & International Organization for Standardization. (2008). Building code requirements for structural concrete (ACI 318 08) and commentary. American Concrete Institute. Akon, A. F., & Kopp, G. A. (2016). Mean pressure distributions and reattachment lengths for roof separation bubbles on low rise buildings. Journal of Wind Engineering and Industrial Aerodynamics 155 115 125. Antonia, R. A., & Atkinson, J.D. (1973). High order moments of Reynolds shear stress fluctuations in a turbulent boundary layer. Journal of Fluid Mechanics 58 (03), 581 593. Arora, J. (2004). Introduction to optimum design Academic Press. Arora, J. S. (2000). Methods for discrete variable structural optimization. In Advanced Technology in Structural Engineering (pp. 1 8). Azad, S. K., & Hasanebi, O. (2015). Computationally efficient discrete sizing of steel frames via guided stochastic search heuristic. Computers & Structures 156 12 28. Balderrama, J. A., Masters, F. J., Gurley, K. R., Prevatt, D.O., Aponte Bermdez, L. D., Reinhold, T. A., et al. (2011). The Florida Coastal Monitoring Program (FCMP): A Review. Journal of Wind Engineering and Industrial Aerodynamics, 99 (9), 979 995. Balde rrama, J. A., Masters, F. J., & Gurley, K. R. (2012). Peak factor estimation in hurricane surface winds. Journal of Wind Engineering and Industrial Aerodynamics 102 1 13. Baldocchi, D. D., & Hutchison, B. A. (1987). Turbulence in an almond orchard: verti cal variations in turbulent statistics. Boundary Layer Meteorology 40 (1 2), 127 146. Baldocchi, D. D., & Meyers, T. P. (1988). Turbulence structure in a deciduous forest. Boundary Layer Meteorology 43 (4), 345 364. Bendat, J. S., & Piersol, A. G. (2000). Random data analysis and measurement procedures. Blocken, B. (2014). 50 years of computational wind engineering: past, present and future. Journal of Wind Engineering and Industrial Aerodynamics 129 69 102.

PAGE 146

146 Bhm, M., Finnigan, J. J., Raupach, M. R., & Hu ghes, D. (2013). Turbulence structure within and above a canopy of bluff elements. Boundary layer meteorology 1 27. Camp, C. V., Bichon, B. J., & Stovall, S. P. (2005). Design of steel frames using ant colony optimization. Journal of Structural Engineerin g 131 (3), 369 379. Camp, C. V. (2007). Design of space trusses using Big Bang Big Crunch optimization. Journal of Structural Engineering 133 (7), 999 1008. Chan, C. M., & Grierson, D. E. (1993). An efficient resizing technique for the design of tall steel buildings subject to multiple drift constraints. The Structural Design of Tall and Special Buildings 2 (1), 17 32. Chan, C. M., Grierson, D. E., & Sherbourne, A. N. (1995 ). Automatic optimal design of tall steel building frameworks. Journal of Structural Engineering 121 (5), 838 847. Chan, C. M., Huang, M. F., & Kwok, K. C. S. (2010). Integrated wind load analysis and stiffness optimization of tall buildings with 3D modes. Engineering structures 32 (5), 1252 1261. Chen, F. (1990). Turbulent characteristics over a rough natural surface part I: turbulent structures. Boundary Layer Meteorology 52 (1 2), 151 175. Counehan, J. (1971). Wind tunnel determination of the roughness l ength as a function of the fetch and the roughness density of three dimensional roughness elements. Atmospheric Environment (1967) 5 (8), 637 642. Cuevas, E., Echavarra, A., & Ramrez Ortegn, M. A. (2014). An optimization algorithm inspired by the States of Matter that improves the balance between exploration and exploitation. Applied intelligence 40 (2), 256 272. Davenport, A. G. (1964, June). Note on the distribution of the largest value of a random function with application to gust loading. In ICE Proceedings (Vol. 28, No. 2, pp. 187 196). Thomas Telford. Davenport, A. G. (1977). Wind Loads on Low Rise Buildings: Final Report of Phases I and II. Part I: Text and Figures Davenport, A. G. (2007). Wind tunnel testing: a general outline. The University of Western Ontario, Faculty of Engineering Science London, Ontario, Canada Eberhart, R. C., & Kennedy, J. (1995, October). A new optimizer using particle swarm theory. In Proceedings of the sixth international symposium on micro machine and human science (Vol. 1, pp. 39 43). Erol, O. K., & Eksin, I. (2006). A new optimization method: big bang big crunch. Advances in Engineering Software 37 (2), 106 111.

PAGE 147

147 ESDU (1974). Characteristics of atmospheric turbulence near the ground. Part I: definitions and general information, Engineering Sciences Data Unit Itm. No. 74030, 74031, London, UK. ESDU (1983). Strong winds in the atmospheric boundary layer, Part 2: Discrete gust speeds, Engineering Sciences Data Unit Itm. No. 83045, London, UK. Fenton, R. G. (1974). A comparison of numerical optimization methods for engineering design. CONSTRAINTS 5 (5), 5. Feury, C., & Geradin, M. (1978). Optimality criteria and mathematical programming in structural weight optimization. Computers & Structures 8 (1), 7 17. Finnigan, J. (2000). Turbulence in plant canopies. Annual review of fluid mechanics 32 (1), 519 571. Foken, T., & Nappo, C. J. (2008). Micrometeorology Springer Science & Business Media. Gartshore, I. S. (1984). Some effects of upstream turbulence on the unsteady lift forces imposed on prismatic two dimensional bodies. J. Fluids Eng 106 (4), 418 424. Haftka, R. T. (1985). Simultaneous analysis and design. AIAA journal 23 (7), 1099 1103. Ha ftka, R. T., & Grdal, Z. (2012). Elements of structural optimization (Vol. 11). Springer Science & Business Media. Harris, R. I., & Deaves, D. M. (1980, November). The structure of strong winds. In Proceedings of the CIRIA conference held on (pp. 12 13). high rise steel buildings using an evolution strategy integrated parallel algorithm. Computers & Structures 89 (21), 2037 2051. Hillier, R., & Cherry, N. J. (1981). The effe cts of stream turbulence on separation bubbles. Journal of wind engineering and industrial aerodynamics 8 (1 2), 49 58. Ho, T. C. E., Surry, D., & Morrish, D. P. (2003). NIST/TTU cooperative agreement windstorm mitigation initiative: Wind tunnel experiment s on generic low buildings. The Boundary Layer Wind Tunnel Laboratory, The University of Western Ontario, London, Ontario, Canada Holmes, J. D. (1981). Non Gaussian characteristics of wind pressure fluctuations. Journal of Wind Engineering and Industrial Aerodynamics 7 (1), 103 108. Holmes, J. D. (1983). Wind loads on low rise buildings: A review Commonwealth Scientific and Industrial Research Organization, Division of Building Research.

PAGE 148

148 Holmes, J. D., & Cochran, L. S. (2003). Probability distributions of extreme pressure coefficients. Journal of Wind Engineering and Industrial Aerodynamics 91 (7), 893 901. Huang, M. F., Chan, C. M., & Lou, W. J. (2012). Optimal performance based design of wind sensitive tall buildings considering uncertainties. Computers & Structures 98 7 16. Hunt, A. (1982). Wind tunnel measurements of surface pressures on cubic building models at several scales. Journal of Wind Engineering and Industrial Aerodynamics 10 (2), 137 163. Haftka, R. T., & Grdal, Z. (2012). Elements of stru ctural optimization (Vol. 11). Springer Science & Business Media. Irwin, P. A. (2006). Exposure categories and transitions for design wind loads. Journal of Structural Engineering 132 (11), 1755 1763. Jackson, P. S. (1981). On the displacement height in th e logarithmic velocity profile. Journal of Fluid Mechanics 111 15 25. Jensen, M. (1958). The model law for phenomena in natural wind. Ingenioren 2 (4), 121 128. Jensen, M., & Franck, N. (1963). Model scale tests in turbulent wind Danish Technical Press. Kareem, A., & Zhao, J. (1994). Analysis of non Gaussian surge response of tension leg platforms under wind loads. Journal of Offshore Mechanics and Arctic Engineering 116 (3), 137 144. Karimpour, A., Kaye, N. B., & Baratian Ghorghi, Z. (2012). Modeling t he neutrally stable atmospheric boundary layer for laboratory scale studies of the built environment. Building and Environment 49 203 211. Kaveh, A., & Talatahari, S. (2009). Particle swarm optimizer, ant colony strategy and harmony search scheme hybridized for optimization of truss structures. Computers & Structures 87 (5), 267 283. Kaveh, A., & Talatahari, S. (2010). A discrete big bang big crunch algo rithm for optimal design of skeletal structures. Asian Journal of Civil Engineering 11 (1), 103 122. Kaveh, A., & Abbasgholiha, H. (2011). Optimum design of steel sway frames using Big Bang Big Crunch algorithm. Asian J Civ Eng 12 (3), 293 317. Kicinger, R ., Arciszewski, T., & DeJong, K. (2005). Evolutionary design of steel structures in tall buildings. Journal of Computing in Civil Engineering 19 (3), 223 238.

PAGE 149

149 Khan, M. R. (1984). Optimality criterion techniques applied to frames having general cross sectio nal relationships. AIAA journal 22 (5), 669 676. Krishna, P. (1995). Wind loads on low rise buildings A review. Journal of wind engineering and industrial aerodynamics 54 383 396. Lettau, H. (1969). Note on aerodynamic roughness parameter estimation on t he basis of roughness element description. Journal of applied meteorology 8 (5), 828 832. Li, L. J., Huang, Z. B., & Liu, F. (2009). A heuristic particle swarm optimization method for truss structures with discrete variables. Computers & Structures 87 (7), 435 443. Lieblein, J. (1974). Efficient Methods of Extreme Value Methodology. Technical Report NBSIR 74 602, National Bureau of Standards, Washington, D.C. Lin, J. X., Surry, D., & Tieleman, H. W. (1995). The distribution of pressure near roof corners of flat roof low buildings. Journal of wind engineering and industrial aerodynamics 56 (2 3), 235 265. LRFD, A. (2010). Manual of steel construction, load and resistance factor design. Chicago: American Institute of Steel Construction Lu, S. S., & Willmarth, W. W. (1973). Measurements of the structure of the Reynolds stress in a turbulent boundary layer. Journal of Fluid Mechanics 60 (03), 481 511. Macdonald, R. W., Griffiths, R. F., & Hall, D. J. (1998). An improved method for the estimation of su rface roughness of obstacle arrays. Atmospheric environment 32 (11), 1857 1864. Maitani, T., & Shaw, R. H. (1990). Joint probability analysis of momentum and heat fluxes at a deciduous forest. Boundary Layer Meteorology 52 (3), 283 300. Marshall, R. D. (19 75). A study of wind pressures on a single family dwelling in model and full scale. Journal of Wind Engineering and Industrial Aerodynamics 1 177 199. Meecham, D., Surry, D., & Davenport, A. G. (1991). The magnitude and distribution of wind induced press ures on hip and gable roofs. Journal of Wind Engineering and Industrial Aerodynamics 38 (2 3), 257 272. Mirjalili, S., Mirjalili, S. M., & Lewis, A. (2014). Grey wolf optimizer. Advances in Engineering Software 69 46 61. Nakagawa, H., & Nezu, I. (1977). Prediction of the contributions to the Reynolds stress from bursting events in open channel flows. Journal of fluid mechanics 80 (01), 99 128.

PAGE 150

150 NRCC (National Research Council Canada, 1995), National Building Code of Canada 1995 (NBCC (199 NBC 1995 Structural Commentaries (Part 4), NRCC, Ottawa, Canada. Oikawa, S., & Meng, Y. (1995). Turbulence characteristics and organized motion in a suburban roughness sublayer. Boundary Layer Meteorology 74 (3), 289 312. Perez, R. E., & Behdinan, K. (2007). Particle swarm approach for structural design optimization. Computers & Structures 85 (19), 1579 1588. Pierre, L. S., Kopp, G. A., Surry, D., & Ho, T. C. E. (2005). The UWO contribution to the NIST aerodynamic database for wind loads on low buildings: Part 2. Comparison of data with wind load provisions. Journal of Wind Engineering and Industrial Aerodynamics 93 (1), 31 59. Raupach, M. R. (1981). Conditional statistics of Reynolds stress in rough wall and smooth wall turbulent boundary layers. Journal of Fluid Mechanics 108 363 382. Raupach, M., Finnigan, J. J., & Brunei, Y. (1996). Coherent eddies and turbulence in vegetation canopies: the mixi ng layer analogy. Boundary Layer Meteorology 78 (3 4), 351 382. Rotach, M. W. (1993). Turbulence close to a rough urban surface part I: Reynolds stress. Boundary Layer Meteorology 65 (1 2), 1 28. Rotach, M. W. (1999). On the influence of the urban roughnes s sublayer on turbulence and dispersion. Atmospheric Environment 33 (24), 4001 4008. Rotach, M. W., Vogt, R., Bernhofer, C., Batchvarova, E., Christen, A., Clappier, A., ... & Mitev, V. (2005). BUBBLE an urban boundary layer meteorology project. Theoretica l and Applied Climatology 81 (3 4), 231 261. Safari, D., Maheri, M. R., & Maheri, A. (2013). On the performance of a modified multiple deme genetic algorithm in LRFD design of steel frames. Iranian Journal of Science and Technology. Transactions of Civil Engineering 37 (C2), 169. Saka, M. P. (1984). Optimum design of space trusses with buckling constraints. In Third International Conference on Space Structures, Elsevier Applied Science Publ., London (pp. 656 660). Shaw, R. H., Tavangar, J., & Ward, D. P. (1983). Structure of the Reynolds stress in a canopy layer. Journal of Climate and Applied Meteorology 22 (11), 1922 1931. Spence, S. M., & Gioffr, M. (2012). Large scale reliability based design optimization of wind excited tall buildings. Probabilistic Engineering Mechanics 28 206 215.

PAGE 151

151 Spence, S. M., & Kareem, A. (2013). Data enabled design and optimization (DEDOpt): tall steel building frameworks. Computers & Structures 129 134 147. Stathopoulos, T (1979). Turbulent wind action on low rise buildings Ph.D. Thesis, The University of Western Ontario, London, Ontario, Canada. Stathopoulos, T. (1980). PDF of wind pressures on low rise buildings. Journal of the structural Division 106 (5), 973 990. Stathopoulos, T., & Surry, D. (1983). Scale effects in wi nd tunnel testing of low buildings. Journal of Wind Engineering and Industrial Aerodynamics 13 (1 3), 313 326. Stathopoulos, T. (1984). Adverse wind loads on low buildings due to buffeting. Journal of Structural Engineering, 110 (10), 2374 2392. doi:10.1061 /(ASCE)0733 9445(1984)110:10(2374) Stathopoulos, T. (2003). Wind loads on low buildings: in the wake of Alan Davenport's contributions. Journal of wind engineering and industrial aerodynamics 91 (12), 1565 1585. Tabak, E. I., & Wright, P. M. (1981). Optimality criteria method for building frames. Journal of the Structural Division 107 (7), 1327 1342. Tieleman, H. W., Reinhold, T. A., & Marshall, R. D. (1978). On the wind tunnel simulation of the atmospheric surface layer for the study of wind loads on low rise buildings. Journal of Wind Engineering and Industrial Aerodynamics 3 (1), 21 38. Tieleman, H. W. (1992). Problems associated with flow modelling procedures for low rise structures. Journal of Wind Engineering and Industrial Aerodynamics 42 (1), 923 934. Tieleman, H. W., Surry, D., & Mehta, K. C. (1996). Full/model scale comparison of surface pressures on the Texas Tech experimental building. Journal of Wind Engineering and Industrial Aerodynamics 61 (1), 1 23. Tieleman, H. W ., Reinhold, T. A., & Hajj, M. R. (1997). Importance of turbulence for the prediction of surface pressures on low rise structures. Journal of wind engineering and industrial aerodynamics 69 519 528. eaching learning based optimization. Engineering Structures 34 225 232. Troldborg, N., Srensen, J. N., Mikkelsen, R., & Srensen, N. N. (2014). A simple atmospheric boundary layer model applied to large eddy simulations of wind turbine wakes. Wind Energy 17 (4), 657 669.

PAGE 152

152 Uematsu, Y., & Isyumov, N. (1999). Wind pressures acting on low rise buildings AMSTERDAM: Elsevier Ltd. doi: 10.1016/S0167 6105(99)00036 7 Vickery, P. J., & Skerlj, P. F. (2005). Hurricane gust factors revisited. Journal of Structural Engineering Von Karman, T. (1948). Progress in the statistical theory of turbulence. Proceedings of the National Academy of Sciences 34 (11), 530 539. Walls, R., & Elvin, A. (2010). Optimizing structures subject to multiple deflection constrai nts and load cases using the principle of virtual work. Journal of structural engineering 136 (11), 1444 1452. Weber, R. O. (1999). Remarks on the definition and estimation of friction velocity. Boundary Layer Meteorology 93 (2), 197 209. Yang, X., Yuan, J ., Yuan, J., & Mao, H. (2007). A modified particle swarm optimizer with dynamic adaptation. Applied Mathematics and Computation 189 (2), 1205 1213. Wind tunnel testing for buildings and other structures (2012). Portland: Ringgold Inc. Walls, R., & Elvin, A (2010). Optimizing structures subject to multiple deflection constraints and load cases using the principle of virtual work. Journal of structural engineering 136 (11), 1444 1452. Wheat, D. L., & Cramer, S. M. (2006). Wood Design Package: National design specification for wood construction with commentary and supplement (Vol. 1). American Forest & Paper Association. Zhu, W., van Hout, R., & Katz, J. (2007). On the flow structure and turbulence during sweep and ejection events in a wind tunnel model canopy Boundary layer meteorology 124 (2), 205 233. Zou, X. K., & Chan, C. M. (2005). An optimal resizing technique for seismic drift design of concrete buildings subjected to response spectrum and time history loadings. Computers & Structures 83 (19), 1689 170 4.

PAGE 153

153 BIOGRAPHICAL SKETCH Pedro Luis Fern ndez Cab n was born in San Juan, Puerto Rico in 1987. He graduated with a Bachelor of Science in civil engineering from the University of Puerto Rico at Mayagez in May 2013. Du ring his time at Mayagez, Fern ndez worked alongside Dr. Luis Aponte in the validation of local wind models. Fern ndez moved to Gainesville, Florida in the fall of 2013 to jo pursue a Ph.D. in civil engineering at the University of Florida. In May 2017, he received a Masters of Engineering in civil engineering ( structural ). Fern ndez secured a post doctorate fellowship at the University of Maryland at College Park that will effectively combine the works presented in this dissertation to coordinate the first cyber physical (CPS) modeling in a boundary layer wind tunnel.