<%BANNER%>

Predicting the Vulnerability of Typical Residential Buildings to Hurricane Damage


PAGE 1

PREDICTING THE VULNERABILITY OF TYPICAL RESIDENTIAL BUILDINGS TO HURRICANE DAMAGE By ANNE D. COPE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

Copyright 2004 by Anne D. Cope

PAGE 3

This work is dedicated to those who have gi ven their lives in the defense of our freedom.

PAGE 4

iv ACKNOWLEDGMENTS I would like to extend si ncere thanks to the many people who made this accomplishment possible. First, I would like to thank my advisor, Dr. Kurt Gurley for his encouragement, support, and guidance, especially when recent world political events became very personal. I would also like to th ank Dr. Jean-Paul Pinelli for his leadership of the engineering team for the Public Loss Hurricane Projection Model. For sincere critique and professional advi ce, I would like to thank Dr Emil Simiu, Dr. Tim Reinhold, and Dr. Peter Vickery. Many thanks also go to the members of my committee and to fellow researchers (especially Liang Zhang, Luis Aponte, and Josh Murphree). I thank the providers of the University of Florida Al umni Scholarship for financial support of my education, and the Florida Department of In surance for funding this research. For their unwavering support and loving advice, I thank my husband, my parents, and my extended family. Lastly, I would like to thank Ad rianne Pickett for her support and warm hospitality during the completion of this project.

PAGE 5

v TABLE OF CONTENTS Page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES.............................................................................................................ix LIST OF FIGURES...........................................................................................................xi ABSTRACT.....................................................................................................................xi x CHAPTER 1 INTRODUCTION........................................................................................................1 Research Hypothesis.....................................................................................................4 Goals and Objectives....................................................................................................5 Summary of Dissertation..............................................................................................5 2 SUMMARY OF PREVIOUS RESEARCH.................................................................7 Background Information on Structural Wind Loads....................................................7 Efforts to Quantify Extreme Wind Loads...................................................................12 Defining the Behavior of Near-Surface Hurricane Winds..................................12 Characterizing Surface Pressures on Structures..................................................14 Characterizing and Codifying Structural Loads..................................................17 Summary of Efforts to Quantify Extreme Wind Loads.......................................17 Post-Damage Investigations.......................................................................................18 Damage Prediction Models.........................................................................................20 Fundamental Concepts in Damage Prediction....................................................20 Damage Prediction Models in the Public Domain..............................................24 Proprietary Damage Prediction Models..............................................................26 Public Loss Hurricane Projection Model............................................................28 3 RESIDENTIAL STRUCTURES IN FLORIDA........................................................30 Sources of Information...............................................................................................31 Florida Hurricane Catastrophe Fund Exposure Database...................................31 County Property Appraiser Databases.................................................................32 Manufactured Home Builder Literature..............................................................33 Post-Damage Investigations................................................................................34 Results of the Building P opulation Investigation.......................................................34

PAGE 6

vi Characterization of Site-Built Homes.................................................................34 Characterization of Manufactured Homes...........................................................40 Building Component Selection............................................................................41 4 STRUCTURAL WIND LOADS FOR TYPICAL.....................................................44 Use and Modification of the ASCE 7-98 Code Provisions to Represent Load Conditions during Ex treme Wind Events..............................................................45 Modifications to Surface Pressure Equations......................................................46 Use and Modifications to Exte rnal Pressure Coefficients...................................48 Main Wind Force Resisting System external pressure coefficients.............49 Component and Cladding extern al pressure coefficients.............................50 Use and Modifications to Inte rnal Pressure Coefficients....................................54 Application of the Modified ASCE 798 Code Provisions to Produce Extreme Wind Event Load Conditions on Se lected Building Components.........................54 Roof Cover and Roof Sheathing Loads...............................................................55 Roof-to-Wall Connection Loads.........................................................................56 Wall Loads...........................................................................................................57 Load Conditions for Openings............................................................................59 Load Conditions for Tie-Down Anchors.............................................................65 Summary of Wind Load Conditions Used in the Simulation Engine.........................65 5 PROBABILISTIC WIND RESISTANCE CAPACITIES FOR RESIDENTIAL DWELLING COMPONENTS...................................................................................67 Fundamental Concepts Applied During the Selection of Load Resistance Values....................................................................................................................68 Site-Built Home Resistance Values............................................................................71 Wind Resistance Capacity of R oof Cover on Site-Built Homes.........................71 Wind Resistance Capacity of Roof Sheathing on Site-Built Homes...................75 Wind Resistance Capacity of Roof -to-Wall Connections on Site-Built Homes..............................................................................................................76 Wind Resistance Capacity of Site-Built Home Walls.........................................79 Wood shear wall capacity............................................................................80 Wood frame out-of-plane load capacity.......................................................81 Wood frame uplift capacity..........................................................................83 Wood frame sheathing capacity...................................................................83 Masonry shear wall capacity........................................................................84 Masonry out-of-plane load capacity.............................................................85 Masonry uplift capacity................................................................................86 Wind Resistance Capacity of Site-Built Home Openings...................................86 Wind resistance capacity of doors for site-built homes...............................87 Wind resistance capacity of garage doors for site-built homes....................87 Wind resistance capacity of windows for site-built homes..........................88 Manufactured Home Resistance Values.....................................................................89 Wind Resistance Capacity of Roof Sheathing and Cover on Manufactured Homes..............................................................................................................90

PAGE 7

vii Wind Resistance Capacity of Roof-t o-Wall Connections for Manufactured Homes..............................................................................................................91 Wall Capacity for Manufactured Homes.............................................................92 Wind Resistance Capacity of Manufactured Home Openings............................93 Wind Resistance Capacity of Tie-Down Anchors...............................................93 Summary of Resistance Values Used in Structural Damage Simulation...................94 6 SIMULATION ENGINE............................................................................................97 Selection of Structural Type and Definition of Geometry..........................................97 Variables for Site-Built Homes...........................................................................98 Variables for Manufactured Homes..................................................................100 Loop for Angle of Incidence.....................................................................................101 Loop for Wind Speed................................................................................................102 Loop for the Simulated Homes.................................................................................102 Randomization of Wind Speed a nd Pressure Coefficients................................103 Initial Load Calculations...................................................................................105 Sampling of Resistances....................................................................................105 Roof cover and roof sheathing resistance sampling...................................106 Roof-to-wall connection resistance sampling............................................107 Wall resistance sampling............................................................................109 Opening resistance sampling......................................................................111 Tie-down anchor resistance sampling........................................................112 Initial Failure Check..........................................................................................112 Initial failure check for roof sheathing.......................................................112 Initial failure check for walls.....................................................................113 Initial failure check for openings...............................................................115 Internal Pressure Evaluation and Recalculation of Loads.................................116 Final Failure Check and Damage Tally.............................................................117 Structural Damage Output Files...............................................................................122 Summary...................................................................................................................123 7 STRUCTURAL DAMAGE VALIDATION AND RESULTS................................124 Structural Damage Validation..................................................................................126 NAHB Report on Hurricane Andrew................................................................127 Application of the NAHB Report Data as a Validation Tool............................128 Validation of Indi vidual Components...............................................................130 Validation of window damage...................................................................132 Validation of masonry wall damage...........................................................135 Validation of wood frame wall damage.....................................................137 Validation of roof-to-wall connection damage..........................................138 Validation of roof sheathing damage.........................................................140 Validation of roof cover damage................................................................142 Investigation of Selected Topics...............................................................................144 Investigation of the Batch Sele ction Method for Roof Sheathing.....................144 Investigation of the Batch Selectio n Method for Roof-to-Wall Connections...146

PAGE 8

viii Investigation of the Difference between Hip and Gable Roofs.........................148 Structural Damage Results.......................................................................................150 Results for Site-Built Homes in the S outh Florida and Florida Keys Region...151 Results for Manufactured Homes......................................................................153 Summary...................................................................................................................156 8 APPLICATION OF RESULTS AND CONCLUSION...........................................157 Relating Structural Damage to Monetary Loss........................................................158 Cost Estimate Model.........................................................................................159 Insured Loss Model...........................................................................................162 Research Contributions.............................................................................................164 Future Uses of the Structural Damage Model..........................................................165 APPENDIX A SOUTH / KEYS REGION CONCR ETE BLOCK GABLE ROOF (CBG) HOMES....................................................................................................................167 B SOUTH / KEYS REGION CONCRETE BLOCK HIP ROOF (CBH) HOMES.....175 C SOUTH / KEYS REGION WOOD FRAM E GABLE ROOF (WG) HOMES........183 D SOUTH / KEYS REGION WOOD FR AME HIP ROOF (WH) HOMES...............191 E FLORIDA MANUFACTURE D SINGLEWIDE HOMES......................................199 F FLORIDA MANUFACTURED DOUBLEWIDE HOMES....................................205 G FLORIDA PRE-HUD CODE MANUFACTURED HOMES.................................211 LIST OF REFERENCES.................................................................................................217 BIOGRAPHICAL SKETCH...........................................................................................222

PAGE 9

ix LIST OF TABLES Table page 3-1. Four most common structural types.........................................................................36 3-2. Population of most comm on structural types in de fined geographic regions..........36 3-3. Additional structural types.......................................................................................37 3-4. Population of additional structural types in defined geographic regions.................37 3-5. Structural type models for each geographic region..................................................38 4-1. Zones 1-6 MWFRS pressure coefficients................................................................49 4-2. Zones 1E-6E MWFRS pressure coefficients...........................................................49 4-3. Roof zone C&C pressure coefficien t values for selected roof pitches.....................53 4-4. Wall C&C pressure coefficient values.....................................................................54 4-5. Summary of load conditions applie d to simulate extreme wind events...................66 5-1. Manufacturer’s uplift capacity fo r typical roof-to-wall connections.......................78 5-2. Mean failure pressures fo r typical unprotected windows.........................................89 5-3. Site-built home summary of wind resistance capacities..........................................95 5-4. Manufactured home summary of wind resistance capacities...................................96 6-1. Site-built home dimensions....................................................................................100 6-2. Manufactured home dimensions............................................................................101 7-1. Modeled structural types........................................................................................125 7-2. Structural types with damage base d on combinations of modeled buildings........125 7-3. Hurricane Andrew damages surveyed in the 1993 NAHB report..........................127 7-4. Wood frame home damages surveyed in the 1993 NAHB report..........................128

PAGE 10

x 7-5. Window damage from Hurrican e Andrew vs. simulated data...............................132 7-6. Masonry wall damage from Hurricane Andrew vs. simulated data.......................135 7-7. Wood frame wall damage from Hurricane Andrew vs. simulated data.................137 7-8. Roof-to-wall connection damage from Hurricane Andrew vs. simulated data......139 7-9. Roof sheathing damage from Hurri cane Andrew vs. simulated data.....................141 7-10. Roof cover damage from Hurricane Andrew vs. simulated data...........................143 8-1. Structural repair cost ratios for Central Florida masonry homes...........................160 8-2. Non-structural repair cost ratio s for Central Florida masonry homes....................160

PAGE 11

xi LIST OF FIGURES Figure page 2-1. Wind speed vs. height profiles.................................................................................8 2-2. Pressure locations for the differentia l pressure calculation in Equation 2-8.........10 2-3. Pressure tap locations and wind angles..................................................................15 2-4. Ratio of aggregate pressure to maximum uplift capacity......................................16 2-5. Example probability distribution func tion of damage at a given wind speed........21 2-6. Vulnerability curve generation..............................................................................21 2-7. Fragility curve generation for 60% overall structural damage..............................22 2-8. Fragility curve for the damage st ate of 60% overall structural damage................23 2-9. Family of fragility curves fo r a particular structural type......................................23 3-1. Regional boundaries for building classification....................................................35 3-2. Distribution of conventio nal (site-built) home roof pitch values according to the National Association of Home Builders Research Center...............................39 3-3. Distribution of manufactured home roof pitch values according to the National Association of Home Builders Research Center.....................................41 3-4. Structural components selected for modeling in the hurricane damageprediction simulation engine..................................................................................42 4-1. MWFRS pressure zones.........................................................................................50 4-2. C&C roof pressure zones.......................................................................................51 4-3. C&C wall pressure zones.......................................................................................51 4-4. Roof pressure zones for winds perpendicular to the ridgeline...............................52 4-5. Roof pressure zones for wi nds parallel to the ridgeline.........................................52

PAGE 12

xii 4-6. Roof pressure zones for cornering winds...............................................................52 4-7. Method of determining shear wa ll loads from MWFRS pressures........................57 4-8. Tributary area for C&C pressures tr ansferred into lateral connections on wood frame walls...................................................................................................58 4-9. Tributary area after significant r oof-to-wall connection damage for C&C pressures transferred into latera l connections on wood frame walls.....................58 4-10. Values of the parameter A used in the determination of missile impact................61 4-11. Values of the parameter B used in the determination of missile impact................62 4-12. Values of the parameter D used in the determination of missile impact...............64 4-13. Probability of missile strike caus ing breakage of a medium (3.5 x 5 ft) window on a 44 ft long windward wall.................................................................65 5-1. Gaussian distributions with a mean of 100 units and varying coefficients of variation.................................................................................................................69 5-2. Lognormal vs. Gaussian for a mean of 100 units and coefficient of variation of 0.2......................................................................................................................70 5-3. Truncated Gaussian distribution with a mean of 100 units and a COV of 0.4.......71 5-4. Typical arrangement of tie-down anchors for manufactured homes.....................94 6-1. Structural damage simu lation engine flowchart....................................................98 6-2. Angles of wind incidence used for each wind speed...........................................102 6-3. Flowchart for realizations of a structural type.....................................................103 6-4. Modeled structural components...........................................................................106 6-5. Batch sampling method for roof-to-wall connections.........................................109 6-5. Location of forces for the overturni ng failure check on manufactured homes....121 7-1. Histograms of window damage on South/Keys CBG homes..............................133 7-2. Window damage vulnerabil ity of South/Keys CBG homes................................134 7-3. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys CBG homes..........................................................................................................134 7-4. Wall damage vulnerability of South/Keys CBG homes......................................136

PAGE 13

xiii 7-5. Fragility curves for 1, 2, 3, a nd 4 damaged walls for South/Keys CBG homes...................................................................................................................136 7-6. Wall damage vulnerabilit y of South/Keys WG homes........................................138 7-7. Fragility curves for 1, 2, 3, and 4 damaged walls for South/Keys WG homes....138 7-8. Roof-to-wall connection damage vul nerability of South/Keys CBG homes.......140 7-9. Fragility curves for 2%, 5%, 10% 25%, and 50% roof-to-wall connection damage for South/Keys CBG homes...................................................................140 7-10. Roof sheathing vulnerability of South/Keys CBG homes...................................141 7-11. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof sheathing damage for South/Keys CBG homes......................................................................................142 7-12. Roof cover vulnerability of South/Keys CBG homes..........................................143 7-13. Fragility curves for 2%, 5%, 10% 25%, and 50% roof cover damage for South/Keys CBG homes......................................................................................144 7-14. Histograms of roof sheathing damage on South/Keys CBG homes....................145 7-15. Histograms of roof-to-wall connec tion damage on South/Keys CBG homes.....147 7-16. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof-to-wall connection damage on South/Keys CBG homes....................................................................147 7-17. Histograms of roof-to-wall connecti on damage on South/Keys concrete block homes.........................................................................................................149 7-18. Histograms of roof sheathing damage on South/Keys concrete block homes.....149 7-19. South/Keys CBG homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls.............................................................................151 7-20. South/Keys CBH homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls.............................................................................152 7-21. South/Keys WG homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls.............................................................................152 7-22. South/Keys WH homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls.............................................................................153 7-23. Singlewide manufactured homes m ean damages for roof cover, roof sheathing, roof-to-wall connections, and walls....................................................154

PAGE 14

xiv 7-24. Doublewide manufactured homes mean damages for roof cover, roof sheathing, roof-to-wall connections, and walls....................................................155 7-25. Pre-HUD Code singlewide manufactur ed homes mean damages for roof cover, roof sheathing, roof-towall connections, and walls.................................155 8-1. Preliminary results of the relation of structural damage to insurable content loss compared with insurance claims data from Hurricane Andrew...................161 A-1. Concrete block gable roof South/Ke ys Region home comparative levels of roof cover, roof sheathing, connect ions, wall, and gable end sheathing damage.................................................................................................................168 A-2. Vulnerability to roof cover da mage for South/Keys CBG homes.......................168 A-3. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof cover for South/Keys CBG homes......................................................................................169 A-4. Vulnerability to roof sheathing damage for South/Keys CBG homes.................169 A-5. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof sheathing for South/Keys CBG homes.................................................................................170 A-6. Vulnerability to roof-to-wall c onnection damage for South/Keys CBG homes...................................................................................................................170 A-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region CBG homes................................................171 A-8. Vulnerability to wall damage for South/Keys Region CBG homes....................171 A-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region CBG homes..........................................................................................................172 A-10. Vulnerability to window damage for South/Keys Region CBG homes..............172 A-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region CBG homes.............................................................................................173 A-12. Vulnerability to exterior door dama ge for South/Keys Region CBG homes......173 A-13. Fragility curves for 1 and 2 damaged exterior doors for South/Keys Region CBG homes..........................................................................................................174 A-14. Vulnerability to garage door dama ge for South/Keys Region CBG homes........174 B-1. Concrete block hip roof South/Ke ys Region home comparative levels of roof cover, roof sheathing, connect ions, wall, and gable end sheathing damage.................................................................................................................176

PAGE 15

xv B-2. Vulnerability to roof cover da mage for South/Keys CBH homes.......................176 B-3. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof cover for South/Keys CBH homes......................................................................................177 B-4. Vulnerability to roof sheathing damage for South/Keys CBH homes.................177 B-5. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof sheathing for South/Keys CBH homes.................................................................................178 B-6. Vulnerability to roof-to-wall c onnection damage for South/Keys CBH homes...................................................................................................................178 B-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region CBH homes................................................179 B-8. Vulnerability to wall damage for South/Keys Region CBH homes....................179 B-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region CBH homes..........................................................................................................180 B-10. Vulnerability to window damage for South/Keys Region CBH homes..............180 B-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region CBH homes.............................................................................................181 B-12. Vulnerability to exterior door da mage for South/Keys Region CBH homes......181 B-13. Fragility curves for 1 and 2 damage d exterior doors for South/Keys Region CBH homes..........................................................................................................182 B-14. Vulnerability to garage door dama ge for South/Keys Region CBH homes........182 C-1. Wood frame gable roof South/Keys Region home comp arative levels of roof cover, roof sheathing, connections, wall, and gable end sheathing damage........184 C-2. Vulnerability to roof cover damage for South/Keys WG homes.........................184 C-3. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof cover for South/Keys WG homes........................................................................................185 C-4. Vulnerability to roof sheathing damage for South/Keys WG homes..................185 C-5. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof sheathing for South/Keys WG homes..................................................................................186 C-6. Vulnerability to roof-to-wall c onnection damage for South/Keys WG homes...................................................................................................................186

PAGE 16

xvi C-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region WG homes.................................................187 C-8. Vulnerability to wall damage for South/Keys Region WG homes......................187 C-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WG homes...................................................................................................................188 C-10. Vulnerability to window damage for South/Keys Region WG homes................188 C-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region WG homes...............................................................................................189 C-12. Vulnerability to exterior door dama ge for South/Keys Region WG homes........189 C-13. Vulnerability to exterior door dama ge for South/Keys Region WG homes........190 C-14. Vulnerability to garage door dama ge for South/Keys Region WG homes..........190 D-1. Wood frame hip roof South/Keys Re gion home comparative levels of roof cover, roof sheathing, connections, wall, and gable end sheathing damage........192 D-2. Vulnerability to roof cover da mage for South/Keys WH homes.........................192 D-3. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof cover for South/Keys WH homes........................................................................................193 D-4. Vulnerability to roof sheathing damage for South/Keys WH homes..................193 D-5. Fragility curves for 2%, 5%, 10%, 25 %, and 50% damage to roof sheathing for South/Keys WH homes..................................................................................194 D-6. Vulnerability to roof-to-wall c onnection damage for South/Keys WH homes...................................................................................................................194 D-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region WH homes.................................................195 D-8. Vulnerability to wall damage for South/Keys Region WH homes......................195 D-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WH homes...................................................................................................................196 D-10. Vulnerability to window damage for South/Keys Region WH homes................196 D-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region WH homes...............................................................................................197 D-12. Vulnerability to exterior door da mage for South/Keys Region WH homes........197

PAGE 17

xvii D-13. Vulnerability to exterior door da mage for South/Keys Region WH homes........198 D-14. Vulnerability to garage door dama ge for South/Keys Region WH homes..........198 E-1. Singlewide manufactured home comparative levels of roof cover, roof sheathing, connections, wall, and gable end sheathing damage..........................200 E-2. Vulnerability to roof cover dama ge for singlewide manufactured homes...........200 E-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for singlewide manuf actured homes..........................................................................201 E-4. Vulnerability to roof sheathing da mage for singlewide manufactured homes....201 E-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for singlewide manufactured homes....................................................................202 E-6. Vulnerability to roof-to-wa ll connection damage for singlewide manufactured homes............................................................................................202 E-7. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof-to-wall connections for singlewid e manufactured homes................................................203 E-8. Vulnerability to wall sheathing da mage for singlewide manufactured homes....203 E-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for singlewide manufactured homes....................................................................204 F-1. Doublewide manufactured home compar ative levels of roof cover, roof sheathing, connections, wall, and gable end sheathing damage..........................206 F-2. Vulnerability to roof cover damage for doublewide manufactured homes.........206 F-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for doublewide manufactured homes........................................................................207 F-4. Vulnerability to roof sheath ing damage for doublewide manufactured homes...................................................................................................................207 F-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for doublewide manufactured homes...................................................................208 F-6. Vulnerability to roof-to-wall connection damage for doublewide manufactured homes............................................................................................208 F-7. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof-to-wall connections for doublewide manufactured homes...............................................209

PAGE 18

xviii F-8. Vulnerability to wall sheathi ng damage for doublewide manufactured homes...................................................................................................................209 F-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for doublewide manufactured homes...................................................................210 G-1. Pre-HUD Code manufactured home comp arative levels of roof cover, roof sheathing, connections, wall, and gable end sheathing damage..........................212 G-2. Vulnerability to roof cover damage for pre-HUD Code manufactured homes...................................................................................................................212 G-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for pre-HUD Code manufactured homes...................................................................213 G-4. Vulnerability to roof sheathi ng damage for pre-HUD Code manufactured homes...................................................................................................................213 G-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for pre-HUD Code manufactured homes.............................................................214 G-6. Vulnerability to roof-to-wa ll connection damage for pre-HUD Code manufactured homes............................................................................................214 G-7. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof-to-wall connections for pre-HUD Code manufactured homes.........................................215 G-8. Vulnerability to wall sheathing damage for pre-HUD Code manufactured homes...................................................................................................................215 G-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for pre-HUD Code manufactured homes.............................................................216

PAGE 19

xix Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PREDICTING THE VULNERABILITY OF TYPICAL RESIDENTIAL BUILDINGS TO HURRICANE DAMAGE By Anne D. Cope August 2004 Chair: Kurtis Gurley Cochair: Gary Consolazio Major Department: Civil and Coastal Engineering Hurricanes have caused billions of dollars in losses in the United States and could devastate up to $1.5 trillion worth of existi ng structures in Florida alone. The population density on Florida’s 1200-mile coastline co ntinues to grow, and potential losses will continue to mount. The insurance industry a nd the Florida insuran ce regulatory agency both require a means of estimating these expect ed losses. Only a handful of studies exist in the public domain to predict aggregate hurricane damage. Most published studies use regression techniques with post-disaster inve stigations or claims data to develop vulnerability curves. This approach is hi ghly dependent on the type of construction common to the areas represented in the data, thus limiting the predictive capabilities to regions of similar construction. A promising approach used by one commercial model estimates vulnerability by e xplicitly accounting for the re sistance capacity of building components and load produced by wind. This so-called component approach applies

PAGE 20

xx claims data from previous storms as a valida tion (rather than calibra tion) tool, and can be readily adapted to differe nt regions with varying predominant construction. The Florida Department of Insurance (FDOI) sponsored the development of a public hurricane risk model. The goal of this ongoing project is to predict hurricane windinduced insurance losses by zip code for the State of Florid a, on an annualized basis and for predefined scenarios. The engineering t eam is responsible for relating specific wind speeds to predicted losses for typical resident ial buildings in the state of Florida. Our study developed a probabilistic model predicting structural damage from hurricane winds in Florida. The core of this model is a Monte Carlo Simulation engine that generates damage information for typical Florida homes, using a component approach. The simulation compares deterministic wind loads, and the probabilistic capacity of vulnerable building components to resist these loads, to de termine the probability of damage. In this manner, probabilistic struct ural damage is identified over a range of assigned wind speeds. Monetary loss associ ated with structural damage and the likelihood of occurrence for discrete wind speeds will be determ ined by models under development by other groups in the project.

PAGE 21

1 CHAPTER 1 INTRODUCTION Windstorms produce billions of dollars in property and other economic losses annually in the United States. Before Hurrican e Andrew struck Flor ida and Louisiana in 1992, many insurance-industry experts thought the worst possibl e windstorm would cause no more than $8 billion in insured property damage (Insurance Information Institute May 2001 Update). Hurricane Andrew resulted in $15.5 billion in insured property losses, $26.5 billion in total losses, and 61 fatalities [1]. Before Hurricane Hugo’s landfall in 1989, no hurricane had result ed in claims in excess of $1 billion (Insurance Information Institute May 2001 Upda te). Hugo resulted in $7 billion in total losses, and 86 fatalities. In 1999, Hurricane Floyd resulted in $6 bill ion in total losses, and 56 fatalities [1]. According to th e National Oceanographi c and Atmospheric Administration, wind-related disasters far outpace other natural disast ers in total loss in the United States. In light of these facts, efforts to estimate expected losses and mitigate damage to residential structures from highwind events are necessary to maintain the viability of the increasing coastal population and infrastructure along the coastal United States. The effort to predict and mitigate hurricane damage is of particular importance in the state of Florida (w hich lies in an area vulnerable to these high-wind events, and has a large and increasing coastal population). Both the insuran ce industry and the Florida insurance regulatory agency require a means of predicting future losses. In the public domain, only a handful of studies predict aggregate hurricane damage. Most published

PAGE 22

2 studies use regression technique s to develop vulnerability curves from post-disaster investigations or available insurance claims da ta. Several of these st udies are detailed in Chapter 2. This approach of using data from previous storms is highly dependent on the type of construction common to the areas repr esented in the data. Thus, the vulnerability curves developed in the studies are limited to predicting damage for regions of similar construction. For example, the observed-dama ge studies and claims data from Hurricane Andrew can be used to develop a relations hip between wind speed and probable damage to homes of typical South Fl orida construction (mostly maso nry). These relations would not be suitable for predicting damage from hurricane winds to homes in North Florida, where timber construction is more comm on. Thus, regression t echniques must be enhanced with methodologies that do not require large observed-damage data sets. A promising approach used by one commercial model estimates vulnerability by explicitly accounting for the resistance cap acity of individual bu ilding components and load produced by wind, within a probabilis tic framework. This so-called component approach applies claims data from previ ous storms as a validation (rather than calibration) tool, and can be readily adapted to differe nt regions with varying predominant construction. While the overall framework of the Federal Emergency Management Agency sponsored HAZUS model has been discussed in public literature [2-4], the complex wind-structure interacti on choices and assumpti ons involved in this commercial model are not presented in full de tail. Because this particular model and other commercial models sponsored by the in surance industry are largely proprietary, many of the details and assumptions used in th eir analysis are not available for public use or critique.

PAGE 23

3 In response to the need for a public model to predict hu rricane wind-induced insurance losses, the Florida Department of Financial Services sponsored a multiuniversity project coordinated by the Intern ational Hurricane Research Center, and involving meteorological, engi neering, actuarial, and com puter-resource components. The product of this effort is the prediction of hurricane wind-induced insurance losses for residential structures by zip code in Florida, on both an annualized basis and for predefined scenarios (specific hurricanes). The engineering team is responsible for relating specific wind speeds to predicted losses, for typical residential buildings in the state of Florida. The Un iversity of Florida’s contribution to the project, presented in this dissertation, is the development of the model that defines the complex relationships betw een hurricane wind speed and the resultant structural damage, in a probabilistic framewor k. The core of this model is a Monte Carlo simulation engine that uses a component appr oach to generate damage information for typical Florida homes. The Monte Carlo simulation compares deterministic wind loads and the probabilistic load-re sistance capacity of building components to determine the probability of damage. In this manner, probabili stic structural damage is identified over a range of assigned wind speeds. This compone nt approach may be developed based on laboratory studies of the capacity of indi vidual components, and proper accounting of load paths and load sharing among components. This allows great flexibility with regard to the types of structures th at can be modeled. The development of damage relations is not dependent on the existence of observed hurricane wind damage, but such information can be used to validate and refine the model. The component approach also allows the incorporation of future know ledge (such as additional capacity information on various

PAGE 24

4 components) and the effects of mitigation m easures (such as gable end bracing). All of the data, decisions, and assumptions used in the model development are available for public critique. Research Hypothesis The complex wind-structure interaction that leads to damage of typical residential buildings during hurricane events can be broke n down into three main components: local wind field acting on the buildi ng, structural loads caused by the wind field, and resistance capacity of the building compone nts. If the relationship betw een the local wind field and the structural loads is defined, then the pr oblem of quantifying the risk of wind damage can be addressed by applying a probabilistic framework to the structural loads and resistance capacities of the building components. The level and likelihood of st ructural damage will depend on parameters describing the probabilistic representation of loads and resistance. Significant information on probabilistic wind loading is available through wind tunnel and full-scale data sets, provided the assumption that hurricane wind fi elds can be modeled by the log-law or power-law holds true. These two modeling laws are described in Chapter 2. Laboratory testing and post-storm damage reports pr ovide valuable information on structural resistance. Using this information to simu late the occurrence of hurricane events on typical residential buildings will provide a m easure of the ability of current typically constructed residential buildings to withst and hurricane-force winds. Incorporating new construction practices and retr ofits (which alter component resistances) into the same probabilistic framework will provide a means of calculating the benefit to homeowners of adding hurricane-damage mitigation features to their homes.

PAGE 25

5 Goals and Objectives The focus of the research is the developm ent of a simulation engine that provides the probability of structural damage for typical Florida residential structures as a function of peak gust wind speeds. Structural-damag e information provided by this simulation engine will serve as the backbone for the engineering component of the first publicly available hurricane-wind damage-prediction mode ls for residential structures. This focus can be represented by four research objectives: Select residential models repr esentative of the current building stock in the state of Florida, and identify components of t hose structures for damage-prediction modeling. Quantify the wind-induced loads on the identified components, and select appropriate load paths. Identify the probabilistic cap acities of individual component s to resist wind loads. Create a probability-based system-res ponse model that will simulate the performance and interaction of the co mponents of typical Florida homes, and evaluate their vulnerability during interaction with hurricane winds. Summary of Dissertation The research objectives described above are detailed in Chapters 3 through 8, following a brief summary in Chapter 2 of pr evious work in the field of hurricane damage mitigation. Specificall y, Chapter 3 presents the resu lts of a survey of current building stock to select typi cal residential build ing types and structural components necessary to predict wind damage. Chapters 4 and 5 provide background information and final decisions for the structural wind loads and building component capacities, respectively. Chapter 6 describes the Mont e Carlo simulation engine which uses the determined loads and capacities to predict th e vulnerability of typical Florida homes to hurricane damage. Results obtained from the simulation process are presented in Chapter

PAGE 26

6 7 and will be used to further develop the public hurricane risk model sponsored by the Florida Department of Financial Services a nd coordinated by the In ternational Hurricane Research Center. Conclusions about the model, modeling process, and potential for future use are discussed in Chapter 8.

PAGE 27

7 CHAPTER 2 SUMMARY OF PREVIOUS RESEARCH Previous research in the area of hurrican e damage mitigation can be divided into three main groups: efforts to quantify extrem e wind loads, post-damage investigations, and damage-prediction models. While numerous articles provide accounts of damage from individual storms, few ar ticles exist on the accurate pr ediction of hurricane damage before a storm occurs. Most of the damage-p rediction models that currently exist are proprietary and unavailable to the public Information available from post-damage investigations and current methods of predicti ng future hurricane damage are detailed in the following paragraphs, after an introductory section describing structural wind loads. Background Information on Structural Wind Loads At any given instant, a snapshot of wi nd speed vs. height at a location near a building might resemble the curve in Figure 2-1 A. Removing the turbulent component to consider the mean wind speed over some averaging time at each height increment provides a smooth curve that might resemble the one in Figure 2-1 B. This curve is typically modeled using one of two methods: the log law or the power law. Each method results in a curve similar to the one in Figur e 2-1 B, which has a mean wind speed of zero at the ground surface and a constant mean wind speed at a distance above the ground referred to as gradient height Typically at elevations of 200 meters, the gradient (or reference) height is the level at which th e wind speed is no longer affected by the surface roughness.

PAGE 28

8 Gradient Height Gradient Height Figure 2-1. Wind speed vs. height profiles. A) Typical profile at any given time. B) Mean wind speed profile. The log-law and power-law equations used to model the mean wind-speed profile are given in Equations 2-1 through 2-3 [5]. Equations 2-1 and 2-2 define the log law, while Equation 2-3 defines the power law. 0 *ln 1 ) ( z z u z U (2-1) 0 u (2-2) ref refz z z U z U ) ( ) ( (2-3) In Equations 2-1 and 2-2, U(z) is the mean wind speed at height z, is the Von Karman constant (approximately 0.4), z0 is the roughness length of the terrain over which the wind acts, and *u is the friction velocity (defined by a ratio of the shear stress at the ground surface, 0, and the density of air, ). The roughness length represents the size of a characteristic vortex created as the wind moves over the terrain. The parameters *u and z0 are modified for each type of terrain [5]. In the power-law equation, U(z) is the mean wind speed at height z, is a parameter modified for the type of terrain, and U(zref) is the mean wind speed at reference (or gradient) height, zref. The two methods provide nearly A B

PAGE 29

9 identical results for the mean wind speed at heights above ground where low-rise structures exist. The turbulent component of the wind is most often represented as a Gaussian random variable, with a zero mean and a sta ndard deviation that varies with height. Experimentation reveals that th e standard deviation remains c onstant over the height at which most structures and all lo w-rise structures exist [5]. The standard deviation of the turbulence component in th e direction of wind flow, u; and the turbulence intensity as a function of height, Iu(z), can be calculated using Equations 2-4 and 2-5 (where A is a constant that varies with the roughness length, z0, and has a value of approximately 2.5 for open-country terrain) [5]. *Auu (2-4) ) ( ) (z U z Iu u (2-5) Assuming that mean wind-speed profiles fit the models described above, one can find a relationship between the mean wind speed and the pressure ac ting on areas of the structure. Generally, the effect of the pressu re on the structure is assumed to have two parts: one from the mean wind speed, and one from the gusty or turbulent component. The maximum pressure, pmax, that a component will experien ce as a result of both of these portions can be expresse d as the mean response, pavg, multiplied by a gust factor, G, as shown in Equation 2-6 [6]. avgGp p max (2-6) The most common approach in determini ng design pressures is to place a model building in a wind tunnel, and conduct pressure coefficient studies. This approach was

PAGE 30

10 used to develop the wind loading provisions for the American Society of Civil Engineer’s Minimum Design Loads for Bu ildings and Other Structures (ASCE 7-98) [7]. Roughness elements are placed in the section of the wi nd tunnel preceding the model building such that the mean wind speed vs. height matches th at predicted by either the log law or power law, and the turbulence intensity matche s that predicted by the equation for Iu(z). Pressure at a given location along a stream line can be found using Bernoulli’s equation for steady, inviscid, incompressi ble flow (Equation 2-7), where p is pressure, is the density of the fluid (air in this case), V is the upstream velocity, is the specific weight of the fluid, and z is the depth or height with respect to a known reference [8]. z V p 2 2 1constant along a streamline (2-7) For the case of differential pressure betw een a point on the surface of the model building and a point just in front of the buildi ng at mean roof height (Figure 2-2), Equation 2-8 provides relative change in pressure. V Mean Roof Height 1 2 V Mean Roof Height 1 2 Figure 2-2. Pressure locations for the differential pressure calculation in Equation 2-8 2 2 1 1 2V p p (2-8) In the wind tunnel, differentia l pressure is measured at locations of interest on the building with respect to a reference pressure (u sually located at gradient height). The raw values of differential pressure are converted to pressure coe fficients with respect to the mean roof height of the building by multiply ing by a correction factor taken from the simplified version of Bernoulli ’s equation in Equation 2-8. This process is shown in

PAGE 31

11 Equation 2-9, where Cp is the pressure coefficient at an individual location on the building referenced to mean roof height, P is the measured differential pressure between Locations 2 and 1 as shown in Equation 2-10, Vmn_roof_height is the mean velocity at mean roof height, and Vgradient_height is the mean velocity at gradient height. 2 _ 2 2 2 1 height roof mn height gradient height gradient pV V V P C (2-9) 1 2p p P (2-10) Fluctuating time histories of Cp at the same location on the building for various angles of wind provide a probabilistic descript ion of the pressure coefficient at that location. This information leads to the sele ction of pressure coefficient values for component design. Equations 2-11 and 2-12 from ASCE 7-98 illustrate the calculation of design pressure for components and cladding on low-rise structures [7]. Equation 2-11 shows the calculation of velocity pressure at mean roof height, qh, which is a function of the density of air. The 0.00256 value in Equation 2-11 is for air in English units, Kh is a terrain exposure coefficient, Kzt is a topographic effect fact or to account for speed up over hills, Kd is a directionality factor, V is the design wind speed, and I is an importance factor for the building. The velo city pressure can be thought of as the pressure measured at Location 1 in Figure 2-2 for the true geographic location of the structure being designed. Multiplying qh by a wind tunnel generated pre ssure coefficient provides the pressure acting on the face of the structure at a particular location. The design pressure, p, for each piece is found by multiplying qh by the difference between the external and internal pressure coefficients with the gust factors built in, GCp and GCpi, respectively. This process is shown in Equation 2-12.

PAGE 32

12 I V K K K qd zt h h200256 0 (ASCE 7-98 Eq 6-13) (2-11) pi p hGC GC q p (ASCE 7-98 Eq 6-18) (2-12) The parameters for Equations 2-11 and 212 provided in ASCE 7-98 are intended to envelope the realistic worstcase scenarios that might occu r for the building, so that it will be designed to withstand winds from any angle using a factor of safety worthy of the importance of the building to the community. Efforts to Quantify Extreme Wind Loads The ASCE 7-98 design equations for wind pressure on the surface of a building presented in the previous sect ion are intended to envelope a realistic worst-case scenario. Studies have show that this approach can lead to designs that are st ill un-conservative, or conversely over conservative [9, 10]. Addre ssing the complexities of wind loading and structural response in a more case-specific manner can rectify these design problems. The accuracy of predicting structural wind loads is directly related to both the exactness with which the behavior of near-surface winds can be predicted and the precision of modeling the wind-structure interaction. The behavior of near-surface winds is highly variable and sensitive to numerous localized characteris tics (such as terrain). Moreover, the windstructure interaction is a highly comple x and nonlinear problem, making detailed characterization of wind loads difficult. Efforts to quantify extreme wind loads on structures have sought to increase our understanding of hurricane wind behavior, structural surface pressure characteriza tion, and structural loading effects. Defining the Behavior of Near-Surface Hurricane Winds The ASCE 7-98 design equations provided in the previous section are based on the assumption that the winds encountered by a building will behave in a manner predictable

PAGE 33

13 by either the log law or power law previously described. There is some evidence, though, that this assumption can lead to non-conser vative predictions of maximum gusts [11]. That is, gust factors used to account for dyna mic fluctuations from the mean wind speed may not be suitable when applied to hurri cane winds. In-field hurricane wind data collection is a critical component to characterizing gust st ructure behavior in hurricanes. Extensive data collection has been conduc ted for normal weather prediction, and the operation of aircraft and airports. Unfortunately, this information does not provide adequate data for the characterization of near-surface hurricane winds. Recently, several institutions have begun effort s to collect ground level wind data during hurricane landfall. Some of these include the National Hurrica ne Center, Texas Tech University, Johns Hopkins University, University of Oklahoma, and Clemson University in conjunction with the University of Florida. The University of Florida and Clemson University have begun a hurricane data collection project known as the Florida Coas tal Monitoring Program (FCMP), sponsored by the Florida Department of Community Affairs. One of the main goals of this project is to help improve the fundamental understand ing of the dynamic and turbulent action of high-speed hurricane gusts. This is done through the collection of high-resolution wind velocity, pressure, rainfall, humidity, and temperature data using custom-built instrumentation set in the path of a land-fa lling hurricane. A set of ten and five meter portable towers, equipped with vane and gill anemometers, barometers, hygrometers, and rain gauges, are used to collect the data at se veral sites within the radius of influence of the land falling hurricane. The FCMP has also instrumented houses in South Florida and the Florida Panhandle area with removable pressure transducer s to collect information on

PAGE 34

14 wind forces in the building envelope. The FCMP has produced data sets from named storms over the past four hurricane seasons Findings from these datasets are still preliminary, but continued efforts will produ ce a large dataset from which conclusions can be drawn about the nature of hurricane winds at ground level. Information gained about the turbulence intensity and gust eddy size may show that hurricane wind behavior is unique at low-rise structure levels. Un til that time, however, the assumption that hurricane winds behave in a manner similar to non-hurricane winds will be used. Characterizing Surface Pressures on Structures The values obtained for pGC in Equation 2-12 are depe ndent on the effective wind area for the structural member in question. As the effective wind ar ea decreases, the value of the coefficient increases. This trend resu lts from the gust struct ure of the wind acting on the structure. The turbulent component of the wind acting on a st ructure results from the buffeting action of wind gusts. These gusts are made of large and small clusters of swirling fluid referred to as eddies. The physical size of eddies is an important characteristic. Small eddies hit the struct ure in an uncoordinated manner, while the correlated winds of a large e ddies can affect the entire effective wind area of some structural components at the same time. The degree of linear correlation between pressures at different locations on model bui ldings in a wind tunnel can be used to determine the size of the gusts. More importa ntly, when the buildings are subjected to wind tunnel tests known to model typical open -country conditions, the degree of linear correlation between pressure tap locations can be used to better charact erize the nature of wind pressures on specific building co mponents, such as roof sheathing.

PAGE 35

15 A study conducted by the author to better ch aracterize loads on low rise roof structures explored both the spatial correlati on and probability characteristics of pressure coefficients acting on the roof and eaves of typical gable roof homes [12]. Investigations were conducted to determine if regions of roof sheathing would have both highly correlated surface pressures and a strong devi ation from Gaussian probability. These conditions represent a departure from the assu mptions used in the gust factor approach, and indicate that the roof sheathing is likely more vulnerable to damage than current design methods suggest. Several standard nonGaussian PDF models were associated with different regions in the building e nvelope using goodness of fit procedures comparing models to wind tunnel data. Signif icant combined effects (non-Gaussian loads and high correlation over a surface) were found for cornering winds and winds perpendicular to the gable end of the structure. A follow-up study was conducted to investigat e the results of the combined effects of spatial correlation and non-Gaussian probabi lity content on the aggregate loading of one 4 ft x 8 ft piece of roof sheathing located at the ridgeline on the gable end of a typical structure [13]. A non-Gaussian simulation algor ithm was used to produce realizations of pressure coefficient time histories at the 14 pressure taps represen ting a single piece of roof sheathing (Figure 2-3). 180 90 Row 1 Row 2 Tap 1 Tap 7 180 90 Row 1 Row 2 Tap 1 Tap 7 Figure 2-3. Pressure tap locations and wind angles

PAGE 36

16 Using the pressure coefficient time histories, realizations of the aggregate pressure on the sheathing panel were obtained for cas es of high, moderate, and low correlation among the pressure taps. The effects of the level of correlation ar e significant, as demonstrated through comparison of highe r-moments of the a ggregate pressure coefficient, as a ratio of aggregate pressu re with ASCE 7-98 load conditions, and as a ratio of aggregate pressure with an experime ntally determined uplift capacity. Figure 2-4 shows the ratios of the aggreg ate pressure resulting from a 150-mph. 3-second gust wind (ASCE 7-98 wind conditions for S outh Florida) to the uplift ca pacity of a typical Douglas fir panel with 6d nails in a 6/12 nail pa ttern [14]. Results i ndicate that regions experiencing highly correlated non-Gaussian pressure fields will frequently see loads greater than the capacity of th e system (a ratio larger than 1), while the assumption that the pressure field is not corre lated, but non-Gaussian result s in loads well within the capacity of the system. Complete results ca n be found in Gioffre, Gurley, and Cope (2002). Figure 2-4. Ratio of aggregate pressure to maximum uplift capacity. A) PDF for three levels of correlation among taps. B) Time history for high correlation A B

PAGE 37

17 Characterizing and Codi fying Structural Loads Some wind engineers seek to incorpor ate the non-Gaussian qualities of wind pressures discussed in the previous section into better building codes by using databaseassisted design methods. Current technol ogy allows design engineers to analyze structural responses with nimble accuracy, yet the wind load provisions remain crudely broad brush. Using wind pressure and climatol ogical databases instead of current wind pressure tables and plots w ould provide a more risk-consis tent design and would allow for the use of the structures own influence lines as opposed to ge neric, cookie-cutter structural influences built in to curren t methods [9]. Studies conducted for the development of database-assisted design softwa re reveal the non-Gaussian nature of wind load effects. Specifically, time histories of th e bending moments in a steel frame low-rise structure indicate that the Ga mma distribution is most appr opriate when selecting the maximum peak load [10]. Additional studi es reveal that the inclusion of wind directionality effects allows for a more risk -consistent design over the current approach of using a global dir ectionality factor, Kd, of 0.85 (Eq. 2-11). In fact, the current approach for wind directionality effects may lead to an underestimation of the structural wind load in approximately 10–15% of buildings designed using the 1998 standard [9]. Summary of Efforts to Quan tify Extreme Wind Loads Investigations into the nature of hurrican e near-surface winds from full-scale data and the nature of wind-structure interaction in the form of pressure coefficient data from wind tunnel testing will conti nue at the University of Florida and other institutions. Synthesis of the information gained from these efforts will lead to the development of better building codes and design practices. The current body of information concerning wind surface loads on low-rise structures is not robust enough to allow full incorporation

PAGE 38

18 in the Monte Carlo simulation developed for the FDOI hurricane loss projection model. The simulation engine described in subseque nt chapters relies on aggregate pressures calculated from pressure coefficient zones. These zones are based on values in ASCE 798, but they are modified for directionality using knowledge gained in the previously described research. The details of modificati on are described in Chapter 4. Inclusion of non-Gaussian behavior and correlation betw een surface pressures is a promising topic that could be incorporated into th e developed model at a later date. Post-Damage Investigations Post-damage investigations provide an a ssessment of how structures perform in extreme wind events and can indicate stre ngths and weaknesses in design codes and construction practices. Numerous papers discuss damage from Hurricanes Alicia, Andrew, Hugo, Iniki, and Op al [15-23]. In general, th e reports contain valuable information on types of failures commonly en countered and recommendations to prevent similar failures in future events, but these observations by experts in the field are not backed by statistically significant numbers of evaluations. For example, the damage to buildings in the Houston-Galveston area duri ng Hurricane Alicia was attributed to the lack of adequate hurricane resist ant construction, rather than to the severity of the storm [17, 18]. A similar conclusion was reached on damage to buildings during Hurricane Hugo [23]. A reliability analysis of roof performance during Hurricane Andrew found actual performance to be bette r than predicted by the governin g building code at the time, although the authors stress the need for furthe r research to quantif y statistically both construction characteristics and damage due to storms [21]. Phang also offers several observations of the damage on low-rise buildings caused by Andrew. He found that plywood sheathing performed remarkably bett er than board sheathing, diagonal bracing

PAGE 39

19 was critical at gable ends, a nd gable roofs showed much more structural damage than hip roofs [22]. Research has also been conducted in Australia by Mahendran who gives an overview of the typical damages encountered by low-rise buildings in the tropics, subjected to either hurricanes or severe storms. In addition, he and the Australian scientific community also stress the fact th at full-scale testing is necessary to better predict to behavior of the entire building system when subjected to high-speed winds. While these studies are extremely valuable fo r the development of safer housing, they do not offer a sufficient basis from which to draw reliable quantitative conclusions [24]. The information obtained from these studies does, however, provide a means of validating the results of a probabilistic appr oach relating peak wind speeds to structural damage. One would expect the most common types of failure s detailed in post-disaster reports to be same as the types of failure obtained from Monte Carlo simulations of hurricane-force winds and structural component resistance. Post-damage studies also provide a mean s of estimating the distribution of the building stock in Florida cities. The most comprehensive studies, undertaken by the National Association of Home Builders ( NAHB) following Hurricanes Andrew and Opal, include information on the sample size and t ypes of homes investigated [19, 20]. This information, in combination with data fr om County Property Appraisers and other resources, is useful for predic ting typical sizes and types of homes in other Florida areas, as detailed in Chapter 3. Furthermore, the storm damage reports serve as a benchmark by which to set priorities for research effo rts since these reports identify the building components that experience the most fr equent or most de bilitating damage.

PAGE 40

20 Damage Prediction Models Damage prediction models make use of the current knowledge base to predict damage in future extreme wind events. While several post-damage reports exist in the public domain, there are few damage predic tion models available for public review. Those that can be found follow one of two pa ths. The most common approach is to use post-damage investigation results to create vulne rability or fragility curves for structures (defined in the following section). A second a pproach is to build a probabilistic model to generate structure fragility curves for damage prediction. This latter approach requires some assumptions about the strength of buildi ngs and type of terrain. Simulations are used to create the curves and data sets to cal ibrate and validate the results. The advantage to this approach is the ability to generate rational approximations of damage curves for structural types that have not yet experienced a major hurri cane. Developing curves based on damage data alone requires the existence of large sets of damage data, while the development of curves based on probabil istic assumptions and simulations can incorporate laboratory data sets and engineering judgment when damage data sets are not available. Fundamental Concepts in Damage Prediction Vulnerability and fragility cu rves are both indicators of the ability of a specified structure to withstand hurricane -force winds. To develop each t ype of curve, the level of damage or damage state must be defined. Fo r instance, one could identify damage states involving roof failure, wall failure, or some other type of failure. For demonstration purposes, damage can be thought of as a percen tage of overall structural damage. Each building will either be undamaged (0% dama ge), partially damaged by some percentage,

PAGE 41

21 or totally destroyed (100% damage). At a gi ven wind speed, there will be a distribution of percent damage to structures of the same type (Figure 2-5). Damage p(D) 0% 100% P(D) = Distribution of damage at a given wind speed for a particular building type Damage p(D) 0% 100% P(D) = Distribution of damage at a given wind speed for a particular building type Figure 2-5. Example probability distributi on function of damage at a given wind speed Once the distribution of damage is known over a range of wind speeds, the vulnerability for that type of structure can be determined. The vulnerability curve is a means of measuring the performa nce of the structure, and is generated from the location of the mean percent damage value from th e damage distribution at each wind speed. Figure 2-6 shows the process of vulnerabili ty curve generation from individual PDFs associated with particular wind speeds. Th e generated vulnerability curve defines the mean damage for a particular structural type as a function of wind speed, where mean is defined as the damage level at which 50% of all structures of that type will be less damaged, and 50% more damaged. V1V=V2V3Mean Wind Speed Vulnerability Curve Mean Damage Factor (%) Mean Mean Mean V1V=V2V3Mean Wind Speed Vulnerability Curve Mean Damage Factor (%) Mean Mean Mean Figure 2-6. Vulnerability curve generation

PAGE 42

22 Fragility curves are another means of desc ribing the performance or reliability of the structure. A fragility curve provides the pr obability that a certain level of damage will be met or exceeded at a given wind speed, and can be used to determine how many buildings of similar type in an area will experi ence at least a certain level of damage. This can be thought of as a conditional probabi lity of exceedence. Given the maximum wind speed for a particular wind event, the fragility curve for a type of structure provides the likelihood of damage exceeding a certain thre shold. Figure 2-7 and Figure 2-8 show how the fragility curve for a given structural t ype is determined from available damage distributions at diffe rent wind speeds. The example dem onstrates how to calculate the fragility curve corresponding to 60% damage by setting a threshold in Figure 2-7 and integrating under each damage distribution from the 60% threshold point to the positive extreme. The integrated values (shaded areas in Figure 2-7) become the data points for the fragility curve at each wind speed (Figure 2-8). The limit on the vertical axis of the fragility curve in Figure 2-8 is 1.0, repr esenting a 100% likelihood of occurrence for the given damage state. V1V2V3Mean Wind Speed Vulnerability Curve Mean Damage Factor (%) 60% 12% 35% 80% V1V2V3Mean Wind Speed Vulnerability Curve Mean Damage Factor (%) 60% 12% 35% 80% Figure 2-7. Fragility curve generatio n for 60% overall structural damage

PAGE 43

23 V1V2V3 Fragility Curve D = 60% Mean Wind Speed Probability of Exceedence V1V2V3 Fragility Curve D = 60% Mean Wind Speed V1V2V3 Fragility Curve D = 60% Mean Wind Speed Probability of Exceedence Figure 2-8. Fragility curve for the damage state of 60% overall structural damage Other damage thresholds can be set to generate a family of fragility curves for this structural type (Figure 2-9). To clarify, the vulnerability curve shows the most likely mean damage that will occur to a given stru cture as a function of mean wind speed, while the fragility curve shows the probability of exceeding a specific level of damage as a function of wind speed. With vulnerability and fr agility curves for structure-type ‘A’, the following types of questions can be answered: 1) for a 90 mph. gust, what is the average expected damage to houses of type ‘A’ (usi ng vulnerability curve) and 2) for a 90 mph. gust, what is the likelihood of seeing 80% damage or greater (u sing fragility curve)? V1V2V3 D = 60% Mean Wind Speed Probability of Exceedance D = 80% D = 40% D = 20% V1V2V3 D = 60% Mean Wind Speed Probability of Exceedance D = 80% D = 40% D = 20% Figure 2-9. Family of fragility curv es for a particular structural type

PAGE 44

24 Damage Prediction Models in the Public Domain Damage prediction models in the public sector using the approach of fitting vulnerability curves from post-damage investig ation results include tw o studies that rely heavily on insurance claim information. Th e first of these studies determined the relationship between insurance claim figures and wind speed for Typhoons Mireille and Flo [25]. The second performed a similar anal ysis for Hurricane Andrew [26]. Since the buildings involved in the first study were residential buildings in Japan, the results are not readily applicable to typical residential structures in Flor ida. The second study used data collected from two large insu rance companies in Dade Count y, Florida to calculate the vulnerability function as a pe rcentage of loss vs. mean wi nd speed at gradient height. This information is clearly helpful in determ ining how residential structures typical of those existing in South Florida in 1992 will perform in a hurricane event of similar magnitude. However, this data is a snapshot in time, capturing the damage on structural types that existed when the extreme wind ev ent took place. The data cannot take into consideration improvements in building construction over ti me, nor can it be readily applied to areas where the terrain and type of construction are notably different. Others in the public sector have pred icted damage using probability-based simulation models to generate the likeli hood of damage vs. wind speed. One such study presents the vulnerability curve for a fully engineered building usi ng the assumption that the resistance capacity of the building is lognormally distributed [27]. Since the model was developed for engineered structures, the approach is not likely to yield the best results for predicting damage to typical resi dential buildings in the state of Florida. Another study presents a met hod of predicting the percentage of damage within an area as a function of the gradient wind speed, gus t factor, average value of the buildings, and

PAGE 45

25 two parameters which govern the rate of damage increase with wind speed [28]. These last two parameters are empirically dete rmined based on experience and knowledge of the area. Since the results of this study were not reproducible, the model is not considered a reasonable approach for the prediction of dama ge to residential buildings in the state of Florida, given the information currently available. Insurance data from Hurricane Hugo is us ed as an example to illustrate the probabilistic approach presente d in a recent study for long-term risk analysis [29]. The authors calculated and published statistics for hurricane simulation parameters based on previous storms that made landfall in Fl orida, North Carolina, and South Carolina. Simulations of hurricane events over 50 year periods and investiga tion into historical wind speed records were used to predict 50 -year mean recurrence interval (MRI) wind speeds at gradient height for selected coastal areas [5]. These 50-year MRI wind speeds at gradient height were c onverted to ground wind speeds ba sed on the type of terrain present and used in conjunction with fragility curves generated from insurance loss data to predict damage in areas of interest. The authors provided a graphi cal representation of the generated 50-year MRI wind speeds at gradie nt height and a fragility curve generated from two sources: insurance claims in Flor ida after Hurricane Andrew, and claims in South Carolina after Hurricane Hugo. The dama ge levels predicted by this method are the amount of damage likely to recur once every 50 years, or that have a 2% chance of being exceeded annually in the area of interest, prov ided the building stock remains relatively unchanged. The difficulty with using this me thod of damage prediction is the reliance on insurance data from only two events to genera te the fragility curve from which losses are predicted for future storms. Unfortunately, information from which to determine accurate

PAGE 46

26 fragility curves for a certain type of structure or family of structures is currently limited. Even if larger data sets were available from other storms, the models would only be valid when used to predict damage to structures of like-construction. This reliance on postdisaster information restricts the ability to project the e ffects of design modifications, code changes, and retrofit measures on the vulnerability of existing structures. This realization has lead to the pursuit of appr oaches that seek to model damage at the component level rather than for the entire stru cture. Structural risk assessment is then a matter of combining the vulnerability of th e individual parts making up a structure. The so-called component approach allows the flexibility to including new components and retrofit measures, provided lab tests are perf ormed to assess their probabilistic resistance. Proprietary Damage Prediction Models Private sector damage prediction models also exist. In the wake of Hurricane Andrew (which generated in surance claims totaling near ly twice the amount thought possible by experts in the fiel d), private sector insurance industry groups contracted damage prediction models from engineering fi rms to develop a bett er understanding of the risks associated with a hurricane strike s in heavily populated areas. Access to this information is limited, since the projects are largely proprietary. Currently, some information has been published describing the strategy used to predict damage for these projects. One such study used a re-arranged version of the design pressure equation from ASCE 7-98 (Equation 2-13) to ca lculate the wind speeds at which individual components will fail [30]. pi p zt h failure failureGC GC I K K p V 00256 0 (2-13)

PAGE 47

27 In Equation 2-13, pfailure is the statistically sampled failure pressure of the component, Kh is an ASCE-defined terrain exposure coefficient, Kzt is an ASCE-defined topographic effect factor to account for speed up over hills, I is an ASCE-defined importance factor, GCp is a statistically sampled external pressure coefficient, and GCpi is an ASCE-defined internal pressure coeffici ent. After calculating failure speeds for each component, the researchers determined damage histories for buildings during simulated hurricane events. The result of this analysis wa s a vulnerability curve for a particular type of building, which can be used with repla cement cost information to determine probable insurance losses. The methodology for this study has been pu blished, but the vulnerability results are not available. The most recent damage prediction mode l is the HAZUS Multi-hazard model, which addresses wind, flood, and earthquake ha zards. Under the dir ection of the National Institute of Building Sciences and the Fe deral Emergency Management Agency, the HAZUS hurricane model was developed by A pplied Research Associates (ARA) over a period of several years. A preview of this model was released to hurricane prone regions of the United States in 2002 th at allows users to estimate and evaluate disaster relief resources and policies through scenario anal ysis [3]. The information supplied by the preview model includes planning for the number of displaced persons, sheltering requirements, and post-storm debris remova l. ARA has published a description of the hurricane model’s six components: hurricane hazard, terrain, wind pressure, wind borne debris, damage, and losses for buildings The distinct advantage of the HAZUS methodology over previous damage prediction me thods lies in the fact that it is a component-based model rather than a re gression curve fitting model. The HAZUS

PAGE 48

28 model explicitly accounts for the resistan ce capacity of individual building components and wind loading, within a probabilistic fr amework. Using information from British, Australian, and American wind loading code s, as well as boundary layer wind tunnel testing, ARA developed an empirical model for the pressure coefficien ts on the surface of typical buildings. Techniques fo r estimating the risk of wind borne debris impact and the effects of sheltering from nearby buildings we re also developed. This information was used to create a computer simulation tool th at would apply a hurricane wind model (also developed by ARA) to a typical building and evaluate the damage accrued every 15 minutes as a result of wind pressure or wind borne debris impact. Monetary losses resulting from structural da mage were obtained by calculating the replacement cost explicitly for the external portion of the building and implicitly for the internal structure and contents. This model has been validated with available insurance records and is considered to be the state of the art in hu rricane damage prediction. While the framework for the model has been well defined in public literature, many decisions and assumptions used in the determination of wind loads remain proprietary. Public Loss Hurricane Projection Model The Public Loss Hurricane Projection Mode l is currently under development for the Florida Department of Financial Services, wi th a scheduled release date of May, 2005. This multi-university project (coordinated by the International Hurricane Research Center) will predict hurricane wi nd-induced insurance losses fo r residential structures by zip code for the State of Florida, on both an annualized basis and for predefined scenarios (specific hurricanes). Since the model is sponsored in the public domain, the data, decisions, and assumptions used will be ava ilable for public critique. The framework of the model includes a meteorology component to generate probabilistic information about

PAGE 49

29 wind speeds on an annualized basis for each zip code in Florida, an engineering component to relate specific wind speeds to physical damage to re sidential structures typical of Florida homes, and a financial co mponent to relate physical damage to both content loss and total insurance dollar loss. S ubsequent chapters out line the strategies employed in the engineering component of calculating physical damage to typical residential buildings in Florida as function of a series of peak 3-second wind speeds. This model, like the HAZUS hurricane model, is component-based, explicitly accounting for resistance capacities of structural compone nts and wind loading within a probabilistic framework. The public model is not as comp lex as the one developed by ARA, foremost in that it does not time step through the entire life cycle of a hurricane. The public model does incorporate, to the extent possible, the current state of the art knowledge in wind pressures, windborne debris and resistance capac ities for typical residential buildings in the state of Florida.

PAGE 50

30 CHAPTER 3 RESIDENTIAL STRUCTURES IN FLORIDA Defining appropriate resident ial structural models for the state of Florida is a critical step in the development of a simulati on engine to predict structural damage in the state as a function of peak gust wind speeds. Wind loading characteristics are heavily dependent on the shape and component ma ke-up of the individual structure under consideration. Thus, the accuracy and reliab ility of the damage-prediction simulation engine is dependent on proper characterization of the build ing population in the state. Additionally, the efficiency of the simulation model relies on correctly identifying building components that are susceptible to wind damage. Finally, the resulting damage predictions will be useful only for statisti cally significant building types. Therefore, knowledge of the types of struct ures, the components of those structures most susceptible to wind damage, and the distribut ion of structural types throug hout the state is critical to the success of each step in the prediction of hurricane damage. Research partners in this joint projec t conducted an in-depth study of building classifications. This chapter summarizes thr ee contributions: statistical analysis of the residential building population of Florida conducted by Liang Zhang of the Florida Institute of Technology, with assistance from the author [31, 32]; manufactured housing research conducted by Luis Aponte of the University of Florida; and a building component investigation conducted by the author. Sources of information for characterizing residential structures in the state of Florida include the Florida Hurricane Catastrophe Fund (FHCF) expos ure database, databases of individual county property

PAGE 51

31 appraiser’s offices, manufactured home builder literature, and post-damage investigations. Sources of Information Florida Hurricane Catastrophe Fund Exposure Database The FHCF exposure database consists of in surance portfolio data for buildings in the state of Florida. At this time, data avai lable to the team of researchers working on the damage-prediction simulation engine consists of a statistical analysis of the FHCF database for single family residences (SFR ) only. Information concerning the population of manufactured homes by ISO clas sification is not available. Unfortunately for wind researchers, the ISO classifications used in insurance portfolios focus largely on fire hazards. This information alone does not provide an adequate structural characteriza tion of Florida residences, with respect to wind loading. It can be used (in combination with other s ources of information) to identify regional boundaries within the state. For example, the population of masonry homes vs. wood frame homes was found to be consistent among groups of counties in the same geographic area. The ISO constr uction classifications (described in greater detail in a master’s thesis written by a research partner [32]) are Frame Joisted Masonry Non-Combustible Masonry Non-Combustible Modified Fire Resistive Fire Resistive Heavy Timber Joisted Masonry Superior Non-Combustible Superior Masonry Non-Combustible Masonry Veneer Unknown

PAGE 52

32 County Property Appraiser Databases The most comprehensive sources of deta iled structural information currently available are the individual county property appraiser databases. Each county gathers residential and commercial pr operty data for tax purposes. Database architecture and contents (beyond those required by the Florid a Department of Revenue) vary, but each database can be separated into four gene ral categories: commercial property, SFRs, condominiums, and manufactured homes. Comm ercial property and condominiums are outside the scope of the current work, so the two categories of interest are SFR and manufactured homes. Nearly all of the SFRs in each county are listed in the county property appraiser’s database. A large number of manufactured home s are taxed through the Department of Motor Vehicles; however, an d are not listed in the property database. Processing database information from each of the 67 counties in Florida is not feasible for the current project; therefore a selection of counties spread throughout the state is used to obtain information about th e characteristics of typi cal Florida homes. The team was able to gather databases from several counties, but approximately half were unusable because files did not match the database layout provided by the property appraiser’s office. The nine counties that supp lied databases from which useful structural information was gained are Brevard County Broward County Escambia County Hillsborough County Leon County Monroe County Palm Beach County Pinellas County Walton County

PAGE 53

33 From the databases of the nine listed countie s, the type of roof, type of roof cover, exterior wall material, stories, square footage, and year bui lt are investigated for SFRs. This information is useful in identifying the most common residential structural types, but is incomplete as a characterization of ho mes, with respect to wind loads. Because the information is used for taxation, databa se categories often describe qualitative information (rather than structural details). For instance, exterior walls may be listed as ‘average’, without indicati ng the building material. Some database fields lump structurally significant details into a single category. Many counties, for example, use a single designation of ‘hip or gable roof’ inst ead of separating the two. This difference is structurally significant, as pos t-damage investigations have noted during past wind events [19]. Additionally, some structur ally significant information is not listed in the databases, such as the presence of a garage. In spite of these limitations, the databases supplied by the nine listed counties allowed the research team to develop models representative of typical Florida homes. Manufactured Home Builder Literature Information about manufactured homes c ould not be easily discerned from the individual county property appr aiser databases. Since many of these homes are taxed by the Department of Motor Vehicles, construc tion information from the tax authority is limited. Manufactured home information ha s been obtained from a report compiled by the National Association of Home Builders ( NAHB) Research Center for the Department of Housing and Urban Development compari ng site-built and manuf actured housing [33] and from contacts with the Partnership for Advancing Technology in Housing (personal correspondence by a research partner, June 2003) and Nobility Homes (personal correspondence by a research partner, June 2003).

PAGE 54

34 Post-Damage Investigations Literature searches of post-damage reports reveal that observa tions by experts in the field are useful in supporting the st atistical information on building population characteristics gained from other sources of data. However, the post-damage reports themselves usually do not contain enough bu ilding evaluations to be considered a statistically significant databa se from which to characteri ze Florida’s building population. The one exception is the NAHB Research Center report that describes the damage in South Florida after Hurricane A ndrew [19]. In this report, th e damage to residences is provided within a statistical framework. Unfo rtunately, this information is available for only one small geographic region following one storm. Though they are not a sour ce of statistical inform ation about the building population, post-damage reports are vital in determining which building components to model in a hurricane damage simulation engi ne. The expert opinions in post-damage reports indicate where severe wind damage o ccurs in typically constructed homes and, therefore, where the most benefit is to be gained from mitigation efforts. Results of the Building Population Investigation The information gained in researching the FHCF database, individual county property appraiser databases, manufactured home builder literature, and post-damage reports is detailed in this section. The disc ussion is divided into two sections: site-built home information is presented first, and manufactured home data follows. Characterization of Site-Built Homes The results gained from the nine indivi dual county property appraiser’s databases can be generalized to four regions of the state. The choice of regional boundaries is governed in part by the statistics of wood fram e houses in each county (an analysis of the

PAGE 55

35 FHCF database conducted by the meteorology team). Additional selection criteria included having at least two representative counties in each defined region and following the population density trends in South Florida. The resulting regions (defined as North, Central, South, and Florida Keys) are outlin ed on the county map of Florida shown in Figure 3-1. The shaded counties indicate the location of the nine from which property appraiser database information was obtaine d and successfully processed. A master’s thesis written by a research partner details th e process of determining the regional borders shown in Figure 3-1 [32]. North Keys South Central North Keys South Central Figure 3-1. Regional boundaries for building classification. Review of each processed county property appraiser database and the post-Andrew NAHB report [19] indicate th at the most common structures in the state can be summarized into four types, provided in Table 3-1. Table 3-2 shows the estimated percentage,p of each type per region and the m ean square footage of temperature controlled area, A, for each case. The areas provided for the Keys Region are marked with an asterisk due to their large standard deviation [32]. Because the average home size in the Keys is likely affected by a few gra nd estates, values from the South Region are used.

PAGE 56

36 Table 3-1. Four most common structural types Structural Type Characteristics CBG Concrete block gable roof one st ory home with shingles or tile CBH Concrete block hip roof one st ory home with shingles or tile WG Wood frame gable roof one stor y home with shi ngles or tile WH Wood frame hip roof one stor y home with shingles or tile Table 3-2. Population of mo st common structural types in defined geographic regions North Region Central Region South Region Florida Keys Structural Type p A (ft2) p A (ft2) p A (ft2) p A (ft2) CBG 12% 42%46%23% CBH 6% 1702 22% 2222 23% 2147 11% 3295* WG 39% 12%4%12% WH 20% 1908 6% 1941 2% 2022 6% 2771* Sum of most common 77% 82%75% 52% Unknown 14% 13%11% 23% Total coverage 91% 95%86% 75% Large standard deviation from observed data The third row from the bottom of Table 32 represents the percentage of the SFR population covered by the most common structur al types. Those not covered include two story homes, unusually constructed homes, and homes of unknown structural type. Unfortunately, the percentage of homes lis ted in available data as having an unknown structural type is significant in each region, as shown in the next to last row of Table 3-2. Since these homes cannot be classified, th e population represented by this category will be assigned an average value of structural wi nd damage obtained from an investigation of other structural types in that region. Further details con cerning this process are provided in Chapters 7 and 8, in which the structural damage results and conclusions are presented. The population of SFRs covered by the four most common structural types is adequate in the North, Central, and South Re gions, but the Keys Re gion has a significant number of homes not represented in Table 3-2. Additional structural types are listed in

PAGE 57

37 Table 3-3. The site-built home population represented by these additional groups is provided in Table 3-4, rounded to the near est whole percent. Using these additional categories, the portions of the building populatio n not counted in Table 3-2 are covered. Table 3-3. Additiona l structural types Structural Type Characteristics 2CBWG Concrete block 1st story, wood frame 2nd story, gable roof home with shingles or tile 2CBWH Concrete block 1st story, wood frame 2nd story, hip roof home with shingles or tile 2WG Wood frame two story gable roof home with shi ngles or tile 2WH Wood frame two story hip roof home with shingles or tile 2Keys Two story home of unsp ecified frame and roof cover CBGM Concrete block gable roof one story home with metal roof CBHM Concrete hip gable roof one story home with metal roof WGM Wood frame gable roof one story home with metal roof WHM Wood frame hip roof one story home with metal roof Table 3-4. Population of additional struct ural types in defined geographic regions Structural Type North Region p Central Region p South Region p Florida Keys p 2CBWG 1%2%8% 2CBWH 1%1%4% 2WG 5%1%1% 2WH 2%1%1% 2Keys 3% CBGM 8% CBHM 4% WGM 7% WHM 3% Sum of most common types (from Table 3-2) 77%82%75%52% Unknown 14%13%11%23% Total 100%100%100%100% For the Keys Region, a significant porti on of the population previously uncounted in Table 3-2 is listed in the categories with an ‘M.’ These match descriptions of the four most common structural types with the exception of the type of roof cover. For the North, Central, and South Regions, two story homes make up the difference. The population of

PAGE 58

38 individual types of two story homes shown in Table 3-4 is small in comparison to the overall population in these three larger regions Additionally, the enti re population of two story homes in the Keys represents only 3% of the population of this smaller region of the state. Given the contribution of two story homes relative to the overall SFR population, separate models are not developed for each two story type listed in Table 3-3. Instead, the performance of two story SFRs is predicted us ing the one story models in each region. In the North region, the WG and WH models ar e used as a framework for determining two story damages. Two story homes in the Cent ral and South Regions are based on the CBG and CBH models. A two story model for the Fl orida Keys uses information from each of the single story models. Plan dimensions are selected for each type of single story home such that the square footage remains close to the mean ar ea plus an unheated garage of approximately 400 ft2, while providing the largest number of wh ole sheathing panels on the roof surface. Unusually shaped sheathing panel cuts are avoided. The resulting site-built models for single story homes are describe d in Table 3-5, where the pl an dimensions represent the wall lengths. An overhang of two feet on each si de adds a total of four feet to both plan dimensions to give the size of the roof surface. Table 3-5. Structural type models for each geographic region North Region Central Region South Region and Florida Keys Region Structural Type Plan (ft) Area (ft2) Plan (ft) Area (ft2) Plan (ft) Area (ft2) CBG or CBH 56x38 212860x44264060x44 2640 WG or WH 60x38 228060x38228056x44 2464 The information presented in Tables 31 through 3-5 represents the bulk of information from the available property appraise r databases useful to the structural wind load characterization of SFRs. Unfortunately additional information critical in the

PAGE 59

39 determination of wind loading conditions is re quired, but generally not available, from this source. As a result, some structurally de scriptive characterizations must be made on a statewide basis, rather than regionally. One of these criti cal structural characteristics not obtained from the county property appraiser’s databases is the sl ope (or pitch) of the roof, a critical factor in the determination of wi nd loads on roof surfaces and in the sizing of roof components. A national distribution of t ypical roof pitch values is presented in Figure 3-2, where the numerator and denominator represent the number of inches of rise and run, respectively. The data for the figur e is taken directly from the NAHB Research Center’s 1998 report comp aring factory built and site-built housing. up to 4/12, 16% 5/12 and 6/12, 40% 7/12 and up, 43% flat up to 1/24, 1% Figure 3-2. Distribution of conventional (site -built) home roof pitch values according to the National Association of Home Builders Research Center. From the national information presented in Figure 3-2, and discussions with Dr. Leon Wetherington of the Univ ersity of Florida College of Building Construction (personal correspondence, September, 2002) a pi tch of 5 on 12 (5 inches of rise to the linear foot), correspond ing to a roof slope, of approximately 23, is selected as the most representative value for the population of site-built homes in Florida. This choice

PAGE 60

40 becomes an integral part of the wind load cr iteria for the structure. One section of the wind load provisions of the American Society of Civil Engineers re quires interpolation by roof slope, while the other divides structures into three categories: 10 30 10 and 30 [7]. Thus, a 5 on 12 pitch falls near the middle of the second category. The wide range of roof slopes cove red in this category certainly covers the majority of typical site-built homes. Given the sparseness of data with which to validate separate models, a single representative roof pitch is assigned to th e entire population of SFRs in lieu of using the statis tics in Figure 3-2 to determ ine what population of Florida homes should be modeled with sepa rate values of roof pitch. Characterization of Manufactured Homes The common structural types presented in Tables 3-1 through 3-5 represent the most prevalent site-built homes in Florida. A similar categorizati on cannot be made for manufactured homes, given the lack of in formation about these residences on regional basis. Instead, three types of manufactured home are used for the entire state. The two models representing typical modern manufact ured homes are referred to as MH 1 and MH 2 for singlewide and doublewide homes, respectively. Additionally, a separate model, MH-pre, is created to represent ol der manufactured homes that pre-date the changes in building requirements for these hom es enacted in 1975. All three are modeled with gable roofs, in accordance with NAHB Research Center fi ndings that 97% of manufactured home roofs in the Un ited States are gable type [33]. The national distribution of typical roof pitch values for manufactured homes, taken from the NAHB Research Center’s 1998 re port, is presented in Figure 3-3, where the numerator and denominator represent th e number of inches of rise and run,

PAGE 61

41 respectively. Using this national information, a pitch of 4 on 12, corresponding to a roof slope of approximately 18, is selected to be most repr esentative of the population of manufactured homes in Florida. For the same reasons discussed in the site-built homes section, the roof pitch is selected such that a representative value is applied to population of manufactured homes across the state. up to 4/12, 86% 5/12 and 6/12, 9% 7/12 and up, 3% flat up to 1/24, 2% Figure 3-3. Distribution of manufactured home roof pitch values according to the National Association of Home Builders Research Center. Building Component Selection It would be impractical and inefficien t to model every possible structural component in each of the building types identif ied in the previous section. Post-damage investigation reports are us ed to select building components common to all of the structural types that are sus ceptible to wind damage. In th is manner, all of the most commonly observed forms of damage are incor porated into the simulation model, and the results will be comparable acro ss residential classifications.

PAGE 62

42 In a 1993 report detailing Hu rricane Andrew damage, the three most critical home characteristics were the pr otection of openings (windows and doors), type of roof covering, and roof sheathing attachment [19]. Additiona l post-damage reports and investigations indicate that a reasonable list of wind damage-prone components for typically constructed gable roof residential structures includ es the roof covering, roof sheathing, roof-to-wall connections, wall sy stems, and openings [16-24]. Given this information, the structural building component s selected for modeling site-built homes in the simulation engine are (from top to bottom, not by order of importance) roof covering, roof sheathing, roof-to-wall connections, wa lls, and openings. Th ese broadly defined components are depicted in Figure 3-4. Each of the structural types in Table 3-5 are modeled based on the capacities of these co mponents. Differences among models of the various structural types come from the definitions of capacity load paths, failure mode, and wind loading. For example, concrete block and wood frame home models both include wall components, though the failure mechanisms and capacities of these systems differ. Also, wind loading differs from hip to gable roofs, though th e roof cover capacity is defined as the same. Openings Roof Sheathing Roof Cover Roof to Wall Connections Walls Openings Roof Sheathing Roof Cover Roof to Wall Connections Walls Figure 3-4. Structur al components selected for m odeling in the hurricane damageprediction simulation engine.

PAGE 63

43 The five components shown in Figure 3-4 are also used on the manufactured home model, with the addition of tie-down anchors. Further de tails concerning the building components (specifically the wind loads appl ied during simulated hurricane events and the resistance capacity of each component) are discussed in Chapters 4 and 5. The method by which the simulation engine uses this information to predict probabilistic damage information for each type of structure is detailed in Chapter 6. Validation of the methods using available data from Hurrica ne Andrew is presented in Chapter 7.

PAGE 64

44 CHAPTER 4 STRUCTURAL WIND LOADS FOR TYPICAL This chapter details the loads applied to simulate an extreme wind event on a typical residential structure. The load cases described here are used in the Monte Carlo simulation engine discussed in Chapter 6 to predict the vulnerability of typical Florida homes to structural damage. These loadi ng conditions are not in tended to represent design levels. Instead, load values are select ed to best represent the pressure or uplift acting on each component of the home during an extreme wind event, such as a hurricane or tropical storm. The preferred method woul d be to use wind tunnel data to accurately model the spatial and temporal characteristics of the pressure coeffici ent on the surface of the building as a function of the wind directi on. As discussed in Chapter 2, however, the current body of wind tunnel test data does not support the use of laboratory generated surface pressure characteristics on typical Florid a residences. This conclusion is not at all surprising. Wind tunnel tests have been conducted successfully over the years to determine the envelope of worst-case loads for appropriate wind load codification. The probabilistic character of surface pressures was not investigated to the level of detail necessary to randomly generate appropriately scaled and co rrelated surfac e pressures on all sides of a structure during a hurricane ev ent. Inclusion of non-Ga ussian behavior and correlation between surface pressu res in design loads is a promising topic that is currently being investigated [9, 10], and it might be possible to incorporate this data into the developed model at a later date. At this ti me, however, appropriate wind loads for each building component must be determined from the existing body of data, which includes

PAGE 65

45 current wind load design provisi ons, wind tunnel data, and full scale data sets described in Chapter 2. Engineering judgment, based on this supporti ng body of information, must be used to select the most appropriate external and in ternal pressures to use in the calculation of event-specific wind loads for building com ponents. For this reason, the wind loads selected for each building component are based on a modified version of the 1998 Minimum Design Loads for Buildings and Other Structures (ASCE 7-98) code provisions. Changes made to th e code provisions include modi fying the equations used to calculate surface pressures, re-mapping the pr essure coefficient zones on the roof surface as a function of the wind direction, and recalculating the internal pressure after initial damage has occurred. Details of these modi fications to the code provisions for the purpose of representing storm event loads are disc ussed in the first section of this chapter. Following that is a discussion of the applica tion of the modified code provisions on the structural components of typical Florida hom es. Load conditions placed on roof cover, roof sheathing, roof-to-wall connections, walls, openings, and tie-down anchors (on manufactured homes only) for the purpos e of simulating extreme wind events are identified. A summary table of the wind load conditions applied dur ing the structural damage simulation engine is provid ed at the end of the chapter. Use and Modification of the ASCE 798 Code Provisions to Represent Load Conditions during Extreme Wind Events Wind loads used for the prediction of stru ctural damage in the simulation model must represent surface pressures acting on each component during an extreme wind event. They should not match design pressures that envelope the worst-case scenarios, but should instead be dependent on the directi on of the wind, and representative of the

PAGE 66

46 pressure at a given moment in time. With this in mind, the load cases for the structural damage simulation engine are generated by re moving the conservativism incorporated in the ASCE 7-98 code provisions and by changing the external pressure coefficient zones such that the map of pressures on the roof surface is dependent on the wind direction. Modifications to Surface Pressure Equations Wind pressures on the surface of simulated homes are generated using modified versions of the ASCE 7 design wind pressure equations discussed previously in Chapter 2. Equations 2-11 and 2-12, for calculation of th e velocity pressure at mean roof height, qh, and the design pressure, p are reprinted here for clarity. The value 0.00256 in Equation 2-11 is a function of the density of air in English units, Kh is a terrain exposure coefficient, Kzt is a topographic effect factor to account for speed up over hills, Kd is a directionality factor, V is the design wind speed, and I is the importance factor. Equation 2-12 illustrates the calculation of p from qh where Cp and Cpi are the external and internal pressure coefficients, respectively, and G is the gust factor. I V K K K qd zt h h200256 0 (2-11) pi p hGC GC q p (2-12) Three of the four factors in Equation 211 are removed in the development of the equation for use in the simulation routine. The importance factor, I is discarded because it is used to scale loads according to th e importance of the structure to the local community. This factor plays a critical ro le in design, but does not assist in the determination of actual loads during a hurrican e event. Additionally, the directionality factor, Kd, is removed. This factor reduces design pressures to account for the fact that every section of the building will not be loaded to the design level at a given time. Since

PAGE 67

47 the directionality of the wind will be explicitly accounted for in the re-mapping of the external pressure coefficient discussed in the following section, the reduction factor Kd is removed. Lastly, few places in the state of Florida would warrant an escarpment factor greater than 1.0, therefore Kzt is unnecessary in the curr ent endeavor. The remaining factor, Kh, has a prescribed value of 0. 85 for low-rise structures (h 15 ft) in opencountry terrain (Exposure C). Substituting the value of Kh into Equation 2-11 and removing the three factors I Kd, and Kzt; the resulting equation used to calculate the velocity pressure in the simulation rou tine is provided in Equation 4-1, where V is the maximum 3-second gust wind associated with a particular storm or recurrence interval. 2) 85 0 ( 00256 0 V qh (4-1) The time scale of 3 seconds is selected to match the design wind speeds of ASCE 7-98. Use of a different time scale would n ecessitate additional modifications to the external pressure coefficients used in th e simulated wind load equations. It can be assumed that the maximum 3-second gust wind speed will occur several times over the period of the storm, since hurricanes generally last several hours. Therefore, damage can be assessed using this discrete value without undue concern fo r the length or cyclic nature of the load application. The safety factor embedded in the ASCE Component and Cladding (C&C) pressure coefficients on roof surfaces was determined by experimentation to be 1.25. This number was obtained from an unpublished study c onducted by the author to compare uplift values on a roof shape for which wind tunne l data was available, and through extensive discussions with Dr. Emil Simiu, an expert in the field, about th e codification of wind tunnel pressures and the available damage statistics from Hurrican e Andrew (personal

PAGE 68

48 communication, November 2001). Assuming that th e same level of risk is maintained in the design provisions for all building compone nts, a factor of 0.8 is added to the calculation of surfaces pressures represente d in Equation 2-12. In this manner, the reduction factor of 0.8 is used to remove the ‘safety factor’ embedded in the code provisions for load calculations. A similar pr ocedure described in Ch apter 5 is used to factor resistance values. Factors to increas e expected loads and decrease expected resistances are necessary in the design pro cess to account for the un certainty of each and reduce the risk of failure. Removal of these f actors is necessary such that ‘true’ loads during generated wind events can be compared to probabilistically ‘tru e’ capacities in the process of predicting of struct ural damage vulnerability. The application of the 0.8 factor to rem ove the built in safety value in the code provisions yields Equation 4-2. Together, Equa tions 4-1 and 4-2 are the basis for all wind load calculations used for structural dama ge prediction in the Florida Department of Financial Services sponsored Public Loss Hurricane Projection Model. pi p hGC GC q p ) 8 0 ( (4-2) Use and Modifications to External Pressure Coefficients External pressure coefficients in th e ASCE 7-98 provision s include both Main Wind Force Resistance System (MWFRS) coe fficients and Component and Cladding (C&C) coefficients. For brevity’s sake, a fu ll description of the design process is not provided in this document. Simply put, the MW FRS loads are applied to the structure as a unit (to provide checks for items such as diaphragm shear walls), while the C&C loads are applied to individual memb ers (for single unit capacity ch ecks). It is important to note, however, that both provisions (MWFRS an d C&C) must be satisfied in the design.

PAGE 69

49 The structural damage-prediction model uses a combination of th ese two provisions to best represent the load cases on modeled co mponents during hurricane winds. Table 4-5, at the end of the chapter, provides a summary table of th e load conditions (MWFRS or C&C) applied to each modeled component during the simulation routine. This section details the MWFRS and C&C external pressure coefficients taken from the ASCE 7-98 provisions and the modifications made for use in the damage-prediction model. Main Wind Force Resisting System ex ternal pressure coefficients The ASCE 7-98 MWFRS provisions are wi nd-direction dependent, and require no modification for use in Equation 4-2. Values for the external pressure coefficient, Cp, used in Equation 4-2 for MWFRS conditions ar e provided in Tables 4-1 and 4-2. To each value, a gust factor, G of 0.85 is applied to obtain GCp in Equation 4-2. The location of each pressure zone is provided in Figure 41, taken directly from ASCE 7-98. Values for Case A are interpolated for two roof pitches (5 on 12 for site-built homes and 4 on 12 for manufactures homes) from the values provi ded in ASCE 7-98. The characteristic dimension, a is the lesser of 10% of the smalle st horizontal dimension and 40% of the mean roof height, but no t less than 4% of the smallest horizontal dimension or 3 feet [7]. Table 4-1. Zones 1-6 MWFR S pressure coefficients MWFRS Pressure Zones (shown in Figure 4-1) 1 2 3 4 5 6 Case A for 5 on 12 0.538-0.456-0.467-0.414NA NA Case A for 4 on 12 0.516-0.690-0.469-0.415NA NA Case B (all roof pitches) -0.450-0.690-0.370-0.4500.400 -0.290 Table 4-2. Zones 1E-6E MW FRS pressure coefficients MWFRS Pressure Zones (shown in Figure 4-1) 1E 2E 3E 4E 5E 6E Case A for 5 on 12 0.771-0.722-0.648-0.598NA NA Case A for 4 on 12 0.780-1.070-0.673-0.609NA NA Case B (all roof pitches) -0.480-1.070-0.530-0.4800.610 -0.430

PAGE 70

50 Figure 4-1. MWFRS zones. A) Winds perp endicular to the ridge line through cornering winds. B) Cornering winds through winds parallel to the ridgeline. (ASCE 7-98 Standard, Minimum Design Loads for Buildings and Other Structures, American Society of Civil Engineers, New York, NY. Fig 6-4, p. 43) Component and Cladding external pressure coefficients Modifying the C&C pressure coefficients on the roof surface and walls to account for the directional nature of wind pressures is accomplished by manipulating the mapped zones to represent observed damage patte rns and wind tunnel pressure investigation results. The ASCE 7-98 pressure zones for the design of roof cladding on gable and hip roofs are shown in Figure 4-2. Zone 3 (the hi ghest magnitude of suction) is applied at each corner, and Zone 2 is applied to locations of discontinuity on the roof surface. Zone 1 (the lowest magnitude of su ction) is applied to all areas not covered by Zones 2 and 3. Figure 4-3 indicates the pressure zones for the wall surfaces. Zone 5 (the highest in magnitude) is applied to each corner and Z one 4 is applied to all other surfaces. The ASCE pressure zones provided in Figur es 4-2 and 4-3 envelope the worst-case scenarios for the life of the st ructure. Using these provisions, the designer is not required to determine which way the building will face relative to the most likely wind direction. Components in all corners are designed to th e same wind pressures. Structures will not experience pressures in this manner duri ng actual loading co nditions, however.

PAGE 71

51 Figure 4-2. C&C roof pressure zones. A) Gable roof zones. B) Roof slope diagram. C) Hip roof zones. (ASCE 7-98 St andard, Minimum Design Loads for Buildings and Other Structures, Ameri can Society of Civil Engineers, New York, NY. Fig 6-5B, p. 46) Figure 4-3. C&C wall pressure zones. (ASCE 7-98 Standard, Minimum Design Loads for Buildings and Other Structures, Am erican Society of Civil Engineers, New York, NY. Fig 6-5A, p. 44) Engineering judgment is required to manipul ate the map of desi gn pressures into a layout that is dependent on the wind direction. Modifi cations to the ASCE 7-98 Component and Cladding roof pressure zone s for varying angles of wind are shown in Figures 4-4 through 4-6. The characteri stic dimensions for zone width, a remains as described in ASCE 7-98, with the exception of the cornering wind case. The width of Zone 3 and the width of Zone 2 over much of the windward side are increased to 2 a for AB

PAGE 72

52 cornering winds on gable roof structures. Figur es 4-4 through 4-6 are not drawn to scale. Modifications to the wall pr essure zone layout (no figure) include removing the edge zone on the windward and leeward walls to apply a single uniform pressure across the face of the wall. The leading edge zone on th e side walls is maintained, and the trailing edge zone is removed. Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Figure 4-4. Roof pressure z ones for winds perpendicular to the ridgeline. A) Gable roof zones. B) Hip roof zones. Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Figure 4-5. Roof pressure z ones for winds parallel to the ridgeline. A) Gable roofs zones. B) Hip roof zones. Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Wind ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 ASCE 7-98 Zone 3 ASCE 7-98 Zone 2 ASCE 7-98 Zone 1 Wind Figure 4-6. Roof pressure zones for cornering winds. A) Gable roofs zones. B) Hip roof zones. B A AB AB

PAGE 73

53 In the ASCE design provisions, the gust fact or and external pressure coefficient for C&C loads are combined into one term, GCp, which is dependent on the effective wind area of the component being designed and, in the case of roof components, on the roof pitch as well. The effective wind area fo r components is defined by ASCE as the maximum of two possible values : the tributary area for the component in question, and the span length times an effective width of one-third of the span length. The effective wind area for fasteners is the worst-case tribut ary area for an individual fastener [7]. As the effective wind area decreases, the magnit ude of the external pressure coefficient increases, providing smaller areas with larger magnitude load cases and larger areas with smaller magnitude pressures. Since the entire region of a large tributary area is not likely to be loaded to maximum capacity at the same time, a uniform design load of smaller capacity is applied to the su rface of large areas. In the structural damage simulation program, efforts have already been taken to el iminate the conservatism or ‘safety factor’ built into the design code, and to map the pres sure coefficients such that the layout is dependent on the wind direction. Given this approach, and the reliance of most of the modeled components on fasteners (e.g. sheathing), the values taken from the ASCE 7-98 provisions for C&C external pressure coefficien ts are those with an effective wind area of 10 ft2 or less. Values for roof zone pressure coefficients are provided in Table 4-3, and values for wall surfaces are provided in Table 4-4. The modified location of roof pressure zones is given in Figures 4-4 through 4-6. Table 4-3. Roof zone C&C pressure coe fficient values for se lected roof pitches GCp Zone 1 -0.9 Zone 2 -2.1 Zone 3 -2.1

PAGE 74

54 Table 4-4. Wall C&C pressure coefficient values GCp Windward Wall 1.0 Side Wall Leading Edge (distance of a from the corner) -1.4 Side Wall -1.1 Leeward wall -0.8 Use and Modifications to Internal Pressure Coefficients Since extremely low barometric pressure s mark hurricane events, the internal pressure in modeled homes is assumed to be greater than the outside pressure before any damage occurs to the structure. With this rationale, the default value of internal pressure assigned to all homes in the structural damage simulation model is obtained by setting the internal pressure coefficient in Equation 4-2 equal to +0.18, the value provided in ASCE 7-98 for enclosed structures. As described in Chapter 6 of this document, initial failure checks are conducted to determine whether individual windows, doors, pieces of roof sheathing, or shear walls fail. A subsequent in ternal pressure, dependent on the level of initial damage to the home, is calculated as the weighted average of the pressure at the location of broken doors and windows, to include the garage door. This value is used in the final round of failure checks, descri bed in greater detail in Chapter 6. Application of the Modified ASCE 7-98 Code Provisions to Produce Extreme Wind Event Load Conditions on Selected Building Components The modified external and internal pressure coefficients discussed in the previous section are used with Equations 4-1 and 42 to generate the load conditions which simulate the occurrence of an extreme wind event on both site-built and manufactured Florida homes. In this section, the selection of modified extern al pressure coefficients for load conditions placed on roof cover, roof sheathing, roof-to-wall connections, walls, openings, and tie-down anchors ( on manufactured homes only) are specifically identified.

PAGE 75

55 Resistances to these wind loads are discussed in Chapter 5, and the order of application and failure checking conducted by the simula tion engine are detailed in Chapter 6. Roof Cover and Roof Sheathing Loads Roof covering materials and roof sheathi ng panels are treated as cladding during the structural damage simulation. The most likely sheathing panel arrangement for each of the models described in Table 3-3 and fo r both of the manufactur ed home models is obtained by starting with a full sh eathing panel on one of the lowest corners, and placing additional panels in an offs et pattern. Given the amount of uncertainty in roof cover loading and wind resistance, any efforts to define the area of an individual roof cover unit would not add to the accuracy of the dama ge prediction results. In light of this information, a section of roof cover is assi gned to each sheathing pa nel on the drawn roof sheathing arrangement. The individual sections of roof cover thus have the same square footage as the underlying sheathing panels. Th e resulting model-specific roof layouts are used to obtain aggregate external pressure coefficients for each individual piece of sheathing and roof cover at each wind angle using the pressure coefficient maps of Figures 4-4 through 4-6. Reasons for using the aggregate pr essure over point pressures are discussed in the section of Chapter 5 devot ed to the resistance capacity of roof cover and sheathing. Wind loads for each piece of roof sheat hing are obtained by using the aggregate external pressure from the model-specific la yout with the appropriate internal pressure coefficient for the state of the building in Equa tion 4-2. Since the roof cover is attached to the outside surface of the roof sheathing, it is not subject to the same internal pressure fluctuations. In order to best represent the load case that would occur during an actual storm event, the wind loads for roof cove r areas are obtained by using the aggregate

PAGE 76

56 external pressure from the model-specific roof layout and an internal pressure coefficient of zero in Equation 4-2. Roof-to-Wall Connection Loads Roof-to-wall connections are modeled in tension, using the dead load and windinduced uplift from the roofing system. As described further in Chapter 6, these connections are one of the last building components check ed for failure. The loads applied result from the remaining roof sh eathing. In this manner, overloaded roof sheathing panels are assumed to fail before passing the overloaded condition to the trusses. An assumed dead load of 10 psf (which includes the weight of typical roof cover, roof sheathing, suspended ceiling, insulation, and ductwork) is applied to each sheathing panel that remains on the roof surface afte r the initial failure check. Wind uplift is obtained from the loads previously describe d for the sheathing pa nels, and individual connection loads are calculated using a tribut ary area approach, assuming that trusses are spaced at 2 feet on center in most homes. Gable end trusses are assumed to have a total of eight gable end type connectors. Loads on the two end trusses for gable roof structures are equally distributed to these connections. Roof-to-wall connections are the only build ing component in the developed Public Loss Hurricane Projection Model where the redistribution of load is applied. Redistribution is not appropriate for other co mponents, but is used here to capture the failure mechanism by which the entire roof se parates from the walls [19]. Once a roof-towall connection fails, the load is redistribu ted to the surrounding connections until the system reaches a point of equilibrium. Additional details of this method are provided in Chapter 6.

PAGE 77

57 Wall Loads Walls on site-built homes are modeled in shear, uplift, and bending. The total shear for each wall is computed using the MWFR S pressure coefficients from ASCE 7-98 provided in Tables 4-1 and 42 and the standard practice of modeling the roof diaphragm as a simply supported beam. In this manner, the surface pressures on opposite sides of the house can be multiplied by half of the building he ight to produce a distributed load on the length of the roof diaphragm beam. Shear lo ads in each supporting wall are the reactions to this distributed load. This method is shown in Figure 4-7. During cornering winds, both cases are applied independently. V1 V2 V1 V2 Figure 4-7. Method of determining shear wall loads from MWFRS pressures. A) Winds perpendicular to the ridgeline through cornering winds. B) Cornering winds through winds parallel to the ridgeline. The uplift forces on each wall are obtained per foot of wall by averaging the total uplift from the attached roof-to-wall connect ions over the length of the wall. Lateral pressures for wall surfaces are obtained usi ng Equation 4-2 with C&C coefficients given in Table 4-4 and the appropriate internal pre ssure coefficient for the building. From these lateral pressures, the bending moments per foot of wall for concrete block walls are obtained with the assumption of simple supports at the roof and floor. This assumption is maintained unless more than half of the roof -to-wall connections fail, at which point the B A

PAGE 78

58 bending moment is amplified by a factor of 2.8. This factor (70% of the multiplier between simply supported and cantilevered mome nts) is selected for use over the pure cantilever condition since the wall would retain some support from the side and interior walls, even if the roof-to-wall connections have failed. Wood framed walls exhibit different behavi or when confronted with out of plane load conditions; therefore the bending moment is not calcula ted for these types of walls. Instead, the lateral force at the wall connection that result s from C&C surface pressures on the wall is used. In this procedure, the pres ence of at least one interior wall on each of the four perimeter walls is assumed. Under th is premise, the tributary area of pressure transferred directly into the la teral wall connections for each of the four perimeter walls is represented by the two trapezoid s shown in Figure 4-8. The tr ibutary area represented in Figure 4-8 relies on the assumption that the rest of the building is undamaged. This assumption is maintained unless more than half of the roof-to-wall connections on the depicted wall fail. After significant roof-towall connection damage, the tributary area is adjusted to the two triangles shown in Figure 4-9. h Figure 4-8. Tributary area fo r C&C pressures transferred into lateral connections on wood frame walls. Figure 4-9. Tributary area after significant roof-to-wa ll connection damage for C&C pressures transferred into latera l connections on wood frame walls.

PAGE 79

59 Surface pressures are calculated using Equa tion 4-2 with the C&C coefficients for walls given in Table 4-4 and the appropriate in ternal pressure coefficient for state of the building. The total load calculated by applyi ng these surface pressures to the tributary area shown in Figure 4-8 or Figure 4-9 for each of the four perimeter walls is distributed evenly to all of the lateral connec tions at the base of each wall. An additional wall load check for wood fr amed walls and the primary check for manufactured home walls is the potential loss of wall sheathing. Aggregate panel loads for individual pieces of wa ll sheathing are obtained in much the same way as roof sheathing loads. A length-specific layout is obtained for each wall. For wood frame walls, the layout is obtained by starting at one end with a full-size upri ght (4 ft wide by 8 ft tall) sheathing panel, and adding upright sheathing pa nels along the length of the wall, side by side. For manufactured homes, the layout is ob tained by stacking typically sized pieces of vinyl siding along the wall length. The surface pressures are calculated using Equation 42 with appropriate Component and Cladding external pressure coefficients, and the internal pressure coefficient for the particular building. Load Conditions for Openings This category covers a wide variety of bu ilding components. Incorporated into the simulation program are doors, garage door s, and windows. Each modeled house is assumed to have one front and one back door. The load applied to each is the surface pressure calculated using Equation 4-2 with the appropriate C&C pressure coefficient from Table 4-4, and the internal pressure coe fficient for the current state of the building. Additionally, houses with garage s are assumed to have the garage door on the front wall. The surface pressure applied to the garage door is the same as the pressure applied to the front door.

PAGE 80

60 Unprotected windows are loaded in two distinct ways: pr essure loads and impact loads. The pressure load scenario is sim ilar to that described for the doors. Surface pressures at the window locations are obt ained by using Equation 4-2 with the appropriate external C&C pressure coefficien t from Table 4-4, and the internal pressure coefficient for the building. Impact loads to windows are caused by wi ndborne debris from neighboring homes. To model this behavior, an equation base d on the cumulative exponential distribution (which describes the likelihood of rare and unrelated discrete events) is used to predict missile strikes. In Equation 4-3 below, ) (V pD is the probability of impact causing a broken opening, given the 3-second maximum gust, V A represents the fraction of potential missile objects (e.g., shingles) in the air. AN is the total number of available missile objects (e.g., number of shingles on the nearest house). B is the fraction of airborne missiles that hit the house, C is the fraction of the impact wall that is glass, and D is the probability that the impacting missiles have momentum above damage threshold. ] * * exp[ 1 ) (D C B N A V pA D (4-3) Equation 4-3 can be used to predict the lik elihood of impact fo r several scenarios. This equation can eventually be used to predict the likelihood of impact by several different sources of debris (e.g., shingles wood studs, and grapefruit). These varied results could be superimposed to determine the final tally of total impacts. With the information currently available, Equation 43 is used to determine the likelihood of windborne debris impact on the windows from any potential missile. Specific choices for each parameter and the methods by which thes e parameters could be honed in future work are discussed in the following paragraphs.

PAGE 81

61 Parameters that define objects in the air include AN and A The total number of available missile objects, AN, is related to the type and density of the building population around the modeled house. The curr ent selection for this number is an empirical choice of 100. This number is expected to change region ally in future iterations of the Public Loss Hurricane Projection Model, as the results fr om the initial model guide improvements in future work. A is related to the capacities of th e upwind building components that will become windborne debris, and is thus a func tion of peak wind speed. This parameter is modeled as a Gaussian cumulative density func tion (CDF). That is, at low peak gust wind speeds, relatively few of the available missiles are torn off upwind buildings to become windborne debris. As wind speed increases, mo re available debris is torn off upwind structures at a faster rate, until the functi on levels off at 1.0 (at which point 100% of potential missiles are in the air). In the current version of th e structural damage-prediction model, the Gaussian CDF defining A has a mean value at a peak 3-second gust of 135 mph, and a standard deviation of 15 mph. This function is shown in Figure 4-10. 50 100 150 200 250 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 3 Second Gust Wind Speed, VValue of A Figure 4-10. Values of the parameter A used in the determination of missile impact Parameter B in Equation 4-3 determines how many of the missiles in the air actually strike the modeled home. This pa rameter is dependent on several factors,

PAGE 82

62 including proximity of the missile starting poi nt and the ability of the missile to stay airborne (which is a function of wind sp eed and missile type). Engineering judgment indicates that missiles will fly further and stay in the air longe r with increasing wind speeds. For lack of better information, a linear function describing the parameter B is selected to have a value of zero (no airbor ne missiles striking the building) at 50 mph 3second gusts and a value of 0.40 at 250 mph 3-second peak gusts. Values of B are provided in Figure 4-11. 50 100 150 200 250 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 3 Second Gust Wind Speed, VValue of B Figure 4-11. Values of the parameter B used in the determination of missile impact Of the missiles striking the house, a fract ion will hit windows (rather than other surfaces). This value is described by the parameter C which defines the fraction of the windward wall space that is occupied by unpr otected glass windows. In the current structural damage-prediction model, the wi ndward wall space is the area of one of the perimeter walls except for the case of co rnering winds, when two of the walls are vulnerable to missile strike. As described later in Chap ter 5, windows on the modeled houses are categorized in four sizes. For accounting purposes inside the structural damage model, the four sizes of windows are treated independently at this point in the development of the missile impact equation. A value of C for each size of window is calculated as the area of that type of window divided by th e wall space vulnerable to

PAGE 83

63 impact. The probability of impact, ) (V pD, generated from using these values of C in Equation 4-3 is the likelihood th at a window of a certain size will be impacted and broken by a windborne debris missile, given the peak 3-second gust wind speed, V The parameter that determines whether th e striking missile w ill cause the window to break is D. This value is dependent on the moment um of the impacting missile and the resistance capacity of the window. Shingles a numerous and read ily available windborne missile type, are used to genera te a function for the parameter D The momentum of a windborne object, mp is defined by Equation 4-4, where m is the mass of the object, V is the wind speed, and R is a reduction factor. The value of ) ( R V is then the wind speed at which the missile object is travel ing. (Note that the subscript m for momentum is added by the author to the commonly used variable p to distinguish be tween momentum and pressure.) ) ( R mV pm (4-4) Conservatively assuming that a typical sh ingle weighs 0.06 lbs (a mass of 0.03 kg), and that the maximum value of R for single shaped missiles 0.64 [34], one can determine the momentum of a shingle moving in a wind gust of 110 mph. (49 m/s) to be 0.944 kgm/s. Given the impact resistance capacity of typical glass windows to be 0.025 kg-m/s [3], the momentum of a windborne shingle in a 110 mph 3-second gust wind event would exceed the capacity of typical unprotected window by a factor of approximately 37. It should be noted that additional reduction fact ors might apply, since the shingle might strike at an angle or not reach terminal velocity before hitting the window. However, these additional reduction factors will not overcome the significant difference between the missile’s momentum and the resistan ce capacity of a typical window. Because

PAGE 84

64 Equation 4-3 encompasses all type s of missiles, the shingle example is used to determine likely thresholds for the parameter D, and not specifically used to generate D as a function of wind speed. The values for D used in the current structural damage simulation program are taken from a Gaussian CDF gene rated using a mean value of 70 mph and a standard deviation of 10 mph. Using this f unction, the likelihood of breakage, given the fact that a missile has impacted th e window, is provided in Figure 4-12. 50 100 150 200 250 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 3 Second Gust Wind Speed, VValue of D Figure 4-12. Values of the parameter D used in the determination of missile impact Using the parameters described, the likelihood of an impact causing breakage during a specific wind event represented by a 3-second maximum gust is determined with Equation 4-3. Values are dependent on the size of window and the size of the windward wall. The eight possible angles of wind e xposure create three possible windward wall scenarios: short wall facing the wind, long wall facing the wind, and cornering winds, in which one short and one long wall are both vul nerable to missile impact. The function ) (V pD must be generated for each of the f our window sizes during each windward wall scenario, for a total of 12 functions pe r modeled building. As an example, ) (V pD for a medium sized window on the short side of th e concrete block, gable roof house in the Central Region of Florida is provided in Figure 4-13.

PAGE 85

65 50 100 150 200 250 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 3 Second Gust Wind Speed, VProbability of Missile Strike Causing Breakage Figure 4-13. Probability of missile strike causing breakag e of a medium (3.5 x 5 ft) window on a 44 ft long windward wall. Load Conditions for Tie-Down Anchors The load cases described in previous sections apply to both site-built and manufactured residences. Because of the di fferences in foundations, however, two loads cases are unique to manufactured housing. Th ese are sliding and overturning loads. Both cases are calculated using MWFRS pressure coefficients provided in Tables 4-1 and 4-2, and located in Figure 4-1. The overall lateral sliding force for a part icular manufactured home is calculated as the vector sum of the re sultant wall surface loads. This force will be resisted by the anchors as we ll as the friction between th e house and foundation piles. The overturning moment is calculated about th e leeward wall support pi er and is resisted by the assumed weight of the house as well as the anchor system. Discussion of the resistance to both of these load conditions is described in Chapter 5. Additional details on the overturning and sliding failure checks are provided in Chapter 6. Summary of Wind Load Conditions Used in the Simulation Engine A summary of the wind load conditions a pplied to individual components during simulation is provided in Tabl e 4-5. Sources described as MWFRS and C&C refer to the modified versions of the ASCE 7-98 provi sions for Main Wind Force Resistance System

PAGE 86

66 and Component and Cladding, respectively. Resistance values for each condition are described in Chapter 5 and the process by whic h the simulation engine applies and checks these conditions is detailed in Chapter 6. Table 4-5. Summary of load conditions applied to simulate extreme wind events Building Component Limit State Source of Loads Additional Notes Roof Cover Separation or pull off C&C Pressure coefficients aggregated over the area of the underlying sheathing panel; no internal pressure applied Roof Sheathing Separation or pull off C&C Pressure coefficients aggregated over the area of the individual panel Roof-to-Wall Connections Tension Roof sheathing Dead plus wind; load redistribution applied Shear wall MWFRS Concrete Block Combined uplift and bending C&C Uplift – Roof-to-Wall Connections Bending – C&C Shear wall MWFRS Uplift Roof-toWall Connections Lateral loading C&C Wood Frame Sheathing pull off C&C Pressure coefficients aggregated over the area of the individual panel Walls Manufactured Homes Sheathing pull off C&C Pressure coefficients aggregated over the area of the individual panel Doors and Garage Doors Overpressure C&C Overpressure C&C Openings Windows Impact damage ) (V pD Not an applied load; a probability of impact causing breakage as a function of wind speed Overturn MWFRS Manufactured housing Tie-Down Anchors Sliding MWFRS Manufactured housing

PAGE 87

67 CHAPTER 5 PROBABILISTIC WIND RESI STANCE CAPACITIES FOR RESIDENTIAL DWELLING COMPONENTS This chapter describes the resistance capaci ties selected for use in the structural damage simulation model. Capacities typi cal of the building components in Florida homes are selected from available literature and manufacturer data for each load case described in Chapter 4. Using this informa tion, truncated Gaussian distributions are created to represent pop ulations of typical bu ilding material resistances to the load cases identified in Table 4-5. These resistance dist ributions will be used in conjunction with the load values discussed in Chapter 4 to dete rmine whether individual structural members fail when subjected to extreme wind loadi ng. The operational flow of the simulation routine determining structural damage to typical Florida homes is provided in Chapter 6. Results and validation of the proce ss are discussed in Chapter 7. In this chapter, the details and selecti on process for the distribution of resistance values for each building component load cas e are provided. The first section of the chapter describes choices and arguments co mmon to the selection of all building component resistance values. Following this introductory disc ussion are sections detailing the selected capacities for roof cover, roof sheathi ng, roof-to-wall connections, walls, openings, and tie-down anchors. The chapte r is divided into a section detailing the resistance capacities of typical site-built hom es and a latter section providing information for manufactured homes. At the end of the chapter, Tables 5-3 and 5-4 provide a summary of all resistance values incorporated in the structur al damage simulation model.

PAGE 88

68 Fundamental Concepts Applied During th e Selection of Load Resistance Values Resistance values described in this chapte r represent the un-factored ability of each structural component to with stand loads induced by extreme wind events. As described in Chapter 4, the conservative factor built into the wind loading provisions of ASCE 7 was removed to determine a ‘true’ wind loading cond ition. In this chapter, the safety factors from manufacturer’s recomme ndations are removed to determine ‘true’ resistances. In this manner, the simulation program seeks to accurately assess the vulnerability of typically constructed homes to structural wind damage. If the safety factors were not removed, the program would provide the level of risk inherent in the current codification process, not the level of pot ential structural damage. Available literature and ma nufacturer’s data are used to determine appropriate probability density functions for component re sistances. Typically, the mean failure value for each component is obtained from availa ble information and the coefficient of variation ( COV ) is determined through engineering judgment. A measure of the spread of the distribution, the COV is the standard deviation divi ded by the mean. The effect of varying the COV is shown in Figure 5-1. Each pl ot in the figure shows a Gaussian (normal) distribution with a mean of 100 units The x-axis represents differing values of units, while the y-axis represents the like lihood of occurrence. The area under each curve is unity, though the peak valu e of the distribution with a COV of 0.2 is nearly twice that of the distribution with a mean of 0.4. The distribution with a COV of 0.2 is more closely centered on the mean value. Thus, there is a hi gher probability of selecting a value away from the mean (between 0 and 50, for example) using the distribution with a COV of 0.4, though both of the plotted di stributions have the same mean value of 100 units.

PAGE 89

69 Figure 5-1. Gaussian distribu tions with a mean of 100 units and varying coefficients of variation. Gaussian distributions similar to the ones depicted in Figure 5-1 are used to model populations of component capacities. Due to manufacturing quality control processes, individual components are like ly to have resistance capac ities which follow a lognormal distribution similar to the one shown in Figure 5-2. As compared to a Gaussian distribution with the same mean and COV the lognormal distribution provides a reduced likelihood of occurrence in the low resistance tail region and a slig htly greater likelihood of occurrence in the high resistance tail regi on. These characteristics are representative of manufacturing processes where the minimum allowable capacity is a quality control measure. Gaussian distributions are se lected over lognormal and other alternate distribution choices, however, to incorporate additional variables. The variation in type, quality, size, and installation for buildi ng components on homes of differing plan dimensions increases the variety of the popul ation under consideration. As the number of

PAGE 90

70 variables increases, the central limit theorem leads to the conclusion that the distribution which would best characterize each capacity is Gaussian. Figure 5-2. Lognormal vs. Gaussian for a mean of 100 units and coefficient of variation of 0.2 Chapter 6 describes the process by which th e Gaussian values are sampled and used to simulate individual homes, while the char acteristics of each dist ribution are described in the following sections of this chapter. As demonstrated by the distribution in Figure 51 with a COV of 0.4, this can lead to the possibility of selecting a value less than zero. To avoid the occurrence of physically impossible or impractical resistance values, truncation is applied to each of the capacity distribut ions described in the following sections. Sampled resistance values are bound within tw o standard deviations of the mean. The application of these upper and lower limits resu lts in a distribution similar to the example shown in Figure 5-3 for a mean of 100 units and a COV of 0.4.

PAGE 91

71 Figure 5-3. Truncated Gaussian distri bution with a mean of 100 units and a COV of 0.4. Site-Built Home Resistance Values Building components modeled for typical site -built homes consist of roof covering, roof sheathing, roof-to-wall connections, wa lls, and openings. Thes e are depicted in Figure 3-2. The following paragrap hs detail the resistance values obtained from available literature, manufacturer’s data, and engi neering judgment for each load condition described in Table 4-5 for site-built homes. The values provided in this section are used to characterize capacity distri butions similar to the exampl e provided in Figure 5-2. The sampling process by which the distributions ar e used to create representative Florida homes is discussed in Chapter 6. Wind Resistance Capacity of Roof Cover on Site-Built Homes The resistance capacity of the r oof covering is the ability of the shingles or tiles to stay attached to the roof sheathing, preventi ng rain water from entering and damaging the

PAGE 92

72 contents. In general, there is limited info rmation available about the uplift capacity of shingles and tiles, in spite of the fact that loss of roof covering contri butes significantly to insurance losses. One experimental study provi des an approach for estimating the wind action on shingles, but does not predict fail ure loads, citing the unknown capacity of the adhesive [35]. Factory Mutual (FM), Underwriters Labor atories (UL), and the American Society for Testing and Materials (ASTM) have de veloped test methods for commercial and residential roof coverings. Unfortunately, th e tests do not provide information about the ultimate failure capacity of these building mate rials, nor do they adequately represent the conditions these components would face in hur ricane events. Many use constant pressure systems instead of using a turbulent wind condition. During the standard FM test, a constant pressure is applied to the undersid e of a test specimen to simulate uplift [36]. Products that withstand the pressure for one minute without separating or delaminating are given a rating. FM Class 1-60 indicates a 60 psf test, while FM Class 1-180 indicates a 180 psf test. UL 580, “Standard for Tests for Uplift Resistance of Roof Assemblies,” and UL 1897, “Standard for Uplift Tests for R oof Covering Systems,” also use constant pressure systems to determine ratings [ 37]. These FM and UL static tests do not accurately simulate the wind action on the roof covering that will lead to shingle peeling and nail pull through. The ASTM standard testi ng protocol D3161, “Standa rd Test Method for WindResistance of Asphalt Shingles,” and the UL 997, “Standard for Wind Resistance of Prepared Roof Covering Materials,” specify a horizontal wind create d by a fan, but the required wind speed is only 60 mph, far belo w the design wind speeds for Florida [37]. A

PAGE 93

73 recent provision has been created in Dade C ounty, Florida which is similar to D3161 and UL 997, but uses a 110 mph fan instead of a 60 mph fan for asphalt shingle testing. Tiles and other roofing materials, however, are still tested fo r Dade County approval using static uplift tests [37]. While the Dade County provision for shi ngles does include a wind test using speeds in Category II of the Saffir-Simpson scale, the test is considered a ‘pass or fail’ event. That is, a product eith er qualifies for use in Dade County by passing the test, or does not qualify by failing the test. The provisions do not require determination of the actual failure capacity. Experimental data pred icting the adhesive fail ure or nail pullout of typical roof coverings (shingl es or tiles) of average age is not currently available, and could be the focus of a future research effort. In the absence of experimental data, th e capacity of typical residential roof coverings is estimated from the average of tw o calculations. The basis of the first logical argument is to infer that the majority of roof coverings were originally manufactured to the 1970’s era Southern Building Code Congres s International (SBCC I) requirement that cladding materials withstand an external posit ive or negative pre ssure of 25 psf. An additional necessity for this argument is to assume that, while improvements have been made, the fundamental manufacturing proce ss for shingles and tiles has not changed radically in the last few decades. Given thes e two assumptions, one can predict that 90% of the roof coverings currently on the mark et in Florida would meet or exceed the requirement of withstanding a 25 psf load under typical quality of workmanship in installation. Using a Gaussian distribution to represent the failure strength of all roof covering products used in the state of Florida, the standard distributi on tables can be used

PAGE 94

74 to determine the mean failure strength. E quation 5-1 provides the method of converting a value to the standard Gaussian distribution. In this equation, x is any value in the Gaussian distribution, is the mean of that distribution, COV is the coefficient of variation, and z is the value in a standard Gaussian distribution with the same likelihood of occurrence as the value x COV x z (5-1) The assumptions listed above are represented by setting x equal to 25 psf and obtaining a z value of -1.28 from the standard tables in Ang and Tang [38]. This z value represents a location at whic h 90% of the products would m eet or exceed the capacity. A COV of 0.4 is selected to represent the wide variety of products and quality of workmanship. With these values, Equation 51 can be rearranged to solve for a mean failure capacity. Using the argument pres ented above, 51 psf would be the most reasonable mean failure capacity for typical roof coverings. A second argument begins with the recen t Dade County uplift test for shingles, which uses a 110 mph fan. If the fan speed is used as the design wind speed, V in the ASCE design pressure equations (Equations 2-11 and 2-12) w ith the assumptions that the building is enclosed and in open terrain, a corresponding surface design pressure can be calculated. Interestingly enough, the value obtaine d is 51 psf, the mean failure capacity of typical roof coverings from the previous exer cise. Products are required to pass this test to be certified for use in S outh Florida, which means that the mean failure capacity of roof coverings is likely to be higher than the calculated value. Assuming that 90% of products pass the test, and again selecting a COV of 0.4 to represent the wide variety of products and quality of workmanship, the same procedure described in the first argument

PAGE 95

75 can be used to determine a possible Gaussian mean for South Florida shingles. A value of approximately 104 psf is obtained using this argument. Engineering judgment and knowledge of th e degree of damage following Hurricane Andrew and other past storms [15-23] indicate that the mean value from the second argument is too high, and the mean value from the first argument is too low, for use as a representative mean for all t ypical roof coverings in the st ate of Florida. A value that would best represent the entire population of roof coverings (t o include both shingles and tile products, as well as old and new construction) lies between the two values. Using this conclusion, a value of 70 psf with a COV of 0.4 is selected for the mean failure capacity of typical roof coverings. Wind Resistance Capacity of Roof Sheathing on Site-Built Homes A critical component in the overall vulnerability of a residence to hurricane damage is the ability of the sheathing to remain fasten ed to the trusses or rafters. A considerable body of research has been conducted in this ar ea in the wake of the Hurricane Andrew damage. One such study conducted at Clemson University indicates that the capacity of sheathing panels is best repr esented by treating the panel as a whole, rather than evaluating the capacity of individual fastener s. Sheathing panel failure usually begins with the pullout or pull-through of a single critical in terior fastener, but if any fastener is improperly installed, the failure mechanism is most likely to begin at that location, whether it is interior or not [39]. Given the difficulty in predicting the most probable failure location, the best means of comparing resistance to load is to use the aggregate load on the entire sheathing pa nel, and compare it with averag e failure loads from tests of whole panels, not just single fasteners.

PAGE 96

76 Results from the study discussed in the preceding paragraph and from additional studies provide mean failure pressures and coeffi cients of variation for panels of different wood species with different fastener sizes and schedules [39-41] Unfortunately, the means differ significantly, and considerable unc ertainty exists concerning the species of wood and the most typical fastener type and si ze used in each area of Florida. A simple arithmetic mean of the failure capacities woul d not necessarily best represent the building population of the state. As an alternative to using laboratory data, in situ data exists for a limited number of homes in South Carolina. These homes were flood damaged during Hurricane Floyd in 1999 and were subsequently purchased for th e purpose of testing and evaluation of retrofit measures. The homes varied in age and construction. Approxi mately half of the homes had planked roofs, one had oriented strand board (OSB), and the rest were plywood. After removing one outlier (a planke d roof with a high failure capacity of 450 psf) the sheathing failure test results average to value of approximately 150 psf, with the highest value at 196 psf and the lowest at 105 psf [42]. Though the hous es tested were in South Carolina, they are fairly representati ve of the types and ages of construction present in Florida. Since only a limite d number of homes were tested, the COV obtained from the eight test values is not used as the COV for a distribution representative of typical Florida roof sheathing. Instead, a value of 0.4 is selected to account for differences in workmanship and ma terials throughout the state. Wind Resistance Capacity of Roof-to-Wa ll Connections on Site-Built Homes The link between the roof system and the ex ternal walls occurs at the roof-to-wall connections. Uplift capacities of several types of roof-to-wall connections for light frame wood construction are available [43, 44]. The study conducted by Reed [44] further

PAGE 97

77 investigated the potential for load sharing be tween rafter connections and found that load sharing existed in nailed conn ections, but not in hurricane strap connections. No studies have been located that investigate the possibl e differences in uplift capacity for roof to masonry wall connections, though masonry struct ures make up a considerable portion of the building stock, as described in Chapter 3. In the absence of test data, the uplift capacity for typical roof-to-wall connections on masonry homes can be estimated from manufacturer information. In order to ma intain consistency between types of houses, manufacturer’s data is used for both masonry and wood frame homes. Personal correspondence between the aut hor and Randy Shackelford, a Simpson Strongtie representative, provi ded information about the co nnections most frequently used in the state of Florida and the typi cal factor of safety placed on the capacity specified by the manufacturer (personal correspondence, May 2002). Roof-to-wall connections manufactured by this company vary in uplift strength. Additionally, the same connector has a different strength rating depending on the type of wood used in construction. Generally, roof -to-wall connectors for both wood and masonry construction can be assigned to one of three strength categories: high strength (which includes most hurricane strap connections), medium, and lo w strength. Table 5-1 provides the values obtained by averaging the manuf acturer’s rated uplift capacity of the products available in each generalized category. Only two categor ies were obtained for the case of gable end connectors on masonry walls, and the valu es for wood construction are obtained assuming Spruce Pine Fir (SPF) construction. In Table 5-1, the term ‘side’ is used to describe typical roof-to-wall c onnections at the end of a tr uss. These connections occur on all four perimeter walls of a hip roof home but only on the side wa lls of a gable roof

PAGE 98

78 home. Gable end connections are those that connect the last truss on each end to the wall, and occur only on gable roof homes in the simulation model. Table 5-1. Manufacturer’s uplift capacity for typical roof-to-wall connections Connector Strength Category Construction Connector High (lb) Medium (lb) Low (lb) Side 1240690460 Wood Frame Gable 1260650380 Side 14001065700 Masonry Gable 640225 Discussions with a Bob Carter, a Br evard County architect, (personal correspondence, June 2003) in addition to th e correspondence with the Simpson Strongtie representative indicate that nearly all of th e homes constructed in Florida over the last 1520 years would fall in the category of highstrength roof-to-wall connections. Based on this information, the high-strength values in Table 5-1 are used as the manufacturer’s rated capacity for each type of home constr uction. Specific values for other types of connections (such as toe-naile d connections) are no t incorporated in the model at this time, due to a lack of information con cerning the distribution of connection types throughout the state. According the testing conducted by Simpson St rongtie, a factor of safety of 3 is applied to obtain a mean value of each connect or population. Using this value, a mean of 3720 lbs and 4200 lbs in uplift capacity per connector are obtained for the side connectors of wood and masonry homes, resp ectively. The mean values for gable end connectors are calculated to be 3780 lbs and 1920 lbs in uplift capacity, for wood and masonry homes, respectively. A COV of 0.2 is selected for all roof-to-wall fastener distributions, and bounds are placed such that acceptable values lie within two standard deviations of the mean. These distributions result in capacities highe r than those obtained

PAGE 99

79 through in situ and laboratory testing [ 42-44], but the damages predicted using the manufacturer’s values corre spond well with post-damage in formation surveyed after Hurricane Andrew. These results are provided in Chapter 7. To capture an observed failure mechanism where the entire roof detaches from the walls [19], the roof-to-wall connection strength for each simulated house is batch selected. A representative value is generate d for the entire house from the Gaussian distribution representing the population of roof-to-wall conn ections for that type of structure (wood or masonry). This value beco mes the mean of a Gaussian distribution having a COV of 0.05 from which individual c onnection capacities are randomly generated. This process represents obtain ing the connectors from the same manufacturer or batch, and using the same quality of instal lation for the home. Additional details on the batch selection process are provided in Chapte r 6. Roof-to-wall conne ctions are the only structural components to be se lected in this manner, spec ifically to incorporate the observed damage state of having the entire roof detach from the walls. The method of batch selection is not used for other structur al components because it results in predicted damages that are not observed in post-damage reports [19]. Wind Resistance Capacity of Site-Built Home Walls Wall failures are much less commonly cite d in post-damage reports than roofing system failures. In many cases, wall failures c ould be attributed to improper installation of connections or to the loss of structural integrity of the r oof system [19]. Capacities to resist shear, out-of-plane loading, and up lift are considered for wood frame walls and masonry walls in the following paragraphs. Resistance capacities for wood walls ar e obtained from the 1997 National Design Specification for Wood Construction (NDS) [45] as well as from laboratory tests. Wood

PAGE 100

80 capacities are distinctly difficult to generalize over a large population of homes because the load carrying capability of wood connections varies signif icantly with different types of lumber. To best represent the types of wood typically found in Florida, southern species of wood, such as Spruce Pine Fir (SPF) and Southern Yellow Pine (SYP), are used in resistance calculations. Damage to masonry walls was less prevalent than damage to wood frame walls, and masonry walls were less dependent on the in tegrity of the roof sy stem [19]. However, damage surveys [46] have shown that un-rein forced masonry might be a weak link in the structural system. After the failure of an openi ng, increased internal pressure can lead to the collapse of masonry walls, which trigge r the collapse of the whole structure. One study was obtained predicting the failure pressu re for simply reinforced and pre-stressed wall sections [47]; however, th is study alone is not enough information to adequately predict failure conditions for typical residential structure walls. In the absence of a significant population of la boratory test data, design provisions are used, with adjustments to allow for the best representation of true failure loads. Wood shear wall capacity Shear wall loads are transf erred through the wall sheat hing in wood frame walls. As a result, the capacity of the wall to resist shear wall loads is dependent on the nailing pattern and thickness of the attached plyw ood. Using 3/8 inch pl ywood sheathing with 8d nails spaced at 6 inches on center along sheat hing edges, the shear flow capacity of a typical wall is 310 lbs per linear foot, according to the NDS. A factor of safety of 3.5 is applied to this capacity, to account for both th e safety built into the design code, as well as the uncertainty in the contribution of other building materials. Wood homes are generally covered with some other form of cl adding, which contribute s to the ability of

PAGE 101

81 the wall to resist shear loading. The extent of the load resisting contribution of different materials (e.g. stucco) is difficult to predict. Additionally, the shear walls are tied into other pieces of the structure, su ch as interior walls. These load sharing contributions are not considered in the design process, so th ey must be accounted for when using design loads to predict the true failure capacity. The resulting Gaussian di stribution representing the failure capacity of wood frame shear wa lls has a mean of 1085 lbs per linear foot, with a COV of 0.2. Values are truncated at a di stance of two standard deviations away from the mean. Wood frame out-of-plane load capacity Out-of-plane loading applied to wood fr ame walls is transferred from the wall cladding into the studs, and then into the st ud wall connections. The w eak link in the load path occurs at the connections and not in the stud itself, under most circumstances [19]. Using the minimum nailing requirements from the Florida Building Code presented in Table 2306.1, this weak link occurs at the bo ttom end of the stud, wh ich is toe-nailed to the sill plate with four 8d nails [48]. Though usually not taken into account for design capacity calculations, the connect ion shares this lateral load with the sheathing nails that penetrate the sill plate. To best represen t the true failure capacity, both the toe nail connection and the contribution of some sheathing nails are taken into account. From the NDS, the equation for determining the lateral resistance ( Z ) of the nailed connection is presented in Equation 5-2, where CD is the load duration factor, CM is the wet service factor, Ct is a factor for temperature, Cd is a penetration factor, Ceg is a factor for end grain nailing, Ctn is a factor for toe nailing, N is the number of nails per connection, and z is the lateral capacity of an indivi dual nail in a partic ular species of wood.

PAGE 102

82 z N C C C C C C Ztn eg d t M D (5-2) To best represent Florida construction, the value of z is taken to be 78 lbs, which is the average of the values for SYP, Sout hern Douglas Fir (Sout h DFIR), SPF, and Southern SPF. CD for wind loads is 1.6, and CM is set at 0.85, assuming that rain water has potentially leaked into the walls. The factors Ct, Cd, and Ceg are each assigned a value of 1 because temperatures are not expected to be over 100F, and the conditions of limited nail penetration or e nd grain nailing do not apply. Since the connection is toenailed, Ctn is assigned a value of 0.83. Using these figures with an N of four, a design value of 352.2 lbs per connection is obtained. To this value, a factor of 3.5 is applied to account for the safety factor built into the NDS code. The Florida Building Code dictates th at the minimum nailing pattern for wall sheathing consists of 6d nails at 6 inches on center along edges and 12 inches on center at intermediate supports. For a typical 4 x 8 ft sheathing panel instal led vertically, this arrangement results in 9 nails along the bottom e dge, or 2.25 nails per linear foot of wall. Assuming a specific gravity of 0.45 for typical southern woods, 6d nails have a withdraw capacity of 21 lbs per inch of penetration [45]. Using a CD for wind loads of 1.6 and a penetration of 1.5 inches, the design contri bution for the end of the sheathing panel is found to be 113.4 lbs per linear foot. To this va lue, a factor of 3.5 is applied to account for the safety factor built into the wood design code. The total lateral resistance capacity of a typical wood frame wall is obtained by summing the toe nail and sheathing panel nail contributions. The fi rst value is found using a Gaussian distribution w ith a mean value of 1232 lbs, a COV of 0.25, and truncation at a distance of two standard devi ations away from the mean. The distribution

PAGE 103

83 representing the sheathing na il contribution is found usi ng a typical stud spacing of 24 inches on center. This distribution is al so Gaussian, with a mean of 794 lbs per connection location and a COV of 0.25. Each value is independently obtained from its respective distribution, and then the two are summed to represent the total lateral capacity at a point along a typical wood frame wall. Wood frame uplift capacity The capacity of a wood wall to resist uplift is modeled at the location of the wall to sill plate connection, the same connection wh ich is the weak link for typical wood wall lateral capacity. The to e-nailed configuration of this connection results in an uplift strength that is identical to the lateral (out -of-plane) capacity. Since the nails are toed in at a 45 degree angle, both uplift and out-of-pl ane loads result in a lateral load in the nailed connection. The additional strength pr ovided by the cladding and other attached building materials might vary slightly be tween the uplift and out-of-plane load conditions, but the difference is neglected for modeling purposes. The capacity of the wood wall connections in out-of-plane and uplif t are taken as identical values, though the load conditions are checked individually. Additional de tail on the failure check procedures is provided in Chapter 6. Wood frame sheathing capacity Plywood sheathing attached to wood frame walls behaves similarly to sheathing attached to the roof. The abil ity of plywood sheathing to rema in attached to the framing during wind load conditions is directly related to the type of wood, the type of fastener, and the fastening pattern. Unfortunately, statistics defining the most popular sheathing and fastening characteristics are not availa ble. Additional variables that cannot be adequately considered include architectural building materials that cover the sheathing

PAGE 104

84 and form the exterior wall covering. These ma terials contribute to the sheathing’s ability to remain fastened to the wall frame, but to an unknown extent. Engineering judgment dictates that wall sheathing is typical ly thinner and/or fastened with smaller nails than roof sh eathing. Comparing the baseline withdraw capacity of two typical fasteners from NDS provides a reasonable assumption about the difference in resistance of typical roof and wall sheathing. Using wood with a specific gravity of 0.45, the un-factored withdraw capacities for 8d and 6d nails are 25 lb and 21 lb per inch of penetration, respectively [45] The 6d capacity is approximately 84% of the value of the 8d capacity. Neglect ing the small difference in pe netration length that would result from thinner wall sheat hing, but incorporating the difference in fastener size, one can assume that the typical wall sheathing panel would have a pressure resistance capacity of roughly 84% of a typi cal roof sheathing panel. This factor is used to reduce the mean capacity of 150 psf, found for typical roof sheathing during in situ testing [42], to a mean value for typical wall sheathing of 126 psf. A COV of 0.4 is applied to account for a wide variety of wall coverings over the sheathing, and to account for the differences in workmanship and quality of construction. Masonry shear wall capacity The ability of typical masonry walls to re sist shear wall loads is obtained from the masonry design code. The maxi mum allowable shear stress, VF is defined by Equation 5-3, where mf is the capacity of the mortar [49]. psi f Fm V37 5 1 min3 4 (5-3) Equation 5-3 is not dimensionally correct, because mf is entered under the square root symbol in psi, and the result is obtained in psi. Using a typical value of 1500 psi (for

PAGE 105

85 mortar in residential construction), VF is calculated from Equati on 5-3 to be 49 psi. A factor of safety of 4 (slightly higher than th e value of 3.5 used for wood) is assumed to be built into the code values for masonry, therefore the calculated VF is multiplied by 4 to obtain a mean shear failure stress of 196 psi. A COV of 0.2 is assumed for the Gaussian distribution of shear stress capacity. Details of the comparison made between this capacity (psi) and the total shear load (lbs) are provided in Chapter 6. Masonry out-of-plane load capacity The behavior of masonry walls in out-of-p lane load conditions can be predicted by yield line theory and analysis of crack pa tterns [3]. This combined method requires knowledge of the aspect ratio and end support conditions of each section of wall. Given the uncertainty in predicting detailed aspect ratios, the computing resources necessary to employ a yield line theory method, and the uncertainty involved in predicting insurance losses based on wall damages; a simpler method is selected. The out-o f-plane capacity of a typical masonry wall is modeled on capaci ties obtained from ACI 530 for a one-foot mid-span section. The bending strength of a typical masonr y section is calculated as the section modulus, S, times the allowable tensile stress, Ft [49]. S is obtained from the geometry of typical masonry units, which are nominally 8 x 8 x 16 inches. Actual measurements are slightly lower than these values, and typica l widths for webs and flanges are 1 and 1 inches, respectively. Using these dimensions S for a one foot section of a typical concrete block wall without reinforcement is calculated to be 87 in3. The allowable tensile stress is provided by ACI 530 as 33.3 ps i, thus a typical one-foot section of a concrete block wall has the code based strength to resist a moment of 2897 in lb (241 ft

PAGE 106

86 lb). A factor of safety of 4 is assumed to be built into the mas onry code; therefore the mean capacity of a typical wall in bending is taken as 11,588 in lb. A COV of 0.2 is assumed to create a Gaussian distribution of bending strength. Details are provided in Chapter 6 for the combined failure check of masonry walls in bending and uplift, which incorporates this capacity of masonry walls in out-of-plane load conditions. Masonry uplift capacity The uplift capacity of typical masonry cons truction is obtained similarly to the outof-plane loading capacity. A one-foot mid-span section of concrete block wall is used to determine the strength of a typical wall in uplift, based on values from the ACI code. The value of Ft, the maximum tensile stress allowed by the ACI code, is 33.3 psi. Multiplying this value by a nominal area of 30 in2, a value of approximately 1000 lbs is obtained for the typical uplift design capacity. This value is multiplied by 4 (the factor of safety assumed to be built into the masonry code) to obtain a mean uplift capacity of 4000 lbs. A COV of 0.2 is assumed to create a Gaussian distribution of resist ance to uplift loads. Details are provided in Chapter 6 for the co mbined failure check of masonry walls in bending and uplift. Wind Resistance Capacity of Site-Built Home Openings Openings included in the damage predic tion simulation consist of doors, garage doors, and windows. Each of these is subjecte d to a component and cladding pressure, as described in Chapter 4. Site-built residences modeled in the structural damage simulation engine are assumed to have a wood or metal front door, a glass or mostly glass back door, and a total of 15 windows. The number of windows is obtained from a comparison study between site-built and manuf actured homes, where 15 was found to be the average number of windows for site-built construction [33]. Additionally, homes with garages are

PAGE 107

87 assumed to have a two-car sized garage door The capacity of each to resist pressure loads is described in the following paragraphs. Wind resistance capacity of doors for site-built homes Several types of doors with numerous lo cking mechanisms can be found in the building population of Florida. A statistical analysis of the failure capacity of the different types of doors with many different fastening and locking mechanisms would require resources beyond the scope of this project. In lieu of this information, mean failure capacities of 100 psf and 50 psf are selected for typical front and back doors, respectively. The choice for back doors is dist inctly lower than front doors to incorporate the likelihood of the back door being larger and consisting of unprotected glass. (French doors and sliding glass doors are popular in Florida.) The mean failure capacities are used in a Gaussian distribution with a COV of 0.2 to predict individual door strengths during simulation. Wind resistance capacity of garage doors for site-built homes The ability of typical garage doors to resi st wind pressure loads is obtained from a manufacturer’s trade group. The Door & Access Systems Manuf acturers Association (DASMA) provides testing provi sions for commercial and residential garage doors based on the 1997 Uniform Building Code. Individual tests for positive and negative design pressures include 1-minute duration design lo ad application and 10-second duration of 1.5 times the design pressure [50]. For one story double size (two-car) garage doors, 29.6 psf and -30.8 psf are the assigned design pres sures. Doors pass if they remain operable and recover at least 75% of their maximum deflecti on after the tests [50]. Assuming that 95% garage doors on the ma rket pass the DASMA test described above, a mean failure capacity for garage door s can be calculated using Equation 5-1. For

PAGE 108

88 the case of garage doors, x is given the value of 30 psf, and a z value of -1.645 is obtained from the standard tables in Ang and Tang [38]. This z value represents a location at which 95% of the products would meet or exceed the capacity. A COV of 0.2 is selected to represent the variety of products and quality of workmanship in the building population. With these values, Equation 5-1 is rearranged to solve for a mean 1-minute load capacity of 44.7 psf, and a 10-second load capacity of 67 psf. A corresponding strength to withstand 3-second gust winds w ould be slightly highe r than the 10-second value of 67 psf. This theoretical value would reflect the test criteria of operability and recovery of 75% of the maximum deflection, which does not necessarily indicate whether the door would be replaced as an insured lo ss. Additionally, a deflected door might allow enough wind into the garage to in crease the internal pressure of the house and contribute to the roof sheathing damage. Based on this information, a lower value of 52 psf is selected as the mean pressure at which a garage door would allow wind to penetrate the opening and at which the door would deflect such that it would be replaced under a typical insurance policy. A COV of 0.2 is applied to create a Gaussian distribution of strength. Wind resistance capacity of windows for site-built homes The ability of unprotected windows to resist pressure loads is dependent on the size and thickness of the glass panes. Assuming th at most typical windows are inch thick, the strength chart for annealed glass provi ded in ASTM E1300, “Standard Practice for Determining Load Resistance of Glass in Buildin gs,” is used to determine the strength of typically sized windows. The factor of safety built into the design values provided in the chart is known to be 2.5 (personal corre spondence with Dr. Jim McDonald, July 22, 2002), thus failure capacities are obtained by multiplying the chart value by 2.5. Mean

PAGE 109

89 failure capacities calculated for each of four selected window sizes are provided in Table 5-2. A Gaussian distribution is used for each case, with a COV of 0.2. Table 5-2. Mean failure pressu res for typical unprotected windows Description Size (ft x ft) Mean Failure Capacity (psf) Small 3.5 x 3.5104.4 Medium 3.5 x 5.069.6 Tall 3.5 x 6.552.2 Picture 6.5 x 6.537.2 As described in Chapter 4, load cases fo r windows include both pressure loads and impact loads. Determination of the likeli hood of a piece of windborne debris striking a window with enough momentum to cause the window to break is discussed in Chapter 4. The capacity of the window to resist impact is already factored into the debris model and is not repeated here. Manufactured Home Resistance Values The building components modeled for typical manufactured homes include the five components of site-built homes as well as tie-down anchors. Unfortunately, the term ‘manufactured home’ describes a great variet y of dwellings; a grow ing population that is not well defined in the state of Florida. Determining resistance values suitable for all types of manufactured homes re lies on engineering judgment. According to a 1998 study conducted by the National Association of Home Builders (NAHB) Research Center, the dema nd for manufactured homes more than doubled between 1991 and 1996, and includes homes that are increasingly similar to their site-built counterparts [33]. The study goes on to indica te that the median age of manufactured homes in 1995 was 15 years, compared to 30 years for site-built homes. At the time of the study, approximately 35% of the manufactured homes nationwide predate the 1975 Manufactured Home Construction and Safety Standards (also called the "HUD-

PAGE 110

90 Code," in reference to the Department of Housing and Urban Development). These factors indicate that the nationwide population of manuf actured homes is becoming increasingly more sophisticated than the stereotype of typical trailer parks might allow, though an older population of homes does still exist. As a summary, the NAHB report indicates that manufactures homes are typically made in similar fashion, but with slightly le sser quality or thinner members than site-built homes. With these findings in mind, the se lected capacities for typical manufactured home components are described in the following sections. A distinction is made between the capacity of typical manufactured ho me components and the components of manufactured homes that predate the 1975 HUD-Code. Values described in the following sections are used to character ize distributions of capacity si milar to the example provided in Figure 5-2. The method by which the distribut ions are used to create representative homes is discussed in Chapter 6. Wind Resistance Capacity of Roof Sheat hing and Cover on Manufactured Homes According to the NAHB report comparing manufactured and site -built housing, a surprising 93% of manufactur ed homes were constructed with oriented strand board (OSB) roof sheathing [33]. This wood product will behave in a similar fashion to the plywood typical of sheathing on site-built houses Additionally, roofs sheathed with OSB typically have a roof covering of asphalt shingles. This construction type is identical to that of site-built homes, with slightly different mean capacity values. To represent the selection of less expensive or thinner materials, a 0.9 reducti on factor is applied to the mean capacities of 70 psf and 150 psf for site-built home roof cover and sheathing, respectively. An additional reduction factor of 0.9 is used to represent the difference between current manufactured housing a nd that predating the 1975 HUD-Code. The

PAGE 111

91 resulting mean capacities for current and pre-HUD-Code manufactured home roof cover are 63 and 57 psf, respectively. Roof sheathi ng mean capacity values are 135 and 122 psf. COV values of 0.4 are selected for these dist ributions to represent the wide variety of available products and workmanship. Wind Resistance Capacity of Roof-to-Wa ll Connections for Manufactured Homes Capacity values for typical hardware us ed in the roof-towall connections of manufactured homes are obtaine d from a leading manufacture r’s website [51]. According to this information, the average manufact ured home code-approved value for typical single strap rafter ties is a pproximately 613 lbs per connector. The average strength of a weaker type of connection (the MMH8) is typically 365 lbs per connector. A stronger connection is achieved when a double strap conf iguration is used, resulting in a typical average value of 900 lbs per connection. Using the same factor of safety of 3 discussed previously for site-built home data, these mean capacities are multiplied to generate typical mean uplift failure loads. Under the assumption that pre-HUD Code homes use the weaker MMH8 type of connection, all manufactured homes built prior to 1975 are assigned roof-to-wall connection capacities based on the MMH8 value. With the safety factor of 3, the average failure load is 1095 lbs per connection. A wi de variety of connectors are assumed for typical modern singlewide homes, therefor e the roof-to-wall conn ection capacity for these homes is calculated as the average of typical single strap ra fter ties and typical MMH8 connections. The un-factored value us ed is 490 lbs per connection, which is multiplied by the manufacturer’s safety factor of 3 to obtain the mean failure pressure of 1470 lbs per connection. Modern doublewide homes experience larg er roof loads, therefore the capacity of the roof-to-wall connections for these homes are assigned based

PAGE 112

92 on average double strap connection values. With the safety factor applied, the mean capacity for typical doublewide hom es is 2700 lbs per connection. Roof-to-wall connection capaci ties for all types of manu factured homes are batch sampled, just like their site-built counter part s. For each home, a value is selected from a Gaussian distribution with a COV of 0.2. The mean of this distribution varies by type of manufactured home, as discu ssed in the previous paragraph. The sampled value becomes the mean capacity for an individually simulated home. The distribution of capacities for individual fasteners on the home is based on the sampled mean, with a COV of 0.05. Additional details concerning this pr ocess are provided in Chapter 6. Wall Capacity for Manufactured Homes Under the assumption that roof damage, ove rturning, or sliding failures resulting in significant insurance losses will occur before whole wall failures, the wall damage modeled for manufactured homes consists of siding failure only. The wind pressure capacity of typical vinyl sidi ng is obtained from manufactur er’s websites [52-54]. From the obtained manufacturer’s information a valu e of 44 psf is selected as the typical pressure resistance capacity of vinyl siding in the medium-priced category. To this value, a factor of safety of 1.5 is applied to obtain a mean failure pressure of 66 psf. The applied safety factor is lower than others used in capacity modeling due to the nature of the product. Vinyl siding is regarded as a non-structural element, in spite of the fact that a siding failure allows wind and water to penetrate the building envelope. For this reason, it is assumed that the manufacturing process woul d include a lower factor of safety than structural components. No safety factors we re obtained directly from manufacturers. Distributions for vinyl siding capacities are obtained for manufactured homes using a COV of 0.2. The mean capacity for modern homes is the value of 66 psf described

PAGE 113

93 above. For pre-HUD-Code homes, this mean va lue is reduced by a factor of 0.9 to account for aging and for the difference in products available a few decades ago. Wind Resistance Capacity of Manufactured Home Openings Openings modeled for typical manufactur ed homes include doors and windows. Just like their site-built count erparts, manufactured homes c ontain a wide variety of doors with differing locking mechanisms and windows of different sizes. A statistical analysis of the population of these openings is not feasible, so engine ering judgment is applied to determine the most likely arrangement and capac ity. Capacities for typical front and back doors are selected with the NAHB findings in mind. Specifically, manufactured homes are far less likely to have glass doors, and th e capacity of typical non-glass doors is likely to be lower than those used in site-built hom es [33]. Given these two factors, the front and back doors on each manufactured home model are assigned a capacity based on a Gaussian distribution with a mean value of 80 psf and a COV of 0.2. One notable exception to the typical differences between site-built and manufactured homes is the cap acity of windows. A pane thic kness of inch is selected for both site-built and manufactur ed homes. Thus, the capacity to resist pressure loading is the same for both types of construction. Mean capacities for typically sized windows are presented in Table 5-2. Like site-built homes, manufactured homes are also subject to windborne debris. The likelihood of a piece of debris impacting and breaking a typical window is discussed in Chapter 4. This argument applies to both site-built and manufactured homes, and is not repeated in this section. Wind Resistance Capacity of Tie-Down Anchors Tie-down anchors are used to resist both sliding and overturni ng of manufactured homes. The systems generally consist of an ear th screw attached to the underside of the

PAGE 114

94 home and anchored in the soil. Unfortuna tely, the resources necessary to conduct a thorough study of the population of manufactur ed homes in the State of Florida which would reveal the various types of anchors, in stallation methods, and the variation in soil capacities are beyond the scope of the current e ffort. A general arrangement of anchors is assumed for all of the manufactured homes in Florida, based on work conducted by Marshall and Yokel [55, 56]. Two lines of anchors are assumed, 7 feet apart, as shown in Figure 5-3 (a sketched side view of a typica l manufactured home). Each line consists of five anchors installed at a 45 degree angle, for a total of ten anchors per home. 7 ft Figure 5-4. Typical arrangement of ti e-down anchors for manufactured homes. Tie-down anchors are characterized by a Ga ussian distribution of pull-out capacity with a mean value of 1550 lbs and a COV of 0.4. The mean value is obtained from work conducted by Yokel, but a value of 0.4 is substituted for the reported COV of 0.3 [56]. The increase of COV reflects an additional uncertainty from installation techniques and soil quality not observed duri ng the test. The mean value of pull-out capacity is confirmed for use in typical Florida soils by comparison with a limited test conducted by Hayes for 8-inch helix screws in sand [57]. Summary of Resistance Values Used in Structural Damage Simulation Capacity values described in the preceding sections for site-built and manufactured homes are summarized in Tabl es 5-3 and 5-4. Table 5-3 provides a description of the

PAGE 115

95 limit state, and capacity distri bution characteristics for sele cted building components of site-built homes, while Table 5-4 provides similar informati on for manufactured homes. Values provided in these tables are used to simulate individual homes representative of typical Florida structures. The resistance of these simulated structures is compared to wind loads described in Table 4-5 to determin e if structural failures occur during highwind events. Details of the simulation process are provided in Chapter 6. Table 5-3. Site-built home summary of wind resistance capacities Building Component Limit State Mean Capacity COV Additional Notes Roof Cover Separation or pull off 70 psf0.4 Roof Sheathing Separation or pull off 150 psf0.4 Concrete Block Tensile failure 4200 lbs (side) 1920 lbs (gable) 0.2 Batch selected Roof-to-Wall Connections Wood Tensile failure 3720 lbs (side) 3780 lbs (gable) 0.2 Batch selected Shear wall failure 196 psi0.2 Concrete Block Combined uplift and bending failure 4,000 lbs (uplift) 11,588 in lb (bending) 0.2 0.2 Capacities separate, failure check is combined. Shear wall failure 1085 lb/ft0.2 Lateral failure 1232 lb (connection) 794 lb (additional) 0.25 0.25 Contributions summed for total capacity Uplift Failure 616 lb/ft (connections) 397 lb/ft (additional) 0.25 0.25 Same as lateral Walls Wood Sheathing failure 126 psf0.4 Doors Pressure failure 100 psf0.2 Garage Doors Pressure failure 52 psf0.2 Small104.4 psf0.2 Medium69.6 psf0.2 Tall52.2 psf0.2 Openings Windows Pressure Failure Picture37.2 psf0.2

PAGE 116

96 Table 5-4. Manufactured home summary of wind resistance capacities Building Component Limit State Mean Capacity (pre HUD-Code) COV Additional Notes Roof Cover Separation or pull off 63 psf (57 psf) 0.4 Roof Sheathing Separation or pull off 135 psf (122 psf) 0.4 Roof-to-Wall Connections Tensile failure 2700 lbs (double) 1470 lbs (single) (1095 lbs) 0.2 Batch selected Walls Siding failure 66 psf (59 psf) 0.2 Doors Pressure failure 80 psf0.2 Small 104.4 psf0.2 Medium 69.6 psf0.2 Tall 52.2 psf0.2 Openings Windows Pressure Failure Picture 37.2 psf0.2 Tie Down Anchors Pull out 1550 lbs0.3

PAGE 117

97 CHAPTER 6 SIMULATION ENGINE This chapter details the probability-based system-response model developed for the Florida Department of Financial Services sponsored Public Loss Hurricane Projection Model described in Chapter 2. The deve loped structural damage model is a MatLAB based Monte Carlo Simulation (MCS) engine that uses the structural wind loads discussed in Chapter 4 and th e building component resistance values described in Chapter 5 to simulate the performance and interaction of structural component s of typical Florida homes during hurricane winds. The model is ba sed on a series of three nested loops: an outer loop for angles of inci dence, an intermediate loop for maximum 3-second gust wind speeds, and an inner loop to simulate i ndividual buildings of a structural type. A flowchart of the developed model is shown in Figure 6-1, where shading identifies tasks within the nested loops. Each of the flowchart tasks listed in Fi gure 6-1 is detailed in this chapter. Structural damage results are obtai ned for buildings repres entative of typical Florida homes using the procedures defined by the flowchart. These building types are defined in Chapter 3, and the structural da mage validation and results are presented in Chapter 7. Selection of Structural Type and Definition of Geometry The MCS engine begins by initializing vari ables common to all structural types as well as variables unique to the particular t ype under consideration (e.g. concrete block gable roof home in the Central Region). Current selections for variables of both site-built and manufactured homes are described in the following sections. Values are selected to

PAGE 118

98 best represent the most common structural types of Florida homes, as described in Chapter 3. While the current values are hard wired into the MCS engine, future uses of the program architecture such as an online learning laboratory c ould incorporate user input to change building parameters. Define Angle of Incidence Define Wind Speed Save Damage FileRepeat until all speeds complete, then go to next angle Repeat until all angles complete Simulate an Individual Home: Randomize Wind Speed, Cp’s Sample Resistances Initial Failure Check: Sheathing, Openings, Walls Calculate Internal Pressure from Opening Failures and Recalculate Structural Loads Final Failure Check: Openings, Sheathing Roof Cover Connections, Walls Selection of Structural Type, Definition of Geometry Loop for Angle Loop for Wind Speed Loop for Building Write 1 Row of Damage Array Repeat until all simulations complete Calculate Initial Loads Define Angle of Incidence Define Wind Speed Save Damage FileRepeat until all speeds complete, then go to next angle Repeat until all angles complete Simulate an Individual Home: Randomize Wind Speed, Cp’s Sample Resistances Initial Failure Check: Sheathing, Openings, Walls Calculate Internal Pressure from Opening Failures and Recalculate Structural Loads Final Failure Check: Openings, Sheathing Roof Cover Connections, Walls Selection of Structural Type, Definition of Geometry Loop for Angle Loop for Wind Speed Loop for Building Write 1 Row of Damage Array Repeat until all simulations complete Calculate Initial Loads Figure 6-1. Structural damage simulation engine flowchart Variables for Site-Built Homes Variables common to all simulated site-bui lt homes include a wall height of 10 ft, an eave overhang of 2 ft, a truss spacing of 2 ft on center, and a roof pitch of 5 on 12. The openings on site-built homes are distributed such that three medium-sized windows, a door, and a two-car garage door occupy the fr ont wall. The front windows are assumed to

PAGE 119

99 be on the interior section of the wall, not in the higher-pressure edge zone that occurs when a neighboring wall is the windward wa ll. A glass door and four medium-sized windows occupy the back wall. Of the four windows on the back wall, two are situated such that they lie in the edge zone. The tw o side walls are identical, with four small windows each. Two of these four windows on each side wall lie in the edge zone. Dimensions for each of the single story site-built structural types described in Chapter 3 are provided in Table 6-1, to incl ude the pressure zone width, a. Sheathing patterns on the roof surface, numbers of roof-to-wall connections, and wood wall sheathing patterns are determin ed from these dimensions The designations CB and W refer to concrete block and wood, respectively. G and H are used to denote gable or hip roof types. The dimensions provided for the North and Central Region wood frame homes are the same, since the average values obtained from county property databases were found to be nearly identical. This is al so true for the concrete block homes in the Central Region and the combined South and Ke ys Region. The regional designations for each of these homes are maintained throughout this document for clarity, in spite of the fact that simulations using the data currently available will produce identical results. (As knowledge is gained concerning the regional characteristics of home construction, the resistance values of specific building mate rials, or the interac tion between hurricane winds and low-rise structure, these regionall y defined models are likely to change.) Additional details about the vari ables and matrices used in the MCS engine are provided in the sections describing resistance capacity sampling. Typical window sizes are provided in Table 5-2.

PAGE 120

100 Values are not provided in Table 6-1 for tw o story homes. As discussed in Chapter 3, these homes make up a small percentage of the population. Structural damages for two story homes will be based on the results of the single story homes. Two story homes in the North region are based on the performance of North WG and WH models due to the prevalence of wood construction in that regi on. In the Central and South Regions, the CBG and CBH models are used as a framew ork for determining two story damages. Lastly, two story homes in the Florida Keys ar e based on all four singl e story types in that region. These homes are not described in detail in this chapter. Met hods used to predict structural damage for two story hom es are presented in Chapter 7. Table 6-1. Site-built home dimensions Structural Types Length (ft) Width (ft) a (ft) North Region CBG or CBH 56 38 3.8 North Region WG or WH 6038 3.8 Central Region CBG or CBH 60 44 4.4 Central Region WG or WH 6038 3.8 South and Keys Region CBG or CBH 60 44 4.4 South and Keys Region WG or WH 56 44 4.4 Variables for Manufactured Homes Variables common to all simulated manufactur ed homes include a length of 56 ft, a wall height of 8 ft, a crawl space under the bu ilding of 3 ft, an overhang of 1 ft, a truss spacing of 2 ft on center, and a roof pitch of 4 on 12. The front and back walls of each simulated manufactured home have a door and three windows. On the front wall, two of the windows are medium-sized and one is la rge. The back wall has all medium-sized windows. Side walls have small windows; one each for singlewide homes and two each for doublewide homes. Dimensions of the ma nufactured home models are provided in Table 6-2, where MH 1 and MH 2 refer to single and doublewide homes, respectively. MH-pre is the model which has the same shape as the MH 1 model, but different

PAGE 121

101 component strengths, representative of th e manufactured homes that pre-date the manufactured home building code changes enacted in 1975. The dimensions shown in Table 6-2 are used to determine the roof sheathing pattern, wall siding pattern, and number of roof-to-wall conn ections. Window sizes are desc ribed in Table 5-2, and additional details about the variables and matr ices used in the MCS engine are provided in the section describing resistance capacity sampling. Table 6-2. Manufactur ed home dimensions Structural Type Length (ft) Width (ft) a (ft) MH 1 56 133 MH 2 56263 MH-pre 56 133 Loop for Angle of Incidence After variables have been defined, the simulation engine enters a series of embedded loops for wind direction and speed. Ei ght angles of incidence (depicted on the plan view of a typical hip roof building in Figure 6-2) are run for each wind speed during the damage simulation routine. In this manner, the orientation of the building relative to the wind direction is uniformly distributed while the 3-second gust wind loads are directionally based as in dicated in Chapter 4. This is an important distin ction, since nondirectional loading provides a different re sult than uniformly di stributed orientation combined with directional loading. The curr ent approach uses a uniform distribution of wind angles due to a lack of statistical information on orientation with respect to wind direction during hurricane landfall. Future effo rts could weight the angles presented in Figure 6-2; accounting for the most likely neighborhood layouts and for the most likely direction of approach for hurricane winds w ithin particular areas of the state by using different weighting values be tween regions or zip codes.

PAGE 122

102 Front 0 180 270 90 225 315 45 135 Front Front 0 180 270 90 225 315 45 135 Figure 6-2. Angles of wind inci dence used for each wind speed Loop for Wind Speed In addition to the eight possible wind appr oach angles, the MCS engine can be run for any number of wind speeds. The current choice of 3-second gust wind speeds ranging from 50 mph to 250 mph in increments of 5 mph reflect the requirements of the Florida Department of Financial Services and the i nput of the meteorology team for the Public Loss Hurricane Projection Model. These disc rete values define the storm intensity. Loop for the Simulated Homes For each combination of angle and wind sp eed, the MCS engine simulates a user defined number of realizations of the syst em, which consists of the wind loads and resistances for the particular structural type being investigated. Each realization is created by randomizing the discrete value of the 3-second gust wind speed and pressure coefficients (Cps) defined in Chapter 4, and by sampling from the distributions of component resistances described in Chapter 5. A sequence of failure checks is then used to determine the structural damage for each building simulated at a particular wind speed and angle of incidence. These steps are show n in Figure 6-3, and are described in the following paragraphs as they would occur for a single realization of the structural type (e.g. one of thousands of Central CBG home s). Figure 6-3 represents the building loop

PAGE 123

103 subset of Figure 6-1 (which describes the enti re simulation process). Before this loop is initiated, the 3-second gust speed defining st orm intensity and the a ngle of incidence are already selected. Save Damage File Simulate an Individual Home: Randomize Wind Speed, Cp’s Sample Resistances Initial Failure Check: Sheathing, Openings, Walls Calculate Internal Pressure from Opening Failures and Recalculate Structural Loads Final Failure Check: Openings, Sheathing Roof Cover Connections, Walls Loop for Building Write 1 Row of Damage Array Repeat until all simulations complete Calculate Initial Loads Figure 6-3. Flowchart for realiz ations of a structural type Randomization of Wind Speed and Pressure Coefficients The discrete value of the 3-second wind sp eed defined in the intermediate loop of Figure 6-1 represents the intensity of the storm to which the simulated buildings are subjected. Selected values are defined be tween 50 and 250 mph in relatively narrow increments of 5 mph. These choices meet th e requirements of the Florida Department of Financial Services commission, and allow for th e relation of structural damage results to the meteorology-team-predicted likelihood of maximum 3-second gust wind speeds in different zip codes throughout the state of Fl orida, on an annualized basis. A single pass within the “Loop for Wind Speed” depicted in Figure 6-1 represents exposure of buildings to a storm with ma ximum sustained 3-second gusts within a zone of intensity.

PAGE 124

104 For each of these defined storm exposure categor ies, there exists a degree of uncertainty concerning the exact value of the maximum sustained 3-second gust wind speed observed at the location of the simulate d building. Additionally, there is uncertainty concerning the exact value of the pressure coefficients described in Ch apter 4. Surrounding obstacles may shelter individual houses, or homes may lie in areas prone to slightly higher than average winds. For these reasons, the discrete value of the wind speed and the discrete values of the pressure coefficients are randomized for each simulated home. Randomization of the wind speed and pressure coefficient values is achieved in the MatLAB based code with the function randn(). This command generates a group of numbers randomly sampled from the standard normal distribution, a Gaussian PDF with a mean of zero and a standard deviation of one. The randomly generated numbers are then individually scaled using Equation 6-1, where z is a randomly ge nerated value from the standard normal distribution, is the mean value of desired PDF, and COV is the coefficient of variation of the desired PDF. The resulting x is a value in the desired PDF with the same likeli hood of occurrence as z 1 COV z x (6-1) The COV for wind speed and pressure coefficien t variation is selected to be 0.1 (10% of the mean value). Values of substituted into Equation 6-1 to generate randomized pressure coefficients are the discre te values presented in Tables 4-1 through 4-4, as well as the internal pressure coeffici ent of +0.18. In this manner, the MCS engine samples the pressure coefficients using E quation 6-1 to generate a randomly selected value from a PDF with a mean of the discre te pressure coefficient provided by ASCE 798 and a COV of 0.1.

PAGE 125

105 Values of substituted into Equation 6-1 fo r the generation of the randomized wind speed are the discrete values of the 3-second maximum gust speed, from 50 to 250 mph in 5 mph increments. The same mean va lue is used within a single “Loop for Wind Speed,” generating slightly different maximum 3-second gusts for each building simulated within that category of storm in tensity. The randomly selected 3-second gust wind speed and the randomized pressure coeffi cient values are maintained for the life of an individual building simulation (the sequen ce of steps shown in Figure 6-3). As each new building is simulated, these values are sampled to account for the uncertainty in wind speed and pressure coefficient intensities on the individual building. In this manner, each of the simulated buildings of a structur al type represents a home at a slightly different location within a neighborhood e xposed to the defined storm intensity. Initial Load Calculations Once the wind speed and pressure coeffici ent values for the individual building simulation are determined using Equation 6-1, the initial wind loads are calculated. During this step, the velocity pressure is obtained using Equation 4-1. The surface pressures described in Chapter 4 are calcula ted using the randomly generated pressure coefficients and wind speed described in th e preceding section in Equation 4-2. From these calculated pressures, the structural loads described in Table 4-5 are obtained. Sampling of Resistances The task of sampling resistance values from the distributions described in Chapter 5 is the key that makes the MCS engine a probability-based system-response model. During this step, each indi vidual piece of a typical co mponent is assigned a unique capacity value. In this manner, each of the simulated houses represents one possible realization of the population of typically constructed Florida bu ilding types. Details of the

PAGE 126

106 capacity assigning process are described for each component in the following sections. For clarity, Figure 3-4 (illust rating the components of typi cal site-built homes) is reprinted here as Figure 64. Manufactured homes are modeled with the five components illustrated in Figure 6-4 and a sixt h component, tie-down anchors. Openings Roof Sheathing Roof Cover Roof to Wall Connections Walls Openings Roof Sheathing Roof Cover Roof to Wall Connections Walls Figure 6-4. Modeled st ructural components. Roof cover and roof sheathing resistance sampling Each of the structural types listed in Ta bles 6-1 and 6-2 has a predetermined roof sheathing panel layout, as detailed in Chapte r 4. Variables describing the roof sheathing and cover are represented in the MCS engine as a series of matrices. Separate matrices with similar indexing strategies are created fo r the size, aggregate pressure coefficient, and capacity of both roof sheathing and cover. Using these similarl y indexed matrices, each individual panel of roof sheathing and representative area of roof cover can be investigated. Additionally, the interacti on between sheathing and roof cover can be addressed with the matrix indexing strategy. Wind resistance values for each individual panel or representative area of roof cover are obtained through i ndependent sampling of the PDFs described in Chapter 5. The MatLAB function randn() is used to generate a group of numbers randomly sampled from the standard normal dist ribution. The randomly generate d numbers are individually

PAGE 127

107 scaled using Equation 6-1, such that they represent a random sampling of the PDF of sheathing or roof cover capacity. The scali ng process described by Equation 6-1 is used for all of the sampled resistances in the MC S engine. For the case of roof sheathing and roof cover, two unique matrices of z values are created using the randn() command. The size of the matrix is linked to the predeter mined number of individual sheathing panels on the roof surface. One of the matrices is scaled to represent the capaci ty of roof sheathing, and the other similar, but unique, matrix is scaled to represent the capacity of roof cover for the individual house under simulation. After the capacity matrices are created using the scaling pr ocess represented by Equation 6-1, the values are checked to en sure that they lie within two standard deviations of the mean capacity provided in Table 5-3. This process prevents unrealistic capacities (such as negative valu es) from being used in the simulation process. If a value lies outside the imposed bounds, it is rejected, and a new va lue is generated using the sampling and scaling process. This process cont inues iteratively until all values lie within two standard deviations of the mean capacity. Roof-to-wall connection resistance sampling The number of roof-to-wall connections that must be assigned unique capacity values on each wall of a simulated home is determined by the truss spacing. For gable roof homes, the truss spacing is used to determine the number of intermediate trusses attached to the side walls. Each of these tr usses has two connections, while the gable end trusses are assumed to be conn ected to the end wall in eigh t places. For hip roof homes, each outer wall has a number of connections equa l to the length of the wall divided by the truss spacing.

PAGE 128

108 The capacity of roof-to-wall connections is obtained through a batch sampling process discussed briefly in Chapter 5. This process is used in the MCS engine only for roof-to-wall connections, specifically to acc ount for the observed damage state of the entire roof system separating from the re maining structure. Using the batch sampling process, a house-specific starti ng value is obtained using the randn() command. For hip roof homes, this single generated value is th e starting point for all roof-to-wall connection capacities on the individual structure. For gabl e roof homes, separate starting values are generated for the side connector s and gable end connectors. The starting value(s) for each house are scaled to represent values selected from the PDF of connection capacities us ing the process defined in E quation 6-1. After scaling, the value(s) are checked to ensure that they lie within two standa rd deviations of the mean provided in Table 5-3. If a value lies outside of these bounds, it is rejected, and a new value is sampled and scaled. This proce ss continues iteratively as needed. The scaled value(s) generated in this process represen t the mean resistance for the population of roof-to-wall connections on th e house. Individual connector capacities are obtained using a group of numbers generated by the randn() command and scaled using Equation 6-1, with the generated mean for the popul ation of connectors on the house and a COV of 0.05 The results of this substitution are shown in Equation 6-2, where y is the resistance capacity of one connector, z is a randomly generated value from the standard normal distribution, and ˆ is the generated mean resistance for the connectors on the house. ˆ 1 05 0 z y (6-2) The process of sampling represented in E quation 6-2 is shown in Figure 6-5. The distribution on the left represen ts the PDF of a type of conn ector, as described in Chapter

PAGE 129

109 5. Selected values from this distribution b ecome mean values for individual homes, as shown on the right hand side. Capacity BoundsProbability of Occurrence PDF for Type of Connection Capacity PDF for House 1 Probability of Occurrence Capacity PDF for House 2 Probability of Occurrence Figure 6-5. Batch sampling met hod for roof-to-wall connections The process of batch sampling results in a narrow distribution of connector capacities on an individual building, as depi cted by the more tightly centered PDFs on the right hand side of Figure 6-5. Using this method of sampling, and employing a load redistribution scheme for failed roof-to-wall connections allows for the possibility of the entire roof structure pulling off of the structure, a damage st ate that has been observed in post-damage reports [19]. Wall resistance sampling Wall resistance capacities are dependent on the structural type. As shown in Table 4-5, concrete block homes, wood frame home s, and manufactured homes have different limit states for walls. Structural damage checks common to site-built homes include shear wall failure, uplift failure, and out-of-plane loading failure. Each of the four perimeter

PAGE 130

110 walls on an individual site-built home is assi gned a shear wall, uplift, and out-of-plane capacity using the sampling method common to most components. The randn() command is used to generate three sets of four star ting values. The four valu es represent the four perimeter walls. Each set of four values is sc aled using the appropria te values from Table 5-3 in Equation 6-1: one set for shear wall resi stance, one for uplift, and the third for outof-plane loading capacity. For concrete bl ock walls, the out-of-plane capacity is the allowable bending moment, while wood walls ha ve an out-of-plane capacity associated with the lateral strength of framing connectio ns. Each of the twelve generated resistance values is compared to the appropriate mean from Table 5-3. Indivi dual values that do not lie within two standard deviations of the mean are rejected, and new values sampled until the truncation criteria are met. A structural damage check common to mo st structural types, but conducted on different areas of each, is the wall sheathing failure test. Due to the typical location of plywood or vinyl sheathing, this check is us ed on the entire wa ll surface of wood frame and manufactured homes, but only for the triang ular gable ends of concrete block homes. Predetermined sheathing panel layouts for each of the structural types listed in Tables 6-1 and 6-2 are used similarly to the roof sheathi ng panel layouts. A matrix of values is used to represent the capacity of each individual sheathing panel. The number of panels is determined by placing typical 4 x 8 ft pieces of plywood on site-built homes, and 3 x 12 ft pieces of vinyl siding on manufactured homes. The randn() command is used to generate a starting value for each panel, which is scaled using the appropriate values from Table 5-3 in Equation 6-1. I ndividual values that do not lie within two standard deviations of the mean are rejected, and new values sampled until the limits are met.

PAGE 131

111 Opening resistance sampling Wind pressure resistance values for indi vidual openings are obtained using the number generation and scaling process previous ly defined. Unique values are generated using the randn() command for the front and back entr ance doors and for the garage door on site-built homes. Individual wi ndows are also assigned unique z values using the randn() command. These are stored in matric es designed to maintain both size and location information for each window. All of the generated z values for openings are scaled using Equation 6-1 with the appropriate mean and COV from Table 5.3. For openings, the truncation limits are set at 2.5 ti mes the standard deviation rather than the bounds of two standard deviations used for other components. Discrete resistance values are not required for the case of windborne debris impact. As discussed in Chapter 4, windows that are not impact-resistant cannot usually withstand a direct hit from a piece of typical debris. Instead, parameters are used to determine the likelihood that a window will be struck and broken, given the wind speed, building geometry, and some assumptions abou t the surrounding terrain and availability of missiles. Using this approach, both the lo ading conditions and re sistance capab ility are built into a single distribution, ) ( V pD. As detailed in Chapter 4, the number of window sizes and windward wall scen arios results in twelve ) ( V pD functions per modeled building. A wind speed dependent value for eac h of the twelve functions is generated during the variable definition st ep. These values are used in the initial failure check to determine the number of windows broken by windborne debris.

PAGE 132

112 Tie-down anchor resistance sampling The pull-out capacity of each tie-down anchor on a simulated manufactured home is generated using the random number generation and scalin g process defined for roof sheathing and roof cover. A unique value is generated per tie-down anchor using the randn() command. These are scaled using Equation 6-1 with the mean and COV provided in Table 5.3. Values are screened such that th ey lie within two standard deviations of the mean. Initial Failure Check After the deterministic 3-second gust wi nd loads and probabilistic component resistance values have been determined, th e MCS engine begins a sequence of failure checks to determine the level of structural damage on the individually simulated building. Failure checks are ordered to represent the most likely sequence of events and load paths. Within the initial check, loads and resistance capacities are compared independently for roof sheathing, walls, and openi ngs. Once the damage is calculated from the initial loads, the internal pressure is adjusted. This proce ss is discussed after the initial failure checks are described. Initial failure check for roof sheathing The pressure resistance capacities of r oof sheathing panels are compared to aggregate wind pressure loads using a pane l-by-panel comparison between the load and resistance matrices. Individual sheathing panels with aggregat e wind pressure equal to or greater than the sampled resistance cap acity are marked as failed by changing the capacity value from a generated number to a value of zero. The similar indexing strategy for the roof cover matrix allows the MCS engi ne to fail roof coveri ng locations at failed sheathing panels. In this manner, the roof c over over a piece of sheathing that has pulled

PAGE 133

113 off is also assumed to be pulled away, and is assigned a capacity of zero for use in the next round of failure checks. Initial failure check for walls During the initial failure check, walls on s ite-built homes are evaluated in uplift, out-of-plane loading, and shear wall loading. Given the leve l of uncertainty involved in the loading conditions and streng th of typical residential wall s, as well as the lack of information with which to compare modele d results, the current MCS engine does not discriminate between degrees of damage to wa lls that fail in uplift, out-of-plane loading, or shear wall loading. A more sophisticated de scription of wall failure could be a targeted area of future research for additional itera tions of the structural damage model. The current model uses the procedures detailed in this section to mark perimeter walls as ‘failed’ or ‘un-failed’ fo r these loading conditions. As described in Chapter 5, concrete block walls and wood frame walls behave differently under similar loading conditions. Fo r concrete block walls, a combined case of uplift and bending is used to determine poten tial failure. The unity check performed for this condition is describe d in Equation 6-3, where U is the unity check value, P is the applied uplift per foot of wall fr om roof-to-wall connection loads, Pallow is the sampled value of allowable uplif t per foot of wall, M is the applied bending moment at the center of the wall, and Mallow is the sampled value of allowable bending moment. allow allowM M P P U (6-3) The applied bending moment, M, in Equation 6-3 is obtained using the MWFRS pressures described in Chapter 4 along a 1foot strip of wall with the assumption of simple supports at the top and bottom of the wall. This is an oversimplification of the true

PAGE 134

114 conditions at every point along the wall. Give n the current body of information, however; the unity check of Equation 6-3 is an adequate analysis of conditions at the centermost point on each wall, where the effects of in teraction with other framing elements are minimal. If the value of U obtained for a wall is greater than or equal to one, then that wall is noted as having a structural failure that requires repair. Unlike concrete block walls, wood frame wa lls are checked inde pendently in uplift and out-of-plane loading. The uplift per f oot of wall applied by the roof-to-wall connections to wood frame walls is compared to the sampled value of allowable uplift per foot of wall. Additionally, a lateral load fa ilure check is conducted using the trapezoidal tributary area shown in Figure 4-8. The lateral force per foot of wall created by this wind pressure zone is compared to the sampled value of lateral resistance per foot of wall. If the applied load in uplift or lateral wind pr essure is greater than or equal to the corresponding sampled resistance, then the wall is marked as having a structural failure. The shear wall load case described in Chap ter 4 is common to both concrete block and wood frame homes. For this loading condi tion, the shears V1 and V2 pictured in Figure 4-7 are divided by the length of the wall to which they are applied to obtain a distribution of load per foot of wall. Case A is applied for winds perpendicular to the ridgeline, and Case B is used for winds para llel to the ridgeline. For cornering winds, Cases A and B are applied independently. The calculated value of shear per foot on each of the perimeter walls subject to this lo ading condition is compared to its sampled resistance value. If the applied load is great er than or equal to the sampled resistance, then the wall is marked as having a structural failure.

PAGE 135

115 Initial failure checks for walls on manuf actured homes are conducted for sheathing pull off. The framework of this check is simila r to that of roof sh eathing panels. For each panel of vinyl siding, a compar ison is made between the wind pressure on the wall at that location and the sampled resistance of the indi vidual panel. Individual panels with wind pressures equal to or greater than the sampled resistance capacity are marked as failed by changing the capacity from a genera ted number to a value of zero. Initial failure check for openings Opening failures consist of door, garage door (site-built homes only), and window pressure failures, as well as window imp act failures. For each door (including the garage), the wind pressure on the wall at the location of the door is compared to the sampled resistance capacity. Doors with an app lied pressure greater than or equal to the resistance capacity are marked as failed. A similar process is used for windows, after the windborne debris impact check has been conducted. Windows along the windward wall, or both windward walls in the case of cornering winds, are checked for windborne de bris impact failure. For each window on this windward area, a value is sampled from a uniform distribution with a range of 0–100. This value represents a randomly chosen pe rcentage between 0% and 100% for each of the windows subject to potential windborne de bris impact. The randomly selected values for each windward window are then compared to the appropriate predetermined value of ) ( V pD, the windborne debris function desc ribed in Chapter 4. Values of ) ( V pD represent the likelihood of breakage for specific combina tions of wind angle, wind speed, window size, and building geometry. Thus, a ) ( V pD value of 60% indicates that a window of particular size, on a given area of windward wall, and s ubject to defined wind

PAGE 136

116 loading conditions will be broken by debris si x out of ten times. If the value sampled from the uniform distribution between 0–100 fo r this individual window falls below 60 (which has a six in ten chan ce of occurring), then the wi ndow is designated as failed by impact loading. If the sampled value is equal to or greater than 60, then the window is not broken by impact. Windows failed by impact loading are marked in the simulation routine by setting their pressure resistance capacitie s to a value of zero. Once this step is complete for the windward windows, the MCS engine conducts th e pressure failure check for all windows using a point-to-point method to compare individual window pre ssure loads to the unique sampled resistance values. Individual failures are tallied when the applied pressure is equal to or greater than the resistance capacity. Internal Pressure Evaluation and Recalculation of Loads Following the initial failure check for roof sheathing, walls, and openings, the condition of openings is evaluated to determ ine the effect of damage on the internal pressure. If no openings have failed, then the in ternal pressure is left unchanged. If one or more openings are damaged, then a new internal pressure for the structure is calculated as the weighted average of pressures at locations of failed openings. Given the amount of uncertainty in the m odeling of opening size, location, strength, and loading, the weighting factors are not pr ecise ratios of square footage. Instead, windows and typical front and back doors are gi ven an equal weight, while garage doors are factored in at a rate of four times the contribution of other openings. This process is represented in Equation 6-4, where pin is the new internal pressure, pg is the aggregate pressure on the garage door, pi is the pressure on an indi vidual failed window or door, and n is the total number of failed windows and doors, other than the garage door. gar is

PAGE 137

117 a variable that takes on a value of one when the garage door has failed, and a value of zero if the garage door has not failed. ) ( 4 ) ( 41gar n p p gar pn i i g in (6-4) Once the new internal pressure has been determined, it is checked against the initial value of internal pressure. This initial va lue is obtained by setting the values of pGC and piGCin Equation 4-2 to zero and 0.18, respectively. If the internal pressure remains the same (no openings have failed, or the average internal pressu re happens to agree with the initial value), then the existing component lo ads are carried forward to the final damage check. If the new internal pressure varies fr om the original value, then the pressures applied to the simulated home are recalculate d using Equation 6-5, which is a modified version of Equation 4-2. in p hp GC q p ) 8 0 ( (6-5) Final Failure Check and Damage Tally Final damage checks are conducted using the structural wind loads, after the internal pressure adjustment. This series of checks is structured to take advantage of the load path and dependence of some building components. Openings are re-checked for overpressure failure using the new pressure lo ads. Previously failed openings remained listed as damaged, and a point-by-point compar ison is used to determine if any additional windows or doors fail as a result of the change in internal pressure. Once the re-check is complete, the total number of failed windows and doors are recorded for the simulated home.

PAGE 138

118 Roof sheathing panels are also re-checked ; maintaining previously failed panels and investigating other panels to determine if additional failures result from the change in internal pressure. The total number of faile d sheathing panels is converted into a percentage by summing the square footage of failed panels over the total roof area. Once the roof sheathing check is complete, the roof c over check is conducted for the first time. This component is not examin ed during the initial failure investigation because the roof cover loads described in Chapter 4 are independent of the internal pressure. Roof cover failure is dependent on sheathing failure, however. Locations of roof cover over sheathing panels that have been damaged are automatically assigned a capacity of zero, such that the panel-by-panel comparison of loads to resistances for the roof cover areas will result in the failure of these locations. Fo r this reason, the roof cover check is conducted only after the sheathing pa nels have been investigated. In the same fashion as roof sheathing, the failed locatio ns of roof cover are converted into a percentage by summing the square footage of damaged roof coveri ng over the total area of the roof. Roof-to-wall connection load s are computed using the wind pressures on the roof sheathing with the adjustment to internal pr essure. A tributary area method is used to distribute loads from individual roof sheathi ng panels into the connections. The first failure check is conducted using these initia l connection loads. For each connector, the applied load is compared to the sampled value of resistance. If the applied uplift is greater than or equal to the probabilistically assigne d capacity, the individual connection is listed as failed, and its load is redistributed to in tact connections. The re distribution subroutine searches for the closest two intact connections on either side of a middle roof-to-wall

PAGE 139

119 connection, sharing the load of the failed connection with four neighboring connectors when possible. Specifically, one-third of the load is shed to each of the two closest connectors, and one-sixth of the load is dist ributed to the next cl osest intact connection on either side. When only one intact connection is available to the left or right of the failed connection, it receives ha lf of the load from the fa iled connection. This failure check and load redistributi on process occurs until no ne w connection failures are discovered, or until an entir e side of the roof is unzi pped from the supporting wall. The last components checked on site-built homes are the walls. Each wall is rechecked for failure in uplift, out-of-plane lo ading, and shear wall loading. For the final damage check, wall support is dependent on roof-to-wall connection failure. Walls with half or less of the roof-to-wa ll connections intact are no longer assumed to be simply supported. As described in Chapter 4, the be nding moment for concrete block walls with more than half of the roof-to-wall connections failed is scaled up by a factor of 2.8, and the tributary area for lateral loads on wood fram e walls is increased. Using the new loads, walls are investigated for additional failures with the methods described during the initial failure check. A final wall check common to every structur al type except hip roof concrete block homes, is the investigation of wall sheathi ng. For manufactured homes, this primary check is re-evaluated here using the adjusted wind loads. Panels previously failed are maintained on the list of damaged members, while the MCS engine looks for additional panel failures. For site-built homes, the wall sheathing check is secondary, conducted only after the other components have been reviewed. The gable end panels of concrete block gable roof homes and th e entire wall surface of wood frame homes are checked for

PAGE 140

120 sheathing failure in much the same way that roof sheathing failures are investigated. A panel-by-panel comparison of wall pressure and probabilistically assigned wall sheathing capacity is conducted. Panels with applied loads greater th an or equal to the assigned panel capacity are listed as failed. These damages are converted to a percentage by dividing the square footage of damaged panels by the total area of wall sheathing. The final structural checks for manufactur ed homes relate to tie-down anchors. Limit states for sliding and overturning ar e investigated using the probabilistically assigned pull-out capacity of the anchors and the calculated sliding and overturning loads discussed in Chapter 4. The sliding force on the manufactured home is resisted by the combined capacity of the anchors on the home. If the applied load is greater than or equal to the resultant capacity, then the home experi ences a sliding failure. This initial sliding may break tie-down anchors, but not move the home off of its foundation. To check whether the home has been removed from the foundation, an additional check is conducted. In this case, the static friction of the home on the pile foundation is added to the sliding resistance. For this calculation, the weight of th e home is probabilistically assigned by sampling from a distribution with a COV of 0.25 and mean of 30 psf times the square footage of the home. Weights out side the truncation limits of two standard deviations are re-sampled. The coefficient of fr iction is assumed to be 0.2, a value typical of wood on metals under wet condi tions [58]. During this check, if the resultant sliding force is greater than or equal to 1.2 times th e newly calculated resistance value, then the home is assumed to have major sliding damage. An independent anchor check is conducte d for overturning. In this case, the capacity to resist overturning is calculated by summing the resistive moment from the tie-

PAGE 141

121 down anchors and the structural weight about at the location of th e support pier for the leeward wall. As shown in Figure 6-6, the enti re weight of the home is assumed to act at the centerline. A resultant wind force acting at mid wall height is calculated from the Main Wind Force Resisting System (MWFRS ) wall loads, and the tie-down anchor forces on the windward wall side contribute at a 45 degree angle. Moments are summed about the foundation pier on the leeward side of the home, represented by a small black circle shown on the right hand si de of Figure 6-6. The anchor s on the leeward side of the home do not contribute to th e overturning resistance capacity since they act through the location of the summation of moments. 7 ft Resultant Wind Force Weight Anchor forces Figure 6-5. Location of forces for the overt urning failure check on manufactured homes If the applied overturning moment is greater than or equal to the resistive moment of the weight and tie-down anchors, then th e home has overturned. Due to the geometry of typical manufactured homes, overturning is not expected to occur prior to sliding when the winds approach parallel to the ridgelin e or from the corners. For this reason, overturning is considered only for wind a ngles perpendicular to the ridgeline. In case of overturning and/or sliding, no attempt is made to identify the discrete number of failed tie-down an chors. Given the amount of uncertainty in predicting

PAGE 142

122 damage to manufactured homes and in quant ifying the monetary loss associated with damage, efforts to identify i ndividual tie-down anchor loss would not add to the accuracy of end results. The state of overturning or sliding applies to the entire home, not to individual tie-down anchors. Structural Damage Output Files The results of the structural damage-pre diction model will be used to develop insured loss functions for the Public Loss Hu rricane Projection Model. These functions will include repair and replacement costs as well as additional costs for loss of contents. For ease in post-processing and to maintain th e flexibility of including additional insured loss variables which might be linked to specif ic damages, the structural damage output from the MCS engine is stored in individua l output files for each combination of wind speed, angle, and building type. For example, a typical output file holds the results of several hundred thousand simulations of Sout h Region Concrete Block Hip Roof (CBH) homes at 45 150 mph 3-second gust winds. For site-built homes, the damage information for each simulated home includes Percent (by area) of failed roof sheathing Percent (by area) of failed roof cover Percent (of total) roof-to-wall connections failed Number of damaged walls (4 total) Number of damaged windows (15 total) Number of damage windows that were broken as a result of debris impact Number of failed entr y doors (2 total) Indicator variable for the garage door (0 = unfailed, 1 = damaged) Percent (by area) of damaged gable e nd panels (0 for hip roof buildings) Percent (by area) of damaged wall sheathing panels (for wood homes) The calculated internal pressure

PAGE 143

123 Similar information is stored for manufact ured homes, with the addition of two variables for sliding and ove rturning. Damage information for manufactured homes includes Percent (by area) of failed roof sheathing Percent (by area) of failed roof cover Number of roof-to-wall conn ections failed (58 total) Number of damaged windows (8 total for singlewide, 10 to tal for doublewide) Number of damage windows that were broken as a result of debris impact Number of failed entr y doors (2 total) Percent (by area) of dama ged vinyl siding panels Indicator variable for sliding (0 = no sliding, 1 = minor, 2 = major) Indicator variable for overturning (0 = not overturned, 1 = overturned) Summary The MCS engine described in this ch apter simulates the performance and interaction of components of typical Florida homes during hurricane winds, using the structural loads discussed in Chapter 4 and the building component resistances described in Chapter 5. This model is developed in partial fulfillment of the engineering tasks for the Public Loss Hurricane Projection Model. Re sults of this effort, presented in Chapter 7, will be used to create insu red loss functions for the pred iction of annual risk in the State of Florida. Additional uses for the MCS engine outside the scope of the current project include the development of an on line learning laboratory where engineering students and homeowners can learn about extreme wind loads through the use of a graphical user interface to the MCS engine that allows the user to change building construction and storm parameters, and then s ee a visual representation of the resulting structural damage.

PAGE 144

124 CHAPTER 7 STRUCTURAL DAMAGE VALIDATION AND RESULTS Results detailed in this chapter represent the structural vulnera bility of typically constructed homes to wind damage, using th e Monte Carlo Simulation engine described in Chapter 6. Wind loads and building compone nt capacities incorporated in the damage simulation are described in Chapters 4 and 5, respectively. Structural damage results obtained from the simulation engine are us ed to determine insured losses on an annualized basis or as the result of a sp ecific storm for the Public Loss Hurricane Projection Model (PLHP). A discussion of the methodology and preliminary findings for the relation of physical damage to insure d losses is provided in Chapter 8. A descriptive list of the homes selected to represent the Florida building stock is provided in Tables 7-1 and 72. Homes modeled in the simulation process detailed in Chapter 6 are described in Table 7-1, while Table 7-2 provides a lis t of structures for which the vulnerability is based on the perfor mance of selected types in Table 7-1. The process of selection of structural types and relevant building components is discussed in greater detail in Chapter 3. As described ear lier, two sets of homes selected for modeling have the same dimensions and structural de scription. These are the North and Central wood frame homes and the Central and South/Ke ys concrete block homes. Results for the pairs will be the same because these home s were similarly described in the county property databases. The regional descriptions are maintained throughout this document, however, for clarity of the methodology used to generate insured losses. Additionally, future iterations of the PLHP model are expe cted to incorporate changes to these models.

PAGE 145

125 Table 7-1. Modele d structural types Structural Type Description Roof Type Area (ft2) North CBG 1 Story concrete block in North FL Gable 2128 North CBH 1 Story concrete block in North FL Hip 2128 North WG 1 Story wood frame in North FL Gable 2280 North WH 1 Story wood frame in North FL Hip 2280 Central CBG 1 Story concrete block in Central FL Gable 2640 Central CBH 1 Story concrete block in Central FL Hip 2640 Central WG 1 Story wood frame in Central FL Gable 2280 Central WH 1 Story wood frame in Central FL Hip 2280 South/Keys CBG 1 Story concrete block in South FL or Keys Gable 2640 South/Keys CBH 1 Story concrete block in South FL or Keys Hip 2640 South/Keys WG 1 Story wood frame in South FL or Keys Gable 2464 South/Keys WH 1 Story wood fram e in South FL or Keys Hip 2464 MH 1 Manufactured home Gable 728 MH 2 Manufactured home Gable 1456 MH-pre Pre-HUD Code Manufactured home Gable 728 Table 7-2. Structural type s with damage based on combin ations of modeled buildings Structural Type Models Used to Predict Structural Performance North 2 story North WG and North WH Central 2 story Centra l CBG and Central CBH South 2 story South/Keys CBG and South/Keys CBH Keys 2 story South/Ke ys CBG, South/Keys CBH South/Keys WG, and South/Keys WH The first section of this chapter provide s a discussion of the validation of the system using the limited data available fr om Hurricane Andrew. Results for South/Keys CBG homes are typically used to valida te methodology used for all homes in the simulation process, since the Hurricane Andrew data consists mainly of homes of this type. Following the validation discussion is an investigation of the batch selection method used for roof-to-wall connections and an investigation of the results for different roof shapes. The last secti on provides structural damage prediction results for typical Florida homes. Specifically, results obtained fr om the simulation engine are presented for site built homes in the South/Keys Regi on and for manufactured homes. The results provided in this chapter are limited, for the sake of brevity. Additional results for these

PAGE 146

126 structural models are provided in Appendi ces A through G. Results for the North and Central Regions will be available when the PLHP is released in May, 2005. Structural Damage Validation The structural wind loads described in Chapter 4 and building component capacities described in Chapter 5 are based on building codes, av ailable literature, manufacturer data, and engineer ing judgment. As described in Chapter 6, selected values for loads and resistances are applied within a component-based probabilistic framework for the prediction of damages at varying levels of storm intensity. This approach is also used for the HAZUS model, developed by Applied Research Associates under the direction of the National Institute of Bu ilding Sciences for the Federal Emergency Management Agency [3]. While the component -based method is considered state of the art in hurricane damage predic tion, the specific values selected for loads and resistances must be validated with available hurricane damage reports. Unfortunately, data available for comparis on is limited. Most damage reports from past storms provide expert opinion on the types of damages observed, and potential building code or construction mitigation effo rts, but few reports provide statistically significant numbers of detailed damage result s for individual structures. Only one report has been located to date that offers the type of information necessary to validate choices made in the development of the structural damage simulation engine for the PLHP. Using information provided in the National Associ ation of Home Builder s Research Center (NAHB) 1993 assessment of Hurricane Andrew damage [19], load and resistance values used to determine structural damage are valida ted to the extent possi ble. A description of the data provided by this 1993 report, a discussion of validation techniques, and comparisons of damage results for individual components are presented in this section.

PAGE 147

127 NAHB Report on Hurricane Andrew The report conducted by the NAHB Res earch Center for HUD on the damages observed during Hurricanes Andrew and Inik i was the first post-damage assessment to use standardized forms for damage inform ation collection on a significant number of homes [19]. Data was collected in South Florida and Loui siana for Hurricane Andrew and in Hawaii for Hurricane Iniki. A detailed description of the cl uster sampling process by which the 515 homes in South Florida were selected is provided on pages 18-20 in the NAHB report [19]. Of the 515 homes assessed, 460 damage reports were included in the published volume. Levels of damages observed on surveyed Florida homes are presented in Table 7-3, with values taken from Tabl e D-2 of the NAHB report. Cases in which the level of damage was not speci fied in the NAHB report incl uded homes where tarps or other obstructions prevented observers from adequately characterizing the level of damage. Table 7-3. Hurricane Andrew damage s surveyed in the 1993 NAHB report Level of Damage Windows Walls Roof-to-Wall Connections Roof Sheathing Roof Cover 1/3 or less 33%96%85%57% 18% 1/3 – 2/3 26%1%6%12% 23% 2/3 or more 34%2%6% 36% Not specified 6%3%6%25% 23% The majority of homes discussed in the NAHB report represent a limited number of structural types. Specifically, 99.6% (464 of 466) of the South Florida homes presented in Table 7-3 above are masonry structures, a nd 81% have gable roofs [19]. A limited resurvey of 34 structures to gain more in formation concerning wood frame home damages is summarized in Table D-5 of the NAHB report. Of these wood frame homes, an

PAGE 148

128 unspecified number have masonry first fl oors with wood frame second stories. The results of the re-survey for wood frame damage are presented in Table 7-4. Table 7-4. Wood frame home damage s surveyed in the 1993 NAHB report Level of Damage Windows Walls Roof-to-Wall Connections Roof Sheathing Roof Cover 1/3 or less 65%82%85%56% 41% 1/3 – 2/3 24%18%6%26% 21% 2/3 or more 9%6%18% 38% Not specified 3%3% Damage levels presented in Tables 7-3 a nd 7-4 are used as a means of validating the simulation engine results. Specific uses and limitations of the data provided by the NAHB report are described in the following section. Application of the NAHB Report Data as a Validation Tool The NAHB report provides a unique dataset of structural damage s resulting from hurricane landfall, and it is currently the only source of data with which to compare simulated structural damages for the purpos e of validation. The information reported must be considered within an appropriate framework; however, for use as a method of validating damage predicted by the simula tion engine. Specifically, the wind speeds, angles of approach, and types of homes represented by the data in the NAHB report impose limits on the use of reported damages to validate simulated damages. The implications of the data characteristic s and the validation methodology employed are described in this section. Homes investigated in South Florid a for the NAHB report were closely geographically spaced, indicati ng that the population represen ts a narrow value of storm intensity. Thus, a comparison of recorded to simulated damage is not possible for the wide range of wind speeds for which the simulation engine has been developed.

PAGE 149

129 Additionally, the winds speeds incurred during the passage of Hurricane Andrew through the neighborhoods surveyed for damage have be en a source of controversy in the years following Andrew’s landfall [59]. Therefor e, direct comparisons between reported damages and simulated damages obtained fr om the developed structural damageprediction model can be made only for a narrow range of wind speeds, the value of which is disputed by experts in the field. In addition to representing a single value of storm intensity, ho mes described in the NAHB report most likely represent a limited num ber of wind approach angles. Given that the buildings were typical aligned in nei ghborhood rows; the initial damage on most of the surveyed homes in each neighborhood wa s caused by winds approaching from the same angle with respect to the ridgeline of the roof. The results of the damage-prediction model include eight different wind directi ons, however. Since the damage-prediction model has been developed to encompass nei ghborhood layouts in all Florida zip codes, the scope of the work is necessarily broad. The observed damage as a result of Hurricane Andrew is by definition, a single scenario with a limited scope in terms of wind approach angles. As a result, comparisons between the two efforts at the wind speeds represented by the Hurricane Andrew data should be qualitative. A quantitative comparison such as a calculation of the per cent error between the simulated data and the Hurricane Andrew data should not be used as a measure of simulation validity. Further limitations on the use of the Hu rricane Andrew data in the 1993 NAHB report are incurred as a result of the buildi ng population in the area of landfall. As described in the previous s ection detailing the NAHB report re sults, most of the buildings surveyed were masonry homes with gabl e roofs. Because they were closely

PAGE 150

130 geographically spaced, the homes were mo st likely built around the same time with similar materials and construction crews. I ndividual homes probabl y had different floor plans, however, with differing numbers of windows, outside dimensions, locations of interior walls, etc. As a result, the popul ation of homes described in the NAHB report represents homes of similar age in one local area, while the damage-prediction model encompasses broader regional zones and homes of all ages. The building classification used in the structural damage model that most closely resembles the building population of the homes surveyed for damage after Hurricane Andrew is the South/Keys CBG described in Table 7-1. This building type is modeled to represent typical homes in the entire South Florida and Fl orida Keys Region, however. The dimensions, numbers of windows, and other unique aspects of i ndividual homes surveyed for the Hurricane Andrew report are not likely to match the ch aracteristics of the simulated model. These uncertainties reinforce the need for qualitat ive validation rather than exact numerical comparison of results between the NAHB data and the simulated damage results for South/Keys CBG homes. Validation of Individual Components In spite of the limitations described in the previous section, the NAHB data collected in the wake of Hurricane Andrew remains the only source of statistically significant information with which to compare the simulated damage results. Qualitative validation is presented in the following s ections for individual building components. Specifically, comparisons of trends and range s of predicted damages are made between the NAHB report data and simulated damages for 3-second gust wind speeds representative of Category 4 storms on th e Saffir-Simpson scale. Wind speeds of 160, 175, and 190 mph 3-second gusts are selected for simulated results because these wind

PAGE 151

131 speeds represent low, medium, and high values of Category 4 storm intensity. Given the limitations of comparison, quantitative measur es of similarity between reported damages and simulated damages are typically not calcu lated. Instead, trend and range comparisons between the narrowly defined Hurricane A ndrew population of homes and the more broadly defined South/Keys CBG homes ar e noted as proof of concept for the methodology used in the damage-prediction model. In addition to the validation of damage results with the NAHB data, a brief description of the vulnerability and fragi lity curves are provided for each component discussed in the following sections. Concept de scriptions of vulnerabi lity and fragility are provided in Chapter 2, and shown in Figure 2-6 through 2-8. The mean damage results presented in the vulnerability curves and the probabilities of exceedi ng discrete levels of damage presented in the fragility curves provide detailed information for predicted damages over the entire range of wind speed s for which the simulation model has been developed. These curves cannot be validated with damage results because data is not available at each wind speed. E ngineering judgment is used to determine whether the shape and location of each curve represents a reasonable expectati on of the level of damage for the component in question. It is worthy to note at this point in the discussion of validation concepts that a detailed validation of structural damage is unlikely to be made for commercially developed risk models. Typically, the develope rs of commercial models for the insurance industry have proprietary claims data availa ble for validation purposes. Using the claims data, the methodology of the commercial models is validated at the monetary level. This level of validation will be conducted for the PL HP model before it is released in its

PAGE 152

132 entirety in 2005. At the pr esent time, however, the PLHP is incomplete. The results presented in this dissertation represent one piece of a multi-university project encompassing several fields of study. As a result, the structural damage results presented in this chapter for windows, walls, roof-towall connections, roof sheathing, and roof cover are investigated to a much greater extent than similar results for proprietary models are likely to be scrutinized. Validation of window damage Using the results obtained from 4,00 0 individual buildi ng simulations, the percentage of homes with windows in each damage category specified by the NAHB report is shown in Table 7-5. Values fr om Table 7-3 for window damage observed on homes surveyed in Florida after Hurr icane Andrew for the 1993 NAHB report are reprinted in Table 7-5 for comparison. Table 7-5. Window dama ge from Hurricane Andrew vs. simulated data Simulated Data Window Damage Level Hurricane Andrew Data from NAHB 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 33% 46.6%20.4% 7.0% 1/3 – 2/3 26% 51.6%69.4% 60.6% 2/3 or more 34% 1.8%10.2% 32.4% Not specified 6% Trends in the data presented in Table 7-5 indicate that the values selected for wind loads and resistances on expos ed windows adequately characterize the performance of typical Florida residences. From the Hurrican e Andrew data, damage to windows can be interpreted as being close to evenly distri buted between the three levels. The simulated data does not appear to be as evenly distri buted, but it does indicat e that the mean value of damage to windows during Category 4 st orms lies in the middle third (1/3 – 2/3 damage). Histograms of window breakage prov ided in Figure 7-1 for Category 4 wind

PAGE 153

133 speeds indicate that the window damage is more evenly spread in the simulated data than Table 7-5 would suggest. Histograms provided in Figure 7-1 are normalized such that the area under each curve is unity Specific choices of window size and location in the simulated models result in the lack of smoothne ss in the histograms in Figure 7-1. This is most apparent for the case of 190 mph 3-second gust winds. Th e variation of the likelihood of occurrence for numbers of br oken windows greater than 10 on a given home, however, will not advers ely affect the end product of insurable loss. A qualitative comparison of the damage represented by the hi stograms in Figure 7-1 with the results of the NAHB study presented in Table 7-5 indicat es that the methodology used to determine window breakage in the damage-prediction mo del is an acceptable representation of typical Florida homes. Figure 7-1. Histograms of window damage on South/Keys CBG homes. Additional conclusions c oncerning the methodology employed in the modeling of window loads and capacity values can be dr awn from the simulated vulnerability and fragility curves. Mean window damage (prese nted in Figure 7-2) and probabilities of exceeding discrete numbers of broken window s (provided in Figure 7-3) cannot be

PAGE 154

134 validated with existing data. These curves are provided to indicate the predicted levels of damage over the entire range of wind speeds for which the damage simulation model has been developed. The vulnerability curve for window damage and the fragility curves for varying levels of damage indi cate that the selected loading mechanisms and resistance values produce simulated window damages that increase with reas onable expectation as the 3-second gust wind speed increases. Figure 7-2. Window damage vulnerabi lity of South/Keys CBG homes. Figure 7-3. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys CBG homes.

PAGE 155

135 Validation of masonry wall damage Using the results of 4,000 individual building simulations, the percentage of South/Keys CBG homes with wall damage in each category specified by the NAHB report is shown in Table 7-6. Values fo r Hurricane Andrew provided in the 1993 NAHB report and listed in Table 7-3 are reprinted in Table 7-6 for comparison. Masonry wall damage obtained by simulation appears to be hi gher than the damage observed as a result of Hurricane Andrew. However, possible differe nces in damage tallying methods exist, which could not be verified. The simulati on routine marks a wall as damaged if any masonry wall failure check discussed in Chapte r 6 is exceeded. As a result, if a home has two walls with even the smallest of structural cracks at the center, it will be labeled in the middle third (1/3 – 2/3 damage). The NAH B report provides example photographs of homes with varying levels of damage and indicates methods employed for standardizing the results obtained by different observers, bu t it does not provide a detailed description of the difference between levels of damage specific to walls. Table 7-6. Masonry wall damage from Hurricane Andrew vs. simulated data Simulated Data Masonry Wall Damage Level Hurricane Andrew Data from NAHB 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 96% 76.2%54.0% 26.8% 1/3 – 2/3 1% 23.8%45.8% 72.4% 2/3 or more 0.0%0.2% 0.8% Not specified 3% Additional data for masonry wall performance is presented in Fi gures 7-4 and 7-5. The vulnerability curve shown in Figure 7-4 indicates that the mean damage for Category 4 hurricane winds is, on average, 1.2 walls. Sp ecifically, the values of mean damage at 160, 175, and 190 mph 3-second gust winds are 0.72, 1.26, and 1.69 walls, respectively. Figure 7-5 indicates that damage to three wa lls is unlikely until wind speeds greater than

PAGE 156

136 approximately 165 mph 3-second gusts are experi enced, and damage to four walls is not likely to occur until wind speeds of 200 mph 3-second gusts. For these reasons, the performance of masonry walls using the selected load and resistance values is determined to be adequate for the repres entation of typical Florida homes. The simulation of masonry wall damage is noted, however, as an area th at could be targeted for future research. Figure 7-4. Wall damage vulnera bility of South/Keys CBG homes. Figure 7-5. Fragility curves for 1, 2, 3, and 4 damaged walls for South/Keys CBG homes.

PAGE 157

137 Validation of wood frame wall damage Use of the Hurricane Andrew data for th e validation of wood wa ll results is more difficult than for masonry homes. The num ber of wood frame homes surveyed for damage is small (34 homes), and includes homes with masonry first floors. Since the data presented in the NAHB report is the only be nchmark with which to make comparisons, Table 7-7 provides a means of investigating the differences between the NAHB data and the wood wall damage results obtained from a large set of 40,000 individual building simulations of South/Keys WG homes. Wood wall damage obtained by simulation appears to be higher than the damage observed as a result of Hurricane Andrew. However, the limitations in the data and th e possible differences in damage tallying methods indicate that strict numerical comp arison between the results is not warranted. The methodology used in the simulation of wood wall damage is accepted as an adequate portrayal of the performa nce of typical wood walls. Table 7-7. Wood frame wall damage from Hurricane Andrew vs. simulated data Simulated Data Wood Frame Wall Damage Level Hurricane Andrew Data from NAHB 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 82% 74.4%52.9% 30.6% 1/3 – 2/3 18% 25.6%47.1% 69.4% 2/3 or more 0.0%0.0% 0.0% Not specified Damage to wood walls over the range of wind speeds for which the simulation engine has been developed is presented in Figures 7-6 and 7-7. The vulnerability curve for wall damage to South/Keys WG homes provided in Figure 7-6 and the fragility curves for 1, 2, 3, and 4 walls provided in Figur e 7-7 indicate that the simulation routine provides a reasonably expected level of damage over a wi de range of wind speeds.

PAGE 158

138 Figure 7-6. Wall damage vulnera bility of South/Keys WG homes. Figure 7-7. Fragility curves for 1, 2, 3, and 4 damaged walls for South/Keys WG homes. Validation of roof-to-wall connection damage Observed Hurricane Andrew data indicates a nearly identical pe rformance of roofto-wall connections for masonry and wood fram e homes, in spite of the differences in manufacturer rated capacity. In light of the limited number of wood homes surveyed, the results of the NAHB study are not compared to wood frame roof-towall connections. A comparison is made for the masonry homes, to validate the choices selected for loading

PAGE 159

139 and capacity characteristics. Given that the masonry home me thods are satisfactory, the same methodology is deemed appropr iate for the wood frame homes. Table 7-8 indicates the pe rcentage of 4,000 individua l building simulations of South/Keys CBG homes in each damage state, as compared to the Hurricane Andrew data previously presented in Table 7-3. The Hurricane Andrew data compares closely to the percentage of simulated homes in each category using a 160 mph 3-second gust wind speed. At higher wind speeds within Category 4, the comparison is not as favorable. For 3-second gust wind speeds of 175 and 190 mph, the damage simulation model predicts slightly higher levels of damage than observed after Hurricane Andrew. Given the limitations previously discussed, this qua litative comparison indicates that the methodology selected for the roof-to-wall conn ections in the damage simulation model adequately represents the perfor mance of typical Florida homes. Table 7-8. Roof-to-wall connection damage from Hurricane Andrew vs. simulated data Simulated Data Connection Damage Level Hurricane Andrew Data from NAHB 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 85% 83.2%67.4% 52.0% 1/3 – 2/3 6% 11.0%20.2% 28.6% 2/3 or more 2% 5.8%12.4% 19.4% Not specified 6% Additional roof-to-wall dama ge information, over the broad range of wind speeds for which the damage simulation model ha s been developed, is provided in the vulnerability curves and fragility curves of Figure 7-8 and 7-9. These curves represent a reasonable expectation for the level of damage at varying wind speeds. As a result, the wind load and capacity selections in the damage-prediction model for roof-to-wall connections are accepted as an adequate measure of performance for typically constructed homes.

PAGE 160

140 Figure 7-8. Roof-to-wall connection damage vulnerability of South/Keys CBG homes. Figure 7-9. Fragility curves for 2%, 5% 10%, 25%, and 50% roof-to-wall connection damage for South/ Keys CBG homes. Validation of roof sheathing damage A comparison of roof sheathing damage is made in Table 7-9 for data observed during the simulation of 4,000 individual Sout h/Keys CBG homes vs. reported Hurricane Andrew data previously shown in Tables 7-3 and 7-4. The Hurricane Andrew data compares reasonably well at wind speeds in the vicinity of 175 and 190 mph 3-second gusts, though the simulation engine might pred ict too little damage at wind speeds close

PAGE 161

141 to 160 mph. It is difficult to make this a ssertion from the data av ailable from Hurricane Andrew, however, given the limitations prev iously described and the large number of homes for which roof sheathing damage is unspecified. The qualitative comparison indicates that the damage model predicts sheathing loss consistent with the damages observed during Hurricane Andr ew for typical homes. Table 7-9. Roof sheathing damage from Hurricane Andrew vs. simulated data Hurricane Andrew Data from NAHB Simulated Data Roof Sheathing Damage Level All Homes Wood Only 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 57%56%95.0%76.2% 44.2% 1/3 – 2/3 12%26%5.0%23.2% 50.6% 2/3 or more 6%18%0.0%0.6% 5.2% Not specified 25% The vulnerability of South/Keys CBG ho mes to sheathing damage presented in Figure 7-10 represents a reas onably expected, increasing cu rve over the range of wind speeds for which the structural damage m odel has been develope d. Additionally, the fragility curves shown in Figure 7-11 indica te that the rate of increasing damage is reasonable, though the curves are slightly steeper than desired. Figure 7-10. Roof sheathing vulnerab ility of South/Keys CBG homes.

PAGE 162

142 Figure 7-11. Fragility curves for 2%, 5% 10%, 25%, and 50% roof sheathing damage for South/Keys CBG homes. Validation of roof cover damage Using the results obtained from 4,000 si mulations, the percentage of South/Keys CBG homes with roof cover damage in each category specified by the NAHB report is shown in Table 7-10. Hurricane Andrew result s are reprinted from Tables 7-3 and 7-4. The observations from Hurricane Andrew for a ll Florida homes are difficult to interpret, due to the large number of homes with unspeci fied damages. Given the explanation in the NAHB report that homes with unspecified da mages were most likely obscured by tarps covering damaged areas, or otherwise blocked from view, it is unlikely that all 23% of homes with unspecified damages in the left hand column would have damages less than or equal to 1/3. Some of these would fall into the middle category of damage, though the exact number is impossible to predict. Additi onally, the wood home results in the second column account for a small sample size. As a result, a comparison of simulated damages to those observed after Hurrican e Andrew for roof cover is less exact than the comparison for other building components. In spite of these difficulties, data provided in Table 7-10 indicates that the mean value of simulated roof cover loss lies in the middle damage

PAGE 163

143 category for all but the lowest intensity Ca tegory 4 storms, a favorable comparison with the data observed during Hurricane Andrew. Additionally, the number of homes with undamaged roof cover is zero or nearly zero fo r all three wind speeds, as expected from storms of this intensity. Given the cu rrent body of information, the methodology employed for roof cover failure checking in the simulation engine is a reasonable approach to predicting the beha vior of typical Florida reside nces during hurricane events. Table 7-10. Roof cover damage from Hurricane Andrew vs. simulated data Hurricane Andrew Data from NAHB Simulated Data Roof Cover Damage Level All Homes Wood Only 160 mph 3-sec Gust 175 mph 3-sec Gust 190 mph 3-sec Gust 1/3 or less 18%41%45.2%21.2% 7.4% 1/3 – 2/3 23%21%52.6%66.0% 57.0% 2/3 or more 36%38%2.2%12.8% 35.6% Not specified 23% The roof cover vulnerability of South/ Keys CBG homes presented in Figure 7-12 and the fragility curves of Figure 7-13 i ndicate that the simulation engine provides reasonable results over the broad range of wind speeds for which the model has been developed, though the fragility curves in Figure 7-13 are steeper than desired. Figure 7-12. Roof cover vulnerabi lity of South/Keys CBG homes.

PAGE 164

144 Figure 7-13. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof cover damage for South/Keys CBG homes. Investigation of Selected Topics Using the validated load and resistance me thodology, additional investigations are conducted for the use of batch selection pro cess and the difference between hip roof and gable roof results. The batch selection proce ss is investigated for roof sheathing and for roof-to-wall connections Differences in damages to hip a nd gable roof buildings are also investigated using these two m odeled structural components. Investigation of the Batch Sel ection Method for Roof Sheathing The batch selection method of capacity sampling provides a distribution of resistance unique to an individual home. This process represents the logical argument that individual pieces (sheathing panels, for exam ple) come from the same manufacturer and are installed on a home by the same group of workers. The method is employed by using a baseline from the distribution of roof sh eathing capacities as a mean capacity for the panels on a single house, and then samp ling from a new distribution with a COV of 0.05 to determine individual panel capacities. Uncer tainties in the buildi ng population (such as

PAGE 165

145 the diversity of roof plan layouts) and anom alies which lead to localized damages (such as a row of nails missing a truss on an ot herwise well-constructed home) are removed from the damage prediction process using this method, however. As demonstrated in Figure 7-14 (which provides normalized histograms of roof sheathing damage) the batching method does not adequately model post-storm damage observations for roof sheathing. In fact, the method produces just th e opposite result. The majority of homes should have damages in th e middle third, according to the Hurricane Andrew report. Figure 7-14A indicates that homes modeled with the batching process applied to roof sheathing w ould most likely have less than 20% or greater than 80% sheathing damage, with very few homes in the middle. Histograms of sheathing damage at 160, 175, and 190 mph 3-second gusts provided in Figure 7-14B indi cate that the roof sheathing damage for homes modeled without batching of the roof sheathing capacities agrees well with the Hurrican e Andrew data from Table 7-9. For this reason, batch sampling is not incorporated in the capacity assignments of structural components, with the exception of roof-to-wall c onnections. The physical argument to support this decision is addressed at the end of the next se ction describing roof-to-wall connections. A B Figure 7-14. Histograms of roof sheathi ng damage on South/Keys CBG homes. A) Batch selected method. B) Non-batch selected method.

PAGE 166

146 Investigation of the Batch Selection Method for Roof-to-Wall Connections The batch selection method is demonstrated in the previous sec tion to contradict the observed damage reports for roof sheathi ng failures. Using this component as an example, the method is not selected for us e in capacity definition of structural components, with the exception of roof-to-wall connections. Batch selection of capacity is selected for roof-to-wall connections alone because this component has a unique distribution of observed damage. Specifical ly, roof-to-wall connection damages should display two distinct and well-separated peaks, representing the observed damage states of little damage or catastrophic damage to the roof. To determine the effectiveness of the batch selection method for the roof-to-wall connections, a comparison is made between batch-selected roof-to-wall connection result s and non-batch-selected results. Normalized histograms for connection damage at wind speeds of 160, 175, and 190 mph 3-second gusts are presented in Figure 7-15, and fragili ty curves are presented in Figure 7-16. In each figure, the left hand side represents da ta using the batch selection process for roofto-wall connection capacity. Right hand sides of Figures 7-15 and 7-16 provide data for roof-to-wall connections sa mpled without the batching process. Data for non-batched connections is taken from a smaller data se t of 8000 simulations of individual South/Keys CBG homes. The difference between batch sampling and non-batch sampling is noted clearly in Figure 7-15, where the likeli hood of having only a few damage d roof-to-wall connections using the batch selection method is much hi gher at each wind speeds than the likelihood of having the same number of damaged wa ll connections using the non-batch selected method. The distribution of damages in Fi gure 7-15A indicate th e presence of two primary damage states, where the distribut ion of damages in Figure 7-15B indicate a

PAGE 167

147 single peak value that increases with wind sp eed. Additionally, the steeper slope of the fragility curves for high levels of roof damage on the right hand side of Figure 7-16 provide a different view of the same issue. The distribution of damages on the left of Figure 7-15 and the fragility curves on the left of Figure 7-16 are more representative of the expected physical damages to typical hom es than those appearing on the right hand side of each figure. A B Figure 7-15. Histograms of roof-to-wall conne ction damage on South/Keys CBG homes. A) Batch selected method. B) Non-batch selected method. A B Figure 7-16. Fragility curves for 2%, 5% 10%, 25%, and 50% roof-to-wall connection damage on South/Keys CBG homes. A) Batch selected method. B) Nonbatch selected method. In addition to the vulnerability and fragi lity curve data suppor ting the selection of roof-to-wall connection capaci ties using a batching process, typical construction practice

PAGE 168

148 supports the use of this methodology as well. Roof-to-wall connection capacity data, unlike data for some other building com ponents, has been well identified through correspondence with a leading manufacturer. Also, the as-bui lt capacities for these pieces of hardware are less susceptible to installati on procedure uncertainties than other building components. For these reasons, the batch selec tion process is maintained for roof-to-wall connections. Investigation of the Difference between Hip and Gable Roofs Using the validated methodology, a second co mparison is made for the shape of the roof. The data in Table D-2 of the 1993 NAHB report on Hurricane Andrew is largely for gable roof structures; however, a comparison is made for overall roof damage to hip and gable roof homes in the text of the report. Hip roof homes were more likely than gable roof homes to have low roof damage, though the specific numbers for roof-to-wall connections and roof sheathing losses are not available [19]. As a comparison between simulated data fo r gable roofs and hip roofs, Figure 7-17 provides histograms of roof-to-wall connecti on damage for South/Ke ys homes at 160, 175, and 190 mph 3-second gust wind speeds. Fi gure 7-17A provides data for simulated gable roof homes, while Figur e 7-17B provides the distribu tion of damage on hip roof homes. A clear difference is observed in the results obtained for r oof-to-wall connections on gable and hip roof homes at each of the three wind speeds shown. Higher levels of maximum damages are obtained for connections on gable roof homes than those for hip roof homes, as noted by the location on the xaxis of the highest damage level for each wind speed. Additionally, the li kelihood of only a few connections being damaged is much higher for hip roof homes at each of the investigated wind speeds than the

PAGE 169

149 likelihood of the same number of connections being damaged on gable roof homes at the same wind speeds. A B Figure 7-17. Histograms of roof-to-wall c onnection damage on South/Keys concrete block homes. A) Gable roof ho mes. B) Hip roof homes. An additional comparison is made between simulated data for gable roofs and hip roofs in terms of roof sheathing damage. Fi gure 7-18 provides the histograms for damage to roof sheathing on South/Keys homes at 160, 175, and 190 mph 3-second gust wind speeds. Differences between the curves are s light, indicating that the differences between hip and gable roof performance for roof sheat hing are not fully addressed in the damage simulation model. A B Figure 7-18. Histograms of roof sheathing dama ge on South/Keys concrete block homes. A) Gable roof homes. B) Hip roof homes.

PAGE 170

150 In spite of the lack of difference obtai ned between gable and hip roof sheathing loss, the method of loading is selected as an adequate representation of hurricane conditions, given the current body of knowle dge. Future study to determine the probabilistic character of r oof sheathing loads and other wind load conditions would result in more accurate modeling of the surface wind loads on typical buildings, which would most likely increase the variation in da mage results. Given the current information, the methodology used for roof sheathing in the simulation engine is accepted as a reasonable approach for modeling the behavior of typical resi dential structures. Since the level of roof sheathing damage at which high monetary values of insured loss is incurred is low, the error involved in the lack of difference between gable and hip roof homes in the current model is expected to result in sm all differences in the end product, prediction of insured loss on an annualized basis. Structural Damage Results In this section, limited results are presented for the site-built homes in the South/Keys Region and for manufacture d homes. Specifically, a mean damage comparison is provided for each type of re sidence, illustrating the overall picture of damage at differing wind speeds. For roof cover, roof sheathing, and roof-to-wall connections, the mean damage is presented as a percentage of the total that is damaged. Wall damage is also presented as a percentage, with respect to the total of four walls, and as a percentage of sheathing panels lost on gable ends for those st ructures with gable roofs. A full body of simulated results is provid ed for each type of site-built home in the South/Keys Region and for manufactured homes in the appendices. Damages to homes in the North and Central Regions will be availabl e when the PLHP is released in May, 2005.

PAGE 171

151 The structural damages presented in this sect ion are used to determine the insurable loss, as described in Chapter 8. Results for Site-Built Homes in the So uth Florida and Florida Keys Region Mean damages for South Florida and Flor ida Keys Region concrete block gable roof homes, concrete block hip roof home s, wood frame gable roof homes, and wood frame hip roof homes are presented in Figures 7-19, 7-20, 7-21, and 7-22, respectively. Concrete block homes have lower mean dama ge percentages for roof-to-wall connections than wood frame homes of the same roof shape. Additionally gable roof homes are more likely to experience high levels of damage to walls and roof-to-wall connections than their hip roof counterparts, with one excepti on. For concrete block homes, the damage to walls on hip roof homes surpasses that of gable roof home after 200 mph 3-second gusts are experienced. Figure 7-19. South/Keys CBG homes mean dama ges for roof cover, roof sheathing, roofto-wall connections, and walls.

PAGE 172

152 Figure 7-20. South/Keys CBH homes mean dama ges for roof cover, roof sheathing, roofto-wall connections, and walls. Figure 7-21. South/Keys WG homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls.

PAGE 173

153 Figure 7-22. South/Keys WH homes mean damages for roof cover, roof sheathing, roofto-wall connections, and walls. Results for Manufactured Homes Mean damage levels for singlewide, doubl ewide, and pre-HUD Code manufactured homes are presented in Figures 7-34, 7-35, and 7-36, respectively. Damages to each component are presented in term s of percentages. For roof cover, roof sheathing, and roof-to-wall connections the damages are presented in the same format as previously shown for site-built homes. Wall sheathing is sim ilar to roof sheathing, in that the mean percentage of damage to the total is provide d. For overturning, a percentage representing the likelihood of the home bei ng overturned is provided. In the simulation model a value of one is recorded for overturned homes and a value of zero represents homes that do not overturn. At each wind speed in Figures 734 through 7-36 the mean overturning value is multiplied by 100 to obtain a comparative percent. A similar process is used for sliding, except that there are two possible sliding categor ies. In the simulated data, a value of one

PAGE 174

154 indicates minor sliding, while a value of two represents major sliding. Zero is used for no sliding damage. For each wind speed, the mean sliding value is multiplied by 100 and divided by 2 to obtain a comparative percent. Figures 7-34 through 7-36 indicate the most common failure mechanisms for each type of home. Singlewide homes, for exampl e, experience overturning more frequently than doublewide homes, although th e larger homes are significantly more susceptible to roof pull off (which occurs when most of th e roof-to-wall connections fail). Additionally, Figures 7-34 through 7-36 show that each type of home is more likely to experience a sliding failure than an overtu rning failure. Damages to all co mponents, with the exception of roof-to-wall connections, are higher on pre-HUD Code homes than on modern manufactured homes. Figure 7-23. Singlewide manufactured homes mean damages for roof cover, roof sheathing, roof-to-wall connections, and walls.

PAGE 175

155 Figure 7-24. Doublewide manufactured homes mean damages for roof cover, roof sheathing, roof-to-wall connections, and walls. Figure 7-25. Pre-HUD Code singlewide manu factured homes mean damages for roof cover, roof sheathing, roof-t o-wall connections, and walls.

PAGE 176

156 Summary In this chapter, results obtained from the simulation engine are validated with postdamage information from Hurricane Andrew to the extent possible, given the limitations described. Additionally, the vulnerability (mean damage) and fragility (probability of exceeding discrete levels of damage) are investigated for building components as a means of verifying, through the use of engineering judgment, that the damage prediction engine produces reasonable results. The levels of st ructural damage for each simulated home are not presented, for brevity. Instead, compar ative graphs of mean damage results are presented for each site-built home in the South/Keys Region and for manufactured homes. Additional structural damage results for these homes are provided in Appendices A through G. Results for the North and Central Region homes will be available when the PLHP is released in May, 2005. Simulated stru ctural damage results (presented briefly in this chapter and more thoroughly in the append ices) are used to de termine the insurable loss associated with each type of building on an annualized basis, and for specific storms. As described in Table 7-2, the performance of two story homes will be a function of the performance of single story homes. Chapter 8 describes how the end product of insurable loss is obtained using the damage s calculated during simulation.

PAGE 177

157 CHAPTER 8 APPLICATION OF RESULTS AND CONCLUSION The previous chapters describe the development of a structural damage prediction engine designed for the Public Loss Hu rricane Projection Model (PLHP). Involving meteorological, engineering, actuarial, and comp uter resource components, the PLHP is a multi-university project scheduled for completion in 2005. The end result of this effort is a prediction of hurricane wind-induced insuranc e losses for residential structures by zip code in Florida on both an annualized basi s and for predefined scenarios (specific hurricanes). In support of this goal, three separate models are being developed: a hurricane model to provide probabilistic wind speed information for each zip code in the state, a structural damage model relating specific wind speeds to predicted losses for typical residential buildings in the state of Florida, and a financial model to relate structural damage to insurable losses. Comb ined, these three segments will become the final PLHP model. The University of Florida’s contribution to the project is the development of the structural damage prediction engine rela ting maximum 3-second gust wind speeds into predicted structural damage. The Monte Carl o simulation engine de scribed in previous chapters uses a probabilistic framework ba sed on a component view of typical homes. This approach explicitly accounts for the cap acity of individual structural components, load paths, and load sharing, to the extent possible with availabl e knowledge. Using this method, the developed simulation engine comp ares wind loads with the load-resistance capacity of building components to identify the likelihood of structural damage over a

PAGE 178

158 range of storm intensities. Steps necessary to the development of the component based damage-prediction model include Characterization of homes representative of the residential building stock in the state of Florida and identification of com ponents critical to wind damage-prediction modeling, presented in Chapter 3 Quantification of the wind-induced loads on building components and identification of appropriate load paths and load sharing for modeling purposes, discussed in Chapter 4 Characterization of the proba bilistic capacities of individual components to resist applied wind loads, detailed in Chapter 5 Creation of a probability-based system -response model that will simulate the performance and interaction of the co mponents of typical Florida homes and evaluate their vulnerability during intera ction with hurricane winds, presented in Chapter 6 and validated in Chapter 7. The results of the structural damage-predi ction model (presented in Chapter 7) will be incorporated into the final PLHP mode l. The other two components (the hurricane model and the financial model) are currently being developed by research partners on the PLHP team, and are not detailed in this diss ertation. This chapter briefly describes the process by which the three developed models will interact to predict insurable losses. Also provided in this chapter are a summary of the research contributions made by the author and a description of future us es for the developed simulation engine. Relating Structural Damage to Monetary Loss The Monte Carlo simulation engine presente d in the previous chapters of this dissertation predicts ex terior structural damage on typi cal Florida homes resulting from extreme wind events. Determining annualized an d specific event related insurance losses from the structural damage information provi ded by this model is a two-step process. First, a cost estimate model is used to dete rmine the monetary value of physical damage as a ratio of the value of the home. Second, an insured loss model combines the cost ratio

PAGE 179

159 data and insurance policy feat ures determined by the actuarial team and the probabilistic wind characteristics provided by the meteorology team to determine insured losses, on annualized basis or for specific storms. These two steps are not included in the research work conducted for the completion of this disse rtation, but are pres ented briefly in the following sections for the sake of completeness in the desc ription of the PLHP model. Further details will be available when the PLHP Model is released in 2005. Cost Estimate Model A cost estimate model to relate physi cal damage to monetary loss is being developed by research partners at Florida In stitute of Technology. Preliminary results are available for this model, which define the co st ratio of damage as a percentage of home value. The methodology uses three pieces of in formation to determine a cost ratio for each damaged home: 1) structural damages fr om the Monte Carlo simulation engine, 2) non-structural damage, and 3) replacement cost ratios. The first item (structural damage) is taken directly from the results presented in Chapter 7. Non-structural damages include building components such as ki tchen cabinets, carpeting, in terior walls, interior doors, ceilings, plumbing, mechanical, and electrical assemblies which are not included in the Monte Carlo simulation engine. Damages to these non-structural items must be determined as a function of the level of ex terior damage predicted by the structural simulation model. Once the level of total damage is determined, replacement cost ratios for structural and non-structural building components are used to characterize damage to individual homes. Additionally, damages to personal property (cont ents) in the home are predicted as a percentage of the total value. Tables 8-1 and 8-2 provide estimated repl acement costs, as a percentage of the value of a new home, for subassemblies of a typical masonry house in Central Florida

PAGE 180

160 with a shingle roof and hurricane shutters [60]. Replacement costs for the non-structural elements of the home, shown in Table 8-2, represent a significant portion of insurable losses. Because repairs to existing cons truction are more expensive than new construction, the sum of the structural repl acement cost ratios in Table 8-1 and the nonstructural cost ratios in Table 8-2 exceeds 100%. Table 8-1. Structural re pair cost ratios for Central Florida masonry homes Structural Subassembly Repair Cost Ratio Roof Sheathing 5% Roof Cover 7% Trusses 9% Exterior Walls 22% Windows 4% Shutters 2% Exterior Doors 1% Garage 1% Total51% Table 8-2. Non-structural repair cost ratios for Central Florida masonry homes Non-Structural Subassem bly Repair Cost Ratio Plumbing 10% Mechanical 7% Electrical 7% Other non-structural components 35% Total59% Ratios presented in Tables 8-1 and 8-2 are us ed to determine the total cost ratio of physical damage. If 3 of 15 windows are dama ged, for example, a value of 20% is multiplied by the replacement ratio for window s. Additionally, equa tions currently being developed by research partners at the Florida Institute of Technology (FIT) relate structural and non-structural damages in te rms of percentages, which can then be multiplied by the cost ratios. A sum of the physical damage percentage times the cost ratio for each item listed in Tables 81 and 8-2 provides a building component

PAGE 181

161 replacement ratio conditional upon the damage state obtained from the structural prediction model as well as th e type of home (e.g. wood fram e home in North Florida). In addition to the building damage, the cost of contents is factored into the total cost ratio. Research partners at the FI T are currently developing a framework of equations to relate the structural damages presented in Chapter 7 to insurable contents loss. Preliminary results for the content loss estimation portion of this framework are provided in Figure 8-1, where damages to roof cover, roof sheathing, and openings have been interpreted as a Building Damage Ratio which is used to predict a ratio of loss of insurable contents [61]. Complete details concerning the prediction of damage to nonstructural components will be available wh en the PLHP Model is released in 2005. Figure 8-1. Preliminary results of the relati on of structural damage to insurable content loss compared with insurance claims data from Hurricane Andrew. The content loss ratios show n in Figure 8-1 are used in combination with the building component replacement ratios to develop the damage ratio (DR) for each simulated building. From this database, a vulne rability matrix of damage ratios vs. wind

PAGE 182

162 speeds for each type of structure is obtain ed, where each cell provi des the probability of occurrence of a damage ratio conditional upon the wind speed. Using this format, the matrices represent discretized conditional pr obability distribution functions. By summing the product of the likelih ood of occurrence and the DR for all possible DRs, the vulnerability of each type of home is descri bed. This process is shown in Equation 8-1, where Vulnerability(type m | V) is obtained in terms of a percentage of value conditional upon the building type, shown as type m, and the wind speed, V; DRi is a particular damage ratio; and ) | (m itype V DR P is the likelihood of occurrence of DRi conditional upon V and the type of building. Conditional vulnerability curves produced using Equation 8-1 are the product of the cost estimate model. Vulnerability(type m | V)= P (iDRi | V typem)* DRi (8-1) Insured Loss Model Insured losses for typical Florida structur es are obtained using the cost estimate model discussed in the preceding section in combination with wind speed data provided by research partners. This information must then be filtered usi ng knowledge of typical insurance practices, such as deductible a nd limits. The result is the prediction of insurance risk by zip code on an annualized basis. Within each zip code, the probability dens ity function of the largest yearly wind speed, v, will be defined by the PL HP meteorology team as P(v). It is assumed that the probability of occurrence of particular storms within a specified interval can be defined by Equation 8-2, where Vi is a particular 3-second gust wind speed as discussed in the preceding chapters for the structural damage simulation model and v is the increment of 5 mph in terms of 3-second gust wind speeds.

PAGE 183

163 v v P v V v v V Pi i ) ( ) 2 2 ( (8-2) The mean annual damage equation for a particular structure of type m can then be obtained as the sum of the conditional vulnera bility defined in Equation 8-1 times the likelihood of wind speed occurrence provided in Equation 8-2, summed over all possible wind speeds. This expression is provided in Equation 8-3. Annual_Mean_Damage type m= iVulnerability(type m | V) ) 2 2 ( v V v v V Pi i (8-3) Furthermore, the mean annual damage for a geographic area or a portfolio of homes can be obtained by summing the value of Annual_Mean_Damage type m times the probability of the home being of type m over all possible building types in the given area or insurance portfolio. A statistical analysis of the Flor ida building stock described in Chapter 3 and documented by Pinelli and Zhang [31, 32] provi des the regional likelihood of occurrence of each of the building types as P(type i) Using this information, the expression for the mean annual damage for a geographic area or portfolio is described in Equation 8-4. Annual_Mean_Damage = iAnnual_Mean_Damage type i P(type i) (8-4) Since the cost estimate model described in the preceding section of this chapter focuses on the cost as a ratio of the value of the home, the result obtained from Equation 8-4 will be in the form of a percentage. Th is mean annual damage figure is multiplied to the value of each home in a geographic area or insurance portfolio to obtain a monetary value per home. Using this information, the insured loss func tion is obtained by truncating the distribu tion of monetary losses accordi ng to insurance policy deductibles

PAGE 184

164 and limits. A complete discussion of this pr ocess will be available in 2005, with the release of the PLHP Model. Research Contributions As a multi-university project encompassing a variety of academic fields, the PLHP represents a synthesis of work conducted by se veral researchers. The specific research contributions of the author include buildi ng classification efforts, building component modeling, and conceptual de velopment of the damage-pre diction model, to include structural wind load analysis and limit state definition. Research partners at FIT conducted the building classificatio n study described in Chapter 3, with assistance from the author Specifically, the author was directly responsible for the investiga tion of post-damage reports to determine which building components were susceptible to wind damage, and therefore the most critical to model. Additionally, the author was instrumental in the selection of residential characteristics researched during the classification study. The author is responsible for the concep tual development of the probabilistic framework for the determination of structur al wind damage. The selected network of embedded loops to predict structural damage at various storm intensities for buildings typical of Florida residences is sim ilar to the model developed for the HAZUS project, in that it is a component-based damage-predi ction model. The method varies significantly from the HAZUS method however, in the applicati on of wind loads and tallying of structural damage. A time steppi ng routine is used in the HAZUS damage-prediction model, during which individually simulated hu rricane events are passed over a generated home in a series of fifteen-minute interval s. At each interval the wind field model

PAGE 185

165 defines the wind speed and di rection of action for used in wind load calculations. Additionally, the extent of damage alrea dy suffered by the house is used in the determination of the internal pressure at each time step [3]. The results of this resourceintensive time stepping model are vulnerability curves for typical residential and commercial structures. In a fa st running model, the devel oped vulnerability curves are used to predict insurable loss [3]. The resources necessary to develop a similar time stepping model for the determination of reside ntial structural vulne rability are beyond the scope of the current work. Instead, the component-based probabilistic framework described in Chapter 6 has been developed by the author to model the performance of typical Florida homes during extreme wind even ts. Characterization of the applied wind loads described in Chapter 4 and the typical build ing component resistances described in Chapter 5 as well as decisions concerning load placement, load sharing, applicable limit states, and the inclusion of load paths with in the failure check se quence are contributed by the author. Future Uses of the Structural Damage Model The main goal of the research presented in this dissertation is the relation of specific wind speeds to predicted damages for typical residential buildings in the state of Florida. The simulation engine detailed in this body or work is specifically intended for the development of the PLHP Model, a multiuniversity project sponsored by the Florida Department of Financial Services and c oordinated by the International Hurricane Research Center. The MatLAB based Monte Carlo Simulation (MCS) engine has been created with a capacity for use in future pr ojects, however. Because the model is based on the component approach, it can be refined as the understanding of the complex interaction between hurricane wi nds and structures increases. With additional detail, the

PAGE 186

166 model can also be used to quantify specific hurricane damage mitigation strategies. Furthermore, the MCS engine created for this project can serve as the spring board for a future online learning laboratory to serve as an academic tool for undergraduate civil engineers, or as a public service to home owners in Florida an d other hurricane prone regions of the United States.

PAGE 187

167 APPENDIX A SOUTH / KEYS REGION CONCRETE BL OCK GABLE ROOF (CBG) HOMES This appendix contains simulated struct ural damage to typical concrete block homes in the South Florida and Florida Ke ys Region with gable roofs. Figure A-1 provides a measure of comparison between diffe rent building components. In this figure, the mean damages to roof cover, roof sheathing, and roof-towall connections are presented as a percentage of the total that is damaged. Wall damage is also presented as a percentage, with respect to the total of four walls, and as a percentage of damaged panels on gable ends. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceeding defined damage st ates) for individual building components. In a few cases, the fragility curves for si mulated damages conflict with engineering judgment. One would expect damages to increase with wind speed, yet the damages to exterior doors (in Figure A-13) decrease after reaching 200 mph 3-second gust wind speeds, This anomaly is a function of the na ture of the simulation routine. The damageprediction engine simulates damages that would occur as a result of th e entire storm from one snapshot of wind speed in time. It does not step through the en tire duration of the storm, accumulating damages as the wind speed increases. For this reason, and due to internal pressure effects in the model, th e damages to exterior doors decrease after 200 mph 3-second gusts. The monetary value of da mage at the point at which the drop in damage to these two component s occurs is already substan tial; thus these unusual results will not adversely affect the end product of insurable loss.

PAGE 188

168 Figure A-1. Concrete block gable roof South/Keys Region home comparative levels of roof cover, roof sheathing, connect ions, wall, and gable end sheathing damage. Figure A-2. Vulnerability to roof c over damage for South/Keys CBG homes.

PAGE 189

169 Figure A-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for South/Keys CBG homes. Figure A-4. Vulnerability to roof shea thing damage for South/Keys CBG homes.

PAGE 190

170 Figure A-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for South/Keys CBG homes. Figure A-6. Vulnerability to roof-to-wall connection damage for South/Keys CBG homes.

PAGE 191

171 Figure A-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region CBG homes. Figure A-8. Vulnerability to wall da mage for South/Keys Region CBG homes.

PAGE 192

172 Figure A-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region CBG homes. Figure A-10. Vulnerability to window da mage for South/Keys Region CBG homes.

PAGE 193

173 Figure A-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region CBG homes. Figure A-12. Vulnerability to exterior door damage for South/Keys Region CBG homes.

PAGE 194

174 Figure A-13. Fragility curves for 1 and 2 da maged exterior doors for South/Keys Region CBG homes. Figure A-14. Vulnerability to garage door damage for South/Keys Region CBG homes.

PAGE 195

175 APPENDIX B SOUTH / KEYS REGION CONCRETE BLOCK HIP ROOF (CBH) HOMES This appendix contains figures showing simulated structural damage to typical concrete block homes in the South Florida and Florida Keys Region with hip roofs. Figure B-1 provides a measure of comparison between different building components. In this figure, the mean damages to roof cover, roof sheathing, and r oof-to-wall connections are presented as a percentage of the total that is damaged. Wall damage is also presented as a percentage, with respect to the total of four walls. Additiona l figures provide the vulnerability (mean damage) or fragility (probability of exceeding defined damage states) for individual building components. In a few cases, the fragility curves for si mulated damages conflict with engineering judgment. One would expect damages to increase with wind speed, yet the damages to exterior doors (in Figure B-13) decrease after reaching 200 mph 3-second gust wind speeds, This anomaly is a function of the na ture of the simulation routine. The damageprediction engine simulates damages that would occur as a result of th e entire storm from one snapshot of wind speed in time. It does not step through the en tire duration of the storm, accumulating damages as the wind speed increases. For this reason, and due to internal pressure effects in the model, th e damages to exterior doors decrease after 200 mph 3-second gusts. The monetary value of da mage at the point at which the drop in damage to these two component s occurs is already substan tial; thus these unusual results will not adversely affect the end product of insurable loss.

PAGE 196

176 Figure B-1. Concrete block hi p roof South/Keys Region home comparative levels of roof cover, roof sheathing, connections, wa ll, and gable end sheathing damage. Figure B-2. Vulnerability to roof c over damage for South/Keys CBH homes.

PAGE 197

177 Figure B-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for South/Keys CBH homes. Figure B-4. Vulnerability to roof shea thing damage for South/Keys CBH homes.

PAGE 198

178 Figure B-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for South/Keys CBH homes. Figure B-6. Vulnerability to roof-to-wa ll connection damage for South/Keys CBH homes.

PAGE 199

179 Figure B-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region CBH homes. Figure B-8. Vulnerability to wall da mage for South/Keys Region CBH homes.

PAGE 200

180 Figure B-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region CBH homes. Figure B-10. Vulnerability to window da mage for South/Keys Region CBH homes.

PAGE 201

181 Figure B-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region CBH homes. Figure B-12. Vulnerability to exterior door damage for South/Keys Region CBH homes.

PAGE 202

182 Figure B-13. Fragility curves for 1 and 2 da maged exterior doors for South/Keys Region CBH homes. Figure B-14. Vulnerability to garage door damage for S outh/Keys Region CBH homes.

PAGE 203

183 APPENDIX C SOUTH / KEYS REGION WOOD FR AME GABLE ROOF (WG) HOMES This appendix contains simulated struct ural damage to typical wood frame homes in the South Florida and Florida Keys Re gion with hip roofs. Figure C-1 provides a measure of comparison between different build ing components. In this figure, the mean damages to roof cover, roof sheathing, and roof-to-wall connections are presented as a percentage of the total that is damaged. Wall damage is also presented as a percentage, with respect to the total of four walls, and as a percentage of sheathing panels lost on gable ends. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceeding defined damage st ates) for individual building components. In a few cases, the fragility curves for si mulated damages conflict with engineering judgment. One would expect damages to increase with wind speed, yet the damages to exterior doors (in Figure C-13) decrease after reaching 200 mph 3-second gust wind speeds, This anomaly is a function of the na ture of the simulation routine. The damageprediction engine simulates damages that would occur as a result of th e entire storm from one snapshot of wind speed in time. It does not step through the en tire duration of the storm, accumulating damages as the wind speed increases. For this reason, and due to internal pressure effects in the model, th e damages to exterior doors decrease after 200 mph 3-second gusts. The monetary value of da mage at the point at which the drop in damage to these two component s occurs is already substan tial; thus these unusual results will not adversely affect the end product of insurable loss.

PAGE 204

184 Figure C-1. Wood frame gable roof South/Keys Region home co mparative levels of roof cover, roof sheathing, connections, wa ll, and gable end sheathing damage. Figure C-2. Vulnerability to roof c over damage for South/Keys WG homes.

PAGE 205

185 Figure C-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for South/Keys WG homes. Figure C-4. Vulnerability to roof shea thing damage for South/Keys WG homes.

PAGE 206

186 Figure C-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for South/Keys WG homes. Figure C-6. Vulnerability to roof-to-wall c onnection damage for South/Keys WG homes.

PAGE 207

187 Figure C-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region WG homes. Figure C-8. Vulnerability to wall dama ge for South/Keys Region WG homes.

PAGE 208

188 Figure C-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WG homes. Figure C-10. Vulnerability to window da mage for South/Keys Region WG homes.

PAGE 209

189 Figure C-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region WG homes. Figure C-12. Vulnerability to exterior door damage for South/Keys Region WG homes.

PAGE 210

190 Figure C-13. Vulnerability to exterior door damage for South/Keys Region WG homes. Figure C-14. Vulnerability to garage door damage for S outh/Keys Region WG homes.

PAGE 211

191 APPENDIX D SOUTH / KEYS REGION WOOD FRAME HIP ROOF (WH) HOMES This appendix contains figures showing simulated structural damage to typical wood frame homes in the South Florida and Fl orida Keys Region with hip roofs. Figure D-1 provides a measure of comparison between different building components. In this figure, the mean damages to roof cover, roof sheathing, and roof-to-wall connections are presented as a percentage of the total that is damaged. Wall damage is also presented as a percentage, with respect to the total of four walls. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceeding defined damage states) for individual building components. In a few cases, the fragility curves for si mulated damages conflict with engineering judgment. One would expect damages to increase with wind speed, yet the damages to exterior doors (in Figure C-13) decrease after reaching 200 mph 3-second gust wind speeds, This anomaly is a function of the na ture of the simulation routine. The damageprediction engine simulates damages that would occur as a result of th e entire storm from one snapshot of wind speed in time. It does not step through the en tire duration of the storm, accumulating damages as the wind speed increases. For this reason, and due to internal pressure effects in the model, th e damages to exterior doors decrease after 200 mph 3-second gusts. The monetary value of da mage at the point at which the drop in damage to these two component s occurs is already substan tial; thus these unusual results will not adversely affect the end product of insurable loss.

PAGE 212

192 Figure D-1. Wood frame hip r oof South/Keys Region home co mparative levels of roof cover, roof sheathing, connections, wa ll, and gable end sheathing damage. Figure D-2. Vulnerability to roof c over damage for South/Keys WH homes.

PAGE 213

193 Figure D-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for South/Keys WH homes. Figure D-4. Vulnerability to roof shea thing damage for South/Keys WH homes.

PAGE 214

194 Figure D-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for South/Keys WH homes. Figure D-6. Vulnerability to roof-to-wall c onnection damage for South/Keys WH homes.

PAGE 215

195 Figure D-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for South/Ke ys Region WH homes. Figure D-8. Vulnerability to wall dama ge for South/Keys Region WH homes.

PAGE 216

196 Figure D-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WH homes. Figure D-10. Vulnerability to window da mage for South/Keys Region WH homes.

PAGE 217

197 Figure D-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys Region WH homes. Figure D-12. Vulnerability to exterior door damage for South/Keys Region WH homes.

PAGE 218

198 Figure D-13. Vulnerability to exterior door damage for South/Keys Region WH homes. Figure D-14. Vulnerability to garage door damage for South/Keys Region WH homes.

PAGE 219

199 APPENDIX E FLORIDA MANUFACTURED SINGLEWIDE HOMES This appendix contains figures representi ng simulated structural damage to typical singlewide manufactured homes in the State of Florida. Figure E-1 provides a measure of comparison between different bu ilding components. In this figure, the mean damages to roof cover, roof sheathing, roof-to-wall c onnections, and wall sheathing panels are presented as a percentage of the total that is damaged. For overturning, a percentage representing the likelihood of the home being overturned is provided. In the simulation model a value of one is recorded for overt urned homes and a value of zero represents homes that do not overturn. At each wind sp eed, the mean overturning value is multiplied by 100 to obtain a comparative percent. A simila r process is used for sliding, except that there are two possible sliding ca tegories. In the simulated data, a value of one indicates minor sliding, while a value of two represents major sliding. Zero is used for no sliding damage. For each wind speed, the mean sliding value is multiplied by 100 and divided by 2 to obtain a comparative percent. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceedi ng defined damage states) for individual building components.

PAGE 220

200 Figure E-1. Singlewide manuf actured home comparative leve ls of roof cover, roof sheathing, connections, wall, a nd gable end sheathing damage. Figure E-2. Vulnerability to roof cover damage for singlewide manufactured homes.

PAGE 221

201 Figure E-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for singlewide manufactured homes. Figure E-4. Vulnerability to roof sheathi ng damage for singlewide manufactured homes.

PAGE 222

202 Figure E-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for singlewide manufactured homes. Figure E-6. Vulnerability to roof-t o-wall connection damage for singlewide manufactured homes.

PAGE 223

203 Figure E-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for singlewide manufactured homes. Figure E-8. Vulnerability to wall sheathi ng damage for singlewide manufactured homes.

PAGE 224

204 Figure E-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for singlewide manufactured homes.

PAGE 225

205 APPENDIX F FLORIDA MANUFACTURED DOUBLEWIDE HOMES This appendix contains figures representi ng simulated structural damage to typical doublewide manufactured homes in the State of Florida. Figure F-1 provides a measure of comparison between different building com ponents. In this figure, the mean damages to roof cover, roof sheathing, roof-to-wa ll connections, and wall sheathing panels are presented as a percentage of the total that is damaged. For overturning, a percentage representing the likelihood of the home being overturned is provided. In the simulation model a value of one is recorded for overt urned homes and a value of zero represents homes that do not overturn. At each wind sp eed, the mean overturning value is multiplied by 100 to obtain a comparative percent. A simila r process is used for sliding, except that there are two possible sliding ca tegories. In the simulated data, a value of one indicates minor sliding, while a value of two represents major sliding. Zero is used for no sliding damage. For each wind speed, the mean sliding value is multiplied by 100 and divided by 2 to obtain a comparative percent. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceedi ng defined damage states) for individual building components.

PAGE 226

206 Figure F-1. Doublewide manufactured home co mparative levels of roof cover, roof sheathing, connections, wall, a nd gable end sheathing damage. Figure F-2. Vulnerability to roof cove r damage for doublewide manufactured homes.

PAGE 227

207 Figure F-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for doublewide manufactured homes. Figure F-4. Vulnerability to roof sheathi ng damage for doublewide manufactured homes.

PAGE 228

208 Figure F-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for doublewide manufactured homes. Figure F-6. Vulnerability to roof-t o-wall connection damage for doublewide manufactured homes.

PAGE 229

209 Figure F-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for doublewide manufactured homes. Figure F-8. Vulnerability to wall sheathi ng damage for doublewide manufactured homes.

PAGE 230

210 Figure F-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for doublewide manufactured homes.

PAGE 231

211 APPENDIX G FLORIDA PRE-HUD CODE MANUFACTURED HOMES This appendix contains figures representi ng simulated structural damage to typical manufactured homes in the State of Florid a that pre-date the 1975 changes to the manufactured home building code. Figur e G-1 provides a measure of comparison between different building compone nts. In this figure, the mean damages to roof cover, roof sheathing, roof-to-wall connections, and wall sheathing panels are presented as a percentage of the total that is damaged. For overturning, a percentage representing the likelihood of the home being overturned is prov ided. In the simulation model a value of one is recorded for overturned homes and a value of zero represents homes that do not overturn. At each wind speed, the mean overturning value is multiplied by 100 to obtain a comparative percent. A similar process is used for sliding, except that there are two possible sliding categories. In the simulated data, a value of one indicates minor sliding, while a value of two represents major slidi ng. Zero is used for no sliding damage. For each wind speed, the mean sliding value is multiplied by 100 and divided by 2 to obtain a comparative percent. Additional figures provide the vulnerability (mean damage) or fragility (probability of exceeding define d damage states) for individual building components.

PAGE 232

212 Figure G-1. Pre-HUD Code manufactured home comparative levels of roof cover, roof sheathing, connections, wall, a nd gable end sheathing damage. Figure G-2. Vulnerability to roof cover da mage for pre-HUD Code manufactured homes.

PAGE 233

213 Figure G-3. Fragility curves for 2%, 5%, 10% 25%, and 50% damage to roof cover for pre-HUD Code manufactured homes. Figure G-4. Vulnerability to roof sheath ing damage for pre-HUD Code manufactured homes.

PAGE 234

214 Figure G-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing for pre-HUD Code manufactured homes. Figure G-6. Vulnerability to roof-towall connection damage for pre-HUD Code manufactured homes.

PAGE 235

215 Figure G-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall connections for pre-HUD Code manufactured homes. Figure G-8. Vulnerability to wall sheath ing damage for pre-HUD Code manufactured homes.

PAGE 236

216 Figure G-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing for pre-HUD Code manufactured homes.

PAGE 237

217 LIST OF REFERENCES 1. S. L. McCabe, Testimony of Dr. Steven L. McCabe on behalf of the American Society of Civil Engineers before th e subcommittee on environment, technology and standards of the committee on science, U.S. House of Representatives, October 11, 2001. 2. J. E. Minor, P. J. Schneider, Hurrica ne loss estimation – the HAZUS preview model, Proceedings of the America’s C onference on Wind Engineering, Clemson, SC, 2001, 572–578. 3. Multi-hazard Loss Estimation Methodology Hurricane Model HAZUSMH Technical Manual, Federal Emer gency Management Agency, 2003. 4. F. Lavelle, P. Vickery, B. Schauer, L. Twisdale, E. Latch, The HAZUS hurricane model, Proceedings of the 11th International Conference on Wind Engineering, Lubbock, TX, 2003, 1015–1022. 5. C. Dyrbye, S. Hansen, Wind Loads on Structures, John Wiley & Sons, New York, NY, 1997. 6. E. Simiu, R. Scanlan, Wind Effects on St ructures, Fundamentals and Applications to Design, Third Edition, John Wiley & Sons, New York, NY,1996. 7. ASCE 7-98 Standard, Minimum Design Load s for Buildings and Other Structures, American Society of Civil Engineers, New York, NY. 8. B. Munson, D. Young, T. Okiishi, Fundamentals of Fluid Mechanics, John Wiley & Sons, New York, NY, 1990. 9. A. Rigato, P. Chang, E. Simiu, Database -assisted design, standardization, and wind direction effects, J. Struct. Eng., ASCE 127 (8) (2001) 855–860. 10. F. Sadek, E. Simiu, Peak non-Gaussian wi nd effects for database-assisted low rise building design, J. Eng. Mech., ASCE 128 (5) (2002) 530–539. 11. A. Kareem, Wind effects on structures: a probabilistic viewpoint, Probab. Eng. Mech., 2 (4) (1987) 166–200. 12. A. Cope, K. Gurley, Spatial characteristic s of pressure coefficients on low rise gable roof structures, Proceedings of the America’s Conference on Wind Engineering, Clemson, SC, 2001, 719–728.

PAGE 238

218 13. M. Gioffre, K. Gurley, A. Cope, Stochast ic simulation of correlated wind pressure fields on low-rise gabl e roof structures, 15th ASCE Engineering Mechanics Conference, New York, NY, 2002. 14. T. Cunningham, Roof Sheathing Fastening Schedules for Wind Uplift, APA Report T92-28, American Plywood Associ ation, Tacoma, WA, March 1993. 15. R. Cook, M. Soltani, editors, Hurri canes of 1992: Lessons Learned and Implications for the Future, ASCE, New York, NY, 1994. 16. Building Performance: Hurricane Andrew in Florida Observations, Recommendations, and Technical Gu idance, FEMA, Federal Insurance Administration. 17. A. Kareem, Structural performance and wind speed-damage correlation in Hurricane Alicia, J. Struct Eng., ASCE 111 (12) 2596–2610. 18. A. Kareem, Performance of cladding in Hu rricane Alicia, J. Struct. Eng., ASCE 112 (12) 2679–2693. 19. NAHB Research Center, Assessment of Da mage to Single-Family Homes Caused by Hurricanes Andrew and Iniki, Prepared for U.S. Department of Housing and Urban Development, Office of Policy Development and Research, 1993. 20. NAHB Research Center, Assessment of Damage to Homes caused by Hurricane Opal, Prepared for Florida State Home Builders Association, 1996. 21. NAHB Research Center, Reliabi lity of Conventional Resi dential Construction: An assessment of Roof Component Perform ance in Hurricane Andrew and Typical Wind Regions of the United States, Prepared for U.S. Department of Housing and Urban Development and National Asso ciation of Home Builders, 1999. 22. M. Phang, Wind damage investigation of low rise buildings, Proceedings of the 1999 Structures Congress, ASCE, New Orleans, LA, 1999, 1015–1021. 23. B. Sill, P. Sparks, editors, Hurricane H ugo One Year Later, Proceedings of the Symposium and Public Forum, ASCE, September 1990. 24. M. Mahendran, Wind resistant lo w-rise buildings in the tr opics, J. Perform. Constr. Fac., ASCE 9 (11) (1995) 330–346. 25. Y. Mitsuta, T. Fujii, I. Nagashima, A predicting method of typhoon wind damages, Probabilistic Mechanics and Structural Reli ability: Proceedings of the 7th Specialty Conference, Worcester, MA, 1996, 970–973. 26. S. Bhinderwala, Insurance Loss Analysis of Single Family Dwellings Damaged in Hurricane Andrew, Master’s Thesis, Clemson University, Clemson, SC, Department of Civil Engineering, 1995.

PAGE 239

219 27. J. Holmes, Vulnerability curves for buildings in tropical cyclone regions, Probabilistic Mechanics and Structural Reli ability: Proceedings of the 7th Specialty Conference, Worcester, MA, 1996, 78–81. 28. B. Sill, R. Kozlowski, Analysis of storm damage factors for low-rise structures, J. Perform. Constr. Fac., ASCE 11 (4) (1997) 168–176. 29. Z. Huang, D. Rosowsky, P. Sparks, Ev ent-based hurricane simulation for the evaluation of wind speeds and expected in surance loss, Wind Engineering into the 21st Century, 1999, 1417–1424. 30. J. Sciaudone, D, Freuerborn, G. Rao, S. Daneshvaran, Development of objective wind damage functions to predict wind damage to low-rise structures, Eighth U.S. National Conference on Wind Engineering, Johns Hopkins University, Baltimore, MD, 1997. 31. J-P. Pinelli, L. Zhang, C. Subramanian, A. Cope, K. Gurley, S. Gulati, S. Hamid, Classification of structural models fo r wind damage predictions in Florida, Proceedings of the 11th International Conference on Wind Engineering, Lubbock, TX, June, 2003, 999–1006. 32. L. Zhang, Public Hurricane Loss Proj ection Model: Exposure and Vulnerability Components, Master’s Thesis, Florida Institute of Technol ogy, Melbourne, FL, Department of Civil Engineering, 2003. 33. NAHB Research Center, Factory and Site Built Housing, A Comparison for the 21st Century, Prepared for U.S. Department of Housing and Urban Development, October 1998. 34. J. Wills, B. Lee, T. Wyatt, A model of wind-borne debris damage, J. Wind Eng. Ind. Aerod., 90 (4/5) (2002) 555–565. 35. J. Peterka, J. Cermak, L. Cochran, B. Cochran, N. Hosoya, R. Derickson, C. Harper, J. Jones, B. Metz, Wind uplift mode l for asphalt shingles, J. Archit. Eng., ASCE 3 (4) (1997) 147–155. 36. FM Global Technologies, 2002, Approval Sta ndard for Class 1 Roof Covers, FM Global Technologies, http://www.fmgloba l.com/approvals/resources/approval standards/4470.pdf (retr ieved February 2004). 37. Industry Perspective: Impact Resistan ce Standards, Natural Hazard Mitigation Insights, Institute for Business and Home Safety, (12), February 2000. 38. A. Ang, W. Tang, Probability Concepts in Engineering Planning and Design, John Wiley & Sons, New York, NY, 1975. 39. D. Mizzell, Wind Resistance of Sheathing fo r Residential Roofs, Master’s Thesis, Clemson University, Clemson, SC, Depa rtment of Civil Engineering, 1994.

PAGE 240

220 40. T. Cunningham, Roof Sheathing Fasten ing Schedules for Wind Uplift, APA Report T92-28, American Plywood Asso ciation, Tacoma, WA, March 1993. 41. D. Rosowsky, S. Schiff, T. Reinhold, P. Sparks, B. Sill, Performance of LightFrame Wood Structures Under High Wind Loads: Experimental and Analytical Program, Wind Performance and Safety of Wood Buildings, FPS Specialty Publication, 1998. 42. T. Reinhold, 13 homes destroyed, Disaster Review, Fall 2002, 9–14. 43. L. Canfield, S. Niu, H. Liu, Uplift resi stance of various rafter-wall connections, Forest Prod. J., 41 (7/8) (1991) 27–34. 44. T. Reed, D. Rosowsky, S. Schiff, Structur al Analysis of Ligh t-Framed Wood Roof Construction, a Wind Load Test Facility Report for Blue Sky, PBS-9606-02, 1996. 45. National Design Specification for Wood C onstruction: Allowable Stress Design (ASD) Manual for Engineered Wood C onstruction, American Wood Council, Washington D.C., 1997. 46. J-P. Pinelli, S. O’Neill, Effect of torn adoes on residential masonry structures, Wind Struct. J., 3 (1) (2000) 23–40. 47. J. Dawe, G. Aridru, Prestressed concrete masonry walls subjected to uniform outof-plane loading, Can. J. Civil Eng., 20 (1993) 969–979. 48. Florida Building Code, Tallahassee, FL, 2001. 49. ACI 530-99/ASCE 5-99/TMS 402-99: Buildi ng Code Requirements for Masonry Structures, American Concrete Inst itute, Farmington Hills, MI, 1999. 50. DASMA, 2002, DASMA Garage Door and Commercial Door Wind Load Guide, Technical Data Sheet #155b, Door & A ccess Systems Manufacturer’s Association International, http://www.dasma.co m/PDF/Publications/TechDataSheets/ CommercialResidential/TDS155b. pdf (retrieved March 2003). 51. Simpson Strongtie, 2002, Connectors for F actory Built Homes, Technical Bulletin T-FBS02, Simpson Strongtie, http://www. strongtie.com/ftp/bulletins/T-FBS02.pdf (retrieved August 2003). 52. Owens Corning, 2001, Certificate of Conf ormance, Owens Corning Select Vinyl Siding, Owens Corning, http://www.ow enscorning.com/around/exteriors_new/ pdfs/COCOwensSelect.pdf (retrieved August 2003). 53. Certainteed, 2000, Pro Edition™ 44 Vinyl Siding, Certainteed: http://certainteed.com/NR/r donlyres/E72BFEB7-459C-488D-86F2BFC0955E8EBF/0/410.pdf (retrieved August 2003).

PAGE 241

221 54. Certainteed, 2001, Weather Happens : Test Results, Certainteed, http://www.certainteed.com/cside/csc t01208rig.html (retrieved August 2003). 55. R. Marshall, Manufactured Homes – Probabil ity of Failure and the Need for Better Windstorm Protection Through Improved Anchoring Systems, NISTIR 5370, Building and Fire Research Laboratory, Gaithersburg, MD, for Department of Housing and Urban Development, Washington, D.C., November 1994. 56. F. Yokel, R. Chung, F. Rankin, C. Yan cey, Load-Displacement Characteristics of Shallow Soil Anchors, NBS Building Sc ience Series 142, National Bureau of Standards, Washington D.C., 1982. 57. K. Hayes, Hurricane Data Collection Hard ware: Design, Construction, and Testing, Master’s Thesis, University of Florida, Gainesville, FL, Department of Civil and Coastal Engineering, 2000. 58. Carbide Depot, (undated), Coefficient for Static Friction of Steel Chart, Carbide Depot, http://www.carbidedepot.com/formu las-frictioncoefficient.htm (retrieved September 2003). 59. M. Powell, S. Houston, T. Reinhold, Hu rricane Andrew’s landfa ll in South Florida Part I: standardizing measurements for documentation of surface wind fields, Weather Forecast., 11 (1996) 304–328. 60. J-P. Pinelli, J. Murphree, K. Gurley, A. Cope, S. Hamid, S. Gulati, Hurricane loss estimation, International Conference on Probabilistic Safety Assessment and Management, Berlin, Germany, June, 2004. 61. A. Cope, K. Gurley, J-P. Pinelli, J. Mur phree, S. Gulati, S. Hamid, A probabilistic model of damage to residential structures from hurricane winds, Joint Specialty Conference on Probabilistic Mechanics and Structural Reliability, Albuquerque, NM, July, 2004.

PAGE 242

222 BIOGRAPHICAL SKETCH Anne grew up in Winter Haven, Florida, where she graduated second in her class from Winter Haven High School. As a Nati onal Merit Finalist, she was awarded the William A. Kenyon Scholarship from Clemson Un iversity. During her educational career at Clemson, Anne was a two-time recipient of the E. L. Clarke Award (the Civil Engineering departmental award for academic excellence). As an undergraduate, she completed a 9-month Eisenhower Fellow experience with the Department of Transportation, and participated in th e National Science Foundation’s Research Experience for Undergraduates Program. Anne graduated summa cum laude with senior departmental honors, earning a Bachelor of Science in civil engi neering May 12, 1995. She then worked as a research fellow unde r the direction of Dr. Tim Reinhold, at Clemson, where she conducted wind-tunnel research on low-rise structures. She earned a Master of Science in civi l engineering on August 9, 1997, with a thesis titled Load Duration Effects on Peak Minimum Pressure Coefficients After completing degree requirements, Anne was commissioned as a se cond lieutenant in the United States Army. She spent 3 years on active duty as an officer and paratrooper at Fort Bragg, NC, before pursuing a Ph.D. at the University of Flor ida. She was awarded the University of Florida’s Alumni Scholarship, and studied unde r the direction of Dr. Kurt Gurley. Her primary research interests lie in the area of wind-damage mitigation.


Permanent Link: http://ufdc.ufl.edu/UFE0005820/00001

Material Information

Title: Predicting the Vulnerability of Typical Residential Buildings to Hurricane Damage
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0005820:00001

Permanent Link: http://ufdc.ufl.edu/UFE0005820/00001

Material Information

Title: Predicting the Vulnerability of Typical Residential Buildings to Hurricane Damage
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0005820:00001


This item has the following downloads:


Full Text












PREDICTING THE VULNERABILITY OF TYPICAL
RESIDENTIAL BUILDINGS TO HURRICANE DAMAGE

















By

ANNE D. COPE


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


2004


































Copyright 2004

by

Anne D. Cope
































This work is dedicated to those who have given their lives in the defense of our freedom.















ACKNOWLEDGMENTS

I would like to extend sincere thanks to the many people who made this

accomplishment possible. First, I would like to thank my advisor, Dr. Kurt Gurley for his

encouragement, support, and guidance, especially when recent world political events

became very personal. I would also like to thank Dr. Jean-Paul Pinelli for his leadership

of the engineering team for the Public Loss Hurricane Projection Model. For sincere

critique and professional advice, I would like to thank Dr. Emil Simiu, Dr. Tim Reinhold,

and Dr. Peter Vickery. Many thanks also go to the members of my committee and to

fellow researchers (especially Liang Zhang, Luis Aponte, and Josh Murphree). I thank

the providers of the University of Florida Alumni Scholarship for financial support of my

education, and the Florida Department of Insurance for funding this research. For their

unwavering support and loving advice, I thank my husband, my parents, and my extended

family. Lastly, I would like to thank Adrianne Pickett for her support and warm

hospitality during the completion of this project.
















TABLE OF CONTENTS
Page

A C K N O W L E D G M E N T S ................................................................................................. iv

L IST O F T A B L E S ........................................................................... ........... ix

LIST OF FIGURES ......... ........................................... ............ xi

ABSTRACT .............. .......................................... xix

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

R research H ypothesis............ ... .............................................................. ........ ........
G oals and Objectives .................. ................................... ................ .5
Sum m ary of D issertation ......... .......................................................... .................. 5

2 SUMMARY OF PREVIOUS RESEARCH ......................... .................................7

Background Information on Structural Wind Loads ..........................................7
Efforts to Quantify Extrem e W ind Loads ................................................................ 12
Defining the Behavior of Near-Surface Hurricane Winds ................................12
Characterizing Surface Pressures on Structures ...............................................14
Characterizing and Codifying Structural Loads...............................................17
Summary of Efforts to Quantify Extreme Wind Loads............... ...................17
Post-D am age Investigations ............................................... ............................ 18
D am age Prediction M odels................. ................................ .................. 20
Fundamental Concepts in Damage Prediction ......................................... 20
Damage Prediction Models in the Public Domain ...........................................24
Proprietary Damage Prediction M odels ................................... .................26
Public Loss Hurricane Projection Model .........................................................28

3 RESIDENTIAL STRUCTURES IN FLORIDA ................ .............................. 30

Sources of Inform ation ..................... .... .. .................................. .............. 1
Florida Hurricane Catastrophe Fund Exposure Database ...................................31
County Property Appraiser Databases ............... .....................................32
M manufactured Hom e Builder Literature ................................... .................33
Post-D am age Investigations ........................................ .......................... 34
Results of the Building Population Investigation.....................................................34









Characterization of Site-Built Hom es ...................................... ............... 34
Characterization of Manufactured Homes....................................................40
Building Com ponent Selection....................................... ......................... 41

4 STRUCTURAL WIND LOADS FOR TYPICAL ............................................. 44

Use and Modification of the ASCE 7-98 Code Provisions to Represent Load
Conditions during Extreme W ind Events ................................... .................45
Modifications to Surface Pressure Equations.....................................................46
Use and Modifications to External Pressure Coefficients...............................48
Main Wind Force Resisting System external pressure coefficients............49
Component and Cladding external pressure coefficients.............................50
Use and Modifications to Internal Pressure Coefficients...................................54
Application of the Modified ASCE 7-98 Code Provisions to Produce Extreme
Wind Event Load Conditions on Selected Building Components.........................54
Roof Cover and Roof Sheathing Loads...... .............. ................................55
Roof-to-W all Connection Loads ........................................ ...... ............... 56
W all L o ad s................................................................................. ............... 5 7
Load Conditions for O openings ........................................ ........................ 59
Load Conditions for Tie-Down Anchors................................ ............... .... 65
Summary of Wind Load Conditions Used in the Simulation Engine.........................65

5 PROBABILISTIC WIND RESISTANCE CAPACITIES FOR RESIDENTIAL
DWELLING COMPONENTS ............................................................................67

Fundamental Concepts Applied During the Selection of Load Resistance
V a lu e s .............. ..............................................................6 8
Site-Built H om e R resistance V alues............................................................................71
Wind Resistance Capacity of Roof Cover on Site-Built Homes.........................71
Wind Resistance Capacity of Roof Sheathing on Site-Built Homes .................75
Wind Resistance Capacity of Roof-to-Wall Connections on Site-Built
H om es ................. ......................................... ......................... 76
Wind Resistance Capacity of Site-Built Home Walls......................................79
W ood shear w all capacity ........................................ ........................ 80
W ood frame out-of-plane load capacity.....................................................81
W ood fram e uplift capacity ................................... .................................... 83
W ood fram e sheathing capacity ....................................... ............... 83
M asonry shear w all capacity ................................. ............. .................. 84
Masonry out-of-plane load capacity ....................................................85
M asonry uplift capacity...................... ............................ .................. 86
Wind Resistance Capacity of Site-Built Home Openings ...................................86
Wind resistance capacity of doors for site-built homes ............................87
Wind resistance capacity of garage doors for site-built homes ....................87
Wind resistance capacity of windows for site-built homes ..........................88
Manufactured Home Resistance Values.............................................................89
Wind Resistance Capacity of Roof Sheathing and Cover on Manufactured
H o m e s ..............................................................................................................9 0









Wind Resistance Capacity of Roof-to-Wall Connections for Manufactured
H om es .............................................. ...................... ... ....... 9 1
W all Capacity for M manufactured Hom es................................................... .... 92
Wind Resistance Capacity of Manufactured Home Openings............................93
Wind Resistance Capacity of Tie-Down Anchors.................... ...............93
Summary of Resistance Values Used in Structural Damage Simulation .................94

6 SIM U L A TIO N EN G IN E ................................................................. .....................97

Selection of Structural Type and Definition of Geometry ..........................................97
V ariables for Site-Built H om es ........................................ ....................... 98
Variables for M manufactured Hom es ...................................... ............... 100
Loop for A ngle of Incidence ........................................................ ............... 101
Loop for Wind Speed........................ ....................... ....... 102
L oop for the Sim ulated H om es............................................................ .................. 102
Randomization of Wind Speed and Pressure Coefficients.............................103
Initial Load Calculations ............................................................................105
Sam pling of R esistances................... ............ ................................ ............. 105
Roof cover and roof sheathing resistance sampling................................106
Roof-to-wall connection resistance sampling .........................................107
W all resistance sam pling....................................... ......................... 109
O opening resistance sam pling ......................................................................111
Tie-down anchor resistance sampling .............................. .................... 112
Initial F ailu re C h eck ............................................................... .......... .. .. 112
Initial failure check for roof sheathing ...................................................... 112
Initial failure check for walls ............. ................................ ............... 113
Initial failure check for openings ............................................................115
Internal Pressure Evaluation and Recalculation of Loads............... ...............116
Final Failure Check and D am age Tally...........................................................117
Structural D am age Output Files ........................................ .......................... 122
Sum m ary ............................................................... ..... ..... ......... 123

7 STRUCTURAL DAMAGE VALIDATION AND RESULTS.............................124

Structural D am age V alidation ....................................................... .....................126
NAHB Report on Hurricane Andrew ............................................................ 127
Application of the NAHB Report Data as a Validation Tool............................128
Validation of Individual Components .............. .......................... .................130
V alidation of w indow dam age ......... ............................... ...... ............. 132
Validation of masonry wall damage................................. ................... 135
Validation of wood frame wall damage ..........................................137
Validation of roof-to-wall connection damage ............... ....................138
Validation of roof sheathing damage ................................. ... ................ 140
Validation of roof cover damage............................................ ..........142
Investigation of Selected Topics...................... .................................144
Investigation of the Batch Selection Method for Roof Sheathing................. 144
Investigation of the Batch Selection Method for Roof-to-Wall Connections ...146









Investigation of the Difference between Hip and Gable Roofs .........................148
Structural D am age R results ............................ ... .... ....... .... ................... ....150
Results for Site-Built Homes in the South Florida and Florida Keys Region...151
Results for Manufactured Homes..... .................... ...............153
S u m m a ry ............. ................. ................. .................................................1 5 6

8 APPLICATION OF RESULTS AND CONCLUSION............ .................157

Relating Structural Damage to M monetary Loss ................................ ............... 158
C o st E stim ate M odel ........................................... ........................................ 159
Insured Loss M odel ............ .... ...... ................... ........ .. ........ 162
R research C ontributions.................................................................................... 164
Future Uses of the Structural Damage Model ............................... .................165

APPENDIX

A SOUTH / KEYS REGION CONCRETE BLOCK GABLE ROOF (CBG)
H O M E S .......................................................................... 16 7

B SOUTH / KEYS REGION CONCRETE BLOCK HIP ROOF (CBH) HOMES..... 175

C SOUTH / KEYS REGION WOOD FRAME GABLE ROOF (WG) HOMES........ 183

D SOUTH / KEYS REGION WOOD FRAME HIP ROOF (WH) HOMES ...............191

E FLORIDA MANUFACTURED SINGLEWIDE HOMES ..................................... 199

F FLORIDA MANUFACTURED DOUBLEWIDE HOMES ..................................205

G FLORIDA PRE-HUD CODE MANUFACTURED HOMES ........... ..................211

L IST O F R E F E R E N C E S ...................................................................... .....................2 17

B IO G R A PH IC A L SK E TCH ........................................ ............................................222
















LIST OF TABLES


Table page

3-1. Four m ost com m on structural types ...................................................................... 36

3-2. Population of most common structural types in defined geographic regions ..........36

3-3. A additional structural types ............................................... ............................ 37

3-4. Population of additional structural types in defined geographic regions ...............37

3-5. Structural type models for each geographic region.........................................38

4-1. Zones 1-6 M W FRS pressure coefficients ..................................... .................49

4-2. Zones 1E-6E M W FRS pressure coefficients ................................. ............... 49

4-3. Roof zone C&C pressure coefficient values for selected roof pitches...................53

4-4. Wall C&C pressure coefficient values ........... .............................. ...............54

4-5. Summary of load conditions applied to simulate extreme wind events .................66

5-1. Manufacturer's uplift capacity for typical roof-to-wall connections .....................78

5-2. Mean failure pressures for typical unprotected windows ........................................89

5-3. Site-built home summary of wind resistance capacities ........................................95

5-4. Manufactured home summary of wind resistance capacities...............................96

6-1. Site-built hom e dim tensions ........................................................ ............. 100

6-2. M manufactured hom e dim tensions ........................................ ........................ 101

7-1. M odeled structural types ......................................................... ............... 125

7-2. Structural types with damage based on combinations of modeled buildings ........125

7-3. Hurricane Andrew damages surveyed in the 1993 NAHB report..........................127

7-4. Wood frame home damages surveyed in the 1993 NAHB report..........................128









7-5. Window damage from Hurricane Andrew vs. simulated data .............................132

7-6. Masonry wall damage from Hurricane Andrew vs. simulated data ....................135

7-7. Wood frame wall damage from Hurricane Andrew vs. simulated data...............37

7-8. Roof-to-wall connection damage from Hurricane Andrew vs. simulated data......139

7-9. Roof sheathing damage from Hurricane Andrew vs. simulated data.....................141

7-10. Roof cover damage from Hurricane Andrew vs. simulated data...........................143

8-1. Structural repair cost ratios for Central Florida masonry homes .........................160

8-2. Non-structural repair cost ratios for Central Florida masonry homes.................... 160
















LIST OF FIGURES


Figure pge

2-1. W ind speed vs. height profiles...................................................... ..................8

2-2. Pressure locations for the differential pressure calculation in Equation 2-8 .........10

2-3. Pressure tap locations and w ind angles................................................. ........... 15

2-4. Ratio of aggregate pressure to maximum uplift capacity. .....................................16

2-5. Example probability distribution function of damage at a given wind speed........21

2-6. V vulnerability curve generation ........................................ ......................... 21

2-7. Fragility curve generation for 60% overall structural damage ...........................22

2-8. Fragility curve for the damage state of 60% overall structural damage ...............23

2-9. Family of fragility curves for a particular structural type............... ..................23

3-1. Regional boundaries for building classification. ................................................35

3-2. Distribution of conventional (site-built) home roof pitch values according to
the National Association of Home Builders Research Center..............................39

3-3. Distribution of manufactured home roof pitch values according to the
National Association of Home Builders Research Center................. ........ 41

3-4. Structural components selected for modeling in the hurricane damage-
prediction sim ulation engine ..................................................................... .. ... ..42

4-1. M W FR S pressure zones ................. ........................................... ............... 50

4-2. C& C roof pressure zones. ....................................................................................51

4-3. C & C w all pressure zones................................................................................. 51

4-4. Roof pressure zones for winds perpendicular to the ridgeline.............................52

4-5. Roof pressure zones for winds parallel to the ridgeline............... ... ...............52









4-6. Roof pressure zones for cornering winds............................................ 52

4-7. Method of determining shear wall loads from MWFRS pressures......................57

4-8. Tributary area for C&C pressures transferred into lateral connections on
w ood fram e w alls ........................... ...... .................................... ........58

4-9. Tributary area after significant roof-to-wall connection damage for C&C
pressures transferred into lateral connections on wood frame walls. ..................58

4-10. Values of the parameter A used in the determination of missile impact ...............61

4-11. Values of the parameter B used in the determination of missile impact ...............62

4-12. Values of the parameter D used in the determination of missile impact ..............64

4-13. Probability of missile strike causing breakage of a medium (3.5 x 5 ft)
window on a 44 ft long windward wall. ..................................... ............... 65

5-1. Gaussian distributions with a mean of 100 units and varying coefficients of
variation ...........................................................................69

5-2. Lognormal vs. Gaussian for a mean of 100 units and coefficient of variation
o f 0 .2 .............................................................................. 7 0

5-3. Truncated Gaussian distribution with a mean of 100 units and a COV of 0.4.......71

5-4. Typical arrangement of tie-down anchors for manufactured homes ...................94

6-1. Structural damage simulation engine flowchart ......................................... 98

6-2. Angles of wind incidence used for each wind speed .................................. 102

6-3. Flowchart for realizations of a structural type .....................................................103

6-4. M odeled structural com ponents ........................................................................ 106

6-5. Batch sampling method for roof-to-wall connections ......................................109

6-5. Location of forces for the overturning failure check on manufactured homes.... 121

7-1. Histograms of window damage on South/Keys CBG homes............................133

7-2. Window damage vulnerability of South/Keys CBG homes .............................134

7-3. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys
CBG hom es .............. ...... ..... ... ................. ........... .......... 134

7-4. Wall damage vulnerability of South/Keys CBG homes .....................................136









7-5. Fragility curves for 1, 2, 3, and 4 damaged walls for South/Keys CBG
hom es. .............................................................................136

7-6. W all damage vulnerability of South/Keys W G homes ............... ..................... 138

7-7. Fragility curves for 1, 2, 3, and 4 damaged walls for South/Keys WG homes....138

7-8. Roof-to-wall connection damage vulnerability of South/Keys CBG homes.......140

7-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof-to-wall connection
damage for South/Keys CBG homes. ........................................ ...............140

7-10. Roof sheathing vulnerability of South/Keys CBG homes.............................141

7-11. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof sheathing damage for
South/K eys CBG hom es. ............................................ ............................. 142

7-12. Roof cover vulnerability of South/Keys CBG homes...................................143

7-13. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof cover damage for
South/K eys CBG hom es. ............................................ ............................. 144

7-14. Histograms of roof sheathing damage on South/Keys CBG homes .................145

7-15. Histograms of roof-to-wall connection damage on South/Keys CBG homes.....147

7-16. Fragility curves for 2%, 5%, 10%, 25%, and 50% roof-to-wall connection
damage on South/Keys CBG homes..................... ............ .............. 147

7-17. Histograms of roof-to-wall connection damage on South/Keys concrete
block hom es. ..................................................................... 149

7-18. Histograms of roof sheathing damage on South/Keys concrete block homes.....149

7-19. South/Keys CBG homes mean damages for roof cover, roof sheathing, roof-
to-w all connections, and w alls ................................................................ ....... 151

7-20. South/Keys CBH homes mean damages for roof cover, roof sheathing, roof-
to-w all connections, and w alls ................................................................ ....... 152

7-21. South/Keys WG homes mean damages for roof cover, roof sheathing, roof-
to-w all connections, and w alls ................................................................ ....... 152

7-22. South/Keys WH homes mean damages for roof cover, roof sheathing, roof-
to-w all connections, and w alls ................................................................ ....... 153

7-23. Singlewide manufactured homes mean damages for roof cover, roof
sheathing, roof-to-wall connections, and walls..............................154









7-24. Doublewide manufactured homes mean damages for roof cover, roof
sheathing, roof-to-wall connections, and walls..............................155

7-25. Pre-HUD Code singlewide manufactured homes mean damages for roof
cover, roof sheathing, roof-to-wall connections, and walls.............................155

8-1. Preliminary results of the relation of structural damage to insurable content
loss compared with insurance claims data from Hurricane Andrew. ................161

A-1. Concrete block gable roof South/Keys Region home comparative levels of
roof cover, roof sheathing, connections, wall, and gable end sheathing
d am ag e .......................................................................... 16 8

A-2. Vulnerability to roof cover damage for South/Keys CBG homes.....................168

A-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
South/K eys CBG hom es. ............................................ ............................. 169

A-4. Vulnerability to roof sheathing damage for South/Keys CBG homes...............169

A-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for South/K eys CB G hom es........... ......... .................................. ............... 170

A-6. Vulnerability to roof-to-wall connection damage for South/Keys CBG
hom es. .............................................................................170

A-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for South/Keys Region CBG homes........................ ....... ............ 171

A-8. Vulnerability to wall damage for South/Keys Region CBG homes ..................171

A-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region
CBG homes..................... .. ......... .......... ......... ... .. ......... 172

A-10. Vulnerability to window damage for South/Keys Region CBG homes. .............172

A-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys
R region C B G hom es. .......................... ...... ................ ............... .... .......... 173

A-12. Vulnerability to exterior door damage for South/Keys Region CBG homes. .....173

A-13. Fragility curves for 1 and 2 damaged exterior doors for South/Keys Region
CBG homes. ............. ........ ..... ......... ..... ......... .. .. ......... 174

A-14. Vulnerability to garage door damage for South/Keys Region CBG homes........174

B-1. Concrete block hip roof South/Keys Region home comparative levels of
roof cover, roof sheathing, connections, wall, and gable end sheathing
d am ag e. .......................................................................... 17 6









B-2. Vulnerability to roof cover damage for South/Keys CBH homes.....................176

B-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
South/Keys CBH hom es. ...... ........................... ......................................177

B-4. Vulnerability to roof sheathing damage for South/Keys CBH homes...............177

B-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for South/K eys CBH hom es.......................................... ........................... 178

B-6. Vulnerability to roof-to-wall connection damage for South/Keys CBH
hom es. .............................................................................178

B-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for South/Keys Region CBH homes................... ....... ............ 179

B-8. Vulnerability to wall damage for South/Keys Region CBH homes ..................179

B-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region
C B H h o m e s ...................................... ............. ............. ................ 1 8 0

B-10. Vulnerability to window damage for South/Keys Region CBH homes .............180

B-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys
R region C B H hom es. .......................... ...................... .. .. ....... ........... ..181

B-12. Vulnerability to exterior door damage for South/Keys Region CBH homes .....181

B-13. Fragility curves for 1 and 2 damaged exterior doors for South/Keys Region
CBH hom es.............. ............. ..... .. ........... ......... .. .. ......... 182

B-14. Vulnerability to garage door damage for South/Keys Region CBH homes........182

C-1. Wood frame gable roof South/Keys Region home comparative levels of roof
cover, roof sheathing, connections, wall, and gable end sheathing damage........ 184

C-2. Vulnerability to roof cover damage for South/Keys WG homes......................184

C-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
South/K eys W G hom es ................ .......... .................................. ............... 185

C-4. Vulnerability to roof sheathing damage for South/Keys WG homes. ................185

C-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for South/K eys W G hom es. ........................................................................... 186

C-6. Vulnerability to roof-to-wall connection damage for South/Keys WG
hom es. .............................................................................186









C-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for South/Keys Region W G homes. .............................................. 187

C-8. Vulnerability to wall damage for South/Keys Region WG homes .....................187

C-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WG
h o m e s. ......................................................................... 18 8

C-10. Vulnerability to window damage for South/Keys Region WG homes..............188

C-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys
Region W G homes........... ........................... .. ..... ............... 189

C-12. Vulnerability to exterior door damage for South/Keys Region WG homes........189

C-13. Vulnerability to exterior door damage for South/Keys Region WG homes........190

C-14. Vulnerability to garage door damage for South/Keys Region WG homes..........190

D-1. Wood frame hip roof South/Keys Region home comparative levels of roof
cover, roof sheathing, connections, wall, and gable end sheathing damage........192

D-2. Vulnerability to roof cover damage for South/Keys WH homes....................192

D-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
South/K eys W H hom es ...................... .... ............... ................... ............... 193

D-4. Vulnerability to roof sheathing damage for South/Keys WH homes ..............193

D-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for South/K eys W H hom es. ........................................................................... 194

D-6. Vulnerability to roof-to-wall connection damage for South/Keys WH
hom es. .............................................................................194

D-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for South/Keys Region WH homes. .............................................. 195

D-8. Vulnerability to wall damage for South/Keys Region WH homes...................195

D-9. Fragility curves for 1, 2, 3 and 4 damaged walls for South/Keys Region WH
hom es. .............................................................................196

D-10. Vulnerability to window damage for South/Keys Region WH homes..............196

D-11. Fragility curves for 1, 3, 5, 7, and 10 damaged windows for South/Keys
Region W H homes............ ... .............. ....... ............................. 197

D-12. Vulnerability to exterior door damage for South/Keys Region WH homes........197









D-13. Vulnerability to exterior door damage for South/Keys Region WH homes........198

D-14. Vulnerability to garage door damage for South/Keys Region WH homes..........198

E-1. Singlewide manufactured home comparative levels of roof cover, roof
sheathing, connections, wall, and gable end sheathing damage ........................200

E-2. Vulnerability to roof cover damage for singlewide manufactured homes...........200

E-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
singlewide m manufactured hom es ....................................................................... 201

E-4. Vulnerability to roof sheathing damage for singlewide manufactured homes. ...201

E-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for singlewide manufactured homes. ....................................... ............... 202

E-6. Vulnerability to roof-to-wall connection damage for singlewide
m manufactured hom es. ........................................ .............................................202

E-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for singlewide manufactured homes.........................................203

E-8. Vulnerability to wall sheathing damage for singlewide manufactured homes....203

E-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing
for singlewide manufactured homes. ....................................... ............... 204

F-1. Doublewide manufactured home comparative levels of roof cover, roof
sheathing, connections, wall, and gable end sheathing damage ........................206

F-2. Vulnerability to roof cover damage for doublewide manufactured homes. ........206

F-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
doublewide manufactured homes. ............................................ ............... 207

F-4. Vulnerability to roof sheathing damage for doublewide manufactured
h o m es. ......................................................................... 2 0 7

F-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for doublewide manufactured homes................. .............................................208

F-6. Vulnerability to roof-to-wall connection damage for doublewide
m manufactured hom es. ........................................ .............................................208

F-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for doublewide manufactured homes ..............................................209









F-8. Vulnerability to wall sheathing damage for doublewide manufactured
hom es. .............................................................................209

F-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing
for doublewide manufactured homes................. .............................................210

G-1. Pre-HUD Code manufactured home comparative levels of roof cover, roof
sheathing, connections, wall, and gable end sheathing damage ........................212

G-2. Vulnerability to roof cover damage for pre-HUD Code manufactured
h o m es. ........................................................................ .. 2 12

G-3. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof cover for
pre-HUD Code manufactured homes........................................ ............... 213

G-4. Vulnerability to roof sheathing damage for pre-HUD Code manufactured
h o m e s. ........................................................................ .. 2 13

G-5. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof sheathing
for pre-HUD Code manufactured homes .................................. ............... 214

G-6. Vulnerability to roof-to-wall connection damage for pre-HUD Code
m manufactured hom es. ........................................ ................................ 214

G-7. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to roof-to-wall
connections for pre-HUD Code manufactured homes............... ... .................215

G-8. Vulnerability to wall sheathing damage for pre-HUD Code manufactured
h o m e s. ........................................................................ .. 2 15

G-9. Fragility curves for 2%, 5%, 10%, 25%, and 50% damage to wall sheathing
for pre-HUD Code manufactured homes .................................. ............... 216


xviii















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

PREDICTING THE VULNERABILITY OF TYPICAL RESIDENTIAL
BUILDINGS TO HURRICANE DAMAGE

By

Anne D. Cope

August 2004

Chair: Kurtis Gurley
Cochair: Gary Consolazio
Major Department: Civil and Coastal Engineering

Hurricanes have caused billions of dollars in losses in the United States and could

devastate up to $1.5 trillion worth of existing structures in Florida alone. The population

density on Florida's 1200-mile coastline continues to grow, and potential losses will

continue to mount. The insurance industry and the Florida insurance regulatory agency

both require a means of estimating these expected losses. Only a handful of studies exist

in the public domain to predict aggregate hurricane damage. Most published studies use

regression techniques with post-disaster investigations or claims data to develop

vulnerability curves. This approach is highly dependent on the type of construction

common to the areas represented in the data, thus limiting the predictive capabilities to

regions of similar construction. A promising approach used by one commercial model

estimates vulnerability by explicitly accounting for the resistance capacity of building

components and load produced by wind. This so-called component approach applies









claims data from previous storms as a validation (rather than calibration) tool, and can be

readily adapted to different regions with varying predominant construction.

The Florida Department of Insurance (FDOI) sponsored the development of a

public hurricane risk model. The goal of this ongoing project is to predict hurricane wind-

induced insurance losses by zip code for the State of Florida, on an annualized basis and

for predefined scenarios. The engineering team is responsible for relating specific wind

speeds to predicted losses for typical residential buildings in the state of Florida. Our

study developed a probabilistic model predicting structural damage from hurricane winds

in Florida. The core of this model is a Monte Carlo Simulation engine that generates

damage information for typical Florida homes, using a component approach. The

simulation compares deterministic wind loads, and the probabilistic capacity of

vulnerable building components to resist these loads, to determine the probability of

damage. In this manner, probabilistic structural damage is identified over a range of

assigned wind speeds. Monetary loss associated with structural damage and the

likelihood of occurrence for discrete wind speeds will be determined by models under

development by other groups in the project.














CHAPTER 1
INTRODUCTION

Windstorms produce billions of dollars in property and other economic losses

annually in the United States. Before Hurricane Andrew struck Florida and Louisiana in

1992, many insurance-industry experts thought the worst possible windstorm would

cause no more than $8 billion in insured property damage (Insurance Information

Institute May 2001 Update). Hurricane Andrew resulted in $15.5 billion in insured

property losses, $26.5 billion in total losses, and 61 fatalities [1]. Before Hurricane

Hugo's landfall in 1989, no hurricane had resulted in claims in excess of $1 billion

(Insurance Information Institute May 2001 Update). Hugo resulted in $7 billion in total

losses, and 86 fatalities. In 1999, Hurricane Floyd resulted in $6 billion in total losses,

and 56 fatalities [1]. According to the National Oceanographic and Atmospheric

Administration, wind-related disasters far outpace other natural disasters in total loss in

the United States. In light of these facts, efforts to estimate expected losses and mitigate

damage to residential structures from high-wind events are necessary to maintain the

viability of the increasing coastal population and infrastructure along the coastal United

States.

The effort to predict and mitigate hurricane damage is of particular importance in

the state of Florida (which lies in an area vulnerable to these high-wind events, and has a

large and increasing coastal population). Both the insurance industry and the Florida

insurance regulatory agency require a means of predicting future losses. In the public

domain, only a handful of studies predict aggregate hurricane damage. Most published









studies use regression techniques to develop vulnerability curves from post-disaster

investigations or available insurance claims data. Several of these studies are detailed in

Chapter 2. This approach of using data from previous storms is highly dependent on the

type of construction common to the areas represented in the data. Thus, the vulnerability

curves developed in the studies are limited to predicting damage for regions of similar

construction. For example, the observed-damage studies and claims data from Hurricane

Andrew can be used to develop a relationship between wind speed and probable damage

to homes of typical South Florida construction (mostly masonry). These relations would

not be suitable for predicting damage from hurricane winds to homes in North Florida,

where timber construction is more common. Thus, regression techniques must be

enhanced with methodologies that do not require large observed-damage data sets.

A promising approach used by one commercial model estimates vulnerability by

explicitly accounting for the resistance capacity of individual building components and

load produced by wind, within a probabilistic framework. This so-called component

approach applies claims data from previous storms as a validation (rather than

calibration) tool, and can be readily adapted to different regions with varying

predominant construction. While the overall framework of the Federal Emergency

Management Agency sponsored HAZUS model has been discussed in public literature

[2-4], the complex wind-structure interaction choices and assumptions involved in this

commercial model are not presented in full detail. Because this particular model and

other commercial models sponsored by the insurance industry are largely proprietary,

many of the details and assumptions used in their analysis are not available for public use

or critique.









In response to the need for a public model to predict hurricane wind-induced

insurance losses, the Florida Department of Financial Services sponsored a multi-

university project coordinated by the International Hurricane Research Center, and

involving meteorological, engineering, actuarial, and computer-resource components.

The product of this effort is the prediction of hurricane wind-induced insurance losses for

residential structures by zip code in Florida, on both an annualized basis and for

predefined scenarios (specific hurricanes).

The engineering team is responsible for relating specific wind speeds to predicted

losses, for typical residential buildings in the state of Florida. The University of Florida's

contribution to the project, presented in this dissertation, is the development of the model

that defines the complex relationships between hurricane wind speed and the resultant

structural damage, in a probabilistic framework. The core of this model is a Monte Carlo

simulation engine that uses a component approach to generate damage information for

typical Florida homes. The Monte Carlo simulation compares deterministic wind loads

and the probabilistic load-resistance capacity of building components to determine the

probability of damage. In this manner, probabilistic structural damage is identified over a

range of assigned wind speeds. This component approach may be developed based on

laboratory studies of the capacity of individual components, and proper accounting of

load paths and load sharing among components. This allows great flexibility with regard

to the types of structures that can be modeled. The development of damage relations is

not dependent on the existence of observed hurricane wind damage, but such information

can be used to validate and refine the model. The component approach also allows the

incorporation of future knowledge (such as additional capacity information on various









components) and the effects of mitigation measures (such as gable end bracing). All of

the data, decisions, and assumptions used in the model development are available for

public critique.

Research Hypothesis

The complex wind-structure interaction that leads to damage of typical residential

buildings during hurricane events can be broken down into three main components: local

wind field acting on the building, structural loads caused by the wind field, and resistance

capacity of the building components. If the relationship between the local wind field and

the structural loads is defined, then the problem of quantifying the risk of wind damage

can be addressed by applying a probabilistic framework to the structural loads and

resistance capacities of the building components.

The level and likelihood of structural damage will depend on parameters describing

the probabilistic representation of loads and resistance. Significant information on

probabilistic wind loading is available through wind tunnel and full-scale data sets,

provided the assumption that hurricane wind fields can be modeled by the log-law or

power-law holds true. These two modeling laws are described in Chapter 2. Laboratory

testing and post-storm damage reports provide valuable information on structural

resistance. Using this information to simulate the occurrence of hurricane events on

typical residential buildings will provide a measure of the ability of current typically

constructed residential buildings to withstand hurricane-force winds. Incorporating new

construction practices and retrofits (which alter component resistances) into the same

probabilistic framework will provide a means of calculating the benefit to homeowners of

adding hurricane-damage mitigation features to their homes.









Goals and Objectives

The focus of the research is the development of a simulation engine that provides

the probability of structural damage for typical Florida residential structures as a function

of peak gust wind speeds. Structural-damage information provided by this simulation

engine will serve as the backbone for the engineering component of the first publicly

available hurricane-wind damage-prediction models for residential structures. This focus

can be represented by four research objectives:

* Select residential models representative of the current building stock in the state of
Florida, and identify components of those structures for damage-prediction
modeling.

* Quantify the wind-induced loads on the identified components, and select
appropriate load paths.

* Identify the probabilistic capacities of individual components to resist wind loads.

* Create a probability-based system-response model that will simulate the
performance and interaction of the components of typical Florida homes, and
evaluate their vulnerability during interaction with hurricane winds.

Summary of Dissertation

The research objectives described above are detailed in Chapters 3 through 8,

following a brief summary in Chapter 2 of previous work in the field of hurricane

damage mitigation. Specifically, Chapter 3 presents the results of a survey of current

building stock to select typical residential building types and structural components

necessary to predict wind damage. Chapters 4 and 5 provide background information and

final decisions for the structural wind loads and building component capacities,

respectively. Chapter 6 describes the Monte Carlo simulation engine which uses the

determined loads and capacities to predict the vulnerability of typical Florida homes to

hurricane damage. Results obtained from the simulation process are presented in Chapter






6


7 and will be used to further develop the public hurricane risk model sponsored by the

Florida Department of Financial Services and coordinated by the International Hurricane

Research Center. Conclusions about the model, modeling process, and potential for future

use are discussed in Chapter 8.














CHAPTER 2
SUMMARY OF PREVIOUS RESEARCH

Previous research in the area of hurricane damage mitigation can be divided into

three main groups: efforts to quantify extreme wind loads, post-damage investigations,

and damage-prediction models. While numerous articles provide accounts of damage

from individual storms, few articles exist on the accurate prediction of hurricane damage

before a storm occurs. Most of the damage-prediction models that currently exist are

proprietary and unavailable to the public. Information available from post-damage

investigations and current methods of predicting future hurricane damage are detailed in

the following paragraphs, after an introductory section describing structural wind loads.

Background Information on Structural Wind Loads

At any given instant, a snapshot of wind speed vs. height at a location near a

building might resemble the curve in Figure 2-1 A. Removing the turbulent component to

consider the mean wind speed over some averaging time at each height increment

provides a smooth curve that might resemble the one in Figure 2-1 B. This curve is

typically modeled using one of two methods: the log law or the power law. Each method

results in a curve similar to the one in Figure 2-1 B, which has a mean wind speed of zero

at the ground surface and a constant mean wind speed at a distance above the ground

referred to as gradient height. Typically at elevations of 200 meters, the gradient (or

reference) height is the level at which the wind speed is no longer affected by the surface

roughness.











Gradient
Height




A B


Figure 2-1. Wind speed vs. height profiles. A) Typical profile at any given time. B)
Mean wind speed profile.

The log-law and power-law equations used to model the mean wind-speed profile

are given in Equations 2-1 through 2-3 [5]. Equations 2-1 and 2-2 define the log law,

while Equation 2-3 defines the power law.


U(z) = u -ln Z (2-1)
/ C o )


u -, (2-2)



U(z) =U(zrf) (2-3)
Zref

In Equations 2-1 and 2-2, U(z) is the mean wind speed at height z, K is the Von

Karman constant (approximately 0.4), zo is the roughness length of the terrain over which

the wind acts, and u, is the friction velocity (defined by a ratio of the shear stress at the

ground surface, To, and the density of air, p). The roughness length represents the size of

a characteristic vortex created as the wind moves over the terrain. The parameters u, and

zo are modified for each type of terrain [5]. In the power-law equation, U(z) is the mean

wind speed at height z, ac is a parameter modified for the type of terrain, and U(zref) is the

mean wind speed at reference (or gradient) height, Zref. The two methods provide nearly









identical results for the mean wind speed at heights above ground where low-rise

structures exist.

The turbulent component of the wind is most often represented as a Gaussian

random variable, with a zero mean and a standard deviation that varies with height.

Experimentation reveals that the standard deviation remains constant over the height at

which most structures and all low-rise structures exist [5]. The standard deviation of the

turbulence component in the direction of wind flow, ou; and the turbulence intensity as a

function of height, Iu(z), can be calculated using Equations 2-4 and 2-5 (where A is a

constant that varies with the roughness length, zo, and has a value of approximately 2.5

for open-country terrain) [5].

cr = Au, (2-4)


(z)=
U(z) (2-5)

Assuming that mean wind-speed profiles fit the models described above, one can

find a relationship between the mean wind speed and the pressure acting on areas of the

structure. Generally, the effect of the pressure on the structure is assumed to have two

parts: one from the mean wind speed, and one from the gusty or turbulent component.

The maximum pressure, pmax, that a component will experience as a result of both of

these portions can be expressed as the mean response, pavg, multiplied by a gust factor, G,

as shown in Equation 2-6 [6].

Pmax = GPg (2-6)

The most common approach in determining design pressures is to place a model

building in a wind tunnel, and conduct pressure coefficient studies. This approach was









used to develop the wind loading provisions for the American Society of Civil Engineer's

Minimum Design Loads for Buildings and Other Structures (ASCE 7-98) [7]. Roughness

elements are placed in the section of the wind tunnel preceding the model building such

that the mean wind speed vs. height matches that predicted by either the log law or power

law, and the turbulence intensity matches that predicted by the equation for Iu(z).

Pressure at a given location along a streamline can be found using Bernoulli's equation

for steady, inviscid, incompressible flow (Equation 2-7), where p is pressure, p is the

density of the fluid (air in this case), Vis the upstream velocity, y is the specific weight of

the fluid, and z is the depth or height with respect to a known reference [8].

p + 2 pV2 + Yz = constant along a streamline (2-7)

For the case of differential pressure between a point on the surface of the model

building and a point just in front of the building at mean roof height (Figure 2-2),

Equation 2-8 provides relative change in pressure.


2l*1 V

Mean Roof
/Height

Figure 2-2. Pressure locations for the differential pressure calculation in Equation 2-8

p2 -1 =pV2 (2-8)

In the wind tunnel, differential pressure is measured at locations of interest on the

building with respect to a reference pressure (usually located at gradient height). The raw

values of differential pressure are converted to pressure coefficients with respect to the

mean roof height of the building by multiplying by a correction factor taken from the

simplified version of Bernoulli's equation in Equation 2-8. This process is shown in









Equation 2-9, where Cp is the pressure coefficient at an individual location on the

building referenced to mean roof height, P is the measured differential pressure between

Locations 2 and 1 as shown in Equation 2-10, Vmnroofheight is the mean velocity at mean

roof height, and Vgradient height is the mean velocity at gradient height.

P (YV2
C r 2 _P gVradient height (2-9)
gradient height n roof height

P = p2 (2-10)

Fluctuating time histories of Cp at the same location on the building for various

angles of wind provide a probabilistic description of the pressure coefficient at that

location. This information leads to the selection of pressure coefficient values for

component design. Equations 2-11 and 2-12 from ASCE 7-98 illustrate the calculation of

design pressure for components and cladding on low-rise structures [7]. Equation 2-11

shows the calculation of velocity pressure at mean roof height, qh, which is a function of

the density of air. The 0.00256 value in Equation 2-11 is 12p for air in English units, Kh is

a terrain exposure coefficient, Kzt is a topographic effect factor to account for speed up

over hills, Kd is a directionality factor, Vis the design wind speed, and I is an importance

factor for the building. The velocity pressure can be thought of as the pressure measured

at Location 1 in Figure 2-2 for the true geographic location of the structure being

designed. Multiplying qh by a wind tunnel generated pressure coefficient provides the

pressure acting on the face of the structure at a particular location. The design pressure, p,

for each piece is found by multiplying qh by the difference between the external and

internal pressure coefficients with the gust factors built in, GCp and GCp,, respectively.

This process is shown in Equation 2-12.









qh = 0.00256KhKtKdV2I (ASCE 7-98 Eq 6-13) (2-11)

p = q, [GC, GC, (ASCE 7-98 Eq 6-18) (2-12)

The parameters for Equations 2-11 and 2-12 provided in ASCE 7-98 are intended

to envelope the realistic worst-case scenarios that might occur for the building, so that it

will be designed to withstand winds from any angle using a factor of safety worthy of the

importance of the building to the community.

Efforts to Quantify Extreme Wind Loads

The ASCE 7-98 design equations for wind pressure on the surface of a building

presented in the previous section are intended to envelope a realistic worst-case scenario.

Studies have show that this approach can lead to designs that are still un-conservative, or

conversely over conservative [9, 10]. Addressing the complexities of wind loading and

structural response in a more case-specific manner can rectify these design problems. The

accuracy of predicting structural wind loads is directly related to both the exactness with

which the behavior of near-surface winds can be predicted and the precision of modeling

the wind-structure interaction. The behavior of near-surface winds is highly variable and

sensitive to numerous localized characteristics (such as terrain). Moreover, the wind-

structure interaction is a highly complex and nonlinear problem, making detailed

characterization of wind loads difficult. Efforts to quantify extreme wind loads on

structures have sought to increase our understanding of hurricane wind behavior,

structural surface pressure characterization, and structural loading effects.

Defining the Behavior of Near-Surface Hurricane Winds

The ASCE 7-98 design equations provided in the previous section are based on the

assumption that the winds encountered by a building will behave in a manner predictable









by either the log law or power law previously described. There is some evidence, though,

that this assumption can lead to non-conservative predictions of maximum gusts [11].

That is, gust factors used to account for dynamic fluctuations from the mean wind speed

may not be suitable when applied to hurricane winds. In-field hurricane wind data

collection is a critical component to characterizing gust structure behavior in hurricanes.

Extensive data collection has been conducted for normal weather prediction, and

the operation of aircraft and airports. Unfortunately, this information does not provide

adequate data for the characterization of near-surface hurricane winds. Recently, several

institutions have begun efforts to collect ground level wind data during hurricane landfall.

Some of these include the National Hurricane Center, Texas Tech University, Johns

Hopkins University, University of Oklahoma, and Clemson University in conjunction

with the University of Florida.

The University of Florida and Clemson University have begun a hurricane data

collection project known as the Florida Coastal Monitoring Program (FCMP), sponsored

by the Florida Department of Community Affairs. One of the main goals of this project is

to help improve the fundamental understanding of the dynamic and turbulent action of

high-speed hurricane gusts. This is done through the collection of high-resolution wind

velocity, pressure, rainfall, humidity, and temperature data using custom-built

instrumentation set in the path of a land-falling hurricane. A set of ten and five meter

portable towers, equipped with vane and gill anemometers, barometers, hygrometers, and

rain gauges, are used to collect the data at several sites within the radius of influence of

the land falling hurricane. The FCMP has also instrumented houses in South Florida and

the Florida Panhandle area with removable pressure transducers to collect information on









wind forces in the building envelope. The FCMP has produced data sets from named

storms over the past four hurricane seasons. Findings from these datasets are still

preliminary, but continued efforts will produce a large dataset from which conclusions

can be drawn about the nature of hurricane winds at ground level. Information gained

about the turbulence intensity and gust eddy size may show that hurricane wind behavior

is unique at low-rise structure levels. Until that time, however, the assumption that

hurricane winds behave in a manner similar to non-hurricane winds will be used.

Characterizing Surface Pressures on Structures

The values obtained for GCp in Equation 2-12 are dependent on the effective wind

area for the structural member in question. As the effective wind area decreases, the value

of the coefficient increases. This trend results from the gust structure of the wind acting

on the structure. The turbulent component of the wind acting on a structure results from

the buffeting action of wind gusts. These gusts are made of large and small clusters of

swirling fluid referred to as eddies. The physical size of eddies is an important

characteristic. Small eddies hit the structure in an uncoordinated manner, while the

correlated winds of a large eddies can affect the entire effective wind area of some

structural components at the same time. The degree of linear correlation between

pressures at different locations on model buildings in a wind tunnel can be used to

determine the size of the gusts. More importantly, when the buildings are subjected to

wind tunnel tests known to model typical open-country conditions, the degree of linear

correlation between pressure tap locations can be used to better characterize the nature of

wind pressures on specific building components, such as roof sheathing.









A study conducted by the author to better characterize loads on low rise roof

structures explored both the spatial correlation and probability characteristics of pressure

coefficients acting on the roof and eaves of typical gable roof homes [12]. Investigations

were conducted to determine if regions of roof sheathing would have both highly

correlated surface pressures and a strong deviation from Gaussian probability. These

conditions represent a departure from the assumptions used in the gust factor approach,

and indicate that the roof sheathing is likely more vulnerable to damage than current

design methods suggest. Several standard non-Gaussian PDF models were associated

with different regions in the building envelope using goodness of fit procedures

comparing models to wind tunnel data. Significant combined effects (non-Gaussian loads

and high correlation over a surface) were found for cornering winds and winds

perpendicular to the gable end of the structure.

A follow-up study was conducted to investigate the results of the combined effects

of spatial correlation and non-Gaussian probability content on the aggregate loading of

one 4 ft x 8 ft piece of roof sheathing located at the ridgeline on the gable end of a typical

structure [13]. A non-Gaussian simulation algorithm was used to produce realizations of

pressure coefficient time histories at the 14 pressure taps representing a single piece of

roof sheathing (Figure 2-3).

1800
Row 1 s.....
Row2 *******

900
Tap 1 Tap 7


Figure 2-3. Pressure tap locations and wind angles










Using the pressure coefficient time histories, realizations of the aggregate pressure

on the sheathing panel were obtained for cases of high, moderate, and low correlation

among the pressure taps. The effects of the level of correlation are significant, as

demonstrated through comparison of higher-moments of the aggregate pressure

coefficient, as a ratio of aggregate pressure with ASCE 7-98 load conditions, and as a

ratio of aggregate pressure with an experimentally determined uplift capacity. Figure 2-4

shows the ratios of the aggregate pressure resulting from a 150-mph. 3-second gust wind

(ASCE 7-98 wind conditions for South Florida) to the uplift capacity of a typical Douglas

fir panel with 6d nails in a 6/12 nail pattern [14]. Results indicate that regions

experiencing highly correlated non-Gaussian pressure fields will frequently see loads

greater than the capacity of the system (a ratio larger than 1), while the assumption that

the pressure field is not correlated, but non-Gaussian results in loads well within the

capacity of the system. Complete results can be found in Gioffre, Gurley, and Cope

(2002).



---- no corr
7-5
9 -- lhcor -- high .....rrelaho n






3 0


.\ A B
02 04 06 08 1 12
ratio of aggregate load to capacity 2 0 1000 2000 3000 40 00 WO 600
6000 point simulation of aggregate Ioad

Figure 2-4. Ratio of aggregate pressure to maximum uplift capacity. A) PDF for three
levels of correlation among taps. B) Time history for high correlation









Characterizing and Codifying Structural Loads

Some wind engineers seek to incorporate the non-Gaussian qualities of wind

pressures discussed in the previous section into better building codes by using database-

assisted design methods. Current technology allows design engineers to analyze

structural responses with nimble accuracy, yet the wind load provisions remain crudely

broad brush. Using wind pressure and climatological databases instead of current wind

pressure tables and plots would provide a more risk-consistent design and would allow

for the use of the structures own influence lines as opposed to generic, cookie-cutter

structural influences built in to current methods [9]. Studies conducted for the

development of database-assisted design software reveal the non-Gaussian nature of wind

load effects. Specifically, time histories of the bending moments in a steel frame low-rise

structure indicate that the Gamma distribution is most appropriate when selecting the

maximum peak load [10]. Additional studies reveal that the inclusion of wind

directionality effects allows for a more risk-consistent design over the current approach

of using a global directionality factor, Kd, of 0.85 (Eq. 2-11). In fact, the current approach

for wind directionality effects may lead to an underestimation of the structural wind load

in approximately 10-15% of buildings designed using the 1998 standard [9].

Summary of Efforts to Quantify Extreme Wind Loads

Investigations into the nature of hurricane near-surface winds from full-scale data

and the nature of wind-structure interaction in the form of pressure coefficient data from

wind tunnel testing will continue at the University of Florida and other institutions.

Synthesis of the information gained from these efforts will lead to the development of

better building codes and design practices. The current body of information concerning

wind surface loads on low-rise structures is not robust enough to allow full incorporation









in the Monte Carlo simulation developed for the FDOI hurricane loss projection model.

The simulation engine described in subsequent chapters relies on aggregate pressures

calculated from pressure coefficient zones. These zones are based on values in ASCE 7-

98, but they are modified for directionality using knowledge gained in the previously

described research. The details of modification are described in Chapter 4. Inclusion of

non-Gaussian behavior and correlation between surface pressures is a promising topic

that could be incorporated into the developed model at a later date.

Post-Damage Investigations

Post-damage investigations provide an assessment of how structures perform in

extreme wind events and can indicate strengths and weaknesses in design codes and

construction practices. Numerous papers discuss damage from Hurricanes Alicia,

Andrew, Hugo, Iniki, and Opal [15-23]. In general, the reports contain valuable

information on types of failures commonly encountered and recommendations to prevent

similar failures in future events, but these observations by experts in the field are not

backed by statistically significant numbers of evaluations. For example, the damage to

buildings in the Houston-Galveston area during Hurricane Alicia was attributed to the

lack of adequate hurricane resistant construction, rather than to the severity of the storm

[17, 18]. A similar conclusion was reached on damage to buildings during Hurricane

Hugo [23]. A reliability analysis of roof performance during Hurricane Andrew found

actual performance to be better than predicted by the governing building code at the time,

although the authors stress the need for further research to quantify statistically both

construction characteristics and damage due to storms [21]. Phang also offers several

observations of the damage on low-rise buildings caused by Andrew. He found that

plywood sheathing performed remarkably better than board sheathing, diagonal bracing









was critical at gable ends, and gable roofs showed much more structural damage than hip

roofs [22]. Research has also been conducted in Australia by Mahendran who gives an

overview of the typical damages encountered by low-rise buildings in the tropics,

subjected to either hurricanes or severe storms. In addition, he and the Australian

scientific community also stress the fact that full-scale testing is necessary to better

predict to behavior of the entire building system when subjected to high-speed winds.

While these studies are extremely valuable for the development of safer housing, they do

not offer a sufficient basis from which to draw reliable quantitative conclusions [24]. The

information obtained from these studies does, however, provide a means of validating the

results of a probabilistic approach relating peak wind speeds to structural damage. One

would expect the most common types of failures detailed in post-disaster reports to be

same as the types of failure obtained from Monte Carlo simulations of hurricane-force

winds and structural component resistance.

Post-damage studies also provide a means of estimating the distribution of the

building stock in Florida cities. The most comprehensive studies, undertaken by the

National Association of Home Builders (NAHB) following Hurricanes Andrew and Opal,

include information on the sample size and types of homes investigated [19, 20]. This

information, in combination with data from County Property Appraisers and other

resources, is useful for predicting typical sizes and types of homes in other Florida areas,

as detailed in Chapter 3. Furthermore, the storm damage reports serve as a benchmark by

which to set priorities for research efforts since these reports identify the building

components that experience the most frequent or most debilitating damage.









Damage Prediction Models

Damage prediction models make use of the current knowledge base to predict

damage in future extreme wind events. While several post-damage reports exist in the

public domain, there are few damage prediction models available for public review.

Those that can be found follow one of two paths. The most common approach is to use

post-damage investigation results to create vulnerability or fragility curves for structures

(defined in the following section). A second approach is to build a probabilistic model to

generate structure fragility curves for damage prediction. This latter approach requires

some assumptions about the strength of buildings and type of terrain. Simulations are

used to create the curves and data sets to calibrate and validate the results. The advantage

to this approach is the ability to generate rational approximations of damage curves for

structural types that have not yet experienced a major hurricane. Developing curves based

on damage data alone requires the existence of large sets of damage data, while the

development of curves based on probabilistic assumptions and simulations can

incorporate laboratory data sets and engineering judgment when damage data sets are not

available.

Fundamental Concepts in Damage Prediction

Vulnerability and fragility curves are both indicators of the ability of a specified

structure to withstand hurricane-force winds. To develop each type of curve, the level of

damage or damage state must be defined. For instance, one could identify damage states

involving roof failure, wall failure, or some other type of failure. For demonstration

purposes, damage can be thought of as a percentage of overall structural damage. Each

building will either be undamaged (0% damage), partially damaged by some percentage,










or totally destroyed (100% damage). At a given wind speed, there will be a distribution of

percent damage to structures of the same type (Figure 2-5).

P(D) = Distribution of damage
p(D) at a given wind speed for a
particular building type




00 Damage 100%


Figure 2-5. Example probability distribution function of damage at a given wind speed

Once the distribution of damage is known over a range of wind speeds, the

vulnerability for that type of structure can be determined. The vulnerability curve is a

means of measuring the performance of the structure, and is generated from the location

of the mean percent damage value from the damage distribution at each wind speed.

Figure 2-6 shows the process of vulnerability curve generation from individual PDFs

associated with particular wind speeds. The generated vulnerability curve defines the

mean damage for a particular structural type as a function of wind speed, where mean is

defined as the damage level at which 50% of all structures of that type will be less

damaged, and 50% more damaged.

Mean Damage
Factor (%)

Mean /Vulnerability
Mean Curve




Mean





Sv Mean Wind Speed


Figure 2-6. Vulnerability curve generation










Fragility curves are another means of describing the performance or reliability of

the structure. A fragility curve provides the probability that a certain level of damage will

be met or exceeded at a given wind speed, and can be used to determine how many

buildings of similar type in an area will experience at least a certain level of damage. This

can be thought of as a conditional probability of exceedence. Given the maximum wind

speed for a particular wind event, the fragility curve for a type of structure provides the

likelihood of damage exceeding a certain threshold. Figure 2-7 and Figure 2-8 show how

the fragility curve for a given structural type is determined from available damage

distributions at different wind speeds. The example demonstrates how to calculate the

fragility curve corresponding to 60% damage by setting a threshold in Figure 2-7 and

integrating under each damage distribution from the 60% threshold point to the positive

extreme. The integrated values (shaded areas in Figure 2-7) become the data points for

the fragility curve at each wind speed (Figure 2-8). The limit on the vertical axis of the

fragility curve in Figure 2-8 is 1.0, representing a 100% likelihood of occurrence for the

given damage state.

Mean Damage
Factor (%)
80% Vulnerability
120% 35% 1 Curve


60% -






V 3 Mean Wind Speed


Figure 2-7. Fragility curve generation for 60% overall structural damage










Probability of
Exceedence
Fragility Curve
D =60%









V1 V2 V3 Mean Wind Speed


Figure 2-8. Fragility curve for the damage state of 60% overall structural damage

Other damage thresholds can be set to generate a family of fragility curves for this

structural type (Figure 2-9). To clarify, the vulnerability curve shows the most likely

mean damage that will occur to a given structure as a function of mean wind speed, while

the fragility curve shows the probability of exceeding a specific level of damage as a

function of wind speed. With vulnerability and fragility curves for structure-type 'A', the

following types of questions can be answered: 1) for a 90 mph. gust, what is the average

expected damage to houses of type 'A' (using vulnerability curve) and 2) for a 90 mph.

gust, what is the likelihood of seeing 80% damage or greater (using fragility curve)?

Probability of
Exceedance
D=200
A D= 40% I D 60%



SD= 80%





V, V2 V3 Mean Wind Speed


Figure 2-9. Family of fragility curves for a particular structural type









Damage Prediction Models in the Public Domain

Damage prediction models in the public sector using the approach of fitting

vulnerability curves from post-damage investigation results include two studies that rely

heavily on insurance claim information. The first of these studies determined the

relationship between insurance claim figures and wind speed for Typhoons Mireille and

Flo [25]. The second performed a similar analysis for Hurricane Andrew [26]. Since the

buildings involved in the first study were residential buildings in Japan, the results are not

readily applicable to typical residential structures in Florida. The second study used data

collected from two large insurance companies in Dade County, Florida to calculate the

vulnerability function as a percentage of loss vs. mean wind speed at gradient height.

This information is clearly helpful in determining how residential structures typical of

those existing in South Florida in 1992 will perform in a hurricane event of similar

magnitude. However, this data is a snapshot in time, capturing the damage on structural

types that existed when the extreme wind event took place. The data cannot take into

consideration improvements in building construction over time, nor can it be readily

applied to areas where the terrain and type of construction are notably different.

Others in the public sector have predicted damage using probability-based

simulation models to generate the likelihood of damage vs. wind speed. One such study

presents the vulnerability curve for a fully engineered building using the assumption that

the resistance capacity of the building is lognormally distributed [27]. Since the model

was developed for engineered structures, the approach is not likely to yield the best

results for predicting damage to typical residential buildings in the state of Florida.

Another study presents a method of predicting the percentage of damage within an area

as a function of the gradient wind speed, gust factor, average value of the buildings, and









two parameters which govern the rate of damage increase with wind speed [28]. These

last two parameters are empirically determined based on experience and knowledge of

the area. Since the results of this study were not reproducible, the model is not considered

a reasonable approach for the prediction of damage to residential buildings in the state of

Florida, given the information currently available.

Insurance data from Hurricane Hugo is used as an example to illustrate the

probabilistic approach presented in a recent study for long-term risk analysis [29]. The

authors calculated and published statistics for hurricane simulation parameters based on

previous storms that made landfall in Florida, North Carolina, and South Carolina.

Simulations of hurricane events over 50 year periods and investigation into historical

wind speed records were used to predict 50-year mean recurrence interval (MRI) wind

speeds at gradient height for selected coastal areas [5]. These 50-year MRI wind speeds

at gradient height were converted to ground wind speeds based on the type of terrain

present and used in conjunction with fragility curves generated from insurance loss data

to predict damage in areas of interest. The authors provided a graphical representation of

the generated 50-year MRI wind speeds at gradient height and a fragility curve generated

from two sources: insurance claims in Florida after Hurricane Andrew, and claims in

South Carolina after Hurricane Hugo. The damage levels predicted by this method are the

amount of damage likely to recur once every 50 years, or that have a 2% chance of being

exceeded annually in the area of interest, provided the building stock remains relatively

unchanged. The difficulty with using this method of damage prediction is the reliance on

insurance data from only two events to generate the fragility curve from which losses are

predicted for future storms. Unfortunately, information from which to determine accurate









fragility curves for a certain type of structure or family of structures is currently limited.

Even if larger data sets were available from other storms, the models would only be valid

when used to predict damage to structures of like-construction. This reliance on post-

disaster information restricts the ability to project the effects of design modifications,

code changes, and retrofit measures on the vulnerability of existing structures. This

realization has lead to the pursuit of approaches that seek to model damage at the

component level rather than for the entire structure. Structural risk assessment is then a

matter of combining the vulnerability of the individual parts making up a structure. The

so-called component approach allows the flexibility to including new components and

retrofit measures, provided lab tests are performed to assess their probabilistic resistance.

Proprietary Damage Prediction Models

Private sector damage prediction models also exist. In the wake of Hurricane

Andrew (which generated insurance claims totaling nearly twice the amount thought

possible by experts in the field), private sector insurance industry groups contracted

damage prediction models from engineering firms to develop a better understanding of

the risks associated with a hurricane strikes in heavily populated areas. Access to this

information is limited, since the projects are largely proprietary. Currently, some

information has been published describing the strategy used to predict damage for these

projects. One such study used a re-arranged version of the design pressure equation from

ASCE 7-98 (Equation 2-13) to calculate the wind speeds at which individual components

will fail [30].


V e P(fG]re' (2-13)
""'"" 0.00256KhK ,t(GC )- (GC,,









In Equation 2-13, pfazlure is the statistically sampled failure pressure of the

component, Kh is an ASCE-defined terrain exposure coefficient, Kzt is an ASCE-defined

topographic effect factor to account for speed up over hills, Iis an ASCE-defined

importance factor, GCp is a statistically sampled external pressure coefficient, and GCp, is

an ASCE-defined internal pressure coefficient. After calculating failure speeds for each

component, the researchers determined damage histories for buildings during simulated

hurricane events. The result of this analysis was a vulnerability curve for a particular type

of building, which can be used with replacement cost information to determine probable

insurance losses. The methodology for this study has been published, but the

vulnerability results are not available.

The most recent damage prediction model is the HAZUS Multi-hazard model,

which addresses wind, flood, and earthquake hazards. Under the direction of the National

Institute of Building Sciences and the Federal Emergency Management Agency, the

HAZUS hurricane model was developed by Applied Research Associates (ARA) over a

period of several years. A preview of this model was released to hurricane prone regions

of the United States in 2002 that allows users to estimate and evaluate disaster relief

resources and policies through scenario analysis [3]. The information supplied by the

preview model includes planning for the number of displaced persons, sheltering

requirements, and post-storm debris removal. ARA has published a description of the

hurricane model's six components: hurricane hazard, terrain, wind pressure, wind borne

debris, damage, and losses for buildings. The distinct advantage of the HAZUS

methodology over previous damage prediction methods lies in the fact that it is a

component-based model rather than a regression curve fitting model. The HAZUS









model explicitly accounts for the resistance capacity of individual building components

and wind loading, within a probabilistic framework. Using information from British,

Australian, and American wind loading codes, as well as boundary layer wind tunnel

testing, ARA developed an empirical model for the pressure coefficients on the surface of

typical buildings. Techniques for estimating the risk of wind borne debris impact and the

effects of sheltering from nearby buildings were also developed. This information was

used to create a computer simulation tool that would apply a hurricane wind model (also

developed by ARA) to a typical building and evaluate the damage accrued every 15

minutes as a result of wind pressure or wind borne debris impact. Monetary losses

resulting from structural damage were obtained by calculating the replacement cost

explicitly for the external portion of the building and implicitly for the internal structure

and contents. This model has been validated with available insurance records and is

considered to be the state of the art in hurricane damage prediction. While the framework

for the model has been well defined in public literature, many decisions and assumptions

used in the determination of wind loads remain proprietary.

Public Loss Hurricane Projection Model

The Public Loss Hurricane Projection Model is currently under development for the

Florida Department of Financial Services, with a scheduled release date of May, 2005.

This multi-university project (coordinated by the International Hurricane Research

Center) will predict hurricane wind-induced insurance losses for residential structures by

zip code for the State of Florida, on both an annualized basis and for predefined scenarios

(specific hurricanes). Since the model is sponsored in the public domain, the data,

decisions, and assumptions used will be available for public critique. The framework of

the model includes a meteorology component to generate probabilistic information about









wind speeds on an annualized basis for each zip code in Florida, an engineering

component to relate specific wind speeds to physical damage to residential structures

typical of Florida homes, and a financial component to relate physical damage to both

content loss and total insurance dollar loss. Subsequent chapters outline the strategies

employed in the engineering component of calculating physical damage to typical

residential buildings in Florida as function of a series of peak 3-second wind speeds. This

model, like the HAZUS hurricane model, is component-based, explicitly accounting for

resistance capacities of structural components and wind loading within a probabilistic

framework. The public model is not as complex as the one developed by ARA, foremost

in that it does not time step through the entire life cycle of a hurricane. The public model

does incorporate, to the extent possible, the current state of the art knowledge in wind

pressures, windborne debris and resistance capacities for typical residential buildings in

the state of Florida.














CHAPTER 3
RESIDENTIAL STRUCTURES IN FLORIDA

Defining appropriate residential structural models for the state of Florida is a

critical step in the development of a simulation engine to predict structural damage in the

state as a function of peak gust wind speeds. Wind loading characteristics are heavily

dependent on the shape and component make-up of the individual structure under

consideration. Thus, the accuracy and reliability of the damage-prediction simulation

engine is dependent on proper characterization of the building population in the state.

Additionally, the efficiency of the simulation model relies on correctly identifying

building components that are susceptible to wind damage. Finally, the resulting damage

predictions will be useful only for statistically significant building types. Therefore,

knowledge of the types of structures, the components of those structures most susceptible

to wind damage, and the distribution of structural types throughout the state is critical to

the success of each step in the prediction of hurricane damage.

Research partners in this joint project conducted an in-depth study of building

classifications. This chapter summarizes three contributions: statistical analysis of the

residential building population of Florida conducted by Liang Zhang of the Florida

Institute of Technology, with assistance from the author [31, 32]; manufactured housing

research conducted by Luis Aponte of the University of Florida; and a building

component investigation conducted by the author. Sources of information for

characterizing residential structures in the state of Florida include the Florida Hurricane

Catastrophe Fund (FHCF) exposure database, databases of individual county property









appraiser's offices, manufactured home builder literature, and post-damage

investigations.

Sources of Information

Florida Hurricane Catastrophe Fund Exposure Database

The FHCF exposure database consists of insurance portfolio data for buildings in

the state of Florida. At this time, data available to the team of researchers working on the

damage-prediction simulation engine consists of a statistical analysis of the FHCF

database for single family residences (SFR) only. Information concerning the population

of manufactured homes by ISO classification is not available.

Unfortunately for wind researchers, the ISO classifications used in insurance

portfolios focus largely on fire hazards. This information alone does not provide an

adequate structural characterization of Florida residences, with respect to wind loading. It

can be used (in combination with other sources of information) to identify regional

boundaries within the state. For example, the population of masonry homes vs. wood

frame homes was found to be consistent among groups of counties in the same

geographic area. The ISO construction classifications (described in greater detail in a

master's thesis written by a research partner [32]) are

* Frame
* Joisted Masonry
* Non-Combustible
* Masonry Non-Combustible
* Modified Fire Resistive
* Fire Resistive
* Heavy Timber Joisted Masonry
* Superior Non-Combustible
* Superior Masonry Non-Combustible
* Masonry Veneer
* Unknown









County Property Appraiser Databases

The most comprehensive sources of detailed structural information currently

available are the individual county property appraiser databases. Each county gathers

residential and commercial property data for tax purposes. Database architecture and

contents (beyond those required by the Florida Department of Revenue) vary, but each

database can be separated into four general categories: commercial property, SFRs,

condominiums, and manufactured homes. Commercial property and condominiums are

outside the scope of the current work, so the two categories of interest are SFR and

manufactured homes. Nearly all of the SFRs in each county are listed in the county

property appraiser's database. A large number of manufactured homes are taxed through

the Department of Motor Vehicles; however, and are not listed in the property database.

Processing database information from each of the 67 counties in Florida is not

feasible for the current project; therefore a selection of counties spread throughout the

state is used to obtain information about the characteristics of typical Florida homes. The

team was able to gather databases from several counties, but approximately half were

unusable because files did not match the database layout provided by the property

appraiser's office. The nine counties that supplied databases from which useful structural

information was gained are

* Brevard County
* Broward County
* Escambia County
* Hillsborough County
* Leon County
* Monroe County
* Palm Beach County
* Pinellas County
* Walton County









From the databases of the nine listed counties, the type of roof, type of roof cover,

exterior wall material, stories, square footage, and year built are investigated for SFRs.

This information is useful in identifying the most common residential structural types,

but is incomplete as a characterization of homes, with respect to wind loads. Because the

information is used for taxation, database categories often describe qualitative

information (rather than structural details). For instance, exterior walls may be listed as

'average', without indicating the building material. Some database fields lump

structurally significant details into a single category. Many counties, for example, use a

single designation of 'hip or gable roof instead of separating the two. This difference is

structurally significant, as post-damage investigations have noted during past wind events

[19]. Additionally, some structurally significant information is not listed in the databases,

such as the presence of a garage. In spite of these limitations, the databases supplied by

the nine listed counties allowed the research team to develop models representative of

typical Florida homes.

Manufactured Home Builder Literature

Information about manufactured homes could not be easily discerned from the

individual county property appraiser databases. Since many of these homes are taxed by

the Department of Motor Vehicles, construction information from the tax authority is

limited. Manufactured home information has been obtained from a report compiled by

the National Association of Home Builders (NAHB) Research Center for the Department

of Housing and Urban Development comparing site-built and manufactured housing [33]

and from contacts with the Partnership for Advancing Technology in Housing (personal

correspondence by a research partner, June 2003) and Nobility Homes (personal

correspondence by a research partner, June 2003).









Post-Damage Investigations

Literature searches of post-damage reports reveal that observations by experts in

the field are useful in supporting the statistical information on building population

characteristics gained from other sources of data. However, the post-damage reports

themselves usually do not contain enough building evaluations to be considered a

statistically significant database from which to characterize Florida's building population.

The one exception is the NAHB Research Center report that describes the damage in

South Florida after Hurricane Andrew [19]. In this report, the damage to residences is

provided within a statistical framework. Unfortunately, this information is available for

only one small geographic region following one storm.

Though they are not a source of statistical information about the building

population, post-damage reports are vital in determining which building components to

model in a hurricane damage simulation engine. The expert opinions in post-damage

reports indicate where severe wind damage occurs in typically constructed homes and,

therefore, where the most benefit is to be gained from mitigation efforts.

Results of the Building Population Investigation

The information gained in researching the FHCF database, individual county

property appraiser databases, manufactured home builder literature, and post-damage

reports is detailed in this section. The discussion is divided into two sections: site-built

home information is presented first, and manufactured home data follows.

Characterization of Site-Built Homes

The results gained from the nine individual county property appraiser's databases

can be generalized to four regions of the state. The choice of regional boundaries is

governed in part by the statistics of wood frame houses in each county (an analysis of the









FHCF database conducted by the meteorology team). Additional selection criteria

included having at least two representative counties in each defined region and following

the population density trends in South Florida. The resulting regions (defined as North,

Central, South, and Florida Keys) are outlined on the county map of Florida shown in

Figure 3-1. The shaded counties indicate the location of the nine from which property

appraiser database information was obtained and successfully processed. A master's

thesis written by a research partner details the process of determining the regional borders

shown in Figure 3-1 [32].





North



Central

"South


Keys

Figure 3-1. Regional boundaries for building classification.

Review of each processed county property appraiser database and the post-Andrew

NAHB report [19] indicate that the most common structures in the state can be

summarized into four types, provided in Table 3-1. Table 3-2 shows the estimated

percentage, p of each type per region and the mean square footage of temperature

controlled area, A, for each case. The areas provided for the Keys Region are marked

with an asterisk due to their large standard deviation [32]. Because the average home size

in the Keys is likely affected by a few grand estates, values from the South Region are

used.









Table 3-1. Four most common structural types
Structural
Type Characteristics
CBG Concrete block gable roof one story home with shingles or tile
CBH Concrete block hip roof one story home with shingles or tile
WG Wood frame gable roof one story home with shingles or tile
WH Wood frame hip roof one story home with shingles or tile

Table 3-2. Population of most common structural types in defined geographic regions
North Region Central Region South Region Florida Keys
Structural p A p A (ft2) p A p A (ft2)
Type (ft2) (ft2)
CBG 12% 42% 46% 23%
1702 2222 2147 3295*
CBH 6% 22% 23% 11%
WG 39% 12% 4% 12%
1908 1941 2022 2771*
WH 20% 6% 2% 6%
Sum of most
77% 82% 75% 52%
common
Unknown 14% 13% 11% 23%
Total
Total 91% 95% 86% 75%
coverage
Large standard deviation from observed data

The third row from the bottom of Table 3-2 represents the percentage of the SFR

population covered by the most common structural types. Those not covered include two

story homes, unusually constructed homes, and homes of unknown structural type.

Unfortunately, the percentage of homes listed in available data as having an unknown

structural type is significant in each region, as shown in the next to last row of Table 3-2.

Since these homes cannot be classified, the population represented by this category will

be assigned an average value of structural wind damage obtained from an investigation of

other structural types in that region. Further details concerning this process are provided

in Chapters 7 and 8, in which the structural damage results and conclusions are presented.

The population of SFRs covered by the four most common structural types is

adequate in the North, Central, and South Regions, but the Keys Region has a significant

number of homes not represented in Table 3-2. Additional structural types are listed in









Table 3-3. The site-built home population represented by these additional groups is

provided in Table 3-4, rounded to the nearest whole percent. Using these additional

categories, the portions of the building population not counted in Table 3-2 are covered.

Table 3-3. Additional structural types
Structural
Type Characteristics
2CBWG Concrete block 1st story, wood frame 2nd story, gable roof home with
shingles or tile
2CBWH Concrete block 1st story, wood frame 2nd story, hip roof home with
shingles or tile
2WG Wood frame two story gable roof home with shingles or tile
2WH Wood frame two story hip roof home with shingles or tile
2Keys Two story home of unspecified frame and roof cover
CBGM Concrete block gable roof one story home with metal roof
CBHM Concrete hip gable roof one story home with metal roof
WGM Wood frame gable roof one story home with metal roof
WHM Wood frame hip roof one story home with metal roof

Table 3-4. Population of additional structural types in defined geographic regions
North Central South Florida
Structural Type Region p Region p Region p Keys p
2CBWG 1% 2% 8%
2CBWH 1% 1% 4%
2WG 5% 1% 1%
2WH 2% 1% 1%
2Keys 3%
CBGM 8%
CBHM 4%
WGM 7%
WHM 3%
Sum of most common 77% 82% 75% 52%
types (from Table 3-2)
Unknown 14% 13% 11% 23%
Total 100% 100% 100% 100%

For the Keys Region, a significant portion of the population previously uncounted

in Table 3-2 is listed in the categories with an 'M.' These match descriptions of the four

most common structural types with the exception of the type of roof cover. For the North,

Central, and South Regions, two story homes make up the difference. The population of









individual types of two story homes shown in Table 3-4 is small in comparison to the

overall population in these three larger regions. Additionally, the entire population of two

story homes in the Keys represents only 3% of the population of this smaller region of the

state. Given the contribution of two story homes relative to the overall SFR population,

separate models are not developed for each two story type listed in Table 3-3. Instead, the

performance of two story SFRs is predicted using the one story models in each region. In

the North region, the WG and WH models are used as a framework for determining two

story damages. Two story homes in the Central and South Regions are based on the CBG

and CBH models. A two story model for the Florida Keys uses information from each of

the single story models.

Plan dimensions are selected for each type of single story home such that the

square footage remains close to the mean area plus an unheated garage of approximately

400 ft2, while providing the largest number of whole sheathing panels on the roof surface.

Unusually shaped sheathing panel cuts are avoided. The resulting site-built models for

single story homes are described in Table 3-5, where the plan dimensions represent the

wall lengths. An overhang of two feet on each side adds a total of four feet to both plan

dimensions to give the size of the roof surface.

Table 3-5. Structural type models for each geographic region
South Region and
Structural North Region Central Region Florida Keys Region
Type Plan (ft) Area (ft2) Plan (ft) Area (ft2) Plan (ft) Area (ft2)
CBG or CBH 56x38 2128 60x44 2640 60x44 2640
WG or WH 60x38 2280 60x38 2280 56x44 2464

The information presented in Tables 3-1 through 3-5 represents the bulk of

information from the available property appraiser databases useful to the structural wind

load characterization of SFRs. Unfortunately, additional information critical in the









determination of wind loading conditions is required, but generally not available, from

this source. As a result, some structurally descriptive characterizations must be made on a

statewide basis, rather than regionally. One of these critical structural characteristics not

obtained from the county property appraiser's databases is the slope (or pitch) of the roof,

a critical factor in the determination of wind loads on roof surfaces and in the sizing of

roof components. A national distribution of typical roof pitch values is presented in

Figure 3-2, where the numerator and denominator represent the number of inches of rise

and run, respectively. The data for the figure is taken directly from the NAHB Research

Center's 1998 report comparing factory built and site-built housing.

flat up to 1/24,
1%
Lp to 4/12, 16%



7/12 and up,
43%





"5/' /12 and 6/12,
40%
Figure 3-2. Distribution of conventional (site-built) home roof pitch values according to
the National Association of Home Builders Research Center.

From the national information presented in Figure 3-2, and discussions with Dr.

Leon Wetherington of the University of Florida College of Building Construction

(personal correspondence, September, 2002) a pitch of 5 on 12 (5 inches of rise to the

linear foot), corresponding to a roof slope, 0, of approximately 23 is selected as the

most representative value for the population of site-built homes in Florida. This choice









becomes an integral part of the wind load criteria for the structure. One section of the

wind load provisions of the American Society of Civil Engineers requires interpolation

by roof slope, while the other divides structures into three categories: 0 < 10,

10 < 0 < 30', and 30 < 0 [7]. Thus, a 5 on 12 pitch falls near the middle of the second

category. The wide range of roof slopes covered in this category certainly covers the

majority of typical site-built homes. Given the sparseness of data with which to validate

separate models, a single representative roof pitch is assigned to the entire population of

SFRs in lieu of using the statistics in Figure 3-2 to determine what population of Florida

homes should be modeled with separate values of roof pitch.

Characterization of Manufactured Homes

The common structural types presented in Tables 3-1 through 3-5 represent the

most prevalent site-built homes in Florida. A similar categorization cannot be made for

manufactured homes, given the lack of information about these residences on regional

basis. Instead, three types of manufactured home are used for the entire state. The two

models representing typical modem manufactured homes are referred to as MH 1 and

MH 2 for singlewide and doublewide homes, respectively. Additionally, a separate

model, MH-pre, is created to represent older manufactured homes that pre-date the

changes in building requirements for these homes enacted in 1975. All three are modeled

with gable roofs, in accordance with NAHB Research Center findings that 97% of

manufactured home roofs in the United States are gable type [33].

The national distribution of typical roof pitch values for manufactured homes,

taken from the NAHB Research Center's 1998 report, is presented in Figure 3-3, where

the numerator and denominator represent the number of inches of rise and run,









respectively. Using this national information, a pitch of 4 on 12, corresponding to a roof

slope of approximately 18 ', is selected to be most representative of the population of

manufactured homes in Florida. For the same reasons discussed in the site-built homes

section, the roof pitch is selected such that a representative value is applied to population

of manufactured homes across the state.

7/12 and up,
3%
flat up to 1/24,
5/12 and 6/12, 2%
9%




\





up to 4/12, 86%

Figure 3-3. Distribution of manufactured home roof pitch values according to the
National Association of Home Builders Research Center.

Building Component Selection

It would be impractical and inefficient to model every possible structural

component in each of the building types identified in the previous section. Post-damage

investigation reports are used to select building components common to all of the

structural types that are susceptible to wind damage. In this manner, all of the most

commonly observed forms of damage are incorporated into the simulation model, and the

results will be comparable across residential classifications.










In a 1993 report detailing Hurricane Andrew damage, the three most critical home

characteristics were the protection of openings (windows and doors), type of roof

covering, and roof sheathing attachment [19]. Additional post-damage reports and

investigations indicate that a reasonable list of wind damage-prone components for

typically constructed gable roof residential structures includes the roof covering, roof

sheathing, roof-to-wall connections, wall systems, and openings [16-24]. Given this

information, the structural building components selected for modeling site-built homes in

the simulation engine are (from top to bottom, not by order of importance) roof covering,

roof sheathing, roof-to-wall connections, walls, and openings. These broadly defined

components are depicted in Figure 3-4. Each of the structural types in Table 3-5 are

modeled based on the capacities of these components. Differences among models of the

various structural types come from the definitions of capacity, load paths, failure mode,

and wind loading. For example, concrete block and wood frame home models both

include wall components, though the failure mechanisms and capacities of these systems

differ. Also, wind loading differs from hip to gable roofs, though the roof cover capacity

is defined as the same.

Roof Cover Roof Sheathing
Roof Cover



Roof to Wall
Connections




Walls
Openings

Figure 3-4. Structural components selected for modeling in the hurricane damage-
prediction simulation engine.






43


The five components shown in Figure 3-4 are also used on the manufactured home

model, with the addition of tie-down anchors. Further details concerning the building

components (specifically the wind loads applied during simulated hurricane events and

the resistance capacity of each component) are discussed in Chapters 4 and 5. The

method by which the simulation engine uses this information to predict probabilistic

damage information for each type of structure is detailed in Chapter 6. Validation of the

methods using available data from Hurricane Andrew is presented in Chapter 7.














CHAPTER 4
STRUCTURAL WIND LOADS FOR TYPICAL

This chapter details the loads applied to simulate an extreme wind event on a

typical residential structure. The load cases described here are used in the Monte Carlo

simulation engine discussed in Chapter 6 to predict the vulnerability of typical Florida

homes to structural damage. These loading conditions are not intended to represent

design levels. Instead, load values are selected to best represent the pressure or uplift

acting on each component of the home during an extreme wind event, such as a hurricane

or tropical storm. The preferred method would be to use wind tunnel data to accurately

model the spatial and temporal characteristics of the pressure coefficient on the surface of

the building as a function of the wind direction. As discussed in Chapter 2, however, the

current body of wind tunnel test data does not support the use of laboratory generated

surface pressure characteristics on typical Florida residences. This conclusion is not at all

surprising. Wind tunnel tests have been conducted successfully over the years to

determine the envelope of worst-case loads for appropriate wind load codification. The

probabilistic character of surface pressures was not investigated to the level of detail

necessary to randomly generate appropriately scaled and correlated surface pressures on

all sides of a structure during a hurricane event. Inclusion of non-Gaussian behavior and

correlation between surface pressures in design loads is a promising topic that is currently

being investigated [9, 10], and it might be possible to incorporate this data into the

developed model at a later date. At this time, however, appropriate wind loads for each

building component must be determined from the existing body of data, which includes









current wind load design provisions, wind tunnel data, and full scale data sets described

in Chapter 2.

Engineering judgment, based on this supporting body of information, must be used

to select the most appropriate external and internal pressures to use in the calculation of

event-specific wind loads for building components. For this reason, the wind loads

selected for each building component are based on a modified version of the 1998

Minimum Design Loads for Buildings and Other Structures (ASCE 7-98) code

provisions. Changes made to the code provisions include modifying the equations used to

calculate surface pressures, re-mapping the pressure coefficient zones on the roof surface

as a function of the wind direction, and recalculating the internal pressure after initial

damage has occurred. Details of these modifications to the code provisions for the

purpose of representing storm event loads are discussed in the first section of this chapter.

Following that is a discussion of the application of the modified code provisions on the

structural components of typical Florida homes. Load conditions placed on roof cover,

roof sheathing, roof-to-wall connections, walls, openings, and tie-down anchors (on

manufactured homes only) for the purpose of simulating extreme wind events are

identified. A summary table of the wind load conditions applied during the structural

damage simulation engine is provided at the end of the chapter.

Use and Modification of the ASCE 7-98 Code Provisions to Represent Load
Conditions during Extreme Wind Events

Wind loads used for the prediction of structural damage in the simulation model

must represent surface pressures acting on each component during an extreme wind

event. They should not match design pressures that envelope the worst-case scenarios,

but should instead be dependent on the direction of the wind, and representative of the









pressure at a given moment in time. With this in mind, the load cases for the structural

damage simulation engine are generated by removing the conservativism incorporated in

the ASCE 7-98 code provisions and by changing the external pressure coefficient zones

such that the map of pressures on the roof surface is dependent on the wind direction.

Modifications to Surface Pressure Equations

Wind pressures on the surface of simulated homes are generated using modified

versions of the ASCE 7 design wind pressure equations discussed previously in Chapter

2. Equations 2-11 and 2-12, for calculation of the velocity pressure at mean roof height,

qh, and the design pressure, p, are reprinted here for clarity. The value 0.00256 in

Equation 2-11 is a function of the density of air in English units, Kh is a terrain exposure

coefficient, K, is a topographic effect factor to account for speed up over hills, Kd is a

directionality factor, Vis the design wind speed, and I is the importance factor. Equation

2-12 illustrates the calculation ofp from qh, where Cp and Cp, are the external and internal

pressure coefficients, respectively, and G is the gust factor.

qh = 0.00256KhKtKdV21 (2-11)

p =qh [GC, -GCp, (2-12)

Three of the four factors in Equation 2-11 are removed in the development of the

equation for use in the simulation routine. The importance factor, I, is discarded because

it is used to scale loads according to the importance of the structure to the local

community. This factor plays a critical role in design, but does not assist in the

determination of actual loads during a hurricane event. Additionally, the directionality

factor, Kd, is removed. This factor reduces design pressures to account for the fact that

every section of the building will not be loaded to the design level at a given time. Since









the directionality of the wind will be explicitly accounted for in the re-mapping of the

external pressure coefficient discussed in the following section, the reduction factor Kd is

removed. Lastly, few places in the state of Florida would warrant an escarpment factor

greater than 1.0, therefore Kt is unnecessary in the current endeavor. The remaining

factor, Kh, has a prescribed value of 0.85 for low-rise structures (h < 15 ft) in open-

country terrain (Exposure C). Substituting the value of Kh into Equation 2-11 and

removing the three factors I, Kd, and Kzt; the resulting equation used to calculate the

velocity pressure in the simulation routine is provided in Equation 4-1, where Vis the

maximum 3-second gust wind associated with a particular storm or recurrence interval.

qh = 0.00256(0.85)V2 (4-1)

The time scale of 3 seconds is selected to match the design wind speeds of ASCE

7-98. Use of a different time scale would necessitate additional modifications to the

external pressure coefficients used in the simulated wind load equations. It can be

assumed that the maximum 3-second gust wind speed will occur several times over the

period of the storm, since hurricanes generally last several hours. Therefore, damage can

be assessed using this discrete value without undue concern for the length or cyclic nature

of the load application.

The safety factor embedded in the ASCE Component and Cladding (C&C) pressure

coefficients on roof surfaces was determined by experimentation to be 1.25. This number

was obtained from an unpublished study conducted by the author to compare uplift

values on a roof shape for which wind tunnel data was available, and through extensive

discussions with Dr. Emil Simiu, an expert in the field, about the codification of wind

tunnel pressures and the available damage statistics from Hurricane Andrew (personal









communication, November 2001). Assuming that the same level of risk is maintained in

the design provisions for all building components, a factor of 0.8 is added to the

calculation of surfaces pressures represented in Equation 2-12. In this manner, the

reduction factor of 0.8 is used to remove the 'safety factor' embedded in the code

provisions for load calculations. A similar procedure described in Chapter 5 is used to

factor resistance values. Factors to increase expected loads and decrease expected

resistances are necessary in the design process to account for the uncertainty of each and

reduce the risk of failure. Removal of these factors is necessary such that 'true' loads

during generated wind events can be compared to probabilistically 'true' capacities in the

process of predicting of structural damage vulnerability.

The application of the 0.8 factor to remove the built in safety value in the code

provisions yields Equation 4-2. Together, Equations 4-1 and 4-2 are the basis for all wind

load calculations used for structural damage prediction in the Florida Department of

Financial Services sponsored Public Loss Hurricane Projection Model.

p = qh(0.8)[GCp GCp (4-2)

Use and Modifications to External Pressure Coefficients

External pressure coefficients in the ASCE 7-98 provisions include both Main

Wind Force Resistance System (MWFRS) coefficients and Component and Cladding

(C&C) coefficients. For brevity's sake, a full description of the design process is not

provided in this document. Simply put, the MWFRS loads are applied to the structure as

a unit (to provide checks for items such as diaphragm shear walls), while the C&C loads

are applied to individual members (for single unit capacity checks). It is important to

note, however, that both provisions (MWFRS and C&C) must be satisfied in the design.









The structural damage-prediction model uses a combination of these two provisions to

best represent the load cases on modeled components during hurricane winds. Table 4-5,

at the end of the chapter, provides a summary table of the load conditions (MWFRS or

C&C) applied to each modeled component during the simulation routine. This section

details the MWFRS and C&C external pressure coefficients taken from the ASCE 7-98

provisions and the modifications made for use in the damage-prediction model.

Main Wind Force Resisting System external pressure coefficients

The ASCE 7-98 MWFRS provisions are wind-direction dependent, and require no

modification for use in Equation 4-2. Values for the external pressure coefficient, Cp,

used in Equation 4-2 for MWFRS conditions are provided in Tables 4-1 and 4-2. To each

value, a gust factor, G, of 0.85 is applied to obtain GCp in Equation 4-2. The location of

each pressure zone is provided in Figure 4-1, taken directly from ASCE 7-98. Values for

Case A are interpolated for two roof pitches (5 on 12 for site-built homes and 4 on 12 for

manufactures homes) from the values provided in ASCE 7-98. The characteristic

dimension, a, is the lesser of 10% of the smallest horizontal dimension and 40% of the

mean roof height, but not less than 4% of the smallest horizontal dimension or 3 feet [7].

Table 4-1. Zones 1-6 MWFRS pressure coefficients
MWFRS Pressure Zones (shown in Figure 4-1)
1 2 3 4 5 6
Case A for 5 on 12 0.538 -0.456 -0.467 -0.414 NA NA
Case A for 4 on 12 0.516 -0.690 -0.469 -0.415 NA NA
Case B (all roof pitches) -0.450 -0.690 -0.370 -0.450 0.400 -0.290

Table 4-2. Zones 1E-6E MWFRS pressure coefficients
MWFRS Pressure Zones (shown in Figure 4-1)
1E 2E 3E 4E 5E 6E
Case A for 5 on 12 0.771 -0.722 -0.648 -0.598 NA NA
Case A for 4 on 12 0.780 -1.070 -0.673 -0.609 NA NA
Case B (all roof pitches) -0.480 -1.070 -0.530 -0.480 0.610 -0.430


















CASE A


Figure 4-1. MWFRS zones. A) Winds perpendicular to the ridgeline through cornering
winds. B) Cornering winds through winds parallel to the ridgeline. (ASCE
7-98 Standard, Minimum Design Loads for Buildings and Other Structures,
American Society of Civil Engineers, New York, NY. Fig 6-4, p. 43)

Component and Cladding external pressure coefficients

Modifying the C&C pressure coefficients on the roof surface and walls to account

for the directional nature of wind pressures is accomplished by manipulating the mapped

zones to represent observed damage patterns and wind tunnel pressure investigation

results. The ASCE 7-98 pressure zones for the design of roof cladding on gable and hip

roofs are shown in Figure 4-2. Zone 3 (the highest magnitude of suction) is applied at

each comer, and Zone 2 is applied to locations of discontinuity on the roof surface. Zone

1 (the lowest magnitude of suction) is applied to all areas not covered by Zones 2 and 3.

Figure 4-3 indicates the pressure zones for the wall surfaces. Zone 5 (the highest in

magnitude) is applied to each comer and Zone 4 is applied to all other surfaces.

The ASCE pressure zones provided in Figures 4-2 and 4-3 envelope the worst-case

scenarios for the life of the structure. Using these provisions, the designer is not required

to determine which way the building will face relative to the most likely wind direction.

Components in all corners are designed to the same wind pressures. Structures will not

experience pressures in this manner during actual loading conditions, however.






51









1 B









Buildings and Other Structures, American Society of Civil Engineers, New
r, Fig 6 p




























Figure 4-3. C&C wall pressure zones. (ASCE 7-98 Standard, Minimum Design Loads
I I i :
I I J I









CFigure 4-2 C&l roof pressure zones. A) G abe roof zo es of winde diagram.


York, NY. Fig 6-5B, p. 46)















Figure 4-3. C&C wall pressure zones. (ASCE 7-98 Standard, Minimum Design Loads
for Buildings and Other Structures, American Society of Civil Engineers,
New York, NY. Fig 6-5A, p. 44)

Engineering judgment is required to manipulate the map of design pressures into a

layout that is dependent on the wind direction. Modifications to the ASCE 7-98

Component and Cladding roof pressure zones for varying angles of wind are shown in

Figures 4-4 through 4-6. The characteristic dimensions for zone width, a, remains as

described in ASCE 7-98, with the exception of the cornering wind case. The width of

Zone 3 and the width of Zone 2 over much of the windward side are increased to 2a for








cornering winds on gable roof structures. Figures 4-4 through 4-6 are not drawn to scale.
Modifications to the wall pressure zone layout (no figure) include removing the edge
zone on the windward and leeward walls to apply a single uniform pressure across the
face of the wall. The leading edge zone on the side walls is maintained, and the trailing
edge zone is removed.

Wind Wind

ASCE 7-98 Zone 3
D ASCE 7-98 Zone 2
D ASCE 7-98 Zone 1
_______A /____B_____ B
Figure 4-4. Roof pressure zones for winds perpendicular to the ridgeline. A) Gable roof
zones. B) Hip roof zones.
Wind Wind


ASCE 7-98 Zone 3
D ASCE 7-98 Zone 2
D ASCE 7-98 Zone 1
A B_______


Figure 4-5. Roof pressure zones for winds parallel to the ridgeline. A) Gable roofs
zones. B) Hip roof zones.



ASCE 7-98 Zone 3
D ASCE 7-98 Zone 2
D ASCE 7-98 Zone 1
A B
SWind -Wind

Figure 4-6. Roof pressure zones for cornering winds. A) Gable roofs zones. B) Hip roof
zones.


B









In the ASCE design provisions, the gust factor and external pressure coefficient for

C&C loads are combined into one term, GCp, which is dependent on the effective wind

area of the component being designed and, in the case of roof components, on the roof

pitch as well. The effective wind area for components is defined by ASCE as the

maximum of two possible values: the tributary area for the component in question, and

the span length times an effective width of one-third of the span length. The effective

wind area for fasteners is the worst-case tributary area for an individual fastener [7]. As

the effective wind area decreases, the magnitude of the external pressure coefficient

increases, providing smaller areas with larger magnitude load cases and larger areas with

smaller magnitude pressures. Since the entire region of a large tributary area is not likely

to be loaded to maximum capacity at the same time, a uniform design load of smaller

capacity is applied to the surface of large areas. In the structural damage simulation

program, efforts have already been taken to eliminate the conservatism or 'safety factor'

built into the design code, and to map the pressure coefficients such that the layout is

dependent on the wind direction. Given this approach, and the reliance of most of the

modeled components on fasteners (e.g. sheathing), the values taken from the ASCE 7-98

provisions for C&C external pressure coefficients are those with an effective wind area of

10 ft2 or less. Values for roof zone pressure coefficients are provided in Table 4-3, and

values for wall surfaces are provided in Table 4-4. The modified location of roof pressure

zones is given in Figures 4-4 through 4-6.

Table 4-3. Roof zone C&C pressure coefficient values for selected roof pitches
GC,
Zone 1 -0.9
Zone 2 -2.1
Zone 3 -2.1









Table 4-4. Wall C&C pressure coefficient values
GC,
Windward Wall 1.0
Side Wall Leading Edge -1.4
(distance of a from the comer) 1
Side Wall -1.1
Leeward wall -0.8

Use and Modifications to Internal Pressure Coefficients

Since extremely low barometric pressures mark hurricane events, the internal

pressure in modeled homes is assumed to be greater than the outside pressure before any

damage occurs to the structure. With this rationale, the default value of internal pressure

assigned to all homes in the structural damage simulation model is obtained by setting the

internal pressure coefficient in Equation 4-2 equal to +0.18, the value provided in ASCE

7-98 for enclosed structures. As described in Chapter 6 of this document, initial failure

checks are conducted to determine whether individual windows, doors, pieces of roof

sheathing, or shear walls fail. A subsequent internal pressure, dependent on the level of

initial damage to the home, is calculated as the weighted average of the pressure at the

location of broken doors and windows, to include the garage door. This value is used in

the final round of failure checks, described in greater detail in Chapter 6.

Application of the Modified ASCE 7-98 Code Provisions to Produce Extreme Wind
Event Load Conditions on Selected Building Components

The modified external and internal pressure coefficients discussed in the previous

section are used with Equations 4-1 and 4-2 to generate the load conditions which

simulate the occurrence of an extreme wind event on both site-built and manufactured

Florida homes. In this section, the selection of modified external pressure coefficients for

load conditions placed on roof cover, roof sheathing, roof-to-wall connections, walls,

openings, and tie-down anchors (on manufactured homes only) are specifically identified.









Resistances to these wind loads are discussed in Chapter 5, and the order of application

and failure checking conducted by the simulation engine are detailed in Chapter 6.

Roof Cover and Roof Sheathing Loads

Roof covering materials and roof sheathing panels are treated as cladding during

the structural damage simulation. The most likely sheathing panel arrangement for each

of the models described in Table 3-3 and for both of the manufactured home models is

obtained by starting with a full sheathing panel on one of the lowest corners, and placing

additional panels in an offset pattern. Given the amount of uncertainty in roof cover

loading and wind resistance, any efforts to define the area of an individual roof cover unit

would not add to the accuracy of the damage prediction results. In light of this

information, a section of roof cover is assigned to each sheathing panel on the drawn roof

sheathing arrangement. The individual sections of roof cover thus have the same square

footage as the underlying sheathing panels. The resulting model-specific roof layouts are

used to obtain aggregate external pressure coefficients for each individual piece of

sheathing and roof cover at each wind angle using the pressure coefficient maps of

Figures 4-4 through 4-6. Reasons for using the aggregate pressure over point pressures

are discussed in the section of Chapter 5 devoted to the resistance capacity of roof cover

and sheathing.

Wind loads for each piece of roof sheathing are obtained by using the aggregate

external pressure from the model-specific layout with the appropriate internal pressure

coefficient for the state of the building in Equation 4-2. Since the roof cover is attached to

the outside surface of the roof sheathing, it is not subject to the same internal pressure

fluctuations. In order to best represent the load case that would occur during an actual

storm event, the wind loads for roof cover areas are obtained by using the aggregate









external pressure from the model-specific roof layout and an internal pressure coefficient

of zero in Equation 4-2.

Roof-to-Wall Connection Loads

Roof-to-wall connections are modeled in tension, using the dead load and wind-

induced uplift from the roofing system. As described further in Chapter 6, these

connections are one of the last building components checked for failure. The loads

applied result from the remaining roof sheathing. In this manner, overloaded roof

sheathing panels are assumed to fail before passing the overloaded condition to the

trusses. An assumed dead load of 10 psf (which includes the weight of typical roof cover,

roof sheathing, suspended ceiling, insulation, and ductwork) is applied to each sheathing

panel that remains on the roof surface after the initial failure check. Wind uplift is

obtained from the loads previously described for the sheathing panels, and individual

connection loads are calculated using a tributary area approach, assuming that trusses are

spaced at 2 feet on center in most homes. Gable end trusses are assumed to have a total of

eight gable end type connectors. Loads on the two end trusses for gable roof structures

are equally distributed to these connections.

Roof-to-wall connections are the only building component in the developed Public

Loss Hurricane Projection Model where the redistribution of load is applied.

Redistribution is not appropriate for other components, but is used here to capture the

failure mechanism by which the entire roof separates from the walls [19]. Once a roof-to-

wall connection fails, the load is redistributed to the surrounding connections until the

system reaches a point of equilibrium. Additional details of this method are provided in

Chapter 6.









Wall Loads

Walls on site-built homes are modeled in shear, uplift, and bending. The total shear

for each wall is computed using the MWFRS pressure coefficients from ASCE 7-98

provided in Tables 4-1 and 4-2 and the standard practice of modeling the roof diaphragm

as a simply supported beam. In this manner, the surface pressures on opposite sides of the

house can be multiplied by half of the building height to produce a distributed load on the

length of the roof diaphragm beam. Shear loads in each supporting wall are the reactions

to this distributed load. This method is shown in Figure 4-7. During cornering winds,

both cases are applied independently.

V2


V2 V1



A B
V1


Figure 4-7. Method of determining shear wall loads from MWFRS pressures. A) Winds
perpendicular to the ridgeline through cornering winds. B) Cornering winds
through winds parallel to the ridgeline.

The uplift forces on each wall are obtained per foot of wall by averaging the total

uplift from the attached roof-to-wall connections over the length of the wall. Lateral

pressures for wall surfaces are obtained using Equation 4-2 with C&C coefficients given

in Table 4-4 and the appropriate internal pressure coefficient for the building. From these

lateral pressures, the bending moments per foot of wall for concrete block walls are

obtained with the assumption of simple supports at the roof and floor. This assumption is

maintained unless more than half of the roof-to-wall connections fail, at which point the









bending moment is amplified by a factor of 2.8. This factor (70% of the multiplier

between simply supported and cantilevered moments) is selected for use over the pure

cantilever condition since the wall would retain some support from the side and interior

walls, even if the roof-to-wall connections have failed.

Wood framed walls exhibit different behavior when confronted with out of plane

load conditions; therefore the bending moment is not calculated for these types of walls.

Instead, the lateral force at the wall connection that results from C&C surface pressures

on the wall is used. In this procedure, the presence of at least one interior wall on each of

the four perimeter walls is assumed. Under this premise, the tributary area of pressure

transferred directly into the lateral wall connections for each of the four perimeter walls is

represented by the two trapezoids shown in Figure 4-8. The tributary area represented in

Figure 4-8 relies on the assumption that the rest of the building is undamaged. This

assumption is maintained unless more than half of the roof-to-wall connections on the

depicted wall fail. After significant roof-to-wall connection damage, the tributary area is

adjusted to the two triangles shown in Figure 4-9.




/2h
V ___________________

Figure 4-8. Tributary area for C&C pressures transferred into lateral connections on
wood frame walls.







Figure 4-9. Tributary area after significant roof-to-wall connection damage for C&C
pressures transferred into lateral connections on wood frame walls.









Surface pressures are calculated using Equation 4-2 with the C&C coefficients for

walls given in Table 4-4 and the appropriate internal pressure coefficient for state of the

building. The total load calculated by applying these surface pressures to the tributary

area shown in Figure 4-8 or Figure 4-9 for each of the four perimeter walls is distributed

evenly to all of the lateral connections at the base of each wall.

An additional wall load check for wood framed walls and the primary check for

manufactured home walls is the potential loss of wall sheathing. Aggregate panel loads

for individual pieces of wall sheathing are obtained in much the same way as roof

sheathing loads. A length-specific layout is obtained for each wall. For wood frame walls,

the layout is obtained by starting at one end with a full-size upright (4 ft wide by 8 ft tall)

sheathing panel, and adding upright sheathing panels along the length of the wall, side by

side. For manufactured homes, the layout is obtained by stacking typically sized pieces of

vinyl siding along the wall length. The surface pressures are calculated using Equation 4-

2 with appropriate Component and Cladding external pressure coefficients, and the

internal pressure coefficient for the particular building.

Load Conditions for Openings

This category covers a wide variety of building components. Incorporated into the

simulation program are doors, garage doors, and windows. Each modeled house is

assumed to have one front and one back door. The load applied to each is the surface

pressure calculated using Equation 4-2 with the appropriate C&C pressure coefficient

from Table 4-4, and the internal pressure coefficient for the current state of the building.

Additionally, houses with garages are assumed to have the garage door on the front wall.

The surface pressure applied to the garage door is the same as the pressure applied to the

front door.









Unprotected windows are loaded in two distinct ways: pressure loads and impact

loads. The pressure load scenario is similar to that described for the doors. Surface

pressures at the window locations are obtained by using Equation 4-2 with the

appropriate external C&C pressure coefficient from Table 4-4, and the internal pressure

coefficient for the building.

Impact loads to windows are caused by windborne debris from neighboring homes.

To model this behavior, an equation based on the cumulative exponential distribution

(which describes the likelihood of rare and unrelated discrete events) is used to predict

missile strikes. In Equation 4-3 below, pD (V) is the probability of impact causing a

broken opening, given the 3-second maximum gust, V. A represents the fraction of

potential missile objects (e.g., shingles) in the air. NA is the total number of available

missile objects (e.g., number of shingles on the nearest house). B is the fraction of

airborne missiles that hit the house, C is the fraction of the impact wall that is glass, and

D is the probability that the impacting missiles have momentum above damage threshold.

PD(V)= 1- exp[-A *NA *B* C D] (4-3)

Equation 4-3 can be used to predict the likelihood of impact for several scenarios.

This equation can eventually be used to predict the likelihood of impact by several

different sources of debris (e.g., shingles, wood studs, and grapefruit). These varied

results could be superimposed to determine the final tally of total impacts. With the

information currently available, Equation 4-3 is used to determine the likelihood of

windborne debris impact on the windows from any potential missile. Specific choices for

each parameter and the methods by which these parameters could be honed in future

work are discussed in the following paragraphs.







61


Parameters that define objects in the air include NA and A. The total number of


available missile objects, NA, is related to the type and density of the building population


around the modeled house. The current selection for this number is an empirical choice of


100. This number is expected to change regionally in future iterations of the Public Loss


Hurricane Projection Model, as the results from the initial model guide improvements in


future work. A is related to the capacities of the upwind building components that will


become windbome debris, and is thus a function of peak wind speed. This parameter is


modeled as a Gaussian cumulative density function (CDF). That is, at low peak gust wind


speeds, relatively few of the available missiles are torn off upwind buildings to become


windborne debris. As wind speed increases, more available debris is torn off upwind


structures at a faster rate, until the function levels off at 1.0 (at which point 100% of


potential missiles are in the air). In the current version of the structural damage-prediction


model, the Gaussian CDF defining A has a mean value at a peak 3-second gust of 135


mph, and a standard deviation of 15 mph. This function is shown in Figure 4-10.


09
08
07
06
S05
04
03
02
01

50 100 150 200 250
3 Second Gust Wind Speed, V


Figure 4-10. Values of the parameterA used in the determination of missile impact

Parameter B in Equation 4-3 determines how many of the missiles in the air


actually strike the modeled home. This parameter is dependent on several factors,











including proximity of the missile starting point and the ability of the missile to stay

airborne (which is a function of wind speed and missile type). Engineering judgment

indicates that missiles will fly further and stay in the air longer with increasing wind

speeds. For lack of better information, a linear function describing the parameter B is

selected to have a value of zero (no airborne missiles striking the building) at 50 mph 3-

second gusts and a value of 0.40 at 250 mph 3-second peak gusts. Values of B are

provided in Figure 4-11.

04
0 35
03
0 25
S 02
015
01
0 05

50 100 150 200 250
3 Second Gust Wind Speed, V


Figure 4-11. Values of the parameter B used in the determination of missile impact

Of the missiles striking the house, a fraction will hit windows (rather than other

surfaces). This value is described by the parameter C, which defines the fraction of the

windward wall space that is occupied by unprotected glass windows. In the current

structural damage-prediction model, the windward wall space is the area of one of the

perimeter walls except for the case of cornering winds, when two of the walls are

vulnerable to missile strike. As described later in Chapter 5, windows on the modeled

houses are categorized in four sizes. For accounting purposes inside the structural

damage model, the four sizes of windows are treated independently at this point in the

development of the missile impact equation. A value of C for each size of window is

calculated as the area of that type of window divided by the wall space vulnerable to









impact. The probability of impact, pD (V), generated from using these values of C in

Equation 4-3 is the likelihood that a window of a certain size will be impacted and broken

by a windborne debris missile, given the peak 3-second gust wind speed, V.

The parameter that determines whether the striking missile will cause the window

to break is D. This value is dependent on the momentum of the impacting missile and the

resistance capacity of the window. Shingles, a numerous and readily available windborne

missile type, are used to generate a function for the parameter D. The momentum of a

windborne object, p,, is defined by Equation 4-4, where m is the mass of the object, Vis

the wind speed, and R is a reduction factor. The value of V(R) is then the wind speed at

which the missile object is traveling. (Note that the subscript m for momentum is added

by the author to the commonly used variable p to distinguish between momentum and

pressure.)

p= mV(R) (4-4)

Conservatively assuming that a typical shingle weighs 0.06 lbs (a mass of 0.03 kg),

and that the maximum value of R for single shaped missiles 0.64 [34], one can determine

the momentum of a shingle moving in a wind gust of 110 mph. (49 m/s) to be 0.944 kg-

m/s. Given the impact resistance capacity of typical glass windows to be 0.025 kg-m/s

[3], the momentum of a windborne shingle in a 110 mph 3-second gust wind event would

exceed the capacity of typical unprotected window by a factor of approximately 37. It

should be noted that additional reduction factors might apply, since the shingle might

strike at an angle or not reach terminal velocity before hitting the window. However,

these additional reduction factors will not overcome the significant difference between

the missile's momentum and the resistance capacity of a typical window. Because











Equation 4-3 encompasses all types of missiles, the shingle example is used to determine


likely thresholds for the parameter D, and not specifically used to generate D as a


function of wind speed. The values for D used in the current structural damage simulation


program are taken from a Gaussian CDF generated using a mean value of 70 mph and a


standard deviation of 10 mph. Using this function, the likelihood of breakage, given the


fact that a missile has impacted the window, is provided in Figure 4-12.


09
08
07
06
05
04
03
02
01

50 100 150 200 250
3 Second Gust Wind Speed, V


Figure 4-12. Values of the parameter D used in the determination of missile impact

Using the parameters described, the likelihood of an impact causing breakage


during a specific wind event represented by a 3-second maximum gust is determined with


Equation 4-3. Values are dependent on the size of window and the size of the windward


wall. The eight possible angles of wind exposure create three possible windward wall


scenarios: short wall facing the wind, long wall facing the wind, and cornering winds, in


which one short and one long wall are both vulnerable to missile impact. The function


PD (V) must be generated for each of the four window sizes during each windward wall


scenario, for a total of 12 functions per modeled building. As an example, pD (V) for a


medium sized window on the short side of the concrete block, gable roof house in the


Central Region of Florida is provided in Figure 4-13.












09
08
07
S06
o05
04
003
02
-01
0
50 100 150 200 250
3 Second Gust Wind Speed, V


Figure 4-13. Probability of missile strike causing breakage of a medium (3.5 x 5 ft)
window on a 44 ft long windward wall.

Load Conditions for Tie-Down Anchors

The load cases described in previous sections apply to both site-built and


manufactured residences. Because of the differences in foundations, however, two loads


cases are unique to manufactured housing. These are sliding and overturning loads. Both


cases are calculated using MWFRS pressure coefficients provided in Tables 4-1 and 4-2,


and located in Figure 4-1. The overall lateral sliding force for a particular manufactured


home is calculated as the vector sum of the resultant wall surface loads. This force will be


resisted by the anchors as well as the friction between the house and foundation piles.


The overturning moment is calculated about the leeward wall support pier and is resisted


by the assumed weight of the house as well as the anchor system. Discussion of the


resistance to both of these load conditions is described in Chapter 5. Additional details on


the overturning and sliding failure checks are provided in Chapter 6.


Summary of Wind Load Conditions Used in the Simulation Engine

A summary of the wind load conditions applied to individual components during


simulation is provided in Table 4-5. Sources described as MWFRS and C&C refer to the


modified versions of the ASCE 7-98 provisions for Main Wind Force Resistance System









and Component and Cladding, respectively. Resistance values for each condition are

described in Chapter 5 and the process by which the simulation engine applies and checks

these conditions is detailed in Chapter 6.

Table 4-5. Summary of load conditions applied to simulate extreme wind events
Limit Source of
Building Component State Loads Additional Notes
Roof Cover Separation C&C Pressure coefficients
or pull off aggregated over the area


Roof Sheathing


Roof-to-Wall Connections


Concrete
Block


Wood Frame


Manufactured
Homes


Separation
or pull off


C&C


Tension Roof
sheathing
Shear wall MWFRS
Combined C&C


uplift and
bending
Shear wall
Uplift


Lateral
loading
Sheathing
pull off

Sheathing
pull off


of the underlying
sheathing panel; no
internal pressure applied
Pressure coefficients
aggregated over the area
of the individual panel
Dead plus wind; load
redistribution applied

Uplift Roof-to-Wall
Connections
Bending C&C


MWFRS
Roof-to-
Wall
Connections
C&C


C&C


C&C


Pressure coefficients
aggregated over the area
of the individual panel
Pressure coefficients
aggregated over the area
of the individual panel


Openings Doors and Over- C&C
Garage Doors pressure
Windows Over- C&C
pressure
Impact PD (V) Not an applied load; a
damage probability of impact
causing breakage as a
function of wind speed


Tie-Down Anchors


Overturn
Sliding


MWFRS
MWFRS


Manufactured housing
Manufactured housing


Walls














CHAPTER 5
PROBABILISTIC WIND RESISTANCE CAPACITIES FOR
RESIDENTIAL DWELLING COMPONENTS

This chapter describes the resistance capacities selected for use in the structural

damage simulation model. Capacities typical of the building components in Florida

homes are selected from available literature and manufacturer data for each load case

described in Chapter 4. Using this information, truncated Gaussian distributions are

created to represent populations of typical building material resistances to the load cases

identified in Table 4-5. These resistance distributions will be used in conjunction with the

load values discussed in Chapter 4 to determine whether individual structural members

fail when subjected to extreme wind loading. The operational flow of the simulation

routine determining structural damage to typical Florida homes is provided in Chapter 6.

Results and validation of the process are discussed in Chapter 7.

In this chapter, the details and selection process for the distribution of resistance

values for each building component load case are provided. The first section of the

chapter describes choices and arguments common to the selection of all building

component resistance values. Following this introductory discussion are sections

detailing the selected capacities for roof cover, roof sheathing, roof-to-wall connections,

walls, openings, and tie-down anchors. The chapter is divided into a section detailing the

resistance capacities of typical site-built homes and a latter section providing information

for manufactured homes. At the end of the chapter, Tables 5-3 and 5-4 provide a

summary of all resistance values incorporated in the structural damage simulation model.









Fundamental Concepts Applied During the Selection of Load Resistance Values

Resistance values described in this chapter represent the un-factored ability of each

structural component to withstand loads induced by extreme wind events. As described in

Chapter 4, the conservative factor built into the wind loading provisions of ASCE 7 was

removed to determine a 'true' wind loading condition. In this chapter, the safety factors

from manufacturer's recommendations are removed to determine 'true' resistances. In

this manner, the simulation program seeks to accurately assess the vulnerability of

typically constructed homes to structural wind damage. If the safety factors were not

removed, the program would provide the level of risk inherent in the current codification

process, not the level of potential structural damage.

Available literature and manufacturer's data are used to determine appropriate

probability density functions for component resistances. Typically, the mean failure value

for each component is obtained from available information and the coefficient of

variation (COV) is determined through engineering judgment. A measure of the spread of

the distribution, the COVis the standard deviation divided by the mean. The effect of

varying the COVis shown in Figure 5-1. Each plot in the figure shows a Gaussian

(normal) distribution with a mean of 100 units. The x-axis represents differing values of

units, while the y-axis represents the likelihood of occurrence. The area under each curve

is unity, though the peak value of the distribution with a COV of 0.2 is nearly twice that

of the distribution with a mean of 0.4. The distribution with a COV of 0.2 is more closely

centered on the mean value. Thus, there is a higher probability of selecting a value away

from the mean (between 0 and 50, for example) using the distribution with a COV of 0.4,

though both of the plotted distributions have the same mean value of 100 units.





























150 200 250 300


Figure 5-1. Gaussian distributions with a mean of 100 units and varying coefficients of
variation.

Gaussian distributions similar to the ones depicted in Figure 5-1 are used to model

populations of component capacities. Due to manufacturing quality control processes,

individual components are likely to have resistance capacities which follow a lognormal

distribution similar to the one shown in Figure 5-2. As compared to a Gaussian

distribution with the same mean and COV, the lognormal distribution provides a reduced

likelihood of occurrence in the low resistance tail region and a slightly greater likelihood

of occurrence in the high resistance tail region. These characteristics are representative of

manufacturing processes where the minimum allowable capacity is a quality control

measure. Gaussian distributions are selected over lognormal and other alternate

distribution choices, however, to incorporate additional variables. The variation in type,

quality, size, and installation for building components on homes of differing plan

dimensions increases the variety of the population under consideration. As the number of










variables increases, the central limit theorem leads to the conclusion that the distribution

which would best characterize each capacity is Gaussian.


0.025
-- Lognormal
---- Gaussian

0.02



0.015


0.01



0.005


0
-50 0 50 100 150 200 250


Figure 5-2. Lognormal vs. Gaussian for a mean of 100 units and coefficient of variation
of 0.2

Chapter 6 describes the process by which the Gaussian values are sampled and used

to simulate individual homes, while the characteristics of each distribution are described

in the following sections of this chapter. As demonstrated by the distribution in Figure 5-

1 with a COV of 0.4, this can lead to the possibility of selecting a value less than zero. To

avoid the occurrence of physically impossible or impractical resistance values, truncation

is applied to each of the capacity distributions described in the following sections.

Sampled resistance values are bound within two standard deviations of the mean. The

application of these upper and lower limits results in a distribution similar to the example

shown in Figure 5-3 for a mean of 100 units and a CO V of 0.4.










0.012


0.01


0.008


0.006


0.004


0.002


-50 0 50 100 150 200 250


Figure 5-3. Truncated Gaussian distribution with a mean of 100 units and a COV of 0.4.

Site-Built Home Resistance Values

Building components modeled for typical site-built homes consist of roof covering,

roof sheathing, roof-to-wall connections, walls, and openings. These are depicted in

Figure 3-2. The following paragraphs detail the resistance values obtained from available

literature, manufacturer's data, and engineering judgment for each load condition

described in Table 4-5 for site-built homes. The values provided in this section are used

to characterize capacity distributions similar to the example provided in Figure 5-2. The

sampling process by which the distributions are used to create representative Florida

homes is discussed in Chapter 6.

Wind Resistance Capacity of Roof Cover on Site-Built Homes

The resistance capacity of the roof covering is the ability of the shingles or tiles to

stay attached to the roof sheathing, preventing rain water from entering and damaging the









contents. In general, there is limited information available about the uplift capacity of

shingles and tiles, in spite of the fact that loss of roof covering contributes significantly to

insurance losses. One experimental study provides an approach for estimating the wind

action on shingles, but does not predict failure loads, citing the unknown capacity of the

adhesive [35].

Factory Mutual (FM), Underwriters Laboratories (UL), and the American Society

for Testing and Materials (ASTM) have developed test methods for commercial and

residential roof coverings. Unfortunately, the tests do not provide information about the

ultimate failure capacity of these building materials, nor do they adequately represent the

conditions these components would face in hurricane events. Many use constant pressure

systems instead of using a turbulent wind condition. During the standard FM test, a

constant pressure is applied to the underside of a test specimen to simulate uplift [36].

Products that withstand the pressure for one minute without separating or delaminating

are given a rating. FM Class 1-60 indicates a 60 psf test, while FM Class 1-180 indicates

a 180 psf test. UL 580, "Standard for Tests for Uplift Resistance of Roof Assemblies,"

and UL 1897, "Standard for Uplift Tests for Roof Covering Systems," also use constant

pressure systems to determine ratings [37]. These FM and UL static tests do not

accurately simulate the wind action on the roof covering that will lead to shingle peeling

and nail pull through.

The ASTM standard testing protocol D3161, "Standard Test Method for Wind-

Resistance of Asphalt Shingles," and the UL 997, "Standard for Wind Resistance of

Prepared Roof Covering Materials," specify a horizontal wind created by a fan, but the

required wind speed is only 60 mph, far below the design wind speeds for Florida [37]. A









recent provision has been created in Dade County, Florida which is similar to D3161 and

UL 997, but uses a 110 mph fan instead of a 60 mph fan for asphalt shingle testing. Tiles

and other roofing materials, however, are still tested for Dade County approval using

static uplift tests [37].

While the Dade County provision for shingles does include a wind test using

speeds in Category II of the Saffir-Simpson scale, the test is considered a 'pass or fail'

event. That is, a product either qualifies for use in Dade County by passing the test, or

does not qualify by failing the test. The provisions do not require determination of the

actual failure capacity. Experimental data predicting the adhesive failure or nail pullout

of typical roof coverings (shingles or tiles) of average age is not currently available, and

could be the focus of a future research effort.

In the absence of experimental data, the capacity of typical residential roof

coverings is estimated from the average of two calculations. The basis of the first logical

argument is to infer that the majority of roof coverings were originally manufactured to

the 1970's era Southern Building Code Congress International (SBCCI) requirement that

cladding materials withstand an external positive or negative pressure of 25 psf. An

additional necessity for this argument is to assume that, while improvements have been

made, the fundamental manufacturing process for shingles and tiles has not changed

radically in the last few decades. Given these two assumptions, one can predict that 90%

of the roof coverings currently on the market in Florida would meet or exceed the

requirement of withstanding a 25 psf load under typical quality of workmanship in

installation. Using a Gaussian distribution to represent the failure strength of all roof

covering products used in the state of Florida, the standard distribution tables can be used









to determine the mean failure strength. Equation 5-1 provides the method of converting a

value to the standard Gaussian distribution. In this equation, x is any value in the

Gaussian distribution, / is the mean of that distribution, COV is the coefficient of

variation, and z is the value in a standard Gaussian distribution with the same likelihood

of occurrence as the value x.


z = (5-1)
COVpu

The assumptions listed above are represented by setting x equal to 25 psf and

obtaining a z value of -1.28 from the standard tables in Ang and Tang [38]. This z value

represents a location at which 90% of the products would meet or exceed the capacity. A

COV of 0.4 is selected to represent the wide variety of products and quality of

workmanship. With these values, Equation 5-1 can be rearranged to solve for a mean

failure capacity. Using the argument presented above, 51 psf would be the most

reasonable mean failure capacity for typical roof coverings.

A second argument begins with the recent Dade County uplift test for shingles,

which uses a 110 mph fan. If the fan speed is used as the design wind speed, V, in the

ASCE design pressure equations (Equations 2-11 and 2-12) with the assumptions that the

building is enclosed and in open terrain, a corresponding surface design pressure can be

calculated. Interestingly enough, the value obtained is 51 psf, the mean failure capacity of

typical roof coverings from the previous exercise. Products are required to pass this test

to be certified for use in South Florida, which means that the mean failure capacity of

roof coverings is likely to be higher than the calculated value. Assuming that 90% of

products pass the test, and again selecting a COV of 0.4 to represent the wide variety of

products and quality of workmanship, the same procedure described in the first argument









can be used to determine a possible Gaussian mean for South Florida shingles. A value of

approximately 104 psf is obtained using this argument.

Engineering judgment and knowledge of the degree of damage following Hurricane

Andrew and other past storms [15-23] indicate that the mean value from the second

argument is too high, and the mean value from the first argument is too low, for use as a

representative mean for all typical roof coverings in the state of Florida. A value that

would best represent the entire population of roof coverings (to include both shingles and

tile products, as well as old and new construction) lies between the two values. Using this

conclusion, a value of 70 psf with a COV of 0.4 is selected for the mean failure capacity

of typical roof coverings.

Wind Resistance Capacity of Roof Sheathing on Site-Built Homes

A critical component in the overall vulnerability of a residence to hurricane damage

is the ability of the sheathing to remain fastened to the trusses or rafters. A considerable

body of research has been conducted in this area in the wake of the Hurricane Andrew

damage. One such study conducted at Clemson University indicates that the capacity of

sheathing panels is best represented by treating the panel as a whole, rather than

evaluating the capacity of individual fasteners. Sheathing panel failure usually begins

with the pullout or pull-through of a single critical interior fastener, but if any fastener is

improperly installed, the failure mechanism is most likely to begin at that location,

whether it is interior or not [39]. Given the difficulty in predicting the most probable

failure location, the best means of comparing resistance to load is to use the aggregate

load on the entire sheathing panel, and compare it with average failure loads from tests of

whole panels, not just single fasteners.









Results from the study discussed in the preceding paragraph and from additional

studies provide mean failure pressures and coefficients of variation for panels of different

wood species with different fastener sizes and schedules [39-41]. Unfortunately, the

means differ significantly, and considerable uncertainty exists concerning the species of

wood and the most typical fastener type and size used in each area of Florida. A simple

arithmetic mean of the failure capacities would not necessarily best represent the building

population of the state.

As an alternative to using laboratory data, in situ data exists for a limited number of

homes in South Carolina. These homes were flood damaged during Hurricane Floyd in

1999 and were subsequently purchased for the purpose of testing and evaluation of

retrofit measures. The homes varied in age and construction. Approximately half of the

homes had planked roofs, one had oriented strand board (OSB), and the rest were

plywood. After removing one outlier (a planked roof with a high failure capacity of 450

psf) the sheathing failure test results average to value of approximately 150 psf, with the

highest value at 196 psf and the lowest at 105 psf [42]. Though the houses tested were in

South Carolina, they are fairly representative of the types and ages of construction

present in Florida. Since only a limited number of homes were tested, the COV obtained

from the eight test values is not used as the COV for a distribution representative of

typical Florida roof sheathing. Instead, a value of 0.4 is selected to account for

differences in workmanship and materials throughout the state.

Wind Resistance Capacity of Roof-to-Wall Connections on Site-Built Homes

The link between the roof system and the external walls occurs at the roof-to-wall

connections. Uplift capacities of several types of roof-to-wall connections for light frame

wood construction are available [43, 44]. The study conducted by Reed [44] further









investigated the potential for load sharing between rafter connections and found that load

sharing existed in nailed connections, but not in hurricane strap connections. No studies

have been located that investigate the possible differences in uplift capacity for roof to

masonry wall connections, though masonry structures make up a considerable portion of

the building stock, as described in Chapter 3. In the absence of test data, the uplift

capacity for typical roof-to-wall connections on masonry homes can be estimated from

manufacturer information. In order to maintain consistency between types of houses,

manufacturer's data is used for both masonry and wood frame homes.

Personal correspondence between the author and Randy Shackelford, a Simpson

Strongtie representative, provided information about the connections most frequently

used in the state of Florida and the typical factor of safety placed on the capacity

specified by the manufacturer (personal correspondence, May 2002). Roof-to-wall

connections manufactured by this company vary in uplift strength. Additionally, the same

connector has a different strength rating depending on the type of wood used in

construction. Generally, roof-to-wall connectors for both wood and masonry construction

can be assigned to one of three strength categories: high strength (which includes most

hurricane strap connections), medium, and low strength. Table 5-1 provides the values

obtained by averaging the manufacturer's rated uplift capacity of the products available

in each generalized category. Only two categories were obtained for the case of gable end

connectors on masonry walls, and the values for wood construction are obtained

assuming Spruce Pine Fir (SPF) construction. In Table 5-1, the term 'side' is used to

describe typical roof-to-wall connections at the end of a truss. These connections occur

on all four perimeter walls of a hip roof home, but only on the side walls of a gable roof









home. Gable end connections are those that connect the last truss on each end to the wall,

and occur only on gable roof homes in the simulation model.

Table 5-1. Manufacturer's uplift capacity for typical roof-to-wall connections
Connector Strength Category
Construction Connector High (lb) Medium (lb) Low (lb)
d F Side 1240 690 460
Wood Frame
Gable 1260 650 380
Side 1400 1065 700
Masoy Gable 640 225

Discussions with a Bob Carter, a Brevard County architect, (personal

correspondence, June 2003) in addition to the correspondence with the Simpson Strongtie

representative indicate that nearly all of the homes constructed in Florida over the last 15-

20 years would fall in the category of high-strength roof-to-wall connections. Based on

this information, the high-strength values in Table 5-1 are used as the manufacturer's

rated capacity for each type of home construction. Specific values for other types of

connections (such as toe-nailed connections) are not incorporated in the model at this

time, due to a lack of information concerning the distribution of connection types

throughout the state.

According the testing conducted by Simpson Strongtie, a factor of safety of 3 is

applied to obtain a mean value of each connector population. Using this value, a mean of

3720 lbs and 4200 lbs in uplift capacity per connector are obtained for the side

connectors of wood and masonry homes, respectively. The mean values for gable end

connectors are calculated to be 3780 lbs and 1920 lbs in uplift capacity, for wood and

masonry homes, respectively. A COV of 0.2 is selected for all roof-to-wall fastener

distributions, and bounds are placed such that acceptable values lie within two standard

deviations of the mean. These distributions result in capacities higher than those obtained









through in situ and laboratory testing [42-44], but the damages predicted using the

manufacturer's values correspond well with post-damage information surveyed after

Hurricane Andrew. These results are provided in Chapter 7.

To capture an observed failure mechanism where the entire roof detaches from the

walls [19], the roof-to-wall connection strength for each simulated house is batch

selected. A representative value is generated for the entire house from the Gaussian

distribution representing the population of roof-to-wall connections for that type of

structure (wood or masonry). This value becomes the mean of a Gaussian distribution

having a COV of 0.05 from which individual connection capacities are randomly

generated. This process represents obtaining the connectors from the same manufacturer

or batch, and using the same quality of installation for the home. Additional details on the

batch selection process are provided in Chapter 6. Roof-to-wall connections are the only

structural components to be selected in this manner, specifically to incorporate the

observed damage state of having the entire roof detach from the walls. The method of

batch selection is not used for other structural components because it results in predicted

damages that are not observed in post-damage reports [19].

Wind Resistance Capacity of Site-Built Home Walls

Wall failures are much less commonly cited in post-damage reports than roofing

system failures. In many cases, wall failures could be attributed to improper installation

of connections or to the loss of structural integrity of the roof system [19]. Capacities to

resist shear, out-of-plane loading, and uplift are considered for wood frame walls and

masonry walls in the following paragraphs.

Resistance capacities for wood walls are obtained from the 1997 National Design

Specification for Wood Construction (NDS) [45], as well as from laboratory tests. Wood









capacities are distinctly difficult to generalize over a large population of homes because

the load carrying capability of wood connections varies significantly with different types

of lumber. To best represent the types of wood typically found in Florida, southern

species of wood, such as Spruce Pine Fir (SPF) and Southern Yellow Pine (SYP), are

used in resistance calculations.

Damage to masonry walls was less prevalent than damage to wood frame walls,

and masonry walls were less dependent on the integrity of the roof system [19]. However,

damage surveys [46] have shown that un-reinforced masonry might be a weak link in the

structural system. After the failure of an opening, increased internal pressure can lead to

the collapse of masonry walls, which trigger the collapse of the whole structure. One

study was obtained predicting the failure pressure for simply reinforced and pre-stressed

wall sections [47]; however, this study alone is not enough information to adequately

predict failure conditions for typical residential structure walls. In the absence of a

significant population of laboratory test data, design provisions are used, with

adjustments to allow for the best representation of true failure loads.

Wood shear wall capacity

Shear wall loads are transferred through the wall sheathing in wood frame walls.

As a result, the capacity of the wall to resist shear wall loads is dependent on the nailing

pattern and thickness of the attached plywood. Using 3/8 inch plywood sheathing with 8d

nails spaced at 6 inches on center along sheathing edges, the shear flow capacity of a

typical wall is 310 lbs per linear foot, according to the NDS. A factor of safety of 3.5 is

applied to this capacity, to account for both the safety built into the design code, as well

as the uncertainty in the contribution of other building materials. Wood homes are

generally covered with some other form of cladding, which contributes to the ability of