Assessing the Wind Resistance of Sectional Door Systems for Facilities in Hurricane-Prone Areas through Full- and Compon...

MISSING IMAGE

Material Information

Title:
Assessing the Wind Resistance of Sectional Door Systems for Facilities in Hurricane-Prone Areas through Full- and Component-Scale Experimental Methods and Finite Element Analysis
Physical Description:
1 online resource (220 p.)
Language:
english
Creator:
Shen, Yan
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Civil Engineering, Civil and Coastal Engineering
Committee Chair:
MASTERS,FORREST J
Committee Co-Chair:
CONSOLAZIO,GARY R
Committee Members:
PREVATT,DAVID
GURLEY,KURTIS R
FERRARO,CHRISTOPHER CHARLES
SULLIVAN,JAMES G
UPJOHU,HENRY

Subjects

Subjects / Keywords:
analysis -- composite -- door -- element -- experimental -- finite -- full -- garage -- scale -- structure -- testing
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre:
Civil Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Wind-induced failure of garage doors can create a largebreach in the building envelope, which can lead to an adverse internalpressurization and cause cascading failures in the building envelop. This studyaddresses the performance of these systems to fill a critical knowledge gap. Sectionalgarage doors were subjected to wind pressure using a new large-scale dynamicwind pressure simulator that was developed to conduct full-scale experimentaltesting. The Simulator was specifically designed to compensate for high airflowleakage (= 50,000 CFM) during testing. Five commercial doors were tested underquasi-static pressure loading until failure. Additionally, finite element analysis(FEA) was used to model two of the five sectional doors for more thorough analysis.Results from FEA and experimental testing matched well. Assessment resultssuggest two main failure mechanisms of current sectional doors under windpressure, namely, local buckling of U-bars and disengagement from door tracks.Excessive deflection is the main reason for their failure. To improve the wind resistance of sectional doors, thisresearch investigated the applicability of a new sandwich panel comprised offiber reinforced plastic face sheets and a polyol-isocyanate foam core.Mechanical properties of the sandwich panel were obtained through directuniaxial compressive, direct uniaxial tensile and four-point bending tests.Foam density and thickness were varied to determine the effect on the sandwichpanel behavior. A full-size panel was subjectedto uniform out-of-plane loading in a pressure loading actuator to simulateservice conditions. Measured strains and displacements closely match resultsfrom finite element modeling using the constitutive properties of the foam obtained from the uniaxial compression andtension tests as inputs. The findings indicate that the sandwich panel issuitable for lightweight building cladding systems used in high wind areas.Deflection is the controlling factor for out-of-plane loading.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Yan Shen.
Thesis:
Thesis (Ph.D.)--University of Florida, 2013.
Local:
Adviser: MASTERS,FORREST J.
Local:
Co-adviser: CONSOLAZIO,GARY R.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2013
System ID:
UFE0046181:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 ASSESSING THE WIND RESISTANCE OF SECTIONAL DOOR SYSTEMS FOR FACILITIES IN HURRICANE PRONE AREAS THROUGH FULL AND COMPONENT SCALE EXPERIMENTAL METHODS AND FINITE ELEMENT ANALYSIS By YAN SHEN 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 2013

PAGE 2

2 2013 Yan Shen

PAGE 3

3 ACKNOWLEDGMENTS My deepest gratitude goes to my advisor, Dr. Forrest Masters, for creating an opportunity to study at the University of Florida, providing academic guidance, and mentoring me to become a passionate caring and responsible researcher I am grateful for his support an d patience when this journey became tough at times during the past four and half years. I am very thankful for Special Lite Inc. T heir financial support made it possible for me to come to the United States Particularly, I would like to thank Henry L. Upj ohn II and Ken Bowditch for their support at all aspects of this research. I also want to give my great appreciation to Dr. Gary Consolazio for always being a go to professor whenever I ha d questions. I am also very grateful for him for being a role mode l for a responsible and caring professor. I would also like to extend my thanks to Dr. Chris Ferraro for his patience, instruction and assistance for the experimental testing on the composite panels. Additionally, I am grateful for Dr. David Prevatt, witho ut given to Dr., Dr. Kurt Gurley and Dr. James Sullivan for all their help and support in this work. I appreciate Jimmy Jesteadt, Alex Esposit o for all the tremendous work and contribution in this research. I would also like to thank Scott Bolton and Cedric Adam for their assistance. In addition, I am grateful to Jordan Nelson, Michael Willis, Abraham Alende, Justin Henika, Kyle Watson and Daniel Getter for their assistance with the experimental testing. I would also like to thank all my officemates, Sylvia Laboy in particular, for their sup p ort in this research.

PAGE 4

4 My time as a doctoral student would have be en more difficult without the love and support of numerous friends. I am deeply warmed by the friendship of Laura Li and Stanley Ling for their prayers and being always available whenever I need them. I also want to give my thanks to Si Chen, Sherry Huo and Jie Li for their support and caring. As for my family, I believe there is much more than I can express my gratitude towards them here. Their unconditional love, caring and support always remind and comfort me that my acceptance by them is independent of my accomplishment s Ultimately, to God, I owe endless grati tude for His faithfulness and grace through the whole process.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 3 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIA TIONS ................................ ................................ ........................... 16 ABSTRACT ................................ ................................ ................................ ................... 18 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 20 1.1 Background ................................ ................................ ................................ ....... 20 1.2 Problem Statement ................................ ................................ ........................... 21 1.3 Research Approach ................................ ................................ .......................... 22 1.4 Organization of the Dissertation ................................ ................................ ........ 23 2 BACKGROUND ................................ ................................ ................................ ...... 26 2.1 Wind Load Rated Garage Doors ................................ ................................ ...... 26 2.1.1 Introduction of Wind Load Rated Garage Doors ................................ ..... 26 2.1.2 Observed Failure of Garage Doors in Hurricanes ................................ ... 26 2.1.3 Studies on Wind Induced Performance of Garage Doors ........................ 26 2.1.4 Current Design Philosophy of Sectional Doors ................................ ........ 28 2.2 Three Layer Composite Panel ................................ ................................ .......... 28 2.3 Methods to Investigate Structural Performance in Hazard Engineering ............ 30 2.4 Full/Large Scale Testing Facility ................................ ................................ ....... 32 2.4.1 BRERWULF ................................ ................................ ............................ 33 2.4.2 3LP PLAs ................................ ................................ ................................ 33 2.4.3 HAPLA ................................ ................................ ................................ ..... 33 2.5 Summary ................................ ................................ ................................ .......... 34 3 METHODOLOGY ................................ ................................ ................................ ... 43 3.1 Full Scale Experimental Testing ................................ ................................ ....... 43 3.1.1 Determining the Design Wind Pressure of the Simulator ......................... 44 3.1.2 System Design Algorithm ................................ ................................ ........ 46 3.1.2.1 An overview of the Simulator ................................ ......................... 46 3.1.2.2 Fan system design ................................ ................................ ......... 47 3.1.3 Introduction of the Simulator ................................ ................................ .... 50 3.1.3.1 Technical information about the Simulator ................................ ..... 51 3.1.3.2 Control system ................................ ................................ ............... 51

PAGE 6

6 3.1.4 Current Status of the Simulator ................................ ............................... 52 3.2 Finite Element Analysis Method ................................ ................................ ........ 52 3.3 Summary ................................ ................................ ................................ .......... 53 4 FORENSIC ASSESSMENT OF CURRENT GARAGE DOORS USING FULL SCALE EXPERIMENTAL AND NUMERICAL METHODS ................................ ...... 68 4.1 Full Scale Experimental Testing ................................ ................................ ....... 69 4.1.1 Materials ................................ ................................ ................................ .. 69 4.1.2 Experimental Assumptions ................................ ................................ ...... 69 4.1.2.1 Concerning the loading direction ................................ .................... 69 4.1.2.2 Concerning the attachment of test specimens to buildings ............ 70 4.1.2.3 Definition of door failure ................................ ................................ 70 4.1.3 Testing instrumentation ................................ ................................ ........... 70 4.1.4 Testing Load ................................ ................................ ............................ 71 4.1.5 Testing Procedure ................................ ................................ ................... 71 4.2 Finite Element Analysis ................................ ................................ ..................... 72 4.2.1 Finite Element Model o f Door 2 ................................ ............................... 72 4.2.1.1 Assumptions and simplifications in FE model of door 2 ................. 72 4.2.1.2 Element components in FE model of door 2 ................................ .. 72 4.2.1.3 Boundary conditions and constraints ................................ ............. 73 4.2.1.4 Material properties ................................ ................................ ......... 73 4.2.1.5 Loads ................................ ................................ ............................. 74 4.2.2 Finite Element Model of Door 3 ................................ ............................... 74 4.2.2.1 Assumptions and simplifications in FE model of door 3 ................. 74 4.2.2.2 Element components in FE model of door 3 ................................ .. 75 4.2.2.3 Boundary conditions and constraints ................................ ............. 75 4.2.2.4 Material properties ................................ ................................ ......... 76 4.2. 2.5 Loads ................................ ................................ ............................. 76 4.3 Results and Discussions ................................ ................................ ................... 77 4.3.1 Full Scale Experimental Results ................................ .............................. 77 4.3.1.1 Failure mechanisms ................................ ................................ ....... 77 4.3.1. 2 Displacement and strain responses ................................ ............... 78 4.3.2 FEA Results of Door 2 and Comparison to Full Scale Experimental Results ................................ ................................ ................................ .......... 79 4.3.2.1 Failure mechanisms comparison ................................ ................... 79 4.3.2.2 Displacement and strain results comparison ................................ .. 80 4.3.2.3 Effects of construction errors ................................ ......................... 80 4.3.3 FEA Results of Door 3 and Comparison to Full Scale Experimental Re sults ................................ ................................ ................................ .......... 82 4.3.3.1 FEA results of the local model ................................ ....................... 82 4.3.3.2 Failure mechanisms comparison of door 3 ................................ .... 83 4.3.3.3 Displacement and strain results comparison ................................ .. 83 4.4 Summary ................................ ................................ ................................ .......... 84 5 MECHANICAL RESISTANT PROPERTIES OF FRP/FOAM COMPOSITE STRUCTURE ................................ ................................ ................................ ........ 132

PAGE 7

7 5.1 Experimental Investigation on the Foam Material Properties .......................... 134 5.1.1 Test Specimens ................................ ................................ ..................... 134 5.1.2 Testing Apparatus and Instrumentation ................................ ................. 134 5.1.3 Experimental Testing ................................ ................................ ............. 135 5.1.3.1 Direct uniaxial compressive tests ................................ ................. 135 5.1.3.2 Direct uniaxial tensile tests ................................ ........................... 136 5.1.3.3 Four point bending tests ................................ .............................. 136 5.1.4 Experimental Results ................................ ................................ ............. 137 5.1.4.1 Compression testing results ................................ ......................... 137 5.1.4.2 Tension testing results ................................ ................................ 138 5.1.4.3 Four point bending testing results ................................ ................ 139 5.2 Experimental and Numerical Investigation on A Full Length Composite Panel ................................ ................................ ................................ ................. 140 5.2.1 Experimental Testing ................................ ................................ ............. 141 5.2. 1.1 Test specimen ................................ ................................ .............. 141 5.2.1.2 Testing apparatus and instrumentation ................................ ........ 141 5.2.1.3 Testing load ................................ ................................ ................. 142 5.2.2 Finite Element Method ................................ ................................ ........... 142 5.2. 2.1 Finite element model components ................................ ............... 142 5.2.2.2 Boundary conditions and constraints ................................ ........... 142 5.2.2.3 Material properties in FE model ................................ ................... 143 5.2.2.4 Pressure load in FE model ................................ ........................... 143 5.2.3 Results and Discussion ................................ ................................ ......... 143 5.2.4 Methods to Increase Stiffness ................................ ............................... 144 5.2.4.1 Changing dimensions of the sandwich panel ............................... 145 5.2.4.2 Inserting through thickness FRP stiffeners ................................ .. 145 5.3 Summary ................................ ................................ ................................ ........ 146 6 DESIGN AND DEVELOPMENT OF SANDWICH DOOR PANEL ......................... 180 6.1 Design of Sandwich Door Panel ................................ ................................ ..... 180 6.1.1 Description of sandwich door panel ................................ ....................... 180 6.1.2 Finite element (FE) model of composite door panel .............................. 181 6.1.2.1 FE model components ................................ ................................ 181 6.1.2.2 Boundary conditions ................................ ................................ ..... 181 6.1.2.3 Material properties in FE model ................................ ................... 182 6.1.2.4 Pressure load in FE model ................................ ........................... 182 6.1.3 FEA results and discussions ................................ ................................ 182 6.2 Comparison between Composite Door Panel and Current Door Panel .......... 183 6.2.1 Comparison methodology ................................ ................................ ...... 183 6. 2.2 Comparison results and discussions ................................ ..................... 184 6.3 Summary ................................ ................................ ................................ ........ 185 7 CONCLUSIONS AND RECOMMENDATIONS FOR FURTURE RESEARCH ..... 198 7.1 Forensic Assessment of Current Garage Doors using Full Scale Experimental and Numerical Methods ................................ ............................... 198

PAGE 8

8 7.1.1 Conclusions and Contributions ................................ .............................. 198 7.1.2 Recommendations fo r Future Research ................................ ................ 2 00 7.1.2.1 Full scale experimental testing ................................ ..................... 200 7.1.2.2 Finite element model ................................ ................................ .... 201 7.2 Mechanical Resistant Properties of FRP/ Foam Composite Structure ............ 202 7.2.1 Conclusions and Contributions ................................ .............................. 202 7.2.2 Recommendations for Future Research ................................ ................ 203 7.3 Research and Development of Sandwich Garage Door Pan els ..................... 204 7.3.1 Conclusions and Contributions ................................ .............................. 204 7.3.2 Recommendations for Future Research ................................ ................ 205 APPENDIX A CONVERTING WIND TUNNEL DATA TO FULL SCALE DATA .......................... 206 B BACKWARD DIFFERENCE APPROXIMATION METHOD ................................ .. 208 LIST OF REFERENCES ................................ ................................ ............................. 212 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 220

PAGE 9

9 LIST OF TABLES Table page 4 1 Description of test specimens ................................ ................................ ............. 86 4 2 Main components of the FE model of door 2 ................................ ...................... 92 4 3 Components of the local model of roller and track ................................ .............. 96 4 4 Main components of the global model of door 3 ................................ ................. 98 4 5 Summary of the pressure at failure of test specimens ................................ ...... 109 4 6 Difference percentage of strain responses between full scale testing and FE model of door 2 ................................ ................................ ................................ 116 4 7 Difference percentage of displacement responses between full scale testing and FE model of door 2 ................................ ................................ .................... 117 4 8 Difference percentage of strain responses between full scale testing and FE model of door 3 ................................ ................................ ................................ 129 4 9 Difference percentage of displacement responses between full scale testing and FE model of door 3 ................................ ................................ .................... 130 5 1 Physical properties of Glasbord REI FRP ................................ ......................... 148 5 2 Type of tests and specimen geometry ................................ .............................. 149 5 3 Mean values of the polyol isocyanate foam properties under uniaxial compression ................................ ................................ ................................ ..... 154 5 4 Experimental mean values of the polyol isocyanate foam properties in tensile tests ................................ ................................ ................................ .................. 158 5 5 Experimental mean values of shear strength for the polyol isocyanate foam ... 159 5 6 Main finite element (FE) model components ................................ .................... 168 5 7 Comparison between flexural responses from pressure testing and FE model 171 5 8 Deflection limits for structural members subjected to wind loads ..................... 173 5 9 Maximum deflections of sandwich panels with FRP stiffeners ......................... 176 6 1 Physical properties of aerodynamic side skirt ................................ ................... 186

PAGE 10

10 6 2 FE model components of a sandwich door panel ................................ ............. 189

PAGE 11

11 LIST OF FIGURES Figure page 1 1 Hurricane prone regions in the USA. ................................ ................................ 24 1 2 Failure of garage doors on a fire station caused the loss of the entire roof during Hurricane Charley ................................ ................................ .................... 25 2 1 Wind rated garage doors.. ................................ ................................ ................. 35 2 2 Failures of garage doors in hurricanes. ................................ ............................. 36 2 3 Experimental and numerical studies on a 10 ft. x 10 ft. rolling door. .................. 38 2 4 Typical sectional garage door details. ................................ ............................... 39 2 5 A photograph of the composite structure sample studied in this research ......... 40 2 6 3LP PLAs developed at the University of Western Ontario .............................. 41 2 7 High Airflow Pressure Loading Actuator (HAPLA). ................................ ............ 42 3 1 Wind tunnel models employed for determination of the design wind pressure for the Simulator discussed in this research ................................ ....................... 54 3 2 Pressure tap layout of the four wind tunnel models. ................................ .......... 55 3 3 An illustration on the procedure to find out the worst case of C P for the design wind pressure. ................................ ................................ ......................... 56 3 4 The worst case of from wind tunnel model m11, m21, m31 and m41. ......... 57 3 5 The worst case scenario of wind pressure time histories for the design pressure of the Simulator. ................................ ................................ .................. 58 3 6 3D drawings of the Simulator with the major components. ................................ 59 3 7 A sketch of air slug moving in/out of a chamber through the opening. .............. 60 3 8 Sketches of the fan design concept accounting for leakage and flexibility of specimens.. ................................ ................................ ................................ ........ 61 3 9 A photograph showing the fan syst em and duct silencers of the Simulator ....... 62 3 10 Photographs of the prime mover (engine) and air clutch of the Simulator. ........ 63 3 11 The control system of the Simulator. ................................ ................................ 64

PAGE 12

12 3 12 Photographs of the airbox. ................................ ................................ ................ 65 3 13 The reaction frame system of the Simulator ................................ ...................... 66 3 14 A schematic of a typical fan curve and several system resistance curves. ........ 67 4 1 One sacrificial door installed to the Simulator at the opening of the airbox ....... 87 4 2 Failures observed from sacrificial tests ................................ .............................. 88 4 3 Instruments used in full scale experimental testing. ................................ .......... 89 4 4 A conceptual sketch of the loading during full scale experimental tes ting. ........ 90 4 5 Half symmetry model of door 2 in ADINA. ................................ ......................... 91 4 6 Boundary conditions and constraints in FE model of door 2.. ............................ 93 4 7 A FE mo del of a roller using three dimensional solid elements. ........................ 94 4 8 The local model of a roller and 21 in. long track. ................................ ............... 95 4 9 The global FE model of door 3 ................................ ................................ .......... 97 4 10 Boundary conditions and constraints in the local FE model of roller and track. ................................ ................................ ................................ ................. 100 4 11 Boundary conditions and constraints in the global FE model of door 3 ........... 102 4 12 Interaction between the local model and global model of door 3. .................... 103 4 13 Failure mechanisms of door 1 ................................ ................................ ......... 104 4 14 Failure mechanisms of door 2 ................................ ................................ ......... 105 4 15 Failure mechanisms of door 3 ................................ ................................ ......... 106 4 16 Failure mechanisms of door 4 ................................ ................................ ......... 107 4 17 Failure mechanisms of door 5 ................................ ................................ ......... 108 4 18 Characteristics of the U bars and rol lers of the five test specimens ................ 110 4 19 Measured strain suction relationships at multiple locations during experimental testing. ................................ ................................ ........................ 111 4 20 Measured deflection suction relationships at multiple locations during experimental testing.. ................................ ................................ ....................... 112

PAGE 13

13 4 21 Failure mechanism comparison between experimental testing and FEA results of door 2. ................................ ................................ ............................... 113 4 22 Strain comparisons between FEA results and experimental measurements of door 2. ................................ ................................ ................................ .......... 114 4 23 Deflection comparison between FEA results and experimental measurements of door 2. ................................ ................................ .................. 115 4 24 A sketch showing 1 in. overlap at the end of door 2 required by the manufacturer. ................................ ................................ ................................ ... 118 4 25 The effects of missing fasteners conne cting U bars to stiles on structural responses of door 2 ................................ ................................ .......................... 119 4 26 The effects of missing fasteners connecting hinges to stiles on structural responses of door 2. ................................ ................................ ......................... 120 4 27 The effects of missing 1 in. overlap at the end of door 2 on structural responses. ................................ ................................ ................................ ........ 121 4 28 FEA results of the interaction between the roller and the track under prescribed displacement. ................................ ................................ .................. 122 4 29 Different contact forces at different locations on the track. .............................. 123 4 30 The relationship between contact force and roller displacement. .................... 124 4 31 Failure mechanisms of door 3 in experimental testing and FEA. ..................... 125 4 32 Roller displacement pressure relationship. ................................ ...................... 126 4 33 Strain comparisons between FEA results and experimental measurements of door 3. ................................ ................................ ................................ .......... 127 4 34 Displacement comparisons between FEA results and experimental measurements of door 3. ................................ ................................ .................. 128 4 35 A photograph showing that the top part of the track is curved. ........................ 131 5 1 Instron 3384 universal testing machine and data acquisition system. ............. 150 5 2 Examples of compressive testing. ................................ ................................ ... 151 5 3 Direct uniaxial tensile testing configuration. ................................ ..................... 152 5 4 FPB testing setup and dimensions of test specimens. ................................ .... 153

PAGE 14

14 5 5 The effects of foam density (A, B and C) and thickness (D, E and F) on its compressive stress strain responses. ................................ .............................. 155 5 6 Failure mechanisms of the composite structure in tensile testing ................... 156 5 7 The effects of foam thickness (A, B and C) and density (D, E, and F) on tensile stress strain responses. ................................ ................................ ........ 15 7 5 8 Failure mechanisms observed in FPB testing. ................................ ................. 160 5 9 Load deflection diagrams for FPB testing. ................................ ....................... 161 5 10 A photograph of the full size sandwich panel. ................................ ................. 162 5 11 Experimental design of the uniform out of plane pressure testing. .................. 164 5 12 The pressure loading applied on the full size sandwich panel during experimental testing. ................................ ................................ ........................ 165 5 13 FE model of the full length panel and details of constraints in the FE model. .. 167 5 14 FE model of the core FRP skin interface and material models of the FRP and foam. ................................ ................................ ................................ ......... 169 5 15 Visual comparison of deformed shape between the experimental testing and FE model. ................................ ................................ ................................ ......... 170 5 16 Expe rimental and numerical results on the flexural behavior of the large scale composite panel. ................................ ................................ ..................... 172 5 17 Numerical evaluation of the ef fects of foam and FRP thicknesses on the allowable pressure of the composite panel based on material strength and deflection limits of serviceability. ................................ ................................ ....... 174 5 18 Numerical evaluation of the effects of inserting through thickness FRP stiffeners in the sandwich panel studied herein. ................................ ............... 175 5 19 Three point testing on sandwich panels with stiffeners. ................................ .. 177 5 20 Deformed shapes of panel samples at failure. ................................ ................ 178 5 21 Applied load mid span deflection relat ionships of panel samples with longitudinal and transverse stiffeners ................................ ............................... 179 6 1 Half symmetry models of sandwich door panels of different width .................. 187 6 2 FE model Components of a sandwich door panel ................................ ........... 188 6 3 Boundary conditions of FE model of the sandwich door panel. ....................... 190

PAGE 15

15 6 4 The load applied to the sandwich door panel. ................................ ................. 191 6 5 Maximum deflections of the sandwich door panels under different applied pressure. ................................ ................................ ................................ .......... 192 6 6 A sketch of the Euler Bernoulli beam deformation under uniformly distributed load. ................................ ................................ ................................ .................. 193 6 7 The allowable pressure for the sandwich door panels considering strength and serviceability requirements. ................................ ................................ ....... 194 6 8 FE models of a current garage door panel and the sandwich door panel to be compared on their performance under wind pressure loading. .................... 195 6 9 Comparison of maximum deflections between door 2 panel and sandwich door panel. ................................ ................................ ................................ ........ 196 6 10 Comparison of the catena ry force maximum mid span deflection relationship between door 2 panel and the sandwich door panel. ................................ ....... 197 B 1 Flow chart for Computing the airflow movements and external pressure at the opening of the airbox. ................................ ................................ ................. 211

PAGE 16

16 LIST OF ABBREVIATIONS ASCE American Society of Civil Engineers ADINA Automatic Dynamic Incremental Nonlinear Analysis ASTM American Society for Testing and Materials BRE Building Research Establishment BRERWULF BRE Real time Wind Uniform Load Follower C P Wind Pressure Coefficient E C Elastic Modulus in Compre ssion E FRP_flexural Flexural Modulus of FRP E T Elastic Modulus in Tension FBC Florida Building Code FE Finite Element FEA Finite Element Analysis FEMA Federal Emergency Management Agency FPB Four Point Bending FRP Fiber Reinforced Plastics HAPLA High Airf low Pressure Load Actuator NIST National Institute of Standard and Technology NOA Notice of Acceptance NOAA National Oceanic and Atmospheric Administration PLA Pressure Loading Actuator Simulator A New Large Scale Dynamic Wind Load S imulator TAS Testing Application Standard TPB Three Point Bending UWO University of Western Ontario

PAGE 17

17 Strain u_t Ultimate Strain in Tension yc Yielding Strain in Compression Stress u_t Ultimate Strength in Tension u_FRP Flexural Strength of FRP ub Ultimate Bonding Strength ult Ultimate Shear Strength

PAGE 18

18 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 ASSESSING THE WIND RESISTANCE OF SECTIONAL DOOR SYSTEMS FOR FACILITIES IN HURRICANE PRONE AREAS THROUGH FULL AND COMPONENT SCALE EXPERIMENTAL METHODS AND FINITE ELEMENT ANALYSIS By Yan Shen December 2013 Chair: Forrest J. Masters Cochair: Gary R. Consolazio Major: Civil Engineering Win d induced failure of garage doors can create a large breach in the building envelope, which can lead to an adverse internal pressurization and cause cascading failure s in the building envelop This study addresses the performance of these systems to fill a critical knowledge gap. S ectional garage doors were subjected to wind pressure using a new large scale dynamic wind pressure simulator that was developed to conduct full scale experimental testing. T he Simulator was specifically designed to compensat e for 0,000 CFM) during testing. Five commercial doors were tested under quasi stati c pressure loading until failure Additionally finite element analysis ( FEA ) was used to model two of the five sectional doors for more thorough analys is. Results from FEA and experimental testing matched well. Assessment results suggest two main failure mechanisms of current sectional doors under wind pressure, namely, local buckling of U bars and disengagement from door tracks. Excessive deflection is the main reason for their failure.

PAGE 19

19 To improve the wind resistance of sectional doors, this research investigated the applicability of a new sandwich panel comprised of fiber reinforced plastic face sheets and a polyol isocyanate foam core. Mechanical prop erties of the sandwich panel were obtained through direct uniaxial compressive, direct uniaxial tensile and four point bending tests. Foam density and thickness were varied to determine the effect on the sandwich panel behavior. A full size panel was subje cted to uniform out of plane loading in a pressure loading actuator to simulate service conditions. Measured strains and displacements closely match results from finite element modeling using the constitutive properties of the foam obtained from the uniaxi al compression and tension tests as inputs The findings indicate that the sandwich panel is suitable for lightweight building cladding systems used in high wind areas. Deflection is the controlling factor for out of plane loading.

PAGE 20

20 CHAPTER 1 INTRODUCTIO N 1.1 Background In the United States, extreme winds cause a significant portion of the total damage due to natural hazards During 1990 to 2005, hurricanes caused approximately $ 10 billion in damage in the continental United States annually Approximately 85% of the damage is attributed to Saffir Simpson Hurricane Wind Scale (SSHWS) Categories 3, 4, and 5 (major) hurricanes, although they only account for 24% of the U.S. land falling tropical cyclones (Pielke et al. 2008) For example, H urricane Andrew in 1992 caused mo re than $ 2 5 billion (1992 USD, $41.4 billion 2012 USD) of damage alone in southern Dade County, Florida (Rappaport 1994) An important reason accounting for such immense hurricane induced damages is that hurricane prone regions in the U.S. (Fig. 1 1) conta in many highly developed cities such as Miami, Tampa, Houston, New Orleans Furthermore the losses are expected to be exacerbated by projected increases in 1) t he population and infrastructure density in coastal areas, and 2) t he anthropogenic warming that will cause hurricanes to be more frequent and more destructive (Bender et al. 2010; Emanuel 2005) A significant part of hurricane induced damage cost is due to the failure and damage of buildings and related structur al damage (Mehta 1984; Pinelli et al. 2003) For instance, during H urricane Andrew, more than 25 ,000 homes were destroyed and another 101 ,000 more were damaged in southe rn Dade County, Florida (Rappaport 1993) Historically, wind resistant design has focused on the main load resisting frame However, hundreds of investigation into windstorm induced damage

PAGE 21

21 have emphasized that building envelope system should receive as much attention as main load resisting frame (Minor 2005) Building envelope system s including roofs, windows, do ors and walls, are directly exposed to severe wind conditions, and their failure or d amage in a hurricane (e.g. the loss of panels of roof sheathing, the breakage of window glass, the falling of a garage door, etc ) can result in water ingress, internal pressure increasing, and ultimately lead to tremendous damage to buildings and interior properties. Therefore, a better design of building envelope system is of significant importance to reduce losses caused by hurri canes The research presented herein contributes to mitigating wind induced damage to building envelope system. 1.2 Problem Statement Post hurricane assessment indicates that the failure of garage doors is common (Croft et al. 2006; FEMA 2005a; b) Garage doors are usually the largest movable component of building envelope system. Their failure can create a larg e breach on the building envelope, which allows wind and rain to enter the building and can result in a significant increase the internal pressure. The internal pressure can increase to the same magnitude of the external suction above the roof (Irwin and Sifton 1998) Fig. 1 2 shows this cascading failure due to the failure of sectional garage doors used in a fire station. With regard to the performance of these systems, this study seeks to answer the followi ng questions: What are the common failure mechanisms of garage doors subjected to hurricane wind loads? What are the reasons accounting for the failure? Can composite materials improve the wind resistance of these systems?

PAGE 22

22 1.3 Research Approach The research objectives are to : 1) advance the understanding of structural behavior of current sectional garage doors under hurricane wind loads, and 2) investigate the applicability of a novel FRP/ foam sandwich panel use in garage doors in hurricane prone a reas. The technical approach is described below. Developing a large scale dynamic wind pressure simulator The methodology applied full scale experimental testing and finite element analysis ( FE A ) This project contributes to full scale experimental testin g by developing a new testing system, which is the large scale dynamic wind pressure simulator (Simulator). The Simulator is designed to replicate time varying wind loads associated with intense hurricane conditions (e.g., Saffir Simpson Hurricane Scale Ca tegory 5 Hurricane) in a laboratory environment to test full scale building components and cladding system and is designed to compensate for high airflow leakage through test specimens during testing. Chapter 3 provides more details on the design an d devel opment of the Simulator. Investigating the performance of current sectional garage doors under wind pressure loading This study (Chapter 4) addresses the structural behavior of sectional doors subjected to wind pressure loading, investigating the reasons accounting for their failure and providing a direction for design of a hurricane resistant garage door. The newly developed Simulator wa s utilized for full scale experimental testing on five commercial sectional doors. FEA models were developed for two doo rs and validated with observed failure mechanisms and strain and displacement responses. Additionally, c onstructions errors were simulated in the FEA models to assess the effect of improper fastening.

PAGE 23

23 Characterizing the mechanical resistance of a FRP/ foa m sandwich panel The overarching goal of this study (Chapter 5) wa s to investigate the appropriateness of a novel FRP/ foam sandwich panel to be used as larger component and cladding systems in high wind areas. The first part of this study focuses on a series of material testing to obtain mechanical properties of the sandwich panel Various foam thickness and density are tested to evaluate th eir effects on mechanical properties of the sandwich panel The second part of this study uses full scale experimental testing and FEA on a full size panel subjected to uniform pressure to determine the applicability of the sandwich panel used for garage d oor in hurricane prone regions. At the end, FEA wa s employed to investigate different optimizations to increase the stiffness of the sandwich panel. 1.4 Organization of the Dissertation T his dissertation consists of seven chapters. Chapter 1 introduces th e background, problem statement and research approach. Chapter 2 discusses relevant background material The methodology comprised of full scale experimental testing and numerical methods is described in Chapter 3. Chapter 4 presents the study of forensic assessment of current sectional garage doors using full scale experimental and numerical methods. Chapter 5 focuses on investigating the applicability of a novel FRP/ foam sandwich panel used for garage doors in hurricane prone areas. Chapter 6 describes a preliminary investigation on design and development of sandwich panels used for garage doors. Finally, Chapter 7 summarizes the doctoral work and provides recommendations for future research.

PAGE 24

24 Figure 1 1 Hurricane prone regions in the USA (Photo courte sy of NOAA )

PAGE 25

25 Figure 1 2. Failure of garage doors on a fire station caused the loss of the entire roof during Hurricane Charley (Photo courtesy of FEMA 2005b )

PAGE 26

26 CHAPTER 2 BACKGROUND 2.1 Wind Load Rated Garage Doors 2.1.1 Introduction of Wind Load Rated Garage Doors There are two main types of wind load rated doors existing, i.e., sectional doors and rolling doors (Fig. 2 1). Wind load rated door systems are designed to resist out of plane component and claddin g loads defined in ASCE 7 ( 2010 ) Typical product approval requirements include uniform static air pressure loading (e.g., ANSI/DASMA 108 ) and cyclic pressure and impact resistance ( e.g., TAS 202 ) This study addresses the out of plane wind resistance of large scale sectional door systems. 2.1.2 Observed Failure of Garage Doors in Hurricanes Post storm assessments have found that failure of garage doors is a recurring issues in hurricanes ( e.g., Croft et al. 2006; FEMA 2005a; b) Garage doors are the largest opening in a building and thus play an important role in the structural integrity of windstorm investigation s have found damage from windborne debris but the majority of damage w as caused by lack of reinforcement and track bracing for deign pressures (Reinhold 2005) Fig. 2 2 gives examples of observed failure of gar age doors in hurricanes. The garage door failure s in hurricanes emphasize the need to 1) study garage door performance under hurricane wind conditions, and 2) develop solutions to improve their wind resistant performance. These two issues are to be address ed in this study. 2.1.3 Studies on Wind Induced Performance of Garage Doors In the area of wind engineering, a majority of research on building envelope components has focused on the performance of roof ing system and components ( e.g.,

PAGE 27

27 Ahmed et al. 2011; Cook 1992; Edmonson et al. 2012; Ellingwood et al. 2004; Morrison and Kopp 2011; Prevatt et al. 2008; Rosowsky et al. 1998; Shanmugam et al. 2009 ). Studies on win dow glass under hurricane wind conditions are also well documented (e.g., Calderone and Melbourne 1993; Gavanski and Kopp 2011; Gavanski 2009; Holmes 1985; Masters et al. 2010; Reed and Simiu 1984 ). Several studies have addressed the issue of water penetration through windows and window wall interface (e.g., Blackall and Baker 1984; Lacasse et al. 2007; Lopez et al. 2011; RD H Building Engineering Limited 2002; Salzano et al. 2010 ). L ittle research has been conducted on the performance of garage doors subjected to wind loads. Only one such study is available in the public domain. Gao and Moen ( 2009) conducted experimental testing on a 10 ft x 10 ft rolling door used for vehicular access. Both positive pressure and negative pressure were applied on the door panel using a custom vacuum chamber (Fig. 2 3A) at Doors & Building C omponents Inc. in Douglasville GA The door specimens were mou nted to a frame using cold formed steel C sections braced by Z section girts (Fig. 2 3B) The supporting frame was designed to simulate a flexible door jamb. They concluded from the experimental testing that catenary forces and deflections of the rolling d oor are sensitive to jamb stiffness, and this stiffness needs to be considered when designing vehicular access doors (i.e., garage doors) To understand the structural performance of the vehicular door, Gao and Moen ( 2011) developed a three dimensional finite element (FE) model of the door. A rational approximation of the jamb stiffness using hand calculations was performed Spring models employing the calculated jamb stiffness were incorporated into a finite element model of the rolling door (Fig. 2 3C). The responses of

PAGE 28

28 the door deflection with applied pressure from FEA were consistent with the experimental results (Fig. 2 3D). In contrast to the Gao and Moen ( 2009, 2011) studies this research focuses on sectional garage doors A major difference between these sys tems is the catenary forces that act in the membrane. Rolling doors can develop large in plane forces due to the end rest raints caused by wind locks. S ectional doors do not have this feature thus the performance of door is usually of greater interest than the door jamb. 2.1.4 Current Design Philosophy of Sectional Doors In addition to the components of a normal garage door (e.g., door panels, track, end support components and etc.), wind load rated garage doors usually include additional components for wi nd resistance, e.g., rolling doors use wind locks, and sectional doors add reinforcing beams attached to door panels. C urrently, the material mostly used in a garage door is metal (e.g., steel, aluminum) Fig. 2 4 shows the main components of a wind load r ated garage door, sectional garage doors in particular. Terminologies commonly used to name these components are also shown in Fig. 2 4 (marked in red), and these terminologies are widely used in Chapter 4. Different from current design philosophy of secti onal doors (i.e., using metal material and containing many components), this study explores a new design philosophy of sectional doors using sandwich composite panel with the purpose of developing a simple (i.e., fewer components) door with high resistance to hurricane wind loads. 2.2 Three Layer Composite Panel Composite panels are becoming more prevalent in civil engineering applications particularly three layer structural insulated panels comprised of (1) two thin outer

PAGE 29

29 layers (also skins ) and (2) a thick core layer (Fig. 2 5). Benefits of this type of system include long term durability, a high strength to weight ratio, excellent impact energy absorption, and good temperature and moisture insulation (Khemani Kishan C. 1997; Klempner et al. 2004) The skin material is normally a stiff material such as glass laminates or a carbon fiber reinforced thermo plastics The core material, common ly composed of foam, is a low strength and low density material. Belingardi et al. ( 2003) investigated the material properties of the skin (glass fiber) and core material s (polymeric foam) of a sandwich panel structure through a series of static and dynamic test s, and t h e testing results showed that the strength of the core material significantly affects the behavior of the sandwich panel F oam can be categorized into ope n cell and closed cell foam. Usually, open cell foam is flexible, while closed cell foam is more rigid (Sivertsen 2007) Because foam can be manufactured to obtain different types of mechanical properties, it can be us e d as structural insulted panels for a wide range of applications. However, the mechanical properties of foam have a significant dependence on density, loading strain rate, temperature, and humidity and f oam is a highly anisotropic material (Maji et al. 1995) In addition, several uncert ainties involved in foam material the void contents in particular (Wouterson et al. 2005) considerably affect the material properties. T herefore it is necessary to understand the mechanical performances of th e material before putting it into a wide variety of construction applications. In general, foam behaves differently in te nsion and compression: U nder uniaxial compression, foam exhibits three stress strain behaviors: a linear elastic region under low loads, a plastic plateau at higher loads, and a densification region due to a large number of microspheres being crushed and c ompacted (Wouterson et al. 2005) prior to failure

PAGE 30

30 Under uniaxial tension, the foam material fails in a brittle manner. Many studies have been performed to investigate the influence of different factors on the mechanical properties of different kinds of foams (Lin 1997; Viot et al. 2005; Wouterson et al. 2005; Zhang et al. 1998) Lin ( 1997) studied two groups of foams: PVC commercial foam and polystyrene foam. For the PVC commercial foam, Lin ( 1997) investigated the effects of foam density and strain rate, and he observed a higher density produced a higher strength of the foam, and an increasing in strain rates also increased the yielding strength in compression. For polystyrene foam, temperature effect was studied and results showed that the compressive strength decreased with an increasing temperature Zhang et al. ( 1998) studied the constitutive behaviors of low density polymeric foam materials under different strain rates and temperatures, and observed that the constitutive behavior of the studied foam was significantly affected by strain rate and temperature. Viot et al. ( 2005) conducted uniaxial compression testing on polypropylene foams at several densities and strain rates This work confirmed previous observation that higher strain rate s resulted in a higher stress threshold (e.g., a higher yielding strength in compression ) Wouterson et al. ( 2005) studied on syntactic foam with three different microstructures and showed that the strength of the foam decreased with increasing the voids in the foam. 2.3 Methods to Investigate Structural Performance in Hazard Engineering The p erformance of critical facilities located in hazard prone (e.g., hurricanes, earthquakes) regions has always been a major concern (e.g., hospitals, schools, fire stations). In order to determine responses of a structure or structural components under extreme load conditions, various structural testing methods have been developed. This

PAGE 31

31 part briefly rev iews different structural testing methods (i.e., full scale structural testing, reduced scale testing and numerical modeling) and the methodology used in this study. Full/Large scale structural testing is an ideal way to understand the performance of a str ucture under a certain load condition (e.g., Baskaran et al. 1999; Boughton 1983; Chowdhury et al. 2013; Morrison 2010; Okamoto 2010; Riley and Sadek 2003; Sinno 2008 etc ). However, due to the cost, limited number of tests and test specimens can be studied usin g this method, and therefore, it is not feasible to perform a statistical analysis of the results and evaluate their uncertainty (Riley and Sadek 2003) Additionally, it is impractical to test numerous scenarios (e.g., various loading conditions) on structures using this m ethod (Blakeborough et al. 2001) structures of large siz e in particular, and therefore, performance based design of structures using this method is unrealistic. Compared to full/ large scale testing, reduced scale testing is more economic and therefore more often employed to study behaviors of structures (e.g. Cholod 1988; Collins 1986; Kind 1986; Meecham 1992; Stopar 1987; Visscher and Kopp 2007 etc ). This testing met hod is able to produce acceptable results on global behaviors of real structure, while it cannot always capture the local behaviors or important behaviors N umerical modeling is also a widely used method today (e.g., He et al. 2001; Kasal et al. 1994; Kumar et al. 2012; Thampi et al. 2011 ). However, considering accuracy, e xperimental validation of numerical models is necessary when using this method (Huang et al. 2010) T he methodology used in this study applies full scale experimental testing and numerical simulatio n. Full scale experimental testing provides a more realistic test

PAGE 32

32 method for building components. However, i t is impossible to conduct full scale experimental testing on a structure in many different scenarios due to the cost. Therefore, numerical method c omplement s this research. N umerical model s are built to recreate the scenario in the full scale experimental testing and then validated using the full scale experimental measurements. The validated numerical model can be modified and tested under various s cenarios (e.g., different hurricane intensities). This approach, combining full scale experimental testing and numerical analysis, offers a reliable and cost efficient way to determine the fragility of building components under various hazard intensities w hich they are likely to meet over their lifetime, thus can be used for the performance based engineering approach (McCullough and Kareem 2011 ) 2.4 Full/Large Scale Testing Facility A full/large scale wind testing facility replicate s wind flow or wind pressure derived from data measured from boundary layer wind tunnel model experiments or full scale testing buildings. By subjecting full/large scale building components or other structures to replicated wind conditions, the facility ser ves as a tool to comprehend structural behaviors in windstorms. However, there are only a few of this kind of facilities Examples of wind load simulation apparatuses include t he BRE Real time Wind Uniform Load Follower (BRERWULF) (Cook et al. 1988) and Pressu re Loading Actuators (3LP PLAs) (Kopp et al. 2010 p. 2010) Examples of wind flow simulation devices include Wall of Wind (Chowdhury et al. 2009) and IBHS Windstorm Simulatio n facility (Liu et al. 2009) In the foll owing section only the wind load simulation apparatuses that are of interest in this study will be discussed.

PAGE 33

33 2.4.1 BRERWULF The BRE Real time Wind Uniform Load Follower (BRERWULF) which was developed by the Building Research Establishment (BRE) in the United Kingdom, is the first notable project aiming to recreate wind loads on building system BRERWULF is mainly used to investigate the performance of multi layer cladding systems subjected to temporally fluctuating wind loads. This system utilized a clo sed loop control system which changed pressure in a test chamber to follow a prescribed stochastic pressure sequence, and it was designed to generate any wind pressure sequence with peak value in the range of 178 psf and frequency not higher than 10 Hz (Cook et al. 1988) 2.4.2 3LP PLAs In the 2000s, the University of Western Ontario (UWO) named the Pressure Loading Actuator (PLA) (Fig. 2 6 A). Each PLA is able to recreate temporally fluctuating wind pressure, and spatial variations of wind pressure can be realized by using many PLAs simultaneously (Fig. 2 6 B). The valve control system was designed to obtain a linear relationship between the pressure and valve position, whic h made control easier (Kopp et al. 2010 p. 2010) This system is able to generate wind pressure of the range from 418 psf to 480 psf and fluctuating frequenc y not higher than 10 Hz (Kopp et al. 2010) 2.4.3 HAPLA Based on the concept of the PLA system, the University of Florida developed the High Airflow Pressure Loading Actuator (HAPLA) (Fig. 2 7). The HAPLA was designed to generate time varying pressure loading to the surfaces of building components (e.g., windows shown in Fig. 2 7) (Lopez et al. 2011) The fan system of the HAPLA is

PAGE 34

34 comprised of two 75 hp centrifugal blowers that can operate in series or parallel. The blowers are connected to a valve with a rotating disk that is actuated by a rotary actuation system. A pressure transducer attached to the airbox provides the pressure feedback. Based on the pressure feedback, t he actuation system controls the position of the valve d isk and therefore adjusts the pressure in the airbox. The airbox is a stiff steel box with a dimension of 8 ft. wide x 8 ft. high x 1 ft. deep 2.5 Summary C hapter 2 presented the background material for the original contributions to be discussed in Chapter 3 through 5. An introduction of wind rated garage doors was discussed with a focus on sectional garage doors. The main components and the terminologies used to name them were introduced, and will be used in Chapter 4. Current design philosophy of sectional garage doors was presented showing its shortcomings that partially inspired the research presented in Chapter 5. Current methods for structural testing subjected t o extreme loading conditions were discussed. Existing systems for full scale structural testing were presented. C hapter 3 presents methodologies used in this research.

PAGE 35

35 Figure 2 1. Wind rated garage doors. A). Sectional garage doors. B) Rolling doors ( Photos courtesy of Google)

PAGE 36

36 Figure 2 2. Failure s of garage doors in hurricanes (Photos courtesy of Croft et al. 2006 FEMA 2008 and FEMA 2008 respectively)

PAGE 37

37

PAGE 38

38 Figure 2 3. Experimental and numerical studies on a 10 ft. x 10 ft. rolling door. A) Experimental setup (Photo courtesy of Gao and Moen 2009 ) B) Door frame details (Photo courtesy of Gao and Moen 2009 ) C) FE model of the rolling door employing jamb stiffness for the boundary conditions (Photo courtesy of Gao and Moen 2011 ) D) Door deflection c omparison between experimental data and FEA results (Photo courtesy of Gao and Moen 2011 )

PAGE 39

39 Figure 2 4. Typical sectional garage door details (Main components and their names are marked in red) (Photo courtesy of Raynor )

PAGE 40

40 Figure 2 5 A photograph of the composite structure sample studied in this research (Photo courtesy of the author)

PAGE 41

41 Figure 2 6 3LP PLAs developed at the University of Western Ontario (Photos courtesy of Kopp et al. 2010 ) A) A CAD model of a single PLA unit with its major components. B) Several PLAs connected to the roof of a test house

PAGE 42

42 Figure 2 7. High Airflow Pressure Loading Actuator (HAPLA) (Photo courtesy of Lopez et al. 2011 ).

PAGE 43

43 CHAPTER 3 METHODOLOGY As discussed in Chapter 2, the major trade off in th e investigat ion of structural performance of full scale test subjects in accuracy vs. cost. Therefore, to address this issue, a methodology combining full scale experimental and numerical methods wa s employed in the present study. C hapter 3 introduces the methodology with a focus on the full scale experimental testing apparatus, namely, a new large scale dynamic wind load s imulator developed in this study. 3.1 Full Scale Experimental Testing To conduct full scale experimental testing, an apparatus wa s re quired to replicate hurricane wind pressure and subject test specimens to the simulated loads. Due to the issue of airflow leakage through test specimens, the existing systems discussed in Section 2.4 i.e., BRERWULF and 3LP PLAs (Cook et al. 1988; Kopp et al. 201 0) are not ideal for this study for testing garage doors. Therefore, a new large scale dynamic wind load s imulator (henceforth Simulator) was developed to recreate extreme wind flow and wind loads to test large/full scale building components and claddi ng systems. This study only focuses on wind pressure simulation, and therefore wind flow simulation part of the Simulator is not discussed herein. The Simulator was designed to : Accurately r ecreate intense hurricane wind load situations, i.e., up to Catego ry 5 hurricanes in a controllable laboratory environmen t. Compensate for high airflow leakage through test specimens during testing. A pply spatially uniform time varying pressures on the surface of large/full scale build ing components to i nvestigate the ir responses under hurricane wind loads. Validate and enhance the numerical simulation of structures, which is the other method used in this study. To design the Simulator in this study the main items involved are:

PAGE 44

44 Determining the design wind pressure ti me history that will be recreated to calculate the pressure requirement of the Simulator Calculating the parameter requirements of the fan system based on the design wind condition and other requirements and select the appropriate equipment meeting the calculated requirements. Designing and develop ing a control system to realize the design wind flow and wind loads. These three items are discussed in detail in the following sections from 3.1.1 through 3.1.3. 3.1.1 De termining the Design Wind Pressure of the Simulator Determining the design wind pressure (i.e., the worst case scenario of wind loads to be replicated) wa s necessary to design and select the components of the Simulator. For example, the required peak press ure determine d the selection of the fan system that was to provide pressure and airflow into the airbox (will be discussed in Section 3.1.3.1) during full scale tests Th e research described in this doctoral work focuses on low rise buildings which are com mon for commercial, residential and industrial use. The US National Institute of Standard and Technology (NIST) created an aerodynamic database of wind induced pressure time histories on the envelope of various low rise buildings (Chen et al. 2003) and was chosen to determine the design wind pressure for this Simulator The wind tunnel data chosen for this research were measured from generic low building models of a 1:100 scale (Ho et al. 2003) The selected building models (m11, m21, m31 and m41) include four models with the same plan dimension of 80 ft. x 125 ft. and the same roof slope of 3:12 but four different eave heights of 16, 24, 32 and 40 ft ( Fig. 3 1 ). Pressure coefficient time series ( ) were obtained from approximately 700 pressure taps equipped on the building envelope of each building model, and these data were measured for 37 wind angles from 180 to 360 at 5

PAGE 45

45 intervals ( Fig. 3 2 ). Of the two exposure conditions (open country and su burban) wind effects are more severe for this exposure. This research focuses on building components and cladding system related to walls of buildings, and therefore the press ure coefficient data collected from the taps on building walls were extracted from the NIST Aerodynamic Database. From the hundreds of wind pressure taps in 37 wind directions of each building model, the pressure coefficient time se ries with the highest value were found out. T he worst case scenario of was chosen from these four pressure coefficient time series. The procedure is illustrated in Fig. 3 3 And Fig. 3 4 shows the worst case of chosen to calculate design pressure for the Simulator. To calculate the design pressure of the Simulator, the following two conversions were necessary The wind pressure coefficient data from wind tunnel model w ere measure d at a reference height in the boundary layer wind tunnel (Ho et al. 2003) and therefore they were converted to wind pres sure coefficient at roof height to be used in the following calculations for the design pressure of the Simulator. Time scale needs to be converted from wind tunnel to full scale. And this is why the time length in Fig. 3 5 is different from that in F ig. 3 4. Details on the conversions are described in Appendix A. For components and cladding systems of low rise buildings, pressure p can be calculated via Equations 3 1 and 3 2 according to ASCE 7 ( 2005) (3 1) (3 2) W here, the dynamic velocity pressure (psf).

PAGE 46

46 the topographic factor, assumed to be 1.0 in this study. the wind directionality factor, assumed t o be 1.0 in this study. 155 mph 1.1, representing 3 s gust wind velocity at mean roof height associated with Category 5 hurricane wind condition. the importance factor, equal to 1.15 in this research to represent for critical facilities. the velocity pressure evaluated at the mean roof height. the external pressure coefficient, is equal to the worst case of full scale wind pressure coefficient the internal pressure coefficient. The wind pressure time histo ry corresponding to the worst case of is shown in Fig. 3 5A and the sampling frequency corresponding to the converted full scale data is 32 Hz. For the design of this Simulator, the wind pressure data was filtered to 10 Hz (Fig. 3 5B), which is well above the upper limit of the frequency domain of wind pressure spectra. 3.1.2 System Design Algorithm 3.1.2. 1 An o verview of the Simulator The Simulator has six main components: an engine, a fan system, ducting, a control system, pressure chamber), and a reaction frame system ( Fig. 3 6 ). The fan is to produce the required pressure and airflow rate, and the engine is to drive the fan. The ducts connecting the fan and the airbox are to convey the airflow. The contro l system is to determine the principal direction of loading (i.e. positive pressure or negative pressure), and realize the desired pressure in the pressure

PAGE 47

47 chamber by changing the system resistance. The pressure chamber with the reaction frame is to be mou nted with test specimens for testing. 3.1.2.2 Fan s ystem d esign An appropriate fan is ought to provide sufficient pressure in the airbox and compensate airflow leakage during testing For this research, the air flow supplied from the fan through the ducts to the pressure chamber (positive pressure situation) or the opposite direction (suction situation), acted like a slug of air moving in or out of the opening of the pressure chamber, which is sketched in Fig. 3 7 S ingle d ischarge e quation The differential equation for the motion of a slug of air moving in and out of a chamber with volume of (Vickery 1986) is given by Equation 3 3 ( 3 3 ) w here, = the air density, = effective length of the opening, = the distance of air slug moving in and out of the volume, = the loss coefficient due to orifices, = polytropi c gas constant (for air is equal to 1.4), = the atmospheric pressure, = the area of the opening, = the internal volume of the pressure chamber = the external pressure, and = the pressure in the test chamber, which is also the target pressure applied on test specimens. Continuity e quation The mass conservation equation in the chamber (Oh 2004; Vickery 1986) is shown in Equation 3 4 ( 3 4 ) Where, = the bulk modulus of air, and represents the number of openings

PAGE 48

48 Modifications a ccounting for the l eakage and f lexibility of s pecimens T his Simulator was designed to account for large leakage and deflections occurring through specimens during hurricane wind pressure testing, which is sketched in Fig. 3 8A Considering openings due to leaky specimens, multiple discharge equations were required; however, to simplify solving the equations for uniform, leaky openings with similar geometry multiple openings at the leeward surface (Fig. 3 8A) were lumped into a large opening sketched in Fig. 3 8B (Vickery 1986) Therefore, two discharge equations were used for the design of the Simulator. As recommended in ASHRAE Handbook: Fundamentals (ASHRAE 2001) a flow exponent can be incorporated in the second term of Eq uat ion 4 3 (i.e., ) to account for leakage in full scale structures, i.e., using to replace To account for the effects of flexible specimens on volume change of the chamber, the origin al volume (Fig. 3 8) was increased to where represents the bulk modulus of the chamber (Vickery 1986) In this study, a value of 30% was used to approximate the value of Therefore, considering both leakage and volume change due to test specimens, the equations to determine the parameters for fan selection are: (3 5) (3 6) (3 7)

PAGE 49

49 w here, is the area of opening 1 (Fig. 3 8B) connected to the ducting system, is the area of opening 2 accounting for the leakage caused by specimens (Fig. 3 8B), and represent effective length s of opening 1 and 2 respect ively, is the air movements into and out of the test chamber through ducts, represents the air movements into and out of the test chamber through the leaky specimen, and represent loss coefficient s due to opening 1 and 2 respectively, a nd they are assumed to be 2.5 (Vickery 1986) represents the pressure at the end of the ducting system, and is equal to the atmospheric pressure out of the specimen. The values and were solved by applying backward difference method ( Appendix B) to Equations 3 5 to 3 7 and then the required airflow was calculated using Equation 3 8 (3 8) Consideration for p ressure l oss from d ucts The pressure loss due to the ducting system was due to f r iction and dynamic loss es Friction loss occur red along the entire ducts, and can be calculated by Equation 3 9 (Vedavarz et al. 2007) : (3 9) (Equa tion source: Equation 7 in Chapter 8 in Vedavarz et al. 2007 ) w here, ducting friction loss, in. of water. friction factor ( Equation 8 in Chapter 8 i n Vedavarz et al. 2007 ). length of ducts, in. hydraulic diameter, in. ( Equation 8 in Chapter 8 in Vedavarz et al. 2007 ).

PAGE 50

50 velocity of air flow, fpm. density of air, lbm/ft 3 Dynamic loss occur red when there were flow disturbances caused by duct fittings which change d n and/or area (e.g. ducting entries, elbows, contraction elements and etc.). Fluid resistance of fittings can be calculated using Equation 3 10: (3 10) ( Equation source: Eq uation 27 in Chapter 8 in Vedavarz et al. 2007 ) w here, ducting dynamic loss, in. of water. fitting loss coefficient (Vedavarz et al. 2007) Summary of f an s ystem d esign According to the previous discussion, the air flow and pressure requirements for the fan system can be calculated using Equation s 3 8 and 3 11: (3 8) (3 11) The pressure and airflow requirements calculated for the fan system were 436 psf and 100,000 CFM, and a 3.1.3 Introduction of the Simulator Fig. 3 6 presented the main components of the Simulator in a 3D drawing. This section presents details on the technical information of these comp onents and the control system.

PAGE 51

51 3.1.3.1 Technical i nformation about the S imulator The technical information of the components of the Simulator is listed as follows. According to calculations from the previous section, the fan of choice for the Simulator is a centrifugal fan ( Fig. 3 9 ) I t is able to generate pressures in the ranges of in. of water (i.e., 468 psf) with 100,000 CFM airflow and compensate for airflow leakage up to 50,000 CFM A Caterpillar 3512 DITA diesel prime mover with a power of 1818 hp was used to drive the fan through engaging the air clutch connecting them ( Fig. 3 10 ). The ducts connecting the fan and the pressure chamber were made of 3/8 in. thick steel. The cir cular parts of the ducting had an internal diameter of 60 in. In addition, two VAW duct silencers were incorporated in the inlet and outlet of the ducting system respectively to isolate the noise while running the Simulator ( Fig. 3 9 ). The control system is comprised of five dampers: four butterfly dampers ( Fig. 3 11A ) and a custom built fast acting opposed blade louver damper ( Fig. 3 11B ). Details on the control system are discussed in detail in Section 3.1.3.2. The airbox has dimensions of 24 ft wide 18 ft high 3 ft deep ( Fig. 3 12A ). It was made of reinforced concrete with approximately 13 tons of mild steel reinforcements ( Fig. 3 12B ). In addition to mild steel reinforcements, 1 in. diameter post tensioning bars ( Fig. 3 12A ) were used to avoid crac ks of the concrete. The reaction frame system consists of primary and secondary reaction frame ( Fig. 3 13A ). During testing, t he test specimen wa s mounted in the primary reaction frame that wa s clamped to the open side of the pressure chamber to close the pressure chamber for testing. The primary frame is to take the catenary forces developed from the test specimen subjected to wind pressure loading. The primary frame was made of HSS 16 X 16 X 3/8, and has a force capacity of 10 kip/ft. The secondary react ion frame can be removed based on the dimension of the specimen ( Fig. 3 13B ) 3.1.3.2 C ontrol s ystem Four butterfly dampers and a custom built fast acting opposed blade damper compose hardware of t he control system. B utterfly dampers 1 and 2 are the exteri or intake and exhaust dampers, and butterfly dampers 3 and 4 are to convey airflow into and out of the test chamber for pressure simulation ( Fig. 3 11A ) T he four butterfly

PAGE 52

52 dampers are employed to set the mode of operation (i.e., positive pressure or negat ive pressure) by changing the flow configuration For example, opening butterfly dampers 1 and 3 allows airflow to be drawn from the outside and enter into the pressure chamber, and thus positive pressure can be achieved and applied to test specimens. Thr ough changing the inclined angle of the blades ( Fig. 3 11B ), t he blade louver damper serves to modulate the system resistance which changes the airflow in/out of the pressure chamber and thus causes a corresponding change in the pressure acting on the spe cimen (i.e., the operating point on the fan curve changes) ( Fig. 3 14 ) T he blade louver damper is activated by a hydraulic servo cylinder implementing the analog proportional integral derivative (PID) technique. During testing, t he pressure is measured in side the test chamber and monitored by a custom analog computer. The computer sends commands to an analog servo that actuates the hydraulic cylinder driving the louver damper. The control/feedback process is continuous; only analog feedback/control was imp lemented (i.e. no A/D or D/A occurs). 3.1.4 Current Status of the Simulator The components were assembled together in the Powell Family Structures and Materials Laboratory at the University of Florida. And the Simulator has been put into use for static pressure testing. The dynamic capacity of the Simulator was calibrated in October 2013 3.2 Finite Element Analysis M ethod The FEA method is widely used in the area of structural engineering. Many programs are available and the program of choice in this study is ADINA (Automatic Dynamic Incremental Nonlinear Analysis) (ADINA 2010)

PAGE 53

53 As discussed at the beginning of C hapter 3 FEA complementing full scale experimental testing is utilized in this study for accuracy and cost. In this study, FEA is employed in the following way: A finite e lement (FE) model simulating the full scale experimental testing is firstly created, including modeling the structure tested, boundary conditions and loads. FEA results (e.g., strains, displacements) are validated through the corresponding full scale exper imental measurements. If the results match experimental measurements, it can be moved to next step, and if not, the FE model needs to be adjusted until the results match experimental measurements. Modifications on the validated FE model are to be conducted and analyzed for various scenarios (e.g., different loads, different designs). 3.3 Summary C hapter 3 presented the methodology employed in this study, i.e., both full scale experimental and numerical methods. The methodology offered us a reliable and cos t efficient way for: 1) conducting structural analysis on complex structures, such as garage doors, which will be discussed in detail in Chapter 4; 2) the R&D process of a new design, which will be discussed in Chapter 5. Successfully using this methodolog y in the studies conducted in Chapter 4 and 5 also indicates its potential for performance based design of building components through testing them under various hazard intensities which they are likely to meet over their lifetime. C hapter 3 also presented a large scale dynamic wind load s imulator developed at the University of Florida for full scale experimental testing. The unique aspect of the Simulator is that it was designed to compensate for high airflow leakage through test specimens up to 50 000 CFM. T he system can be used to test garage doors as well as other building components and cladding systems (e.g., building walls, windows).

PAGE 54

54 F igure 3 1. Wind tunnel models employed for determination of the design wind pressure for the Simulator discussed in this research ( Photo courtesy of NIST Aerodynamic Database)

PAGE 55

55 Figure 3 2. Pressure tap layout of the four wind tunnel models (m11, m21, m31 and m41) (Photo courtesy of Ho et al. 2003 )

PAGE 56

56 Figure 3 3. An illustration on the procedure to find out the worst case of C P for the design wind pressure

PAGE 57

57 Figure 3 4. The worst case of from wind tunnel model m11, m21, m31 and m41

PAGE 58

58 Figure 3 5. The worst case scenario of wind pressure time histories for the design pressure of the Simulator. A ) The frequency of original data is 32 Hz B ) Data were filtered to 10 Hz for the design pressure of the Simulator

PAGE 59

59 Figure 3 6. 3D drawings of the Simulator with the major components. A) An overview of the Simulator located in the Powell Family Structures and Materials Laboratory at the University. B) A zoomed view of the Simulator without the reaction frame and the engine.

PAGE 60

60 Figure 3 7. A sketch of air slug moving in/out of a chamber through the opening

PAGE 61

61 Figure 3 8. Sketches of the fan design concept accounting for leakage and flexibility of specimens. A) The design concept of the Simulator considering leak age and deflections of specimens. B) A simplified model using a concentrated opening replacing distributed openings

PAGE 62

62 Figure 3 9. A photograph showing the fan system and duct silencers of the Simulator (Photo courtesy of the Powell Family Structures a nd Materials Laboratory at the University of Florida )

PAGE 63

63 Figure 3 10. Photographs of the prime mover (engine) and air clutch of the Simulator (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 64

64 Figure 3 11. The control system of the Simulator. A) The four butterfly dampers. B) The louver damper (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 65

65 Figure 3 12. Photographs of the airbox. A). The airbox and post tensioning bars. B). Mild steel reinforcements (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 66

66 Figure 3 13. The reaction frame system of the Simulator (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida) A) Components of the reaction frame system. B) An illustration showing the secondary reaction frame can be removed

PAGE 67

67 Figure 3 14. A schematic of a t ypical fan curve and several system resistance curves

PAGE 68

68 CHAPTER 4 FORENSIC ASSESSMENT OF CURRENT GARAGE DOORS USING FULL SCALE EXPERIMENTAL AND NUMERICAL METHODS As discussed in Chapter s 1 and 2, few research studies ha ve been conducted on the performan ce of garage doors subjected to hurricane wind pressure. C hapter 4 describes a new study intended to fill the gap in understanding of the structural responses and failure mechanisms of sectional door systems In order to explore this research objective, fu ll scale experimental testing and finite element method were used. Five commercial sectional doors were experimental tested using the newly developed hurricane pressure Simulator T wo of the five test specimens were modeled in the finite element analysis p rogram ADINA. Due to the status of the newly developed hurricane Simulator described in Chapter 4, i.e., the dynamic capacity was under calibration when this study was performed only static pressure until failure testing was conducted on the five test specimens C ha p t er 4 presents details on the experimental testing and numerical modeling Results from these two methods are presented and compared to each other. Two princip al failure mechanisms were observed in ex perimental testing, namely, local buckling of reinforcing beams (U bars) and disengagement of the roller s from the tracks. Finite element model results match ed experimental measurements, demonstrating the applicability of using FEA complementing full scale experimental testing as a cost effective and reliable way to evaluate structural performance of garage doors. Then construction errors were evaluated through modifications on FE model of prototypical door. Additionally, the responses of current sectional doors indicate d a potential of introducing a new design philosophy of sectional doors by using composite structures (discussed in Chapter 5).

PAGE 69

69 4.1 Full Scale Experimental Testing 4.1.1 Materials Five commercial sectional doors from different manufacturers were evaluated. Table 1 provides the dimension and design pressure of the five test specimens. The design pressure of the five sectional doors was stipulated in product documentation maintained by the Miami Dade County Product Control Division. 4.1.2 Exp erimental Assumptions 4.1.2.1 Concerning the l oading d irection From Table 4 1, it is observed that the design pressure s of sectional doors are different in positive and negative loading directions To expedite the testing only one loading condition was to be simulated using the Simulator to load the specimens. To determine the pressure direction that would be applied on the five sectional doors, two identical sectional doors (18 ft. wide x 7 ft. high) were tested under dynamic wind pressure using the Simul ator described in Chapter 3 (Fig. 4 1). The two sacrificial doors were tested only for a visual observation on the ir responses under dynamic wind pressure situations, and therefore no data were recorded. We observed that d uring sacrificial tests, there were more failure mode s occu rring in suction than in positive pressure test In positive pressure test, failure occurred by overbending of roller stems (Fig. 4 2) However, in suction negative pressure test other failures (Fig. 4 3) were observed (e .g. local bucking of stiffeners, de bonding between the door panels and the stiles and disengagement of rollers from the tracks ). T he sectional doors have a larger design negative pressure than positive pressure ( Table 4 1 ). Thus all tests applied suctio n to the door specimens

PAGE 70

70 4.1.2.2 Concerning the a ttachment of t est s pecimens to b uildings As mentioned in Chapter 2, this study focuses on structural behaviors of door itself without considering the performance of other related structures, e.g., the stiffness of door jamb (cf. Gao and Moen 2009) Therefore, it was assumed that the structural framing of buildings where the five test doors might be installed would be strong enough to carry the catenary forces and w ould not fail during tests. The primary reaction frame of the Simulator has a load capacity of 10 kips/ft which provides ample stiffness to ju stify this assumption 4.1.2.3 D efinition of d oor f ailure Failure is defined as one of the following conditions: a) the door is not operable any longer, e.g., the door disengages from its track; b) significant local failure (e.g., local buckling of U bars) occurs at some certain components of the door system. 4.1.3 Testing instrumentation To obtain a comprehensive understanding on responses of the five door specimens under simulated pressure the following instrumentations were used during full scale pressur e testing To measure reaction forces, t hree SDI Tri axial 70148 00B EA00C load cells were installed between the door jamb and the reaction frame of the S imulator (Fig. 4 3A). The load cells were located around the rollers at bottom, middle and top of the specimen (Fig. 4 3B ). The LabVIEW data acquisition system was used to collect the measurement data during testing. To measure strain responses, e ight Micro Measurement C2A 06 250LW 350 strain gauges were attached on eight U bars at their mid spans (Fig. 4 3C ) To measure displacement responses, p hotogrammetry with sub millimeter accuracy was used for displacement measurement by recording the coordinates of traceable spheres attached on U bars at their mid spans (Fig. 4 3D )

PAGE 71

71 4.1.4 Testing Load FBC TAS 202 (FBC 2010) ( Criteria for testing impact & nonimpact resistant building envelope components using uniform static air pressure) is generalized below : (1) Apply half of the test load (is equal to 1.5 times of the design load) across the specimen and hold for 30 seconds. (2) Rel ease the load and allow a recovery period of not less than 1 minute no more than 5 minutes. (3) Apply one half the reverse test load and hold for 30 seconds. (4) Release the reverse load and allow a recovery period of not less than 1 minute no more than 5 minutes. (5) Repeat step1 4 with full test load. The purpose of FBC TAS 202 is to determine whether the tested product is able to provide sufficient resistance to wind forces, while this study aims to obtain failure mechanisms and detailed responses of sectional doors Therefore, a step and hold negative pressure sequence with 10 psf decrement was applied to each test specimen until failure (Fig. 4 4). Each load was held for 5 minutes and then released to check the operability of the test specimen. 4.1.5 Testing Proc edure The testing procedure wa s as follows (1) Adjust the opening of the air box for the specimen to be tested. (2) Install three load cells between the reaction frame of the air box and the door tracks at the predetermined locations (3) Attach strain gauges to eight U bars of the test specimen (4) Install the specimen into the opening of the air box according to the installation introduction provided by the manufacturer (5) Instrument the specimen with deflection globes (6) Apply uniform static pre ssure on the specimen in the pattern shown in Fig. 4 4

PAGE 72

72 4.2 Finite Element Analysis Based on the failure mechanism s observed in the full scale experimental testing, doors 2 and 3 were decided to be modeled using finite element (FE) model Three resources used to obtain properties for modeling the sectional doors were : 1) The Notice of Acceptance (NOA) drawing of each specimen, 2) dimensions measured on the physical specimen, and 3) observations from full scale experimental testing. 4.2.1 Finite Element Mo del of Door 2 4.2.1.1 Assumptions and s implifications in FE m odel of d oor 2 For FE modeling of a complex structure, such as a sectional door, it is impractical to model every component of the structure due to the time. Material properties of some component s of door 2 were not provided in the NOA drawing Therefore, the following assumptions and simplifications were made when building the FE model of door 2 in ADINA 8.8 (ADIN A 2011) : (1) Hinges and fasteners were not physically modeled, while their constraining effects were simulated in the FE model. (2) For the components whose material properties were not given in the NOA drawing, their properties were assumed to be those of the commonly used. For example, the material properties of blockings of door 2 were not provided in its NOA drawing, and properties of No. 2 southern yellow pine were used in it FE model. (3) The rollers were not included in the model, and instead the constraints from them were included and treated as boundary conditions along the sides of the model. This simplification can be improved in the future if necessary. (4) The effects of gravity are negligible and were not considered in the model. 4.2.1.2 Elemen t c omponents in FE m odel of d oor 2 Fig. 4 5 shows t h e FE model of d oor 2 created in ADINA 8.8. H alf of the door was modeled due to the symmetry. Table 4 2 summarizes the main components in the

PAGE 73

73 FE model. The FE model considered contact constraints, material nonlinearity and second order effects. 4.2.1.3 Boundary c onditions and c onstraints To simulate the constraints of door 2 in experimental testing, boundary conditions described in Fig. 4 6 were used for the model. S ymmetrical boundary conditions were appl ied along the vertically spanning centerline (Fig. 4 6A) The effects of hinges connecting each two adjacent door panels were modeled using corresponding constraints at the locations where hinges were located Because no adhesive failure was observed for d oor 2 during experimental testing, the bond between stiles and door panels can be reasonably assumed to be perfect, therefore rigid links were used to connect stiles to door panels (Fig. 4 6B). A r igid link in ADINA is similar to a spring with infinite sti ffness at all six degrees of freedom constraining the linked nodes so that they move together when the model deforms. R igid links were used between the astragal retainer (used as the bottom seal of garage doors) and the bottom of the door panel (Fig. 4 6B) The fasteners used in the physical door (e.g., fasteners connecting U bars to stiles, and fastener s connecting blockings to stiles) were also modeled as rigid links. Contact boundary conditions were used between surfaces wherever there were no constraint s but contact might occur, i e ., the bottom of one door panel and the top of the panel below it (Fig. 4 6B) 4.2.1.4 Material p roperties The material properties of door panel s stile s and U bars were taken from the NOA drawing provided by the manufacturer. However, the material properties of the wood blocking and aluminum were not given, and therefore, typical properties were assumed. As mentioned in section 4.2.1.1, f or the blocking material properties of No. 2

PAGE 74

74 souther n yellow pine were used : elastic modulus density F or the aluminum, the elastic modulus density ratio and yielding strength are 10000 ksi, 1676 pcf, 0.33 and 40 ksi res pectively 4.2.1.5 Loads To simulate the experimental testing, the load applied to door 2 in the experimental testing was used in the FE model, i.e., step and hold loading pattern with 10 psf decrement (Fig. 4 4). 4.2.2 Finite Element Model of Door 3 4.2.2.1 Assumptions and s implifications in FE m odel of d oor 3 Similar to door 2, the assumptions 1 to 3 of the FE model of door 2 also applied to door 3. However, different from door 2, door 3 failed by disengagement of the rollers from the tracks, which will be shown in Section 4.3.1.1. Therefore, it wa s necessary to include rollers in the FE model of door 3. To model rollers, three dimensional solid elements w ere used ( Fig. 4 7 ). For door 3, there are twenty rollers in tot al i.e., ten per side If all the ten rollers were modeled in the way shown in Fig. 4 7 in the FE model of door 3, the analysis w ould be extremely time consuming. To reduce the computation time, a strategy of hierarchical modeling was employed for the com putational analysis of door 3: firstly, a local FE model includ ing a roller, the roller stem and 21 in. long track was created in ADINA to determine the constraints from the track to the roller. The length of the track was 21 in. because the roller to roll er spacing is 21 in. Secondly, the constraints were simulated as a nonlinear spring at the location of each roller in the FE model of the whole door 3. T he following assumptions were made for the FE model of door 3 (assumptions 1 to 3 for the FE model of d oor 2 were included as well):

PAGE 75

75 (1) Hinges and fasteners were not physically modeled, while their constraining effects were simulated in the FE model. (2) The effects of gravity are negligible and were not considered in the model. (3) For the components whose mate rial properties were not given in the NOA drawing, their properties were assumed to be those of the commonly used. (4) For the local FE model the constraints from the track to the roller were expressed with the curve representing the relationship between con tact force at the roller track interface and the roller displacement. (5) For the FE model of the whole door 3, the stiffness of the nonlinear springs used the curve output from the local model. 4.2.2.2 Element c omponents in FE m odel of d oor 3 Local Model o f Roller and Track Fig. 4 8 shows t h e local FE model of a roller and 21 in. long track of door 3. The roller stem and roller were modeled with three dimensional solid elements, and the track wa s modeled with shell elements. Table 4 3 summarizes the components in the local FE model. The local FE model considered contact constraints, material nonlinearity and second order effects. Global Model of Door 3 Fig. 4 9 shows t h e FE model of D oor 3 created in ADINA 8.8. H alf of the door was m odeled due to the symmetry. Table 4 4 summarizes the main components in the global FE model. The global FE model accounted for contact constraints, material nonlinearity nonlinear springs and second order effects. 4.2.2.3 Boundary c onditions and c onstrai nts Local m odel including r oller and t rack Fig. 4 10 shows t h e boundary conditions and constraints applied in the local FE model. X y and z translational degrees of freedom of face 2 of the track (Fig. 4 10) were constrained to simulate the constraints from the track bracket connected to structural framing. Contact boundary conditions were applied at the interface of the roller and the track. The roller stem was

PAGE 76

76 applied prescribed displacement at the part that moved with the door panel, i.e., part 2 sho wn in Fig. 4 10. Rigid links were used to connect the roller stem to th e roller at their interface. Global m odel of d oor 3 Fig. 4 11 shows t h e boundary conditions and constraints applied in the global FE model. Similar to door 2, s ymmetrical boundary con ditions were applied along the vertically spanning centerline (Fig. 4 11A) The effects of hinges were modeled using corresponding constraints Rigid links were used to model: 1) t he effects of fasteners, 2) adhesive effects between stiles and door panels; and 3) the connection between door panels and roller stems. Nonlinear springs were modeled to simulate the constraints from the interaction between rollers and the track (Fig. 4 11B). And the stiffness of the nonlinear springs was from the output of the local model. Fig. 4 12 illustrates how the results from the local model were used in the global model. 4.2.2.4 Material p roperties Local m odel including r oller and t rack The material propertie s of the roller stem, the roller and the track were taken from the NOA drawing of door 3. Global m odel of d oor 3 The material properties of door panel s stile s, U bars and roller stems were taken from the NOA drawing. T he material properties of the alumi num were not given, and therefore, typical properties were used in the FE model: the elastic modulus density and yielding strength are 10000 ksi, 1676 pcf, 0.33 and 40 ksi respectively 4.2.2.5 Loads Local m odel inc luding r oller and t rack A prescribed displacement increasing with time was applied on the roller stem at the part that moved with the door panel to

PAGE 77

77 determine the contact force roller displacement curve for nonlinear spring in the global FE model of door 3 Global m odel of d oor 3 The step and hold loading pattern with 10 psf decrement that was used in the experimental testing of door 3 was applied to the global model. 4.3 Results and Discussions 4.3.1 Full Scale Experimental Results 4.3.1.1 Failure m echanisms During full scale testing on the five sectional doors, the following failure mechanism s shown from Fig 4 13 to Fig 4 17 were observed: (1) Door 3 and 5 experienced a small amount of adhesive failure between stiles and door panels before the catastrophic failure, while door 1 showed a large area of adhesive failure (Figs. 4 13, 4 15 and 4 17). (2) Local buckling of U bars occurred at the catastrophi c failure of door 1, 2, 4 and 5 (Figs. 4 13, 4 14, 4 16 and 4 17). (3) Disengagement of the rollers from the tracks was seen at the catastrophic failure of most of the specimens except door 5 (Figs. 4 13 to 4 17) Therefore, local buckling of U bars and disengagement failure are common catastrophic failure modes for the five sectional doors, and these two ty pes of failure usually occur red together. Table 4 5 lists the catastrophic failure mechanisms observed for each test specimen, and the suction pressure at failure compared to their test pressure. From Table 4 5, it can be noted that some doors experienced the catastrophic failure at a pressure higher than their test pressure; however, their performance under dynamic wind pressure is questionable because some components may experience the fatigue failure at a smaller but repeatedly varying wind pressure, an d therefore dynamic

PAGE 78

78 pressure testing is proposed as future work. Additionally, several different failure modes occurring for each specimen calls for the necessity to define the failure/ failures for garage doors. Fig. 4 18 compare s the characteristics of U bars and roller s of the five sectional doors. Considering the testing results listed in Table 4 5, the following can be concl uded. (1) T he length of the U bar contributes significantly to the wind resistance of the door, because door 3, the only one of the five specimens had a catastrophic failure pressure lower than its test pressure, has shorter U bars compared to the other four specimen s. (2) The ratio of U bar length and thickness is an important factor affecting l ocal buckling failure of U bars 4.3.1.2 Displacement and s train r esponses The instruments described in Section 4 .1. 3 were used to measure the structural responses of the five t est specimens during experimental testing. Figs. 4 1 9 and 4 20 show the variation of measured strains and displacements with an increase in the applied pressure. At the ultimate failure of each door, measured strains and displacements are unusable in plot t ing their relationships with the applied load, and therefore, the maximum load plotted is that before catastrophic failure occurred. From plots in Figs. 4 1 9 and 4 20 the following observations are made: (1) The relationship between strains and applied pressu re is almost linear at a large majority of the locations for each door. Only the strains measured on the U bars located at the middle of the first and fourth panels of door 1 show a sudden change in the curves when the suction was increased from 30 psf to 40 psf (Fig. 4 1 9 A ). One of the reasons for the change can be effects from the significant adhesive failure of door 1 at 30 psf (Table 4 5); however, this needs to be validated through more testing. (2) For each door, the displacement load relationship is a lmost linear at all the locations where displacements were successfully measured (note: some of the traceable spheres were not tracked successfully during testing).

PAGE 79

79 (3) Even though responses (strains, displacements, and maximum loads) under wind pressure are different from door 1 to 5, for each door, the measured responses at different locations are similar, particularly responses at the interior locations (i.e., the locations between the top and bottom of each door). (4) Comparing Figs. 4 1 9 A and 4 20 A, it can be noticed that the adhesive failure of door 1 affects the strains, but has negligible influence in displacement responses. 4.3.2 FEA Results of Door 2 and Comparison to Full Scale Experimental Results 4.3.2.1 Failure m echanisms c omparison In the experimenta l testing on door 2 the specimen failed at a suction pressure of around 120 psf, while the FEA was stopped at 115 psf. One of t he reason s for the difference could be: for ADINA, once buckling occur red the program treat ed it as failure and d id not allow the analysis to be continued H owever, in the experimental testing, even though the buckling ha d occurred, the load can still be increased but the U bars did not contribute to stiffness any more. Thus, we hypothesize that is why the door deformed so much that it disengaged from the tracks. Fig. 4 2 1 compare s the catastrophic failure of door 2 in the experimental testing and FE model. It can be concluded that: (1) T he local bucking of U bars can be recreated in the simplified FE model. (2) The failure of rollers disengaging from the tracks was not recreated in the simplified FE model. The reason is because rollers were not physically modeled in ADINA a s mentioned in Section 4.2.1.1. Since the FE model was able to recreate the first catastrophic failure, i.e., local buckling failure, it is satisfactory. Furthermore, the FE model revealed the failure sequence of door 2, i.e., local buckling of U bars led to the disengagement failure. And it indicates that material strength limits the performance of door 2 under w ind pressure loading.

PAGE 80

80 4.3.2.2 Displacement and s train results c omparison The comparisons between FEA and experimental results in strain and displacement responses are shown in Figs. 4 2 2 and 4 2 3 The x coordinates in Figs. 4 2 2 and 4 2 3 were plotted in positive only to make them easier to read and the load is still negative pressure The load range in Figs. 4 2 2 and 4 2 3 is from 0 to 110 psf which is the loading step before the catastrophic failure of door 2 occurred in the experimental testing Table s 4 6 and 4 7 give the difference percentage between the two methods. From the strain and displacement comparisons, the following can be seen: (1) Similar to the experimental measure results, FEA results show linear strain pressure and displacement pressure re lationships of door 2 before catastrophic failure (Figs. 4 2 2 and 4 2 3 ). (2) FE A results match well with experimental measurements in both strain and displacement responses of door 2 (Figs. 4 2 2 and 4 2 3 ). (3) Compared to lower loads, the difference between FEA r esults and experimental measurements at higher loads is larger (Figs. 4 2 2 and 4 2 3 ). The reasons could be: 1) the repeated step and hold loading pattern might have caused fatigue or yielding in some material (e.g., connecting fasteners, roller stems, etc. ) and therefore decreased the structural stiffness; 2) it might be over constrained to fix x and z translational DOFs on the side of the FE model (Fig. 4 6). (4) The average difference percentage is no more than 16% (Tables 4 6 and 4 7), which is satisfactory (5) The good agreement between FEA and experimental testing indicates that this FE model of door 2 is validated, which builds confidence in using the model for future analysis. 4.3.2.3 Effects of c onstruction e rrors During preparation for the experimental te sting, it was observed that there was a high occurrence of construction errors in field installations e.g., missing or stripped fasteners not following the installation manual The validated FE model of door 2 was,

PAGE 81

81 therefore, utilized to investigate the effects of the following construction errors to structural performance. Construction error 1: Missing fasteners connecting U bars to stiles. Construction error 2: Missing fasteners connecting hinges to stiles. Construction error 3: Missing 1 in. overlap at the end of door 2, which is required in the NOA drawing ( Fig. 4 2 4 ). For the first construction error, five fasteners connecting U bars to stiles were randomly chosen using MATLAB, and thirty trials were determined for analysis to obtain statistical data. In FE model, the rigid links representing the chosen five fasteners were released to imitate the error. For the second construction error, three hinges affected by missing fasteners were randomly chosen using MATLAB. Similarly, thirty trials were analyzed for statistical data. In FE model, this error was simulated by releasing constraints for the chosen hinges. For the third construction error simulated in FE model, the width of the model was increased by 1 in. Figs. 4 2 5 to 4 2 7 show the influence of the three construction errors on structural performance of door 2. The influence is similar at each location where responses were measured, and therefore, only one comparison was shown here for strain and displacement. For construction error 1 and 2, average from the thirty trials was used for comparison. It can be observed that: (1) Missing fasteners connecting U bars to stiles caused the buckling of U bars occurred at a lower pressure (around 110 psf) compared to the error free model (around 115 psf) (Fig. 4 2 5 A). Before U bar buckling, this construction error did not change strain and displacement responses (Fig. 4 24). At 110 psf pressure, the displacement response did not change (Fig. 4 2 5 B) as strain responses did, and the reason is probably because disp lacement response was less sensitive to this construction error. (2) Missing fasteners connecting hinges to stiles did not affect the strain and displacement responses of door 2 (Fig. 4 2 6 ).

PAGE 82

82 (3) Missing 1 in. overlap at the end of door 2 had a negative influe nce on strain and displacement responses. It caused door 2 failed at a lower pressure of 110 psf (Fig. 4 2 7 ). 4.3.3 FEA Results of Door 3 and Comparison to Full Scale Experimental Results 4.3.3.1 FEA r esults of the l ocal m odel Under the prescribed displacement applied on part 2 of the roller stem, the track was able to constrain the movement of the roller when the displacement was small. When the displacement reached to a threshold the track started to yield (Fig. 4 2 8 A) and th us was not able to constrain the roller anymore, which led to the disengagement of the roller from track completely (Fig. 4 2 8 B). Fig. 4 2 9 gives an example of the contact force responses on the track under a certain displacement level, and it can be noti ced that contact forces varied with the locations on the track (different colors represent different values). Therefore, effective contact forces were used to determine the spring stiffness for the global model of door 3. The average of contact forces at n ode 152, 153 and 154 were used as the effective contact forces because: 1) maximum contact forces occurred on the three nodes under different load level; and 2) contact forces at other nodes were about ten times smaller and influencing areas were also smal ler. Fig. 4 30 gives the curve representing the relationship between the effective contact force and the displacement of the roller and the simplified curve, which wa s used for the spring stiffness in the global model. From Fig. 4 30 the following can be observed: (1) When the roller displacement was about 0.13 in., the track started to yield. (2) When the roller displacement was about 0.3 in., the contact force at the interface started to decrease, which indicates the roller started to slide out of the track. (3) Whe n the roller displacement was about 1.1 in., the roller completely separated from the track.

PAGE 83

83 4.3.3.2 Failure m echanisms c omparison of d oor 3 Fig. 4 3 1 shows the failure of door 3 observed in the experimental testing and FEA. And Fig. 4 3 2 presents the disp lacement of the roller varying with the applied suction. The following can be drawn: (1) In the experimental testing, disengagement of the roller occurred when the pressure was 55 psf. The disengagement failure was led by the overbending of roller stems ( Fig. 4 3 2 A ). The FE model of door 3 was able to recreate the overbending of roller stems ( Fig. 4 3 1 B ). (2) In the FE model, the overbending first occurred at the 9th roller stem when the load was 53.5 psf and then progressed towards the following roller stems in the order of 8th, 7th, 6th, 5th, 4th, 3rd, and 2nd at 54 psf pressure. (3) From Fig. 4 31, it can be noticed that at around 54 psf suction, roller s 2 to 9 had a displacement of about 0.3 in. According to the local FE model results, it was concluded that wh en the roller displacement was about 0. 3 in, it started to slide out of the track. Therefore, even though the FE model of door 3 did not visually recreate the disengagement failure, results of the roller displacements and orientations of the rollers (Fig. 4 3 1 B ) implied that the disengagement of roller s 2 to 9 w ould occur immediately after the overbending. (4) The agreement between FEA and experimental testing indicates that: (a) The results from the local model are reasonable. (b) Nonlinear spring can be an efficient way to simulate the constraints of the track. (c) The approach of combining local and global models is reasonable and efficient to recreate the failure of door 3. The complementary FEA results revealed the failure sequence of door 3, i.e., the overbending of roller stems led to the disengagement failure. Furthermore, it indicates that material strength is the controlling factor of the performance of door 3 under wind pressure. 4.3.3.3 Displacement and s train r esults c omparison The comparisons between FEA and experimental results in strain and displacement responses are shown in Figs. 4 3 3 and 4 3 4 Similar to Figs. 4 2 2 and 4

PAGE 84

84 2 3 the x coordinate was plotted in positive to be read easier. The load range in Figs. 4 3 3 and 4 3 4 is from 0 to 5 0 psf which is the load before the catastrophic failure of door 3 occurred in the experimental testing Table s 4 8 and 4 9 provide the average difference percentage between these two methods. From the strain and displacement comparisons, the following can be seen: (1) FEA results show linear strain pressure and displacement pressure relationships before catastrophic failure (Figs. 4 3 3 and 4 3 4 ). (2) FEA results match well with experimental measurements in both strain and displacement responses of door 3 (Figs. 4 3 3 and 4 3 4 ). (3) The difference between FEA results and experimental measurements at the top of door 3, i.e., strain results corresponding to strain gauge 8, and deflections at H= 172 in., are a little bit larger compared to the other locations. One of the reasons could be : it is over constraining to fix x translational degree of freedom at the top of door 3 model, because the top part of the track is curved, and therefore z displacement can result in x displacement ( Fig. 4 3 5 ). However, average difference less than 10% ( Ta bles 4 8 and 4 9 ) demonstrated the FEA results are satisfactory. (4) The good agreement demonstrates that combining global and local models is a justified and effective approach to model door 3. 4.4 Summary Failure of garage doors in hurricanes can lead to cas cading failures of other building envelope components. Therefore, understanding of their structural behavior under hurricane wind pressure is important. In this study, quasi static pressure tests were conducted on five commercial sectional doors using the newly developed s imulator which was presented in Chapter 3. Experimental results suggest ed two main catastrophic failure mechanisms of current sectional doors, and they are U bars buckling and disengagement of rollers from the track. Linear relationships b etween structural responses, including strains and displacements, and pressure were found for all the five test specimens. Following experimental testing, finite element models of door

PAGE 85

85 2 and door 3 were built in ADINA. FEA results matched the experimental testing well The complementary FEA revealed the failure sequence of door 2 and 3, and indicated that material strength was the controlling factor for them to be used in high wind areas. Furthermore, the validated FE models can be used in the future for ot her scenarios. Three construction errors were investigated in FEA for their influence on structural performance of door 2. It was observed that missing three hinges did not affect structural behaviors of door 2, while it failed at a smaller pressure loadin g when missing some fasteners connecting U bars to stile s or missing 1 in. overlap at the end. As the first one on structural performance of sectional doors subjected to wind pressure, this study contributes to our fundamental understanding in this area. Furthermore, this study revealed several shortcomings of current sectional doors, namely, too many components and insufficient stiffness. Together, these provide d a direction for a new design philosophy of sectional doors to resist strong winds.

PAGE 86

86 Table 4 1 Description of test specimens Door Model Dimensions: Width x Height Design Pressure ( psf ) a Test pressure ( psf ) (1.5 x Design pressure) Door 1 18 ft. 2 in. x 12 ft. + 42 / 46 + 63 / 69 Door 2 18 ft. 2 in. x 14 ft. + 43.5 / 50 + 65 / 75 Door 3 18 ft. 2 in. x 15 ft. 9 in. + 55 / 64 + 83 / 96 Door 4 16 ft. x 15 ft. 9 in. + 50 / 5 7 + 75 / 86 Door 5 16 ft. x 13 ft. 9 in. + 46 / 52 + 69 / 78 a Test pressure is 1.5 times of design pressure (FBC 2010)

PAGE 87

87 Figure 4 1. One sacrificial door installed to the Simu lator at the opening of the air box (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 88

88 Figure 4 2 Failures observed from sacrificial tests (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida) A) Failure modes in positive pressure testing. B) Failure modes in negative pressure testing

PAGE 89

89 F igure 4 3. Instruments used in full scale experim ental testing (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida) A) A photograph of door 2 showing the locations of traceable spheres and load cells. B) Load cell mounted between wood jamb and reaction f rame. C) An example showing the locations of traceable spheres and strain gauges with respect to U bars. D) An illustration of displacement measurement technique.

PAGE 90

90 Figure 4 4. A conceptu al sketch of the loading during full scale experimental testing

PAGE 91

91 Figure 4 5. Half symmetry model of door 2 in ADINA.

PAGE 92

92 Table 4 2 Main components of the FE model of door 2 Item Material Entity Finite element material model Door panel Steel 4 node shell element Elastic perfect plastic material Stile Steel 4 node shell element Elastic perfect plastic material U bar Steel 4 node shell element Elastic perfect plastic material Blocking Wood 8 node solid element Linear elastic material Astragal retainer Aluminum 4 node shell element Elastic perfect plastic material

PAGE 93

93 Figure 4 6. Boundary conditions and constraints in FE model of door 2. A) Boundary conditions. B) Constraints between different components.

PAGE 94

94 Figure 4 7. A FE model of a roller using three dimensional solid elements.

PAGE 95

95 Figure 4 8. The local model of a roller and 21 in. long track.

PAGE 96

96 Table 4 3. Components of the local model of roller and track Item Material Entity Finite element material model Roller stem Steel 8 node solid element Elastic perfect plastic material Roller Steel 8 node solid element Elastic perfect plastic material Track Steel 4 node shell element Elastic perfect plastic material

PAGE 97

97 Figure 4 9. The global FE model of door 3. (The model is half symmetry model)

PAGE 98

98 Table 4 4 Main components of the global model of door 3 Item Material Entity Finite element material model Door panel Steel 4 node shell element Elastic perfect plastic material Stile Steel 4 node shell element Elastic perfect plastic material U bar Steel 4 node shell element Elastic perfect plastic material Roller stem Steel 2 node beam element Elastic perfect plastic material Astragal retainer Aluminum 4 node shell element Elastic perfect plastic material

PAGE 99

99

PAGE 100

100 Figure 4 10. Boundary conditions and constraints in the local FE model of roller and track. A) Prescribed displacement applied on part 2 of the roller stem. B) X y and z translational DOFs constrained at face 2 of the track and contact applied at the interface of the roller and the track. C) Rigid link connection the roller stem and roller.

PAGE 101

101

PAGE 102

102 Figure 4 11. Boundary conditions and constraints in the global FE model of door 3. A) Boundary conditions applied on the door model. B) Constraints between components. C) Boundary condition applied on the side of the FE model.

PAGE 103

103 Figure 4 12. Interaction between the local model and global model of door 3.

PAGE 104

104 Figure 4 13. Failure mechanisms of door 1 (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 105

105 Figure 4 14. Failure mechanisms of doo r 2 (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 106

106 Figure 4 15. Failure mechanisms of door 3 (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 107

107 Figure 4 16. Failure mechanisms of door 4 (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 108

108 Figure 4 17. Failure mechanisms of door 5 (Photo courtesy of the Powell Family Structures and Materi als Laboratory at the University of Florida)

PAGE 109

109 Table 4 5. Summary of the pressure at failure of test specimens Door Model Test pressure ( psf ) Adhesive failure pressure ( psf ) Catastrophic failure mechanisms c Catastrophic failure pressure P f ( psf ) Door 1 + 63 / 69 ~ 30 a Failure 1 and b 2 ~ 85 Door 2 + 65 / 75 Failure 1 and 2 ~ 120 Door 3 + 83 / 96 ~ 50 Failure 2 ~ 55 Door 4 + 75 / 86 Failure 1 and 2 ~ 130 Door 5 + 69 / 78 ~ 80 Failure 1 and 2 ~ 95 a Failure 1 represents the local buckling of U bars b Failure 2 represents the disengagement of the door panel from the tracks. c The catastrophic failure pressure is the load when the experimental testing was stopped, because w hen both failure mechanisms oc curred, the time interval was short, and it was difficult to capture the failure sequence.

PAGE 110

110 Figure 4 18. Characteristics of the U bars and rollers of the five test specimens. A) Door 1. B) Door 2. C) Door 3. D) Door 4. E) Door 5. (Note: the dimensions of the roller of door 4 w ere not provided in the NOA drawing, and therefore, they are not shown in Fig. 4 18.)

PAGE 111

111 Figure 4 1 9 Measured strain suction relationships at multiple locations during experimental testing. A) Door 1. B) Door 2. C) Do or 3. D) Door 4. E) Door 5.

PAGE 112

112 Figure 4 20 Measured deflection suction relationships at multiple locations during experimental testing. A) Door 1. B) Door 2. C) Door 3. D) Door 4. E) Door 5.

PAGE 113

113 Figure 4 2 1 Failure mechanism comparison between experimental testing and FEA results of door 2.

PAGE 114

114 Figure 4 2 2 Strain comparisons between FEA results and experimental measurements of door 2.

PAGE 115

115 Figure 4 2 3 Deflection comparison between FEA results and experimental measurements of door 2.

PAGE 116

116 Table 4 6 Difference percentage of strain responses between full scale testing and FE model of door 2 ( a b ) / Strain gauge 1 3.4 % Strain gauge 2 0.8 % Strain gauge 3 0.6 % Strain gauge 4 0.1 % Strain gauge 5 0.8 % Strain gauge 6 0.4 % Strain gauge 7 2.6 % Strain gauge 8 0.8 % Average 1.2% a Strain results from FE model b Strain measurements from full scale testing

PAGE 117

117 Table 4 7 Difference percentage of displacement responses between full scale testing and FE model of door 2 ( a b ) / H = 78 in. 13.8 % H = 98 in. 18.5 % H = 120 in. 15.5 % H = 141 in. 16.3 % Average 16% a Displacement results from FE model b Displacement measurements from full scale testing

PAGE 118

118 Figure 4 2 4 A sketch showing 1 in. overlap at the end of door 2 required by the manufacturer.

PAGE 119

119 Figure 4 2 5 The effects of missing fasteners connecting U bars to stiles on structural responses of door 2. A) Strain responses comparison. B) Displacement responses comparison.

PAGE 120

120 Figure 4 2 6 The effects of missing fasteners connecting hinges to stiles on structural responses of door 2. A) Strain responses comparison. B) Displacement responses comparison.

PAGE 121

121 Figure 4 2 7 The effects of missing 1 in. overlap at the end of door 2 on structural responses. A) Strain responses comparison. B) Displacement responses comparison.

PAGE 122

122 Figure 4 2 8 FEA results of the interaction between the roller and the track under prescribed displacement. A) Yielding of the track. B) Separation of t he roller from the track.

PAGE 123

123 Figure 4 2 9 Different contact forces at different locations on the track.

PAGE 124

124 Figure 4 30 The relationship between contact force and roller displacement.

PAGE 125

125 Figure 4 3 1 Failure mechanisms of door 3 in experimental testing and FEA. A) Overbending of roller stems in experimental testing (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida). B) Overbending of roller stems in FEA.

PAGE 126

126 Figure 4 3 2 Roller displacement pre ssure relationship.

PAGE 127

127 Figure 4 3 3 Strain comparisons between FEA results and experimental measurements of door 3.

PAGE 128

128 Figure 4 3 4 Displacement comparisons between FEA results and experimental measurements of door 3.

PAGE 129

129 Table 4 8 Difference percentage of strain responses between full scale testing and FE model of door 3 ( ) / Strain gauge 1 4.3 % Strain gauge 2 6 .2 % Strain gauge 3 5 8 % Strain gauge 4 1.7 % Strain gauge 5 6.0 % Strain gauge 6 6.6 % Strain gauge 7 1 1 .1% Strain gauge 8 7.5 % Average 6.2%

PAGE 130

130 Table 4 9 Difference percentage of displacement responses between full scale testing and FE model of door 3 ( ) / H = 9 in. 7.1 % H = 42 in. 1.8 % H = 68 in. 1.6 % H = 89 in. 2.5 % H = 110 in. 4.5 % H = 131 in. 9.5 % H = 152 in. 15.3 % H = 172 in. 21.3 % Average 8.0%

PAGE 131

131 Figure 4 3 5 A photograph showing that the top part of the track is curved (Photo courtesy of the Powell Family Structures and Materials Laboratory at the University of Florida)

PAGE 132

132 CHAPTER 5 M ECHANICAL RESISTANT PROPERTIES OF FRP/FOAM COMPOSITE STRUCTURE From the structural performance of current garage doors discussed in Chapte r 4, it was observed that insufficient material strength was the primary reason accounting for the catastrophic failure of garage doors under wind pressure loading. Additionally, during preparing for the experimental testing of the five sectional garage doors, it was noticed that many components of a current garage door (e.g., stiles, U bars, and a large number of small fasteners) resulted in : a complicated installat ion high occurrence of construction errors during shipping and installation Considering the disadvantages of current design philosophy of sectional doors, this research (Shen et al. 2013) develops a new design philosophy of sectional doors using a FRP/fo am sandwich panel to replace commonly used metal material for sectional doors, because sandwich panels usually have a high strength to weight ratio. Additionally, by using this new sandwich panel a simple design of sectional doors without stiles and U bar s is desired. The application of FRP/foam sandwich panels is growing rapidly in the area of civil engineering. The sandwich panel of interest in this research is comprised of: 1) two thin skins made of Glasbord REI FRP; and 2) a polyol isocyanate foam cor e. The composite structure is currently used in architectural components (e.g., entry doors, partitions and etc.), and this research aims to investigate its applicability for larger component and cladding systems intended for use in high wind areas, sectio nal doors in particular. For the sandwich panel studied in this research, the material properties of the FRP were provided by the manufacturer, but those of the foam were unknown.

PAGE 133

133 Therefore, the first half of this study addresses the experimental investiga tion on the constitutive properties of the polyol isocyanate foam. Direct uniaxial compressive, direct uniaxial tensile and four point bending rests were performed to characterize the in plane and out of plane responses of the sandwich panel The strength of the foam core material usually limits the overall behavior of a composite structure, and therefore the polyol isocyanate foam density and thickness were varied in the range of possible manufacturing specifications to investigate their effects on the foa m properties. The remainder of this study focuses on the experimental and numerical investigation on the performance of a full size panel made of the sandwich composite subjected to out of plane loads simulating service conditions. During numerical investi gation, the material properties of the foam obtained from the first part of the study were employed in the finite element (FE) model of the full size panel. This part of the study was conducted to: Investigate the applicability of the sandwich panel used f or large building component and cladding systems in hurricane prone areas. Evaluate the material properties of the foam obtained from the first part of the study through comparing the FE model results to experimental measurement. The findings validate the material properties of the foam obtained from experimental testing, suggest that the sandwich panel is suitable for lightweight building cladding systems, and indicate that deflection is the controlling design factor for out of plane loading. At the end, different strategies including increasing panel thickness and adding stiffeners were tried using FE method and proposed to increase the stiffness of the panel. Adding stiffeners were also tested experimentally. A paper (Shen et al. 2013) describing the study included in Chapter 5 was published in May 2013 in the Journal of Composite Structures

PAGE 134

134 5.1 Experimental Investigation on the Foam Material Properties 5.1.1 Test Specimens The sandwich panel studied in this research is comprised of 1) two thin skins made of Glasbord REI FRP; and 2) a polyol isocyanate foam core. Unlike most other sandwich panel s using adhesive to bond FRP skins and foam core, the bonding between FRP skin and foam core of this sandwich panel is produced by the chemical interaction and heat evolution during the recovery of the foam after being injected into the chamber with w ood around and FRP skins at top and bottom. The material properties of the FRP are listed in Table 5 1 provided by the Manufacturer (CRANE Composites 2011) To investigate the influence of foam density and thickness to the material properties of the foam, nine combinations of foam density and thickness were evaluated: = 5.0, 5.5, and 6.0 pcf; = 1.0, 2.0, and 3.0 in. The bounds were set based on manufacturing capabilities. Table 5 2 summarizes the geometry of the test specimens for the compressive, tensile and flexure testing Five replicate s of each of the nine specimen types were tested. 5.1.2 Testing Apparatus and Instrumentation Static uniaxial compression, direct uniaxial tensile and four point bending tests were performed using an Instron 3384 universal testing machine (UTM) located in the Weil Hall Structures Laboratory in the Department of Civil and Coastal Engineering at the University of Florida (Fig 5 1 ). The UTM re corded load and crosshead displacement at 10 Hz on a standalone computer system (Fig 5 1 ). During four point bending tests, f lexure induced strain s in the FRP sheet w ere measured at a 1.7 Hz sampling rate using pre wired Omega KFG 3 120 C1 11L1M2R strain gauges. Strain data were recorded using LabVIEW software and later synchronized with the UTM load and

PAGE 135

135 displacement measurements. Compressive and tensile strains in the foam were computed by dividing the crosshead displacement by the thickness of the foam. Attempts to measure strain directly on the foam with gauges were unsuccessful due to the local stiffening effects from the epoxy for attaching strain gauges to a low modulus material (Ajovalasit et al. 2007) And a ttempts to correct for the local reinforcement effect did not produce usable results. The local stiffening effects of the epoxy used to attach the strain gauge to the foam caused strain measurements to vary significantly from one test to another and therefore engineering strain was used as a n approximation This approach was deemed reasonable because deflections were on the order of centimeters, and the tests produced highly repeatable results in the elastic region. 5.1.3 Experimental Testing 5.1.3.1 Direct u niaxial c ompressive t est s Cylindrical specimens were cut from large panels using a Cincinnati Bickford radial drill at a con stant low speed to avoid notching and roughening of the surfaces. Specimens were oriented upright in the UTM with the face sheets parallel to the compression platens. The geometry of the test specimens was determined following ASTM C365 (ASTM 2005) which stipulates that: 1) the facing area shall be not smaller than 1 in 2 and not larger than 1 6 in 2 and 2) the loading rates shall produce failure within 3 to 6 min. The foam specimens with = 1.0, 2.0, and 3.0 in. were loaded under displacement control at rates of 1.0 2 .0, and 3 0 in /min, respectively, to meet the requirements. Fig 5 2A give s examples of 1.0 in. thick specimens, and Fig. 5 2 B shows 5.0 pcf specimens during and after loading.

PAGE 136

136 5.1.3.2 Direct u niaxial t ensile t est s Tensile specimens were carefully cut to avoid roug h or uneven surfaces and de laminations. Loading plates were bonded to the FRP face sheets using a high modulus, high strength, structural epoxy paste adhesive (Sikadur 30) to transmit uniform tension to the specimen. The center line s of the specimen and th e loading plates were aligned to avoid eccentric loading Loads were increased at a speed of 0. 02 in. /min until failure. Fig 3 3 shows an example of the direct uniaxial tensile test configuration. The conditions for failure were based on ASTM C297 (ASTM 2004) which only acceptable failure modes for this test method are those internal to the sandwich the core failure and the core to FRP skin bond failure). 5.1.3.3 Four p oint b ending t ests Fig 5 4A shows the configuration for the four point bending (FPB) testing. The distance between the supporting and loading cylinders was 12 in and 4 in respectively. FPB tests were conducted according to ASTM C393 (ASTM 2006) on beam like specimens, which requires that the wid th of the specimen should be set to at least twice and at most six times the total thickness. Spec imen dimensions are shown in Fig. 5 4B The 1.0 in. and 2.0 in. thick specimens were 4 in wide The 3.0 thick specimen s were 6 in wid e The length was set t o 14 in which includes 1 in margins out of the supporting cylinders for each group. The crosshead displacement rate was 0.25 in/min. Strain gauges were attached to the top and bottom FRP skins to measure the compressive and tensile strains.

PAGE 137

137 5.1.4 Experi mental Results 5.1.4.1 Compression t esting r esults Similar to other foams described in Chapter 2, the polyol isocyanate foam exhibited ductile behavior under compression. And the compressive stress strain curves also followed the pattern: 1) an initial li near elastic stage, followed by 2) a plastic plateau region, and 3) a stage of densification at the end of loading (Fig. 5 5). Table 5 3 reports the results of elastic modulus E C yielding stress yc and yielding strain yc in compression, which are averages of five samples for each case according to calculation recommended in ASTM C365 (ASTM 2005) Figs. 5 5A to 5 5C and 5 5D to 5 5F show the effects of foam density and thickness to compressive stress strain responses. The influence of f oam d ensity A n increase of foam density does not change the yielding strain yc noticeably (Figs. 5 5A to 5 5C), but produces a rise in the yielding strength yc and therefore increases the elastic modulus E C in compression. From the results listed in Table 5 3, it can be seen that in all but one case, both E C and yc increase with foam density. The influence of f oam t hickness (1) Using a thicker foam does not necessarily produces a higher yielding stress. In effect, 2.0 in. thick foam gives rise to the lowest yielding stress for the three density conditions, while 1.0 in. and 3.0 in. thick foams show similar yielding stresses for 5.0 and 5.5 pcf density conditions (Figs. 5 5D to 5 5F). This phenomenon may be attributed to the variations in the air void content of the foam, but due to the uncertainty in the number and size of the air voids, their effects cannot be accurately evaluated herein. (2) 3.0 i n. thick foam produces the highest yielding strength for the 6.0 pcf density condition ( Figs 5 5D to 5 5F). (3) For the three density conditions, 3.0 in. thick foam shows the largest elastic modulus represented by the slope of the elastic region (Figs. 5 5D t o 5 5F).

PAGE 138

138 5.1.4.2 Tension t esting r esults In contrast to the ductile response observed in compression testing, the composite showed a brittle failure either in the foam or at the core skin interface during the tension testing. The results from the tensile testing were not as repeatable as those from compressive testing For example, the coefficient of variation (CoV) for yielding strain in compression and ultimate strain in tension are 0.06 and 0.17 respectively, and this can be attributed to greater sensitivity of the composite to the air void content under tensile loading. Figs. 5 6A, 5 6B, and 5 6C show failure examples of tensile specimens with 1.0, 2.0 and 3.0 in. thick foam respectively. The effects of foam thick ness and density to the tensile stress strain curves are plotted in Fig. 5 7, and Table 5 4 summarizes the results of the tensile testing, which are also calculated from the mean values of the five replicates for each case. The influence of f oam t hickness (1) Failure modes of the composite structure in tension depend on the foam thickness: specimens with 1 in. thick foam failed at the core FRP skin interface regardless of density (Fig. 5 6A), while most of the specimens with 2 and 3 in. thick foams exhibited a core fracture, shown in Figs. 5 6B and 5 6C. Fig. 5 6D shows that the core fracture occurred at a cross section where voids are present. Void contents can be the primary reason for these phenomena: 1 in. thick foam is expected to have fewer voids (Nambiar and Ramamurthy 2007) inside compared to 2. 0 and 3.0 in. thick cases, and therefore the debonding between the foam and FRP skins occurred before the core fracture; however, for 2.0 and 3.0 in. thick foam, high void contents may become the weak points and cause core fracture earlier than the bonding failure at the interface. (2) Different foam thicknesses contribute to the evaluation of two physical properties: (a) Bonding strength from 1.0 in. thick foam case. (b) Tensile properties of foam from 2.0 and 3.0 in. thick foam cases. (3) Comparing 2.0 and 3.0 in. thick foams (Figs. 5 7A to 5 7C), the 3.0 in. thick foam is shown to have a higher elastic modulus for the three density conditions, and in all but one case, it also produces a higher ultimate tensile strength.

PAGE 139

139 (4) From Table 5 4, it can be observed that the bondin g strength ( ub ) at interface is higher than the ultimate tensile strength ( u t ) of the foam for the three density conditions. The influence of f oam d ensity (1) Fig. 5 7D presents the tensile stress strain curves for 1 in. thick foam case, and therefore gives the bonding strength at the interface as discussed in the foam thickness influence (2). The comparison indicates a negligible change in bonding strength with the foam density. (2) The ultimate tensile strength and elastic modulus of the foam increase with the foam density (Figs. 5 7E and 5 7F). 5.1.4.3 Four p oint b ending t esting r esults Fig 5 8 shows the observed failur e mechanisms during FPB tests: 1) unsymmetrica l shea r failure in the foam, 2) symmetri cal shear failure in the foam, 3) de bonding at t he core to FRP skin interface, 4) local co llapse at the top FRP surface, 5) loc al buckling of the top FRP and 6) local de bonding between the top FRP skin and the foam. Unsy mmetrical shear failure in the foam (Fig 5 8A ) occurred more than 90% of the time, and it typically led to a further failure of de bonding between the FRP skin and the foam. Therefore, shear failure was assumed to be the dominant failure mechanism for FPB tests. Fig 5 9 presents load deflection curves of the test specimens. All data are shown except for three anomalous variations observed for two samples in the = 1.0 in. and 5.0 pcf case and one samp le for the 2.0 in. and 5.0 pcf case. From these plots, the following observations were made (1) The peak load capacities were similar for the five replicates for almost every combination of density and thickness; however the corresponding deflections at failure vary significantly (2) The ultimate shear strength is approximately equal to the yielding strength, especially for the smaller foam thickness cases (3) The linear region of the load deflection relationship curve is nearly identical for each replicate in a test series

PAGE 140

140 (4) The variability between the five replicates after the load drop can be attributed to: (a) Randomness in the air voids size distribution and location in each sample (b) The deflection recorded is that under the loading elements, and most specimens failed at unsymmetrical shear cracking, and after failure, the two loading elements were not on the same elevation any more Knowing the load at shear failure (i.e., ) the core shear ultimate stres s can be calculated from Equation 5 1 ( 5 1 ) W here = the maximum force prior to failure lb f = the thickness of the composite panel, = the core thickness, and = the composite width. Table 5 5 report s the average values for and for the nine combinations of thickness and density. And the influence of foam density and thickness to the shear behaviors of the foam can be summarized as: (1) The ultimate shear strength increases slightly with the foam density (2) A higher foam thickness does not produce higher shear strength. The thinnest foam ( 1.0 in. ) gives the highest shear strength for the three densities, and 2.0 in. thick foam gives the lowest shear strength 5.2 Experimental and Numerical Investigation on A Full Length Composite Panel During the FPB testing in the Section 5.1.3, core shear failure was the dominant failure mode. Flexure i nduced failure of the FRP did not occur, most likely because the length of the support span was insufficient. ASTM D 6272 02 (ASTM 2008) requires the support span to be at least 16 times the depth of the test specimen to achieve flexural failure, which was not possible with the UTM setup for FPB testing shown in Fig. 5 12A. Therefore, uniform out of plane pressure testing was conducted on a full length composite panel experimentally and numerically to:

PAGE 141

141 (1) Investigate the bending behavior of a panel of sufficient size to be used in a building system (2) Verify the material properties obtained from the material testing through comparing numerical r esults and experimental measurements 5.2.1 Experimental Testing 5.2.1.1 Test s pecimen A full length composite panel was fabricated by the manufacturer to the following dimensions : 95.75 in. long x 23.75 in. wide x 1.74 in thick The specimen had a 1.5 in thick polyol isocyanate foam core with density of 6.0 pcf and two 0.12 in thick Glasbord REI FRP skins. 5.2.1.2 Testing a pparatus and i nstrumentation Uniform out of plane pressure testing was conducted using the HAPLA discussed in Chapter 2 (Fig 5 11A ) in the Powell Family Structures and Material Laboratory at the University of Florida. The specimen was mounted to a steel frame of the test chamber of the HAPLA (Fig 5 11B ). To simulate one way bending condition s, the top and bottom ends of the pane l were constrained from translating by 2 x 2 x 3/16 in steel angles connected to the steel frame with 1/2 20 x 6 bolts, and the two sides of the panel were free to move (Fig 5 11C ). The horizontal straps shown in Fig 5 11C did not make contact with the specimen, and were only in place to restrain the panel if the connections failed. A Balluff BOD 63M LA02 S115 distance sensor (Fig 5 11D ) measured deflections at the center of the panel, and an Omegadyne PX409 001CG5V pressure sensor measured the pressure inside the test chamber. TML PF 30 11 strain gauges were attached to the FRP at the 1/4, 1/2, and 3/4 points along the centerline of the compressive and tensile FRP skins in the vertical direction to measure the longitudinal strain (Fig s. 5 11B and 5 11C )

PAGE 142

142 5.2.1.3 Testing l oad The HAPLA applies user defined spatially uniform load conditions to the specimen. The pressure was increased stepwise up to 145 psf as shown in Fig. 5 12. LabVIEW software was used to record the data measurements during the testing. 5.2.2 Finite Element Method 5.2.2.1 Finite e lement m odel c omponents Numerical simulation of the out of plane pressure testing on this full length composite panel was performed using the experimental measurements of the flexural response of the panel as a validation measure. The finite element (FE) model (Fig 5 13A ) was created in ADINA 8.7 (ADINA 2010) The FRP skins were modeled using four node shell elements with inco mpatible modes that function to increase the flexibility of the element in bending situations (ADINA 2010) and the foam core was modeled using eight node brick element s. Perfect bonding was assumed at the skin core interface, and therefore this region was modeled with rigid links (Fig 5 13A ). Table 5 6 summarizes the main components of the FE model. 5.2.2.2 Boundary c ondition s and c onstraints An overview of the constr aint system is shown in Fig 5 13B The steel frame was modeled to be infinitely stiff i.e., all degrees of freedom were fixed ( Fig 5 13B ). The steel angles were included in the model to capture partial moment constraint. The steel bolts ( Fig 5 13C ) wer e treated as rigid links between the contacting surfaces of the steel frame and the steel angles ( Fig 5 13C ). The rigid links constrain ed the linked nodes so that they move d together when the model deform ed Surface to surface contact was imposed at the f rame panel and steel angle panel interfaces ( Fig 5 13D ).

PAGE 143

143 5.2.2.3 Material p roperties in FE m odel The FRP was modeled as isotropic linear elastic material with (Table 5 1 and Fig ure 5 14A ); and the foam was modeled as bilinear elastic material employing compressive modulus and tensile modulus measured from the material testing (Fig 5 14B ). Additionally, t he material properties of 1.5 in thick foam were approximated by interpolation between 1.0 and 2.0 in. thick foam properties. 5.2.2.4 P ressure l oad i n FE m odel To reproduce the load sequence applied in the experimental testing, the measured pressure history ( Fig 5 12 ) durin g the experimental testing was imported into ADINA to be applied to the surface of the FE model (Fig 5 13A ). 5.2.3 Results and Discussion Figs. 5 15A and 5 15B show the deformed panel in the experimental testing and FE model. Fig. 5 16 gives comparisons between FE results and experimental measurements on mid span deflection, mid span longitudinal strain, and quarter point strain respectively. And Table 5 7 provides the average difference percentage between these two methods. The following conclusions can be drawn (1) A linear displacement pressure relationship can be observed at the mid span of the panel under uniform pressure from the experimental testing and FE model (Fig. 5 16A). (2) The strains at the three locations (mid span, 1/4 and 3/4 point) varied linea rly with the applied pressure (Figs. 5 16B and 5 16C). (3) The compressive strain at the front FRP skin matched the tensile strain at the back FRP skin, which suggests that a linear strain profile along the thickness of the composite panel is a reasonable assu mption (Figs. 5 16B and 5 16C). (4) FE results showed an excellent agreement with the experimental measurements on displacements and strains (Fig. 5 16), differing by less than 15 % (Table 5 7 )

PAGE 144

144 (5) The good match between FE model and experimental testing indicates the material properties obtained from the material testing are satisfactory. Both the experimental and FE model results demonstrate that the wind resistance of this panel is expected to exceed strength limit state requirements for low rise buildings in hu rricane prone areas. To put it in perspective, U.S. minimum design wind load requirements according to ASCE 7 ( 2010) are nominally 144 psf for components and cladding on a low rise (six story) building that is regular shaped, non essential, enclosed and located in flat, open terrain anywhere in the United States. H owever, the results suggest that serviceability limit state requirement is like ly to be the controlling factor for the FRP/polyol isocyanate foam panel s Most building codes require that the deflections shall not exceed L/ 12 0 for exterior walls and interior partitions with flexible finishes (Table 5 8 ) (IBC 2009) (Note: The requirements are us ed as a reference even though they may not apply to doors). Although the Finite element analysis ( FEA ) derived maximum pressure for this panel based on the material strength was found to be 315 psf the load causing a L/120 deflection was 60 psf or approxi mately 1/ 5 of the ultimate structural load. To further investigate this issue FEA was employed to investigate possible methods to increase the stiffness of the sandwich panel, which is discussed in the following. 5.2.4 Methods to Increase Stiffness In ord er to decrease deflections of the sandwich panel subjected to out of plane pressure loading, two approaches were investigated to increase the stiffness of the sandwich panel: 1) changing the dimensions of the sandwich panel e.g., FRP and foam thicknesses (Dawood et al. 2010) ; and 2) inserting through thickness FRP stiffeners (Kim et al. 1999) Because the FE model of the prototype design of the sandwich panel

PAGE 145

145 wa s validated through the experimental testing, it is therefore reasonable to use FE method to conduct this part of the study through modifications on the validated FE model. Samples with stiffeners added were also tested experimentally. 5.2.4.1 Changing d i mensions of the sandwich panel Four combinations of foam and FRP thicknesses with six different span lengths (4, 5, 6, 8, 10 and 12 ft.) were evaluated (Fig 5 17A ): 1.5 and 3.0 in, and 0.12 and 3/16 in. The total twenty four cases were modeled in ADINA through modifying the FE model discussed in Section 5.2.2. Considering material strength and serviceability requirements for deflection (IBC 2009) the allowable pressure for each case was determined and shown in Fi g. 5 17B. F ig 5 17B shows that span length, foam thickness and FRP thickness can easily be adjusted to increase the design load c apacity of the composite panel. 5.2.4.2 Inserting t hrough t hickness FRP s tiffeners Finite e lement m ethod Two configurations of stiffener inserted through sandwich panel thickness were evaluated in ADINA: longitudinal and transverse stiffeners (Fig. 5 18A). And the FRP material used for sandwich panel skins was used for stiffeners. FE models of the two configu rations of stiffeners were added into the prototype FE model discussed in Section 5.2.2, and rigid links were used to attach the stiffener sides to the FRP skins (Fig. 5 18B). The two new FE models were subjected to the same load applied to the prototype F E model (Fig. 5 12).Table 5 9 gives the maximum deflection comparison. It can be observed that adding longitudinal FRP stiffeners is more efficient than transverse stiffeners, and it decreases the maximum deflection by a factor of two.

PAGE 146

146 Experimental t esting Three point bending testing was performed on sandwich panel samples with longitudinal and transverse stiffeners (Fig. 5 19). The material (FRP and foam) of the sandwich panel samples tested here was different from that of the sandwich panel studied in C h apter 5 and therefore the tests were conducted only to compare the effects of the two different stiffener configurations. For each configuration, five replicates were tested until failure (Fig. 5 20). Fig. 5 21 shows the relationships between applied load s and mid span deflections from the experimental testing. And it is easily seen that adding longitudinal stiffeners produces a much stiffer panel than adding transverse stiffeners, whi ch agrees with the FEA results. 5.3 Summary C hapter 5 detailed the exper imental investigation of the constitutive properties of the polyol isocyanate foam employed in the composite structure of interest in this research. From a series of experimental testing, modulus of elasticity and yielding strength under compression, modul us of elasticity, bonding strength at the core FRP skin interface and ultimate tensile strength of the foam under tension, and the ultimate shear strength of the foam were obtained. The experimental results were highly repeatable in the elastic linear regi on which is the focus of this study. T he effects from foam density and thickness to material properties were evaluated. An increase in foam density usually produces better performance of the foam, except that the core FRP skin bonding strength does not cha nge with the foam density. The influence of foam thickness cannot be sufficiently evaluated due to the air voids effects, and the air voids show a significant negative influence on the foam properties. A dditionally, the flexural behavior of the FRP/foam co mposite structure was investigated through a uniform pressure test on a large scale panel of the study composite and a FE model simulating

PAGE 147

147 the experimental testing was developed to verify the material properties obtained from the material testing. The resu lts from FE model matched well with the experimental testing and therefore verified the material properties of the foam experimentally obtained. The full length composite panel exhibited a linear elastic behavior over a wide range of wind loading, but defl ection appears to be the controlling factor for the potential of this composite structure used for building cladding systems in high wind areas. According to FEA results on this issue, it indicates that the composite panel can be used for a wide range of d esign considerations in high wind areas through adjusting foam thickness, FRP thickness and span length. Additionally, FEA and experimental results on inserting through thickness FRP stiffeners shows that adding longitudinal stiffeners is another efficient solution for this issue. O ther solution s include stiffening boundary conditions and/or using FRP with a uniform fiber orientation (multi directional fibers were used in the test specimens). The responses of the composite structure used in large scale buil ding components (e.g., garage doors) will be investigated in the future of this research.

PAGE 148

148 Table 5 1. Physical properties of Glasbord REI FRP Property Glasbord REI 0.12 in Test Method Density 1 00 pcf N/A Flexural Strength 21 10 3 psi ASTM D790 Flexural Modulus 0.7 10 3 psi ASTM D790 Tensile Strength 13 10 3 psi ASTM D638 Tensile Modulus 1.2 10 6 psi ASTM D638 Coefficient of Linear Thermal Expansion 1.3 10 5 in / in F ASTM D696 Barcol Hardness 55 ASTM D2583 Izod Impact 14.0 ft lb / in notched ASTM D256 Gardner Impact Strength 120 in lb ASTM D3029 Water Absorption 0. 20 %/24hrs@ 77 F ASTM D570

PAGE 149

149 Table 5 2. Type of tests and specimen geometry Test Geometry of C ross S ection Nominal T hickness of F oam t = 1 in t = 2 in t = 3 in Compression Circular . 3 .25 in Tension Circular . Flexure Rectangular 14 in 4 in 14 in 4 in 14 in 6 in Note: The dimensions and geometric shapes of the specimens were determined according to requirements from the associated ASTM Standards.

PAGE 150

150 Figure 5 1. Instron 3384 universal testing machine and data acquisition system (Photo courtesy of the author)

PAGE 151

151 Fig ure 5 2. Examples of compressive testing (Photo courtesy of the author) A) Examples of compression specimens with 1.0 in. thick foam B) 5.0 pcf dense foam specimens during testing (top) and after testing (bottom)

PAGE 152

152 Figure 5 3. Direct uniaxial ten sile testing configuration (Photo courtesy of the author)

PAGE 153

153 Figure 5 4. FPB testing setup and dimensions of test specimens. A) A photograph of FPB testing configuration (Photo courtesy of the author) B) A sketch of FPB testing load locations and dimensions of test specimens.

PAGE 154

154 Table 5 3 Mean values of the polyol isocyanate foam properties under uniaxial compression Thickness ( in. ) Density ( pcf ) Elastic M odulus E c ( psi ) Yielding S tre ngth y ( psi ) Yielding S train y (%) a CoV of E c CoV of y CoV of y 1.0 5.0 3080 96 3.1 0.02 0.02 0.02 5.5 3530 113 3.2 0.06 0.04 0.07 6.0 3585 115 3.2 0.04 0.08 0.09 2.0 5.0 2625 78 3.0 0.05 0.01 0.08 5.5 2915 90 3.1 0.04 0.05 0.03 6.0 3640 105 2.9 0.03 0.03 0.02 3.0 5.0 3255 94 2.9 0.03 0.07 0.09 5.5 3945 110 2.8 0.03 0.02 0.09 6.0 4315 125 2.9 0.04 0.01 0.08 a CoV: Coefficient of variation, = standard deviation ( ) / mean value ( )

PAGE 155

155 Figure 5 5. The effects of foam density ( A, B and C ) and thickness ( D, E and F ) on its compressive stress strain responses

PAGE 156

156 Fig ure 5 6. Failure mechanisms of the composite structure in tensile testing (Photo courtesy of the author) A) Failure of specimens with 25 mm thick foam under tension B) Failure of specimens with 50 mm thick foam under tension C) Failure of specimens with 75 mm thick foam under tension D) Air voids (marked in blue) observed at the cr oss section where core fracture occurred in tensile testing

PAGE 157

157 Figure 5 7. The effects of foam thickness ( A B and C ) and density ( D E and F ) on tensile stress strain responses.

PAGE 158

158 Table 5 4. Experimental mean values of the polyol isocyanate foam properties in tensile tests Thickness ( in. ) Density ( pcf ) Tensile E lastic M odulus E t ( psi ) Ultimate S trength a or b ( psi ) CoV of E t CoV of t u 1.0 5.0 4400 1 45 0.05 0.06 5.5 4720 1 50 0.04 0.06 6.0 4820 1 50 0.06 0.06 2.0 5.0 3240 95 0.02 0.07 5.5 3570 100 0.05 0.03 6.0 4200 110 0.05 0.10 3.0 5.0 3733 93 0.09 0.07 5.5 4637 120 0.05 0.17 6.0 5100 120 0.06 0.10 a : Bonding strength at FRP/foam interface. b : Ultimate tensile strength of foam.

PAGE 159

159 Table 5 5. Experimental mean values of shear strength for the polyol isocyanate foam Density ( pcf ) Foam T hickness ( in. ) Width ( in. ) ( lbf ) a ( psi ) 5.0 1.0 4.0 810 96 2. 0 4.0 1055 65 3.0 6.0 2920 81 5.5 1.0 4.0 855 100 2. 0 4.0 1180 73 3.0 6.0 2920 81 6.0 1.0 4.0 900 105 2. 0 4.0 1300 78 3.0 6.0 3145 87 a : Ultimate shear strength of beam like sandwich panels

PAGE 160

160 Figure 5 8. Failure mechanisms observed in FPB testing (Photo courtesy of the author) A) Unsymmetrical shear failure in the foam B) Symmetrical shear failure in the foam C) De bonding at the core to facing interface D) Local coll apse at the top FRP surface E) Local buckling of the top FRP F) Local de bonding of the FRP skin

PAGE 161

161 Figure 5 9. Load deflection diagrams for FPB testing

PAGE 162

162 Figure 5 10. A photograph of the full size sandwich panel (Photo courtesy of the author)

PAGE 163

163

PAGE 164

164 Fig ure 5 11. Experimental design of the uniform out of plane pressure testing. A) A p hotograph of the High Airflow Pressure Load Actuator (HAPLA) (Photo courtesy of Lopez et al. 2011 ) B) A p hotograph of the panel mounted in the test chamber (Photo courtesy of the author). C) C onstraint details and dimensions of the panel (only the front is s hown herein and the back has the same constraint mechanism) (Photo courtesy of the author). D) The distance sensor for deflection measurements (Photo courtesy of http: //www.cbtcompany.com/products/BrowseResults.aspx?CatID=282174&P arent= )

PAGE 165

165 Fig ure 5 12. The pressure loading applied on the full size sandwich panel during experimental testing.

PAGE 166

166

PAGE 167

167 Fig ure 5 13. FE model of the full length panel and details of constraints in the FE model. A) FE model of the full size composite panel in ADINA B) An overview of the constraints on the FE model C) Rigid links simulating the effects of anchor bolts. D) Contact boundary conditions used to simulate constraints from t he steel frame and steel angles to the sandwich panel.

PAGE 168

168 Table 5 6. Main finite element (FE) model components Item Material Entity Finite element model (ADINA) Steel frame Steel Solid Isotropic linear elastic material Steel angle Steel Solid Isotropic linear elastic material Composite panel FRP Shell Isotropic linear elastic material Foam Solid Nonlinear elastic material FRP to foam interface Constraint Surface s urface rigid link Composite panel to steel frame interface Contact Surface s urface contact Composite panel to steel angle interface Contact Surface s urface contact

PAGE 169

169 Fig ure 5 14. FE model of the core FRP skin interface and material models of the FRP and foam. A) M a terial model for the FRP. B) M aterial model for the foam

PAGE 170

170 Fig ure 5 15. Visual comparison of deformed shape between the experimental testing and FE model. A) Deformed shape in the experiment testing (Photo courtesy of the author) B) Deformed shape in FE model.

PAGE 171

171 Table 5 7. Comparison between flexural responses from pressure testing and FE model Flexural response ( a b ) / Mid span deflection 6.8% 1/4 point strain (compressive) 5.6% 1/2 point strain (compressive) 10.9% 3/4 point strain (compressive) 7.8% 1/4 point strain (tensile) 9.9% 1/2 point strain (tensile) 8.1% 3/4 point strain (tensile) 12.1% a : Strain results from FE model. b : Strain measurements from e xperimental testing

PAGE 172

172 Fig ure 5 16. Experimental and numerical results on the flexural behavior of the large scale composite panel A) Mid span deflection vs. applied pressure ( Note: the disturbance in deflection responses when the load was between 75 and 100 psf was due to the fluctuations in applied pressure shown in Fig. 5 1 2. ) B) Compressive and tensile strain at mid span vs. applied pressure C) Compressive and tensile strain at quarter span vs. applied pressure

PAGE 173

173 Table 5 8 Deflection limits for structural members subjected to wind loads Construction Wind Load Exterior walls and interior partitions: With brittle finishes L/240 with flexible finishes L/120 Source: International Building Code 2009 Table 1604.3 (IBC 2009)

PAGE 174

174 Figure 5 17. Numerical evaluation of the effects of foam and FRP thicknesses on the allowable pressure of the composite panel based on material strength and deflection limits of serviceability A) Four combinations of foam and FRP thickness in FE models. B) Allowable pressure panel span curves for the four cases shown in (A).

PAGE 175

175 Figure 5 18. Numerical evaluation of the effects of inserting through thickness FRP stiffeners in the sandwich panel studied herein. A) FE models of the original panel and panels with two configurations of stiffeners added. B) Rigid links connecting stiffeners to FRP skins in FE models.

PAGE 176

176 Table 5 9 Maximum d eflection s of sandwich panels with FRP stiffeners Stiffener Configuration Maximum Deflection at p = 145 psf No stiffener 4.7 Longitudinal stiffener (spacing =1.5 in.) 2.3 Transverse stiffener (spacing =1.5 in.) 4.4

PAGE 177

177 Figure 5 19. Three point testing on sandwich panels with stiffeners (Photo courtesy of the author) A) Longitudinal stiffeners. B) Transverse stiffeners.

PAGE 178

178 Figure 5 20. Deformed shapes of panel samples at failure (Photo courtesy of the author) A) Unseeable deformation of panel sample with longitudinal stiffeners. B) Obvious deformation of panel sample with transvers stiffeners.

PAGE 179

179 Figure 5 21. Applied load mid span deflection relationships of panel samples with longitudinal and transverse stiffeners. (Note: The slope of the first linear part for each curve represents its stiffness.)

PAGE 180

180 CHAPTER 6 DESIGN AND DEVELOPMENT OF SANDWICH DOOR PANEL The studies on a full size sandwich panel described in Chapter 5, suggest that composite panels are highly suitab le for use in large building components ( e.g., garage doors ) in hurricane prone areas if deflections are controlled C hapter 6 discusses the design of sandwich door panels using the polyol isocyanate foam core characterized in Chapter 5 but with different FRP skin option s. Various width cases were studied based on the standard garage door width existing currently. Their performance under various wind pressure loads was examined to provide reference for future design. In determining the design pressure for the sandwich door panels, s erviceability limit state requirement according to IBC 2009 w as considered At the end, the performance of a current garage door panel (i.e., door 2 studied in Chapter 4) was compared to the sandwich door panel with an equivalent width and the results showed that the sandwich door panel had a smaller deflection tha n door 2 panel Finite element method was used in th e stud y C hapter 6 is a preliminary study for sandwich garage door panel design And the aim is to provide guidance for future work. 6 .1 Design of Sandwich Door Panel 6.1.1 Description of sandwich doo r panel The sandwich panel studied in C hapter 6 is comprised of: 1) two 0.125 in. thick FRP skins and 2) a polyol isocyanate foam core. The bonding between FRP skins and foam core of this sandwich panel is the same as that of the sandwich panel studied in Chapter 5. The material properties of the aerodynamic side skirts were listed in Table 6 1 provided by the Manufacturer (CRANE Composites 2012) The material properties of the foam core were identical to those employed in the sandwich panel discussed in

PAGE 181

181 Section 5.2, i.e., = 6.0 pcf and = 1.5 in. In addition to composite panel itself, aluminum covers around the four edges of the panel were included in this study to simulate the realistic condition. The aluminum covers are to be mounted with hinges, roller assemb lies, etc. to connect adjacent panels and to mount the door into its tracks. The thickness of the aluminum cove rs was assumed to be 0.075 in. Additionally, s ix different width case s ( 8 10 12 14 1 6 and 1 8 ft.) were evaluated (Fig. 6 1 ) 6.1.2 Finite ele ment (FE) model of composite door panel 6.1.2.1 FE model components Due to symmetry, half symmetry model of composite door panel 1 ( Fig. 6 1 ) was built in ADINA 8. 8 (ADINA 2011) T he FRP skins and aluminum covers were modeled using four node shell elements with incompatible modes, and the foam core was modeled using eight node bric k elements (Fig. 6 2 ). R igid links were used to model: 1) the bonding between the FRP skins and foam core; and 2) bonding between the aluminum covers and FRP skins ( Fig. 6 2 ). Table 6 2 summarizes the main components of the FE model of composite panel 1 6.1.2.2 Boundary conditions Fig. 6 3 shows the boundary conditions of the FE model The end of the FE model was constrained in x z translational, x and y rotational degree of freedoms (DOFs). To approximate the constraints from rollers, two points at th e back of the composite panel were constrained in y translational DOF in addition to the constrained DOFs at the end. The symmetry surface and lines were constrained by symmetrical boundary conditions.

PAGE 182

182 6.1.2.3 Material properties in FE model Material properties of FE model of composite door panel 1 are similar to those described in Section 5.2.3, except the FRP skins used and for aerodyn amic side skirts ( Table 6 1 ). For aluminum, the material properties are: , and 6.1.2.4 Pressure load in FE model A ramp pressure up to 200 psf was applied on the surface of t he FE model ( Fig. 6 4 ) to understand its performance under a wide range of wind pressure loading. And the load direction is negative to keep consistent with that applied to current garage doors discussed in Chapter 4. 6.1. 3 FEA results and discussions Fig. 6 5 compares maximum deflections of the six sandwich panels under various wind pressure loads, i.e., 50, 100, 150 and 200 psf For all the six sandwich panels of different width, the maximum deflection occurred at their mid span. It can be observe d that maximum deflections showed an almost linear relationship with the panel width Different from Euler Bernoulli beam theory i.e., mid span deflection is proportional to for uniformly distributed load ( Fig. 6 6 ), the FE model of the sandwich pan els considered large displacement analysis, i.e., the geometry of the structure is taken into account to update the coordinate system of each element during the deformations (ADINA 2011) and therefore the stiffness matrix is constantly updated. As discussed in Section 5.2.3, deflection is the controlling factor of using the sandwich panel for large building cladding systems in high wind areas. Considering material strength and deflection limits based on serviceability requirements (IBC 2009)

PAGE 183

183 Fig. 6 7 sh ows the allowable pressure for each panel width scenario and it can be noticed that the allowable pressure decreases with the panel width. However, when assembling several panels for a garage door, the allowable pressure can increase due to constraining effects between adjacent panels. 6.2 Comparison between Composite Door Panel and Current Door Panel Studies in Section 6.1 provided a preliminary research for design of the sandwich panel for garage doors. To obtain a n understanding of the difference between the sandwich door panel and current garage door panel, the research presented in this secti on compares the performance of a current garage door panel to the sandwich door panel with the equivalent width. Specifically, door 2 was selected to conduct the comparison, because it showed the best performance of the five door specimens evaluated in Cha pter 4. 6.2.1 Comparison methodology To conduct the comparison, FE method was used and summarized as follows The FE model of one panel was taken out from the global FE model of door 2 (Fig. 6 8) The FE model of the sandwich door panel with the same widt h as door 2 (i.e., 18 ft. sandwich door panel ) from the studies described in Section 6.1 was used (Fig. 6 8) The load applied on the FE model of door 2 panel used the same pattern that employed for FE models of sandwich door panels described in Section 6. 1.2.4 ( Fig. 6 4) The same boundary conditions were used for the two FE models (Fig. 6 8) .

PAGE 184

184 6.2.2 Comparison results and discussions Fig. 6 9 compares the relationship of the applied pressure and maximum deflections between door 2 panel and the sandwich door panel. The following conclusions can be drawn. (1) When the suction pressure was less than 110 psf, door 2 panel had smaller deflections, however; when the load was more than 110 psf, sandwich door panel showed a better performance. (2) Through the load rang e, door 2 panel showed a linear behavior; while the sandwich door panel showed a nonlinear behavior, and the stiffness was higher under a larger load. The possible reasons for the difference are as follows. (a) For the sandwich door panel, larger catenary forc es were developed at large r deflections (Fig. 6 10) T h us the panel behaved in catenary action and its axial restraint stiffness became d ominant in the resistance of the sandwich panel subjected to out of plane pressure loading (b) Compared to the catenary forces of the sandwich door panel, the catenary forces of door panel were larger at smaller deflections, but became much smaller at larger deflections (Fig. 6 10). Therefore, the resistance to out of plane pressure loading for door 2 was mainly from the be nding stiffness of stiles and U bars. The reason for the change in the catenary force of door 2 panel at around 1.5 in. could be the local yielding of the door skin around the location where axial constraints were applied. Even though door 2 panel deflect ed less than the sandwich door panel when the pressure load was small, hurricane pressure traces are highly fluctuating, and therefore, the sandwich door panel can be a high potential to be used in hurricane prone regions. Additionally, much fewer components of the sandwich door panel contribute to an easier installation compared to current garage doors This comparison was only conducted using FEA method, and therefore the results are recommended to be validated using experimental testing in the f uture.

PAGE 185

185 6.3 Summary C hapter 6 described the preliminary design and development of sandwich door panel s Based on the standard garage door width, sandwich door panels of various widths were studied using FEA FEA results gave maximum deflections of the sand wich door panels under different applied load. At the same applied pressure, the maximum deflection increase d linearly with panel width Additionally, considering both strength and s erviceability requirement per IBC 2009 the allowable pressure of the sandwich door pa nels decreased nonlinearly with the panel width. This part of the study provided a preliminary guide before fabricating the sandwich panels for ga rage doors of different sizes. The second part of the study in Chapter 6 compared the performance of door 2 pa nel investigated in Chapter 4 and the sandwich door panel of an equivalent width. FEA results showed that door 2 panel had a smaller deflection when the applied pressure was low, but the sandwich door panel deflected less under a higher pressure. Given hig h fluctuations of hurricane wind pressure the sandwich door panel should be a better potential for use in hurricane prone regions. Additionally, compared to a current garage door panel, the sandwich door panel has fewer components (e.g., there are no stiles, U bars and related fasteners), and this contributes to an easier installation and lower occurrence of construction errors.

PAGE 186

186 Table 6 1. Physical properties of aerodynamic side skirt Property Aerodynamic Side Skirt Test Method a Specific Gravity 1 .6 ASTM D790 Flexural Strength 2 5 10 3 psi ASTM D790 Flexural Modulus 0. 9 10 3 psi ASTM D790 Tensile Strength 28 10 3 psi ASTM D638 Tensile Modulus 1.2 10 6 psi ASTM D638 Coefficient of Linear Thermal Expansion 1. 5 10 5 in / in F ASTM D696 Barcol Hardness 5 0 ASTM D2583 Izod Impact 25 ft lb / in notched ASTM D256 Thermal Conductivity 0.4 Btu in /hr ft 2 F ASTM C177 Water Absorption 0. 3 %/24hrs@ 77 F ASTM D570 a Specific gravity is the ratio of the density of a material to the density of a reference material, which is water for this case

PAGE 187

187 Figure 6 1. Half symmetry models of sandwich door panels of different width (8 ft., 10 ft., 12 ft., 14 ft., 16 ft., and 18 ft. from left to right)

PAGE 188

188 Figure 6 2. FE model Components of a sandwich door panel (14 ft. wide panel is shown as an example)

PAGE 189

189 Table 6 2 FE model components of a sandwich door panel Item Entity Finite element model (ADINA) FRP Shell Isotropic linear elastic material Foam Solid Nonlinear elastic material Aluminum cover Shell Isotropic linear elastic material FRP to foam interface Constraint Surface s urface rigid link Aluminum cover to FRP interface Constraint Surface s urface rigid link

PAGE 190

190 Figure 6 3. Boundary conditions of FE model of the sandwich door panel

PAGE 191

191 Figure 6 4. The load applied to the sandwich door panel

PAGE 192

192 Figure 6 5. Maximum deflections of the sandwich door panels under different applied pressure. A) p = 50 psf. B) p = 100 ps f. C) p = 150 psf. D) p = 200 psf.

PAGE 193

193 Figure 6 6. A sketch of the Euler Bernoulli beam deformation under uniformly distributed load

PAGE 194

194 Figure 6 7. The allowable pressure for the sandwich door panels considering strength and serviceability requirements

PAGE 195

195 Figure 6 8. FE models of a current garage door panel and the sandwich door panel to be compar ed on their performance under wind pressure loading

PAGE 196

196 Figure 6 9. Comparison of maximum deflections between door 2 panel and sandwich door panel

PAGE 197

197 Figure 6 10. Comparison of the catenary force maximum mid span deflection relationship between door 2 panel and the sandwich door panel.

PAGE 198

198 CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS FOR FURTURE RESEARCH The central theme of this research focuse d on mitigating wind damage to sectional garage doors and assessment of composite panel systems for door systems Chapter 1 through 6 presented the studies involved in 1) investigating structural behaviors of current sectional doors under simulated wind pressure, and 2) characterizing mechanical resistance of a new FRP/ foam composite structure, which will be used for a new design of se ctional doors in hurricane prone regions. The following sections recapitulate the conclusions and contributions of the studies documented in this dissertation and present recommendations for future research in this area. These sections are organized based on research topic. 7 .1 Forensic Assessment of Current Garage Doors using Full Scale Experimental and Numerical Methods 7 .1.1 Conclusions and Contributions This study was conducted to advance fundamental understanding of wind induced performance of curre nt sectional doors, a rarely studied topic in wind engineering field. Full scale experimental testing and finite element method were employed to conduct this research for the sake of accuracy and cost efficiency. Destructive tests on five commercial sectio nal doors were conducted experimentally with quasi static pressure loading. Based on the experimental observations, two of the five doors were modeled using FE method to obtain a more detailed understanding of their structural performance. A new large scal e dynamic wind pressure simulator (Simulator) was developed to conduct full scale testing. The Simulator is designed to be able to replicate wind loads associated with SSH W S Category 5 hurricane conditions. Particularly, the

PAGE 199

199 Simulator is designed to compen sate for high airflow leakage through test specimens up to 100, 000 CFM, which is the first of this kind. Additionally, the Simulator is able to test a wide range of building envelope components as long as the dimension is smaller than 24 ft. wide x 18 ft. high. Full scale experimental tests were conducted on five commercial sectional doors using the newly developed Simulator. Experimental results indicated a linear relationship between structural responses (including strains and displacements) and applied suction before catastrophic failure of each door specimen. During experimental testing, adhesive failure was observed for three door specimens, but they were still operable. Experimental testing revealed two main catastrophic failure mechanisms of the five doors under suction pressure, i.e., local buckling of U bars and disengagement of rollers from the door tracks. The two failure mechanisms occurred sometimes toge ther and sometimes separately. Following the experimental testing, FE models of two doors, na mely, door 2 and door 3, were developed in ADINA. FEA results matched experimental testing well, which validated the FE models and indicated their potential to be used for other scenarios in the future. The good agreement also demonstrated the success of t he methodology, combining full scale experimental testing and numerical analysis, to analyze such complex structure. From FEA results, it was noticed that disengagement failure was led by other failures, namely, buckling of U bars for door 2 and overbendin g of roller stems for door 3. This indicated that material strength was controlling factor for the performance of current garage doors under wind pressure loading Additionally, three construction errors were simulated in the validated FE model of door 2 t o

PAGE 200

200 investigate their effects on the structural performance. It was found out missing fasteners connecting U bars to stiles and 1 in. overlap at the end of the door caused local buckling failure of U bars occurred at a lower load, while missing fasteners con necting hinges to stiles did not affe ct the structural performance. Investigations of current sectional doors revealed their insufficient stiffness under wind pressure loading. Additionally, preparation for experimental testing also addressed other shortco mings of current sectional doors, e.g., 1) a large weight, 2) time consuming installation and 3) high probability of construction errors in shipping and installation due to too many components. These findings provided a direction for a new design philosoph y of sectional door s for use in strong wind areas. 7 .1.2 Recommendations for Future Research Suggestions for future research on current sectional doors are provided in two areas: experimental testing and FE model. 7 .1.2.1 Full s cale e xperimental t esting In this study, the testing conducted on the five sectional doors was quasi static. According to testing results, for some doors, the load at catastrophic failure was lower than the test pressure, while it was higher for others, which was unexpected before th e experimental testing. However, their performance under dynamic wind pressure is questionable. Therefore, it is proposed to conduct dynamic pressure testing as future work after the Sim ulator is finished calibration. Additionally, adhesive failure occurre d for several sectional doors before their catastrophic failure. Even though the adhesive failure did not prohibit their operability, it calls for the necessity to define the failure / failures for garage doors.

PAGE 201

201 7 .1.2.2 Finite e lement m odel FE m odel of d o or 2 Limitations of current FE model of door 2 are listed below. (1) Rollers were not included in current FE model of door 2. Therefore, if other loading scenarios (e.g. dynamic wind pressure loading) cause failure of the roller components leading to disengagement failure before buckling of U bars, current FE model will be unable to recreate this failure. (2) Adhesive between door skins and stiles was modeled as rigid links. Therefore, if future testing involves adhesive failure, this model is not appropri ate. (3) Fixing x and z translational DOFs might over constrain the side of the FE model. Based on the limitations, the following suggestions are provided for future to modify FE model of door 2 (1) Use the method employed for the FE model of door 3, including l ocal and global models, to modify the FE model of door 2: A local model, including roller stem, roller and track, can be developed to determine the constraints from the track to door 2. The constraints can be modeled as nonlinear springs and then be incorp orated into the global model of door 2. This modification, on one hand, can capture the failure of roller components; on the other hand, can develop more realistic boundary conditions on the side of the FE model. Therefore, limitations 1 and 3 can be addre ssed. (2) Adhesive strength should be obtained through experimental testing, and spring elements employing the testing results should be included in FE model of door 2. With this modification, the adhesive failure can be recreated, and therefore, limitation 2 can be eliminated. FE m odel of d oor 3 Limitations of current FE model of door 3 are listed below. (1) Similar to FE model of door 2, adhesive between door skins and sti les was modeled as rigid links. (2) The local model of roller and stem only considered the nega tive pressure situation. Therefore, the current model is limited for negative pressure loading. According to the limitations, the following modifications are suggested for FE model of door 3 (1) Similar to door 2, adhesive strength should be obtained through experimental testing, and spring elements employing the testing results should be included in

PAGE 202

202 FE model of door 3. With this modification, the adhesive failure can be recreated, and therefore, limitation 1 can be eliminated. (2) The local model should be loaded with prescribed displacement representing positive pressure applied to the door, so that constraints of the track in both positive and negative pressure conditions can be obtained. With this modification, the model can be applied for a wider use in the future. Types of Analysis In addition to suggestions on FE models, the following kinds of analysis are also recommended to be conducted using FEA to obtain a more thorough understanding of curr ent sectional doors (1) Analysis of the effects of adhesive failure to structural behaviors. (2) Analysis of the influence of construction errors to structural performance of door 3. (3) Dynamic analysis. 7 .2 Mechanical Resistant Properties of FRP/ Foam Composite Str ucture 7 .2.1 Conclusions and Contributions This study was performed to investigate the applicability of a FRP/polyol isocyanate foam sandwich panel for use as larger component and cladding systems intended in high wind areas. The first part of this study c onducted a series of material testing to obtain mechanical properties of the sandwich panel Foam thickness and density were varied to evaluate their effects on mechanical properties of the sandwich panel The second part of this study employed full scale experimental testing and FE method on a full size panel subjected to uniform pressure to determine the appropriateness of the sandwich panel used in high wind areas. At the end, different optimizations were proposed to increase its stiffness using FE metho d. Constitutive properties of the sandwich panel were obtained f rom the material testing This contributes the knowledge base on sandwich panel systems using foam

PAGE 203

203 cores. An increase in foam density was found to improve the performance of the sandwich panel except that the core FRP skin bonding strength does not change with the foam density. Studying the influence of foam thickness indicated that the air voids produced a significant negative influence on the foam properties. Results from FE model and experi mental testing of the full size panel subjected to uniform wind pressure matched well, which verified the material properties of the foam core obtained experimentally. The good agreement also implied that the validated FE model could be used for future opt imization. The full size panel presented a linear elastic behavior over a wide range of wind loading, and deflection appeared to be the controlling factor for using this sandwich panel for building envelope components in hurricane prone regions. FEA result s on this issue indicated that the sandwich panel can be used for a wide range of design considerations in high wind areas through adjusting foam thickness, FRP thickness, span length and adding FRP stiffeners between two skin layers of the sandwich panel 7 .2.2 Recommendations for Future Research The work included in this study presents the most current knowledge of the new FRP/polyol isocyanate foam sandwich panel To further understand the novel sandwich panel the following research activities are sugge sted (1) For material testing of the sandwich panel the influence of foam thickness was not sufficiently evaluated due to the air voids effects. Therefore, more thickness scenarios should be tested to possibly evaluate its effects to the material properties of the sandwich panel (2) The proposed solutions to increase the panel stiffness were evaluated using FEA, and adding FRP stiffeners were evaluated in a small size. A full size panel with FRP stiffeners should be experimentally tested to verify the FEA resul ts.

PAGE 204

204 7.3 Research and Development of Sandwich Garage Door Panels 7.3.1 Conclusions and Contributions Based on the findings through characterization of the novel sandwich panel t his study was conducted as an initial step to design and develop hurricane resistant garage doors using sandwich panels. The first part of this study used FE A method to estimate the performance of the sandwich door panels of interest in this research. Thes e sandwich door panels were different in width and t h e width scenarios were chosen based on the common widths used for current garage doors. The second part of this study compared behaviors of a current garage door panel (i.e., the panel of door 2 studied in Chapter 4) and the sandwich door panel of interest in this study. Maximum deflections of the sandwich door panels under different applied pressure were calculated. It was observed that the maximum deflection increased linearly with the door width. Considering strength and serviceability requirements, the allowable pressu re for the sandwich door panels was calculated. And the allowable pressure decreased nonlinear ly with the door width. This part of the study as a preliminary investigation of the novel sandwich door panels, can be used for performance based design of the panels. Additionally, FE A method used in this study contributes a cost efficient method. Comparison between door 2 panel and the sandwich door panel indicated that the sandwich door panel could perform better when subjected to hurricane wind pressure, whic h is highly fluctuating. This part of the study provides a direct understanding of the performance of the novel sandwich panel compared to a current garage door panel. The comparison results on deflection responses (i.e., the sandwich panel showed a smalle r deflection than door 2 panel under higher wind pressure

PAGE 205

205 loading) demonstrated the potential of using this sandwich panel for garage doors in hurricane prone regions. Additionally, compared to a current garage door panel, fewer components of the sandwich panel added another advantage. 7.3.2 Recommendations for Future Research The work conducted on design and development of sandwich panels for garage doors is at a preliminary level. Future research activities on this topic are suggested as below. (1) Optimizati on of the sandwich door panel s should be conducted considering structural performance and cost. Several iterations of the design should be investigated until the performance and cost are satisfactory. Considering time and cost, FEA should be employed durin g the iterations. (2) In addition to static pressure testing, dynamic testing on the sandwich panels should be conducted employing realistic hurricane wind pressure data. (3) Considering accuracy and cost, t he satisfactory design of the sandwich panel should be f abricated and then tested experimentally to validate FEA results. Both static and dynamic pressure testing should be conducted, and the d ynamic testing could be conducted once finishing calibrating dynamic capacity of the Simulator

PAGE 206

206 APPENDIX A CONVERTING WIND TUNNEL DATA TO FULL SCALE DATA In the wind tunnel data file extracted from the NIST Aerodynamics Database, those data related to the conversion discussed here are list below: = the mean wind velocity measured at the reference height in the wind tunnel where upper level wind speed is taken, is equal to 45 mph (Ho et al. 2003) Here, ms represents model scale. = the wind pressure coefficient at the reference height. = the sampling frequency of wind tunnel data, is equal to 500 in the study. = the conversion factor. = the length scale, is equal to 1:100 in the study. Here, f s represents full scale. = the roof height, is equal to 16, 24, 32, and 40 ft. in the study. Additionally, full scale data related to the conversion are: = the full scale 3 s gust wind speed at the reference height (10 m). = the gust factor, is equal to 1.5 for the conversion between 3 s gust to 1 hour mean wind speed. To use the wind tunnel data in full scale, the following two conversions are to be done Wind pressure coefficient conversion, i.e., converting wind tunnel pressure coefficient at the reference height (i.e., ) to roof height (i.e., ) using Equation A 1. (A 1)

PAGE 207

207 Time scale c onversion, i.e., converting wind tunnel frequency (i.e., ) to full scale frequency (i.e., ). And the procedure is presented in the following: (1) Convert full scale 3s gust wind speed at 10 m (i.e., ) to 1hr wind speed at 10 m (i.e ., ) via Equation A 2. (A 2) (2) Convert full scale 1hr wind speed at 10 m (i.e., ) to 1hr wind speed at roof height H (i.e., ). Wind speed at a ce rtain height z can be calculated using Log Law (Ishizaki 1983) (A 3) Where represents shear velocity, and is the von Krmn constant Because shear velocity and the von Krmn constant keep unchanged, therefore, 1 hr. wind speed at roof height H can be calculated via Equation A 4. (A 4) (3) Convert wind tunnel frequency to full scale frequency. This step is based on that the Strouhal number ( fL / v ) does not change, where f is the frequency, L is the characteristic length (e.g., building dimensions) and v is the velocity. Therefore, full scale frequency and time scale can be ca lculated by Equations A 5 and A 6 (A 5) (A 6) It is noted that the velocity in wind tunnel and full scale must be referenced to the same height.

PAGE 208

208 APPENDIX B BACKWARD DIFFERENCE APPROXIMATION METHOD The equations to determine the pressure and airflow requirements of the fan system for the Simulator we re given in Chapter 3, and are listed below as well (3 5) (3 6) (3 7) Where p ext2 is equal to the atmospheric pressure out of the specimen as discussed in Chapter 3, and theref ore p ext2 p airbox = p airbox which is gauge pressure, and p airbox is the target pressure (also gauge pressure here) time series in the airbox. Because pressure can be expressed in the term of pressure c oefficient by Equation B 1 (B 1) W here U H is the full scale 1hr wind velocity at roof height. Therefore, Equations 3 5 to 3 7 can be expressed as below (B 2) (B 3) (B 4) Where C pext1 is the wind pressure coefficient at the opening connected to the ducting system, which is to be calculated here, and C pairbox is the internal pressure coefficient in

PAGE 209

209 the airbox To simplify the calculation process, Equation B 2 to B 4 are expressed as below (B 5) (B 6) (B 7) W here Equations B 5 to B 7 are solved using backward differencing approximation (B 8)

PAGE 210

210 (B 9) Where i represents the i th time step, i.e., i th iteration. Therefore, at ith time step, discharge equations B 5 to B 7 can be written as : (B 10) (B 11) (B 12) a nd the Initial conditions are assumed x (0) = x ( 1) = 0 (Oh 2004) Fig. B 1 presents the flow chart to calculate the pressure coefficient procedure.

PAGE 211

211 Figure B 1. Flow chart for Computing the airflow movements and external pressure at the opening of the airbox.

PAGE 212

212 LIST OF REFERENCES Ahmed, S., Canino, I., Chowdhury, A., Mirmir of the Capability of Multiple Mechanical Fasteners in Roof to Wall Connections of Practice Periodical on Structural Design and Construction 16(1), 2 9. Strain 43(4), 299 305. Rolling Doors: Determination of Structural Performance under Uniform Static Air Approved American National Standard DASMA, Cleveland, OH, 17. ASCE 7. (2005). ASCE standard, minimum design loads for buildings and other structures ASCE Standard 7 05, Reston, Va ASCE 7. (2010). ASCE standard, minimum design loads for buildings and other structures ASCE Standard 7 10, Reston, Va. ASHRAE. (2001). 2001 ASHRAE Handbook Fundamentals, Inch Pound Edition ASHRAE, Atlanta, GA. ASTM C 297M 04 ASTM International, Philadelphia, 16 21. ASTM C 365M 05 ASTM International, Philadelphia, 34 39. ASTM C 393M 06 ASTM International, Philadelphia. reinforced plastics and electrical insulating materials by four ASTM D 6272 02 ASTM International, Philadelphia.

PAGE 213

213 Journal of Architec tural Engineering 5(1), 16 24. composite Composite Structures 61(1 2), 13 25. Bender, M. A., Knutson, T. R., Tuleya, R. E., Sirutis, J. J., Vecchi, G. A., Garner, S. T., Science 327(5964), 454 458. Blackall, T. N., an 8. development of real Philosophica l Transactions of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences 359(1786), 1869 1891. Journal of Wind Engineering and Industrial A erodynamics 14(1 3), 103 112. Journal of Wind Engineering and Industrial Aerodynamics 48(1), 81 94. pressure time series in an Journal of Wind Engineering and Industrial Aerodynamics 91(6), 737 765. Cholod, M. (1988). Failure of Wood Framed Houses Faculty of Engineering Science, The University of Western Ontari o, London, Ont., Canada. Loading Effects on Roof to Journal of Engineering Mechanics 139(3), 386 395. Chowdhury, A., Natural Hazards Review 10(1), 1 10. Collins, S. D. (1986). Design and Scaling of Models Work term Rep., Faculty of Engineering, U niversity of Waterloo.

PAGE 214

214 Journal of Wind Engineering and Industrial Aerodynamics 42(1 3), 1525 1536. Journal of Wind Engineering and Industrial Aerodynamics 29(1 3), 99 107. c. Composites, Inc. Croft, P., Drgger, P., Hardy Pierce, H., Moody, R., Olson, R., Robertson, R., Roodvoets, D., Shoemaker, L., Smith, R., and Wilson, J. (2006). Hurricanes Charley and Ivan Wind Investigation Report RICOWI. way bending behavior of 3 D GFRP sandwich panels with through Composite Structures 92(4), 950 963. Edmonson, W., Schiff, S., Framed Wood Roof to Wall Connectors Using Aged Lumber and Multiple Connection Journal of Performance of Constructed Facilities 26(1), 26 37. Ellingwood, B., Rosowsky, D., Li, Y., and Kim, J. (2004). Journal of Structural Engineering 130(12), 1921 1930. 30 Nature 436(7051), 686 688. FBC TAS 202 94 International Code Council Inc., Washington, DC. FEMA. (2005a). Mitigation Assessment Te am Report: Hurricane Charley in Florida FEMA. FEMA. (2005b). Summary Report on Building Performance: 2004 Hurricane Season FEMA. FEMA. (2008). Summary report on building performance: Hurricane Katrina FEMA.

PAGE 215

215 Gao, T., and Moen, C. D. (2009). Experiment al Evaluation of A Vehicular Access Door Subjected to Hurrican Force Win Pressures Virginia Polytechnic Institute and State University. Gao, T., and Moen, C. D. (2011). Vehicular Access Doors Under Hurricane Force Wind Pressure: Analysis Methods and A De sign Tool Virginia Polytechnic Institute and State University. of Western Ontario, London, Ont., Canada. under Fluctuating Wind Journal of Architectural Engineering 17(1), 34 41. Dimensional Timber Light Journal of Structural Engineering 127(8), 901 913. Ho, T. C. E., Surry, D., and Morrish, D. (2003). NIST/TTU Cooperative Agreement Windstorm Mitigation Initiative: Wind Tunnel Experiments on Generic Low Buildings The Boundary Layer Wind Tunnel Laboratory, The University of Western Ontario, 106. Engineering Structures 7(4), 226 230. Huang, H. physical systems for real Proceedings of the 1s t ACM/IEEE International Conference on Cyber Physical Systems ACM, New York, NY, USA, 69 78. 2009 IBC International Code Council Inc., Washington, DC, 305 307. Irwin, P. A., and Sifton, V Journal of Wind Engineering and Industrial Aerodynamics 77 78, 715 723. typhoon Journal of Wind Engineering and Industrial Aerodynamics 13(1 3), 55 66. Element Model of Complete Light Frame Wood Structures. Journal of Structural Engineering 120(1), 100 119. Khemani Kishan C. Polymeric Foams ACS Symposium Series, American Chemical Society, 1 7.

PAGE 216

216 strength of stitched foam Composite Structures 47(1 4), 543 550. Journal of Wind Engineering and Industrial Aerodynamics 23, 223 235. Klempner, D., Sendijarevi V., and Aseeva, R. M. (2004). Handbook of Polymeric Foams and Foam Technology Hanser Verlag, Munchen, Germany. Wind Tunnel and Full Natural Hazards Review 11(4), 151 161. framed low rise Engineering Structures 39, 79 88. Lacasse M. A., Manning, M., Rousseau, M., Cornick, S. M., Plescia, S., Nicholls, M., Window Lin, H. roperty relationships of commercial foamed Polymer Testing 16(5), 429 443. The Seventh As ia Pacific Conference on Wind Engineering Taipei, Taiwan. residential window and wall systems subjected to steady and unsteady wind Building and Environment 46( 7), 1329 1342. Properties of Polyurethane Journal of Engineering Mechanics 121(4), 528 540. Masters, F. J., Gurley, K. R., Shah, N., and Fernandez, Engineering Structures 32(4), 911 921. Based Engineering in Multi Hazard Coastal Structures Congress 2011 American Society of Civil Engineers, 1961 1972.

PAGE 217

217 A Journal of Wind Engineering and Industrial Aerodynamics 43(1 3), 1717 172 6. Engineering Structures 6(4), 242 247. Journal of Architectur al Engineering 11(1), 10 13. Story Residential House Under Realistic Ont., Canada. nce of toe nail connections under Engineering Structures 33(1), 69 76. void characterisation of foam concrete. Cement and Concrete Research 37(2), 221 230. Proceedings of the INEL Severe Windstorm Testing Workshop Idaho. induced internal pressure in low University of Western Ontario, London, Ont., Canada. Earth and Space 2010 American Society of Civil Engineers, 2179 2187. Pielke, R., Gratz, J., Landsea, C., Collins, D., Saunders, M., and Musulin, R. (2008). Natural Hazards Review 9(1), 29 42. Pinelli, J. P., Subramanian, C., Zhang, L., Gurley, K. R., Cope, A., Simiu, E., Diniz, S., sses for Proceedings of the European Safety and Reliability Conference Netherlands. Mechanically Attached Single Ply Roofing Systems: The Need f or Correction Journal of Structural Engineering 134(3), 489 498. Rappaport, E. (1993). Preliminary Reprot: Hurricane Andrew 16 28 August, 1992 National Hurricane Center.

PAGE 218

218 Weath er 49(2), 51 61. RDH Building Engineering Limited. (2002). Water Penetration Resistance of Windows Study of Codes, Standards, Testing, and Certification RDH Building Engineering Limited. J. Struct. Eng. 110(4), 715 729. Riley, M. A., and Sadek, F. (2003). Experimental Testing of Roof to Wall Connections in Wood Frame Houses Building and Fire Research Laboratory, 79. values for metal connectors i n light Journal of testing and evaluation 26(5), 426 433. of residential window installation options for hurricane Building and E nvironment 45(6), 1373 1388. models for roof components in existing light Engineering Structures 31(11), 2607 2616. Shen, S. Y., Masters, F. resistance properties of FRP/polyol Composite Structures 99, 419 432. Rolla). Journal of Polymer Science Part B: Polymer Physics Stopar, E. M. F. (1987). Wind Damage on Masonry Walls of Low Rise Flat Roof Buildings Faculty of Engineering Science, The University of Western Ontario, London, Ont., Canada. Journal of Wind Engineering and Industrial A erodynamics 32(3), 343 360.

PAGE 219

219 of tornados with a low Journal of Wind Engineering and Industrial Aerodynamics 99(4), 369 377. Vedavarz, A., Kuma r, S., and Hussain, M. (2007). HVAC: The Handbook of Heating, Ventilation and Air Conditioning for Design and Implementation Industrial Press Inc., New York, NY. factors for internal Journal of Wind Engineering and Industrial Aerodynamics J WIND ENG IND AERODYN 23, 259 271. Viot, P., Beani, F., and Lataillade, J. Journal of Materials Science 40(22), 5829 5837. Vissch Journal of Wind Engineering and Industrial Aerodynamics 95(8), 697 713. Wouterson, E. M., Boey, F. Y. C., Hu, X., and Wong, S. an Composites Science and Technology 65(11 12), 1840 1850. reduced scale models of masonry i Earthquake Engineering & Structural Dynamics 30(6), 819 834. polymeric foam material subjected to dynamic crash load International Journal of Impact Engineering 21(5), 369 386.

PAGE 220

220 BIOGRAPHICAL SKETCH Yan Shen was born in Suzhou, China, to Xiuhong Shen and Meifang Zhou, in 1987. She lived in Suzhou until she finished her high school in 2003. In September 2003, she started her undergraduate study in Naval Architecture and Ocean Engineering at Tianjin University. During her undergraduate studies, she kept top five and averaged top 1 in her school studies. In July 2007, she received her Bachelor of Science degree in N aval Architecture and Ocean Engineering. Two months later, she was recommended and enrolled to the graduate school without entrance examination at Tianjin University. During 2007 2009, she studied in the area of Ocean Engineering under the supervision of D r. Yougang Tang. Her master Study on the Snap Tension Developed in the Mooring System of Ocean Platforms Considering Slack Taut Effects Engineering. In May 2008, an earthquake hit and caused tremendous damage to Wenchuan, a city in southwest China. This earthquake propelled her to pursue a PhD with a specialization in hazard mitigation. In July 2009, she was accepted into the PhD program of structural engineering in the Department of Civil and Coastal Engineering at the University of Florida under the guidance of Dr. Forrest Masters. Since then, her work has primarily focused on using full scale experimental and numerical methods to evaluate behaviors of current garag e doors under wind pressure loading conditions and investigate the applicability of a novel FRP/ polyol isocyanate foam sandwich panel to be used as large building components and cladding system in high wind areas.