This item is only available as the following downloads:
1 THE WIND RESISTANCE OF ASPHALT ROOFING SHINGLES By CRAIG ROBERT DIXON 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
2 2013 Craig Robert Dixon
3 To my Mom
4 ACKNOWLEDGMENTS First and foremost, I would like to thank my advisors, Drs. David O. Prevatt, Forrest J. Masters, and Kurtis R. Gurley for their guidance, support, and friendship. This work would not have been accomplished without their mentoring. I would also like to thank my external committ ee me mber Dr. Bhavani Sankar Additionally, I would like to thank my friends in the wind engineering group for their help in various aspects of my research, especially: Dany Romero, Kenton McBride, Daniel Smith, Scott Bolton, David Roueche, Alon Krathammer, Tuan Vo, and Peter Datin. I would also like to thank the research oversight committee for their valuable input, especially: Dr. Jon Peterka, Tom Smith, Dr. Ben Thomas, and Dr. Walt Rossiter. Finally, I wish to thank my mom for being a sou rce of intelligence and strength The financial support for this research is gratefully acknowledged from the Southeast Region Research Initiative under grant #10031592 Residential Roof Covering Investigation of Wind Resistance of Asphalt Shingles. I am also grateful to the financial support of the Florida Building Commission and Florida Depa rtment of Emergency Management.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 16 Knowledge Gaps in the Wind Resistance of Asphalt Shingles ............................... 18 Research Goals and Scope ................................ ................................ .................... 19 Research Impacts ................................ ................................ ................................ ... 21 Layout of the Dissertation ................................ ................................ ....................... 22 2 ASPHALT SHINGLE COMP OSITION AND INSTALLA TION ................................ 24 Shingle Composition ................................ ................................ ............................... 24 Reinforcement Mat ................................ ................................ ........................... 26 Asphalt Coating ................................ ................................ ................................ 26 Roofing Granules ................................ ................................ ............................. 27 Sealant Stri p ................................ ................................ ................................ ..... 28 Fasteners ................................ ................................ ................................ ......... 31 Underlayment ................................ ................................ ................................ ... 31 Three tab Shingles ................................ ................................ ........................... 33 Laminate Shingles ................................ ................................ ............................ 33 Asphalt Shingle Installation ................................ ................................ ..................... 34 3 LITERATURE REVIEW ................................ ................................ .......................... 43 The Early Years (1893 1950) ................................ ................................ ............... 43 Development of the First Test Standards for Wind Resistance (1950 1980) ....... 45 The Development of the Asphalt Shingle Wind Uplift Model (1980 1997) .............. 50 The Modern Era (1997 ) ................................ ................................ ........................ 61 The ASTM D6381 Asphalt Shingle Mechanical Uplift Resistance Test Method ................................ ................................ ................................ .......... 61 In Service Wind Performance of Asphalt Shingles ................................ ........... 64 4 UNSEALED NATURALLY A GED ASPHALT SHINGLES AND THEIR VULNERABILITY IN WIN D ................................ ................................ ..................... 68 Study 1: Survey of Naturally Aged Shingle Roofs for Unsealed Shingles ............... 68
6 Survey Method ................................ ................................ ................................ 70 Potential for Wind Induced Loss of Shingle Sealing ................................ ......... 70 Survey Results ................................ ................................ ................................ 72 Shingles in the field of the roof ................................ ................................ ... 72 Ridge and hip shingles ................................ ................................ ............... 78 Study 2: Full Scale Testing of Asphalt Shingle Roof Systems ................................ 80 Experimental Design ................................ ................................ ........................ 80 Wind Test Sequence and Boundary Layer Simulation ................................ ..... 82 Results ................................ ................................ ................................ ............. 84 Pre wind test unsealed shingle surveys ................................ ..................... 84 Wind performance of shingles installed in the field of the roof ................... 85 Hip shingle wind performance ................................ ................................ .... 88 Discussion ................................ ................................ ................................ ........ 90 5 WIND RESISTANCE OF N ATURALLY AND ARTIFIC IALLY AGED ASPHALT SHINGLES ................................ ................................ ................................ .............. 92 Aging of Asphalt Shingles ................................ ................................ ....................... 92 Research Objectives ................................ ................................ ............................... 94 Study 1: Wind Uplift Capacity of Asphalt Shingles Subjected to Artificial Aging ..... 96 Experimental Setup ................................ ................................ .......................... 96 Shingle sp ecimen specifications ................................ ................................ 96 Thermal aging chamber specifications and protocol ............................... 96 UV Thermal Water aging chamber specifications and protocol .............. 98 ASTM D6381 Mechanical Uplift Test Procedure ................................ ...... 102 Results ................................ ................................ ................................ ........... 104 ASTM D6381 Procedures A and B uplift resistance ................................ 104 Failure modes in uplifted shingles ................................ ............................ 108 ASTM D7158 total wind uplift resistance ................................ ................. 112 Discussion ................................ ................................ ................................ ...... 114 Study 2: Natur ally Aged Shingle Wind Uplift Resistance ................................ ...... 116 Test Sites ................................ ................................ ................................ ....... 116 Portable Mechanical Uplift Apparatus ................................ ............................ 117 In Situ ASTM D6381 Specimen Preparation and Test Procedure .................. 117 Results ................................ ................................ ................................ ........... 119 In situ ASTM D6381 mechanical uplift resistance ................................ .... 119 Failure modes ................................ ................................ .......................... 120 ASTM D7158 total wind uplift resistance ................................ ................. 121 Discussion ................................ ................................ ................................ ...... 122 Discussion of Co mbined Results ................................ ................................ .......... 123 6 THREE DIMENSIONAL MEASUREM ENTS OF WIND FORCE O N ASPHALT SHINGLE SEALANT STRI PS WITH FULL AND PAR TIAL ADHESIO N ............... 125 Knowledge Gaps ................................ ................................ ................................ .. 126 Wind Load Model and Load Path ................................ ................................ ... 126 Partially Unsealed Shingles ................................ ................................ ............ 128
7 Experimental Setup ................................ ................................ .............................. 129 Concept ................................ ................................ ................................ .......... 129 Test Apparatus ................................ ................................ ............................... 130 Introduction ................................ ................................ .............................. 130 Componentry ................................ ................................ ........................... 131 Test Specimens ................................ ................................ .............................. 134 Test Deck Specifications ................................ ................................ ................ 136 Multi Axis Load Cell Specifications ................................ ................................ 139 Velocity Sensor Specifications ................................ ................................ ....... 142 Test Specimen Installation ................................ ................................ ............. 143 Experimental Procedure ................................ ................................ ................. 14 7 Results ................................ ................................ ................................ .................. 149 Phase I: Wind Field Above the Test Specimen ................................ .............. 149 Mean longitudinal velocity ................................ ................................ ........ 149 Longitudinal turbulence intensity ................................ .............................. 151 Discussion on Turbulence ................................ ................................ .............. 153 Static Pressure ................................ ................................ ............................... 154 Phase II: Ef fect of Static Pressure on Shingle Test Specimens ..................... 155 Phase III: Wind Strip ................................ ................................ ................................ ............. 157 Data processing ................................ ................................ ....................... 157 Fully sealed shingle results mean forces and moments ........................ 159 Fully sealed shingle results force coe fficients ................................ ....... 163 Partially unsealed shingle results measured forces and moments ........ 166 Partially unsealed shingle results interfacial forces ............................... 170 Partially unsealed shingle results force coefficients .............................. 174 Discussion ................................ ................................ ................................ ............ 176 Significant Findings ................................ ................................ ............................... 178 7 CONCLUSIONS AND RECOMMENDATIONS ................................ ..................... 181 Conclusions on the Causes of Wind Damaged Asphalt Shingle Roofing .............. 181 Partially Unsealed Shingles ................................ ................................ ............ 181 Effect of Aging on Wind Resistance ................................ ............................... 183 ASTM D7158 and the Load Model for Asphalt Shingles ................................ 185 ASTM D7158 Design Methodology ................................ ................................ 186 Eave and Rake ................................ ................................ ............................... 187 Recommendations for Future Research ................................ ............................... 187 LIST OF REFERENCES ................................ ................................ ............................. 189 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 197
8 LIST OF TABLES Table page 1 1 Oversight committee members ................................ ................................ ........... 22 2 1 Wind classification required for asphalt shingle installed in Florida .................... 36 3 1 Wind tunnel measured three tab shin gle with cutouts ................................ 58 3 2 Summary of standardized test methods to evaluate asphalt shingle wind performance ................................ ................................ ................................ ....... 64 4 1 Estimates of peak instantaneous velocity near the roof plane at each survey location ................................ ................................ ................................ ............... 73 4 2 Wind test sequence duration, wind speeds, and turbulence intensities .............. 83 5 1 Exposure times where ASTM D6381 tests were performed ............................... 98 5 2 Mean resistance in Thermal and UV Thermal Water methods ......................... 108 5 3 Specimen dimensions, ASTM D7158 differ ential pressure coefficients, and ASTM D7158 required resistance. ................................ ................................ .... 113 5 4 Mean and lowest measured wind uplift resistance vs ASTM D7158 Class H required wind resistance. ................................ ................................ .................. 114 5 5 Test site location, age, type, and quantity of ASTM D6381 tests ..................... 116 5 6 Test site differential pressure coefficients, length dimensions, and ASTM D7158 Class H required wind uplift resistance. ................................ ................ 121 6 1 Test specimen ID, type, planform dimensions, and number of specimens ....... 135 6 2 coefficients ................................ ................................ ................................ ........ 136 6 3 Force coefficients and relative contribution of ASTM D6381 Procedures A and B to total uplift ................................ ................................ ............................ 137 6 4 Six axis load cell sensing ranges and resolutions ................................ ............ 140 6 5 Longitudinal integral length scales measured 12 mm above shingle surface ... 153 6 6 Mean forces and moments measured on fully sealed Laminate specimens ..... 160 6 7 Mean forces and moments measured on fully sealed Three Tab specimens ... 162
9 6 8 Laminate force coefficients directly measured vs. ASTM D7158 predicted ...... 164 6 9 Three Tab force coefficients directly measured vs. ASTM D7158 predicted .... 165 6 10 Mea n forces and moments measured on partially unsealed Laminate specimens ................................ ................................ ................................ ........ 167 6 11 Mean forces and moments measured on fully s ealed Three Tab specimens ... 168 6 12 Laminate Specimen 3 measured moments and estimated interface forces ... 172 6 13 Laminate Specimen 4 measured moments and estimated interface forces ... 172 6 14 Three Tab Specimen 4 measured moments and estimated interface forces 173 6 15 Three Tab Specimen 5 measured moments and estimated interface forces 173 6 16 Three Tab Specimen 6 measured moments and estimated interface forces 174 6 17 Laminate force coefficients directly measured vs. ASTM D7158 predicted ...... 176 6 18 Three Tab force coefficients directly mea sured vs. ASTM D7158 predicted .... 176
10 LIST OF FIGURES Figure page 2 1 Asphalt shingles are installed in the field of the roof with additional shingles al ong hip and ridge lines. ................................ ................................ .................... 24 2 2 Plan view of a standard three tab shingle with six fastener locations shown. ..... 25 2 3 Exploded view of a three tab asphalt shingles constitutive materials. ................ 25 2 4 ... 28 2 5 Plan view of typical laminate shingle with six fastener pattern shown. ............... 34 2 6 Exploded view of typical laminate shingle constitutive materials. ....................... 34 2 7 Diagonal installation of three ta b asphalt shingle system. Laminate installation produces a similar pattern. ................................ ............................... 40 2 8 Vertical (racked) installation of three tab asphalt shingle system. Laminate shingles are not installed with this pattern. ................................ ......................... 41 3 1 Pre wind test asphalt shingle test deck ................................ .............................. 51 3 2 Post wind test asphalt shingle test deck ................................ ............................. 51 3 3 Wind load model proposed by Peterka et al. (1983). ................................ .......... 52 3 4 Test setups for w ind testing of asphalt shingles ................................ ................. 56 3 5 Peak pressure coefficients measured on a full scale asphalt shingle subjected to wind flow from varying dir ections. ................................ ................... 60 4 1 Locations of the asphalt shingle surveys conducted in Florida. .......................... 69 4 2 Location of partial unsealing. ................................ ................................ .............. 74 4 3 Location of partially/fully unsealed three tab and laminate shingles (tape marks). ................................ ................................ ................................ ............... 74 4 4 Shingle roofs located in Houston, TX with partially unsealed shingles located by triangular chalk marks and fully sealed shingles located by dash marks ...... 75 4 5 Percent of unsealed shingle strips located in the field of the roof verses roof age. ................................ ................................ ................................ .................... 76 4 6 Boxplot of unsealed shingle strips located in the field of the roof verses roof age at the time of investigation. ................................ ................................ .......... 77
11 4 7 Blown off three tab asphalt shingles. ................................ ................................ .. 78 4 8 Typical condition for partially unsealed ridge and hip shingle ............................ 79 4 9 Percent of fully and partially unsealed hip and ridge shingles ............................ 79 4 10 Wind directions for gable and hip roof specimens. ................................ ............. 81 4 11 Measured and best fit theoretical normalized mean velocity, longitudinal turbulence intensity, and lateral turbulence intensity. ................................ ......... 83 4 12 Normalized wind spectrum of Wind Level 3 (measured at 5 m). ......................... 84 4 13 Hip roof three tab shingle specimen pre and post test conditions ..................... 87 4 14 Shingle roof damage initiated by pre wind test unsealed shingles. .................... 87 4 15 Statistical comparison of roof damage for the roof specimen shown in Figure 4 13 ................................ ................................ ................................ ................... 88 4 16 Characteristic hip shingle blow off patterns. ................................ ....................... 88 4 17 Progression of hip shingle blow off through the wind test sequence for specimen oriented at the 0 wind direction. ................................ ........................ 89 5 1 Wind pressures on shingle roofing. ................................ ................................ .... 94 5 2 Cross ................................ .... 94 5 3 ASTM D6381 specimens ................................ ................................ ................... 97 5 4 UV Thermal Aging chamber components. ................................ ......................... 99 5 5 Measured irradiance, plan view, and temperature time history of one cycle. ... 100 5 6 ASTM D6381 test apparatus, setup, and uplifted specimen. ............................ 103 5 7 ASTM D6381 test results for Product A. ................................ ........................... 105 5 8 ASTM D6381 test results for Product B. ................................ ........................... 105 5 9 ASTM D6381 test results for Product C. ................................ ........................... 106 5 10 Example failure modes observed in mechanically uplifted shingles. ................ 109 5 11 Distribution of failure modes on Product A Procedure A. ................................ 110 5 12 Distribution of failure modes on Product A Procedure B. ................................ 111
12 5 13 Distribution of failure modes on Product C. ................................ ...................... 11 1 5 14 Portable Mechanical Uplift Apparatus compo nents. ................................ ......... 117 5 15 In situ ASTM D6381 test results. ................................ ................................ ...... 119 5 16 Distribution of failure modes for in situ ASTM D6381 tests. .............................. 120 5 17 Wind resistance of naturally aged shingles vs. ASTM D7158 Class H required resistance. ................................ ................................ .......................... 122 6 1 Wind pressures on shingle roofing. ................................ ................................ .. 126 6 2 Rendering of the Dynamic Flow Simulator componentry. ................................ 132 6 3 Dynamic Flow Simulator, profile view, and test section, orthogonal view. ........ 132 6 4 Dynamic Flow Simulator, as constructed. ................................ ......................... 132 6 5 Cross section of DFS test se ction. ................................ ................................ ... 133 6 6 Test deck below opening in test section. ................................ .......................... 133 6 7 laminate shingle instrumented with load cells is shown. ................................ ... 134 6 8 Mean longitudinal velocity across the width of a shingle test specimen ........... 134 6 9 Plan view of Three Tab test deck. ................................ ................................ .... 137 6 10 Plan view Laminate test deck. ................................ ................................ .......... 138 6 11 Cross section of DFS with test deck. ................................ ................................ 138 6 12 Multi axis load cell elevation and plane view. ................................ ................... 139 6 13 Cross section of load cell attachment detail. ................................ .................... 141 6 14 Part 1: load cell arrangement for Laminate speci men showing load cell base connection. ................................ ................................ ................................ ....... 142 6 15 Part 2: load cell arrangement for Laminate specimen showing wood decking s urrounding top plates. ................................ ................................ ..................... 142 6 16 Three Tab specimen. ................................ ................................ ....................... 143 6 17 Plan vie ws for Three Tab fully sealed and partially unsealed arrangements. ... 144 6 18 Laminate specimen. ................................ ................................ ......................... 144
13 6 19 Plan views for Three Tab fully sealed and partially unsealed arrangements. ... 145 6 20 Fully sealed laminate test specimen installed on the test deck. ....................... 147 6 21 Plan view and cross section of velocity measurement locations. ..................... 148 6 22 Velocity sensor test setup. ................................ ................................ ................ 148 6 23 Mean longitudinal velocity profiles Low wind spe ed. ................................ ..... 150 6 24 Mean longitudinal velocity profiles Medium wind speed. ............................... 150 6 25 Mean longitudinal velocity profiles High wind speed. ................................ ..... 151 6 26 Mean longitudinal velocity of all measurement positions and wind speeds normalized by the 152 mm height mean. ................................ .......................... 152 6 27 Mean longitudinal velocity of all measurement positions and wind speeds. ..... 152 6 28 Phase II experimental setup. ................................ ................................ ............ 155 6 29 Relationship between static pressure and mean force. X and Z axes component force are shown for 416 mm (16.375 in) length of sealant strip. .... 156 6 30 Representative time history plot of z axis force between raw and filtered signals and not affected by temperature. ................................ .......................... 158 6 31 Time history plot of z axis force affected by temperature change on the load ................................ ................................ ................................ ........ 158 6 32 Reference frame for fully sealed Laminate and Three Tab specimens. ........... 160 6 33 Partially unsealed Laminate Specimen 3 lifting in 44 m/s (98 mph) mean wind velocity. ................................ ................................ ................................ ............ 169 6 34 Partially unsealed Laminate Specimen 4 lifting in 44 m/s (98 mph) mean wind velocity. ................................ ................................ ................................ ............ 169 6 35 Reference frame for partially unsealed Laminate and Three Tab specimens showing interfacial force location. ................................ ................................ ..... 170
14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THE WIND RESISTANCE OF ASPHALT ROOFING SHINGLES By Craig Robert Dixon December 2013 Chair: David O. Prevatt Cochair: Forrest J. Masters Major: Civil Engineering Asphalt shingle roofing is the leading cause of hurricane wind related insured losses in residential buildings. Damage statistics generated from recent hurricanes indicate shingle roofs sustain da mage in wind velocities below design level with damage frequency increasing with shingle roof age. The objective of this dissertation is the identification of primary mechanisms triggering the failure of shingle roof systems in wind The research goal is to reduce future shingle roof wind damage and improve our ability to predict asphalt shingle wind resistance. Five studies comprising this dissertation address ed the adhesive consistency and strength of aged asphalt shingles, system level wind resistance, and the load model underpinning the ASTM D7158 wind test standard. The most significant and unexpected finding was partially unsealed shingles on field, hip, and ridge locations on Florida and Texas homes. sealant strip where unsealed and failure mode were consistent at each location Total quantity of partially unsealed shingles in the field of the roof significantly increased with ag e, aligning with damage statistics. Full scale wind tunnel tests demonstrate partially
15 unsealed shingles are more vulnerable than fully sealed due to increased distributed force on sealant strip and concentrated force at the adhered and non adhered interfa ce. Uplift resistance was measured in artificially and naturally aged shingles. For artificially aged shingles, one of three products evaluated had statistically significant decreases in mean uplift resistance as exposure time increased. However, resistan ce was above design level at all exposure test intervals. Naturally aged shingles also had resistance above design level. Combined results demonstrate that reduced uplift capacity can occur, but high initial bond strength promotes long term uplift resistan ce. Wind loads exerted on the shingles sealant strip load path were directly measured on fully sealed and partially unsealed three tab and laminate shingles. Results indicate that ASTM D7158 and load model is conservative in force prediction for fully sea led shingles. ASTM D7158 is not conservative for partially unsealed shingles. Research concludes that partially unsealed shingles occur naturally and represent a large contributor to wind damage. Retrofit of existi ng shingle roofs and further work identifying specific cause will provide significant reduction of wind risk in shingle roofing.
16 CHAPTER 1 INTRODUCTION Over the past 50 year s hurricane induced economic losses in the United States (US) have increased from an estimated average annual total loss adjusted for 2006 $ U S of $1.3 billion in 1949 1989, $10.1 billion in 1990 1995 to $35.8 billion in 2001 2006 ( National Science Board 2007 ). The greatest impact from US landfalling hurricanes is felt along the Atlantic and Gulf coastal communities where approximately one third of the US population resides within 100 miles of the coast line ( US Census Bureau 2007) and over $9 trillion of insured property exists ( Liu et al. 2010 ) In Florida alone, insurers lost a n estimated $9.3 billion in 2004 (Charley, Frances, Ivan, and Jeanne) and $3.8 billion in 2005 (Dennis and Wilma) ( Malmstadt et al. 2009 ) Historically, building failures have played a large role in economic losses resulting from hurricanes. For example, in Hurricane Hugo (1989), r esidential building damage accounted for 58% of the $ 5 1 7 billion in total insured losses ( 2012 $ US ) ( Sparks et al. 1994 ) In turn, t hese losses have been absorbed by residents in vulnerable coastal areas in the form of hig her property insurance premiums ( Hamid et al. 2011 ). B uilding failures have a larger societal impact, as well. D isplacement from damaged dwellings is often required while the structure is repaired ( Levine et al. 2007 ) debilitating coastal communities that de pend on its residents for economic stability and cultural vibrance ( Levine et al 2007 ). (1992) winds destroyed an estimated 85,000 homes, leaving 180,000 South Florida residents temporarily homeless ( Mitchell et al. 2012 ) More recently over one half of Galveston single family homes were still vacant two months after Hurricane Ike struck the Texas coast in 2008 ( Mitchell et al. 2012) I ncrease d short and long term psychological stress (e.g. post traumatic
17 stress) can also arise as the result of hurricanes and the subsequent damage that is imposed on the community ( Galea et al. 2005 ). F ailed roof co verings and rooftop accessories continue to be the leading cause of building damage in hurricanes ( FEMA 2005b ) (Figure 1 1) Roof covering damage in Hurricane Charley (2004) accounted for approximately 50% of the direct losses to damaged buildings at all wind speed ranges ( Applied Research Associates 2008 ) Their failure can also contribute to a ge due to rainwater entering the building envelope through the breach opened by the failed roofing In Hurricane Charley, interior damage was the second leading cause of direct loss to the buildings surveyed This dissertation is focused on the wind resist ance of asphalt shingles the predomina nt roof covering material for residenti al buildings. Asphalt shingles were first introduced in the late 19 th century ( Cullen 1992 ) and their use has grown to a 2006 US steep slope roofing market share of 89% predominantly in residential construction ( ARMA 2011 ). The w ind resistance of shingle s has been a critical design consideration throughout its history ( Cullen 1992 ). P rior to the mid 1950s, wind induced blow off was the most common premature failure mode of asphalt shingles ( Cullen 1992 ) In response, manufacturers introduced a thermally activated adhesive strip along the shingle s leading edge to prevent the shingle from lifting in the wind a design that is still used today ( Cullen 1992 ). Research cond ucted in the 1980s and 90s discovered that uplift forces develop on the shingle s upper and lower surface s near its windward edges due to lo calized separations of the near roof surface wind flow ( Peterka et al. 1997 ). Despite this improved knowledge of wind load mechanics and subsequent
18 modifications to standardized test methods and materials (e.g., introduction of polymer modification to bitumen chemistry) reports of wind induced failures of asphalt shingles continue ( FEMA 20 05a; FEMA 2005b; FEMA 2006; RICOWI 2006 ). Approximately 6 0% of shingle roofs over five years old were damaged in below design level winds from Hurricanes Ike and Gustav (2008) (Liu et al. 2010), while Hurricane Charley caused an estimated 78% damage rate t o shingle roofs subjected to near design level winds in Knowledge Gaps in the Wind Resistance of Asphalt Shingles The primary challenge towards improving the wind resistance of asphalt shingles is the lack of substantive data that provides manufacturers, engineers, installers, and researchers the root causes of failure. Current test methods for certifying a shingle ngles that are installed in strict accordance with manufacturer specification and tested in a laboratory setting. The effects of aging, installation errors, and roof edge detailing are not addressed within these wind certification tests. These test assum ptions exist despite repeated observations of a potential reduction in wind capacity as the shingle ages ( Liu et al. 2010 ; Gurley and Masters 2011 ), failures along roof edge det ails ( FEMA 2009 ), and installation techniques that may adversely influence wind performance ( FEMA 2009 ; RICOWI 2006; RICOWI 2007 ). Post hurricane damage assessment reports ( FEMA 2005a; FEMA 2005b; Liu et al. 2010; RICOW I 2006; RICOWI 2007 nts can be useful to manufacturers and building officials in identifying common damage patterns in order to modify current material design and installation specifications. However, these reports are limited in
19 their capacity to identif y the specific cause of failure as the investigations are conducted after failure has occurred on roofs where the pre storm condition of the roof is unknown. sealant strips before the storm can influen wi nd performance Moreover, t he source of shingle damage is inferred by field investigators using forensic investigative methods and previous experience on failed roofs. This complicates the issue further, as it increases the potential for identifying the wrong mode of failure. As will be shown later, over the past 25 years, post hurricane reports have repeatedly shown common damage patterns with commonly associated failure modes. The rate of damage incurred combined with the consistency of damage suffered by asphalt shingle roofs indicates the need to identify why shingle roofs continue to fail in wind that is below their design level despite the presence of standardized test methods and post damage analysis available to all stakeholders in asphalt shingle roofing. Research Goals and Scope The overarching goal of this experimental research is to improve the wind resistance of asphalt shingles through identification of the root causes of continued asphalt shingle failure s in below design level hurricanes This new research builds upon the existing knowledge of the wind load model for asphalt shingles and synthesizes the results of laboratory and field research to fill the critical knowledge gap between laboratory testing m ethods and the in situ wind performance of shingle roofing. More specifically, this research advances our understanding of the impacts that aging and system level components have on the wind load model that functions as the basis for the methods currently available to assess the wind resistance of asphalt shingle roofing
20 The shingle roof covering is a discontinuous roof system with each roof comprised of several thousand water resistant sheets that are designed to act as an integrated system to resist win d forces in order to maintain a watertight roof envelope. Yet wind performance for asphalt shingles is rated by the mean resistance of system components in isolation, rather than as a holistic unit ignoring the effects of load sharing between constituent s Previous post hurricane damage reports ( Liu et al. 2010; Gurley and Masters 2011 ) have observed that new er asphalt shingle roofs perform better in hurricane winds than older. However th ey fail to delineate if the primary cause of this performance gap is due to recent improvements in product design, building code r equirements and test methods or if it is the result of age related degradation in older shingle systems. Unfortunately t he current methods for rating wind resistance cannot predict the perfo rmance of a shingle once it has aged Furthermore, throughout the development of the wind load model, measurements of wind uplift forces directed through the main wind load path the sealant strip were approximated from point measurements on t he shingles surface, rather than directly measured through the load path. This measurement technique is also utilized for the ASTM D7158 (ASTM 2011c) wind test. It is hypothesized that these assumptions confined within the current wind load model and stand ardized wind test methods may adversely affect the ability for all stak eholders to design, manufacture and install asphalt shingle products that are wind resistant throughout their service life. The new research presented in this dissertation comprised of five studies First, field surveys were performed on thirty homes in Florida and Texas to identify whether asphalt shingle remain adhered along their sealant strip throughout service. Second,
21 full scale wind tunnel tests were conducted at the IBHS Resea rch Center on seventeen asphalt shingle roof system s The objective was to evaluate the wind resistance of the system as a holistic unit to identify weaknesses in design and installation not captured by standard test methods. Third, three asphalt shingle p roducts were artificially aged using two techniques that accelerate weathering. Wind resistance was measured on a portion of the shingles at discrete intervals and compared to initial resistance and ASTM D7158 design level requirements (ASTM 2011c) Fourth four naturally aged shingle roof systems were tested for wind uplift resistance in situ with results compared to ASTM D7158 design level requirements (ASTM 2011c) Finally, in and out of plane wind loads were measured on sealant strips of three tab and asphalt shingles. Specimens were fully sealed and partially unsealed. Measured forces were compared to forces predicted by ASTM D7158 (ASTM 2011c) and the theoretical model for wind loads on asphalt shingles (Peterka et al. 1997) Research Impacts At its foundation, t he results from this research impact a broad group of stakeholders that will rely on this information to develop better asphalt shingle products and installation methods, refine the understanding of risk to existing homes, and reduce the econo mic and societal damage incurred from shingle roof failures during hurricanes. To achieve the greatest impact, twenty three experts who represented the inter ests of product manufacturers, wind e ngineers, building officials, governmental agencies, risk anal yzers, insurers, and public interest groups oversaw the research (Table 1 1) The oversight committee provided an external peer review of all experiments during the pl anning and results disseminating e xperiments would provide relevant and impactful results to these critical groups, while,
22 at the same time, fostering by in from the committee due to their external input on experiment topics methods and results reporting. Table 1 1. Oversight committee members Name Affiliation Peter Vickery, Ph.D., P.E. Applied Research Associates Bill Coulbourne, P.E. Applied Technology Council Michael Fischer Asphalt Roofing Manufacturers Association Jon Peterka, Ph.D., P.E. Cermak Peterka Petersen, Inc. John Minor, CGC (FL) Complete General Contractors, Inc. Leslie Chapman Henderson Tim Smail Federal Alliance for Safe Homes John Plisich Andrew Herseth, P.E., S.E. Thomas Smith, A.I.A., R.R.C. a FEMA Rick Dixon Mo Madani Florida Building Commission Miles Anderson Florida Division of Emergency Management Jack Glenn Florida Home Builders Association Anne Cope, Ph.D., P.E. Tim Reinhold, Ph.D., P.E. Insurance Institute of Business & Home Safety Michael Young, P.E. Peter Datin, Ph.D. Risk Management Solution s, Inc. (RMS) Walter Rossiter, Ph.D. RCI (formally the Roof Consultants Institute) Jim Baker Roofing Industry Committee on Weather Issues (RICOWI) Benjamin Thomas Jr., Ph.D. Southeast Region Research Initiative Tom Nichols U.S. Polyco Inc. Julie Serakos BMS Intermediaries, Inc. a Mr. Smith also works as a consultant for TLSmith Consulting Inc. Layout of the Dissertation This dissertation consists of seven chapters. Chapter 1 contains the background and introduction to the research, knowledge gaps in the wind resistance of asphalt shingles, research goals and scope, and research impacts. Chapter 2 provides an
23 overview of the asphalt shingle roof system and its installation. A literature review of the history of the wind resistance of asphalt shingl es is presented in Chapter 3 This is followed by the results of two studies addressing unsealed shingles in Chapter 4 two studies addressing the wind resistance of aged shingle roofing in Chapter 5 and an assessment of the wind load model for asphalt shingles in Chapter 6 The three research chapters (4 6 ) each contain an introduction to the specific research topic, experimental methods, results, and significant findings Chapter 7 concludes the dissertation with a discussion of the combined results of the experiments, recommendations for retrofit solutions, and recommendations for future work.
24 CHAPTER 2 ASPHALT SHINGLE COMP OSITION AND INSTALLATION The primary purpose of the asphalt shingle roof is to waterproof the roof surface installed as individual pieces in the field of the roof over the existing s tructural roof decking and lapped to provide a path for water transpo rt down the roof slope (Figure 2 1 ). Additional shingles are installed over the hip and ridge lines of the roof to prevent moisture ingress at ro of slope intersections (Figure 2 1). As de tailed in Chapter 3 the composition and planform of shingles has varied throughout their history; however, the three tab and laminate shingle styles evaluated in this dissertation and described below have dominated the shingle roofing market over the last 30 years ( Cash 1995 ; Malarkey 2001 ). Figure 2 1 Asphalt shingles are installed in the field of the roof with additional shingles along hip and ridge lines. Shingle Composition The asphalt shingle is a composite material composed of an asphalt impr egnated reinforcement mat, granular rock surface coating, and a thin strip of adhesive know n as
25 the sealant strip (Figure 2 2 ). In the manufacturing of shingles, the reinforcement mat is constructed first, and then passed through a coating machine where ho t modified distributed over the top asphalt layer and embedded by physically pressing the granules into the asphalt. Finally, the sealant strip is applied over the granule sur facing as a continuous or discontinuous strip of adhesive, and the shingle material is cut to form the individual shingle strip that is installed on the roof ( Noone and Blanchard 1993 ). Three tab shingles contain one laye r of shingle composite (Figure 2 3 ), while laminates contain two layers that are bonded using an asphalt based adhesive (Figure 2 6 ). F igure 2 2 Plan view of a standard three tab shingle with six fastener locations shown. Figu re 2 3 Exploded view of a three tab asphalt shingles constitutive materials.
26 Reinforcement M at pull through, and crack penetration ( Noone and Blanchard 1993 ). Fiberglass is the most popular mat material for modern shingles, repr esenting a 95% market share in 2001 ( Malarkey 2001 ). Given this overwhelming market dominance, all shingle products used in this dissertation contain fiberglass mats. The glass fiber used in the mat is made from dispersions of surface treated glass with an approximate fiber length of 1 in by 14 to 16 microns diameter ( Noone and Blanchard 1993 ). The mat is constructed in a wet process where glass fibers are randomly oriented (i.e. non woven) and bound together by resin typically a modified urea formaldehyd e then cured by heat ( Noone and Blanchard 1993 overall weight ( Noone and Blanchard 1993 ), mat weight can be used as a predictor of service stresses (e.g., tear resistance) ( Cash 1995 ). Asphalt C oating The coating applied on the upper and lower surfaces of the reinforcing mat is a composite of a modified asphalt matrix and mineral filler reinforcement ( Noone and Blanchard 1993 ). Virgin asphalt is first extr acted as a by product from crude oil as it is distilled into gasoline ( Berdahl et al. 2008 ). The asphalt is then oxidized into its final state by blowing air into the asphalt; thereby increasing its stiffness and softening point (220 degrees Fahrenheit fin al versus 120 degrees F virgin) to make the material stable for roofing applications (Malarkey 2001 ). Finally, mineral filler most frequently limestone dust is blended into the modified asphalt to produce the final coating material. The filler possesse s a higher density and lower thermal expansion coefficient than the modified asphalt ( Noone and Blanchard 1993 ). Therefore, in the final coating,
27 the modified asphalt provides water resistance, while the mineral filler increases the mixtures resistance to thermal and mechanical stresses by composite action between the asphalt and filler ( Noone and Blanchard 1993 ). Roofing G ranules The primary purpose of the top surface roofing granules is to prevent photo oxidation of the top asphalt coating ( Berdahl et al. 2008 ). Photo oxidation occurs when incident ultraviolet (UV) light waves are absorbed by bare asphalt and energetic UV oxygen groups, including sulfate (SO 4 ). The formation of these solu ble groups makes the top layer susceptible to removal from the roofing during rain events, thereby, exposing a new layer of asphalt for the process to begin anew ( Dutt 1986 ). Physical properties of the asphalt also change as a result of this altered chemic al state due to increased Robertson 1991 ). The result is a stiffer asphalt material that may be more susceptible to cracking or crazing ( Berdahl et al. 2008). Roofing granules placed over the asphalt prevent photochemical degradation by absorbing incident UV before it can penetrate into the asphalt ( Dutt 1986). Typical granule materials include: slate, blast furnace slag, or crushed stone (Dutt 1986). Granule size generally ranges from 0.15 mm (0.006 in) to 3. 3 mm (0.13 in) with the relative contribution of each particle size over the shingle surface designated by the face; however weathering can cause loss of granules and exposure of asphalt (Dutt 1986). While standard tests for granule coverage and long term adhesion do not exist, ASTM D4977 03(2009)
28 Standard Test Method for Granule Adhesion to Mineral Surfaced Roof ing by Abrasion (ASTM 2008a ) does provide a standard test method for manufacturers to evaluate the granular adhesion of new shingle products. Sealant S trip based adhesive that below (Figure 2 4 ). Its purpose is to: 1. Reduce the aerodynamic wind forces that are produced on the shingles upper and lower surfaces by restricting the cross sectional geometry to an aerodynamically efficient shape. 2. Transfer wind forces produced on the shingles surfaces to the shingle course below. The sealant strip was introduced in the mid 1950s in an effort to impro ve the wind resistance of asphalt shingles (Cullen 1992) and is still used today as the shingles primary wind resistance method. Fi gure 2 4
29 Sealant strip design is control led by shingle system wind performance, rather than material composition specifications. Therefore, sealant strip materials and designs vary between manufacturers, even within different product lines and manufacturing locations. The sealant can be placed o s surface, as shown in Figures 2 3, 2 4, and 2 6 Recall, shingles are lapped between courses when they are installed; therefore, a sealant strip manufactured on the top sur face of the shingle will seal the leading edge of the shingle course above, while application of the sealant strip on the bottom surface mm (0.5 in) and the strip may be applied as a continuous or discontinuous strip. manufacturer; however, as shown later, this distance influences the magnitude of wind uplift forces that are generated on th As with the other design parameters, the material composition of the sealant strip varies widely amongst manufactures (Nichols 2010 ). The overarching goal of the strip material is to maintain adhesion between the two shingle courses w hen subjected to in service stresses (e.g., wind uplift and thermal movement) throughout the lifetime of the have a softening point at a temperature that can be achieve d on a roof during a sunny day (Nichols 2010). Despite this ambiguity in performance standard, it is expected that the softening point (i.e., activation) of the sealant should occur at a minimum temperature of 57 C (135 F), as this is the minimum required conditioning temperature set forth in the ASTM D3161 (ASTM 2013) ASTM D6381 (ASTM 2008 b ) ASTM D7158
30 ( ASTM 2011c ) and UL 2390 (UL 2003) wind test standards. Wind induced blow off failures of asphalt shingles with non activated sealant strips due low roof temperature installation have been reported (Fronapfel 2006 ; Nichols 2010), and manufacturers include statements on shingle packaging stipulating that installers hand apply roofing cement under the leading edges of s hingles installed in winter months. Re sin based adhesives were used in the first sealant strip materials and are still used for sealant strips in some modern shingle products ( Nichols 2010). Resin is derived from the extraction of asphalt and represents a less valuable material than the asphal t itself (Nichols 2010). Additives, such as fillers tackifiers, and rubber compounds are frequently blended with resins in order to meet bond strength performanc e specifications [e.g., ASTM D6381 (ASTM 2008 b )] and to promote early tack and long term durab ility ( Nichols 2010). However, resins and their related blended formulations are known for their brittle nature, especially at low temperatures, making them susceptible to fracture under in service dimensional stresses (Nichols 2010). Polymer modification was introduced in 1843 and gained popularity in the US paving industry in the mid 1980s (Yildirim 2007). Today polymer modified asphalts adhesives have emerged as an alternative to the traditional resin based sealant strips (Nichols 2010). The process of manufacturing polymer modified asphalts involves the separation of asphalt into fractions (i.e., hydrocarbon groups) then combining ideal fractions with elastomeric polymers and other additives (Nichols 2010). Polymer modifiers include: styrene butadinene styrene (SBS) block copolymer, styrene butadinene rubber (SBR), and crumb rubber modifier (Yildirim 2007). Within the paving industry, the polymer modified asphalts are known for their increased long term
31 resistance to fatigue, thermal cracking, rutting, stripping, and temperature susceptibility when compared to unmodified asphalts (Yildirim 2007). Fasteners Fasteners are placed in a row perpendicular to the roof slope to secure each shingle strip to the roofing substrate below The type, location, and quantity of fasteners are either specified by local building codes (e.g., Florida Building Code) or product manufacturers. For example, the 2010 F lorida Building Code section R905 .2.5 (Florida Building Commission 2010) requires th at fasteners should consist of: Galvanized steel, aluminum or copper roofing nails, minimum 12 gage [0.105 in (3 mm)] shank with minimum 3/8 in (10 mm) diameter head, ASTM F 1667, of a length to penetrate through the roofing materials and a minimum of in (19 mm) into the roof sheathing. Where the roof sheathing is less than in (19 mm) thick, the fasteners shall penetrate through the sheathing. Florida Building C ode also requires a minimum of four fasteners per strip, and, where the structure is within the High Velocity Hurricane Zone (Broward and Miami Dade, FL counties), a minimum of six nails are required ( Florida Building Commission 2010 ). Fasteners are either placed by hand or pneumatic air pressure guns. Underlayment The main objectives of the und erlayment are: 1) serve as a primary water barrier during re roofing applications and 2) provide a secondary water barrier to the primary asphalt shingle waterproofing during normal in service use (Schaack 2006a). Underlayment is a general term describing an application and purpose; therefore, several products exist with varying degrees of rated water and wind resistance (Schaack 2006a). The most popular underlayments are asphalt saturated felt paper and
32 self adhered polymer modified bitumen sheets Saturat ed felt paper is produced in two grades (Schaack 2006a): 1. [ ASTM D226 Type I ASTM D 4869 Type I ( ASTM 2011a )] 2. ASTM D226 Type II (ASTM 2009)] Duty Shingle Underlaym II ( ASTM 2011a )] In general, the self adhered modified bitumen will provide the best protection from wind and moisture, followed by the Type II and Type I felt papers, respectively. Underlayment is packaged as rolls with a 914 mm (36 in) width and installed parallel with eave edge of the roof. As with asphalt shingles, underlayment are installed with laps on their head region (i.e., lap parallel to the eave) and end region (i.e., lap perpendicular t o the eave) to provide a path for moisture transport and fasteners secure the sheets to the roof substrate (Schaack 2006b). Local building codes control what underlayment products are allowed and installation methods, including: minimum lap distances (head and end), fastener types and fastener patterns (Schaack 2006b). The 2010 Florida Building Code: Residential (Florida Building Commission 2010) requires that all felt pape r underlayment be installed with 483 mm (19 in) head laps (i.e., double layer) for roofs with a slope of 2 units vertical in 12 units horizontal, up to 4 units vertical in 12 units horizontal. When roofs have a slope greater than 4 units vertical in 12 units horizontal, a single layer of underlayment with a minimum 51 mm (2 in) head lap must be installed. Fasteners for both roof slope conditions are a required 914 mm (36 in) on center along the head lap sections. When the residential building is located in a High Velocity Hurricane Zone, all roofs should have a 483 mm (19 in) head lap for Type I felt paper and 102 mm (4 in) head lap for Type II. Fastener density also
33 increases within the high wind regions with a maximum spacing of 305 mm (12 in) throughout the sheet and 152 mm (6 in) along the end and head laps of each sheet. Three tab S h ingles Three tab shingles were among the first shingle designs introduced in the early 20 th century ( Abraham 1920 ). The planform and dimensions of a common three tab shingle strip is shown above in Figure 2 2 These shingles derive their name from the two vertical (i.e., up roof slope) cuts that are made into each shingle strip to produce the appearance of three individual tabs per strip when installed. The planform dimensions of three tab shingles vary slightly between product models and manufacturers. How ever, each strip typically has overall dimensions of 0.3 m (1 ft) (upslope) by 0.91 m (3 ft) (along slope) with 6 mm (0.25 in) wide cuts in the strip on the third points of the along slope dimension; forming three 0.3 m (1 ft) wide tabs. The length of cut, shown in Figure 1 3 as 140 mm (5.5 in), is determined by the manufacturer and represents the total exposed length of the installed shingle. The remaining portion of the shingle above the cut is covered by the shingle on the course above. The sealant strip on a three tab is most frequently placed on the upper surface of t he shingle, as shown in Figure 2 3 Laminate S hingles Laminate shingles, also known as architectural shingles, wer e introduced in the 1970s (Cash 1995). Unlike the three tab design, laminate products do not contain cuts in their strips, rather, contrast is created by adhering a second layer of shingle composite to mimic the appearance of slate roofing tile s (Figures 2 4 and 2 6 ). The planform of la minates is generally larger than three tabs with dimensions of 0.33 m (1 ft 1 in) by 0.99 m (3 ft 3 in). Laminate shingle fastener placement and the location of the sealant along the top or bottom of the strip are similar to three tab shingles.
34 Fi gure 2 5 Plan view of typical laminate shingle with six fastener pattern shown. Asphalt Shingle I nstallation Installing a new shingle roof on a standard single family home can require up to 4000 individual shingle strips, hip/ridge cap shingles, underlayment edge details, and penetration details. Fig ure 2 6 Exploded view of typical laminate shingle constitutive materials. One advantage to shingle roof system is the discontinuous nature of the installation process a single person can install the roof w ithout the assistance of additional manpower or lifting equipment. This may also be seen as a disadvantage due to ease of
35 materials handling, which, can encourage the homeowner or unqualified contracted laborer to perform the roof installation. The compone nts of a shingle roof are designed perimeter, and resist wind forces exerted on the shingle roof covering. As shown in Chapter 3 the wind induced failure of a single comp onent can initiate a progressive system Installation quality, or the lack thereof, is frequently cited as a cause of win d induced shingle failure (FEMA 2005a; FEMA 2005b; FEMA 2009 ; RICOWI 2006; RICOWI 2007); therefore, the goal of this section is to introduce the methods and requirements for asphalt shingle product selection and installation. Selecting an asphalt shingle product can be based on a combination of pricing, contra ctor preference, appearance, minimum material properties, or wind and/or hail impact requirements stipulated by local building codes. As shown by the Tex as Department of Windstorm Insurance (2011 ) survey of manufactured shingles 201 different shingle prod ucts are currently available to homeowners all with varying degrees of material and performance certifications or ratings. For hurricane prone regions of the Southeast US, the wind resistance rating of the shingle is the primary factor for approved produc ts. In Florida, shingles approved for installation on residential structures must be rated for their wind resistance by one of three test methods: ASTM D7158 ( ASTM 2011c ) ASTM D3161 (ASTM 2013) or TAS 107 (Florida Building Commission 2010) The specific test methods, limitations, and research leading to each test method are detailed in Chapter 3 The wind rating requirements outlined in Florida Building Code section R905.2.6.1 (Florida Building Commission 2010) for asphalt
36 shingles installed Florida resid ential structures are given i n Table 2 1 with reference to the Basic Wind Speed where the structure is located. specifies installation guidelines with additional installation requirements dictated by local building codes. New shi ngle roof installations performed in states, counties, or cities located in high wind zones, such as Florida, can require licensed installers and permit submittals that detail the proposed work and certification of approved products for use. The following paragraphs of this section detail a standard new asphalt shingle roof system installation for three tab and laminate shingles developed from guidelines produced by the Asphalt Roofing Manufacturers Association (ARMA) 2006 Asphalt Roofing Residential Manual Manual (CertainTeed 2012 ) with additional shingle installation requirements stipulated in the 2010 Florida Building Code: Residential (FCBR) for both High Velocity Hurricane Zones (HVAZ) and non HVAZ ar eas (Florida Building Commission 2010). Table 2 1. Wind classification required for asphalt shingle installed in Florida Classification of Asphalt Shingles Approved for Use Maximum Basic Wind Speed, V ult Standard Wind Test Method ASTM D7158 ASTM D3161 / TAS 107 110 D, G, H A, D or F 116 D, G, H A, D or F 129 G or H A, D or F 142 G or H F 155 G or H F 168 H F 181 H F 194 H F This installation description assumes that a new asphalt shingle roof system will be installed on a site built single family home with a roof slope greater than 4 units vertical in 12 units horizontal. Roofs with slopes less than this require additional me asures to
37 ensure water resistance, such as increased requirements for secondary water barrier protection. For new construction, the installation process begins with the application of underlayment over the new structural decking, following the materials a nd installation methods detailed in the Underlayment section above. For reroofing applications, the homeowner can choose to tear the existing roof off or leave the existing roof on and roof over. However, there are regulations within the 2010 Florida B uild ing Code Existing Building Section 611.3 (Florida Building Commission 2010) and the ARMA (2006) manual stipulating when a roof over may be allowed. Given previous post hurricane observations of failed roof overs due to uneven substrates causing poor sealan t strip adhesion and lack fastener penetration into th e structural roof decking (FEMA 2005b ; RICOWI, 2006), it appears that removing the existing roof covering first would provide better wind resistance than a roof over. Therefore, it will be assumed that new and reroof applications begin with the underlayment installed directly over the structural deck. Edge metal (i.e., drip edge) is first installed along the rake and eave edges of shingle roofs is required by the 2010 FBCR (Florida Building Commission 2 010) and ARMA (2006) manual with the metal installed over the underlayment along the rake edge and installed over or under the underlayment on the eave edge. Edge metal is an L shaped flashing element aluminum, galvanized steel, or another approved mater ial in 2010 FBCR Table R903.2.1 (Florida Building Commission 2010) that provides a path for moisture to run off of the roof edge. One leg is attached to the wood sheathing using mechanically fastened roofing nails 12 gauge or greater at a spacing of 305
38 mm (12 in) to 102 mm (4 in) on center, contingent on the wind zone where the structure is located. The other leg of the edge metal projects downward. If the underlayment installed over the edge metal on the eave, a continuous strip of 102 mm (4 in) wid e asphalt roof cement is required over the edge metal along the eave. When the structure is located in an HVAZ, a continuous 204 mm (8 in) wide strip of asphalt roof cement is required along all rake and eave edges under the shingle strips. Individual shi ngles are then installed over the underlayment and edge metal following one of two procedures. The first procedure is the diagonal installation method visually depicted in Figure 2 7 This is the preferred method for all shingle installations as it produce s a reliable method for fastening the complete shingle strip on all locations in the field of the roof ( ARMA 2006 ). The second method is the vertical (i.e., racked) installation procedure shown in Figure 2 8 This method is not covered and not recommended in the ARMA (2006) manual and is not allowed for laminate shingle products. However, the vertical method was used extensively on three tab shingles in the hurricane prone Southeast US throughout the 1980s and 90s and is a frequently citied cause of wind re lated shingle failure (FEMA 2005b; FEMA 2006; RICOWI 2007). Three tab product guidelines produced by CertainTeed (2012) present a vertical installation method; therefore, the vertical installation method will be detailed here as well. No matter the instal lation method, starter strips are req ui red along all eaves (Figures 2 7 and 2 8 ). The starter strips can be purchased as a separate item or created from a three tab shingle with the tabs cut from the strip and the sealant strip placed closest to the eave. Shingles along eaves and rakes have an overhang ranging from 6
39 mm (0.25 in) (for HVAZ located residential structures) to 19 mm (0.75 in) ( ARMA 2006 ) to provide an additional flashing for water runoff. Starter strip are fastened 38.1 mm (1.5 in) to 76 mm (3 in) from the eave edge (ARMA 2006) following a four or six nail pattern, The diagonal installation method begins at the corner of the roof at the intersection of the rake and eave. The ins tallation methods for laminate and three tab shingles are similar with the main difference occurring on the amount of offset placed between shingle courses (i.e., rows). The first course starts at the corner with a full width shingle aligned with the edges of the corner starter strip. The number and type of fasteners used in each shingle strip is stipulated by either the manufacturer or local building code. The installation proceeds along the eave installing full width shingles across the entire eave until terminating either at the far gable end of the building or at a valley or roof projection. To start the next course of shingles, the first shingle is cut to create an offset and installed on the rake. For three tab shingles, one half tab width is cut from the far left portion of the leftmost tab. Laminate shingles have more options; ARMA 2006 ). Once offset, full width shingles are used for the remainder of the course. The distance between the leading edge of the first course and the leading edge of the second course (i.e., exposure) is stipulated by the product manufacturer. The next course proceeds with another cut to the rake shingle; for three tab, it is the complete left tab of the str ip and for laminate it is twice the offset distance. The installation proceeds along eave and u p the rake, as shown in Figure 2 7 until the roof slope is completely shingled.
40 For the vertical installation, the first course begins with one full width thre e tab strip. Then, in contrast to the diagonal method, the installation proceeds vertically to the next course with a strip that is cut one half tab width on the left side leftmost tab. Next, a full width three tab shingle strip is fastened to the roof dec king. This proceeds vertically along the rake end switching between full width shingles and shingles with one half tab termination point, the installation proceeds to right a nd continues vertically for one full width shingle t he n again to the right (Figure 2 8 ). As with the diagonal, this continues until the roof slope is completely shingled. Fi gure 2 7 Diagonal installation of three tab asphalt shingle system. Laminate i nstallation produces a similar pattern.
41 Fig ure 2 8 Vertical (racked) installation of three tab asphalt shingle system. Laminate shingles are not installed with this pattern. The placement of the corner fasteners in the shingle strip are often missed by installers using the vertical installation method. As the installation proceeds across the eave, shingles on alternating courses are hidden by shingles already installed on the course above. Post hurricane reports have noted roofs with missing corner na ils that were installed by the vertical method and subsequently failed due to th e missing corner fastener (FEMA 2005a; FEMA 2005b ; RICOWI 2006; RICOWI 2007). This is also the primary reason for ARMA ( 2006 ) not recommending the vertical installation method for three tab shingles. over their roof lines (Figure 2 1 ). The hip/ridge shingle may be purchased as a pre manufactured hip/ridge shingle or may be created by cutting a three tab shingle vertically at the third points of
42 the shingle; thereby, extending the vertical cut already placed in the shingle ( ARMA 2006 ). Both the pre manfacturered and cut three tab hip/ridge shingles are flat materials that are folded over the hip or ridge line with the centerline of the shingle aligning with the centerline of the hip or ridge line (Figure 2 1 ). The hip/ridge shingle seal ant strip seals down the leading edge of the shingle perpendicular to the hip or ridge line with the secured to the structural roof decking with two nails, each driven 140 mm (5.5 in) from the shingles leading edge and 25 mm (1 in) from the edge of the hip/ridge shingle parallel to the hip or ridge line. Hip/ridge shingle failures are frequently observed in post hurricance assessments as a result of the folding technique that p roduces poor adhesion between the shingle and sealant strip near the edges of the shingle (FEMA 2009).
43 CHAPTER 3 LITERATURE REVIEW The objectives of this review: 1. Detail the research that has shaped our understanding of wind loads on asphalt shingles. 2. Present the standardized wind test methods that drive the design and installation The chapter is organized chronologically into four time periods: 1. The Early Years (1983 1950) 2. Development of the First Test Standards for Wind Resistance (1950 1980) 3. The Development of the Asphalt Shingle Wind Uplift Model (1980 1997) 4. The Modern Era (1997 ) Shingle wind performance issues are addressed throughout, with emphasis on the behavior of in service shingle systems during hurricanes p resented in the final section of the review. The Early Years (1893 1950) The first steep slope asphalt roofing system was introduced in 1893. Known as asphalt prepared roofing, it consisted of a thin reinforcing cotton rag felt impregnated with asphalt (Abraham 1920) 1897, top surface mineral granules were added to improve the durability of the product (Cullen 1992) It was later discovered that the mineral granules served an important functio n of absorbing the ultraviolet (UV) light from the sun. UV oxidizes the top surface asphalt coating and leads to an accelerated degradation of asphalt coating (Berdahl et al. 2008) Reprinted with permission from Dixon, C.R., Masters, F.J., Prevatt, D.O. and Gurley, K.R. (2012 RCI Interface (5), 4 14.
44 Asphalt shingles were introduced in the beginning of the 20 th century. Shingles were manufactured by cutting asphalt prepared roo fing into smaller sections in order to create a discontinuous roof covering resembling wood shakes or slate. Similar to tab style, single tab shingles [ typically 228 mm by 406 mm ( 9 in by 16 in) ] and multi tab shingles [ typically 254 mm by 81 6 mm ( 10 in by 32 1/8 in) ] were produced. The individual tab shingles had exposed leading edges that were often designed with interlocking mechanisms to hold the shingles down during wind storms. The multi tab styles had unrestrained leading edges, allowin g the shingle tab to lift in the wind (Abraham 1920; Cash 1995) By the late 1920s, the cotton reinforcing felts were replaced with substitute materials due to a price increase in cotton rags. In 1926, the Asphalt Shingle and Research Institute and the National Bureau of Standards (NBS; now the National Institute of Sta ndards and Technology) jointly investigated the effects of weathering on the newer substitute felts. Results of the three year weathering study showed no adverse aging effects on the shingles containing the substitute materials (Cullen 1992 ) The use of asphalt shingles increased as a result of World War II construction demands, prompting another change of reinforcing felt to a less expensive wood based organic material. Greenfeld (1969) reinforced asph alt shingles performed as well as their predecessors. Blake (1925) developed one of the earliest known shingle attachment schedules for a four tab strip shingle; which called for five, 19 mm ( 3/4 in ) long galvanized clout nails to be placed one half inch a bove each cutout. T he specified fasteners are similar
45 to earlier prepared asphalt roll roofing products. Single and multi tab shingles were installed on the roof in a stair (Abraham 1920) By 1941, three tab stri p shingles came into the market with fastening requirements of four 11 or 12 gauge galvanized nails per shingle. Snoke (1941) notes that three tab shingles with six nails would be more resistant in high wind prone area; a statement that is echoed in today attachment called for six galvanized roofing nails with a minimum 10 mm ( 3/8 in ) diameter head at 25 mm ( 1 in ) from the shingle edge and 38 mm ( 1.5 in ) on either side (Strahan 1947) The most likely premature shingle failure during this era was due to wind, signaling that attachment requirements were inadequate. During moderate wind, continued flexing of the non restrained exposed shingle tabs weakened the nailed connec tion, increasing the vulnerability of the shingle to blow off in strong wind gusts (Cullen 1992) Development of the First Test Standards for Wind Resistance (1950 1 980 ) The 1950s saw the introduction of two tab sealing methods in order to improve perfo rmance under wind loading (Cullen 1960) The first method consisted of a during the manufacturing process. Early tab seals were typically resin based materials, which ar e asphalt byproducts that have a sudden softening point that would adhere the leading edge of the shingle tab to the shingle below. Early formulations of resin based tab seals were susceptible to brittle fracture failures as a result of thermal fluctuation s. Today, tab sealants consist of either limestone or fly ash modified resins sealant or polymer modified bitumen (Nichols 2010) The second method consisted of a field application of asphalt roof cement dollops along the underside of each shingle tab
46 (Cul len 1960) This method is still recommended for steep slope roofs with a slope greater than 60 degrees and for repairs to shingle tabs that have lost adherence of their self seal systems. In the early 1950s, a letter survey was sent to military installati ons along the US East Coast to ascertain the performance of asphalt shingles installed on their buildings. The results were poor; 67% of those surveyed noted that wind damage had occurred with their shingles. The survey results, coupled with increasing ins urance claims on wind damaged shingles, prompted the first investigation on asphalt shingle wind performance in 1955, which was led by NBS. The goal of the investigation was to assess the wind resistance of organic reinforced asphalt shingles through labor atory, simulated service and field performance evaluations (Cullen 1960) There is some evidence that manufacturers were testing asphalt shingle wind performance already prior to this NBS study (Cullen 1960) However, this is the first published study of this kind. The major component of this investigation was the laboratory simulated wind tests of asphalt shingle test decks. There were two goals to this test: 1. Evaluation of the performance of free tab shingles (i.e. unrestrained shingle tabs) and its corr elation to the weight of the strip shingle. 2. Evaluation of the performance of restrained shingle tabs by either self seal or asphalt roof cement methods. At the time, free tab systems were losing popularity to the restrained tab systems. However, given amount of building stock still using the free tab products, it was important to understand how the weight of the shingle affected performance. This would also be useful for later studies on restrain ed tabs that have lost their adhesive bonds. The laboratory tests consisted of bond strength tests on the tab sealants and
47 wind storm simulation tests. In his report of the wind tests, Cullen (1960) notes that laboratory wind tests fell short of completely simulating in service wind behavior, but that these tests may serve as a useful tool when combined with other methods. The simulation was conducted by using an open jet configuration ; however, n o mention of the flow characteristics of the jet (i.e., magni report. A 1.2 m by 0.9 m ( 4 ft by 3 ft ) test deck with a slope of two on twelve was placed in front of the jet The free tab test decks, nine in total, were subjected to a mean wind speed of 13 m/s ( 30 mph ) for an unknown amount of time with the rise of the leading edge measured to describe performance. Good performing free tab shingles were defined as having smaller lifts during wind testing. Not surprisingly, heavier shingles performed the best and a near linea r relationship between performance and weight was identified for a given uniform shingle thickness. The goal of the wind investigation was to assess the sealing characteristics of sealing three tab organic reinforced shingles. Th erefore, the test decks containing self sealing shingles, twelve in total, were subjected to three di fferent curing temperatures [49 C (120 F), 60 C ( 140 F), and 76 C ( 160 F)] for sixteen hours each. The test consisted of four step and hold mean wind veloc ities of 13 m/s ( 30 mph ) 18 m/s ( 40 mph ) 22 m/s ( 50 mph ) and 27 m/s ( 60 mph ) The time held at the first three wind velocities was not reported. The time for the 27 m/s ( 60 mph ) test was two hours. Failure for these tests was defined as failure of the a dhe sion on one shingle tab The tests revealed that nine of the twelve shingle deck specimens could
48 withstand 27 m/s ( 60 mph ) for two hours when they were conditioned at 60 C ( 140 F) for two hours. The remaining three required a 76 C ( 160 F) cure. Bond strength tests correlated well with the wind tests findings. Twelve asphalt shingle products were subjected to the same variation in cure temperature for five and sixteen hours and then tested for uplift resistance of their tab seals (Culle n 1993) Fr om these two tests, it was reported that a cure temperature of 60 C ( 140 F) and time of 16 hours was adequate to evaluate the wind performance of self sealing shingle systems. To validate the findings of the laboratory tests, Cullen investigated the perfor mance of self sealing shingle systems in the natural environment. Twenty two test decks were exposed for a period of one year in Washington D.C. starting in the spring. The tab seals were periodically inspected for adherence and results showed that all dec ks had full adhesion within fifty days. When the tab seal bonds were broken the following December, all shingles resealed the following spring. At the time of the Cullen (1960) report, no standard wind performance tests existed for asphalt shingles, but a s a result of this investigation the Underwriters Laboratory (UL) 997 Wind Resistance of Prepared Roof Covering Materials (1995) was developed. The test is est setup and conditioning. When first drafted in 1960, 27 m/s ( 60 mph ) was near the limit of fan controllability; therefore, the test decks were subjected to a maximum of 27 m/s ( 60 mph ) despite the likelihood of higher in service wind speeds. The America n Society for Testing and Materials (ASTM) D3161 Standard Test Method for Wind Resistance of Asphalt Shingles (2008 b ) was first published in 1972 with an identical test procedure. These
49 standards are based on data from shingles that were developed and manu factured seal organic reinforced asphalt shingle response to higher wind speeds and wind speed fluctuations (Benjamin and Bono 1967) This research was conducted u sing a larger fan system capable of wind speeds up to 44 m/s ( 100 mph ) ; additional tests were conducted on shingles that passed the UL 997 27 m/s ( 60 mph ) wind test. All of the 225 shingle test decks passed a 15 minute, 33 m/s ( 75 mph ) mean wind test and 9 5% of the test decks passed a 5 minute, 44 m/s ( 100 mph ) mean wind speed test. To replicate the fluctuating component of the wind speed, the wind speed was varied between 13 m/s ( 30 mph ) and 44 m/s ( 100 mph ) Each wind speed was held for 60 s for some decks and 30 s for others before a series of wind speed changes cycling from 13 m/s ( 30 mph ) to 44 m/s ( 100 mph ) for a total of 20 oscillations were applied. All test decks passed the wind fluctuation tests. While results of higher wind speed research show ed good asphalt shingle wind performance in simulated hurricane strength wind speeds, concerns surrounding the validity of the ASTM D3161 (ASTM 2013) and UL 997 (UL 2005) test methods soon followed. With the advent of the asphalt shingle self seal tab sys tem and its improved wind resistance, the weight of the asphalt shingle was no longer the main source of wind resistance. This allowed the use of lighter weight and cheaper shingle mats (Cash 1995) In 1960, glass fiber strand based mats were introduced as a replacement to the organic material based mat (Cullen 1992) The drawback to the fiberglass mat is an increase in flexibility of the shingle; that is, if a self seal were to fail the shingle would be
50 more likely to lift in the wind compared to a heavier and stiffer organic reinforced fiberglass shingles contain a chemical saturant that gave the fiberglass reinforced shingles a Class A fire rating. Organic reinforced asphalt shingles typically have a lower class (Class C) rating due to the combustibility of the organic material. The growth of the fiberglass reinforced asphalt shingle market can be partly attributed to the increase in condominium and commercial construc tion that required Class A fire ratings. By 1982, production of fiberglass reinforced asphalt shingles overtook organic reinforced asphalt shingle production; a trend that has continued (Cash 1995) The Development of the Asphalt Shingle Wind Uplift Model (1980 1997) The goal of the UL 997 test was to provide a predictive method for in service asphalt shingle wind performance. However, during in house product testing in the early 1980s using the UL 997 wind test standard, Owens Corning Fiberglas observed no appreciable shingle performance differences between products that should have produced significantly different results. Following this, Drs. Jon Peterka and Jack Cermak of Colorado State University (CSU) were contracted by Owen Corning to reevaluate UL 99 7 and develop a more refined test method that would more accurately simulate in service wind loading conditions This work (Peterka and Cermak 1983) lead UL 997 procedure to include more realistic wind effects. The standard 0.9 m by 1.2 m ( 3 ft by 4 ft ) test deck Layer Wind Tunnel to conduct tests using turbulent boundary layer wind that simulated the natural wind behavior (Figure s 3 1 and 3 2 ). Unsealed organic reinforced and fiberglass reinforced shingles from several
51 manufacturers were subjected to wind speeds up to 36 m/s ( 80 mph ) To evaluate the effect of temperature on shingle performance, the shingles were tested at two temperatures, 24 and 2 C ( 75 and 35 F) At the time, it was thought that lower temperature would increase the brittleness of the shingle, thereby increasing shingles pulling over the fasteners during wind events. Fig ure 3 1 Pre wind test asphalt shingle test deck inside the Colorado State University Meteorological Boundary Layer Wind Tunnel (Figure courtesy of Jon Peterka) Fi gure 3 2 Post wind test asphalt shingle test deck inside the Colorado State University Meteorological Boundary Layer Wind Tunnel (Figure courtesy of Jon Peterka)
52 The goal of the test was to observe how the shingles behave during this new test method and to discern if performance differences could be extrapolated. It was observed that organic reinforced shingles sustained less damage than the fiberglass reinforced shingles, likely due to a higher shingle mass resulting in a greater resistance to uplift. Greater shingle damage was o bserved in tests on the colder [2 C ( 35 F ) ] shingles. The overa ll outcome from the testing was that the performance of the shingles in the new test correlated to the pr edicted quality of the shingle. From wind flow visualization tests a wind uplift mechanism was proposed. It states that as wind flow encounters the lea surface causing a negative pressure relative to ambient in this separated region. A positive pressure relative to ambient is produced at the leading edge and is forced under the shingle. Th e effect of the positive pressure below the shingle and the negative pressure above the shingle produce a net uplift force on the shingle. (Figure 3 3 ). (Peterka et al. 1983). Fig ure 3 3 Wind load model proposed by Peterka et al. (1983).
53 Following the initial shingle uplift experiments, Peterka et al. (1983) experimentally investigated the proposed wind uplift model as well as shingle permeability and the distribution of win d induced uplift pressures on asphalt shingles. Relatively air impermeable roofing materials such as membrane roofing systems are susceptible to uplift pressures developed by the separation of wind flow over the building. The pressure in this separated reg ion is lower than the internal pressure of the building, producing uplift Wind uplift pressure distributions and magnitudes found in building standards such as ASCE 7 10 (ASCE 2010) are for impermeable systems. A permeable roofing system will allow for partial equalization of pressure between the upper and lower surfaces of the system. If the permeability is high enough, the loads developed within the separation region will be of a s mall magnitude due to rapid venting of pressure through the system surface. To examine the permeability of shingles, a box was sealed to an asphalt shingle deck with a vacuum attached to rapidly reduce the pressure within the sealed volume. Two different t ests were conducted: one with shingles installed as per in service installation and the other with all shingle tab joints and deck joints sealed with silicone. Pressure measurements above and below the asphalt shingle revealed that asphalt shingles rapidly vent air between their upper and lower surface. The results suggested that, given a high permeability in asphalt shingles, a significant uplift load will not be generated by the larger scale flow separation region. Rather, the proposed mechanism of locali zed flow separation at the shingle tab leading edge may be the genesis of asphalt shingle uplift Pressure measurements taken simultaneously above and below shingles during wind testing showed that wind flow near the roof surface was correlated to uplift p ressure, further validating the new uplift
54 model. Expected uplift pressures for asphalt shingles subjected to 80 mph wind testing varied from 48 to 144 Pa ( 1 to 3 psf ) significantly lower than pressures found on impermeable roofing systems. The results of these two studies (Peterka and Cermak 1983; Peterka et al. 1983) provided three major conclusions about asphalt shingle wind loading: 1. For wind flowing up the roof slope, localized flow separation at the leading edge of the shingle tab may be the largest contributor to asphalt shingle wind uplift. Asphalt shingles are a relatively permeable material and may not be significantly affected by the larger scale flow separation bubble. 2. Near roof surface wind speed may be used as a prediction for asphalt shingle uplift pressures. 3. Near freezing temperatures may increase the brittleness of fiberglass reinforced asphalt shingles; which, in turn, may increase the vulnerability of the wind related damage. Seeking to develop a more refined predictive method for as phalt shingle wind res istance than the current UL 997 (UL 1995) and ASTM D3161 (ASTM 2013) test standards, ARMA formed the High Wind Task Force in 1990. The goal of the task force was to determine the relationship between wind speeds and asphalt shingle ta b uplift resistance (Shaw 1991). A two phase program was developed: 1) develop a self seal adhesive strip (described in the next section), 2) define the physics of shing le and his colleagues were contracted to perform the wind tunnel and outdoor studies that validated the asphalt shingle load model previously developed (Peterka and Cermak 1983; Peterka et al. 1983). He proposed the asphalt shingle uplift equation shown in Equation 3 1
55 ( 3 1 ) where: is the peak uplift pressure that the shingle must resist is the density of air is the mean approach wind velocity at the eave height of the building is the peak gust wind speed on the roof and is the uplift differential pressure coefficient, unique for each shingle design With thi s equation, the peak uplift pressure exerted on a shingle can be predicted by knowing the approach flow characteristics, the near roof surface wind flow characteristics, and the uplift pressure coefficient that will be unique to each shingle design. Buildi ng standards such as ASCE 7 10 (ASCE 2010) represent uplift loads as pressure coefficients, a dimensionless number that defines the relative pressure for a given flow field independent of the flow velocity. For asphalt shingles, a differential pressure coe fficient is used to describe the net uplift pressure on asphalt shingles. The wind experiments conducted in the early 1990s by Peterka et al. (1997) investigated parameters of the components in this uplift equation by three methods: 1. Magnitude and distribu tion of near roof surface wind flow on model scale residential buildings in a boundary layer wind tunnel. 2. Correlation between near surface roof flow and uplift pressures generated on a full scale asphalt shingle test deck in a boundary layer wind tunnel. 3. Evaluation of uplift pressures and near roof surface wind flow generated on a full scale residential building located outdoors and subject to natural wind. For component one, three 1:25 scale t shape model buildings were constructed (Figure 3 4 a ) with roof slopes of 2:12, 5:12, and 9:12. Also tested was a 1:25 scale model gable roofed building that matched a full scale building constructed for validation of the model scale data (Figure 3 4 b ). Each building was placed inside the boundary
56 layer wind tunnel at CSU and subjected to a wind flow correspo nding to open country exposure [ i.e. ASCE 7 10 Exposure C (ASCE 2010)] Because the flow near the roof surface was of greatest interest for asphalt shingle wind loading, flow measurements were taken over each build 1 mm [ 0.04 in ( 25 mm at full scale) ] above the roof surface. The ratio between the peak observed near roof surface wind speed and the mean wind speed of the upwind a irflow is needed for Equation 3 1 An upper bound ratio of 2.5 was observed in the scaled model wind tunnel tests. The highest observed near roof surface wind speeds for all four buildings were located near the intersection of the roof ridge and gable end (Cochran et al. 1999) A B Fig ure 3 4 Test setups for wind testing of asphalt shingles. A) Model scale wind tunnel tests measuring near roof surface flow. B) Full scale outdoor test measuring simultaneous wind flow and asphalt shingle pressures. (Figures courtesy of Jon Peterka) The design of the leadin g edge of the asphalt shingle plays an important role on the aerodynamics of asphalt shingle uplift These design factors may include: the location of the self seal adhesive, the installed pattern (or distribution) of the self seal adhesive (i.e. a disconti nuous pattern may allow airflow behind the seal, increasing the pressure on the underside; thereby increasing the net uplift on the shingle), and the profile of the leading edge of the shingle tab (i.e. thick butt, sharp edge, etc.). The
57 second component o f the investigation utilized the same elevated 1.2 m by 0.9 m ( 4 ft by 3 ft ) asphalt shingle test deck developed during the 1983 experiments (Peterka and Cermak 1983). The deck was subjected to a boundary layer flow in the CSU wind tunnel with two differen t turbulence intensity levels of 4 and 17%. The CSU wind tunnel was unable to replicate full scale turbulence intensities found in the natural wind; therefore, it was necessary to determine the effect of the magnitude of turbulence intensity on the develop ed shingle uplift pressure coefficients. The shingle tab located in the middle of each deck was instrumented with pressure taps above and below the shingle surface and wind flow measurements were obtained 1 in above the instrumented shingle using either a hot film anemometer or pitot static probe. Mean pressure coefficients captured during this test showed that the uplift force is higher in front of the tab sealant compared to behind (up slope of) the tab sealant (Table 3 1). This likely occurs for three r easons: 1) a separated flow region is generated above the leading edge of the shingle with reattachment occurring a few inches upwind, 2) the tab seal reduces/prevents the positive pressure behind the tab sealant (depending on the sealant design), and 3) t ab cutouts assist in pressure equalization behind the seal strip Therefore, the location of the tab sealant will play a large role in the loading mechanism generated on tab adhesive. Differences also exist in pressure coefficients between the three tab and double thickness shingles ( Table 3 1 ). To investigate the role of near surface wind flow on uplift pressures, the middle shingle tab from a test deck was replaced with a thin rectangular piece of brass that would mimic a shingle tab both in dimension a nd location on the deck with a seal located where one would be on an asphalt shingle. Fifty four pressure
58 taps were installed on the brass shingle (half on the top surface, half on the bottom surface) as this would allow for larger area averages to be dete rmined. The deck was placed on the floor, oriented such that the generated wind flow would travel up the 4:12 sloped test deck with a smooth curved transition between the wind tunnel floor and the test deck. As with the previous pressure measurements, near surface roof flow at 25 mm ( 1 in ) above the brass shingle was captured. From this data it was observed that asphalt shingle uplift pressures correlate with near roof surface flow in flow fluctuations up to 12 Hz for a wind flow of 10 m/s ( 22.5 mph ) Ta bl e 3 1 Wind tunnel measured three tab shingle with cutouts Shingle p art Shingle t hickness Single, ~2.8 in Double, ~5.6 in Seal strip to front edge 0.4 0.8 Top of cutout to seal strip 0.1 0.1 Top of cutout to front edge for unsealed shingle 0.4 0.8 The final component of the investigation was the validation of wind tunnel test data using a full scale, gable roofed building constructed outside in a windy location near Fort Collins, CO. The house consisted of a 7 m by 10.5 m ( 23 ft by 34.5 ft ) one story building with a 5:12 gable roof. Three tab fiberglass reinforced asphalt shingles were installed on the roof with pressure taps installed above and below the shingles at several locations on the roof. To capture simultaneous velocity and pres sure data, unidirectional velocity sensors were installed above the tapped shingles and oriented down the roof slope. The house could be rotated 360 degrees to provide uplift/velocity data for all wind azimuths. To capture the approach flow conditions, a 6 0 m ( 197 ft ) instrumented meteorological tower was located near the house and a 10 m ( 33 ft )
59 meteorological tower was installed upwind of the house. Data from the observation towers, shingle pressure transducers, and roof surface velocity sensors were capt ured during strong wind storms with peak wind gusts ranging from 13 m/s ( 30 mph) to 27+ m/s ( 60+ mph ) Several observations were made from the comparison of the full scale outdoor tests and the wind tunnel experiments: 1. The full scale data appeared to vali date the wind tunnel data, and the highest pressures observed were within the prediction of the uplift model. 2. For wind flow up the roof slope, asphalt shingle uplift pressures correlated well with near surface roof flow. 3. The highest observed shingle uplift pressures corresponded to a 50 degree wind azimuth relative to ridge line of building (Figure 3 5 ). Due to the unidirectional nature of the velocity sensors, only the upslope component of the wind velocity vector could be obtained for this azimuth. 4. Significant shingle uplift pressures were observed for wind flow approaching the leeward side of the roof (with respect to the instrumented shingles). While the wind uplift model only describes wind flow up the roof slope, it may correctly model the local flow at other wind azimuths. From Peterka et al. (1997), ARMA and UL drafted a standard test method to Test Methods for Wind Resistant Asphalt Shingles with Sealed Tabs ( UL 2003) was publi shed in May 2003; while the identical ASTM D7158 Wind Resistance of Asphalt Shingles (Uplift Force / Uplift Resistance Method) ( ASTM 2011c ) was first published in 2005. These standards are based on a standards developme nt report (Peterka and Esterday 2000) that is not publically available, since the provisions are published in the standard. These methods can be broken up into three parts:
60 Fig ure 3 5 Peak pressure coefficients measured on a full scale asphalt shingle subjected to wind flow from var ying directions. 1. Determine uplift rigidity of the shingle through mechanical uplift testing. Shingles will lift in the wind and the magnitude of this lift will depend on the stiffness of the shingles. The aerodynamics of asphalt shingles change as the shi ngle lifts, therefore, the resulting pressures exerted on the shingle can change. The stiffer the shingle, the lower the resulting loads. A conservatively low stiffness (EI) of 0.0072 N m 2 ( 2 .5 lbs in 2 ) may be used as a default. This shingle stiffness valu e is used for Part 2 to determine pressure coefficients. 2. Determine the wind uplift pressure coefficients on the asphalt shingle. Shingles are installed on 1.2 m by 0.9 m ( 4 ft by 3 ft ) test deck with one shingle tab in the middle of the deck containing fo ur pressure taps above and four taps below the shingle. Similar to ASTM D3161 (ASTM 2008b) these decks are conditioned and placed in front of a fan system. However, fan speeds are limited to 15.5 m/s ( 35 mph ) and a small amount of turbulence is introduced into air flow. Mean uplift pressures are captured for shingles lying flat on the deck surface and for shingles that have their edges raised with shims at the leading edge to simulate shingle uplift during strong wind events. The pressure coefficients are u sed in combination with the Peterka wind uplift equation to determine the uplift loading of asphalt shingles at various peak 3 sec gust wind speeds. 3. Determine the uplift resistance of the asphalt shingle tab sealant through mechanical uplift testing (outl ined in the next section). Results of this test are compared to the predicted uplift loads determined in Part 2.
61 From ASTM D7158 ( ASTM 2011c ) asphalt shingles are classified and labeled on their packaging according to their predicted resistance to peak 3 second gust (basic) wind speeds at 10 m ( 33 ft ) in Exposure C (open country) following ASCE 7 05 (ASCE 2005) Adjustment factors are required for various environmental/building factors such as buildings greater than 18 m ( 60 ft ) and if the user is using t he ASCE 7 10 (ASCE 2010) design standard. The shingle classification is shown below: Class D Passed at basic wind speeds up to and including 40 m/s ( 90 mph ) Class G Passed at basic wind speeds up to and including 54 m/s ( 120 mph ) Class H Passed at basic wind speeds up to and including 67 m/s ( 150 mph ) Most United States residential building codes refer to ASCE 7 05 (ASCE 7 05) as their wind load standard; therefore, this classification system provides a direct comparison between shingle requirement s and shingle performance. A 2011 survey of shingle products have been wind tested by ASTM D7158 (ASTM 2011c) and all of those tested were classified wi th Class H rating s ( Texas Department of Wind Insurance 2011). The same survey noted that all products listed have a Class F [49 m/s ( 110 mph) ] ASTM D3161 ( ASTM 2008b ) classification as well. The Modern Era (1997 ) The ASTM D6381 Asphalt Shingle Mechanical Uplift Resistan ce Test Method Prior to the initiation of the Peterka wind load studies, the ARMA task force began development of a test method that would determine the uplift resistance o f a 1991). From the initial Peterka et al. (1983) repor t, it was evident that the greatest uplift loads would occur nearest the leading edge of the shingle. At the time, shingles were typically produced with to 1 in distances between the leading edge of the tab sealant and the leading edge of the shingle tab (Hahn et al.
62 2004). The resultant wind loading on this cantilever span would produce a peel type uplift force on tab sealant. The mechanical uplift test was developed to simulate this loading condition. The test specimen consisted of 89 mm ( 3.5 in ) wide b y 178 ( 7 in ) long asphalt shingle lower piece with a n 89 mm ( 3.5 in ) wide by 102 mm ( 4 in ) long upper tab service tab bond is replicated. Prior to uplift testing the bond between the lower and upper shingle was conditioned at 60 C ( 140 F) for 16 hours; the same as the ASTM D3161 (ASTM 2013) and UL 997 (UL 1995) conditioning procedure. Mechanical uplift testing consisted of the specimen attached to a clamp assembly along the 89mm ( 3.5 in ) edges. The uplift load was generated from a clamp affixed along the leading edge of the shingle specimen. This clamp was connected to a universal testing machine, which provided a constant velocity uplift of 127 mm/min ( 5 in/min ) and simultaneous measurement of uplift load on the shingle tab. Seven testing labs were utilized for round ro bin testing of this draft standard to confirm repeatability of test methods and results (Shaw 1991). After confirmation, the standard published in 1999 and designated as ASTM D6381 Standard Test Method for Measurement of Asphalt Shingle Mechanical Uplift R esistance ( ASTM 2008b ) As described below, recent modifications to the mechanical uplift test have been made in response to changes in the tab sealant design and market trends. Many edge. A decrease in distance between the tab sealant and the leading edge will reduce the total uplift loading generated ahead of the sealant. Therefore, this loading mechanism can change from a peel type to a direct tension type load ing (Hahn et al. 2004) The way an
63 adhesive is loaded (i.e. peeling, direct tension, etc.) is known to have a significant effec t on its strength (Shiao et al. 2003b ). Results of asphalt shingle tab sealant mechanical uplift resistance tests comparing peel, direct tension, and c ombined showed that direct tension produced over double the resistance of the D6381 pee l type resistance (Shiao et al. 2003b ). The combined test consisted of an attachment that mimicked the Peterka wind load model (Peterka et al. 1997) with forces being ge nerated on shingle specimens ahead and behind the tab sealant. The sealant strength for this loading fell between the low peel strength and high direct tension strength, suggesting that the actual loading of a tab seal is a combination of peel and direct t ension. As a result, the 2008 edition of ASTM D6381 ( ASTM 2008b ) test requires direct tension testing be conducted along with the original peel test. Depending upon the magnitude of the pressure coeffi cients obtained from ASTM D7158 ( ASTM 2011c ) or UL 2390 (UL 2003) testing, the results of each test may be used separately or in combination to determine total uplift resistance of a 2004). Questions remain on the applicability of this test method to predicting in service shing le wind resistance. Foremost among them is the loading protocol; which specifies a constant displacement velocity of 127 mm/min (5 in/min) Near roof surface wind flow is turbulent in nature; therefore, the uplift loading from wind will also conta in fluctu ations (Peterka et al. 1997). Shiao et al. (2003b ) has shown that an increase in loading rate correlates to a high er shingle tab seal resistance [ i.e. the current ASTM D6381 ( ASTM 2008b ) loading rate produces conservative resistance results] However, shingles are potentially subjected to thousands of wind gusts throughout their lifetime and the long term performance of shingle tab seal to these fluctuations (i.e. fatigue resistance) has
64 not been quantified. Thus, with the current ASTM D6381 ( ASTM 2008b ) it is difficult to Table 3 2 Summary of standardized test methods to evaluate asphalt shingle wind performance Test method d esignation Year first p ublished Test method o verview UL 997 (UL 1995) 1960 Asphalt shingles are installed on a 0.9 m by 1.2 m ( 3 ft by 4 ft ) test deck, cured for 16 hours at 60 C ( 140 F ) and then subjected to 2 hours of 60 mph winds. Failure is defined as a shingle tab that either loses its tab adhesion or failure of its mechanical interlock. ASTM D3161 (ASTM 2013) 1972 Essentially identical to UL 997 with the exception of the maximum allowable wind speed. D3161 has three classification designations: 1) Class A passed 27 m/s ( 60 mph ) 2) Class D passed 40 m/s ( 90 mph ) 3) Class F passed 49 m/s ( 110 mph ) Note: These wind speeds do not directly correlate to ASCE 7 wind speeds. ASTM D6381 ( ASTM 2008b ) 1999 resistance. Shingle specimens are subjected to a constant rate peel and direct tension testing of the sealant. UL 2390 (UL 2003) 2003 Based on the Peterka wind load model, this test method determines a shingle wind uplift pressure coefficients. The pressu re coefficients can be used to predict the loads that will be exerted on a shingle at various ASCE 7 wind speeds. ASTM D7158 ( ASTM 2011c ) 2005 Identical to UL 2390 in test procedure. References ASTM D6381 to determine the uplift resistance of the tab sealant. Comparison between D7158 predicted uplift force and D6381 measure resistance gives three wind speed classifications: 1) Class D up to 40 m/s ( 90 mph ) resistance, 2) Class G up to 54 m/s ( 120 mph ) resistance, 3) Class H up to 67 m/s ( 150 mph ) resistance. Note: These wind speeds correlate to winds defined by ASCE 7 05 for non critical facilities less than 60 ft tall in Exposure C. In Service Wind Performance of Asphalt Shingles Laboratory wind testing of asphalt shingles provides a relatively simple method for predicting in service wind performance. However, these methods cannot completely
65 replicate the conditions that shingles are subjected to once they are installed. A key comp onent to understanding shingle wind resistance is observations that are made following shingle damage caused by wind events. Since 1989, damage assessments made by organizations and federal agencies such as FEMA and RICOWI have provided sphalt shingle performance. The observations made in these reports provide an opportunity to evaluate deficiencies in products, design, and installation. An overview of selected damage report observations is provided below. Hurricane Hugo made landfall on the east coast of South Carolina in 1989 as a Category 4 hurricane on the Saffir Simpson scale. Damage observations of asphalt shingle roofing by Smith and McDonald (1990) noted highly variable wind uplift performance of shingles, with some houses sustain ing no damage, while others nearby sustained complete shingle loss. The damage was primarily attributed to weak tab seals. Improperly located fasteners were also often observed at damaged roofs. Failure of the roof covering did not only just impact the cov ering itself. Rather, in financial terms, the resulting interior losses caused by roofing failure were often greater than loss from the roof covering. Smith concluded that standardized wind testing of roof coverings, including the ASTM D3161 (ASTM 2013) fo r asphalt shingles, appeared deficient in predicting wind performance. This observation would be repeated after Hurricane Andrew made landfall in South Florida in 1992, also causing damage to asphalt r oofing systems on houses (Smith 1995). Improperly loc ated shingle fasteners have often been obs erved at damaged shingles (FEMA 2005a; Smith 1995; Smith and McDonald 1990). However, the extent to which the installation plays on the wind resistance of the shingle has not yet been
66 quantified. Pull through of th e shingle over the fasteners is often attributed to improper fastener placemen t. Smith and Millen (1999) note tests on unsealed asphalt shingles with misplaced fasteners showed a decrease in wind performance, but no definitive conclusions could be made regarding variations in placement (Smith and Millen 1999). A common observation throughout post storm reports is the failure of roof detail s such as hip, ridge, eave, an d rake shingle conditions (FEMA 2005a; FEMA 2005b; FEMA 2006; FEMA 2009; IBHS 2009). These failures appear to be independent of the age of the roof and more closely tied to the design and installation of these edge conditions. Bonding of the hip and ridge caps appears to be an ongoing issue, and starter courses along the eave were often improperly installed. The implications of failures to these areas of the roof range from a minor exposure of the hip and ridge deck joints to a more widespread failure propagating from eave and rake edge failures. While damage reports continue to be a valuable source of information, more work is necessary to understand the role of installation variability in asphalt shingle wind performance. Thr oughout the 2000s, hurricanes impacted the Southeast and Gulf Coast US causing extensive shingle damage. Shingle p erformance was variable (RICOWI 2006). An Insurance Institute of Business & Home Safety (IBHS 2009) study of shingle damage in Hurricane Ike showed variable performance amongst products with the same ASTM D7158 Class H (150 mph) rating ( ASTM 2011c ) Wind speeds at the investigated site were 49 m/s ( 110 mph ) [ peak 3 second gust at 10 m ( 33 ft ) Exp osu re C] below
67 (2010) conducted an asphalt shingle damage survey in Texas after Hurricane Ike and Gustav in 2008 and found that homes with newer (less than 5 years old) shingle installations performed significantly better than older shingle roofs (greater than 5 years old), although it was not certain if age or changes to the building code aro und 2002 was the cause. This performance gap was also noted by RICOWI (2006) after Hurricane Charley struck Florida in 2004 and by Gurley and Masters (2011) in a post 2004 hurricane season building performance survey. All three studies postulated that, whi le product improvement could be attributed to the better performance of newer roofs, the effects of aging could not be discounted. Experiments by Terrenzio et al. (1997) and Shiao et al. (2003 a ) have noted that the greatest cause of asphalt shingle aging is thermal loading. Over time, the asphalt within the shingle becomes oxidized causing embrittlement of the shingle. Currently, no studies have quantified the effects of aging s based upon an asphaltic formulation, what effects would this potential oxidation reaction have on the tab seals adhesive s trength? The current ASTM D7158 ( ASTM 2011c ), ASTM D6381 ( ASTM 2008b ), and UL 2390 (UL 2003) standard test methods only provide info rmation on the performance of new, laboratory prepared asphalt shingles, making estimation of the long term performance of the tab adhesive difficult.
68 CHAPTER 4 UNSEALED NATURALLY A GED ASPHALT SHINGLES AND THEIR VULNERABIL ITY IN WIND The first of tw o studies presented in this chapter describes a field assessment of 30 single family homes in Florida and Texas to characterize the occurrence of unsealed shingles on field, hip, and ridge roof regions. In the second study, 17 asphalt shingle roof systems were subjected to full scale wind testing at the Insurance Institute of Business & Home Safety (IBHS) Research Center. The findings indicate that unsealing of shingles may have been a contributor to shingle roof cover damage reported in po st hurricane asse ssments (FEMA 2005a; FEMA 2005 b; FEMA 2009; Gurley and Masters 2011; Liu et al. 2010; Rickborn 1992; RICOWI 2006 ; RICOWI 2007). A strong correlation is demonstrated between partially unsealed shingles, resultant wind damage during IBHS wind tests, and dama ge observations in post hurricane reports. Further, new shingles generally appear to remain fully sealed for the first 4 5 years of service life. Beyond that timeframe, the frequency of unsealing trends upward. These results are consistent with post hurric ane assessments by Gurley and Masters (2011) and Liu et al. (2010), which found that shingle roofs with six or more years of weathering were damaged at a 50% higher rate than newer shingle roofs. Study 1: Survey of Naturally Aged Shingle Roofs for Unseale d Shingles The research objective was to assess the adhesion of the shingle sealant strips on in service roofs. The subject roofs were located at single family homes in Florida and Texas. Reprinted with permission from Dixon, C.R., Masters, F.J., Prevat t, D.O., Gurley, K.R., Brown, T.M., and Peterka, J.A. (2013 c The influence of unsealing on the wind resistance of asphalt shingles Journal of Wind Engineering and Industrial Aerodynamics [Article Submitted for Review]
69 In 2012, 27 roofs were surveyed in Altamonte Springs (2 roofs), Gainesville (3 roofs), Volusia County (4 roofs), a nd Sarasota (18 roofs). Figure 4 1 depicts the locations. Roof slopes ranged from 4 units vertical by 12 units horizontal (4:12) to 7:12. Ten roofs were three tab style, and 17 were laminate style. For the Florida surveys, over 6100 m 2 (6 6130 ft 2 ) of shingle roofing was surveyed, corresponding to a sample size of 46,800 shingles. The installation age for 23 of 27 Florida roofs was obtained from the homeowner or roofing permit records. The shingle age was defined as the time from the installation to the survey. The distribution of ages is : 0 6 years (six roofs), 7 13 years (ten roofs), 14 20 years (seven roofs), and unknown (four roofs). Access to these roofs was made possible t hrough a Florida Department of Emergency Management grant or personal contact with the homeowner. Insight Engineering and Cross Pointe Construction provided information about three additional shingle roof systems in the Houston, Texas metropolitan area that were surveyed in February 2013. The roof covers were installed within approximately 4.5 years prior to the survey as part of repairs resulting from Hurricane Ike (2008). One roof consisted of three tab style shingles and two roofs consisted of laminat e style shingles. Figure 4 1 Locations of the asphalt shingle surveys conducted in Florida.
70 Survey Method Individual shingles were manually inspected on each roof using a non destructive technique (Figure 4 2 ). Survey personnel gently applied upward pressure by hand to the leading edge. Shingles were classified as: (1) sealed, (2) partially unsealed, or (3) fully unsealed. A sealed shingle was defined as a shingle with either full adhesion on the sealant strip or the loss of adhesion less than a cont inuous 51 mm (2 in). A partially unsealed shingle was defined as any loss of adhesion on the shingle that was greater than or equal to a continuous 51 mm (2 in) length, whereas a fully unsealed shingle was defined as the loss of adhesion along the entire l ength of the sealant strip. A strip of colored tape or chalk was placed on the top surface of each partially or fully unsealed shingle to assist with pattern recognition and photographic documentation. For each partially or fully unsealed shingle found on the Florida roofs, the following data were recorded on a roof plan: 1. Location on the roof. 2. Unsealed location on the strip (left corner, center, right corner). 3. Unsealed length. 4. Plane within the shingle composite where the loss of adhesion occurred to det ermine the sealant strip failure mode ( cf. Shiao et al. 2003 a ). Potential for Wind Induced Loss of Shingle Sealing Extreme wind climatology in Florida and the Texas coast is predominantly associated with hurricanes, thus the peak wind speeds at each survey location in Florida and Texas were extracted from H*Wind swath datasets published by the Hurricane Research Division (Powell et al., 1998) to assess historical wind events as a potential cause of partially unsealed shingles. Site specific peak gust wind speeds were
71 calculated from all tropical cyclones impacting a given roof from the date of roof installation to the date of survey. Houston, TX was not impacted by tropical cyclones in 2009 through 2012, thus it was not included in the survey. Gust calculations were performed on all tropical cyclones from 1992 to the date of roof survey, which encompasses all but four roof lifespans in the study. H*Wind wind swaths are reported as maximum 60 s wind speeds ( V 60 ) in open exposure at 10 m (33 ft) for all land areas. The exposure condition at the survey sites was suburban. Following the approach of Masters et al. (2010), H*Wind velocities (i.e. the 60 s mean wind speed at 10 m in open country) we re converted to mean wind speeds at 5 m (16 ft), which nominally corresponds to the mean eave height of a single family home in suburban terrain ( z 0 = 0.3 m). The conversion factor was 0.48. Next, the factor was multiplied by a speed up factor of 1.8 to co nvert the mean wind speed to the peak instantaneous velocity expected to occur on the roof deck (Dixon et al., 2013). Thus the total conversion factor was 0.87. Results of the analysis are shown in Table 4 1. Altamonte Springs experienced the highest near roof gust of all locations, 25 m/s (56 mph), during Hurricane Jeanne (2004). The second highest near roof gust occurred in Ormond Beach during Hurricanes Floyd (1999) and Irene (1999), 22 m/s (49 mph). The remaining sites experienced near roof gust wind s peeds ranging from 11 21 m/s (25 47 mph). All of the wind speed estimates are lower than the 27 m/s (60 mph) maximum near roof velocity threshold used in the ASTM D3161 (ASTM 2008b) fan test for shingle wind resistance certification (Dixon et al., 2012), w hich is the lowest threshold used by product approval standards in the last two decades.
72 Based on these assessments, it was concluded that it is unlikely an extreme wind even caused the unsealing, acknowledging that absent a long term monitoring program, it is not possible to prove if wind loads induced at lower wind speeds cause the unsealing. However, the systematic nature of the partially unsealed shingles detailed in the next section and the lack of observed surface cracking and tearing normally associ ated with sh ingle wind damage (FEMA 2005a; FEMA 2005b; FEMA 2009; RICOWI 2006 ; RICOWI 2007) support the assertion that wind was not the cause of the shingle tabs losing adhesion. Survey Results Shingles in the field of the roof More than 99% of the unseal ed shingles found on the Florida roofs were partially unsealed and exhibited the same location of unsealing reported in Marshall et al. (2010). Partially unsealed shingles of this type were found on 8 of 10 three tab shingle roofs and 11 of 17 laminate shi ngle roofs for a total of 19 of 27 surveyed roofs (70%). The unsealed shingles exhibiting locations and length of unsealing that was different from those reported by Marshall et al. (2010) (less than 1% of unsealed shingles) were contained on portions of t he 19 roofs identified above and the other eight surveyed roofs. The results presented below detail the location and plane of failure found on the more frequently observed partially unsealed shingles. The partial unsealing of three tab shingles occurred on the outside end tab of the strip where the end joint of the shingle course below, aligned with the centerline of the tab (Figure 4 2A). Laminate shingles exhibited a similar pattern of unsealing to the three tabs with the unsealed length running from th e end joint of the strip to the end joint of the shingle course below (Figure 4 2B). The unsealed length for laminate shingles
73 appears to be controlled by the horizontal offset selected by the installer typically 102 mm (4 in) to 178 mm (7 in). As shown in Figure 4 3, the resulting alignment of partially unsealed shingle locations produced easily observable patterns that were installation specific, i.e., vertically aligned for vertical (racked) installations and diagonally aligned for diagonal installatio ns. Table 4 1 Estimates of peak instantaneous velocity near the roof plane at each survey location Survey Location Analyzed Hurricane Seasons Peak Wind Speed Above the Roof Plane (m/s) [mph] Tropical Cyclone Name (Year) Altamonte Springs 2002 2011 1 25 [ 56 ] Jeanne (2004) Gainesville 1992 2011 2 11 [ 25 ] Frances (2004) Orange City 2002 2011 1 21 [ 47 ] Jeanne (2004) Ormond Beach 1992 2011 2 22 [49 ] Floyd (1999) and Irene (1999) Sarasota 1992 2011 2 18 [40 ] Frances (2004) Houston 2009 2012 3 None Reported None Reported 1 No roofs installed prior to 2002. 2 Location contains roof(s) with unknown installation date. 3 No roofs installed prior to 2009 In the present study, all partially unsealed shingles had visible adhesive residue of the unsealed portion of the sealant strip on both the bottom surface of the top shingle and top surface of the bottom shingle (i.e. a cohesive failure in the sealant), wh ich indicates that the shingles were initially fully sealed. Fully driven nails were found in the sealant strip on some partially unsealed shingles, however this was determined not to be a controlling factor because there was consistency in failure mode an d unsealed
74 length for shingles with and without fully driven nails in the sealant strip. Marshall et al. (2010) observations did not include sealant failure mode data. A B Figure 4 2 Lo cation of partial unsealing. A) three tab and B) laminate shingle systems. Figure 4 3 L ocation of partially/fully unsealed three tab and laminate shingles (tape marks ) All surveyed roofs in the Houston, TX metropolitan area contained partially unsealed field shingles exhibiting the same location of unsealing and sealant strip failure mode as the Florida roof surveys and Marshall et al. (2010). The engineers conducting th e Texas roof surveys did not quantify the percent of unsealed field
75 shingles on each roof; however, a review of photographs taken during the surveys indicated that partially unsealed shingles represented a majority of all unsealed shingles existing on the roofs. An example of the survey results on a portion of the three tab roof and one la minate roof is shown in Figure 4 4 In this figure, the triangular marks represent the location and length of unsealing on the shingle and dash marks represent shingle str ips or tabs that are fully sealed. Similar to the Florida roof surveys, the patterns of partially unsealed shingles in Texas corresponded to the direction of field shingle installation. A B Fig ure 4 4 Shingle roofs located in Houston, TX with partia lly unsealed shingles located by triangular chalk marks and fully sealed shingles located by dash marks. A) Three tab shingles and B) laminate shingles. Figure 4 5 shows the percentage of unsealed shingle strips on each roof as a function of roof age. The black square markers correspond to roofs with patterns of partially unsealed shingles that exhibited patterns found in Marshall et al. (2010). The gray circle markers depict roof coverings without partially unsealed shingles exhibiting a pattern similar t o Marshall et al. (2010). Roofs with patterned partially unsealed shingles had a range of less than 1% up to 86% of their shingle strips unsealed. The age of the roof with 86% unsealed strips was unknown and, therefore, not shown in Figure 4 5.
76 Every roof that containing unsealed strips with no discernible pattern had less than 1% of their shingle strips unsealed. Figure 4 5 also shows that the percentage of unsealed shingles for all roofs less than six years old is less than 1%, while 14 of 17 roofs older than six years had more than 1% of their shingles partially or fully unsealed. Figure 4 5 Percent of unsealed shingle strips located in the field of the roof verses roof age. Figure 4 6 shows a box plot of the percentage of unsealed shingles as a function of each age group. Roofs were stratified into three age ranges with the following distribution: 0 6 years (6 roofs), 7 13 years (10 roofs), and 14 20 years (7 roofs). The inset shows the result of a single 2004) comparing the mean values among the three groups. A statistically significant increase in the mean percentage of unsealed shingles was established at a 95% confidence level when the 0 6 age range was compared to the 7 13 age range (p value = 0.02) a nd 14
77 20 age range (p value = 0.02). A statistically significant increase was established between the 7 13 and 14 20 age ranges at a 90% confidence level (p value = 0.08). Figure 4 6 Boxplot of unsealed shingle strips located in the field of the roof verses roof age at the time of investigation. In summary, partially unsealed field shingles found on Florida and Texas roofs and reported in Marshall et al. (2010) demonstrate: (a) partially unsealed shingles in the field of the roof exist in hurricane prone Florida and Texas, (b) the nature of the unsealing is systematic and not induced by wind, and (c) the loss of adhesion increases with roof cover age. A relationship between likelihood of wind damage and the pre wind presence of unsealed shi ngles can be drawn when the findings of the roof surveys are combined with the Liu et al. (2010) study that showed a 50% increase in wind damage frequency on shingle roofs greater than six years old. Furthermore, photos of damaged shingle roofs reported in post hurricane damage investigations re veal blow off patterns (Figure 4 7 ) that are strikingly similar to the patterns of partially unsealed shingles obser ved both in this study (Figure 4 3 ) and M arshall et al. (2010) The damage pattern photographs in Fi gure 4 7 were chosen among many that are similar in the nature of the
78 damage pattern, and it is not conclusive that the damaged s hingles in Figure 4 7 were unsealed prior to the wind event. However, the shingle tabs blown off from Hurricane Ike in Figure 4 7B were located above the end joint of the shingle course below, identical to the observed location of partially unsealed shingles ( Figure 4 2A ). A B Figure 4 7 Blown off three tab asphalt shingles. A) Hurricane Katrina in 2005 and B) Hurricane Ike in 2008. Ridge and hip shingles Partially and fully unsealed ridge and hip shingles were found on 20 of the 27 surveyed roofs. Observations of unsealing were concentrated a t the downslope edges of both hip and ridge shingles. Full adhesion was observed elsewhere. Two findings indicate that these unsealed shingles never properly sealed. First, in contrast to field shingles, the unsealed strip on hip and ridge shingles did not show a transfer of sealant from the top surface of the sealant strip and bottom surface of top shingle (Figure 4 8 ), which is consistent with FEMA (2005a) post hurricane damage observations. Second, the percentage of unsealed ridge and hip shingles shows no obse rvable trend with age (Figure 4 9 ).
79 Figure 4 8 Typical condition for partially unsealed ridge and hip shingle with an adhesive failure mode between the top shingle and sealant strip indicated by the lack of sealant residue on the undersid e of the shingle. Figure 4 9 Percent of fully and partially unsealed hip and ridge shingles verses roof age. Two roofs contained hip and ridge shingles without sealant strips and are not shown in figure.
80 One potential source of unsealed ridge and hip shingles arises from their method of installation. Ridge and hip shingles can be purchased either pre manufactured or cut from three tab shingles. Both pre manufactured and cut three tabs are originally flat shingle strips that are folded over the ridge a nd hip roof line and nailed to the substrate with two fasteners per shingle. Once folded, the edges of a ridge and hip shingle will tend to lift to reorient the shingle back to its original geometry. If the sealant strip is unable to bond the edge of the s hingle at the onset of service, the shingle edge is not restrained from rising leaving it partially unsealed at its edges and sealed along its centerline where the crease in the shingle is formed. Study 2: Full Scale Testing of Asphalt Shingle Roof Syste ms In June July 2012, 17 full scale 6 on 12 slope roofs covered with ASTM D7158 Class H (ASTM 2011c) asphalt shingle were subjected to fluctuating winds at the IBHS Research Center in Richburg, SC. The roofs were constructed by licensed roofing contractors and conditioned outdoors 11 months prior to testing. Using the same method outlined in the previous section surveys were performed on each roof specimen just prior to wind testing. The surveys found fully and partially unsealed field shingles on 8 of the 17 roof specimens and partially unsealed hip shingles on all hip roofs. This section focuses on the wind performance differences between the sealed and unsealed field and hip shingles to demonstrate the vulnerability of pre existing unsealed shingle to wi nd. Experimental Design The test matrix consisted of two laminate shingle products and one three tab product classified as ASTM D7158 Class H (ASTM 2011c) and A STM D3161 Class F (ASTM 2013) wind resistant shingles. A licensed roofing contractor installed the asphalt
81 shingle roof systems in conformance to the 2010 Florida Residential Building Code driven 12 gauge electroplate galvanized nails wit h a 9.5 mm (3/8 in) diameter head and 31 mm (1.25 in) shaft length. Three tab specimens were secured with four nails per strip, while laminates were secured with six nails per strip. The roof specimens were placed on a base structure (9.1 m W x 12.2 m L x 2.4 m H) with a permanent half roof on one end (Figure 4 10 ) to form an enclosed test structure. A B C Fig ure 4 10 Wind directions for gable and hip roof specimens Once installed on the test structure, field, hip, and ridge shingles were surveyed following the procedure described in the previous section Painters tape was placed on all shingles containing an unsealed length greater than 51 mm (2 in), and overall photographs of each roof slope were captured to document the location of unsealed shingles. While the wind test sequence was ongoing, seven high definition video cameras above the test structure recorded the roof. Following the test, the roof was inspected for surface cracking, material tears, pull through at fastener heads, shingle blow off, and damage to the edge fastening. Field notes, roof plans, and photographs were used as part of the documentation process.
82 Wind Test Sequence and Boundary Layer Simul ation The full scale test facility at the IBHS Research Center was designed to replicate turbulent boundary layer flows at sufficient scale to evaluate the performance of a single family home. Wind is generated by 105 vaneaxial fans (Liu et al., 2011) grou ped into 15 subarrays under individual fan speed control. Wind speed records are derived from the Davenport (1961) spectrum accounting for desired mean velocity, peak velocity, terrain exposure, and turbulence characteristics. Five records were created for the shingle roof tests: four sequences of 30 minutes each with fluctuating wind replicating the turbulent boundary (henceforth, Wind Levels 1a, 1b, 2, and 3), and a fifth 17 minute sequence corresponding to a step and hold wind velocity up to the maximum wind speed capable at the test section (henceforth, Wind Level 4) corresponding to 3 sec open exposures gust envelope in the ASCE 7 wind load provisions. The first three test roofs were subjected sequentially to Wind Levels 1a, 2, and 3, while the remainin g 14 roofs were subjected sequentially to Wind Levels 1b, 2, 3, and 4. Table 4 2 lists the measured mean/peak velocities and longitudinal/lateral turbulence intensities of the five test sequences. Anemometric data were captured from a Turbulent Flow Inst ruments Cobra Probe three axis velocity sensor mounted 0.3 m upwind of the windward face of the test structure on the centerline of fan opening at a height of 5 m (16.4 ft) above the chamber floor. Additional measurements of velocity were made at heights o f 1.4 m (4.6 ft), 2.8 m (9.2 ft), and 3.9 m (12.8 ft) to produce the normalized mean wind velocity, lateral turbulence intensity, and longitudinal turbulence intensity vert ical profiles shown in Figure 4 11 Theoretical mean velocity profiles were generate d from ESDU (2002) and normalized to 5 m.
83 Table 4 2 Wind test sequence duration, wind speeds, and turbulence intensities Wind Level Test Duration (minutes) Mean Wind Speed 1 (m/s) [mph] Peak Instantaneous Wind Speed 1,2 (m/s) [mph] Longitudinal Turbulence Intensity (%) 1 Lateral Turbulence Intensity (%) 1 1a 30 18  33  23 9 1b 30 23  44  23 9 2 30 28  45  23 9 3 30 28  54  23 9 4 1 41  -14 6 5 48  -14 6 5 50  -14 6 5 54  -14 6 1 Measured at 5 m (16 ft) with velocity sampled at 500 Hz 2 Wind speeds varied approximately +/ 1 m/s (2 mph) per day due to air density fluctuations Figure 4 11 Measured and best fit theoretical normalized mean velocity, longitudinal turbulence intensity, and lateral turbulence intensity. Theoretical longitudinal (Iu) and lateral (Iv) turbulence intensity profiles were generated from ESDU (1983), assuming that ( / u* = 2.5) for Iu and ( / u* = 2.2) for Iv. The the oretical lines shown on Figure 4 11 correspond to the best fit roughness length
84 ( z 0 = 0.06 m). Figure 4 12 shows the normalized longitudinal wind spectrum measured at 5 m (16.4 ft) during the highest wind spee d level (3). Comparisons to von Karmin (1948), Kaimal (1972), and Davenport (1961) model spectra are also shown. Reasonable agreement between the data and model was found except for the lateral turbulence intensity, which was attributed to the limited rang e (~30 degrees) of the rotational vanes. Figure 4 12 N ormalized wind spectrum of Wind Level 3 (measured at 5 m) Results Pre wind test unsealed shingle surveys The percentage of pre wind test unsealed shingles ranged from 0% (9 of 17 roof specimens) up to 12% (one three tab roof specimen). Unsealed shingles on the laminate roofs exhibited adhesive failures between the sealant strip and overlapping shingle, whereas unsealed three tab shingles most frequently failed cohesively within the sealant strip. The location and length of unsealing on the partially unsealed shingles was more random than those observed in the roof surveys of Study 1 Partially unsealed hip shingles were found on all hip roof specimens prior to wind testing. The
85 location of the hip the same as that observed for the partially unsealed hip shingl es in the in situ roof surveys detailed in the previous section Wind performance of shingles installed in the field of the ro of Visible wind induced shingle damage included surface cracking, pull through of shingles over fasteners, and blow off. All damage initiated either from shingles identified as unsealed prior to wind testing the focus of this paper or pull through of e ave or rake roof edge shingles over edge fasteners. The percentag e of damaged roof area on the nine laminate roofs ranged from 0 2.5%, whereas the range on the eight three tab roofs was 1 55%. The laminate roofs sustained less damage than the three tab roo fs due to: (1) lower quantity of pre wind partially/full unsealed shingles and (2) better resistance to progressive lifting where eave and rake shingles suffered fastener head pull through One example of the consequence of pre wind test unsealed shingles is given for a three tab shingle hip roof specimen at the 45 wind orientation (Figures 4 13 and 4 14 ), selected because of its relatively high percentage of pre wind test unsealed shingles. Roofs with lower quantities of pre wind test unsealed shingles h ad similar statistical damage results as those to be presented. The pre wind test roof survey of the hip roof specimen shown in Figure 4 13 found fully or partially unsealed shingle tabs on 9% of the tabs located on the wind ward roof slopes Post test ana lysis of the high definition video captured during the wind tests showed the progression of damage that occurred during the wind test. Beginning in and blow off o ccurring near the ridge where unsealed shingles were adjacent to one
86 another (Figure 4 13 same roof as shown in Figure 4 14 ). Additional shingle tabs lifted throughout Wind Levels 2 4 due to their pre existing unsealing, causing damage to adjacent fully sealed shingles. A second analysis of the wind test footage was conducted to define the damage outcome of all shingle tabs located on the windward r oof slopes, defined in Figure 4 14 Each shingle tab was assigned a color and hatch pattern representing its pre wind test sealed or unsealed condition and post wind test damage outcome. The results of this analysis are shown in Figure 4 14 A statistical comparison of wind damage to pre wind test shingle tab condition is shown in Figure 4 15 Approximately 13% of the windward shingle tabs (147 tabs out of 1102) sustained some form of damage (e.g., blow off or surface cracking) 8% occurred on shingles identified pre wind as fully sealed and the remaining 5.5% occurred on shingles identified pre wind as partiall y/fully unsealed (Figure 4 15A ). Thus, nearly 60% of the pre wind unsealed tabs sustained some form of wind damage. Whereas, only 9% of the pre wind test sealed tabs sustained wind damage, all of which were initiated by either adjacent unsealed shingles or shingles that lifted at the eave. In summary, among the shingle tabs that were wind damaged during testing, none of the damage was initiated by pre wind test fully seal ed field shingle tabs (Figure 4 15 b). The results of this roof test demonstrates that pre existing unsealed shingles in the field of the roof pose a significant threat to both those unsealed shingles and the vulnerability.
8 7 Figure 4 13 Hip roof three tab shi ngle specimen pre and post test conditions with pre wind test unsealed shingles denoted by blue tape in the top left photo and the post test condition summarized in the roof plan at right. A B Figure 4 14 Shingle roof damage initiated by pre wind test unsealed shingles A) Blow off and B) cracking of shingles.
88 A B Figure 4 15 Statistical comparison of roof damage for the roof specimen shown in Figure 4 13 A) Comparison of the post wind test condition of windward shingle tabs stratified by pre wind test sealed/unsealed condition. B) Contribution of each potential initiator of shingle wind damage on the overall damage rate. Hip shingle wind performance Hip shingles blew off of all hip roof specimens. The quantity of blown off h ip shingles ranged from 41% to 86% of the total amount of hip shingles installed on the roof. Wind flow roughly parallel to the leading edge of the hip shingles produced the largest hip shingle loss (Figure 4 16) Loss of pre wind unsealed hip shingles ini tiated damage to sealed hip shingles upslope, as described below. A B C Figure 4 16 Characteristi c hip shingle blow off patterns. A) 0, B) 45, and C) 90 wind directions.
89 One example of the progression of hip shingle blow off is given in Figure 4 17. For this specimen oriented at the 0 wind direction (see Figure 4 10 ), the first loss of hip shingles occurred during Wind Level 1 wit h the blow off of two shingles The first shingle to lift was identified prior to the wind test as partially unseale d on its windward edge, and blow off occurred after the lifted shingle pulled through the fastener head. The blow off then progressed upwards during Wind Levels 2 and 3 on shingles that were adhered directly upslope from the initially blow n off shingle A pre wind partially unsealed hip shingle also blew off towards the bottom of the roof during Wind Level 3 causing progressive blow o ff through Wind Level 4 By the end of the wind test, only 10 out of the 50 hip shingles on the windward hip line remained on the roof. Damage vulnerability is, therefore, magnified for hip shingles that are unsealed on their windward edges, and loss of unsealed shingles instigates progressive failure of upwind adjacent sealed hip shingles. Figure 4 17 Progression of hip shingle blow off through the wind test sequence for specimen oriented at the 0 wind direction.
90 Discussion The results of the two studies demonstrate that unsealed asphalt shingles installed on the field, hip, and ridge locations of re sidential buildings are prone to unsealing over time, and that the unsealed condition increases the vulnerability of shingles to wind damage. As part of this research, 30 roofs in Florida and Texas were surveyed for unsealed shingles. All roofs contained u nsealed shingles with occurrence of unsealing reaching 86% of the total amount of installed shingles. The quantity of unsealed shingles installed in the field of the roof generally increased with roof age, whereas the quantity of unsealed hip and ridge shi ngles showed no discernible relationship to roof age. When unsealed shingles were observed in the field of the roof more than 99% of the shingles were unsealed along a partial length of their sealant strip line. The plane of fracture where unsealed cohe sively in the sealant strip and location of unsealing were consistent in the partially unsealed shingles, indicating a systematic failure of the sealant strip to remain adhered. The specific cause is unknown, but the observed increase in the total amount service age indicates that the effects of natural aging (Berdhal et al., 2008) influence the partial unsealing of field shingles. Blow off patterns of shingle roofs in previous hurricanes were similar to the spa tial patterns that result from partially unsealed field shingles, and experimental results from the wind tests performed at the IBHS Research Center demonstrate that the wind vulnerability of partially unsealed field shingles is greater than that of sealed shingles. Further work remains to identify the specific mechanism(s) that cause unsealing. This knowledge is critical for the development of appropriate retrofit
91 guidelines for existing shingle roofs and for future asphalt shingle design, manufacturing, a nd installation. For hip and ridge shingles, the installation technique combined with improperly placed nails in the sealant strip line are the most likely factors causing partial unsealing at the edge of the shingle. In the wind tests at the IBHS Researc h Center, blow off of hip shingles initiated from the lifting of pre existing partially unsealed hip shingles, then progressed up the roof slope. Retrofit solutions to seal the edges of hip and ridge shingles are available in FEMA (2012), but further work is necessary to quantify the long term durability and increased wind performance of the proposed retrofit.
92 CHAPTER 5 WIND RESISTANCE OF N ATURALLY AND ARTIFIC IALLY AGED ASPHALT SHINGLES T his chapter details t wo interrelated studies that measur ed the bond strength and resultant failure modes of naturally and artificially aged asphalt shingles. Results provide knowledge on the expected performance of fully adhered shingles in service and the predictive capabilities of the ASTM D7158 (ASTM 2011c) and ASTM D6381 (ASTM 2008b) test methods for asphalt shingle wind resistance. In the first study, asphalt shingle specimens were artificially aged using two techniques and wind uplift resistance was measured in a portion of the population at discrete inter vals during aging. In the second study, wind uplift resistance was measure in situ on shingle roofs installed on four Florida homes with more than nine years of service. A literature review of aging effects on asphalt shingles is presented first, fol lowed by the research objective s and study results. The chapter concludes with a discussion of the combined results. Aging of Asphalt Shingles In service asphalt shingles are exposed to diurnal cycles of temperature and sunlight, fluctuations in humidity, intermittent rainfall, and wind ( Berdahl et al. 2008 ). Changes in referred to as aging. Prior research on shingle aging has primarily focused on the lt, while other materials in the shingle have received less attention. Reprinted with permission from Dixon, C.R., Masters, F.J., Prevatt, D.O., and Gurley, K.R. (2013 b Wind uplift resistance of artifically and naturally aged asphalt shingles Jour nal of Architectural Engineering [Article Submitted for Review]
93 Heat is the primary mechanism of aging in asphalt coatings (Marechal et al. 1983; McLintock 1991; Shiao et al. 2003a; Surffleton and Still 1990; Terrenzio et al. 1997; Warford 1998; Wright 1 979; Wypych 1995) as well as asphalt paving (Lia ng and Lee 1996). A heat driven aging model for asphalt shingles was first proposed by Terrenzio et al. (1997) and later expanded upon by Shiao et al. (2003a). The underlying theory is that heat promotes the diffusion of lower molecular weight oils out of the asphalt coating where they are volatized or washed away due to photo oxidization. Oxygen then diffuses into the asphalt coating to produce heavier molecular weight heptane insoluble molecules (asphaltenes ). Given a constant temperature, formation of heptane insoluble molecules increases rapidly at the onset of aging before reaching an equilibrium state. This thermal oxidization produces a stiffer, crack suscep tible coating (Terrenzio et al. 1997). Photo o xidation is a secondary aging mechanism in asphalt coatings, mitigated by ultraviolet (UV) light absorbing roofing granules embedded on the top surface of the shingle coating ( Terrenzio et al. 1997; Dutt 1983 ). A loss of granules can occur over time, cause d by rainwater running over the surface of the shingle or the loss of embedment in thermally oxidized asphalt ( Terrenzio et al. 1997). Once exposed, energetic UV photons diffuse into the top surface of the asphalt, forming sulfur oxygen groups and increase d concentrations of carbonyl groups (Berdahl et al. 2008). The now water soluble top layer is then vulnerable to removal as water passes over the shingle. Once removed, a new layer is exposed for the photo oxidation process to repeat until all coating asph alt is displaced (Terrenzio et al. 1997).
94 The w ind resistance of the shingle is based upon the weakest element of cohesive or adhesive strength in or between the constitutive materials in the load path (Figure s 5 1 and 5 2 ). Although the literature cited widely recognizes that heat and UV light drive chemical and physical changes in the shingles, the potential effects of these For example, the asphalt coating is known to embrittle over exposure time, how does this affect its cohesive or adhesive strength when subjected to wind uplift? The purpose of this chapter is to fill this knowledge gap. Figure 5 1. Wind pr essures on shingle roofing Figure 5 2. Cross titutive materials. Research Objectives The overarching objective studies was to identify whether and to what extent aging reduces the bond strength of the sealant strip. The first study subjected new asphalt shingles to two industry accepted artificial aging protocols: (a) up
95 to 5376 hr of ASTM D5869 04a (ASTM 2004) dark oven heat at 70C (158F) (henceforth, Thermal) and (b) up to 3360 hr of ASTM D4799 ( ASTM 2008a ) continuous cycles of UV light, heat at 70C (158F), and water spray (henceforth, UV Thermal life. Based upon results of Terrenzio et al. (1997) use of the Thermal and UV Thermal Water protocols, coating molecular weight and ~ 150 % modulus (i.e., stiffness). Additionally, 5376 hr of aging produced an equivalent change in asphaltene content to two years of natural aging in Houston, TX. However, the materials used in the current study were not precisely the same as those used in the Terrenzio et al. (1997) study At discrete intervals during each exposure, the uplift resistance of the sealant strip composite bond was measured using the ASTM D6381 (ASTM 2008b) mechanical uplift test, and compared to the baseline mean uplift resistance measured on the new shingles. Modes of failure in the uplifted sh ingles were recorded and stratified by exposure time and type. Total wind uplift resistance across the exposure time was computed and compared to design wind loads specified in ASTM D7158 (ASTM 2011c) In the second study uplift measurements were perform ed in situ on four shingle roof systems installed on four Central Florida homes, each with greater tha n nine years of natural aging. The t otal wind resistance of three of the four systems was calculated via the approach described in Peterka et al. (1997) a nd compared to their theoretical design wind loads
96 Study 1: Wind Uplift Capacity of Asphalt Shingles Subjected to Artificial Aging Experimental Setup Shingle specimen specifications Test specimens were ASTM D7158 Class H (ASTM 2011c) three tab fiberglass reinforced asphalt shingles produced by three manufacturers henceforth, Product s A, B, and Products B and C were purchased from a contract supply store. Within each product, shingles were sourced from the same manufacturing batch to mitigate variability in the t strip chemistry (Shiao et al. 2003b). All shingles were stored in indoor conditioned space between their time of acquisiti on and initiation of testing. Test specimens were constructed following Section 7 in ASTM D6381 (ASTM 2008b) The number of prepared specimens was even split between Procedure A and Procedure B. Specimens were cut to their required dimensions (Figure 5 3 ), then labeled on the lower surface of the bottom shingle with a unique identification number corresponding to their product manufacturer, bundle number, shingle number within each bundle, and specimen number within the individual shingle. The specimen loca tion in the aging chamber and exposure time prior to mechanical uplift testing were randomized. Thermal aging chamber specifications and protocol The thermal aging chamber consisted of an Excal ibur COM2 forced air dark oven [ ASTM E145 (ASTM 2011b), Type IIB]. high by 0.7 m (2.4 ft) wide by 0.6 m (2.2 ft) deep, and specimens are placed on one of forty vertically distributed open wire metal racks. T emperature was measured with four
97 Omega model 5TC GG T 24 72 const antan copper (Type T) thermocouples located in the center of racks 2, 13, 26, and 38. Temperature data w ere recorded using a National Instruments C DAQ module. The oven has a 600 specimen capacity. Additional specimens were added to the oven after removing specimens scheduled for mechanical uplift testing. A B Figure 5 3 ASTM D6381 specimens A) Procedure A specimen and B) Procedure B specimen. Shingle specimens were exposed to a continuous 70 C (158F) 3 C (5 F) air circulated heat. At the exposure time intervals shown in Table 5 1 ten shingle specimens per test procedure per shingle product were removed from the oven. Product C was only exposed to 3360 hr. One exception to this procedure occurred at the final t ime interval, where 20 specimens per test procedure per product were removed and tested Once removed, the shingles were air cooled to 21C (70 F) 3 C (5 F) then tested for their ASTM D6381 mechanical uplift resistance (ASTM 2008b)
98 Table 5 1 Exposure times where ASTM D6381 tests were performed Exposure Time (hr) 16 84 168 252 336 504 840 1176 1512 2016 2688 3360 4032 4368 4704 5040 5376 Thermal UV Thermal Water UV Thermal Water aging chamber specifications and protocol UV Thermal Water exposure was performed in a custom built weathering chamber that conformed to the requirements for accelerated weathering of bituminous materials in ASTM D4799 ( ASTM 2008a ) ASTM G151 (ASTM 2000), and AS TM G154 (ASTM 2006) (Figure 5 4 ). The chamber has plan dimensions of 1.4 m by 4.9 m (4 ft 6 in by 16 ft) and a sloping profile ranging from 0.3 m to 0.5 m (1 ft to 1 ft 6 in). The chamber houses 240 ASTM D6381 ( ASTM 2008b ) test specimens. Environmental con ditions inside chamber were controlled, monitored and recorded using a National Instruments Labview 8.5 and a National Instruments CompactDAQ data acquisition system. The UV light system consists of 1.2 m (4 ft) long UVA 340 lamps manufactured by Q Lab, located 102 mm (4 in) above the specimens at an on center spacing of 102 mm (4 in) to ensure irradiance uniformity (Figure 5 5 ). The lamps produced peak irradiance at a wavelength of 340 nm, and were powered by fluorescent light ballasts using an overdrivi ng technique to produce a maximum irradiance at the specimen level of 0.72 W/m 2 at 340 nm. The irradiance output was 0.04 W/m 2 (at 340 nm) greater than the irradiance of the sun at noon on a clear day ( Fedor and Brennan 1996 ). The lamps were supplied with the constant current throughout the experiment, irrespective of the irradiance output.
99 Figure 5 4 UV Thermal Aging chamber components. ASTM G151 (ASTM 2000) Section 5.1.2 specifies that the irradiance at any point in the specimen area must be within 70% of the maximum irradiance measured in the same area. Irradiance produced by the lamps is inversely correlated to the ambient air temperature near the lamp and over time, the irradiance output can decrease due to a Therefore, the irradiance of the UV light system was periodically recorded at 25 mm (1 in) increments using an Apogee SU 100 radiometer attached to a gantry affixed to the centerline of the chamber. Prior to the experiment, the Apogee radiometer was calibrated to a Q Lab CR 10 radiometer. Figure 5 5 shows the irradiance data measurement captured at the initiation of the aging experiment Uniformity of the irradiance thr o ughout the chamber [ASTM G151 (ASTM 2000) Section 5.1.2] was met during the entire test H owever the average irradiance values decreased
100 from 0.72 W/m 2 to 0.20 W/m 2 at the end of the experiment. The uniformity along each confirmed prior to the initiation of the experiment. Figure 5 5 Measured irradiance, plan view, and temperature time history of one cycle. To ensure uniformity of the water spray rate over the specimen area spray nozzles were located 152 mm (6 in) abo ve the specimen surface between the lighting system at an on center spacing of 305 mm (12 in) (Figure 5 4 ) The temperature of water spray was approximately 21C (70C). A n in line water filtration system meeting the specifications of ASTM D4799 Section 22.214.171.124 ( ASTM 2008b ) was used to reduce the concentration of cations, anions, organics, and silica in the water used for the water spray.
101 Heat was supplied to the chamber interior via ducting with an in line blower and duct heater to create a forced air oven condition During the experiment copper con stantan (Type T) thermocouples located at the specimen level continuously monitored and recorded the chamber temperature (Figure 5 5) Additional thermocouples were installed above the specimen area in the center of the chamber to provide feedback to the air heating control system. Figure 5 5 presents a temperature time history of one complete 5 hour 15 minute cycle used in the aging process Starting at 0 minutes t he chamber began to heat after greatest rate of heating due to its location just below the entrance of the heating duct into the chamber. The entire heating process took approximately 7 0 minutes to reach temperature stabilization. The chamber then held a constant temperature for the remaining 220 minutes of the heating/light cycle. At 300 minutes into the cycle, the heating and light system shut off and the water spray cycle began. Once completed, the cycle restarted. The two extreme end thermocouple temperatures of TC1 and TC4, are approximately 10C and 7 C less than TC2. The observation for TC1 is the result of an he lab due to an opening at the end of the chamber that provides access for the radiometer. Thus, the extreme ends of the specimen area were moved away from the chamber ends to ensure that the air temperature on the sp ecimens was within 3C ( 5F ) of th e set point temperature of 70 C (158 F), as required by ASTM G1 51 Section 5.2.8 (ASTM 2000)
102 Due to the capacity limits of the weathering chamber, only Products A and B were subjected to UV Thermal Water exposure. S pecimens were conditioned in the force d air dark oven for 16 hours at 70C (158F) 3 C (5 F) p rior to their UV Thermal Water exposure This e nsure d tab seal activation resulting in bonded shingles prior to their exposure to water spray Following this, a ll specimens were inspected by hand fo r adhesion, then placed inside the UV Thermal Water chamber and exposed to continuous cycles of five hours of 70C (158F) heat with UVA 340 ultraviolet light, followed by 15 minutes of water spray At the times shown in Table 5 1 twenty predetermined shi ngle specimens per test procedure per shingle product were removed from the chamber. T he entire specimen inventory could not fit in the chamber at one time; therefore, specimens scheduled for removal at 168, 2688, and 3360 hr were placed in the chamber fir st. After removal of the 168 hr exposure specimens, 2016 hr specimens were placed in the camber Following 2016 hr removal specimens scheduled for 840 hr of exposure were placed in the chamber Similar to the thermal exposure procedure, specimens were air cooled to 21C (70 F) 3 C (5 F) after their removal from the chamber, then tested for their mechanical uplift resistance described in the next section. ASTM D6381 Mechanical Uplift Test Procedure The ASTM D6381 test method consists of two test procedures (A and B) that simulate the individual wind loading components of asphalt shingles (ASTM 2008b) The results of each test ( R A and R B ) were combined using rational engineering analysis [per ASTM D7158 (ASTM 2011c)] to genera te the total wind uplift resistance, as shown in Equation 5 1 The force variables shown in the equation represent the windward ( F F ),
103 leeward ( F B ), and total ( F T ASTM D7158 (ASTM 2011c) (5 1) Procedures A and B simulate a peel sealant strip, respectively (Figure 5 6 ). In product approval, this method is performed on ten specimens per test pro cedure after they are conditioned in a forced air dark oven at a 57 to 60C (135 to 140F) for 16 hr and air cooled to 21C (70 F) 3 C (5 F) Figure 5 6 ASTM D6381 test apparatus, setup, and uplifted specimen. Upon removal from the aging chamber, all shingle seals were inspected for adhesion along the complete sealant strip length Shingles lacking full adhesion were not observed in the experiment. As per ASTM D6381 (ASTM 2008b) all P rocedure B test specimens w ere affixed with an aluminum section using a low exotherm epoxy and cured overnight prior to mechanical uplift testing. Procedure A specimens were invalidated due to broken seals caused during the process of attachin S ubsequent Product C Procedure A specimens were affixed with a small steel channel
104 prior to uplift testing, after ASTM D6381 Section 7.7.3 (ASTM 2008b) It was not necessary to use the steel channel o n Products A and B. Uplift tests were performed in an air conditioned laboratory on shingle specimens with a 21C (70F) 3C (5F) surface temperature. The test apparatus was an Instron 3367 Un iversal Testing Machine (UTM) set to a n ASTM D6381 (ASTM 2008b) specified constant uplift displacement rate o f 127 mm/min (5 in/min). The test stopped when: (a) the two shingles lost adhesion along a complete length of their sealant strip line or (b) the instantaneous force measurement was less than 30% o f the maximum force measured during the test. Case (b) frequently occurred when the mode of failure was at to glass mat interface (rightmost photo of Figure 5 6 ). Following the test, the operator recorded the peak force applied to the shingle and the shingle temperature. Uplift resistance of the specimen was defined as the peak force applied to the shingle during the uplift test. The mode of failure was determined by an analysis of high resolution post test photographs o Results ASTM D6381 Procedures A and B uplift resistance Figure s 5 7 through 5 9 present the complete dataset of ASTM D6381 test results stratified by product, mechanical uplift test procedure, artificial aging protocol, and exposure time. The mean uplift resistance, one standard deviation, and 95% confidence interval of the mean are provided. The black dashed boxes surrounding individual confidence intervals indicate where mean resistance was significantly lower than mean res istance measured at the 16 hr Thermal exposure period.
105 Figure 5 7 ASTM D6381 test results for Product A Figure 5 8 ASTM D6381 test results for P roduct B
106 Figure 5 9 ASTM D6381 test results for P roduct C A t test was performed at to determine whether a statistically significant degradation of strength occurred as a function of aging (Ott and Longnecker 2004). Product A was the only product where statistically significant reductions in mean up lift resistance were observed. For Product A Procedure A, significant reductions were observed at 12 of 16 Thermal test intervals, and four of five UV Thermal Water test intervals. Three distinct phases occurred over the 5376 hr Thermal exposure in Product A Procedure A: 1. t = 16 840 hr: stable mean uplift resistance 2. t = 1176 4032 hr : decreasing mean uplift resistance 3. t = 4368 5376 hr: stable mean uplift resistance at a decreased capacity relative to the 16 hr mean resistancde
107 By the last test interval in the two aging methods for Product A, mean Procedure A uplift resistance reduced by 36% in Thermal ( p value = 0.00) and 39% in UV Thermal Water (p value = 0.00) when compared to t = 16 hr mean Procedure A resistance. For Product A Procedure B, significant reductions were observed at 3 of 16 Thermal test intervals, and all five UV Thermal Water test intervals. The reduction in uplift resistance observed at the final Procedure B test interval was 15% in Thermal ( p value = 0.00 ) and 20% in UV Therm al Water ( p value = 0.00 ). For Products B and C, mean Procedures A and B resistances measured at t > 16 hr were either statistically similar to or greater than their corresponding t = 16 hr mean resistance. Product C had the least variability, with all 11 test intervals statistically similar to t = 16 hr in Procedure A mean resistance, and 7 of 11 in Procedure B. For Product B Procedure B, similarity to t = 16 hr was only observed in two of the sixteen Thermal test intervals, and one of the five UV Thermal Water test intervals. A 53% increase in mean Procedure B resistance was observed in Product B between the 16 hr and 84 hr test intervals (p resistance increased by 9% (p value = 0.534). Terrenzio et al. (1997) observed similar changes in chemical and physical structure of the coating asphalt subjected to the Thermal and UV Thermal Water protocols. By extension, mean uplift resistance in specimens exposed to the Thermal protocol should be statistically si milar to specimens exposed to the UV Heat Water protocol at the five common test intervals (Table 5 1 t test ( ) was used in the analysis. The results shown in Table 5 2 indicate no strong trends to reject the hypothesis. Statistical simil arity was observed at 8 of the 20 comparisons (40%).
108 Product B Procedure B contained the only occurrence of consistent outcome across all five test intervals dissimilar (p Table 5 2 M ean resistance in Thermal and UV Thermal Water m ethods Product ASTM D6381 Test Procedure Mean Resistance at Common Exposure: Thermal = UV Thermal Water? (p value 1 ) 168 hr 840 hr 2016 hr 2688 hr 3360 hr A A No 2 (0.00) Yes (0.81) Yes (0.09) Yes (0.16) No 2 (0.01) B No 2 (0.00) No 2 (0.02) Yes (0.77) Yes (0.43) No 2 (0.01) B A Yes (0.26) Yes (0.90) No 3 (0.00) Yes (0.19) No 3 (0.00) B No 2 (0.00) No 2 (0.00) No 3 (0.01) No 2 (0.00) No 2 (0.03) 1 t 2 Thermal > UV Thermal Water 3 Thermal < UV Thermal Water Failure modes in uplifted shingles Five modes of failure were observed in the mechanically uplifted shingles, referring to Figure 5 10 : 1. Adhesive failure at the b ottom shingle asphalt coating to glass mat interface 2. Cohesive failure in the sealant strip 3. Adhesive failure at the seala nt strip to top shingle interface 4. Cohesive failure in the bottom shingle asphalt coating 5. Adhesive failure at the top shingle asphalt coating to glass mat interface The majority of tested shingles had more than one mode of failure along a given sealant to quantify the relative contribution of each failure mode. As shown in top two photos of Figure 5 10 if a mode contributed to more than half of the sealant strip area, that was unable to be determined (i.e., bottom photograph of Figure 5 10 ), the failure mode
109 Figure 5 10 Example failure modes observed in mechanically uplifted shingles. Figure 5 1 1 displays the results of this analysis in Product A Procedure A as a stacked bar chart of the distribution of failure modes stratified by exposure time. Only the dominant and mixed modal failure modes are shown. Recall, this particular product and test pr ocedure combination had three distinct phases of uplift resistance across Thermal exposure time. The same three phases were observed in the failure mode distribution. From t = 16 840 hr, the distribution is mostly composed of adhesive failures at the bot tom shingle to glass mat interface and mixed modal failures. From t =
110 1176 3360 the amount of bottom shingle interfacial failures decreased, while the amount of cohesive failures in the top shingle coating increased. Beyond t = 4032 hr, the majority of f ailures were cohesive in the top shingle coating. Dominant failures at the bottom shingle to glass mat interface were not observed. This indicates that the period, resulti ng in a reduction of the shingle uplift resistance. Figure 5 11 Distribution of failure modes on Product A Procedure A. Product A Thermal Water distributions show poor agreement at t = 168 hr and 840 hr and close agreement at t > 840 hr. Product A Procedure B shows better agreement between the two aging methods at all common time intervals (Figure 5 12 ). The dissimilarity in average failure mode distribution in nd B results agrees with Shiao et al. (2003 b ).
111 Figure 5 12 Distribution of failure modes on Product A Procedure B Figure 5 13 Distribution of failure modes on Product C. The distribution of failure modes in Products B and C were more consistent over time than Product A Procedure A, as demonstrated in Figure 5 13 for Product C. The
112 same observation was made in the uplift resistance analysis. Product B had a similar average failure mode distribution in Procedures A and B: ~50% adhesive failures at the bottom shingle asphalt coating to glass mat interface, ~35% mixed modal, and minor contributions from the remaining failure modes. As shown in Figure 5 13 distributions showed strong similarity between Procedures A and B with adhesive fail ures at the bottom shingle asphalt coating to glass mat interface representing roughly 75% of the failure modes observed at all time intervals. ASTM D7158 total wind uplift resistance Table 5 3 gure 5 1 ), ASTM D7158 leeward and windward mean differential pressure coefficients (ASTM 2011c) and ASTM D7158 Class H required total resistance (ASTM 2011c) calculated by Eq. 5 2 derived from ASTM D7158 (ASTM 2011c) The manufacturers of Products A and C provided the authors with their respective mean differential pressure coefficients, coefficients were used as a proxy for Product B due to the match in exposed length di mensions. The ASTM D7158 Class H (ASTM 2011c) required resistance shown in Table 5 3 represents the peak design load exerted on a 95 mm (3.75 in) width of the roof wind speed location of a s teep slope roof (Peterka et al. 1997), and subjected to an ASCE 7 02 (ASCE 2002) 67 m/s (150 mph) peak 3s gust ( V ) in an Open Country exposure. Recall, all products used in the experiment were certified as ASTM D7158 Class H wind resistant (ASTM 2011c) (5 2)
113 Table 5 3 Specimen dimensions, ASTM D7158 differential pressure coefficients, and ASTM D7158 required resistance. Product Exposed Length Dimensions (mm) ASTM D7158 Differential Pressure Coefficients ASTM D7158 Class H Required Total Resistance (N) L 1 L 2 DCp(f ) DCp(b ) A 32 95 0.48 0.14 25 B 32 95 0.48 1 0.14 1 25 C 25 102 0.53 0.09 20 1 Coefficient from Product A The total uplift resistance ( R T ) was calculated using Equation 5 1 As with the mean differential pressure coefficients, the design forces in Products A and C were Product B. The resistance variables in Equation 5 1 repres ent the experimental values obtained by ASTM D6381 (ASTM 2008b) Procedure A ( R A ) and Procedure B ( R B ). ASTM D7158 stipulates that R A and R B equal the average resistance of ten samples per test procedure. Thus, R T in ASTM D7158 (ASTM 2011c) represents the a verage wind resistance to a peak design wind load. Table 5 4 presents the mean and minimum total uplift resistance for each product Class H required resistance (ASTM 2011c) Mean R T in each product was computed at each time interval, and then averaged across all time periods to produce the mean results shown in column three of Table 5 4 From this analysis, the three products would be expected, on average, to resist five to eleven ti mes the load required to meet ASTM D7158 Class H (ASTM 2011c)
114 Table 5 4 Mean and lowest measured wind uplift resistance vs. ASTM D7158 Class H required wind resistance Product Aging Method Mean Resistance Over All Time Intervals (N) Lowest Uplift Resi stance ASTM D7158 Class H Required Uplift Resistance (N) Value (N) Time (hr) A Thermal 146 86 5376 25 UV Thermal Water 123 61 2016 B Thermal 168 1 70 1 16 1 25 UV Thermal Water 162 1 109 1 16 1 C Thermal 226 102 16 20 1 mean differential pressure coefficients. The minimum total wind resistance of the test specimens at each time interval was obtained by reassigning R A and R B as the minimum Procedures A and B test results at each time interval. R T was then recalculated by Equation 5 2 using the reassigned R A and R B variables. Table 5 4 provides the lowest value obtained across the complete exposure time stratified by product and aging method and the time corresponding to this minimum total resi stance. The lowest value of total uplift occurred uplift occurred at later intervals in both aging methods. Product A exposed to UV Thermal Water had the lowest value o f uplift (61 N) among the three products. This resistance is 2.4 times the ASTM D7158 Class H design resistance (ASTM 2011c) Discussion The central question motivating this study was the potential relationship between aged uplift capacity and previously reported differences in new and aged shingle wind performance. Six to seven years of natural exposure was a breakpoint between low ( ~5% ) and high ( ~65% ) damage rate to shingle roofs in recent hurricanes (Liu et al., 2010). The results of this experiment in dicate that a reduction in mean wind resistance
115 may occur in shingle roofs as result of heat driven aging. Yet, the capacity of aged shingles at the reduced resistance exceeds design level both in a mean and minimum sense. 5376 hr of Thermal aging was roug hly equivalent to two years of natural exposure in Houston, TX (Terrenzio et al., 1997); t herefore it is reasonable to assume that six years of natural aging was not achieved in either aging protocol. The most significant finding of the experiment involve s the two distinct responses to artificial aging observed in the three shingle products evaluated. Product A demonstrated a statistically significant reduction in mean uplift resistance over time. Conversely, all mean resistance measurements of Products B and C were either statistically similar or greater than the mean resistance of the baseline new installation capacity at 16 hr of Thermal exposure. We speculate that variation in strength across specimens is relate d tion which was not studied during these experiments Further, the wind uplift capacity of all three products at all time intervals and at all possible combinations of Procedures A and B resistances exceeded the required capacity to meet ASTM D7158 Class H (ASTM 2011c) ASCE 7 02 (ASCE 2002) design wind of 67 m/s (150 mph) in open country terrain. The reduction in the bond strength did not exceed the reserve capacity, which suggests that increasing the strength requirement may not be warranted. Additional research is required to determine if this conclusion is extensible to the entire roof system, which is comprised of thousands of individual shingles. Failures were most frequently observed either at the interface of the top coating of the bottom shingle t o glass mat or cohesively in bottom coating of the top shingle. Cohesive failures in the sealant strip and interfacial failures at the sealant strip to top
116 shingle asphalt coating were rarely observed. Thus, the cohesive strength of sealant strip and the a dhesive strength of the sealant strip to top shingle interface were rarely the weak link in the composite. Gel permeation chromatography, SARA fractionation, and dynamic shear rheometer tests were performed on the three products exposed to the Thermal prot ocol at ten intervals. A future paper will address these results. Study 2: Naturally Aged Shingle Wind Uplift Resistance Test Sites In situ ASTM D6381 mechanical uplift tests (ASTM 2008b) were performed on four existing homes in Central Florida that were p rovided by the State of Florida Departmen t of Emergency Management The list of the homes is given in Table 5 5 along with their respective shingle roof age shingle type and the number of Procedure A and B mechanical uplift tests The four roofs were also inspected for non adhered asphalt shingles prior to the uplift tests. Results of the four roofs tested are discussed below. The complete dataset can be found in Chapter 4 of this dissertation. Table 5 5 Test site location, age, type, and quantity of ASTM D6381 tests Test Site ID Test Site (Florida city) Shingle Roof Age (years) Shingle Type ASTM D6381 Test Procedure Number of Tested Specimens ORM 01 Ormond Beach N/A 1 Three Tab A 25 ORM 02 Ormond Beach 13 Three Tab A 39 B 40 ORM 03 Ormond Beach 10 Three Tab A 36 B 40 ORG 01 Orange City 9 Laminate B 81 1 Building permit records indicate that the shingle roof is greater than 13 years old. Roof age was defined as the year of uplift test minus the year of installation, as determined by permit records. One roof age could not be determined (ORM 01);
117 however, the lack of permits from 1999 to 2012 indicates that its age is in excess of 13 years. The wind resistance classifications of the four products are unknown. Portable Mechanical Uplift Apparatus The author and one additional graduate student designed and constructed a Portable Mechanical Uplift Apparatus (PMUA) to perform in situ ASTM D6381 mechanical uplift tests (ASTM 2008b) Hardware components include a Tritex TLM20 electric linear actuator affixed to an aluminum frame an Xplor iX104C tablet computer running a custom National Instruments Labview 2010 program, and an 890 N (200 lb) capaci ty Futek model LRF 350 in line load cell (Figure 5 14 ). Control and feedback o f actuator arm position and velocity are sent and received as digital input and output A National Instruments 6211 data acquisition hardware mo unted on the side of the frame translates the signals between the computer, load cell, and actuator. Figure 5 14 Portable Mechanical Uplift Apparatus components. In Situ ASTM D6381 Specimen Preparation and Test Procedure S pecimens were randomly selected throughout the roof area, and then checked by hand for adhesion prior to their preparation for uplift testing. For Procedure A
118 specimens, two 114 mm (4.5 in) upslope cuts were made in the upper shingle strip starting from the leading edge of the s strip Additional upslope cuts we re made in the adjacent shingle material to prevent inte rference between the test clamp and external shingles during testing. For Procedure B specimens, two 38 mm (1.5 in) upslope cuts spaced 95 mm (3.75 in) apart were made in the upper shingle starting then isolated from the surrounding shingle by a 95 mm (3.75 in) horizontal cut to form a rectangular piece of top shingle with dimensions of 95 mm (3.75 in) wide by 38 mm (1.5 in) long. An ASTM D6381 (ASTM 2008b) specified aluminum tee shaped section was adhered on the top shingle and cure d overnight prior to uplift testing ASTM D6381 Procedures A and B had similar test procedu res. First, the PMUA was placed over the specimen and positioned to align the arm with leading edge of the shingle (Procedure A) or the aluminum tee (Procedure B). The load cell reading was then zeroed to account for the self weight of the clamp (Procedure A) or hook and chains (Procedure B). As with the thermal aging experiment, the Procedure A clamp was affixed to the leading edge of the specimen, while, in Procedure B, one hook was placed on either side of the aluminum tee. The mechanical uplift test was then initiated using the ASTM D6381 (ASTM 2008b) specified 2.11 mm/s (0.083 in/s) constant displacement rate, and uplift ceased once the specimen lost adhesion. Personnel then visually inspected and recorded the failure mode of the test specimen along wit h the maximum force produced during the test.
119 Results In situ ASTM D6381 mechanical uplift resistance Figure 5 1 5 depicts the in situ ASTM D6381 (ASTM 2008b) test results stratified by test site and test procedure. The statistical parameters shown in this figure are mean uplift resistance, one standard deviation from the mean, 95% confidence interval of the mean, and the range of uplift resistance data. The mea n Procedure A uplift resistance on ORM 01 shingles was the lowest of the three sites where Procedure A tests were t test comparing the three Procedure A sites shows that ORM 01 shingles had a statistically significantly lower mean upli ft resistance compared to ORM 02 and ORM 03 at a 95% confidence. ORM 01 shingles are the oldest in the data set. Figure 5 15 In situ ASTM D6381 test results Excluding ORM 1, the in situ test results show similarity in mean resistance between roofs. P aired comparisons of mean Procedure B uplift resistance in ORM 02, ORM 03, and ORG t
120 test). Mean Procedure A uplift resistance in ORM 02 and ORM 03 are statistically equivalent with 95% confidence. Failure modes The distribution s of observed failure modes at the four sites are shown in Figure 5 16 ORM 01 and ORM 03 Procedure A specimens predomi nantly failed cohesively within the sealant strip, a mode that was rarely observed during the artificial aging experiments. For ORM 02 Procedure A specimens, interfacial failures at the top shingle coating to glass mat occurred in 67% of the specimens. Thi s was also rarely observed in the artificially aged shingles. The in situ and artificially aged shingles were produced with different asphalts, sealant strips, and fiberglass mats so this finding may be related to materials more than differences in aging t echniques. The distribution of failure modes in Procedure B was more similar than Procedure A amongst the four roofs. Mixed modal failures were the most prevalent for Procedure B than Procedure A, occurring at a rate ranging from 58 to 85% of in situ speci mens. Figure 5 1 6 Distribution of failure modes for in situ ASTM D6381 tests.
121 ASTM D7158 total wind uplift resistance Figure 5 17 compares the measured wind uplift resistance of ORM 02, ORM 03, and ORG 01 to the uplift resistance required by each system to meet ASTM D7158 Class H (ASTM 2011c) The mean resistance, standard deviation, 95% confidence The measured total wind uplift resistances of ORM 02 and ORM 03 were ca lculated for all combinations of Procedure A and B test results using Equation 5 1 For ORG 01, Procedure B uplift resistance was the only data required to satisfy the Peterka et al. (1997) uplift model due the location of ORG 01 sealant strip relative to the leading edge of the shingle. ORG 01 wind uplift capacity equaled its Procedure B resistance. ORM 01 wind resistance could not be calculated because Procedure B tests were not differential pressu re coefficients (ASTM 2011c) differential pressure coefficients published in Peterka et al. L 1 and L 2 ), Peterka et al. (1997) differential pressure coefficients ( DC p ), and ASTM D7158 Class H required uplift resistance, calculated by Equation 5 2 (from ASTM 2011c) are provided in Table 5 6 Table 5 6 Test site differential pressure coefficients, length dimensions, and ASTM D7158 Class H required wind uplift resistance. Site ID Windward DC p 1 Leeward DC p 1 L1 2 (mm) L2 2 (mm) ASTM D7158 Class H Required Wind Uplift Resistance (N) ORM 02 0.4 0.1 32 95 20 ORM 03 0.4 0.1 32 95 20 ORG 01 0.8 0.1 13 127 19 1 Peterka et al. (1997) positive values act upwards on the shingle 2 Refer to Figure 5 1
122 ORM 03 contained the only combination of Procedure A and B resistance (18 N) below ASTM D7158 Class H resistance (20 N) (Figure 5 16 ). All other combinations of Procedure A and B uplift resistance on ORM 03 and all combinations on ORM 02 and ORG 01 would be expected to resist an ASCE 7 02 (ASCE 2002) design 3 s wind speed of 150 mph in Open Country exposure. The measured mean wind resistances of ORM 02 and ORM 03 were six times greater than ASTM D7158 Class H required (ASTM 2011c) while ORG ured resistance was eight times greater than ASTM D7158 Class H required (ASTM 2011c) Figure 5 1 7 Wind resistance of naturally aged shingles vs. ASTM D7158 Class H required resistance. Discussion The dataset of naturally aged homes represents a small sampling of the residential building stock, however these results strongly suggest that individual shingles exposed outdoors for more than nine years have sufficient residual capacity to resist the highest design wind loads. This finding only applies to sh ingles with fully
123 adhered sealant strips, however. Another experiment conducted on these roofs reveals a significant problem with the consistency of sealant strip adhesion along the complete The four roofs were survey ed by hand for the presence of adhesion on all shingles installed in the field, hip, and ridge locations of the roofs. Only those with full adhesion were mechanically tested for uplift capacity. On three of the four homes, shingles were found without seala nt strip adhesion along a partial length of their leading edge (i.e., partially unsealed). The consistent location of unseal and failure mode where unsealed indicate a systematic failure of the sealant strip. Partially unsealed shingles were found on 7.7% of ORM ORG consisted of 27 inspected roofs in Florida, ranging in age from two months to over twenty years. Many had a signifi cant portion of their shingles partially unsealed. The most extreme case was a roof with over 86% of its shingles lacking full adhesion. Significantly, partially unsealed shingles were only found on Florida roofs with greater than six years of natural expo sure, aligning with the performance breakpoint reported in Liu et al. (2010). Discussion of Combined Results The experimental objective was to identify whether and to what extent aging reduces the wind uplift capacity of fully sealed asphalt shingles. For the first experimental investigation, new ASTM D7158 Class H (ASTM 2011c) three tab shingles from three different manufacturers were artificially aged by two methods, and wind uplift resistance was quantified in a sample of the population at pre determine d exposure times. For the second experimental investigation, wind uplift resistance was measured
124 in situ on four shingled roofs in Central Florida exposed to natural aging for more than nine years. Significant finding are detailed below. 1. Weather induced r eductions in uplift resistance cannot be ruled out as a potential contributor in wind damage to shingle roofing, however this study found high uplift resistance in naturally and artificially aged shingles relative to design level load. Reduced uplift resis tance caused by aging appears to be a secondary contributor. The major conclusion is that shingles with fully adhered sealant strips provide excellent long term resistance to uplift loads. 2. (ASTM 2011c) is limited to the resistance of new (non aged) shingle products to design level wind force. Given the results of the first experiment, the resistance measured on a new shingle should not be extrapolated to resistance once in service and weathered. Additio nal exposure times are recommended beyond the 16 hr condition time currently specified in ASTM D7158 to improve the predictive capabilities of ASTM D7158 (ASTM 2011c) Chapter 4 of this dissertation presents a more likely primary contributor to shingle roo f wind damage i.e. the existence of partially unsealed shingles prior to wind. The study demonstrated that (a) partially unsealed shingles have a significant reduction in uplift capacity, and (b) their observed occurrence is a significant percentage of th e total roof cover that becomes more prominent with age.
125 CHAPTER 6 THREE DIMENSIONAL MEASUREM ENTS OF WIND FORCE O N ASPHALT SHINGLE SEALANT STRI PS WITH FULL AND PARTIA L ADHESION Th us far, the new research presented in this dissertation has addressed the material performance and wind resistance of shingle roofing in its n ew and aged states. In Chapter 4 partially unsealed shingles were observed on homes in Florida and Texas and the wind vulne rability of fully sealed and partially unsealed shingles was addressed in full scale wind tunnel tests. In Chapter 5 the wind resistance of fully sealed aspha lt shingles was measured on artificial ly and natural ly aged specimens to evaluate the effect of w eathering on wind resistance The purpose of the research presented in this chapter was to measure the wind load mechanism acting on the sealant strips of asphalt singles in fully sealed and partially unsealed conditions. In this research wind was passed over fully sealed and partially unsealed three tab and laminate style asph alt shingles while force and moment were measured in three The tests were conducted in a newly constructed wind tunnel at Powell Famil y Structures and Materials Testing Laboratory at the University of Florida designed for the specific purpose of replicating near roof wind velocity above discontinuous roofing materials. T he mean wind induced in and out of plane force acting on the seala nt strip and rotational moments acting about the sealant strip were measured. Measured mean forces in both adhesion conditions are compared herein to forces predict ed by the ASTM D7158 test standard (ASTM 2011c) F or partially unsealed shingles, the wind f orce distributed along the sealant strip and concentrated at the adhered and non adhered interface are estimated to further define the vulnerability of partially unsealed shingles to wind induced damage.
126 Knowledge Gaps Wind Load Model and Load Path The distribution of wind induced uplift pressure and resultant force on an asphalt shingle as described by the Peterka et al. (1997) wind load model is shown in cross section in Figure 6 1 P ressure s P F and P B arise from the interaction of near roof wind o leading edge is the only direction considered in the model N et uplift pressure is uniformly applied over the width of the shingle (cross flow) and on the flow wise lengths L 1 and L 2 with magnitude P F and P B M agnitude and location of the net force vectors ( F F and F B in Figure 6 1 ) are therefore, a function of L 1 and L 2 the width of interest and the applied pressure ( P F and P B ) Figure 6 1. Wi nd pressures on shingle roofing One gap in knowledge in the Peterka et al. (1997) model is the magnitude of in and out of plane wind force acting on the sealant strip. Measurements of wind induced Due to the positioning of the sealant strip relative to the pressure distribution, the model assumes wind uplift pressure i s completely transferred through the sealant strip a nd into the downslope fasteners.
127 For adhesives such as the sealant strip, resistance varies as a function of applied force ( Gay and Leiber 1999 ), thus, an accurate representation of force acting on the s trip is require for accurate prediction of its resistance. This was exemplified in Chapter 5 where peel strength of the sealant strip was approximately one half its tensile strength. Shiao et al. (2003a) found similar results using the same ASTM D6381 test method (ASTM 2008b) on non aged three tab and laminate shingles. Equations 6 1 and 6 2 are used within ASTM D7158 (ASTM 2011c) to calculate the wind resistance of an asphalt shingle product ( R T ASTM D6381 Procedures A ( peel R A ) and B (tensile R B ) mechanical uplift testing are F F ) and leeward ( F B ) forces at design level wind conditions. The forces are a calculated using the differential pressure coefficients measured on the product during an ASTM D7158 (ASTM 2011c) wind test and the exposure lengths L1 and L2 (see Figure 6 1). If F F > F B Equation 6 1 is used, while F F < F B requires Equation 6 2. (6 1) (6 2) E quation s 6 1 and 6 2 provide an estimate of the amount of peel and tensile force exerted on the sealant strip in design level wind velocity T he corresponding peel ( R A ) and tensile ( R B ) uplift resistance are each weighted by the relative contribution of F F and F B acting on the sealant strip then summed to form the overall wind resistance of sealant strip To understand the eq uations meaning, one could say that F F is the only force that exists. Thus, using Equation 6 1, peel resistance ( R A ) is the only contributing uplift resistance in the total uplift resistance of the shingle ( R T ). The same is true if only
128 leeward force ( F B ) is exerted on the shingle (Equation 6 2). From Peterka et al. (1997), standard th ree tab and laminate shingles have a mixture of windward and leeward wind forces so total uplift resistance is most often a mixture of peel and tensile resistance. The measurements of in and out of strip made in this research expand the knowledge base with respect to the reaction forces developed on the primary wind load path of asphalt shingles. Ultimately, a better understanding of reaction forces will improve the predict ive capabilities of standard test methods either through new measurement techniques or modifications to existing design equations Partially Unsealed Shingles The second knowledge gap relates to the wind induced force on partially unsealed shingles. Chapter 3 identified partially unsealed shingle s as a potentially significant contributor to wind damage in shingle roofing. One significant finding from the research was the performance gap between fully and partially unsealed shingles subjected to equivalent wind speeds From this work, t wo failure m echanisms contributing to the vulnerability of partially unsealed shingles to wind induced damage are proposed, both caused by the lifting of the unsealed portion of the shingle First, an increased distributed load on the adhered length of the sealant strip arises from increased bottom surface wind pressurization Second, a concentrated force is applied at the interface of the adhered and non adhered portions of the sealant strip as wind force is exerted on the lifte d section of shingle, acting as a peel force on the sealant strip at the interface. The research presented in the chapter places actual values related to the two proposed mechanisms. The research addresses vulnerability by comparing
129 wind induced forces on partially unsealed to those measured on fully sealed shingles and those predicted in ASTM D7158 (ASTM 2011c) Experimental Setup Concept The study was modeled after full scale boundary layer wind tunnel tests used in the development of the load model pro posed in Peterka et al. (1997) and after wind tunnel methods specified in ASTM D7158 (ASTM 2011c) The unique features differentiating this work from previous research were the direct measurement s of three ant strip and the use of partially unsealed shingles in the test matrix. W ind force was measured on o ne three tab and one laminate style shingl e product. ASTM D7158 d esign differential pressure coefficients indicate that both products are representative of three tab and laminate shingles commonly in service today (ASTM 2011c) In this study, a sphalt shingle s were installed on a 2.43 m (8 ft) long by 1.82 m (6 ft) wide planform test deck to model an installation of shingles in the field of the roof. For comp arison to Peterka et al. ( 1997) and ASTM D7158 (ASTM 2011c) t he leading edges of the shingles were oriented perpendicular to mean wind flow. One shingle towards the leeward portion of the deck was designated as the test specimen, and up to three multi axi s load cells were attached to the test specimen below along the The remaining shingles were fixed along the leading edges to maintain a fully sealed condition. Using a newly constructed wind tunnel designed for replicating wind above the roof plane, wind was passed over the top surface of the test deck, while wind speed and sealant strip forces and moments were measured. Three mean wind speeds were used
130 in this t est, 15 m/s (33 mph) sampled at 1250 Hz was 12 14% near the instrumented specimen surface. Wind was he ld constant for a minimum 180 s at each wind speed. A vertical plane of the wind field above the test specimen was first quantified and then a minimum of three replications of each wind speed was applied on each test specimen. For fully sealed shingles, wind induced force acting on the sealant strip was determined from the resolved vectors of m ean in plane streamwise and mean out of plane ( vertical ) forces Sealant strip force was then normalized at each wind speed by velocity pressure measured 25 mm (1 in), forming force coefficients. Measured f orce coef ficients were then compared to force coefficients predicted by product specific ASTM D7158 differential pressure coefficients (ASTM 2011c). For partially unsealed shingles, mean in plane flow wise and out of plane vertical forces acting at the interface b etween the adhered and non adhered portions of the sealant strip were estimated from rotational moments measured by the multi axis load cell. From this, mean distributed force on the sealant strip was estimated by subtracting the estimate of interfacial fo rces from forces measured by the load cell. Force coefficients derived from the velocity pressure normalized distributed sealant strip force are then compared to forces coefficients pre dicted by ASTM D7158 differential pressure coefficients (ASTM 2011c) me asured on fully sealed shingles. Test Apparatus Introduction To achieve the objectives and future research objectives relating to the wind resistance of discontinuous roofing systems a new wind tunnel henceforth,
131 Dynamic Flow Simulator or DFS was constructed at the University of Florida. Test specimens inside the DFS are subjected to the mean and turbulent components of near roof wind, providing a new tool to accurately measure the wind load and resistance of discontinuous roofing systems. Th e unique features of the DFS include the maximum velocity at the test section approximately 95 m/s (212 mph) and the ability to replicate up to 5 Hz waveforms in wind velocity The present study was the first performed in the DFS Componentry The DFS is one part of the test ap paratus detailed in Shen et al. ( 2013 ). Fi gure 6 1 provides a rendering of the seven main components of DFS. A 1340 kW (1800 HP) centrifugal blower draws air through a 1.5 m (5 ft) diameter inlet where it passes through op posed blade damper system for active wind speed control. On the leeward side o f the blower, air is pushed through two 90 elbow bends before traveling into a settling chamber consisting of a wide angle diffuser, turbulence screens, honeycomb (68% porosity) and duct contraction. undesired fine scale turbulence and improve flow uniformity across the duct cross section (Figures 6 2, 6 3, and 6 4) The duct contraction on the leeward side of the settling chamber ca uses the wind to accelerate to its target velocity by the time it enters the test section. Passive turbulence generation devices (e.g., spires, grids, or roughness blocks) were not installed upwind of the test section. After passing through the test sectio n, air exhausts to the free atmosphere through a diffuser at the exit.
132 Figure 6 2 Rendering of the Dynamic Flow Simulator componentry. Figure 6 3 Dynamic Flow Simulator, profile view, and test section, orthogonal view. Figure 6 4 Dynamic Flow Simulator, as constructed.
133 The cross sectional area at the entrance to the test secti on is 2.13 m (7 ft) wide by 0.31 m (1.25 ft) tall (Figure 6 5 ) The width of the test section does not vary, however the height is adjusted to regain static pressure lost by friction. In other words, the size of the test deck used in the experiment was 2.43 m (8 ft) long by 1.82 m (6 ft). A pneumatic lift raised the test deck into place through an opening in the bottom floor of the test section (Figure 6 6 ) form ing a continuous lower surface An interior view of the test deck during one of the wind tests used for this study is shown in Figure 6 6 The wind field generated in a ver tical plane above the shingle test specimen is discussed in a later section, but an example mean longitudinal velocity profile measured over the test specimen is shown in Figure 6 8 Figure 6 5 Cross section of DFS test section. Figure 6 6 Test d eck below opening in test section.
134 Figure 6 7 laminate shingle instrumented with load cells is shown. Figure 6 8 Mean longitudinal velocity across the width of a shingle test specimen located 1.9 m (6.2 ft) downwind of the windward edge of a shingle roof test deck. Test S pecimens Test specimens consisted of one three tab asphalt shingle product (henceforth, Three Tab) and one laminate asphalt shingle product (henc eforth, Laminate). Both
135 products were fiberglass reinforced and complied with ASTM D3462 (ASTM 2010). The Three Tab product was same product used in the artificial aging experiment detailed in Chapter 5 identified in that chapter as Product A. The Lamina te product was procured from a Gainesville, FL contractor supply store. Table 6 planform dimensions and the number of individual shingles used to measure wind load in the experiment. Both products were certified ASTM D7158 Class H (ASTM 2011c) and ASTM D3161 Class F (ASTM 2013) D7158 differential pressure coefficients (ASTM 2011c) directly from the manufacturer (Table 6 2) for use as the baseline comparison to wind loads measu red along the differential pressure coefficient on the windward [ DCp(f ) ] side of the sealant strip on Laminate is approximately double Three Tab, while closer agreement is found in the DCp(f ) ]. Therefore, the shingle products used in this study are representative of typical three tab and laminate style shingles, as reported in Peterka et al. (1997). Table 6 1. Test specimen ID, type, planform dimensions, and number of specimens Shingle ID Shingle Type Planform Dimensions (mm) Number of Tested Shingles Width Height Fully Sealed Partially Unsealed Three Tab Three tab 914 305 3 3 Laminate Laminate 984 337 2 2
136 Table 6 2. ASTM D7158 differential pressure coefficients Shingle ID Exposed Length Dimensions (mm) ASTM D7158 Differential Pressure Coefficients L 1 L 2 DC Pf DC Pb Three Tab 32 95 0.48 0.14 Laminate 22 121 1.03 0.23 Given the information of Table 6 2, an equivalent wind speed should produce a 3 displays Laminate and Three ( F F ) and leeward ( F B the length and pressure coefficients reported in Table 6 2. The coefficient of force represents the force per unit length along the sealant strip normalized by the velocity p ressure. Also sh own in Table 6 3 are the relative contributions of ASTM D6381 Procedures A ( R A ) and B ( R B ) that would be used for each product, as calculated by Equation 6 1 from ASTM (2008b) The calculated contributions indicate that Procedure B (tensile) resistance rep resents the dominant force exerted on both Three Tab (64%) and Laminate (76%). Therefore, the predominate wind induced force on the sealant strip should be an out of plane vertical force (e.g., tensile). In plane flow wise and rotational moments about the axis oriented lengthwise with the sealant strip should are also expected. Test Deck Specifications A 2.43 m (8 ft) long by 1.82 m (6 ft) planform test deck was constructed as the substrate for the shingle test specimens and reaction frame for the multi axi s load cells.
137 The test deck consisted of a frame, structural decking, base sheet, and surrounding shingles corresponding to the same product as the test specimen. Table 6 3 Force coefficients and relative contribution of ASTM D6381 Procedures A and B to t otal uplift Product Force Coefficient (mm) Contribution of ASTM D6381 Procedures A and B Resistances to Total Uplift Resistance (%) F F F B R A R B Three Tab 15 7 36 64 Laminate 23 14 24 76 The load cells where wind load was measured were also housed in the interior of the test deck. P lan view s of the Three Tab and Laminate test deck s mounted to the lower floor of the DFS test section are shown in Figure s 6 9 and 6 10 respectively Shown also in the figure are the locations of the test specimens, wind force measurement, and reference velocity sensor (discussed later). Figure 6 11 shows a cross section of the test deck mounted to the lower floor of the DFS test section. Figure 6 9 Plan view of Three Tab test deck.
138 Figure 6 10 Plan view Laminate test deck. Figure 6 11 Cross section of DFS with test deck. The frame was 2 x 8 dimensional lumber with 16 mm (0.625 in) thick structural top surface. The base sheet was a single layer of ASTM D226 (ASTM 2009) Type II (No. 30) asphalt impregnated felt underlayment, secured to the sheathing using hand driven 31 mm (1.25 in) long galvanized steel button cap type fasteners. The surrounding shin gles were installed
139 courses. Each shingle strip was secured to the deck using six pneumatically driven, 12 gauge, 9.5 mm (3/8 in) diameter head, 31 mm (1.25 in) length, galvan ized steel roofing prevent the leading edge from lifting during the wind test sequence. The load cell chamber referenced in Fi gure 6 11 housed the multi axis load cells used to measure sealant strip forces and moments. Multi Axis Load Cell Specifications Three ATI Industrial Automation model Nano25 IP65 six axis loads cells were used as the wind load measurement devices (Figure 6 12 ) The load cells resolve forces and moment s in the X Y and Z planes. Six axis load cells, as opposed to single or three axis lo ad cells, were chosen because forces were expected on three measurement planes and the attachment detail between the top of the load cell and the bottom of the shingle which introduced a moment arm. Figure 6 12 Multi axis load cell elevation and plane view.
140 Each load cell is a 28 mm (1.1 in) diameter by 28 mm (1.1 in) tall stainless steel cylinder with silicon strain gauges fixed on the interior face. The sensing ranges and resolutions for each load axis are given in Table 6 4 Load cell calibrations were performed by ATI Industrial Automation and are National Institute of Standards and Technology traceable. The load readings were captured via National Instruments Labview 2010 and a National Instruments 6218 DAQ analog to digital converter. Prior to the experiment the accuracy of input force to output reading on each measurement axis was verified by the author using known weights at known distances from the load ce sensing reference frame origin Table 6 4 Six axis load cell sensing ranges and resolutions Measurement Force Moment Axis X Y Z X Y Z Sensing Range 222 N 222 N 890 N 5.6 N m 5.6 N m 3.4 N m Sensing Resolution 1/112 1/112 3/112 1/80 1/19 1/60 fixed to t slot aluminum tube. The steel plate was 76 mm (3 in) by 38.1 mm (1.5 in) by 6 mm (0.25 in) thick and screwed into t 13 ). The plate was affixed on a 25 mm (1 in) square t slot aluminum tube that ran a continuous length along the sealant strip. The slotted feature of the tube allowed the load cells to slide ree Tab shingles, but were fixed in place during wind testing. Figure 6 13 illustrates a cross section of the load cell attachment to the test deck and the shingle test specimen. The centerline of the load cells was 1.88 m (6.17 ft) downwind of the leadi ng windward edge of the test deck and the test specimen was
141 aligned to the centerline of the test decks cross flow length. The base of the cavity formed where the test specimen overlaid the load cells was sealed using an adhesive modified bitumen sheet to restrict the passage of air from ambient into the cavity. The implications of this air passage are discussed later in this chapter. Figure 6 13 Cross section of load cell attachment detail. The plates screwed to the top of the load cell were 38.1 mm ( 1.5 in) wide by 6 mm (0.25 mm) thick steel with length corresponding to the measurement length selected for the given load cell and test specimen combination. A photo of the plate arrangement for the Laminate specimens is shown in Figure s 6 14 and 6 15 Th e attachment of the test shingle to the load cell is discussed in the next section.
142 Figure 6 14 Part 1: load cell arrangement for Laminate specimen showing load cell base connection. Figure 6 15 Part 2: load cell arrangement for Laminate specimen showing wood decking surrounding top plates. Velocity Sensor Specifications Velocity was measured in the DFS using Turbulent Flow Instrument Cobra Probes. Characteristics of the Cobra Probe includes an ability to measure flow within a +/ 45 degrees cone, a maximum sampling frequency of 2000 Hz and a general accuracy of better than 0.5 m/s and 1 degree yaw up to 30% turbulence intensity. Four probes were used in the study : three to measure a vertical plane of the wind field
143 above the test specimen and a fourth to serve as a reference probe during the sealant strip measurement tests. Test Specimen Installation Each load cell measured a portion of th distribution of load cells for Th ree Tab is shown in Figures 6 16 and 6 17 The distribution of load cells for Laminate is shown in Figure 6 18 and 6 19 The numbers below each shingle represent the load cell (LC) number corresponding to the measurement location. In plane coordinate axis alignmen t for each type of specimen is shown in Figures 6 1 7 and 6 19 Interfacial forces are discussed later in the chapter. For a fully sealed test specimen, the entire sealant strip of the test specimen was fixed to the top plates of the load cells. For a parti ally unsealed shingle, the sealant strip of the test specimen on Load Cell 1 was not attached, while the remaining sealant strip line was attached to the load cells. Figure 6 16 Three Tab specimen.
144 Figure 6 17 Plan views for Three Tab fully sealed a nd partially unsealed arrangements. Figure 6 1 8 Laminate specimen.
145 Figure 6 19 Plan views for Three Tab fully sealed and partially unsealed arrangements. Prior to the installation of the specimen on the test deck, a 25 mm (1 in) wide by 3 mm (0.125 in) thick flexible strip magnet was adhered to the specimen using a ~ 1 mm (0.04 in) layer of low exotherm epoxy and cured overnight. The windward edge of the magnet aligned with the windward edge of the sealant strip line to maintain the distance from th an important parameter for wind loads on shingles (Peterka et al. 1997). The polarity of the magnet was aligned to attract to a metal plate attached to the top of the load cell. The attraction force was approximately 360 N/m (25 lbf/ft) and exceeded the predicted maximum distributed wind and plate connection, and into the load cell measured only where a magnet was locate d.
146 For the Three Tab specimens, a fully sealed shingle contained a continuous partially sealed Three Tab shingle contained a magnetic strip along the right half of the center tab as viewed from plan. For the Laminate specimens, a fully sealed shingle contained a magnetic strip along the entire 984 mm (38.75 in) width. A partially unsealed Laminate shingle contained an 832 mm (32.75 in) length of magnetic strip, leaving the left most 152 mm (6 in) of the test specimen without a magnet. magnetic strip aligned to the centerline of the top plates. Due to the relatively low peel resistance of the magnets, one fastener was installed 13 mm (0.5 in) from each edge of magnet strip. For the fully sealed case, this occurred at the edges of the test specimen in the Laminate specimens and the edges of the center tab in the Three Tab specimens. For the partially unsealed case, this occurred at the joint between the (Laminate) or tab (Three Tab). The fastener was a 6 mm (0.25 in) shaft diameter low profile head bolt w ith 13 mm (0.5 in) outer diameter washer screwed through the top of The test specimen was secured to the deck using six fasteners, located as shown above in Figure 6 16 and 6 18 The fasteners were 19 mm (0.75 in) shaft length mechanically driven wood screws. Once secured, a single row of shingles was installed downwind of the test specimen to form the standard exposure length on the test specimen. The leading edge of the last ro w of shingles was secured using the same
147 wood screws distributed as required to prevent lifting of the shingle. A completely integrated fully sealed Laminate tes t specimen is shown Figure 6 20 Figure 6 20 Fully sealed laminate test specimen installed on the test deck. Experimental Procedure The experiment was performed in three phases. Phase I measured a vertical plane of the wind field above the sealant strip line of the test specimen. Phase II quantified the effect of static pressure measured in the DFS on the shingle test specimen load cell readings. Phase III measured the wind induced forces and moments In Phase I, velocity measurements were taken at six heights ranging from 12 mm (0.5 in) to 154 mm (6 in) at eight different locations in the cross flow direction (refer to Figure 6 2 1 ) The sampling frequency of the Cobra Probes was 1250 Hz. A laminate style shingle roof was ins tall ed on the test deck (Figu res 6 21 and 6 22 ) and measurements were taken 25 mm (1 in) leeward where the leading edge of the instrumented test shingle was to be located following ASTM D7158 (ASTM 2011c) Velocity was measured at each location at the three wind speeds used in Phase II for a minimum of 180 s.
148 Figure 6 21 Plan view and cross section of velocity measurement locations. Figure 6 22 Velocity sensor test setup. One additional Cobra Probe was fixed to th e position shown in Figures 6 21 and 6 22 152 mm (6 in) above the shingle surface. Wind velocity was simultaneously measured at the various points above the shingle surface and at the reference location to establish the relationship between velocity above the instrumented shingle and velocity at the reference locati on. Following this, the reference probe was the only probe in use during Phase I I I.
149 In Phase II the pressure difference measured between the ambient and interior of the test section during Phase I was replicated on the shingle test specimen using a Press ure Loading Actuator affixed to a chamber over the test specimen. Forces and moments measured by the load cell were related to static pressure in the chamber. In Phase II I w ind was passed across the test s pecimen for a minimum of 180 s while velocity at the reference location and sealant strip force were measured simultaneously. Test specimens were subjected, sequentially, to the Low, Medium, and High wind speeds with a 30 s period of ~3 m/s (7 mph) wind between each speed. Five cycles of the Low/Medium/H igh sequence were used on the each Laminate specimen, while three cycles were used for each Three Tab specimen. Results Phase I: Wind Field Above the Test Specimen Mean longitudinal v elocity Vertical profiles of mean longitudinal velocity shown Figure s 6 23 through 6 25 with each figure representing one wind speed As expected, a boundary layer above the specimen is present in all mean longitudinal velocity profiles. Recall, the most critical parameter for the wind loading of shingles is the mean longitu dinal velocity measured For this study, mean longitudinal velocity measured above the specimen varied by 0.7 m/s in the Low s peed (15 m/s) 1.6 m/ s in the Medium speed ( 30 m/s ) and 2.1 m/s in the High speed (44 m/s)
150 Figure 6 23 Mean longitudinal velocity profiles Low wind speed. Figure 6 24 Mean longitudinal velocity profiles Medium wind speed.
151 Figure 6 25 Mean longitudinal vel ocity profiles High wind speed. Figure 6 26 illustrates the mean and range of longitudinal velocity profile measurements at the eight positions across the specimen at all three wind speeds normalized by the 152 mm (6 in) measurement height mean value. Black squares in the figure are the mean value and the line and vertical ticks are the range of measurements. A smooth boundary layer profil e is shown in the mean measurements. T he 25 mm (1 in ) mean longitudinal velocity is approximately 80% its counterpart at 152 mm (6 in). Longitudinal turbulence intensity passage through the DFS system upstream of the test section a nd flow over the shingle test deck. Vertical profiles of longitudinal turbulence intensity ( Iu ) at the three wind speeds are shown in Figure 6 27 The raw (1250 Hz sampling frequency) Iu is compared to a third order Butterworth filtered Iu set to cutoff fr equency of 12 Hz. Each point represents the average Iu
152 width at the given height. Measurements at 25 mm (1 in) above the shingle indicate a 10% reduction in Iu from the raw to filtered measurements. Fi gure 6 26 Mean longitudinal velocity of all measurement positions and wind speeds normalized by the 152 mm height mean. Figure 6 27 Mean longitudinal velocity of all measurement positions and wind speeds. Table 6 5 provides the mean and standard dev iation of longitudinal integral length scales ( Lx (Figure 6 21 ) at the smallest measurement height 12 mm (0.5 in). Values in the table were calculated from the
153 unfiltered 1250 Hz records Mean Lx ranged from 115 mm (4.5 in) in the Low wind speed to 135 mm (5.4 in) in the High wind speed with standard deviations roughly 30 mm (1.1 in). Table 6 5 Longitudinal integral length scales measured 12 mm above shingle surface Wind Speed a Longitudinal Integral Length Scale Lx (mm) b Mean Standard Deviation Low (15 m/s) 115 29 Medium (30 m/s) 126 26 High (44 m/s) 136 34 a Speed measured 25 mm (1 in) above shingle surface b Sampling frequency = 1250 Hz Discussion on Turbulence Pet erka et al. (1997) state that 17% Iu and 500 mm (20 in) Lx in the approach wind field for a full scale bo undary layer wind tunnel provides sufficient gust size to accurately predict the mean differential pressure expected on a shingle subjected natural wind. Iu and Lx near the r oof surface and sampling frequency of the velocity sensor were omitted in Peterka et al. (1997) Assuming similitude between the approach and near roof flow in Peterka et al. (1997), mean Iu and Lx measured at the three wind speeds in the DFS were ~6% in Iu and ~375 mm (15 in) below those recommended by Peterka et al. (1997). To understand the effect of underrepresenting Iu and Lx on mean shingle surface pressure s in the DFS experiment relative to Peterka et al. (1997) the cross section of a shingle was approximated as a flat bar with height ( D ) residing on horizontal ground with wind passing normal mm (0.12 in) to 6 mm (0.25 in), thus D took on the same range of values.
154 Setting Lx equivalent to th e average value shown in Table 6 4 [126 mm (5 in)] produced an Lx / D ratio of 21 to 42. Values of Lx / D from Peterka et al. (1997) range from 83 to 167. From Li and Melbourne ( 1999 ), Lx / D ratios greater than 12 at equivalent Iu pro duce similar mean surface pressure distributions due to the reduced influence of length scales on mean flows over bluff bodies. Thus, underrepresenting Lx in the present study is not expected to have an effect on mean pressure distribution on the shingle s urface. Li and Melbourne (1999) do, however, demonstrate that a reduction in Iu reduce s the magnitude of mean pressure s nearest the leading edge of the bluff bo dy (e.g., shingle). Mean force on the leading edge measured in this study may be lower than those predicted by the Peterka et al. (1997) load model. A further question pertains to the equivalence between the turbulence structure of the present study and that in the ASTM D7158 tes t method (ASTM 2011c) Longitudinal turbulence intensity at the test section of an ASTM D7158 shingle deck is unknown (ASTM 2011c) although the method incorporates a turbulence generation grid at the fan exit and roughness strips upwind of test section. F uture work should address this knowledge gap. Static Pressure The average static pressure between the test section and ambient 25 mm (1 in) above the test specimen was 100, 400, and 800 Pa ( 0.01, 0.06, and 0.12 PSI) for the three wind speed levels, respectively. Thus, during wind testing, an upward pressure was exerted on the underside of test deck exposed to the ambient pressure. Recall from Figure 6 12 the load cells were fixed to the test specimen below in a cavity formed in the test section. If ambient air freely entered the cavity, it would produce an
155 sealed prior to the experiment to prevent air passage into the cavity. The seal, however, did allow a relatively small amount of ambient air into the cavity. A separate e xperiment detailed in Phase II measured the effect of static pressure on the force and moment recorded on the sealant strip. Phase II: Effect of Static Pressure on Shingle Test Specimens A separate e xperiment was performed prior to Phase III to quantify the load exerted on a test specimen as the result of a pressure difference between the top surface of the shingle and ambient. One Laminate test specimen was installed on the test deck and a rectangula r box with one open surface was installed over the complete planform area of the test specimen to form a sealed chamber ( Figure 6 28 ). Air was exhausted out of the chamber using a Pressure Load Actuator (Kopp et al. 2010) and held for a minimum of 60 s at four differential pressures levels ranging from 0 to 430 Pa (0 to during the test, producing a relationship between average static pressure in the DFS test section and average fo rce or moment on the sealant strip. The process was repeated for a Three Tab specimen. A B Figure 6 28 Phase II experimental setup. A) Laminate specimen with pressure box. B) Sealed box with air hose and pressure sensor attached.
156 Results from this experiment show: a) X and Z axes loads occur on the Laminate specimen and b) Z axis loads occur on the Three Tab specimen. Moments and all other for ces were negligible. Figure 6 29 illustrates the resultant relationship between static pressure and X and Z axes forces on a 416 mm (16.375 in) length of sealant strip for the Laminate specimen. The figure indicates a linear relationship between pressure and force (R 2 = 0.99). The resultant equations relating static pressure ( P ) and X axis and Z axis fo rce ( F Static ) for the Laminate s hingle is given in Equations 6 3 an d 6 4 respectively. A similar linear relationship was found between Z axis force and static pressure in the Three Tab shing le, producing Equation 6 5 The variable ( L ) in all three equatio ns represents the measurement length along the sealant strip. Figure 6 29 Relationship between static pressure and mean force. X and Z axes component force are shown for 416 mm (16.375 in) length of sealant strip. (6 3 ) (6 4 ) (6 5 )
157 Phase II I : Wind Induced Data p rocessing S ealant strip force s and moments measured during Phase III of the study were processed to the final force and moment data reported below using the procedure outline in this section R aw 100 Hz sampled load cell records were first digitally low pass filtered using a third order Bu tterworth filter set to a 12 Hz cutoff frequency Figure 6 29 shows a time history comparing the raw and filtered signals for uplift (+ Z axis) on a fully sealed Lamainate specimen Next, force and moment data were visually inspected as time history plots in Matlab 2009 b software. The goal was to identify: a) the typical force s and moment s in the sealant strip as wind passes over the shingle, and b) where output forces and moments. The load c ells d o not have internal temperature. The temperature of the air passing over the deck in the DFS during wind testing was greater than ambient at all times; therefore, the l oad cells exhibited temperature induced ch anges in early wind speed tests. Conversely, later tests showed no effects from temperature likely due to the equilibrium reached between the l Figures 6 30 and 6 31 provide a comparison between the typical force time histories on the Z axes for outputs th at were (Figure 6 31) and were not affected (Figure 6 30) by temperature. The records where temperature induced changes were not observed show a constant mean force on all three axes during the constant speed wind test. The records where temperature induce d changes were observed show a non constant mean force during the constant wind speed test. During this time, the load causing an expansion in the load cells body producing a
158 change in strain reading of the load cell and, subsequ ently, a change in the force and moment recorded by the computer. Figure 6 30 Representative time history plot of z axis force between raw and filtered signals and not affected by temperature. Figure 6 31 Time history plot of z axis force affecte d by temperature change on the To account for temperature effects, all force and moment re cords identified as temperature influenced were inspected as time history plots in Matlab 2009b software An example of the estimation process for m ean force and moment is given in the follow
159 sentences. The initial change in force at t he initiation of the wind test 1 in Figure 6 31 The change in force at 2 in Figure 6 31 1 was within 2 as it was in Figure 6 11 the mean force or moment was recorded as 1 2 Where not within 10%, the record was not used for result s reporting. For records not influenced by temperature, the mean force or moment was calculated as the average value measured during the >180 s wind test All mean force data were then processed to account for the static pressure effect quantified in Phase II Static pressure in the DFS test section relative to ambient was measured during each wind test using the static pressure port of the reference Cobra Probe velocity sensor and translated to an equivalent static pressure 25 mm (1 in) above the shingle surface. For the Three Tab specimens, mean force measured in the Z axis was subtracted by static pressure i nduced force predicted by Equation 6 5 Fo r the Laminate specimens, mean force measured in the X and Z axes were subjected by their corresponding static pressure induced forces predicted by Equations 6 3 and 6 4 respectively. In terms of actual values, the amount of force in the Z axis subtracted from a Laminate specimen measured during the High speed level was ~25% of the originally m easured mean Z axis force Fully sealed shingle r esults mean forces and moments Tables 6 6 and 6 7 present the experim ental results of f orce and moment acting on the fully sealed Laminate and Three Tab sealant strips, respectively, stratified by measurement axis specimen number and wind speed level. Figure 6 3 2 prov ides the in plane coordinate axe s for the fully sealed specimens Z axis is positive out of the page. The load cells did not measure rotations about X axis; therefore translation of the three independent reference frames (load cell) to the single frame shown in Figure 6 3 2
160 was not an issue. The mean values on the ent ire length of the sealant strip. The values were calculated by averaging the results of all tests performed on the specimen at the given wind speed. T he data shown in Table 6 6 had a variation of m ean longitudinal wind 25 mm (1 in) above the specimen of 0.6 m/s (Low), 1.4 m/s (Medium), and 0.3 m/s (High). Thus, some variability in the force and moment data can be attributed to variation in the wind field between test runs. Figure 6 32 Referen ce frame for fully sealed Lam inate and Three Tab specimens. Table 6 6. Mean forces and moments measured on fully sealed Laminate specimens Wind Speed Specimen Force (N) Moment (N mm) X Y Z X Y Z Low 1 0.8 (0.3) 0 (0) 3.7 (0.2) 0 (0) 0 (0) 0 (0) 2 0 (0) 0 (0) 4.4 (0.3) 0 (0) 12 (6) 0 (0) Medium 1 3.4 (0.9) 0 (0) 13.4 (0.2) 0 (0) 24 (16) 0 (0) 2 1.5 (1) 0 (0) 15.4 (0.5) 0 (0) 37 (26) 0 (0) High 1 9.8 (0.7) 0 (0) 28.9 (0.4) 0 (0) 84 (14) 0 (0) 2 7.9 (1.8) 0 (0) 33.9 (0.3) 0 (0) 74 (46) 0 (0) *Averaged from five tests at each wind speed
161 For the Laminate specimens, force was measured as acting only in the X and Z planes. As expected (ASTM 2011c) force was predominantly uplift ( + Z axis) on the sealant strip and coefficients of variations of the Z axis force on the range of 1 5%, indicating consistency between test runs at all wind speeds In plane loads measured on the X axis are also logical as a negativ e X axis force is oriented in same direction as the mean wind vector. Resolving X and Z axis loads into a single vector produces an average vector angle of 78 relative plane of the shingle surface (i.e., horizontal) The resolved mean force vectors of S pecimen 2 were consistently higher than Specimen 1 0.7 N higher in the Low speed, 1.6 N in Medium, and 4.3 N in High The effect of such variability on the normalized force coefficient is explored later in this section. Moments measured on the Laminate specimens were highly variable in the Y axis and non existent in the X and Z axes. A n eccentricity is formed where X axis forces are applied to the sealant strip caused by the steel plate located on the top surface of the load cell. Therefore, Y axis m oment measurements represent eccentric applications of Z axis force and X axis force on the sealant strip. Positive Y ax is moments were consistently measured on Specimen 2 with standard deviations greater than one half the mean moment. Negative Y axis mome nts were measured on Specimen 1 in the Medium and High wind speeds with less deviation from the mean than Spec imen 2.
162 Table 6 7. Mean forces and moments measured on fully sealed Three Tab specimens Wind Speed Specimen Force (N) Moment (N mm) X Y Z X Y Z Low 1 0.2 (0.4) 0 (0) 1 (0.1) 0 (0) 9 (15) 0 (0) 2 0 (0) 0 (0) 0.6 (0.5) 0 (0) 0 (0) 0 (0) 3 0 (0) 0 (0) 1.2 (0.2) 0 (0) 2.4 (4.2) 0 (0) Medium 1 0.5 (0.9) 0 (0) 3.1 (0.2) 0 (0) 7 (19) 0 (0) 2 0.1 (0.2) 0 (0) 3.2 (0.1) 0 (0) 0 (0) 0 (0) 3 0.1 (0.4) 0 (0) 3.1 (0.2) 0 (0) 1.5 (2.6) 0 (0) High 1 1.5 (2.5) 0 (0) 6.2 (0) 0 (0) 2 (42) 0 (0) 2 0 (0) 0 (0) 7.6 (0.1) 0 (0) 20.5 (4.1) 0 (0) 3 0 (0) 0 (0) 6.3 (0.1) 0 (0) 1.5 (2.6) 0 (0) *Averaged from three tests at each wind speed For the Three Tab specimens, forces were only measured on the X and Z axes (Table 6 7 ) A greater proportion of force was measured as uplift (Z axis) on the Three Tab specimens than the Laminate. Comparisons of mean to standard deviation in Z axis forces indicate consistency in Z axis measurements between run s (COV = 1 16%) similar to the Laminate specimens. With the exception of the outlier X axis measurement of Specimen 3 during the Medium wind speed, the average angle of the X and Z axes force vector was 86 relative to the plane of the shingle surface. The average total force vectors of Specimens 1, 2 and 3 equate to 1.00.4 N in the Low speed, 3.10.3 N in the Medium speed, and 6.80.8 N in the High speed. Similar with the Laminate spec imens, moments measured on the Three Tab specimens were highly variable on Y axis and non existent on the X and Z axes.
163 Fully sealed shingle results force coefficients Directly measured mean sealant strip forces were converted to force coefficients and compared to force coefficients predicted by ASTM D7158 (ASTM 2011c) For this study, t he force coefficient represents the relationship between an applied distributed force on the sealant strip and velocity pressure in the wind field above the shingle. Therefore, force coefficients presented below have a unit of length. For each wind test, mean X and Z axes forces acting along the entire measured length of the sealant strip were resol ved into a single force vector, and then divided by the sealant strip measurement length to produce a distributed force on the sealant strip ( F D ). Force coefficients ( F CMeasured ) were calculated by normalizing F D by the mean velocity pressure measured 25 mm (1 in) above the test specimen duri ng the corresponding wind test Equation 6 6 provides the normalization scheme. Recall, the wind speed measured during each test was located at the ref erence location shown in Figure 6 5 For the velo city pressure reference wind speed was rescaled using the previously calculated relatio nship to produce an estimated wind speed 2 5 mm (1 in) above the specimen ( V ). (6 6 ) ASTM D7185 differential pressure coefficients ( DCp ) (ASTM 2011c) were also converted into force coefficients ( F CPredicted ). E quation 6 7 was developed from the shingle wind load model proposed by Peterka et al. (1997). Values of DC P L 1 and L 2 for the Laminate and Three Tab specimen are given in Table 6 2 The resultant forces coefficients predicted by ASTM D7158 (ASTM 2011c) a re 37 mm ( 1.4 in) for Laminate and 22 mm (0.9 in) for Three Tab.
164 (6 7 ) The comparison of measured and predicted force coefficients are provided in Tables 6 8 and 6 9 Significantly, all force coefficients on the Laminate specimens were below ASTM D7158 predicted (ASTM 2011c) Similar results were obtained on the Three Tab specimens at the Medium and High wind speeds with exception of one measured value at the Medium speed. Good agreeme nt is shown in the Laminate specimens between average values of force coefficients at the three wind speeds. The Low, Medium, and High wind speeds produced average forces coefficients of 29 mm, 28 mm, and 28 mm with standard deviat ions at each wind speed o f 1 mm (COV = 3%). Three Tab specimens produced a similar outcome at the Medium (20 mm) and High (19 mm) wind speeds with equivalent standard deviation s of 2 mm. The data, therefore, indicates consistent force coefficients were obtained at the three wind s peeds. T able 6 8 Laminate force coefficients directly measured vs. ASTM D7158 predicted Specimen Test Fc (mm) Low Speed Medium Speed High Speed ASTM D7158 Predicted 1 1 29 26 27 37 2 30 27 27 3 29 27 27 4 30 27 27 5 27 27 27 2 1 30 28 29 2 27 28 28 3 31 29 30 4 30 29 29
165 Table 6 9 Three Tab force coefficients directly measured vs. ASTM D7158 predicted Specimen Test Fc (mm) Low Speed Medium Speed High Speed ASTM D7158 Predicted 1 1 24 21 21 22 2 26 18 17 3 27 24 17 2 1 -* 19 20 2 21 20 21 3 20 20 21 3 1 22 20 17 2 28 19 17 3 31 18 17 *Load cells did not record Five of the eight Low speed force coefficients on the Three Tab specimens exceeded force coefficient predicted by ASTM D7158 (ASTM 2011c) This finding is less significant because 17 of 18 measurements on the same Three Tab specimens at the higher wind speeds produced force coefficients below ASTM D7185 predicted (ASTM 2011c) Moreover, the Low wind speed (~15 m/s) combined with relatively short measurement length on the Three Tab specimen yields a higher degree of measurement error when compared to Laminate force coefficients at all wind speeds and Three Tab force coefficients at higher wind speeds. As described in the previous section, the total wind force measured on Laminate Specimen 2 was consistently higher than Specimen 1. The force coefficients measured on Specimen 2 are also consistently higher than Specimen 1; though, this difference is small relative to difference between measured and predicted force coefficients. The Three Tab dataset produced a similar finding. The 18 measurements at the Medium
166 and High wind speeds yielded a 10% coefficient of variation about the overall mean force coefficient calculated at the Medium and High win d speeds (19.2 mm). Thus, force coefficients do not appear to be heavily influenced individual test specimen. Additional specimens are necessary to strengthen this conclusion. As expected from the ASTM D7158 differential pressure coefficients stated for ea ch shingle type (ASTM 2011c) the Three Tab specimens produced lower average force coefficients than the Laminate at all three wind speeds. This demonstrates that the directly measured results align with the load mechanism proposed by Peterka et al. (1997) Figure 3 7 indicates that peak pressures exerted on a shingle change as a function of approach wind direction; therefore, the reader is cautioned from The differenc e s b etween measured and predicted likely relate to the method for measuring differential pressure coefficient measurement in ASTM D7158 (ASTM 2011c) Recall from Chapter 3, ASTM D7158 uses fully sealed shingles with artificially raised leading edges that e qual the vertical displacement estimated in a design level wind speed. Wind speeds at the ASTM D7158 Class H design level are lowe r than those used in this study (ASTM 2011c) ; t herefore, th e differential pressure coefficients measured in ASTM D7158 are hig her than those likely achieved in this study. Partially unsealed shingle results measured forces and moments Tables 6 10 and 6 11 present the force and moment acting on the partially sealed Laminate and Three Tab sealant strips, respectively, stratif ied by measurement axis specimen number, and wind speed level. T average force or moment acting on the adhered length of the sealant strip. Therefore, the values are not directly comparable to the forces and moments of the ful ly sealed
167 shingles (Tables 6 6 and 6 7 ) due to the shorter measurement length in the partially sealed dataset. The differences in load mechanisms between the fully sealed and partially unsealed data are immediately visible. Significant contributions of X axis force, relative to Z axis, are present both partially unsealed datasets. Moments generated at the load cells are also greater in the partially unsealed specimens than fully sealed. The concentrated load at the interface of the adhered and non adhered portions of the sealant strip likely influence these findings. Recall, a fastener was installed at the edge of this interface on the adhered portion of the strip to prevent the shingle from further loss of adhesion and to measure wind force acting at the i nterface. Table 6 10 Mean forces and moments measured on partially un sealed Laminate specimens Wind Speed Specimen Force (N) Moment (N mm) X Y Z X Y Z Low 3 2.8 (2.2) 3.1 (3.4) 4.7 (1.6) 134 (85) 27 (26) 27 (39) 4 0.5 (0.6) 2.0 (2.3) 4.5 (0.7) 81 (54) 1 (3) 2 (4) Medium 3 13.0 (9.4) 10.5 (3.3) 17.5 (2.3) 478 (79) 165 (125) 242 (209) 4 3.2 (1.3) 9.7 (7.2) 16.4 (1.9) 408 (252) 25 (47) 18 (12) High 3 29.7 (16.8) 26.2 (7.5) 38.6 (4.5) 1285 (541) 391 (150) 535 (311) 4 13.3 (5.3) 28.7 (16.4) 45.0 (7.6) 1736 (1034) 18 (189) 134 (85) *Averaged from four tests at each wind speed
168 Table 6 11 Mean forces and moments measured on fully sealed Three Tab specimens Wind Speed Specimen Force (N) Moment (N mm) X Y Z X Y Z Low 4 0.0 (0.0) 1.4 (1.2) 1.2 (0.8) 41 (39) 18 (18) 0 (0) 5 0.0 (0.0) 1.5 (1.3) 9.8 (0.6) 25 (27) 14 (13) 14 (24) 6 0.5 (0.8) 1.5 (1.3) 1.0 (0.3) 8 (13) 18 (16) 9 (16) Medium 4 2.4 (1.6) 1.8 (0.2) 4.0 (3.2) 169 (153) 99 (81) 36.4 (59) 5 4.3 (3.6) 6.1 (2.1) 5.1 (2.8) 191 (176) 93 (66) 65 (42) 6 1.7 (0.5) 8.6 (3.3) 3.5 (2.0) 70 (68) 68 (61) 102 (57) High 4 8.5 (2.3) 0.6 (3.7) 8.7 (1.0) 492 (378) 274 (162) 2.3 (155.6) 5 6.6 (3.1) 11.7 (4.9) 9.8 (3.1) 648 (115) 292 (74) 117 (36) 6 5.3 (2.1) 8.6 (24.2) 17.6 (3.3) 600 (163) 364 (50) 272 (75) *Averaged from three tests at each wind speed Once loaded by wind, the non adhered portion of the shingle lifted, producing a force at the fastener location (Figures 6 33 and 6 34). The eccentricity between the fastener and Z axes moments shown in Tables 6 8 and 6 9. The force reported on the Y axis for the partially unsealed Three Tab and Laminate specimens makes less physical sense than the forces reported on the X and Z axes. Thirteen of the fourteen reported Y direction forces in Tables 6 10 and 6 11 are negative. Yet, as the partially unsealed portion of the shingle rises, the force exerted on the interface fastener is expected be positive or neutral in the Y direction. The reader is cautioned, therefore, from making conclusions on Y axis fo rce exerted on a
169 partially unsealed sealant strip. The interfacial and force coefficient data presented in the next two sections will not include Y axis force due to this uncertainty. Figure 6 33 Partiall y unsealed Laminate S pecimen 3 lifting in 44 m/s (98 mph) mean wind velocity. Figure 6 34 Partia lly unsealed Laminate Specimen 4 lifting in 44 m/s (98 mph) mean wind velocity.
170 Partially unsealed shingle results interfacial forces The force exerted at the interface of the adhered and non adhered sealant strip was calculated under the assumption that the moments g enerated on Load Cell 2 about the X and Z axes were caused by a concentrated force at the interface (Figure 6 35) T he lack of X and Z axes moments reported in the fully sealed shingle specimens strengthens this assumption (Tables 6 6 and 6 7 ). Moment about the Y axis has other force influences, such as the eccentric application of upward force (Z axis) generated on the shingle caused by additional wind flow entering through the non adh ered portion of the shingle. Therefore, Y axis moment will not be used in the calculation of interfacial force. Figure 6 3 5 Reference frame for partially unsealed Laminate and Three Tab specimens showing interfacial force location.
171 Tables 6 12 through 6 16 summarize the measured mean X and Z axes moments origin at Load Cell 2 and the estimate of interface forces on the X and Z axes origin at interface shown in Figure 6 35 for all tests on the La minate and Three Tab specimens. The analysis confirms that the loss of adhesion along a partial length of the sealant strip introduces significant wind force at the interface between the adhered and non adhered portions of the sealant strip. Reported estimates of interfacial force make physical sense as force is consistently reported in the vertical (+Z ) and in plane flow wise ( X ) directions. Given this, the interfacial peel force mechanism proposed in the Knowledge Gaps section exist s on partially unsealed shingles. The variability of interfacia l forces generated during similar wind speeds is likely caused by differences in material flexibility For example, Laminate Specimen 4 was two times thicker than Laminate Specimen 3, due to the location of the relief pattern used on Laminate shingles. As wind was passed over the shingles at the High speed, the non adhered portion of Specimen 4 was deflected upward, while the more flexible non adhered portion of Specimen 3 folded backwards. For Laminate Specimen 3, the High wind speed (~44 m/s) produced an average X axis interface force of 3.70.2 N and the average Z axis interface force of 3.90.3 N. Conversely, Laminate Specimen 4 at the High wind speed had an average X axis interface force of 1.00.1 N and the average Z axis interface force of 10.6 1.7 N. Thus, less total force was exerted at the interface of the relatively flexible Laminate Specimen 3 in comparison to the more rigid Specimen 4.
172 Table 6 12 Laminate Specimen 3 measured moments and estimated interface forces Wind Speed Test Measu red Moments (N mm) Estimated Interface Forces (N) X Z X Z Low 1 62 0 0.0 0.3 2 236 47 0.2 1.1 3 209 87 0.4 1.0 Medium 1 557 263 1.3 2.7 2 498 499 2.4 2.4 3 395 442 2.1 1.9 High 1 824 723 3.5 4.0 2 863 763 3.7 4.1 3 754 797 3.8 3.6 Table 6 13 Laminate Specimen 4 measured moments and estimated interface forces Wind Speed Test Measured Moments (N mm) Estimated Interface Forces (N) X Z X Z Low 1 35 0 0.0 0.2 2 107 0 0.0 0.5 3 113 0 0.0 0.5 4 116 0 0.0 0.6 Medium 1 322 23 0.1 1.5 2 447 0 0.0 2.1 3 441 34 0.2 2.1 4 624 45 0.2 3.0 High 1 1555 147 0.7 7.5 2 1876 177 0.9 9.0 3 2158 198 1.0 10.4 4 2565 207 1.0 12.3
173 Table 6 14 Three Tab Specimen 4 measured moments and estimated interface forces Wind Speed Test Measured Moments (N mm) Estimated Interface Forces (N) X Z X Z Low 1 0 0 0.0 0.0 2 46 0 0.0 1.0 3 77 0 0.0 1.0 Medium 1 11 28 0.4 0.1 2 317 89 1.2 4.2 3 177 46 0.6 2.3 High 1 926 182 2.4 12.2 2 319 93 1.2 4.2 3 230 82 1.1 3 .0 Table 6 15 Three Tab Specimen 5 measured moments and estimated interface forces Wind Speed Test Measured Moments (N mm) Estimated Interface Forces (N) X Z X Z Low 1 0 0 0 .0 0 .0 2 21 0 0 .0 0.3 3 54 41 0.5 0.7 Medium 1 23 40 0.5 0.3 2 176 42 0.6 2.3 3 375 113 1.5 4.9 High 1 613 153 2 .0 8 .0 2 776 81 1.1 10.2 3 554 118 1.5 7.3
174 Table 6 16 Three Tab Specimen 6 measured moments and estimated interface forces Wind Speed Test Measured Moments (N mm) Estimated Interface Forces (N) X Z X Z Low 1 0 0 0 .0 0 .0 2 0 0 0 .0 0 .0 3 23 27 0.4 0.3 Medium 1 0 53 0.7 0 .0 2 73 89 1.2 1 .0 3 136 165 2.2 1.8 High 1 415 260 3.4 5.4 2 659 393 5.2 8.6 3 725 203 9.5 2.7 Partially unsealed shingle results force coefficients Force coefficients ( Fc ) were calculated for the partially unsealed shingles and compared to force coefficients predicted by ASTM D7158 (ASTM 2011c) For each wind test, the X and Z axes interfacial force s estimated from the analysis of the previous section were subtracted from the ir corresponding mean X and Z axes forces measured by the load cell The forces resulting from this subtraction repr esent the wind force acting on the sealant strip less the estimated interfacial force. The two component sealant strip force was then resolved into a single force vector, and then divided by the length of the adhered sealant strip. Similar to the fully sea led shingle procedure, the force coefficient measured on each partially unsealed shingle specimen was then calculated by Equation 6 6 The comparison between the measured and predicted force coefficients are presented in Tables 6 17 and 6 18 for the Lamin ate and Three Tab specimens, respectively. S ignificantly 12 of 21 force coefficients on the Laminate specimens and 23 of 27 force coefficients on the Three Tab specimens exceeded ASTM D7158 predicted
175 force coefficients (ASTM 2011c) Laminate Specimen 1 pr oduced the most consistent results, showing the distributed force s exerted on the sealant strip at the M edium and High wind speeds were, on average 44% higher than predicted by ASTM D7158 (ASTM 2011c) for a fully sealed Laminate shingle Test 1 on Lamina te Specimen 3 exemplifies the mechanism driving the increase in distributed force. For this case, the non adhered portion of the shingle remained horizontal during the Low wind speed test. During the Medium wind test, the non adhered portion lifted, allowi ng wind to enter underside of the test speci men leeward the sealant strip where pressure then developed on the underside of the test specimen. The resultant force coefficients, therefore, increased from 16 mm during the Low wind speed test to 58 mm in the High wind speed test. M ore confidence can be given to the Laminate force coefficients than Three Tab. The average Laminate specimen force coefficients at each wind speed were 28 mm (Low wind speed), 37 mm (Medium wind speed), and 53 mm (High wind speed). L aminate s tandard deviations at each wind speed were 13 mm, 12 mm, 6 mm. Contrast this with the Three Tab specimens, where average force coefficients measured at the three wind speeds were 39 mm (Low wind speed), 4 3 mm (Medium wind speed), and 51 mm (High w ind speed). Yet, standard deviations were 19 mm (Low wind speed ), 29 mm (Medium wind speed ), and 22 mm (High wind speed ). One potential source of variability in the Three Tab dataset comes from the subtraction of interfacial forces during the computation of force coefficients. The magnitude of interfacial forces r elative to the magnitude of the total force exerted on the sealant strip was higher in the Three Tab specimens than the magnitude relativities in the Laminate specimen s Thus,
176 estimation error of the interfacial forces has a larger impact on the force coefficients reported for Three Tab than Laminate. Table 6 17 Laminate force coefficients directly measured vs. ASTM D7158 predicted Specimen Test Fc (mm) Low Speed Medium Speed High Speed ASTM D7158 Predicted 3 1 16 49 58 37 2 45 52 55 3 47 48 57 4 1 19 24 45 2 25 27 45 3 17 27 53 4 29 34 57 Table 6 18 Three Tab force coefficients directly measured vs. ASTM D7158 predicted Specimen Test Fc (mm) Low Speed Medium Speed High Speed ASTM D7158 Predicted 4 1 17 6 67 22 2 20 68 55 3 41 53 51 5 1 24 26 12 2 27 105 46 3 50 33 39 6 1 75 27 47 2 44 24 50 3 53 46 95 Discussion asphalt shingle wind load model for fully sealed shingles and the vulnerability of partially
177 unsealed shingles in wind. For fully sealed shingles the average force coefficient measured on the Laminate specimens were 23% lower than predicted by ASTM D7158 (ASTM 2011c) while the average force coefficient measured on the Three Tab specimens at the Medium and High wind speeds were 12% lower than predicted. Moreover, as predicted by the Peterka et al. (1997) load model, the average force coefficients measured on the fully sealed Laminate specimens were greater than Three Tab. The conservatism of ASTM D7158 displayed in this study is expected, as instrumented test specimens in ASTM D7158 contain artificially raised leading edges corresponding to the tip deflect ion at design level wind speed (ASTM 2011c) This, in turn, leads to differential pressure coefficients reported in ASTM D7158 (ASTM 2011c) that are higher th an what would be expected at lower wind speeds used in this study. The reader is cautioned from ex wind. As shown in Figure 3 7 peak differential pressures exerted on a n instrumented shingle installed on a test home varied as a function of wind direction. Furthermore, the peak differential pr essure measured on the shingle did not align with the wind direction used in this study. A dditional work is rec ommended to measure sealant strip wind force at other wind directions. The second major finding of this study is the further confirmation of the vulnerability of partially unsealed shingles in wind As was demonstrated in Chapter 4 and shown in the present study, two damaging mechanisms occur in wind for shingles with a partial loss of adhesion. First, a measurable force is exerted at the interfac e between the adhered and non adhered portions of the sealant strip. Interfacial force
178 was estimated in this study in the flow wise and vertical directions. Results indicate interfacial forc e exists in both planes. Second, the distributed force exerted on the adhered portion remaining on the sealant strip i ncrease s triggered by the vertical deflection of the unsealed portion of the shingle. Once deflected, wind flow enters the underside of the shingle tly, the distributed force measured on the partially unsealed Laminate specimens at the High wind speed was 41% greater than predicted by ASTM D7158 for a fully sealed shingle (ASTM 2011c) Significant Findings In this study, the force and moment on a shingle s sealant strip was directly measured in three dimensions as wind was passed over shingle test specimens. Wind was oriented perpendicular to the leading edge Test specimens included one three tab style shingle (Three Tab) and one laminate style shingle (Laminate). Two adhesion conditions were tested: (1) adhered along a complete length of sealant strip (fully sealed) and (2) adhered along a partial length of sealant strip (partially unsealed). The length and location where not adhe red in condition (2) matched the length and location of partially unsealed shingles identified in Chapter 4. Three objectives were achieved with this research First, the load exerted on sealant strip is predominantly vertical wit h minor contributions of force in plane with the sealant strip in the flow wise direction. The average resolved force vector was 78 from horizontal on Laminate and 86 from horizontal on Three Tab. The load exerted on the adhered portion of a partially un sealed shingle sealant strips is characterized by an increased contribution of
179 force in plane in the stream wise direction. Additionally, a force is exerted on the sealant strip at the interface of the adhered and non adhered portions of the sealant strip. Vertical and in plane flow wise forces were measured at this interface. Second, measured values of total force on sealant strip length were compared to total forces predicted by the ASTM D7158 shingle wind design standard (ASTM 2011c) Results demonstrate that ASTM D7158 (ASTM 2011c) and the underlying Peterka et al. (1997) load model are conservative in sealant strip force prediction for shingles with full adhesion along their length The average force measured on Laminate was 23% low er than predicted by ASTM D7158 (ASTM 2011c) while the average force on Three Tab at the two highest mean wind speeds used in the study was 12% lower than predicted by ASTM D7158 (ASTM 2011c) However, results demonstrate that ASTM D7158 test standard (ASTM 2011c) is not conservative when predicting the distributed force exerted on partially unsealed shingles. For example, t he average distributed force exerted on the partially unsealed Laminate specimens in approximately 44 m/s wind was 44 % higher than the distributed force predicted by ASTM D7158 (ASTM 2011c) The increased load is caused by the vertical deflection of the unsealed portion shingle allowing wind flow into the underside of the shingle. Third the effect of changes in mean wind speed on induced loads through the sealant strip was quantified by passing wind over the specimens at t hree mean speeds ( 15 m/s 30 m/s and 44 m/s). Measured distributed sealant strip for ces were normalized by wind velocity 25 mm above the test specimen to prod uce a coefficient of force independent of wind speed. For the fully sealed Laminate specimens, the three wind speeds produced average forces coefficients of 29 mm, 28 mm, and 28 mm with
180 standard deviations at each wind speed of 1 mm. Three Tab specimens pr oduced a similar outcome at the 30 m/s (20 mm) and 44 m/s (19 mm) wind speeds with equivalent standard deviations of 2 mm. The data, therefore, indicates consistent force coefficients were obtained at the three wind speeds. The results of the present study and Chapter 4 demonstrate that partially unsealed shingles exist on current homes in the residential building stock and that such homes are more vulnerable to wind damage than homes without partially unsealed shingles Two damage mechanisms are prop osed for partially unsealed asphalt shingles subjected to wind. The first mechanism involves the application of force at the interface between the adhered and non adhered portions of the sealant strip. The second mechanism involves the increased distribute d load exerted on the sealant strip
181 CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS The research presented in this dissertation focuses on the identification of mech anisms initiating the continued failure of asphalt shingle roof systems in wind. The mechanisms evaluated in this research are described below with analysis based on the findings from the five studies comprising this dissertation. Recommendati ons for futur e research on the wind resistance of shingle roofing and other discontinuous roofing materials are provided. Conclusions on the Causes of Wind Damaged Asphalt Shingle Roofing Partially Unsealed Shingles The most significant and unexpected finding from this research is the occurrence and wind vulnerability of partially unsealed shingles installed in the field, hip, and ridge locations of the roof. research, there is a high degree of confidence that partially unsealed shingles installed in the field, hip and ridge locations of the roof contribute significantly to the wind damage of asphalt shingle roofs. As part of this research, thirty homes were surveyed in Florida and Texas for un sealed asphalt shingles. Twenty two of the homes (73%) contained partially unsealed shingles in the field locations of the roofs All roofs surveyed Florida roofs had partially unsealed hip or ridge shingles P artially unsealed shingles observed in the fie ld locations of the roofs had systematic locations of unseal (see Figures 4 2 and 4 3 ) and failure mode where unsealed (cohesive fracture in the sealant strip). Partially unsealed shingles in the hip and ridge loc ations were also systematic, but their mode of failure
182 where unsealed was different from those in the field of the roof (adhesive failure between the sealant strip and overlying hip or ridge shingle). The cause of partially unsealed shingles in the field of the roof was not fully identified in thi s research; though, long term weathering effects as opposed to wind, are the likely cause. First, the failure mode of partially unsealed shingles indicates that shingles were fully adhered prior to the loss of adhesion. Second, a statistically significant difference in the mean total quantity of unsealed shingles was established between roofs with ages 0 6 and those 7 13 and 14 20. Thus, the total quantity of partially unsealed shingles on a given roof increased with roof age. Third, H*Wind analysis of his torical tropical cyclone induced gust wind speeds on each survey roof indicated that peak near roof speeds were below wind speeds used in historical product approval tests. Future work related to the identification of specific mechanisms causing partially unsealed shingles in the field of the roof are given in the recommendations section. The vulnerability of partially unsealed shingles installed in the field of the roof to wind damage was demonstrated in two studies detailed in this dissertation. First, 1 7 full scale asphalt shingle roof systems were subjected to a turbulent boundary layer wind flow with speeds up to 54 m/s. Prior to the test, roofs were surveyed for unsealed shingles and, where unsealed, location on the roof was recorded. Wind test result s showed that all damage in the field of the roof initiated from shingles that were found unsealed prior to the wind test or lifted shingles at eave and rake roof locations. Fully sealed shingles remained fully sealed and undamaged in all tests, unless dam aged by
183 adjacent pre wind unsealed or lifted eave and rake shingles. Damage to the pre wind unsealed shingles consisted of blow off, folding, and tearing. A second study measured the wind force exerted on sealant strip of partially unsealed three tab and laminate style asphalt shingles. Two load conditions were measured on the partially unsealed shingles that likely contribute to an increased vulnerability to wind damage. First, a concentrated force was measured at the interface between the adhered and non adhered portions of the sealant strip. This force existed in the vertical and in plane flow wise directions. Second, the distributed force on the adhered portion of the sealant strip was measured as greater than measurements on specimens with complete adh esion along their leading edge length and distributed force predicted by ASTM D 7158 (ASTM 2011c) For shingles on hip and ridge locations, a relationship between age and the amount of unsealed hip and ridge shingles was not established from the survey results The findings of this research indicate that unsealed hip and ridge shingles are caused b y folding the shingle over the hip or ridge line during installation. Post hurricane damage reports from the past decade have found similar ad hesion issu es with hip and ridge shingles Partially unsealed hip and ridge shingles were found on eight hip roof specimens used as part of the full scale experiment described above. Blow off of hip shingles occurred on all eight roofs. Significantly, bl ow off on all hip roofs initiated from the lifting and subsequent blow off of pre wind partially unsealed shingles. Effect of Aging on Wind Resistance Two studies detailed in this dissertation were devoted to the measurement of wind resistance of asphalt shingles exposed to weathering. The findings of this research
184 indicate that w eather induced reductions in uplift resistance cannot be ruled out as a potential contributor in wind damage to shingle roofing. However, its contribution to historical wind dama ge is not as likely as partially unsealed shingles naturally occurring on the roof. Supporting analysis is provide below and recommendations for strengthen this conclusion are provided in the recommendations section. W eather induced reductions cannot be ruled out because a statistically significant reduction in mean ASTM D6381 uplift resistance (ASTM 2008b) was observed between non aged and artificially aged shingles in one of the three products used in the research. For the same product where significant reductions were observed, the predominate mode of failure in the uplifted shingles also changed throughout the exposure time. This finding is tempered, though, due to the response of other two products evaluated in the artificial aging study. S tatistically similar or significant increases in mean ASTM D6381 Procedures A and B uplift resistance (ASTM 2008b) were found in the two products when the non aged and artificially aged specimens were compared. Significantly, the mean and minimum ASTM D71 58 wind resistance (ASTM 2011c) of all three artificially aged products at all exposure time intervals in both artificial methods exceeded the wind resistance required to meet the most stringent ASTM D7158 wind resistance classification (H) (ASTM 2011c) T he second study performed as part of this research topic addressed the wind resistance of four asphalt shingle roofs exposed to over nine year s of natural weathering. Similar to the findings of the artificial study, the naturally aged roofs had sufficient uplift capacity to meet an ASTM D7158
185 Class H (ASTM 2011c) as estimated from differential pressure coefficients reported in Peterka et al. (1997). ASTM D7158 and the Load Model for Asphalt S hingles AST M D7158 (ASTM 2011c) is the most commonly used asphalt shingle wind resistance metric in industry product approval (e.g., Miami Dade), and building codes (e.g., International Residential Code) The work presented in this dissertation and in companion document by the author (Dixon et al., 2013) represents the first independent study of A STM D7158 (ASTM 2011c) and the load model underpinning the test method (Peterka et al., 1997) C onclusions presented below are positive in regards to the accuracy of ASTM D7158, though, they should not be regarded as a fu ll validation of ASTM D7158 and the load model. Results from this initial work indicate ASTM D7158 (ASTM 2011c) produces an accurate and potentially conservative portrayal of peak wind speeds above a common shingle roof and wind induced forces on the shing i.e., the wind load path). Distribution of peak wind speeds measured in full scale 25 mm (1 in) above a 6:12 single story building in the IBHS Research Center aligned with those collected in model scale presented in Cochran et al. (199 9 ) and Peterka et al. (1997). Dixon et al. (2013 a ) found an upper bound ratio of peak near roof wind velocity to mean approach velocity of 2.5, equivalent to the bound value proposed by Cochran et al. (1999) and Peterka et al. (1997). More importantly, the 2.5 factor serves as the gust factor for near roof wind speeds in ASTM D7158 (ASTM 2011c) In other words, peak wind speeds critical to the definition of wind forces on asphalt shingles appear accurate for simple shaped roof systems.
186 The second major find ing from this work reported in Chapter 6 indicates ASTM D7158 (ASTM 2011c) produces a conservative estimate of distributed wind force acting on the sealant strip of a fully sealed common three tab and laminate asphalt shingles. Moreover, this research measured a force mechanism acting on the sealant strip similar to that predicted by ASTM D7158 (ASTM 2011c) Thus, the magnitude and application of wind force acting on a fully sealed asphalt shingle appears accurately defined within ASTM D7158 (ASTM 2011c ) The research conducted by the author represents the first step towards a better understanding of wind loads on shingle roofing. The initial work address ed the accuracy of ASTM D7158 (ASTM 2011c) within the scope defined by the test methods underlying l oad model proposed in Peterka et al. (1997). That is, wind over a common shape Wind directions other than normal to the leading edge were not addressed herein, but are recommended in future work based upon findings of peak pressures on shingle roofing occurring in wind roughly parallel to the leading edge, as reported in Peterka et al. (1997). Additional work also remains on wind speeds over complex roof shapes common to modern homes. ASTM D 7158 Design M ethodology For shingles in the field of the roof, ASTM D7158 (ASTM 2011c) classifies wind resistance based upon a mean ASTM D6381 (ASTM 2008b) wind uplift resistance (20 specimens) to a peak design level wind force. As shown in Chapter 5, the resistance of mean. The question that remains is whether mean resistance in comparison to peak force produces an acceptable definition of resistance in a shingle r oof system. Moreover, r esistance at hip, ridge, eave, and rake locations on the roof are not
187 considered in ASTM D7158. Repeated observations of wind damage at these locations found in this research and in post hurricane damage reports motivate the incorpor ation of all system components in wind resistance classification process. A unified method for system resistance and a precise definition of acceptable performance based upon a Load and Resistance Factored Design methodology is recommended Eave and R ake Eave and rake shingles are among the most common forms of wind damage to asphalt shingle roofing. Installation error is most frequently ci ted cause in post hurricane reports The wind resistance of eave and rake edge details was assessed in this research as part of the full scale wind tunnel study at the IBHS Research Center. Results were discussed briefly in Chapter 4 and are expand upon in Dixon et al. (2013). E ave and rake shingle pull through at attachment points were observed on nearly all roofs eval uated at IBHS This was typically initiated by installation error (i.e., nails placed outside of manufacturer specification). However, edge details are prescriptively specified by shingle manufacturers and/or local building codes, rather than directly eval uated through performance testing methods (e.g., ASTM, UL, etc.). It is therefore difficult to assess proper installation guidelines for rake and eave attachment details. details, such as the addition of asphalt roof cement along the rake a nd eave, are available, but, to the a knowledge, their performance has not been measured. Recommendations for Future Research The findings culled from this research strongly indicate that without action, loss of shingle roof coverings will continue as the leading cause of residential property damage in hurricane winds The most significant opportunity for risk reduction identified from this
188 research involves p arti ally unsealed shingles in the field, hip, and ridge locations Two parallel efforts mus t take place to abate the ir occurrence. First, identify the cause of partially unsealed shingles. Second, develop reliable retrofit techniques for existing roofs containing partially unsealed shingles. For shingles in the field of the roof natural weather ing is the most likely source of partial unsealing. term expansion and contraction in the shingle system is a starting place for future research. Part of this work should include an improved understanding of the long term constitutive elements. Emphasis should be placed on the sealant strip, as the mode of failure in partially unsealed field shingles was a cohesive fracture in the sealant strip. For partially unsealed hip and ridge shingles future research should focus on installation methods and new products that increase the likelihood of adhesion on the downslope edge of the shingle. Retrofit techniques for unsealed field, hip, and ridge shingles are given in FEMA P 499 (2012). T he technique involves the application of asphalt roofing cement in 25 mm (1 in) dollops along the leading edges of unsealed asphalt shingles. Large scale implementation is questioned due to known blistering effects when excessive application occurs.
189 LIST OF REFERENCES Abraham, H. (1920). Asphalts and Allied S ubstances D Va n Nostrand Company, New York, Ny Applied Research Associates (2008). 2008 Florida Residential Wind Loss Mitigation S tudy Florida Office of Insurance Regulation Tallahassee, Fl ARMA (2006). Asp halt Roofing Residential Manual The Asphalt Roofing Manufacturers Association, Washington, D.C. ARMA (2011). The Bitumen Roofing Industry A Global Perspecitve The Asphalt Roo fing Manufactuers Association Washington, D.C. ASCE 7 02 Reston, Va ASCE 7 05 Reston, Va. um design loads for buildings and other ASCE 7 10 Reston, Va. Standard practic e for exposing nonmetallic materials in accelerated test devices that use laboratory light sources. G151 00 West Conshohocken, Pa ASTM. (2004). Standard practice for dark oven heat exposure of bituminous materials D5869 04, West Conshohocken, Pa G154 06, West Conshohocken, Pa ASTM ( 2008 a ). Standard practice for accelerated weathering test conditions and procedures for biduminous materials (fluorescent uv, water spray, and condensation method ). D4799 08 West Conshohocken, Pa ASTM (2008 b ). Standard test method for measu rement of asphalt shingles mechanical uplift resistance D6381 08 West Conshohocken, Pa Standard specification for asphalt saturated organic felt used in roofing and waterproofing D226 09 West Conshohocken, Pa Standard specification for asphalt shingles made from glass felt and surfaced with mineral granules D3462 10a West Conshohocken, Pa Standard specification for asphalt saturated organic felt underlayment used in steep slope roofing D4 869 05(2011)e1 West Conshohocken, Pa
190 ASTM (201 1 b ). Standard specification for gravity convection and forced ventilation ovens E145 94(2011) West Conshohocken, Pa Standard test method for wind resistance of asphalt shingles (uplift force/uplift resistance method). D7158 11 West Conshohoc ken, Pa est method for wind resistance of asphalt shingles (fan induced method) D3161 13, West Conshohocken, Pa resistance tests of roof shingles and roof ASTM Special Technical Publication 409 thering of roofing materials a Co nstruction and Building Materials 22, 423 433. Blake, E.G. (1925). Roof Coverings: Their Manufacture and Application Van Nostrand Co., New York, Ny 11th Conference on Roofing Technology Gaithersburg, Md CertainTeed (2007). CertainTeed Corporation, Blue Bell, Pa. distributions on low Architectural Science Review 42(3), 151 160 Fall Conferences of the Building Research Institute National Academy of Sciences, Washington D.C., 33 42. Professional Roofing Magazine R4 R8. 10 th Conference on Roofing Technology Gathersburg, Md 6 12. Quarterly Journal of the Royal Meteorological Society 87, 194. Dixon, C.R., Masters, F.J., Prevatt, D.O., and Gurley, K.R. (2012 perspective on t RCI Interface (5), 4 14.
191 Dixon, C.R., Masters, F.J., Prevatt, D.O., and Gurley, K.R. (2013a ). investigation of the wind resistance of asphalt shingle roof coverings SERRI Report 02 901 00 Southeast Region Resarch Initiative. Dixon, C.R., Masters, F.J., Prevatt, D.O., and Gurley, K.R. (2013 b Wind uplift resistance of artifically and naturally aged asphalt shingles Journal of Architectural Engineering [Article Submitted for Review] Dixon, C.R., Masters, F.J., Prevatt, D.O., Gurley, K.R., Brown, T.M., and Peterka, J.A. (2013 c The influence of unsealing on the wind resistance of asphalt shingles Journal of Wind Engineering and Industrial Aerodynamics [Article Submitted for Review] Meteorological Magazine 80, 181 187. Durability of Building Materials, 3, 213 223. ESDU (1 983). Data Item 83045 Strong Winds in t he Atmospheric Boundary Layer Part 2: Discrete Gust Speeds Engineering Science Data Unit. ESDU (2002). Data Item 82026 Strong Winds in the Atmospheric Boundary Layer Part 1: Hourly Mean Wind Speeds Engineering Scie nce Data Unit. fluorescent UV exposures: UVA Durability Testing of Non Metallic Materials (1294), 91 105. rmance: 2004 hurricane s FEMA 490 Federal Emergency Management Agency Washington, D.C FEMA 488, Federal Emergency Management Agency Washington, D.C mance: h FEMA 548. Federal Emergency Management Agency Washington, D.C FEMA P 757, Federal Emergency Management Agency Washington, D.C principles and practices of planning, siting, designing, constructing, and maintaining residential buildings in coastal areas FEMA P 55 Federal Emergency Management Agency Washington, D.C
192 FEMA P 499 Federal Emergency Management Agency Washington, D.C Florida Building Comission (2010) Florida Building Cod e Florida Building Comission, Tallahassee, Fl. RCI Interface, 9 40 46. traumatic Epidemiologic R eviews 27, 78 91. Physics Today 52(11), 48 52. Greenfeld, S.H. (1969). Performance of Roofing Felts Made w ith Asplund Fibers National Bureau of Standards Washington, D.C Gurley, K. R. and Masters, F. J. (2011). e sidential Nat. Hazards Rev 12(4), 177 183. Hahn, L.T., Roodvoets, D.L. RCI Interface Magazine 7, 36 42 Ham id, S.S., Pinelli, J.P., Chen, S., and based assessment of hurricane risk and estimates of potential insured losses for the Nat. Hazards Rev 12, 171 176. IBHS (2009). orces vs. Structural Strength Insurance Insti tute for B usiness & Home Safety, Tampa, Fl of surface Quarterly Journal of the Royal Meteorological Society 98, 563 589. Journal of Planning Literature 22(1), 3 15 scale turbu lence on J. Wind Eng. And Ind. Aerodyn. 83, 159 169. term and long term aging behavior of rubber Journal of the Transportat ion Research Board 1530, 11 17.
193 Liu, Z., Pogorzelski, H., Masters, F.M., Tezak, S. and Reinhold, T.A. (20 nature's fury: p RCI Interface Magazine RCI. RCI Interface, 4, 5 9. Southeastern Geographer 49(2), 108 131. Marechal, J., Ghaleb, G., and Bonnet, D. (1 983). Aging of Waterproofing Coatings Made From SBS Elastomeric Asphalt CTSB Studies and Research. 29 th Conference on Hurricanes and Tropical Meteorology Tucson, Az Bulletin of the American Meteor logical Society 91, 1665 1682. Contractors Guide (4), 48 49. displacement, and sheltering provision in the Un Nat. Hazards Rev., 13(2), 151 160. National Science Board (2007). "Hurricane warning: the critical need for a national hurricane research initiative ." National Science Board of the National Science Foundation Arlington, VA. Nichols, T. (2 Professional Roofing Magazine a century of success and 10 th Conference on Roofing Technology Gathersburg, Md 23 33. Ott, R. L. and Longnecker, M.T. (2004). A First Course in Statistical Methods Brooks/Cole, Belmont, Ca Parker, D., Sherwin, J. (1998). "Comparative summer attic thermal performance of six roof constructions." The 1998 ASHRAE Annual Meeting Toronto, Can ada. Peterka, J .A. and Cermak, J.E. (1983). Rpt. Wind Tunnel Study of Wind Resistance of Roofing Shingles Colorado State University, Fort Collins, Co
194 Peterka, J.A., Cermak, J.E., Cochran, L.S., Cochran, B.C., Hosoya, N., Derickson, R.G., Jones, J., and Journal of Architectural Engineering 3(4), 147 155. Peterka, J.A., Cerma k, J.E. and Hosoya, A. (1983). Rpt. Wind Tunnel Study of Wind Pressures on Roofing Shingles Colorado State University, Fort Collins, Co Peterka, J.A. and Este rday J. (2000). Development of a Shingle Uplift Coefficient Test S tandard Cermak Pe terka Petersen, Fort Collins, Co Powell, M.D., Houston, S.H., Amat, L.R., and Morisseau Leroy, N. (1 real J. Wind Eng. and Ind. Aerodyn. 77, 53 64 Clemson University, Clemson, S c RICOWI (2006). Hurricanes Charle y and Ivan I nvestigation R eport Roofing Industry Committee on Weather Issues, Powder Springs Ga RICOWI (2007). Hurricane Katrina Investigation Report Roofing Industry Committee on Weather Issues, Powder Springs, Ga. Robertson, R.E. (1991). Chemical properties of asphalt and their relati onship to pavement performance Strategic Highway Research Program Washington, D.C. RCI Interface, 2, 20 28. RCI Interface 4, 37 46. Shaw, D.E. ( ew approach for evaluation of asphalt shingle wind 1991 International Symposium on Roofing Technology Gaithersburg Md scale simulation of dynamic wind and pressure effects on full 12 th Americas Conference on Wind Engineering Seattle, Wa. Shiao, M.L., Nester, D.A. and Terrenzio, L.A. (2003 a Roofing Research and Standards Development 5, 119 135
195 Shiao, M.L., Snyder, R.A., Livsey, R.D., and Kalkanoglu, H.M. (2003b Roofing Research and Standards Development 5, 3 18. RSI (5), 44 48. 11th Conference on Roofing Technology Gaithersburg, Md 39 48. Smith, T.L. and McDonald, l essons from Professional Roofing Magazine North American Conference on Roofing Technology Snoke, H.R. (1941). Asphalt P repa red Roll Roofings and Shingles Nat ional Bureau of Standards, Washington, D.C. J of Wind Eng and Ind Aerodyn. 53(1 2), 145 155. Strahan, J.L. (1947). Asphalt Roofing and A ccessories: A Discussion Covering the Manufacture, S election and Application of Asphalt R oofing P roducts Asphalt Roofing Industry Bureau. Terrenzio, L.A., Harrison, J.W., l vs. artificial aging: u se of diffusion theory to model asphalt and fiberglass reinfoced Fourth International Symposium on Roofing Technology Gaithersburg, Md Texas Department of Wind In surance. the 2008 International Residential Code (IRC) and 2006 International Building < http://www.tdi.texas.gov/wind/documents/asphltshnglslst.pdf > (Oct 3, 2013) UL (1995). aterials Standard UL 997 Northbrook, Ill. UL (2003). est methods for wind resistant asphalt shingles with sealed tabs Standard UL 2390 Northbrook, Ill US Census Bureau. (2007). "200 7 population estimates ."
196 Proceedings of the National Academies of Sciences 34(11), 530 539. Roofer (7), 12 16. Bituminous Materials: Asphalts, Tars, and Pitches 249 306. Wypych, G. (1995). Handbook of Materials Weathering 2 nd ed, William Andrew Publishing, New York, Ny Construction and Building Materials, 21, 66 72.
197 BIOGRAPHICAL SKETCH Craig R. Dixon received his undergraduate degree in civil engineering in the fall of 2008 from the University of Florid a and in the fall of 2009 he joined Dr. David O. assistant. During his undergraduate work, Mr. Dixon spend six summers interning for Gale Associates in Orlando, FL, performing roof observation and roof assessment resistance of wood roof sheathing, standing seam metal roofing, and asphalt shingles. Mr. Dixon received his Ph.D. from the University of Florida in the fall of 2013.