Investigating a wind tunnel method for determining wind-induced loads on roofing tiles

J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Contents lists available at ScienceDirect

Journal of Wind Engineering and Industrial Aerodynamics journal homepage: www.elsevier.com/locate/jweia

Investigating a wind tunnel method for determining wind-induced loads on roofing tiles Daniel J. Smith a,b,n, Forrest J. Masters a, Arindam G. Chowdhury c a

Department of Civil and Coastal Engineering, University of Florida, 365 Weil Hall, Gainesville, FL 32611, USA College of Science, Technology, and Engineering, Cyclone Testing Station, James Cook University, Townsville, QLD 4811, Australia c Department of Civil and Environmental Engineering and International Hurricane Research Center, Florida International University, 10555 W. Flagler St. Miami, FL 33174, USA b

art ic l e i nf o

a b s t r a c t

Article history: Received 28 April 2015 Received in revised form 16 May 2016 Accepted 16 May 2016

Current design loads for roofing tile systems in the U.S. are determined based on a standardized wind tunnel testing method developed in the 1990s to examine wind-induced pressures on the upper and lower surfaces of the tile. The method neglects several key parameters that are well known to affect wind loading (e.g. wind angle, specimen shape, etc.). The research objective of this study is to investigate this method by [1] characterizing wind-induced surface pressure distributions on field tiles for varying wind angles of attack and [2] measuring load path intensity through mechanically fastened tile attachments. Surface pressure distributions were measured on three full-size, rapid prototyped roofing tile models with 256 pressure taps immersed in wind flows. The models are geometrically identical to low-, medium- and high-profile concrete roofing tiles that are widely used in high wind areas. Additionally, their real counterparts were instrumented with load cells to measure reaction forces at mechanical fasteners. The results highlight areas of the method that are lacking in specificity and shows that low-resolution pressure measurement may yield conservative parameters for low- and medium-profile tiles but is potentially not conservative for asymmetric (s-shaped) high-profile tiles. & 2016 Elsevier Ltd. All rights reserved.

Keywords: Roofing tiles Roof damage Florida Building Code International Building Code Residential buildings Surface pressure measurements Wind loading

1. Introduction Annual hurricane-induced economic losses have increased steadily in the U.S. during the past 50 years, averaging $1.3 billion (in constant 2006 dollars) from 1949 to 1989, $10.1 billion from 1990 to 1995, and $35.8 billion from 2001 to 2006 (National Science Board, 2007). Wind related damage on the East and Gulf coasts of the U.S. alone averaged $5 billion in annual economic losses as of 1998 (Pielke and Landsea, 1998). Florida is a particularly strong contributor to annual losses, largely due to the high likelihood of hurricane landfall in the Southeast U.S. Insured losses related to roof cover damage (i.e. shingles, roofing tiles, metal) accounted for over 50% of insured losses in Florida during the 2004–2005 hurricane seasons (ARA, 2008). Although asphalt shingles are the most prevalent form of roof cover in Florida, roofing tile systems represent significant market share, particularly in the South Florida region n Corresponding author at: College of Science, Technology, and Engineering, Cyclone Testing Station, James Cook University, Townsville, QLD 4811, Australia. Tel.: þ 61 7 4781 5512. E-mail addresses: [email protected] (D.J. Smith), [email protected]fl.edu (F.J. Masters), chowhur@fiu.edu (A.G. Chowdhury).

http://dx.doi.org/10.1016/j.jweia.2016.05.006 0167-6105/& 2016 Elsevier Ltd. All rights reserved.

due to their esthetic appeal, ventilating characteristics, and durability. As of 2011, tile roofing market share was 56%, 36%, and 24% for Broward, Palm Beach, and Lee Counties respectively (FPHLM, 2011). Insurance claim analysis post-2004 hurricane season in Florida suggested insured losses for tiled roofs were greater than asphalt shingles when wind speeds were greater than 54 m/s (120 mph). Replacement costs for roofing tiles are a key factor, exceeding those of asphalt shingles in some cases by 400% (ARA, 2008). Post-2004 hurricane damage assessments indicated that in several instances, tiles did not perform as predicted (i.e. failed at less than design level wind speeds) by the 2001 FBC and FRSA/TRI Concrete and Clay Tile Installation Manual (FEMA, 2005). In addition to cladding replacement costs, damage to any roof covering system during hurricane events increases likelihood of water ingress related damages, making roof cover loss a leading cause of building performance issues during hurricanes (FEMA, 2005). Field tiles are those in the zone that generally make up the largest surface area of the roof (i.e. the “field” of the roof) and include all tiles not within edge/corner zones or along ridge, hip, and gable end lines. Several full-scale studies have examined wind loading on roofing field and ridge tiles (i.e. Robertson et al., 2007; Laboy-Rodriguez et al., 2013; Tecle et al., 2013a, 2013b; Li et al.,

48

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

2014), but not directly examined the methodology employed by design standards. Internationally, the majority of standards for wind resistance of roofing tiles systems were introduced or underwent significant development in the 1990s, including British Standard BS 5534 (BSI, 2015), Dutch Standard NEN 6707 (Netherlands Standardization Institute, 2011), European Standard EN 14437 (CEN, 2004), Australian Standard AS 2050 (Standards Australia, 2002), and the USbased standard SSTD-11 (SBCCI, 1999). The British and U.S. methods were both heavily influenced by research conducted in the UK in the late 1980s and early 1990s by Redland Technology. All of the standards are based on wind tunnel tests of some form, however, the U.S. standard is the only one to include a wind tunnel testing method for determining local wind-induced pressure distributions on roofing tiles. This paper briefly reviews a Redland Technology (1991) study, which serves as the basis for current roofing tile design provisions in the U.S.. The study was based on several of the first published works on wind loading mechanisms of roofing elements (i.e. Kramer et al., 1979; Hazelwood, 1980, 1981; Kramer and Gerhardt, 1983). A review of the current design provisions, including limitations, is also presented. Finally, the first two of four experiments recently conducted to investigate the wind resistance of roofing field tiles are discussed. Experimental findings are discussed in context of the 2014 Florida Building Code (FBC) and 2015 International Building Code (IBC) provisions for roofing tiles. The third and fourth experiments in the series are discussed in a separate companion paper.

is provided in Table 1 (Note: Table 1 provides the first publication year while references throughout the paper provide the most current publication year). The focus of this paper is limited to the wind tunnel testing method described by TAS 108, C1569, and SSTD 11 (see Eqs. (2)–(4)). Key limitations of the Redland study include: [1] the approximated relationship between near-roof flow and approach flow conditions (originally proposed by Hazelwood, 1981), [2] limited wind angles of attack (perpendicular to the leading edge only), and [3] use of low-resolution surface pressure measurements to determine design parameters (e.g., lift coefficients) for roofing tiles. Since the study, progress has been made and it is known that the method was not fully representative of wind loading mechanisms for roofing tiles. While the mechanisms are briefly discussed by Smith et al. (2014), and are known to include cavity pressure effects and local effects on the external surface of tiles, a critical review of the mechanisms and the progression of related research to date is not yet available in the literature.

3. 2010 Florida Building Code In 2002, the 2001 FBC officially superseded all local codes in the State of Florida. The design guidelines for roofing tiles were modeled after SBC provisions (based on Redland Technology work). The wind-load interaction for tiles is expressed in the 2010 FBC (Building) chapters (appears as Eqs. (16)–(33)) as an overturning moment: M a ¼ qh C L bLLa ð1  GC p Þ

2. The “Redland” study In 1991, the Southern Building Code Congress International (SBCCI) commissioned Redland Technology (a UK-based company) to investigate wind loads on roofing tile systems and develop a code consistent design methodology. Two experiments were performed by Redland to develop their design method: [1] wind loads were estimated from wind tunnel tests where surface pressures on medium and high profile roofing tiles were measured as wind was blown across a tile array and [2] wind uplift resistance was estimated from constant displacement rate uplift tests that quantified the uplift resistance of roofing tiles with various attachment methods. The resulting method was incorporated into the Standard Building Code (SBC), and later the Florida Building Code (FBC) and International Building Code as Equation 16–33 and Equation 16–34 respectively (see Eq. (1)), FBC Testing Application Standards (TAS) 101, 102, 102A, 108 (FBC, 2014a, 2014b, 2014c, 2014d, 2014e) and FBC Roofing Application Standard (RAS) 127 (FBC, 2014f). The study is also the basis for SBCCI 11-99 (SBCCI, 1999) and ASTM International standards C1568, C1569, and C1570 (ASTM, 2003a, 2003b, 2003c). A brief overview of these standards

ð1Þ

where, b is the exposed width of the roofing tile, C L is the tile lift coefficient, GC p is the roof pressure coefficient for each applicable roof zone determined by ASCE 7 (not adjusted for internal pressure), L is the length of the roofing tile, La is the moment arm from the axis of rotation to the point of uplift on the roofing tile, M a is the aerodynamic uplift moment acting to raise the tile leading edge, and qh is the wind velocity pressure determined from ASCE 7 (ASCE, 2010). Eq. (1) is also used as the design equation for roofing tiles in the 2012 IBC (appears as Eq. (16)–(34)). In design, the aerodynamic uplift moment is compared to the attachment resistance moment (M f ), determined via TAS 101/102/102A testing and included in product approval documentation for the tile. Eq. (1) may be used for design in areas outside of the HVHZ. The lift coefficient (C L ) is determined by laboratory testing per TAS 108 (FBC) or SSTD-11 (IBC). The method of these standards is generally the same as both are derived from the work of Redland Technology. TAS 108 is the more detailed document and will be used for discussion in this paper. Tile specimens are fitted with pressure taps along the centerline and subjected to wind loading from the longitudinal direction (perpendicular to the leading

Table 1 Progression of standardized test methods for roofing tiles from the Redland Technology (1991) study to the present (Smith et al., 2014). Designation

First Publication Basis

SSTD 11

1993

FBC TAS 101 1995 FBC TAS 102 1995 FBC TAS 102A 1995 FBC TAS 108 1995 FBC TAS 116 1995 ASTM C1568 2003 ASTM C1569 2003 ASTM C1570 2003

Redland

Overview

Includes methods for determining uplift capacity of mechanical, mortar, and adhesive attachments. Air-permeability method added in 1999 revision. Redland Static uplift capacity of mortar or adhesive tile attachments. Redland Static uplift capacity of mechanical tile attachments. Redland Static uplift capacity of mechanical tile attachments with clips. Redland Wind tunnel test for determining overturning moment coefficients and aerodynamic load multipliers for tiles. BS5534/Redland Procedure for determining air permeability of rigid, discontinuous, roofing systems. SSTD 11/Redland Mechanical uplift resistance testing. Derived from SSTD 11, essentially a combination of TAS 101, 102, 102A. SSTD 11/Redland Wind tunnel test for determining overturning moment coefficients. Derived from SSTD 11, similar to TAS 108. SSTD 11/Redland Test for determining air permeability of a roofing tile system. Derived from SSTD 11–99 update, similar to TAS 116.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

49

moment (M f ), determined via TAS 101/102/102A and also included in product approval documentation for a tile. Eqs. (4) and (5) are not used in the 2015 IBC. The TAS 108 method employs two key assumptions: [1] spatial variation of surface pressures is negligible across the width of tiles, including asymmetric tile profiles (e.g. high-profile s-shapes) and [2] flow perpendicular to the leading edge results in the largest wind uplift forces.

4. Current research focus This study investigated the wind resistance of roofing tile systems, specifically field tiles, emphasizing systems typical to the State of Florida. Key objectives were to [1] develop a better understanding of the wind-induced load interaction created via the traditional wind tunnel test and [2] examine the Florida Building Code (and IBC) provisions for roofing tile systems. 4.1. Experiment one: characterization of the pressure field acting on the roofing tile Fig. 1. Schematic of 2010 Florida Building Code (TAS 108) and 2012 International Building Code (SSTD-11) methods for computing roofing tile lift coefficients (CL ) (lower tile surface not shown).

edge). Measured surface pressures are used to compute the lift coefficient by assigning tributary areas to each pressure tap location. Each tributary area extends the entire width of the tile (Fig. 1). P P C pt δt i C pbi δbi CL ¼  i i i ð2Þ lb lb where C pb i is the ith pressure coefficient on the bottom surface of the tile, δbi is the tributary area of corresponding bottom surface pressure coefficient, C pt i is the ith pressure coefficient on the top surface of the tile, δt i is the tributary area of corresponding top surface pressure coefficient, l is the length of the tile, and bis the exposed width of the tile. The design method required within the HVHZ is presented in RAS 127 and includes two additional parameters computed per TAS 108 (C Ma ,λ). The coefficient of moment, C Ma , is expressed as follows: P P 0 0 i C pbi δbi lb i i C pt i δt i lt i C Ma ¼  ð3Þ 2 2 l b l b 0

where lb i is the moment arm acting from the head of the tile (or 0 batten) at the ith pressure tap on the bottom surface of the tile, lt i is the moment arm acting from the head of the tile (or batten) at the ith pressure tap on the top surface of the tile, and all other variables are as defined in Eq. (2). The aerodynamic multiplier is an alternative form of the coefficient of moment computed by normalizing C Ma by the length and width of the tile as follows:

λ ¼ C Ma  b  l2

ð4Þ

where all variables are defined previously. The aerodynamic multiplier must be included in the product approval documentation for a tile and is used with the following RAS 127 design equation to compute the wind-induced moment that tile attachments must be able to resist:   M r ¼ P asd1;2;3  λ  M g ð5Þ where λ is the aerodynamic multiplier determined via TAS 108, M g is the self-weight moment, P asd is the minimum design wind uplift pressure computed in accordance with ASCE 7 (P asd ¼0.6P ult per specification by RAS 127) for field (P asd1 ), perimeter (P asd2 ), and corner (P asd3 ) areas of the roof, and M r is the required moment of resistance that must be exceeded by the attachment resistance

The research goal for experiment one was to characterize the wind pressure distribution on low-, medium-, and high-profile field tiles for multiple wind angles using a high-resolution array of pressure tap measurements. Experiment one was designed as a modified version of TAS 108. Key modifications include a larger distribution of pressure taps (256 vs 27) and multiple wind angles of attack. Additional deviations from TAS 108 include: choice of underlayment (self-adhered vs. standard two-ply 30/90 system for TAS), absence of plenum chamber below test deck to simulate internal pressure, number of tile courses per specimen (5–6 vs 9 for TAS) and reference wind speed range (18–35 m/s vs 31–49 m/ s for TAS). It was assumed that underlayment type (and varied surface roughness between different types) would have negligible impact on wind-induced pressures along the tiles above. The effects of simulated internal pressure were found negligible by Redland (1991) for this type of experimental configuration. The dimensions of the testing apparatus allowed for installation of five or six tile courses as opposed to the nine courses required by TAS 108. Although this limits the length for flow development upwind of pressure measurements, the analysis for this experiment primarily makes comparisons within the data set (i.e. varied wind angle, number of pressure taps) as opposed to comparisons with loads on in situ tiles. Further, reattachment of the separated flow near the leading edge (i.e. first course) of the arrangement occurred before the start of the second course and therefore did not have a significant effect on measurements in fourth and fifth courses. The reference wind speed range was chosen primarily due to the fragile nature of the tile models. Wind conditions were simulated by the University of Florida Dynamic Flow Simulator (DFS), which is capable of generating wind speeds up to 100 m/s (224 mph) across a test deck. Tile array specimens were attached to the rectangular test deck either directly or above battens. To characterize wind loading, full-size model replicas of low-, medium- and high-profile tiles were used to measure surface pressures through 256 taps connected to a Scanivalve pressure scanning unit. Three-axis wind velocities were measured inside the DFS test section using a Turbulent Flow Instruments Cobra Probe mounted above the deck and upwind of specimens. Fig. 4 shows typical arrangements for experiment one inside the DFS. Each of the six tile profile/attachment combinations were subjected to three different reference wind speed levels at each of three wind angles of attack. Exact velocity at each of the three

50

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

levels varied by 72 m/s depending on DFS engine conditions. Wind speed levels 1, 2, and 3 correspond to 18–22 m/s (40– 49 mph), 26–30 m/s (58–67 mph), and 31–35 m/s (69–78 mph) respectively. The three wind angles of attack included 0° (perpendicular to the leading edge), 45°, and 90° (parallel to the leading edge). A single model tile was installed within each array for pressure measurements (Fig. 5). 4.1.1. Model roofing tiles Replicas of low-, medium-, and high-profile roofing tiles approved for use in the State of Florida were rapid prototyped at the University of Western Ontario. Hollow shells of the tiles were fabricated from resin and fitted with 256 pressure taps distributed throughout the upper and lower surfaces (Fig. 2). Each tap connects to a 1.6 mm (0.063 in.) diameter vinyl tube inside of the model. Tubes exit the model through the rear face and connect to ports on a pneumatic bulkhead connector. The bulkhead connectors provide an air-tight connection to the scanning unit. Corrections for dynamic response characteristics of the tubing were considered but not necessary, the analysis was limited to mean pressures only.

4.1.2. Pressure scanning system The Scanivalve Corporation unit is capable of 625 Hz measurement at 256 channels. The unit consists of four electronic pressure scanning modules and a single processing Ethernet remote A/D. Each module contains 64 silicon pressure transducers calibrated to a maximum range of 71245 Pa ( 75 in water column) and a high speed multiplexer (45 kHz). The digital signal processor utilizes a pressure temperature look-up table to compensate the pressure sensors for temperature change. 4.1.3. Testing apparatus The Dynamic Flow Simulator (DFS) was used to generate wind above the tile array specimens (Fig. 3). Air enters through a 1.52 m (5 ft) diameter inlet and passes through an actively controlled opposed-blade damper system, which regulates the system resistance. The air is then pulled through an 1800 HP centrifugal blower which can accelerate the air to velocities of up to 100 m/s (224 mph) once it passes through a settling chamber consisting of wide angle diffuser turbulence screens, a honeycomb, and a 5:1 ratio contraction duct that connects to the test section. The cross sectional area at the entrance to test section is 213 cm (7 ft) wide by 38 cm (1.25 ft) tall. At the exit, the width is 213 cm (7 ft); however, the height may be adjusted to regain static pressure lost

Fig. 2. Low-profile (upper left), medium-profile (upper right), high-profile (lower left), and medium-profile interior (lower right) rapid prototype models of typical Florida roofing tiles manufactured at University of Western Ontario and made from hollow resin shells fitted with 256 pressure taps each.

Fig. 3. Dynamic flow simulator at University of Florida configured for experiment one and two testing.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

51

due to friction across the test section. The test decks were attached to the 243 cm (8 ft) long by 182 cm (6 ft) wide opening in the bottom floor of the test section using a hydraulic lift located below the test section. The top surface of the test deck sits at the same horizontal plane as the bottom floor of the test section. Turbulence intensity in the test section at the reference location upwind of specimens was measured via cobra probe consistently in the range of 1–3% (analysis was limited to mean values only). 4.1.4. Velocity measurement system A Turbulent Flow Instrument Cobra Probe was used to measure three-axis wind velocities inside the test chamber. These are multi-hole pressure probes capable of resolving the three components of velocity and measuring local static pressure. Probe characteristics include an ability to measure flow within a 745° cone, at a maximum sampling frequency of 2000 Hz and accuracy of 70.5 m/s and 7 1° yaw up to 30% turbulence intensity. Sampling frequency for this study was 1250 Hz based on previous work by the authors in full-scale flow fields. TAS 108 requires velocity measurement at a height of 10.1 cm (4 in.) above the tile specimens, which sit at a height of approximately 6 cm (2.4 in.) above the test deck (varies with tile profile and presence of battens). Therefore, the probe was used to record velocity and local static pressure 16 cm (6.3 in.) above the test deck (for all configurations) at a location 10.1 cm (4 in.) downwind from the leading edge of the test deck (still upwind of the specimens) and 10.1 cm (4 in.) from the lateral edge of the test deck to avoid disrupting specimen approach flow (Fig. 4). 4.1.5. Test deck preparation A wood frame test deck was constructed as the substrate for tile array attachment. The rectangular opening in the floor of the DFS test section is 183  244 cm2 (72  96 in.2). A flange was installed around the perimeter of the opening by welding 2.5  2.5  0.6 cm3 (1  1  0.25 in.3) steel angles to each of the four sides. The flange is oriented such that a 2.5  0.6 cm2 (1  0.25 in.2) portion of the steel angle protrudes from the top

Fig. 4. Internal views (looking downwind) of the dynamic flow simulator configured for experiments one and two with a typical high-profile (upper) tile setup showing instrument locations and a typical low-profile (lower) tile setup showing dimensions. Note: self-adhered underlayment not shown.

Fig. 5. TFI Cobra Probe setup at 16 cm reference height for a low-profile tile array attached over battens and oriented for a 45° wind angle of attack inside the Dynamic Flow Simulator (DFS) during experiment one.

edges of the opening, level with the floor of the test section. This configuration allowed the test deck to be lifted into the test section from below and pressed against the flange, minimizing the transition between test section floor and test deck to a 0.6 cm (0.25 in.) step (see Fig. 4). The test deck consisted of 5  15.2 cm2 (2  6 in.2) wood framing at 61 cm (2 ft) on center. Two saw-cut 121  180  1.1 cm3 (47.5  71  0.451 in.3) sheets of wood sheathing were fastened to the test deck framing. A self-adhered underlayment was installed per manufacturer specification over the test deck sheathing. 4.1.6. Specimens Each array specimen consisted of a low-, medium-, or highprofile replica tile installed within an array of concrete tiles of matching profile (Fig. 5). Arrays were typically five tiles wide by five or six courses long and attached either directly to the test deck or over nominal 2.5 cm (1 in.)  5 cm (2 in.) wooden battens using two #8  7.6 cm2 (3 in.) screws. In order to simulate variation in wind attack angle, arrays were installed with orientation of 0° (wind perpendicular to tile leading edge), 45° or 90° (wind parallel to tile leading edge). Some arrays installed at 45° orientation, included only five courses due to the size restrictions of the test section (Fig. 6). In all cases the replica tile was installed at the center of the array and in the second to last course. Tubing from the replica was passed through a 5 cm (2 in.)  15.2 cm (6 in.) cutout in the test deck located just downwind of the replica's trailing edge below the overlapping tile in the last course. 4.1.7. Procedure Specimens were constructed and lifted into the DFS test section from below. The cobra probe was installed at the reference location near the leading edge of the deck. Tubing from the replica tile was connected to the pressure scanning unit located below the test section. Vinyl tubing from both the cobra probe and the pressure scanning unit were connected to a reference pressure apparatus located at ground level below the test section and designed to minimize fluctuations in local ambient pressure. The apparatus was designed by Scanivalve and consisted of a 10 cm (3.9 in.) diameter plastic tube filled with a material for damping and sealed on each end with an instrument tubing port and a small (less than 1 mm diameter) pressure port that opened to ambient. Pressure time-histories were recorded at each of three wind velocity levels for each tile configuration. Each pressure scan was 120 s with 625 Hz sampling.

52

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Fig. 6. High-profile experiment one tile array specimen configured for 0° wind angle of attack (left), tubing from the tile array specimen is connected to the pressure scanning system located below the DFS test section (center), a low-profile specimen configured for 45° wind angle of attack is being prepared for testing inside the DFS (right).

Fig. 7. ATI Nano25 (IP65) six-axis load cell (left), withdrawal (z-direction) and shear (x–y directions) reference axes where wind azimuth is aligned in the x-direction (center), and typical load cell arrangement for battened tile attachment (right).

4.2. Experiment two: tile attachment reaction measurement Research goal of this experiment was to directly quantify the reaction forces of field tile mechanical fastener attachments so that results could be compared to experiment one findings. Concrete field tiles were affixed to a six-axis load cell at the mechanical fastening location and subjected to wind loading in the DFS test section. The load cell provided direct measurement of wind generated withdrawal and shear forces at the fastener. For each array specimen, a single instrumented tile was installed in the center of the fifth (or fourth for 45° cases) course. Test deck preparation and velocity measurement were as described for experiment one (4.1.4 and 4.1.5).

4.2.1. Instrumentation An ATI Industrial Automation model Nano25 (IP65) six-axis load cell measured wind induced reaction forces on roofing tile mechanical fasteners. The load cell is capable of resolving forces in the x- (shear), y- (shear), and z-axes (withdrawal) (Fig. 7) and consists of a 28 mm (1.1 in.) diameter by 28 mm (1.1 in.) steel cylinder with silicon strain gauges fixed on the interior face. A custom-machined 3.81 cm (1.5 in.) by 3.81 cm (1.5 in.) by 1.27 cm (0.5 in.) steel block, with threaded #8 screw hole in the center, was secured to the upper face of the load cell to allow mechanical fastener attachment. For each tile array, the top of the load cell arrangement was installed level with the test deck (or batten) and directly below the tile installation hole nearest to the under lapping edge (left side from leading edge view of the tile). Load readings were captured with 100 Hz sampling rate via National Instruments Labview 2010 and a DAQ analog-to-digital converter.

4.2.2. Specimens Each specimen consisted of a low-, medium-, or high-profile array of concrete tiles attached either directly to the deck or over nominal 2.5 cm (1 in.)  5 cm2 (2 in.) timber battens. A single tile in the center of the second to last course of each array was affixed to a load cell below using a single #8 machine screw (Figs. 8 and 9) of variable length depending on the configuration. All other tiles in the array were mechanically fastened using one #8  6.4 cm2 (2.5 in.) Quik Drive tile roofing screw. In order to adjust wind angle of attack, tile arrays were installed with orientation of 0° (wind perpendicular to tile leading edge), 45°, or 90° (wind parallel to tile leading edge). 4.2.3. Procedure For each test, data acquisition was initiated before wind loading (i.e. 0 m/s). After 10 s, velocity in the test section was increased to reference level 1 (18–22 m/s) for 60 s (note: a sampling time of 60 s was chosen to expedite testing since load cell force readings were relatively constant once wind load was applied). Finally, velocity was reduced to 0 m/s and after an additional 10 s, data acquisition was terminated to complete a “step-up” and “stepdown” time history sequence. The sequence was repeated for three trials at each of the three reference wind speed levels (nine total time histories) for each of the 18 combinations of tile profile (high-, medium-, low-), attachment (direct, battens), and wind angle of attack (0°, 45°, and 90°). Load cell readings within the first 10 s represent the effect of installation (e.g., screw tightness, tile overlapping, etc.) in the arrangement prior to wind loading and were generally in the order of 1–5 N. The mean value from this 10 s portion of the record was used in post-processing to tare the wind load portion of the time history.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

53

Fig. 8. Six-axis load cell arrangement for single-screw mechanically attached roofing tiles (direct to deck) in experiment two (Note: direction of wind loading is right to left in the figure.)

Fig. 9. Experiment two single-screw mechanically fastened tile load cell arrangement with cover plate removed showing mounting configuration for load cell below tile arrays (left), the cover plate installed with load cell in place and ready for tile installation (center), and tile array installed with the specimen tile fastened to load cell arrangement (right).

Fig. 10. High-profile tile geometry (left) and example upper surface C P distribution (right) with schematic of integration element location, i, and local 4-node C 0pt i values and slopes φi and θi for calculation of lift and moment coefficients using the 256 tap methodology.

54

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

5. Results Tile surface pressure data from experiment one were analyzed to assess the sensitivity of roofing tile lift coefficients and aerodynamic multipliers to number of pressure taps (256 vs 27), installation configuration (direct vs battens), and wind angle of attack and to investigate pressure equalization. Mechanical fastener shear and withdrawal reaction forces from experiment two were analyzed for comparison to experiment one findings. The results are intended for research discussion as outlined in this paper and should not be used directly for design purposes. 5.1. Design parameter sensitivity to number of pressure taps The mean surface pressure for the sample period was computed at each tap and converted to a dimensionless pressure coefficient using the following equation: CP ¼

P  P1 q

ð6Þ

where C P is the coefficient of pressure, P is the mean local pressure at the tap, P 1 is the mean free stream static pressure at 16 cm reference height, q is the mean velocity pressure at 16 cm reference height. The tile models were used for all experiment one testing, however, the lift coefficient (C L ) and aerodynamic multiplier (λ) for each tile configuration were computed using two different methods for comparison. The first method represents the design parameters as calculated per TAS 108 (Eq. (2) and Fig. 1) with 27 centerline surface pressure measurement locations. Since pressure taps on the models were not located at all 27 TAS 108 locations, in some cases surfaces pressures were interpolated for TAS 108 locations using actual centerline taps on the models. The second method incorporates all 256 taps on the upper and lower surfaces of the models. Measured pressures were used to fit gridded pressure surfaces for the upper and lower faces of the tile with dimensions extending to the edges of the tile (Fig. 10). This was achieved using a custom routine developed via Matlab R2011a. The fitting component of the routine uses a distance-based Greens’ function approach. A pressure coefficient was computed for each rectangular element in the grid by averaging pressures at the four nodes (Eq. (7)) that make up the element and multiplying by the elemental area dxdy. C 0pt i ¼ ½C 0pt i11 þ C 0pt i12 þC 0pt i21 þ C 0pt i22=4

ð7Þ

C 0pt i

is the 4-node average pressure coefficient at an element where on the top surface of the tile, C 0pb i is the 4-node average pressure coefficient at an element on the bottom surface of the tile (note: lower surface equation is identical but not shown), and C 0pt i11; C 0pt i12; C 0pt i21; C 0pt i22 is the pressure coefficients on the interpolated surface at the four nodes that make up element i. In order to account for the curvature of tile profiles (i.e. when normal forces are not in the vertical direction), a set of threedimensional geometric points (1 cm resolution) for the outer shell of each tile were either provided by the manufacturer or manually recorded using a three-dimensional scanner. These points were imported to Matlab and used to compute the slope at each element location in the gridded pressure surfaces. Slope data were combined with pressure data at each element to calculate the lift coefficient using the following equation: P 0 P 0 i C pb iδi cos ðφi Þ cos ðθ i Þ i C pt iδi cos ðφi Þ cos ðθ i Þ C 0L ¼  ð8Þ lb lb where C 0pb i and C 0pt i are defined previously, δi is the area (dxdy) of a single element, l is the length of the tile, and b is the exposed width of the tile, θi is the slope (dz=dx) at an element in the

Fig. 11. Lift coefficients for roofing tiles at three reference wind speed levels and 0° wind angle of attack. H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 256, 027 indicate the number of pressure measurement locations used to compute C L respectively.

Fig. 12. Aerodynamic multipliers for roofing tiles at three reference wind speed levels and 0° wind angle of attack. H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 256, 027 indicate the number of pressure measurement locations used to compute λ respectively.

direction parallel to the leading edge, φi is the slope (dz=dy) at an element in the direction perpendicular to the leading edge (see Fig. 10 for axes designations). The decrease in lift coefficient due to slope consideration for low-, medium-, and high-profile tiles is approximately 9%, 14% and 16% respectively. Similarly the coefficient of moment was computed by incorporating the distance of each elemental areaδi , from the axis of rotation for the tile as follows:     P 0 P 0 0 0 i C pb iδi lδ i cos ðφi Þ cos ðθ i Þ i C pt iδi lδ i cos φi cos θ i 0  ð9Þ C Ma ¼ 2 2 l b l b 0

where lδ i is the moment arm acting from the head of the tile (or batten) to the ith elemental area and all other variables are as defined previously. The aerodynamic multiplier is computed using Eq. (4). Lift coefficients (Fig. 11) and aerodynamic multipliers (Fig. 12) were computed for a 0° wind angle of attack using 27 pressure locations along the centerline (i.e. the approach of standardized tests) and all 256 distributed pressure taps. In all testing configurations, lift coefficients computed using the 27 centerline locations were larger than those computed with the 256 tap data.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

High-, and medium-profile configurations show greater variation between the methods, likely due to the variation in profile (and therefore reduced vertical component of uplift force) from the centerline of the tile outward for those two profiles. The variation between methods was least significant for low-profile configurations. Attachment method (i.e. direct to deck vs. battens) did not appear to have significant effect on variation between the two methods. In order to use Eq. (1), C L must be computed per TAS 108 (SSTD-11 for IBC) or 0.2 can be used as a default value. All configurations except medium-profile direct and low-profile with battens suggest 0.2 is a conservative estimate for design. However, future research should seek to verify these estimates with testing in full-scale flow fields. Computed aerodynamic multipliers (λ) suggest that the 27 centerline tap configuration underestimates the lifting moment for high-profile tiles. For both direct and battened attachment configurations, λ-values were larger when computed using 256 taps for this profile. The 256 tap configuration values were only slightly larger for low-profile configurations and were smaller for both medium-profile configurations. The results are related to differences in the two analysis methods. The TAS 108 method (i.e. 27 taps) uses Eq. (3) with centerline taps only and the moment arm for each tap from the axis of rotation (head of tile or batten). The alternative method (i.e. 256 taps) described in Eq. (9), follows a similar logic except that elemental areas of surface pressure and their moment arms cover the entire width and length of the tile as opposed to the centerline only. The result is that variations in surface pressure when moving outward from the tile centerline are included in calculation of the moment coefficient and have increasingly significant effect approaching the leading edge of the tile due to longer moment arms. For example, the asymmetric sshape of high-profile tiles means that surface pressures will vary significantly when moving outward from the centerline, in particular near the leading edge region. Therefore the contribution of surface pressures at the leading edge is different than if only the centerline pressure were to be considered. In contrast, the lowprofile tile is relatively flat, surface pressures do not change dramatically when moving away from the centerline, and therefore lift and moment coefficients are more similar in magnitude for both the TAS 108 method and the 256 tap method. Manufacturer product approval values (calculated per TAS 108 by a certified laboratory) for each tile were also plotted in Fig. 12. Values for medium-profile directly attached and low-profile battened configurations were larger than product approval values for both computation methods. However, the opposite was observed for high-profile configurations; observed λ-values were

55

significantly lower than product approval values and other profile values despite similar lift coefficient magnitudes for all profiles. Specifications for the wind tunnel apparatus, flow regime, and leading edge conditions are limited in detail for TAS 108. The most recent TAS 108 product approval testing for the three products tested was conducted (circa 1994) in the Redland Technology wind tunnel (resultant values are in Fig. 12). Variations induced by testing in the DFS, as opposed to the Redland tunnel, in addition to the other deviations from TAS 108 (mentioned previously), may have had a more dramatic effect on surface pressures for highprofile specimens. Regardless of parameter magnitude, the analysis emphasizes the relative variation between the two computation methods, rather than the magnitude of computed parameters. For the current experiment, the aerodynamic multiplier computed using the 256 tap method versus the standardized method (centerline taps only), had a significantly higher magnitude. As this parameter is moment-based, the effects of asymmetry and hence surface pressure variations across the highprofile tile are more pronounced than for other profiles. These effects are not accounted for using the standardized method but can be significant as indicated by the 256 tap method. Lift coefficients showed slight dependence on the reference wind speed (an average of 710% from 18 to 35 m/s). In general, higher wind speeds (i.e. levels 2 and 3) resulted in larger lift coefficients. Longitudinal and vertical components of turbulence intensity in the test section were found to be inversely proportional to wind speed, decreasing from 2.7% to 1.0% and 1.8% to 1.0%, respectively, with an increase in mean wind speed from 19 m/s to 35 m/s. Turbulence is known to promote earlier reattachment (a reducing effect) in flow separation zones. Higher levels of turbulence at comparatively low wind speeds in the DFS test section indicate a less coherent flow pattern, which may have reduced the effect of separated flow (and therefore uplift) at the leading edge of the tiles. In contrast, the effect of turbulence was lessened at higher wind speeds which may explain slightly higher lift coefficients in those cases. 5.2. Pressure equalization Fig. 13 shows typical pressure time histories for taps on the upper and lower surface for a high-profile arrangement in experiment one, demonstrating the pressure equalization process. Equalization occurred for all configurations due to negative pressure coefficients on both the upper and lower surfaces of the tiles. In all cases, the magnitude of pressure was larger on the upper surface than the lower surface which produced a net lifting effect on the tile. Li et al. (2014) hypothesized and confirmed experimentally via full-scale testing that pressures on the underside of the tile are more uniform due to pressure equalization in the open space of the cavity. The current study confirmed that underside pressures are generally uniform except near the leading edge where positive pressure builds up (for wind perpendicular to leading edge) in the 7.6 cm (3 in.) tile overlap region (Fig. 14). A pressure equalization factor, C eq , has been previously discussed for porous cladding (e.g., Geurts, 2000) and defined as a ratio of the pressure difference (net pressure) normalized by the pressure on the outer surface of the roof. This was also discussed by Stenabaugh et al. (2015) in the context of PV panels and is computed for experiment one data by normalizing the net pressure on the tile by the pressure on the outer surface as follows: C eq ¼ P

C 0L  δ cos ðφi Þ cos ðθi Þ

C0 i i pt i

ð10Þ

lb

Fig. 13. Typical pressure time history for experiment one high-profile tile arrangement relative contributions of lower and upper surface pressures (i.e. equalization) on lift coefficient.

The results are shown in Fig. 15. In contrast to the findings of Li et al. (2014), cavity pressure coefficients were negative in most

56

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Fig. 14. Upper (left), lower (center), and net (right) C p distributions (interpolated using 256 taps) along the longitudinal dimension (cm, where tile leading edge is at ~43 cm) of a high-profile tile for wind perpendicular to the leading edge (experiment one) at approximately 20 m/s (45 mph). Note: the curve thickness in each plot represents variation in the range of C p values across the width of the tile (cross-wind direction).

Fig. 15. Pressure equalization factors for experiment one computed as the ratio of the pressure difference (net pressure) normalized by the pressure on the outer surface of the tile where H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 00, 45, 90 indicate wind angle of attack.

Fig. 16. Resolved lift coefficients for roofing tiles at three reference wind speed levels. H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 00, 45, 90 indicate wind angle of attack.

Fig. 17. Measured z-direction withdrawal force (N) at a single screw fastener (installed near the underlapping edge) for roofing tiles at three reference wind speed levels where H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 00, 45, 90 indicate wind angle of attack. Note: Data sets L/B/45 and L/D/90 are not shown due to relatively high signal to noise ratios.

Fig. 18. Total measured x–y plane shear force (N) at a single screw fastener (installed near the underlapping edge) for roofing tiles at three reference wind speed levels where H, M, L indicate high-, medium-, or low-profile tile respectfully, B, D indicate battened or direct to deck installation respectfully, and 00, 45, 90 indicate wind angle of attack. Note: Data sets L/B/45 and L/D/90 are not shown due to relatively high signal to noise ratios.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Fig. 19. Typical load cell time histories (experiment two) for z-direction withdrawal force (N) at 0, 45, and 90 degree wind angles of attack for medium-profile tile arrays installed over battens.

configurations. Equalization factors less than unity imply cavity pressure coefficients acting to restrain the tile and equalize a portion of the pressure on the upper surface of the tile. C eq values were higher in general for low-profile arrangements suggesting the smaller cavity volume and more flow restrictive leading edge (see Fig. 4) for that profile played a significant role in pressures developed in the cavity. A similar trend is shown for battened installations at 0°, suggesting the battens also play a role in flow restriction in the cavity space. The experimental configuration used by Li et al. (2014) consisted of a foam adhesive-set high-profile tile roof installed on a building model measuring 2.74  2.13 m2 in plan with 2.13 m eave height and immersed in a turbulent wind field 4.9 m high  7.3 m wide. The reduction of cavity space volume due to foam, tighter system sealing allowed by a full-scale building mockup, and sloped roof (relative to the mean flow vector), likely explain positive cavity pressures in contrast with negative cavity pressure coefficients measured within the DFS in experiment one. Values for C eq are highly dependent on pressure build up in the cavity space. Experiment one included unsealed edges around tile arrays that likely prohibited pressure build up. For sealed edge conditions (including leading and trailing edges of the arrangement), C eq values produced using the test method setup will likely be higher than those presented in Fig. 15. In particular when flow is perpendicular to the tile leading edge and impinges on the overlap between courses. It is important to note that the experiment one configuration more closely resembles the testing standard, while the Li et al. (2014) configuration is arguably more realistic. The flow characteristics for each configuration were also significantly different. Internal cavity pressures can have a dramatic effect on overall lift coefficients for design and thus should be investigated with study in full-scale flow fields in order to provide for appropriate simulation by the test method. 5.3. Lift coefficient dependence on install configuration and wind angle To examine combined effects of wind attack angle, attachment, and tile profile on lift coefficients, C L was plotted for each of the 18 specimen configurations in experiment one (Fig. 16). Each lift coefficient was computed using all 256 pressure taps. For all configurations, except medium-profile tiles attached over battens, there is strong lift coefficient dependence on wind angle of attack. Values for high-profile tiles, whether attached directly to the deck or over battens, were 50–90% greater for 45° and 90° wind angles than for a 0° wind angle. Values for all low-profile tile configurations and medium-profile tiles attached directly to the deck were

57

30–100% greater for a 0° wind angle compared to 45° and 90° wind angles. Medium-profile tiles attached over battens exhibited only slight dependence on wind angle of attack. All 45° and 90° wind angles for low- and medium-profile tiles resulted in lift coefficients less than the default value (0.2) specified by the FBC and IBC. Lift coefficients for 0° wind angles were slightly larger than the default values for medium-profile direct and low-profile battened attachments. Relatively large aerodynamic multipliers compared to product approval values were also observed for these two cases. Lift coefficients computed for high-profile configurations were greater than 0.2 for 45° and 90° wind angles. High-profile tiles subjected to 45° or 90° wind angles of attack resulted in the largest overall lift coefficient magnitude. Lift coefficient magnitudes and dependence on wind angle of attack were similar for low- and medium-profiles with the exception of lowprofile tiles attached over battens and subjected to a 0° wind angle of attack. This configuration and attack angle produced the greatest overall lift coefficient for low- and medium-profile testing scenarios. 5.4. Fastener reaction force dependence on installation configuration and wind angle Mean withdrawal (z-direction) (Fig. 17) and shear forces (x- and y-directions) (Fig. 18) at the mechanical fastener were computed from experiment two load cell data. The data indicate that withdrawal force is dependent on wind angle of attack. For high-profile configurations, there was an increase in withdrawal load with increasing wind angle, which is consistent with lift coefficient data from experiment one. Fig. 19 shows a typical set of load cell time histories for a medium-profile configuration with battens. Withdrawal forces were much greater for the 0° wind angle and consistent with experiment one data. Increase in withdrawal force during wind loading was not discernable from signal noise for two low-profile data sets (L/B/45 and L/D/90), which indicated relatively small withdrawal loads but did not allow computation of mean load increase. Low-profile data for battened attachments indicate that withdrawal loads were greater for a 0° wind angle than for a 90° wind angle (consistent with experiment one C L data) while direct attachment configurations indicate higher loads for a 45° wind angle than for a 0° wind angle (not consistent with experiment one C L data). Withdrawal loads were weakly dependent on installation configuration for both medium-, and low-profile cases. Highprofile cases indicate slightly larger withdrawal forces at the highest wind speed level for directly attached configurations. Shear forces at the mechanical fastener were computed by combining x-direction and y-direction forces using the following equation: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi V xy ¼ V 2x þ V 2y ð11Þ where V xy is the total resolved shear force measured by the load cell in the x- and y-directions, V x is the x-direction shear force measured by the load cell, and V y is the y-direction shear force measured by the load cell (see Fig. 7 for axes schematic). V xy is plotted for each installation combination in Fig. 18. The shear forces are, in general, of lower magnitude than withdrawal (z-direction) forces. The only case that suggests otherwise is the high-profile, battened installation with 45° wind angle of attack. Shear loads for this case were significantly higher than all others while withdrawal loads (Fig. 17) for this particular case were relatively low and did not increase with increasing wind speed. The irregular data for this test case in both shear and withdrawal likely indicates presence of an irregularity in the array installation.

58

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Variation in fastener seating during specimen array installation was a potential source of error. The load cell was in a fixed position on the deck and four courses of tiles in the array had to be installed before attaching the specimen tile to the load cell in the second to last course. The interlocking nature of the tile arrays required that all courses be properly aligned in order for the specimen tile (located in center of second to last course) to be positioned with its installation hole directly above the load cell arrangement. In some cases, where tile installation holes did not align perfectly with the load cell, the tile had to be slightly shifted in the array. Once the remaining tiles around the instrumented tile were installed (with interlocking), frictional forces may have developed within the array and near the load cell fastener prior to wind loading, which may have affected load path and intensity behavior during wind loading.

6. Discussion Lift coefficients and aerodynamic multipliers were computed for a 0° wind angle from experiment one data using 27 centerline locations and also using 256 taps distributed throughout the tile surfaces. In all six test configurations, the standardized method produced higher lift coefficients than the 256 tap method. The aerodynamic multiplier is an indicator of the overturning moment induced on a tile (normalized by tile surface area) and accounts for the relative locations of pressure variations on the tile surfaces. The 256 tap method produced larger multipliers than the standardized method for high-profile configurations. This suggests that the variation in surface pressures across the tile are not adequately accounted for by using centerline taps only for moment-based calculations. The standardized test method resulted in larger aerodynamic multipliers than the 256 tap method for medium-profile configurations while the two methods produced similar results for low-profile tiles. This was expected (and consistent with experiment one data) due to the flat profile of the tile, which minimizes variation in surface pressures across the tile. The standardized method requires testing for a 0° wind angle of attack only. The implied assumption is that this presents the strongest loading condition for roofing tiles, although (Tecle et al., 2013a) suggest oblique angles may be more critical. The aerodynamic multipliers suggest that 0° is the worst case scenario for low- and medium-profile tiles and 45°/90° is the worst case for high-profile tiles for the experiment one setup. Fastener withdrawal load data from experiment two support these conclusions (except for L/D/ 45). However, the ability of the testing configuration to simulate realistic conditions e.g., flow interaction at boundary conditions, pressure equalization, etc. is limited and needs to be validated with high-resolution testing in full-scale flow fields. The method also does not currently include simulation or analysis of peak pressures, which are known to modulate wind loads on roofing tiles. It is recommended that future studies in full-scale flow fields and using the wind tunnel configuration emphasize comparative analysis of peak effects to determine the extent to which the wind tunnel testing methodology can and should simulate these effects. Pressure equalization was shown to have a significant impact on net uplift, highlighting the importance of further investigation in this area. Specifically, the height of the high- and mediumprofile tiles allows air to flow below the tile arrangement at the leading edge. Li et al. (2014) found that cavity pressure coefficients (in full-scale) will be only be positive when tile joints are facing the wind (e.g., near 0 degrees when wind is perpendicular to eave) and that at other wind angles there will be negative cavity pressure coefficients, having an equalization effect. Although, cavity pressure coefficients were negative at 0° for the current experiment, the magnitude (and equalization effect) was much lower

than for other wind angles, supporting the full scale trend. It is likely that positive cavity pressure was unable to develop for the 0° degree case due to relatively unrestricted flow below the tile arrays. The fact that pressure equalization trends are similar to full scale but not raw values suggest that with modification (e.g. sealed boundary conditions, etc.), the test method may produce more realistic pressure data. Although outside the current scope, future study should also aim to develop a wind testing method that incorporates hip and ridge tiles that are frequently damaged during high-wind events. A comprehensive full-scale testing program that considers validation of the current study on field tiles (i.e. high-resolution of taps, load cells at attachments, multiple profiles and wind angles, etc.) and includes hips, ridges, and cornering zones would be a valuable contribution to the literature.

Acknowledgments This paper was written through the support of the Florida Building Commission, the Florida Department of Emergency Management, and the International Hurricane Research Center (FIU). Special thanks to Joseph Esposito and Matthew Terza for their invaluable contributions in the laboratory. The authors also thank the following groups for additional support and guidance: Tile Roofing Institute, Eagle Roofing Company and Technical Representative Manual Oyola, Boral Roofing, 3M, and the American Plywood Association. Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors and do not necessarily reflect the views of the sponsors, partners, or contributors.

References ARA, 2008. 2008 Florida Residential Wind Loss Mitigation Study. Applied Research Associates Inc., Florida Office of Insurance Regulation, Tallahassee, FL. ASCE, 2010. ASCE 7-10: Minimum Design Loads for Buildings and Other Structures. American Society of Civil Engineers, Reston, VA. ASTM, 2003a. C1568-03 Test Method for Wind Resistance of Concrete and Clay Roof Tiles (Mechanical Uplift Resistance Method). C15 Committee. ASTM International. ASTM, 2003b. C1569-03 Test Method for Wind Resistance of Concrete and Clay Roof Tiles (Wind Tunnel Method). C15 Committee. ASTM International. ASTM, 2003c. C1570-03 Test Method for Wind Resistance of Concrete and Clay Roof Tiles (Air Permeability Method). C15 Committee. ASTM International. British Standards Institution, 2015. BS 5534:2014þ A1:2015 Slating and Tiling for Pitched Roofs and Vertical Cladding – Code of Practice BSI Standards Limited. CEN, 2004. EN 14437: Determination of the Uplift Resistance of Installed Clay or Concrete Tiles for Roofing - Roof System Test Method. European Committee for Standardization B-1050, Brussels, p. 34. FBC, 2014a. Testing Application Standard No. 101-95 Test Procedure for Static Uplift Resistance of Mortar or Adhesive Set Tile Systems. 2014 Florida Building Code. FBC, 2014b. Testing Application Standard No. 102-95 Test Procedure for Static Uplift Resistance of Mechanically Attached, Rigid Roof Systems. 2014 Florida Building Code. FBC, 2014c. Testing Application Standard No. 102A-95 Test Procedure for Static Uplift Resistance of Mechanically Attached, Clipped, Rigid, Roof Systems. 2014 Florida Building Code. FBC, 2014d. Testing Application Standard No. 108-95 Test Procedure for Wind Tennel Testing of Air Permeable, Rigid, Discontinuous Roof Systems. 2014 Florida Building Code. FBC, 2014e. 2014 Florida Building Code. Florida Building Commission, International Code Council, Inc.. FBC, 2014f. Roofing Application Standard No. 127 Procedure for Determining the Moment of Resistance and Minimum Characteristic Load to Install a Tile System on a Building of a Specified Roof Slope and Height. 2014 Florida Building Code. FEMA, 2005. Hurricane Charley in Florida: Observations, Recommendations, and Technical Guidance. Mitigation Assessment Team Report, Federal Emergency Management Agency, Mitigation Assessment Team, U.S. Department of Homeland Security, Washington, D.C.. FPHLM. Survey of Single Family Residential Buildings in Florida. Florida Public Loss Model: Engineering Team Report, 2011.

D.J. Smith et al. / J. Wind Eng. Ind. Aerodyn. 155 (2016) 47–59

Geurts, C.P.W., 2000. Wind loads on permeable roof covering products. In: Proceedings of the Fourth Colloquium on Bluff Body Aerodynamics and Applications, pp. 511–514. Hazelwood, R.A., 1980. Principles of wind loading on tiled roofs and their application in the British standard BS5534. J. Wind. Eng. Ind. Aerodyn. 6 (1–2), 113–124. Hazelwood, R.A., 1981. The interaction of the two principal wind forces on roof tiles. J. Wind. Eng. Ind. Aerodyn. 8 (1–2), 39–48. Kramer, C., Gerhardt, H.J., 1983. Wind loads on permeable roofing systems. J. Wind. Eng. Ind. Aerodyn. 13 (1), 347–358. Kramer, C., Gerhardt, H.J., Kuster, H.W., 1979. On the wind-loading mechanism of roofing elements. J. Wind. Eng. Ind. Aerodyn. 4 (3–4), 415–427. Laboy-Rodriguez, S.T., Smith, D., Gurley, K.R., Masters, F.J., 2013. Roof tile frangibility and puncture of metal window shutters. Wind Struct. 17 (2), 185–202. Li, R., Chowdury, A., Bitsuamlak, G., Gurley, K., 2014. Wind effects on roofs with high-profile tiles: experimental study. J. Architect. Eng. Hous. Resid. Build. Constr. National Science Board, 2007. Hurricane Warning: The Critical Need for a National Hurricane Research Initiative. National Science Board. Netherlands Standardization Institute, 2011. NEN 6707: Fixing of Roof Coverings – Requirements and Determination Methods, Delft, Netherlands. Pielke Jr., R.A., Landsea, C.W., 1998. Normalized hurricane damages in the United States: 1925–95. Weather. Forecast. 13 (3), 621–631.

59

Redland Technology, 1991. Fixing Studies for MRTI Normal Weight Tiles – SBCCI Submission. Redland Technology. Robertson, A.P., Hoxey, R.P., Rideout, N.M., Freathy, P., 2007. Full-scale study of wind loads on roof tiles and felt underlay and comparisons with design data. Wind Struct. 10 (6), 495–510. SBCCI, 1999. SSTD 11-99: Test Standard for Determining Wind Resistance of Concrete or Clay Roof Tiles. Southern Building Code Congress International. Smith, D.J., Masters, F.J., Gurley, K.R., 2014. A historical perspective on the wind resistance of concrete and clay roofing tiles. RCI Interface 32 (10). Standards Australia, 2002. AS 2050: Installation of Roof Tiles SAI Global Limited, Sydney, Australia. Stenabaugh, S.E., Iida, Y., Kopp, G.A., Karava, P., 2015. Wind loads on photovoltaic arrays mounted parallel to sloped roofs on low-rise buildings. J Wind Eng. Ind. Aerodyn. 139, 16–26. Tecle, A., Bitsuamlak, G.T., Chowdhury, A.G., 2013a. Wind load on ridge and field tiles on a residential building: a full scale study. Adv. Hurric. Eng.—Learn. Our, 506–516. Tecle, A., Bitsuamlak, G.T., Suskawang, N., Chowdhury, A.G., Fuez, S., 2013b. Ridge and field tile aerodynamics for a low-rise building: a full-scale study. Wind Struct. 16 (4), 301–322.