Performance of hurricane shutters under impact by roof tiles

Engineering Structures 32 (2010) 3384–3393

Contents lists available at ScienceDirect

Engineering Structures

journal homepage: www.elsevier.com/locate/engstruct

George Fernandez, Forrest J. Masters

, Kurtis R. Gurley

∗

Department of Civil and Coastal Engineering, University of Florida, 365 Weil Hall, Gainesville, FL 32611, United States

a r t i c l e i n f o

Article history:

Received 6 April 2010

Received in revised form

4 July 2010

Accepted 5 July 2010

Available online 11 August 2010

Keywords:

Windborne debris

Missile impact

Roof tile

Shutters

Storm panels

Hurricane

Window protection

a b s t r a c t

This paper presents an experimental investigation of the performance of shutter systems designed to

protect windows from windborne debris. Observations from post hurricane damage investigations have

found that a wide variety of windborne debris types cause damage to buildings, including roof tiles

in residential neighborhoods. This investigation subjected steel and aluminum storm panel shutters to

impact from concrete roof tiles commonly used in hurricane prone regions. 4.1 kg (9 lb) tiles were

launched at 15.25 m/s (50 fps) using a custom apparatus, duplicating the 2× 4 (in.) lumber impact product

certification test in both missile weight and impact speed. The tests were then repeated using 2×4 lumber,

providing a comparison of performance as a function of debris type. The test matrix included steel panels

of three different thicknesses and aluminum panels of two different thicknesses. Three manufacturers

of each of the five storm panel types were tested, each using two common installation methods. Tests

were conducted with tiles impacting on their edge and impacting flat. Results demonstrate a significant

difference in both total and plastic shutter deflection for tile impacts vs. 2 × 4 lumber. With regard to

the vulnerability of the glass being protected, the results suggest that the current standards may not be

conservative under circumstances likely to occur in tile roof residential neighborhoods.

1. Introduction

Windborne debris can cause significant damage to building

envelopes in high wind events. Post storm investigation studies

have noted a high degree of roof cover loss, indicating that roof

cover is a primary source of potentially damaging windborne

debris in residential areas (e.g. [1–8]). Building codes along much

of the world’s hurricane prone coastline have evolved to address

debris impact on the building envelope, requiring the testing of

window protection systems, fenestration, wall and roof claddings

in order to certify a minimum impact resistance standard.

Although it was recognized by the developers that roof cover

material is a primary debris source, US and Australian standards

(e.g. [9–13]) specify testing with 2×4 (in.) lumber and/or steel balls

as the debris for the purpose of maintaining a repeatable procedure

with controlled representative impact momentum [14]. Due to

the limited relationship between the debris used in certification

tests and that most commonly observed after hurricane events,

the actual level of protection being evaluated by the existing

certification methods is not well defined.

The design-level event Hurricane Charley (2004) produced

statistical evidence that the use of window protection systems is an

∗

Corresponding author. Tel.: +1 352 392 9537x1508; fax: +1 352 392 3394.

E-mail addresses: [email protected] (F.J. Masters), [email protected]

(K.R. Gurley).

Tel.: +1 352 392 9537x1505.

effective mitigation, reducing the probability of window damage

by at least a factor of 2.5 relative to unprotected windows [1].

However, evidence collected during this same study demonstrated

that some code approved opening protection systems experienced

failures due to debris impact from roofing tiles. Tile is a commonly

used roof cover in many hurricane prone regions in the US,

Australia and elsewhere. For example, the proportion of residential

roofs with tile exceeds 20% in some Florida counties [15].

Methods to model probable losses and to determine the cost

effectiveness of various mitigation measures can benefit from

specific information about the frequency and severity of damage

to the building envelope, with the latter being the focus of this

study. For example, risk models that project losses from hurricane

winds address the vulnerability of both protected and unprotected

glazed openings to windborne debris [16,17]. Recent advances in

debris flight and impact probability modeling (e.g. [18–29]) can

be enhanced with test data quantifying the vulnerability of the

building envelope to debris typical of the hurricane environment.

The body of knowledge concerning the vulnerability of fenes-

tration to impact includes numerous studies on the impact resis-

tance of annealed, tempered, and laminated glass [30–43], includ-

ing a recent study presenting the momentum threshold required

to damage unprotected annealed residential window glazing when

impacted by asphalt roof shingles and wooden dowels [44]. The

current study continues this focus on physical testing of the build-

ing envelope with realistic debris by documenting the performance

of metal shutter systems impacted by concrete roof tiles. The study

also provides results from impact testing using 2 × 4 lumber as a

doi:10.1016/j.engstruct.2010.07.012

G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393 3385

benchmark representing current certification methods. This origi-

nal contribution to the body of knowledge presents a clear contrast

between building envelope protection performance as evaluated

via (a) current test standards, and (b) experiments that better rep-

resent field conditions with regard to debris type.

All tests were conducted at a speed of 15.25 m/s to provide

a direct comparison to the speed currently used in certification

standards. At an elevation of 10 m, a 33.1 m/s sustained (one

minute average) marine exposure wind speed defines the lower

bound of a Saffir–Simpson Hurricane Wind Scale Category 1

event [45,46]. In overland open exposure conditions the expected

three second gust is approximately ten percent higher than the

sustained marine winds [47], or 36.4 m/s. The speed of 15.25 m/s,

used for certification standards and for all tests in this study, is 42%

of the 36.4 m/s three second gust of a minimal category 1 hurricane,

and less than 33% of the gusts associated with a minimal category

2 hurricane. Recent studies (e.g. [26–29]) suggest that 0.33 is less

than the median for the ratio of debris to gust wind speed, and

post storm damage studies did not commonly find roof tile debris

in regions that experienced less than category 2 winds. Therefore,

winds strong enough to generate roof tile debris are likely to impart

a speed to that debris that is higher than the 15.25 m/s used for the

current standard and duplicated in this study. The 15.25 m/s test

speed represents a lower bound of speeds likely to occur in real

conditions.

2. Description of impact test methods and test matrix

This investigation subjected steel and aluminum panel shutters

to impact from concrete roof tiles. 4.1 kg (9 lb) tiles were launched

at 15.25 m/s (50 fps) using a custom apparatus, duplicating the 2×4

lumber impact test used for product certification in both missile

weight and impact speed.

Permanent (plastic) and total (plastic and elastic) deflections

were recorded using a high speed camera. These tests were

then repeated using 2 × 4 lumber, providing a comparison of

performance as a function of debris type. The test matrix included

steel panels of three different thicknesses and aluminum panels of

two different thicknesses. Products from three manufacturers of

each of the five shutter types were tested, each using two common

installation methods. Tests were conducted with tiles impacting on

edge and impacting flat. Each tile impact test was conducted twice.

A total of 180 tile impact and 60 2× 4 impact tests were conducted.

New specimens were used for every test.

Apparatuses were constructed to launch roof tile debris with

controlled accuracy, flight mode and speed, and to launch 4.1 kg

2 × 4 lumber in a manner consistent with current debris impact

certification standards. Both apparatuses are based upon the same

air cannon platform. All testing was performed at the Powell

Family Structures and Materials Laboratory located on the Eastside

Campus of the University of Florida.

2.1. Launching apparatus

The tile launching apparatus is comprised of four major

components; the pneumatic ram that propels the tile projectile

along a guided track toward the target, the air reservoir tank

and barrel that supply the propulsion force to the ram, the

electronic butterfly valve that releases the pressure from the tank

to the barrel, and the integrated electronic system which monitors

projectile speed and maintains the desired tank pressure. Fig. 1

presents an illustration of the launching apparatus.

The air reservoir is coupled to a 10.16 cm diameter schedule 40

steel pipe connected to a 120 V AC electronic valve with a 7.62 cm

diameter inlet and outlet. The ram fits within the barrel and

harnesses the released pressure. The far end of the ram is located

at the exit end of the barrel, and bears against the tile projectile. A

launch deck (guided track) supported by extruded T-slot aluminum

rails extends from the exit of the barrel towards the target. The tile

projectile sits upon this deck and is propelled toward the target

via the ram. The deck is lined with a polyoxymethylene sheet to

reduce sliding friction. The ram forward motion is arrested as the

tile and ram approach the end of the deck, putting the tile in free

flight towards the target.

The electronic system includes feedback control for filling,

purging, monitoring and maintaining air reservoir pressure, con-

trol of the electronic valve for projectile launch, and measure-

ment of the projectile speed using two infrared beams. This

system is controlled from a custom National Instruments LabVIEW

application.

A wood frame bolted to the strong floor supports the storm

panels (impact targets) during testing. The frame is constructed of

4 × 4 (in.) timber and is 1.70 m high and 1.52 m wide. The frame

is sheathed with 1.9 cm thick plywood to act as both a diaphragm

for stability and a barrier to stop any debris that penetrates the

storm panels. All joints of the frame were reinforced by metal

brackets. The inner dimensions of the frame opening are 160.02 cm

(63 in.) wide by 156.84 cm (61.75 in.) tall. The span of the shutter

system varied due to different keyhole spacing and the sheet metal

stamping process; however the number of panels and the number

of studs used to fasten all systems remained constant.

The 2×4 lumber testing was conducted after the tile testing was

completed. The same apparatus described above was modified to

accommodate 2 × 4 lumber. The launch deck and pneumatic ram

were removed, resulting in a traditional air cannon lumber launch

configuration, with the lumber missile taking the place of the push

ram within the barrel.

2.2. Tile projectile

A nominal 4.1 kg (9 lb) common concrete S-shape roofing tile

was used as the projectile (Fig. 2). Each test used a new undamaged

tile. Every test was conducted at a launch speed of 15.25 m/s (50

fps). Thus, every tile test has a nominally equivalent momentum

to a standard 2 × 4 impact test. The speed of the tile launch was

controlled via air reservoir pressure and a series of calibrations

using infrared sensors and a high speed camera to document

launch speed.

2.3. Impact orientation of tile projectile

The flight mode of the tile, and therefore its orientation upon

target impact, can be controlled by a combination of launch deck

length and free flight distance to the target. The two impact

orientations considered were edge impact and flat impact. For edge

impact, the plane of the tile is perpendicular to the plane of the

target, the normal to the tile plane is vertical, and the short edge

of the tile impacts the target. For flat impact the tile and target

are in parallel planes upon impact, with the long edge of the tile

vertical. These repeatable flight modes are imparted by controlling

the length of the launch deck. When the pneumatic ram motion is

arrested prior to the tile exiting the deck, the tile travels toward

the target normal to the tile plane vertical (edge impact). When

the launch deck is shortened such that the front end of the tile

exits the launch deck prior to arresting the push ram motion, a

forward rotation is imparted to the tile. A calibration of the free

flight distance to the target allows this rotation to result in a flat

impact of the tile with the target. The free flight distance for the

tile was 3.92 m for edge impact and 4.39 m for flat impact.

2.4. Impact location of tile and 2 × 4 projectiles

The shutter systems were impacted in one of two locations, in

the center of the shutter assembly (opening mid-span) and in the

3386 G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393

Computer

Valve

Barrel

Tank

Pressure Relief

Ram

Tile

Speed Monitor

2×4 Configuration

Guided Track/

Launch Deck

Speed Monitor

Tile Configuration

Storm Shutter Pannel

Target Frame

Fig. 1. Testing apparatus illustration.

Fig. 2. Concrete tile used as windborne debris.

upper corner within 15.24 cm (6 in.) of either edge. Both flat and

edge tile orientations were used for center impacts. Only edge tile

orientation was used for corner impacts to avoid striking the frame.

2 × 4 testing was conducted at both center and corner locations.

Each shutter sample was only impacted once.

2.5. Data recorded from impact tests

Multiple variables were recorded for each impact test, including

shutter and mounting type. The speed of the projectile was

recorded. Photographs of the test specimen before and after impact

were taken. The post-impact condition of the installation was

recorded (tearing, pulling out of track, etc.). Missile impact location

and orientation (for tile) was recorded. The maximum permanent

(plastic) deflection of the panel was measured and recorded. A high

speed camera was used to view the backside plane of the shutter,

and recorded the total deflection (elastic plus plastic). Thus, the

potential damage to glass behind the shutter was determined for

every impact test.

2.6. Velocity sensor

The velocity sensor system consists of photoelectric diodes and

a National Instruments USB data acquisition unit. The photoelectric

diodes have a response time that exceeds the 0.15 m/s required

by ASTM E 1886-05 [9]. The diodes are mounted at the end of the

guided track for the tile launch configuration, and near the barrel

for the 2 × 4 launch configuration. In both cases, the system is

triggered after the pressure cap on the pneumatic ram or 2 × 4

has passed the pressure relief ports at the end of the cannon barrel

(see Fig. 1). This ensures that the speed is measured after the ram or

2×4 has stopped accelerating and has reached its peak velocity. For

both configurations the sensor system was calibrated using both a

high speed camera and radar gun monitoring the launch speed of

the projectile. Thus, the impact speeds were carefully controlled

and independently validated.

2.7. Shutter type and mounting

Steel shutters of three thicknesses (20, 22, and 24 gauge) and

aluminum shutters of two thicknesses (0.050 and 0.060 in., or

0.127 and 0.152 cm) were tested. For each of these five shutter

types, products from three different manufacturers were tested.

Shutters were purchased from seven different manufacturers, and

all products are approved for use in Florida. Panels measured

between 33.02 and 37.78 cm wide and each was 167.64 cm

(66 in.) tall. Four overlapping panels were installed in a vertical

orientation, bearing against the frame at the top and bottom with

no bearing along the vertical edges.

Each of the five shutter types were tested using both direct

mount and track mount installation methods. In the direct mount

configurations, the wood frame has threaded studs installed at a

spacing of 15.24 cm (6 in.) on center or 15.87 cm (6.25 in.) on

center across the top and bottom horizontal framing boarders. The

studs pass through the panels and the panel is fastened by a 1/4-

20 washered wing nut on each stud (Fig. 3). For track mounting

a 5.1 cm (2 in.) wide h-header track guide is installed along the

top border of the frame. The panels are slid into position with

the track along the top restraining motion and studs and wing

nuts along the bottom similar to the direct mount (Fig. 4). All

installation hardware was purchased from an independent third

party vendor, including the ‘‘h’’ headers and studded angles for the

top and bottom installations of the track mount.

3. Results and discussion

3.1. Results presentation

All impact data are presented in a series of six graphs

(Figs. 6–11). Each graph presents either edge tile impact and 2 × 4

impact, or edge tile impact and flat tile impact, represented by wide

and thin bars, respectively. Each graph presents the results for each

of the three manufacturers for every product tested (three steel

thicknesses, two aluminum thicknesses). All tile and 2 × 4 impacts

were performed using a nominal 4.1 kg (9 lb) object traveling at

G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393 3387

Fig. 3. Direct mount installation.

Fig. 4. Track mount installation using h-header along the top and stud mount along

the bottom.

15.25 m/s (50 fps). Every tile test was conducted twice for every

condition, and each graph presents the results of both tests. Every

2 × 4 test was conducted once. In the four graphs that present

2 × 4 impact tests (Figs. 8–11), the 2 × 4 results are repeated

for direct comparison to both tile tests. The horizontal axis in

each graph indicates the anonymous manufacturer designator

(A-G), product type, and test number using the key provided in

Fig. 5. Fig. 5 also illustrates the proper interpretation of the plastic

(permanent), elastic (recovered) and total (plastic plus elastic)

deflections presented in Figs. 6–11.

In many tests, damage occurred at the installation (the interface

between the panel and frame). For direct mount installation,

damage is defined as the stud tearing through the panel at one

stud or more. For track mount installation, damage is defined as

either a pull-out or push-through of the panel from the track, such

that the top edge of the panel is partially or fully separated from

the track. Examples of tearing for direct mount and pull-out and

push-through for track mount installation are shown in Fig. 12.

Installation interface damage occurred for a small number of 2 × 4

tests, and a larger portion of the tile tests. In most cases, damage

at the interface is associated with a more severe panel deflection.

Each of the individual tests that resulted in installation damage

is indicated in the six graphs (Figs. 6–11) with an icon above the

deflection results for a given test. These icons are defined in Fig. 5.

3.2. Discussion: tile edge vs. tile flat impact (Figs. 6 and 7)

Fig. 6 provides the results of the flat tile and edge tile impacting

the center of the direct mounted panel system, and Fig. 7 presents

the results of the flat tile and edge tile impacting the center of the

track mounted panel system. Flat tile impacts were not conducted

for corner shots. These two figures contrast the deflection caused

by a tile impacting on edge (wide bars) and flat (thin bars). In

Fig. 6 the missing thin bar for the second 24 gauge steel product

indicates that the high speed camera did not function for that test.

Notable observations from Figs. 6 and 7 are provided below. Refer

to the accompanying figures for specific values that support these

observations:

• For any given product, the two repeated tests yield very

similar results in most cases, while results among different

manufacturers of the same product type vary widely. For

example, the 20 gauge steel product (first 6 columns in Fig. 6)

shows strong repeatability of results for a given manufacturer,

while the results among manufacturers varied.

• The total deflection from edge tile impact is larger than that

from flat tile impact in most cases.

• In every case, the edge tile impact total deflection exceeded

7.62 cm (3 in.). This is significant, as the tested products

required a minimum distance of 7.62 cm between product

and window glass for installation based on the results of the

certification testing.

The remaining four results’ figures each compare deflection from

2 × 4 and edge tile impacts. Figs. 6 and 7 provide a means to

compare 2 × 4 with flat tile impact results.

3.3. Discussion: direct mount impact with edge tile and 2 × 4 (Figs. 8

and 9)

Fig. 8 presents the results of the 2 × 4 (thin bars) and edge

tile (wide bars) impacting the corner of the direct mounted panel

system, and Fig. 9 presents the results of the 2 × 4 and edge tile

impacting the center of the direct mounted panel system. Notable

observations are provided below. Refer to the accompanying

figures for specific values that support these observations:

• Deflections are larger in almost all cases for the center shots

compared to corner shots.

• As was the case for Figs. 6 and 7, the two repeated tile tests

yield very similar results in most cases, with exceptions for the

22 gauge steel corner impacts (attributable to pull-out of the

shutter from its track in the second test).

• In most cases the edge tile impact yielded considerably larger

deflection than the 2 × 4 impact.

• Plastic deflection from the 2 × 4 tests did not exceed the critical

setback value of 7.62 cm for any test in Figs. 8 and 9, and in only

a few cases the total deflection from 2 × 4 impact exceeded that

value. This seems to conform to the setback recommendations

based on plastic deflection from certification tests.

3388 G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393

Fig. 5. Guide for Fig. 6 through 11: horizontal axis labels (top left), icons indicating installation damage (bottom left), and a guide to interpreting deflection results (right).

Fig. 6. Center impact, direct mount, tile edge and tile flat.

• In contrast, plastic deflection from edge tile impact in the corner

(Fig. 8) exceeded 7.62 cm for the thinnest steel and aluminum

products, while total deflection exceeded this value in more

than half of the corner tests.

• Tile impacts at the center of the direct mount panels (Fig. 9)

produced plastic deflections in excess of 7.62 cm for half of the

tests, while total deflection exceeded this value in almost every

test.

3.4. Discussion: track mount impact with edge tile and 2 × 4 (Figs. 10

and 11)

Fig. 10 presents the results for the 2 × 4 and edge tile impacting

the corner of the track mounted panel system, and Fig. 11 presents

the results for the 2 × 4 and edge tile impacting the center of

the track mounted panel system. Tile impacts are the wide bars,

and 2 × 4 impacts are the thin bars. For the track mounted

system, it can be observed that many corner shots yielded more

deflection than the center shots. This is attributable to the pull-

out of the shutter from the upper track mount, which was more

common for corner shots than center shots, and permitted more

deflection. Other notable observations are provided below. Refer

to the accompanying figures for specific values that support these

observations:

• The two repeated tests typically yield similar results in most

cases, with a few exceptions.

• In most cases the edge tile impact yielded considerably larger

deflection than the 2 × 4 impact.

• Plastic and total deflection from the 2×4 tests exceeded 7.62 cm

for a minority of tests.

• In contrast, plastic deflection from edge tile impact in the corner

(Fig. 10) exceeded 7.62 cm for the thinnest steel and aluminum

products, while total deflection exceeded this value in more

than half of the corner tests.

G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393 3389

Fig. 7. Center impact, track mount, tile edge and tile flat.

Fig. 8. Corner impact, direct mount, tile edge and 2 × 4.

• Tile impacts at the center of the track mount panels (Fig. 11)

produced plastic deflections in excess of 7.62 cm for more than

half of the tests, while total deflection exceeded this value in

every test.

3.5. Discussion: track and direct mount installations

Typically the deflections in the track mount installation (Figs. 10

and 11) were greater than the deflection in the direct mount

installation (Figs. 8 and 9). This difference is minor for 2 × 4

impacts, but much more pronounced for tile impacts. This is mainly

attributable to the much higher rate of occurrence of installation

damage for tile impacts. Installation damage is strongly associated

with greater deflection, as observed by the icons above the

individual tests. Ten percent of the tile tests produced installation

damage to the direct mount systems for both corner and center

impacts, while the tile tests damaged 73% and 33% of the track

mount installations from corner and center impacts, respectively.

3.6. Average statistics: ratio of tile to 2 × 4 deflection and total to

plastic deflection

The results in Figs. 8–11 were analyzed to provide some

statistical quantification of shutter deflection. Table 1 presents the

ratio of total deflection from edge tile impact to total deflection

from 2×4 impact, stratified by product type, mounting, and impact

location. The total deflection from each edge tile test is taken in

ratio with that of the comparable 2 × 4 test, and these ratios are

then averaged over the given stratification. Each value in the table

is therefore an average from six tile tests (three manufacturers,

two tests each). As an example to aid in interpretation, the direct

mount corner impact ratio value of 11.32 for 22 ga steel in

3390 G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393

Fig. 9. Center impact, direct mount, tile edge and 2 × 4.

Fig. 10. Corner impact, track mount, tile edge and 2 × 4.

Table 1 can be seen graphically in Fig. 8, columns 7 through 12, by

comparing the wide bars to the thin bars. In particular the product

by manufacturers C & D (columns 9 through 12) shows a very large

ratio due to small deflections from the 2 × 4 tests. In all cases in

Table 1 the average ratio exceeds 1.0. The deflection from edge tile

impact is more severe than deflection from a 2 × 4 of identical

weight and speed, regardless of impact location, mounting type,

or product material or thickness. Referring to Figs. 8–11, only 7.5%

of the individual test cases produced a 2×4 induced deflection that

exceeded the tile induced deflection.

The dark and light stacked bars in Figs. 6–11 visually

demonstrate that for any given test the ratio of total to plastic

deflection varies with test conditions and product. Interpretation

of the ASTM test [10] is based upon measured plastic deflection and

some assumptions about the total deflection (recording the total

deflection is not required in the test standard). Table 2 presents

Table 1

Average ratio of total deflection from edge tile to 2 × 4 impacts.

Mount Location 20 ga

steel

22 ga

steel

24 ga

steel

.06

alum

.05 alum

Direct

Corner 1.45 11.32 1.45 5.66 5.15

Center 3.95 3.58 5.10 1.88 2.88

Track

Corner 2.53 1.80 2.74 3.68 3.22

Center 3.75 4.18 5.98 2.53 1.72

the ratio of the total deflection to plastic deflection, stratified by

debris type, product type, mounting, and impact location. The ratio

of total to plastic deflection for each impact test is calculated, and

these values are then averaged over the given stratification. Each

value in the table is the average from either six or three tests for

tile or 2 × 4 results, respectively. For the 2 × 4 tests the average

ratio ranges from 1.28 to 5.22. The average of the 2 × 4 ratios for

G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393 3391

Fig. 11. Center impact, track mount, tile edge and 2 × 4.

Fig. 12. Examples of direct mount tear (left) and track mount pull-out (right).

Table 2

Average ratio of total to plastic deflection for edge tile and 2 × 4 impacts.

Mount Location 20 ga steel 22 ga steel 24 ga steel .06 alum .05 alum

Tile 2 × 4 Tile 2 × 4 Tile 2 × 4 Tile 2 × 4 Tile 2×4

Direct

Corner 2.34 1.96 1.46 2.43 1.57 3.14 1.56 5.22 1.43 2.32

Center 2.63 1.28 2.16 1.43 1.80 1.34 1.94 1.59 1.79 1.81

Track

Corner 2.36 3.23 1.33 1.79 1.28 1.56 1.24 2.09 1.11 2.06

Center 2.29 3.34 1.90 2.88 1.20 1.92 1.63 1.67 1.60 1.57

direct mount is 2.25, and the average of the 2 × 4 ratios for track

mount is 2.21. For the edge tile tests the average ratio ranges from

1.11 to 2.63. The average of the edge tile ratios for direct mount is

1.87, and the average of the edge tile ratios for track mount is 1.59.

The ratios reported in Table 2 are not consistently higher for

one debris type, mounting, or impact location, but it is clear

that the elastic portion of the deflection is significant. For the

products tested, the total deflection can exceed the permanent

deflection by a factor of more than two. With regard to the existing

2 × 4 standard, setback recommendations based on permanent

deflection measurements should account for the elastic portion

of the deflection to prevent glass breakage. The measurement of

elastic deflection may be cost prohibitive for test certification labs.

These results provide some guidance for the development of a

multiplication factor for the permanent deflection.

3.7. Significance of results

The focus of this study is on the performance of certified

window protection systems when impacted by roof tiles at a mo-

mentum value that corresponds to the current standards that use

lumber as the debris. The results clearly demonstrate a statistically

significant difference in deflection as a function of debris types

with identical mass, speed and impact location. As a complement

to these findings, recent studies (e.g. [26–29]) present evidence

that roof cover debris speeds are very likely to exceed 15.25 m/s.

3392 G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393

The results in this study indicate that the current 2 × 4 lumber

based test standards do not offer a performance evaluation that is

conservative with regard to the protection of windows from debris

commonly observed in post hurricane damage studies. This is by no

means a rejection of the value of using window protection. Studies

clearly demonstrate the effectiveness of window protection [1],

and even damaged shutters can continue to provide resistance to

envelope failure. Much of the debris documented in post storm

studies are lighter and more flexible than a 4.1 kg roof tile, such

as roof shingles, and much less likely to cause severe deflections.

However, the study presents evidence that questions the efficacy

of applying the current debris impact test standard to evaluate

shutter systems intended for use in neighborhoods where tile roof

cover is dominant.

Field studies have documented that tile debris is often released

from rooftops in fragments, which presents an additional mode

of failure for window protection via puncture. Puncture failures

were not observed in the current study using full tiles. A follow-

up experimental study will quantify momentum thresholds for

puncture of window protection from tile fragments.

4. Conclusions

This paper presents an experimental investigation of the

performance of shutter systems designed to protect windows

from windborne debris. Steel and aluminum storm panel shutters

were subjected to impact from concrete roof tiles and 2 × 4

lumber. The results of this study indicate that there is a significant

difference in the plastic and total deflection of the tested panels

when impacted by roof tiles and 2 × 4 lumber of identical weight

and speed. The major implication is that impact momentum

alone is not a sufficient metric on which to base performance

criteria. The deflection of the metal panel window protection

system is highly sensitive to impact location (which is currently

addressed in product approval testing), and also to debris type

and impact orientation. In most tests, the deflection imparted

by a tile impacting on its edge exceeded the specified setback

from the glass, while only a few such cases were found for the

2 × 4 tests. Although the tested products did conform to the

performance requirements for product certification, it is evident

that the current product testing using a 2 × 4 missile does

not provide an adequate evaluation of the expected performance

of shutter systems subjected to tile debris. The tested window

protection products are likely to allow glass breakage if impacted

by roof tiles. This is by no means a rejection of the value of

using window protection. Rather, the study presents evidence

that questions the efficacy of applying the current debris impact

test standard to evaluate shutter systems intended for use in

neighborhoods where tile roof cover is dominant.

Acknowledgements

The authors thank the Florida Building Commission for

sponsoring this research. The contributions of Mr. Jim Austin and

Mr. Scott Bolton are gratefully acknowledged. Special thanks to

Jaime D. Gascon and Helmy Makar at the Miami-Dade Building

Code Compliance Office for their review of the manuscript prior

to submission.

References

[1] Gurley K, Davis R, Ferrera S-P, Burton J, Masters F, Reinhold T, Abdullah M.

Post 2004 hurricane field survey — an evaluation of the relative performance

of the standard building code and the florida building code. ASCE structures

congress. St. Louis; 2006.

[2] Federal emergency management agency (FEMA). Hurricane Andrew in

florida: building performance observations, recommendations, and technical

guidance Rep. No. FEMA P-22. Washington (DC): 1993.

[3] Federal emergency management agency (FEMA). Mitigation assessment team

report: Hurricane Charley in Florida Rep. No. FEMA 488, 5.1–5.68. Washington

(DC): 2005.

[4] Federal emergency management agency (FEMA). Mitigation assessment team

report: Hurricane Ivan in Alabama and Florida Rep. No. FEMA 489, 5.1–5.65.

Washington (DC): 2005.

[5] Meloy N, Sen R, Pai N, Mullins G. Roof damage in new homes caused by

Hurricane Charley. J Perform Constr Facil 2007;21(2):97–107.

[6] Beason WL, Meyers GE, James RW. Hurricane related window glass damage in

Houston. J Struct Eng 1984;110(12):2843–57.

[7] Oliver C, Hanson C. Failure of residential building envelopes as a result

of hurricane Andrew in dade county. In: Cook RA, Soltani M, (editors).

Florida Hurricanes of 1992: lessons learned and implications for the future.

Proceedings of an ASCE symposium. 1994. p. 496–508.

[8] Ayscue JK. Hurricane damage to residential structures: risk and mitigation

1996. Retrieved October 20, 2008 from natural hazards research and

applications information center

http://www.colorado.edu/hazards/publications/wp/wp94/wp94.html.

[9] ASTM E 1886–05. Standard test method for performance of exterior windows,

curtain walls, doors, and storm shutters impacted by Missile(s) and exposed

to cyclic pressure differentials. American society for testing and materials, 100

Barr Harbor Drive, PO Box C700. West Conshohocken (PA): 2005. p. 19428.

[10] ASTM E 1996-09. Performance of exterior windows, curtain walls, doors,

and impact protective systems impacted by windborne debris in Hurricanes.

American society for testing and materials, 100 Barr Harbor Drive, PO Box

C700. West Conshohocken (PA): 2005. p. 19428.

[11] TAS 201-94. Impact test procedures. Florida building code test protocols

for high-velocity hurricane zones, department of community affairs building

codes and standards, 2555 Shumard Oak Boulevard. Tallahassee (FL): 1994.

p. 32399.

[12] AAMA 506-05. Voluntary specification for hurricane impact and cyclic testing

of fenestration products. American architectural manufacture association,

1827 Walden office square, Suite 550. Schaumburg (IL): 2005. p. 60173-4268.

[13] Australian/New Zealand standard on wind actions AS/NZS1170.2.

[14] Minor J. Windborne debris and the building envelope. J Wind Eng Ind Aerodyn

1994;53(1–2):207–27.

[15] Gurley K, Pinelli JP, Subramanian C, Torkian BB. Survey of single family

residential buildings in Florida, and development of new shapes. A research

report on the development of the Florida Public Hurricane Loss Model,

International Hurricane research center; 2009.

[16] Vickery PJ, Skerlj PF, Lin J, Twisdale Jr LA, Young MA, Lavelle FM. HAZUS-MH

Hurricane model methodology. ii: Damage and loss estimation. Nat Hazards

Rev 2006;7(2):94–103.

[17] Pita GL, Pinelli J-P, Gurley K, Weekes J, Subramanian CS, Hamid S. Vulnerability

of mid-high rise commercial–residential buildings in the Florida Public

Hurricane Loss Model. In: Proceedings European safety and reliability

conference ESREL. 2009.

[18] Twisdale LA, Vickery PJ, Steckley AC. Analysis of hurricane windborne debris

risk for residential structures. Raleigh (NC): Applied Research Associates Inc.;

1996.

[19] HAZUS-MH technical manual. Washington (DC): Federal Emergency Manage-

ment Agency (FEMA); 2003.

[20] Lin N, Vanmarcke E. Windborne debris risk assessment. Probab Eng Mech

2008;23(4):523–30.

[21] Wills JAB, Lee BE, Wyatt TA. A model of windborne debris damage. J Wind Eng

Ind Aerodyn 2002;90(4–5):555–65.

[22] Holmes JD. Wind loading of structures. New York (NY): Spon Press; 2002.

[23] Holmes JD. Trajectories of spheres in strong winds with application to wind-

borne debris. J Wind Eng Ind Aerodyn 2004;92(1):9–22.

[24] Tachikawa M. A method for estimating the distribution range of trajectories of

wind-borne missiles. J Wind Eng Ind Aerodyn 1988;29(1–3):175–84.

[25] Lin N, Letchford CW, Holmes JD. Investigations of plate-type windborne debris.

Part I. Experiments in wind tunnel and full scale. J Wind Eng Ind Aerodyn 2006;

94(2):51–76.

[26] Lin N, Holmes JD, Letchford CW. Trajectories of windborne debris and

applications to impact testing. J Struct Eng 2007;133(2):274–82.

[27] Holmes JD, Letchford CW, Lin N. Investigations of plate-type windborne debris.

II. Computed trajectories. J Wind Eng Ind Aerodyn 2006;94(1):21–39.

[28] Kordi B, Kopp GA. The effect of local flow field on the flight on wind-borne

debris. In: 11th Americas Conference on Wind Engineering. 2009.

[29] Kordi B, Traczuk G, Kopp G. Effects of wind direction on the flight trajectories

of roof sheathing panels under high winds. Wind Struct 2010;13(2):145–67.

[30] Beason WL. Breakage characteristics of window glass subjected to small

missile impacts. Thesis, Civil engineering department, Texas Tech University;

1974.

[31] Harris PL. The effects of thickness and temper on the resistance of glass

to small missile impact. thesis, Civil engineering department, Texas Tech

University;1978.

[32] Pantelides CP, Horst AD, Minor JE. Post breakage behavior of heat strengthened

laminated glass under wind effects. J Struct Eng 1993;119(2):454–67.

[33] Behr RA, Kremer PA. performance of laminated glass units under simulated

windborne debris impacts. J Architect Eng 1996;2(3):95–9.

[34] Ji FS, Daharani LR, Behr RA. Damage probability in laminated glass subjected

to low velocity small missile impacts. J Mater Sci 1998;33(19):4775–82.

[35] Saxe TJ, Behr RA, Minor JE, Kremer PA, Dharani LR. Effects of missile size and

glass type on impact resistance of sacrificial ply laminated glass. J Architect

Eng 1992;8(1):24–39.

G. Fernandez et al. / Engineering Structures 32 (2010) 3384–3393 3393

[36] Dharani LR, Ji F, Behr RA, Minor JE, Kremer PA. Breakage prediction of

laminated glass using the sacrificial ply design concept. J Architect Eng 2004;

10(4):126–35.

[37] Minor JE, Beason WL, Harris PL. Window glass failures in windstorms. J Struct

Div 1976;102(ST1):147–60.

[38] NAHB research center. WindBorne Debris – impact resistant of residential

glazing. Report prepared for the US department of housing and Urban

development; 2002 (Partnership for advancing technology in housing – PATH

program).

[39] Minor JE. Window glass failures in windstorms. J Struct Div 1976;102:147–60.

[40] Bole SA. Investigations of the mechanics of windborne missile impact on

window glass. Thesis, Civil engineering department, Texas Tech University;

1999.

[41] Ball A, McKenzie HW. On the low velocity impact behavior of glass plates.

J Phys 1994;4(8): c8-783–c8-788.

[42] Wilson JF. Similitude experiments on projectile induced fracture of monolithic

glass. Int J Impact Eng 1996;18(4):417–24.

[43] Wiederhorn SM, Lawn BR. Strength degradation of glass impacted with sharp

particles: i, annealed surfaces. J Am Ceram Soc 1979;63(1–2):66–70.

[44] Masters FJ, Gurley KR, Shah N, Fernandez G. The vulnerability of residential

window glass to lightweight windborne debris. Eng Struct 2009;32(4):

911–921.

[45] Simpson RH. (Attributed): the hurricane disaster potential scale. Weatherwise

1974;27:169–86.

[46] Saffir H. 1975: Low cost construction resistant to earthquakes and hurricanes.

ST/ESA/23, United Nations; 1975.

[47] Simiu E, Vickery PJ, Kareem A. Relation between Saffir–Simpson hurricane

scale wind speeds and peak 3-s gust speeds over open terrain. J Struct Eng

2009;133(7):1043–5.