A Novel Approach to Estimate the Wind Uplift Resistance of Roofing Systems

lable at ScienceDirect

Building and Environment 44 (2009) 723–735

Contents lists avai

Building and Environment

journal homepage: www.elsevier .com/locate/bui ldenv

A novel approach to estimate the wind uplift resistance of roofing systems

Bas A. Baskaran*, Steven Kee Ping Ko, Suda MolletiInstitute for Research in Construction, National Research Council of Canada, 1200 Montreal Road, Ottawa, ON, Canada K1A OR6

a r t i c l e i n f o

Article history:Received 15 January 2008Received in revised form 17 June 2008Accepted 18 June 2008

Keywords:DecksFastenersWind upliftDesign alternatives

* Corresponding author. Tel.: þ1 613 990 3616; faxE-mail address: [email protected] (B.A. Baskara

0360-1323/$ – see front matter Crown Copyright � 2doi:10.1016/j.buildenv.2008.06.024

a b s t r a c t

Roof wind design consist of three parts: determination of wind loads, evaluation of wind uplift resistanceand correlating the resistance with the design load such that the resistance is higher than the loadrequirement. Wind uplift resistance of a system with its respective components is evaluated in laboratorytesting. This paper presents a novel approach to estimate wind uplift resistance when components aresubstituted during field application. Wind dynamics, on a mechanically attached single-ply roofingassembly, lift the membrane and cause fluttering, introducing stresses at the attachment locations. Insuch assemblies, the fastener–deck interface is a critical design factor. First, by taking steel deck asa component this paper systematically characterizes the various steel decks that are commonly used inlow slope application. Second component, namely the fasteners and its engagement strength with deckhave been quantified for variations of its design, size and sources. Based on this component character-ization, fastener pullout resistance (FPR) is identified as a verification factor for system wind resistanceestimation. When variations occur in the fastener–deck interface between the proposed and the existingconfigurations, the present research through case studies has proved that: ‘‘as long as the FPR of theproposed configuration is higher than the existing configuration then wind uplift ratings can be maintained’’.This is valid as long as both the configurations have all the remaining roofing components similar withcomparable layout. Based on this verification, the study recommends that the testing lab should reportthe FPR along with the wind uplift resistance such that FPR can be used as a verification factor to acceptdesign/field alternatives.

Crown Copyright � 2008 Published by Elsevier Ltd. All rights reserved.

1. Introduction

Recent, natural wind hazard damages are dramatic, incurringlosses of life and property around the world. Wind-induced rooffailure is one of the major contributors to insurance claims, and it isrising. Recently, members of the Roofing Committee on WeatherIssues (RICOWI) completed two major wind investigation projectson Hurricane Charley and Katrina, documenting extensive factualdata on roof damages, which once again proved that wind-inducedroof failures are still a major concern [1,2].

In North America over 50% of low slope applications aremechanically attached single-ply roof systems (MAS). Marketsurvey indicates that there is continuous growth for the MAS. InMAS the waterproof membrane is attached to the deck usingmechanical fasteners. Steel deck is commonly used as deckingmaterial for the MAS. Waterproofing membrane attachmentmechanism varies depending on the type of membrane beingused. Fig. 1 illustrates the different modes of membraneattachments. MAS with thermoset membranes, the attachment

: þ1 613 998 6802.n).

008 Published by Elsevier Ltd. All

can be done two ways: as inseam attachment or as topside (barover) attachments. MAS with thermoplastic membranes, theattachment can be done either as one side weld or as double-side weld attachment.

The wind dynamics on MAS causes deflection of the membraneand thereby inducing tension in the membrane. This membranetension is resisted by the fastener–deck attachment. The windloading on the fastener can be asymmetrical as in spot – fastenedsystems or symmetrical as in bar-over attachment as shown inFig. 1. When the fasteners are exposed to multi-directional windforces, rotation and rocking occurs. These effects cause fatiguedeformation on the metal deck. Ultimately, this prolonged fatigueweakens the deck–fastener engagement and leads to a reductionin the fastener pullout resistance (FPR). When the resistance rea-ches below the design load requirement, failure occurs on theassembly [3].

This paper presents a novel approach to estimate wind upliftresistance when components are substituted during field appli-cation. In MAS, the fastener–deck interface is a critical designfactor. First, by taking steel deck as a component this papersystematically characterize the various steel decks that arecommonly used in the low slope application. Second components,namely the fasteners and its engagement strength with deck have

rights reserved.

mailto:[email protected]

www.sciencedirect.com/science/journal/03601323

http://www.elsevier.com/locate/buildenv

Fig. 1. Wind-induced fastener forces on various mechanically attached roofing system.

B.A. Baskaran et al. / Building and Environment 44 (2009) 723–735724

been quantified for variations of its design, size and sources.Based on this component characterization, fastener pulloutresistance (FPR) is identified as a verification factor for systemwind resistance estimation. Through case studies, it has beenproved that: ‘‘as long as the FPR of the proposed configuration ishigher than the existing configuration then wind uplift ratings can bemaintained’’. Based on this verification, the study recommendsthat the testing lab should report the FPR along with the winduplift resistance such that FPR can be used as a verification factorto accept design/field alternatives.

1 Imperial units as the default units with SI units in bracket.

2. Evaluation of steel deck properties

The two common properties of steel deck that influences FPRare gauge/thickness and tensile strength. Steel decking is availablein 22 and 20 gauge thickness at Grade 33 and 80. These gradesrepresent the yield strength of the decks, i.e., 33 ksi and 80 ksi.These are commonly used decks in low slope roofing applications.Present study, through component evaluation, characterizes thephysical and mechanical properties and compares them withpublished (nominal) values.

2.1. Experimental setup and data

The experiments were carried out as per the following threestandards:

� ASTM E8-03 – Standard test methods for tension testing ofmetallic materials [4]� ASTM A370-03a – Standard test methods and definitions of

mechanical testing of steel products [5]� ASTM A653/A 652M-03 – Standard specifications for steel

sheet, zinc coated (galvanized) or zinc–iron-alloy-coated(galvannealed) by the hot-dip process [6]

As shown in Fig. 2, the steel deck coupons in the shape of dog-bone with dimensions of1 800 � 3/400 (203 mm� 19 mm) havinga middle tapering section of 1/200 (13 mm) were cut from the maleflutes (top flange) of the steel decks. Ten coupons for each of thedeck types, namely 22 Ga-33, 22 Ga-80, 20 Ga-33 and 20 Ga-80,

Fig. 2. Steel deck coupons for property evaluation.

B.A. Baskaran et al. / Building and Environment 44 (2009) 723–735 725

were cut. All the four decks were obtained from the samemanufacturer.

Gauge (thickness) of the steel deck coupons was measured usingvernier calipers as shown in Fig. 3. The vernier caliper had anaccuracy of 0.0100 (0.254 mm). Thickness measurements were madeat three locations for each sample and the obtained values wereaveraged to obtain the mean thickness of the sample. Fig. 3 presentsthe comparison between the averaged measured thickness and thenominal (published) thickness of the tested samples. As shown inFig. 3, for 22 Ga-33 and 20 Ga-80, the measured and the nominalthickness are similar. However, for 22 Ga-80, the measured was 10%greater than the nominal and for 20 Ga-33, the measured was 6%lower than the nominal thickness.

Strength is a general term that refers to the capacity of thestructure to resist loads. The common parameter with which thesteel deck is referenced in the roofing industry is the yield strength.Yield strength is the magnitude of the load required to cause

0.0270.0280.0290.03

0.0310.0320.0330.0340.0350.0360.037

22-33-06 22-80-06 20-33-06 20-80-06

Sample ID

Th

ickn

ess, in

ch

es

Measured Thickness (1 inch = 25.4 mm)

Measured Nominal

Fig. 3. Measured vs. nominal thickness of the steel decks.

yielding in the deck or the maximum stress (force/area) that can beapplied without permanent deformation of the steel deck. In otherwords, it is the limit strength on the stress–strain curve, i.e., thestress level at which the linear portion of the curve ends. The yieldstrength is different from the ultimate strength, which representsthe ultimate load or failure load. Present practices use the yieldstrength to represent the tensile strength. To demonstrate thedifferences, the tensile strength of the above four steel deck typeswere evaluated and compared with the published (nominal) values.

Steel deck coupons, as shown in Fig. 4, were subjected to tensilestress using INSTRON Model 4502 machine. As shown in the figure,ends of the coupon having wider widths were gripped on the topand bottom grips of the computer – controlled INSTRON. Thebottom jaw is fixed. The movable top jaw applies a force at the rateof 674 lbf/min (3000 N/min). While the experiment is running, thesoftware records the stress vs. strain curve. At the completion of thetest, data were stored in an ASCII file for further analysis.

Fig. 4 also shows two typical stress–strain curves obtained fromthe tensile testing of the two deck types. Clear differentiation canbe observed from the stress–strain plots between the Grade 33 andGrade 80 decks. In the case of the Grade 33 deck, the stress–strainplot has three distinctive characters identified as: yield strength,ultimate tensile strength and fracture/failure point. The yield pointis followed by large plastic strains. This is an important character-istic of structural steel. It is used to classify the steel deck as ductileor brittle. This phenomenon of ductility can be observed from thefailure coupons of Grade 33 decks also shown in Fig. 4, where thenarrow region of the coupon has been stretched/necked thatresulted in the reduction of the area of the coupon. However, in thecase of the Grade 80 decks, this pronounced differentiationbetween the yield strength and ultimate strength is not observed.The reduction in area in the narrow region of the coupon is insig-nificant and the failure happened close to the tapering section ofthe narrow part of the coupon. It failed very abruptly without anyprior warning. This failure mode was quite contrary as compared to

Instron Machine

Region ofnecking

Region offracture

Region offracture

22-Ga (33 ksi) 22-Ga (80 ksi)

0

10

20

30

40

50

60

70

Tensile Strain, % Tensile Strain, %

Ten

sile S

tress, ksi

Ten

sile S

tress, ksi

0 10 20 30 40

22 Ga -33 ksiFracture

Ultimate Tensile.

Strength

Yield Strength

0

20

40

60

80

100

120

0 0.5 1 1.5 2 2.5 3 3.5 4

0.2% Offset

Yield Strength

22 Ga –80 ksi

Ultimate Tensile

Strength

Fig. 4. Yield strength measurement of steel decks.

0

10

20

30

40

50

60

70

80

90

22-33-06 22-80-06 20-33-06 20-80-06Sample ID

Yield

S

tren

gth

, ksi

(1 ksi = 6.9 Mpa) Measured Nominal

Fig. 5. Measured vs. nominal yield strength of the steel decks.


Grade 33 deck coupons. To determine the yield point of the Grade80 deck, as suggested in Ref. [18], the theory of 0.2% offset yieldpoint was applied. The yield stresses thus calculated were around80 ksi for different coupons. Comparing the physical propertiesbetween the two deck grades, the Grade 80 deck has very high yieldstress as shown in the Y-axis (80 ksi vs. 45 ksi) and relatively lowvalues of strain (3.5% vs. 30%) as shown on the X-axis. Thiscomparison clearly indicates that the Grade 80 decks are not asductile as the Grade 33 decks.

The yield strength measured from the 10 coupons of each decktype has been averaged and the mean values of the yield strengthare compared to their nominal (published) values as shown inFig. 5. The nominal yield strength values of the Grade 33 steel decksare lower than the measured data. In the case of the Grade 33 deck,the measured yield strength was on average 51% greater than theirnominal value. For the Grade 80 deck, the measured yield strengthis almost similar to their nominal value.

3. Evaluation of fastener pullout resistance

In MAS, membranes and insulations are attached to the steeldeck using fasteners. The strength of fastener–deck interface isa very important parameter in the design of roofing assembly. Toevaluate this interface three fastener sizes (#14, #15 and #21) areobtained from four sources (S1, S2, S3 and S4). Fig. 6 presents thephysical characteristics of the tested fasteners in terms of fastenerhead, tip and shank thread. For quick reference, the fastener–sourcecombination can be named as #14-S1, #15-S2, #21-S2, #15-S3 and#21-S4.

� #14-S1 had a thread diameter of 0.17000 (4.37 mm) and headdiameter of 0.4300 (11.0 mm);� #15-S2 had a thread diameter of 0.28100 (7.13 mm) and head

diameter of 0.4400 (11.2 mm);� #15-S3 had a thread diameter of 0.26100 (6.62 mm) and head



diameter of 0.6700 (16.9 mm).

Fastener

Type &

Source

Head Tip Thread

Source 1

#14

Source 2

#15

Source 2

#21

Source 3

#15

Source 4

#21

Fig. 6. Physical characteristics of the tested fasteners.


As shown in Fig. 6, differences exist among the fasteners in the head,thread, and tip design that could contribute to the pullout resistance.Closer examination reveals the differences among the fasteners’ threaddesign, namely distance between the thread and its angle to the shank.This observation is not only true for different fastener sizes but alsovalid for the same fastener size from different sources.

3.1. Experimental setup and data

To quantify the FPR, experiments were carried out as per ‘‘ANSI/SPRI FX-1’’ [7]. Steel deck samples were prepared by cutting thedecks in dimensions of 7.500 (191 mm) in width and 1200 (305 mm) inlength as shown in Fig. 7. It was necessary to use these dimensions,since samples tested with shorter lengths 3.500 � 600 and 600� 600

(89�152 mm and 152�152 mm) provided inaccurate resultsowing to excessive sheet deformation. On the other hand, samplestested with 1200 �1200 (305� 305 mm) do not affect the pullout

resistance of the fasteners. The fasteners were centered withrespect to length and installed with a minimum penetration depthof 3⁄4

00 (19 mm) to a maximum depth of 100 (25 mm). Testing wascarried out on INSTRON Model 4502. To accommodate the samplesin the INSTRON a steel attachment base was designed and thesample was held down along its female ribs through the predrilledholes. With the sample in position, a constant upward force wasapplied by pulling the fastener at a speed of 200/min (51 mm/min).All the samples were tested following the same procedure. Whilethe experiment is running, the software records the load vs.displacement curve in kilograms and millimeters. At the comple-tion of the test, data were stored in an ASCII file for further analysis.

Fig. 7 also shows conditions of 22 Ga-33 and 80 decks after thepullout of #21 fasteners. In the case of grade 33 decks, the fastenerwas gradually pulled out of the deck causing sheet deformationalong the fastener’s perimeter or area of influence in a conicalshape as shown in Fig. 7. The sheet deformation around the fastener

80 ksi Steel Deck

33 ksi Steel Deck

Fastener Under

Pullout

80 ksi Steel Deck33 ksi Steel Deck

Fig. 7. Experimental setup for fastener pullout measurements.


was approximately 0.1500 (4 mm) vertically. However, in the grade80 decks, the fastener was pulled out with a small conicaldeformation around the fastener’s perimeter without any sheetdeformation of the deck. This was the same trend observed for allthe tested fastener types when tested on grade 33 and 80 decks.This phenomenon of sheet deformation in grade 33 decks can berelated to their ductility property as discussed in Section 2. Due tothis ductile property, the fastener while pulling out pulls andstretches the steel deck along with it, which leads to thedeformation of the steel deck. For the fasteners tested on grade 80decks, the failure pattern was sudden and the fastener pulls out ona snap without deforming the steel deck.

Fig. 8 shows the typical load vs. displacements recorded duringthe fastener pullout testing. The graph refers to results obtainedfrom testing fasteners #15-S2 and #21-S2 on 22 Ga-33 ksi and22 Ga-80 ksi steel decks. The dotted lines refer to the load–displacement curves on grade 33 decks and the solid lines refer tograde 80 decks. A quick inference that can be made from theseload–displacement curves is, irrespective of the fastener type,fastener pullout on grade 33 decks has lower load while fastenerpullout on grade 80 decks has higher load.

Average FPR along with deviation bars are shown in Fig. 9. Theplotted FPR data represent the average value of 10 independentpullout tests for each fastener–deck combination. The standarddeviations are also calculated and shown in the figure. From theplotted graph the following observations are noted:

� Irrespective of the deck type and fastener source, fastener #21yields the highest FPR followed by #15 and #14, respectively,indicating that greater thread diameter resulted in higherfastener pullout resistance.� Within the steel decks having same gauge the measured data

indicate that higher FPR is obtained on higher yield strengthdecks. For example, #15-S2 fasteners showed an increase of60% in pullout resistance when engaged with 22 Ga-80 deckscompared to 22 Ga-33 decks. This was the same trend for allthe other cases with the variation in magnitude from onefastener–deck combination to the other.� The effect of deck gauge on FPR is very clear for the grade 33

decks, where the data indicated that by increasing the deckthickness one could obtain higher FPR. For example, #14-S1showed an increase of 15% from 22 Ga-33 to 20 Ga-33, while#15-S2 and #21-S2 measured an increase of 21%, and #15-S3and #21-S4 showed an increase of 12%, respectively.� Similar trend in the FPR is also observed in the case of the grade

80 decks with a maximum increase of 30%.� Within the same fastener type but from two different sources

there was a marked difference in the measured FPR values. Forexample, fastener #21 was obtained from S2 and S4. Themeasured FPR values on 22-80 decks indicate that #21-S2measured 15% higher than that of #21-S4. Though both thesefasteners have industry classification of #21 the tip designs ofthese fasteners are quite different as shown in Fig. 6, which

0

1

2

3

4

5

6

7

0 5 10 15 20 25Cross-Head Displacement, mm

0 5 10 15 20 25Cross-Head Displacement, mm

FP

R, kN

0

1

2

3

4

5

6

7

FP

R, kN

#15-S2: 22 Ga - 33 ksi#15-S2: 22 Ga - 80 ksi#21-S2: 22 Ga - 33 ksi#21-S2: 22 Ga - 80 ksi

(1 kN= 228 lbf)

#15-S2: 20 Ga - 33 ksi#15-S2: 20 Ga - 80 ksi#21-S2: 20 Ga - 33 ksi#21-S2: 20 Ga - 80 ksi

Fig. 8. Load vs. displacements curves during the fastener pullout testing.


could be the reason for this variation in the FPR values. Simi-larly, the #15 fasteners obtained from S2 and S3 had differentdimensional characteristics especially the thread diameter andtip design. Fastener #15 from S2 has a greater thread diameter(0.28100) when compared to S3 (0.26100), which could bethe contributing factor for measuring higher FPR values with#15-S2.

From the above discussion, one can conclude that the compo-nent evaluation data revealed that the FPR could vary whencomponent type or component physical property or source varies.

4. A novel approach to verify system wind uplift performance

Section 2 of this paper documented the variation in deckproperties by comparing the nominal (published) values with themeasured data. These comparisons considered four different decks

0

200

400

600

800

1000

1200

1400

#14-S1 #15-S2 #15-S3 #21-S2 #21-S4Fastener ID

FP

R, lb

f

22-33 ksi20-33 ksi22-80 ksi20-80 ksi

1 lbf = 4.4 N

Fig. 9. Comparison of the fastener pullout resistance for different deck types.

that are commonly used in low slope application with mechanicallyattached roof assemblies. It concluded that irrespective of the decktypes the measured data are always higher than the publishedvalues. Section 3, on the other hand, investigated the fastenerpullout resistance (FPR). Three different fasteners from fourdifferent sources are considered that are commonly used in lowslope application with mechanically attached roof assemblies. Itconcluded that the higher the fastener shank diameter, the higherthe FPR. This is true irrespective of the deck types.

These component investigations provided valuable data,however, they are limited when one wants to quantify their influ-ence on system performance. These limitations can impact thewind uplift rating design of roof systems in several ways:

� It is rather challenging for a designer to account for similaritiesfor all components between design and construction. As onecan expect, the probabilities are high during the constructionto substitute similar components.� Normally, roof systems are tested in the lab conditions to

obtain wind uplift rating [8,9]. However, during the fieldinstallation the deck type may be different from that of the oneused in the lab. Question arises whether the deck differencebetween the field and lab will influence the wind uplift rating.If so, how much?� Along the similar thought process, roof systems are tested in

the lab with a particular fastener type, and situations in thefield may warrant use of a different fastener type or fastenerfrom different source or fasteners with different design. Thiscan challenge the system supplier to justify that the winduplift rating will not be compromised due to this change.� Also, it would not be economical for system suppliers to test for

wind uplift rating each time when components are varied orsubstituted.� The above discussions can get complicated in the case of re-

roofing/recovering situations. Normally, only above deckcomponents will be re-roofed/recovered by keeping theexisting deck. How much the wind uplift rating is affected if theexisting deck is not same as that of the tested ones? How onecan avoid deck replacement at the same time maintain thesimilar wind uplift ratings? Also, in re-roofing there is a need toreplace only certain segments of the degraded decks. Then alsothe above questions are equally valid.

Proposing a practical design solution that can account forthese variations would be beneficial for the roofing communityduring the wind uplift design. Based on the extensive experi-mental research carried out at the Dynamic Roofing Facility(DRF) of the National Research Council of Canada, the presentstudy identifies a single parameter, namely, the fastener pulloutresistance (FPR) as a verification parameter when variationsoccur in the fastener–deck interface. If the components of thefastener–deck interface are varied, one should compare the FPRbetween the two sets. Through case studies, it has been provedthat: ‘‘as long as the FPR of the proposed configuration is higherthan the existing configuration then wind uplift ratings can bemaintained’’. This is valid as long as both the configurations haveall the remaining roofing components similar with comparablelayout. For example, a system is tested in a lab obtained a winduplift rating of 60 psf (2.9 kPa). For this system, the FPR of thefastener–deck combination is 400 lbf (1760 N). During the fieldinstallation the deck is different from the one used in the lab.Nevertheless, the FPR of the field fastener–deck combination ishigher than 400 lbf (1760 N). Then irrespective of the deckvariation, the designer can expect similar wind uplift rating inthe field. This has been demonstrated through case studies inthe following section.

Fig. 10. CSA dynamic load cycles.


5. Proposal validation through case studies

5.1. Design dilemma 1

� A roofing system tested at the DRF in accordance to the CSAA123.21-04 [10] obtained a wind uplift rating of 90 psf(4.3 kPa) as an indication of its wind uplift performance. Let theLab System be referred as LS1.� A building needs re-roofing and based on the wind load

calculation in accordance to the NBCC [11], it requires a systemwith a wind uplift rating of 90 psf (4.3 kPa).� Manufacturer X proposes the LS1 as a viable option to meet the

design requirement.� When reviewing the lab report, the designer noticed that

the LS1 was tested on 22 Ga-33 decks. However, theexisting deck in the field has a 22 Ga-80 deck. This raisesthe question whether this variation in deck can affectthe wind uplift rating. Let this Field System be referred asFS1.

5.2. Solution 1

� According to the present study’s practical solution, ‘‘as long asthe FPR of the proposed configuration is higher than the existingconfiguration then wind uplift ratings can be maintained’’.� Based on this design solution, the verification factor FPR was

determined for the fastener–deck interface of the LS1 andFS1 in accordance to the ANSI/SPRI FX 1 [7]. LS1 yielded an

FPR of 420 lbf (1820 N). Whereas the FPR of FS1 was 660 lbf(2900 N).� Even though there is a variation in the deck type, the FPR

of FS1 is higher than the LS1, therefore it is expected thatwind uplift rating can be maintained between these twosystems.

5.3. Validation experiment 1

All the experimental works were carried out at the DRF. The DRFfeatures are given in Baskaran and Lei [12]. All assemblies weresubjected to the simulated wind dynamics in accordance to CSA A123.21-04 standard [10]. Fig. 10 shows the CSA load cycle and it hasfive rating levels (A–E). The test pressure on the Y-axis correspondsto the design pressure, in accordance with local building codes orwind standards. The CSA load cycle is more detailed in Baskaranet al. [13].

Using a thermoset waterproofing membrane, two full-scalesystem mockups, LS1 and FS1, were constructed and tested. Thesystem layout details are shown in Fig. 11. LS1 had 22 Ga-33 ksideck as the structural substrate. In the DRF bottom frame, two fullsheet of 3600 (914 mm) wide and a cut sheet of 2400 (610 mm) widesteel decks were installed along the length of the test frame. A 6-milpolyethylene sheet was installed as the barrier over the deck inloose laid condition. The insulation configuration comprised ofa layer of 200 (51 mm) thick boards. Four full boards of 4800 � 7900

(1219 mm� 2006 mm) and one partial board of 4400 �7900

(1118 mm� 2006 mm) were installed with the long edges

Fig. 11. System details for benchmarking experiment 1.


perpendicular to the steel deck flutes. The boards were mechan-ically fastened to the steel deck with 8 fasteners/board. Themembrane layout comprised of three EPDM membrane sheets,which were laid on top of the insulation and were fastened with#14 fasteners at a fastener row spacing (Fr) of 7800 (1981 mm) andfastener spacing (Fs) of 1200 (305 mm) along the seam. The fasteningplates were 200 (51 mm) diameter polymer plates. The membraneseam, which had an overlap of 600 (152 mm), was primed andadhered to form overlap seam with an adhesive tape as shown inFig. 11. This type of attachment is known as inseam attachment asreferenced in Fig. 1.

To duplicate the field condition, all the above components weremaintained constant as that of the LS1 with the exception of thedeck. A second mockup with 22 Ga-80 decks was installed in theDRF bottom frame and labeled as Field System – FS1.

Both LS1 and FS1 were subjected to the CSA A123.21-04 [10]load cycle and Fig. 12 shows the system response. The measuredwind uplift resistance data indicate that LS1 passed the first threelevels (60, 75 and 90 psf – 2.9, 3.6 and 4.3 kPa) of the CSA load cycle.Since the system failed at Level D sequence 4, LS1 obtained a winduplift rating of 90 psf (4.3 kPa). In the case of FS1, the wind upliftrating is 105 psf (5 kPa) as the FS1 successfully sustaining all fourlevels (60, 75, 90 and 105 psf – 2.9, 3.6, 4.3 and 5 kPa) and failedduring the sequence 4 at Level D of the CSA load cycle. As proposedby the present study, since the verification factor FPR of the FS1 ishigher than the LS1, the deck variation does not influence therequired 90-psf (4.3 kPa) wind uplift rating.

To generalize the concept of the FPR as verification factor, it hasbeen decided to investigate further on systems with differentwaterproofing membrane, namely a system with thermoplasticmembrane. The primary difference in the system constructionusing a thermoset or a thermoplastic membrane is the seamattachment mechanism as discussed in Fig. 1. The following casestudies demonstrate the viability of the verification factor FPR forsystems having thermoplastic membrane.


� A roofing system with thermoplastic membrane tested atthe DRF in accordance to the CSA A123.21-04 [10] obtaineda wind uplift rating of 157 psf (7.5 kPa). This is referred asLS2.� Similar to the ‘‘Design Dilemma 1’’, this roof system needs to be

installed in field, which has requirement of 150 psf (7.2 kPa).� Comparing the components used in the lab with the

arrangements available in the field, it has been noticed thatvariation exists in the thickness (gauge) of the deck, i.e., thedeck used in the lab was 22 Ga and whereas the deck in thefield is 20 Ga. Note that the as discussed in Section 2,difference in thickness of deck can result difference in yieldstrength. Nevertheless, the question is whether this compo-nent variation will affect the overall system wind upliftrating.� Let the thermoplastic field system be referred as FS2

5.5. Solution 2

� The fastener–deck interface of LS2 (#15–22 Ga) yielded an FPRof 685 lbf (3014 N), while FS2 (#15–20 Ga) had an FPR value of890 lbf (3930 N)� Even though there is a variation in the deck thickness, the

verification factor FPR of FS2 is higher than the LS2, therefore itis expected that similar wind uplift rating can be maintainedbetween these two systems.


As shown in Fig. 13, a mechanically attached thermoplasticsystem (TPO) was selected. In LS2 the steel deck layout comprises of22 Ga. With steel deck installed, self-adhered membranes were laidon top of it as the barrier. A layer of 200 (51 mm) thick insulation with

0 10000 20000 30000 40000Time, sec

0

30

60

90

120

Pressu

re, p

sf

0

30

60

90

120

Pressu

re, p

sf

15

30

46

61

30

46

61

38

57

77

57

77

46

70

93

70

92

54

82

109

Pt = 60psf

070627Pref

0 10000 20000 30000 40000Time, sec

070619Pref

15

30

45

60

30

45

60

38

56

74

56

75

45

67

90

68

90

53

79

104

59

89

Pt = 60psf

Lab System 1

Field System 2

a

b

Fig. 12. Validation of the lab system performance with field system for benchmarkingexperiment 1.


four full boards of 4800 � 7900 (1219 mm� 2007 mm) and one partialboard of 4400 �7900 (1118 mm� 2007 mm) was installed with thelong edges perpendicular to the steel deck flutes. The boards weremechanically fastened to the steel deck with 8 fasteners/board. TheTPO membrane sheets were installed at a row spacing (Fr) of 70.500

(1791 mm) and were fastened at 1200 (305 mm) spacing (Fs) alongthe seam, resulting in 7 fasteners per seam. A typical double-sideweld (DSW) seam as shown in Fig. 1 was adopted as the seamattachment mechanism. It was done by fastening the seam overlapportion of the bottom sheet to the deck using #15-500 long fastenerswith polymer batten bars of 100 (25 mm) wide. Hot air welding was

used to seam the top membrane to the bottom one. FS2 construc-tion and components were similar to that of LS2 with the exceptionof the deck used.

When subjected to CSA dynamic wind loading cycle, themeasured wind uplift resistance data indicate that LS2 sustained90 psf (4.3 kPa) of wind uplift rating while FS2 sustained a winduplift rating of 105 psf (5 kPa). From the system responses shown inFig. 14, the concept of FPR as a verification factor to estimate winduplift performance of a roof system is validated.


� For the above-discussed scenario, during the field construction,due to various reasons the manufacturer was not able to supplythe same type of fastener (#15) for one segment of the roof. Asubstitute fastener type (#21) was proposed and it had an FPRof 1148 lbf (5050 N). Again, the question is how this will impacton the wind uplift rating?

5.8. Solution 3

� Even with the fastener substitute, the verification factorFPR is higher than that of LS2; therefore one can concludethe fastener substitution will not impact the wind upliftrating.


By keeping all the components constant, a third mock up wasconstructed at the DRF. For this mock up – labeled as FS3 – thesubstitute fastener (#21) was used to attach the membrane withthe deck. FS3 was also subjected to the CSA A123.21-04 [10] loadcycle and Fig. 15 compares the system wind uplift ratings. FS3 hada wind uplift rating of 175 psf (8.4 kPa). This once again validatesthe present study proposal of using the FPR as a verificationfactor.

In addition to the above validation it has also been observedthat an increase in the FPR can increases the wind uplift rating.LS2 with FPR of 685 lbf (3014 N) had a wind uplift rating of 157whereas the wind uplift rating is increased to 180 psf as FPRincreases to 1148 lbf (5050 N). To understand this observation,the measured fasteners forces were further analyzed and pre-sented in Fig. 16. Fig. 16 compares the measured fastener forceat various pressure levels for LS2, FS2 and FS3. Data corre-sponding to the Group one of the CSA load sequences (referFig. 10) were used. Irrespective of the deck type, the maximummeasured fastener load is less than the fastener pullout resis-tance. For fastener/deck combination used on LS2 and FS3 the22 Ga deck, the pullout resistance from component testing was685 lbf (3014 N) and 1143 lbf (5050 N), respectively. Themeasured fastener loads were about 640 lbf (2810 N) and 672 lbf(2950 N), respectively, from the system testing. In the case ofFS2, respective value was 893 lbf (3930 N) compares to 706 lbf(3106 N).

Currently, the generic equation force divided by area isa common practice to calculate the system resistance, i.e.

System resistance ¼ fastener pullout value=fastener spacing

� fastener row spacing ½1�

Eq. (1) is not true representation of the wind-induced dynamiceffects on the MAS. To demonstrate this, linear regression lineswere obtained for the LS2, FS2 and FS3 data. As discussed previ-ously, all three systems used same fastener spacing and fastenerrow spacing. In other words, tributary area was constant. However,

Fig. 13. System details for benchmarking experiment 2.


as shown below, regression equations had difference in slope asfollows:

YLS2 ¼ 3:2X (2)

YFS2 ¼ 3:5X (3)

YFS3 ¼ 3:7X (4)

From the above equations, it is clear that as the FPR increases, theslope increases, as does the wind uplift rating. Thus system resis-tance is a combined performance outcome of all components usedin the system.

During the review process of this paper two related issueswere raised. First the influence of the testing table size effect onthe wind uplift resistance of the system or the influence of tableedge effect on the wind uplift rating of the roofing system. Tableedge effect is a phenomenon, which affects the uplift perfor-mance of roofing system when systems are tested in laboratoryconditions using testing table smaller than the required table size.Authors [14,15] and other researchers [16,17] addressed this issuein the past. Fastener load is one of the critical design factors thatwill be influenced when systems are tested in a narrow table.Authors developed correction factors to overcome this issue andthose developments are now part of the Canadian nationalstandard CSA A123.21-04 [10]. The contribution of the presentpaper is to illustrate the relative comparison of the systems’performance. Thus the issue of table edge effect can be eliminatedand the FPR could still be applicable as a verification tool. The

second issue raised focus on the cost benefit analysis. As it can beagreed, documentation of the cost benefit analysis needs exper-tise and knowledge beyond technical investigations and data. Totackle this issue for making some conceptual and qualitativerecommendations, attempts were made by the author elsewhere[9].

6. Conclusion

Fastener Pullout Resistance (FPR) is a function of thestrength of the fastener–deck interface. To characterize thefastener–deck engagement strength, commonly used fastener–deck combinations were investigated by conducting pullouttesting. With the obtained results, further investigation wasconducted on full-scale roof systems with the same fastener–deck combinations, and attempts were made to relate the FPRwith the wind uplift resistance of the roof system. Based on thepresented results and discussions, the following conclusions canbe drawn:

1. FPR is characterized by three parameters namely, deck thick-ness, tensile strength and the physical properties of thefastener. The physical properties of the fastener include: shankdiameter, thread angle and tip design. FPR can be increasedeither by increasing the yield strength and gauge thickness ofthe deck or by engineering the fastener design.

2. FPR is identified as a verification factor for system wind resistanceestimation. When variations occur in the fastener–deck interface

0 10000 20000 30000 40000Time, sec

0 10000 20000 30000 40000Time, sec

0

30

60

90

120

150

180

210

Pressure, p

sf

0

30

60

90

120

150

180

210

Pressure, p

sf

Pt = 90psf

23

46

68

90

46

68

90

59

86

113

86

114

70

102 102

136 136

77

118

160

89

136

180

040107-Pref

Lab System 2 (LS2)

Pt = 90psf

040209-Pref

25

47

71

93 92

47

7160

85 87

113 115

71

102 104

138 138

80

117

160

162

90

137

180

Field System 2 (FS2)

b

a


157

180 175

0

20

40

60

80

100

120

140

160

180

200

Su

stain

ed

P

ressu

re, p

sf

Dec

k Va

riatio

n

Fast

ener

Var

iatio

n

Lab System 2 Field System 2 Field System 3

1 psf = 47.88 Pa)


y = 3.7x

y= 3.2x

y= 3.5x

0

100

200

300

400

500

600

700

800

0 50 100 150 200Pressure, psf

Measu

red

F

asten

er L

oad

, lb

f

LS2

FPR = 685 lbf

FS3

FPR = 1148 lbf

FS2

FPR = 893 lbf

(1 psf = 47.88 Pa)

(1 lbf = 4.4 N)

Fig. 16. Load verification for evaluating systems’ performance.


between the proposed and the existing configurations, thepresent research through case studies has proved that: ‘‘as long asthe FPR of the proposed configuration is higher than the existingconfiguration then wind uplift ratings can be maintained’’. This isvalid as long as both the configurations have all the remainingroofing components similar with comparable layout.

3. During the wind uplift evaluation of a roof assembly, FPR of thefastener–deck interface should be reported along with thesystem wind uplift rating. This will provide designers anunderstanding of the fastener–deck interface properties usedat the system testing in comparison to the fastener–deckinterface available in the field. Based on the FPR, design/fieldalternatives can be accepted.

Acknowledgements

The presented research is being carried out for a consortium –Special Interest Group for Dynamic Evaluation of Roofing Systems(SIGDERS). SIGDERS was formed from a group of partners whowere interested in roofing design. These partners included:

Manufacturers: Atlas Roofing Corporation, Canadian GeneralTower Ltd., Carlisle Syn Tec., GAF Materials Corporation, GenFlexRoofing Systems, Firestone Building Products Co., IKO IndustriesLtd., ITW Buildex, Johns Manville, Sarnafil Roofing, Soprema Can-ada, Stevens Roofing, Tremco and Trufast.

Building Owners: Canada Post Corporation, Department ofNational Defence, Public Works and Government Services Canada.

Industry Associations: Canadian Roofing Contractors’ Associa-tion, Canadian Sheet Steel Building Institute, National RoofingContractors’ Association and Roof Consultants Institute.

References

[1] RICOWI, Inc.. Hurricanes charley and Ivan wind investigation report. McDonough,Georgia: Roofing Committee on Weather Issues Inc.; 2006. p. 286.

[2] RICOWI, Inc.. Hurricane Katrina wind investigation report, powder springs,and Georgia. Roofing Industry Committee on Weather Issues, Inc.; 2007. p 183.

[3] Baskaran A, Smith TL. A guide for the wind design of mechanically attachedflexible membrane roofs. Ottawa, Ontario, Canada, K1A 0R6: NationalResearch Council Canada; 2005.


[4] ASTM E8-03 – standard test methods for tension testing of metallic materials.[5] ASTM A370-03a – standard test methods and definitions of mechanical testing

of steel products.[6] ASTM A653/A 652M-03 – standard specifications for steel sheet, zinc

coated (galvanized) or zinc-iron-alloy-coated (galvannealed) by the hot-dip process.

[7] SPRI. Standard field test procedure for determining the withdrawal resistanceof roofing fasteners, ANSI/SPRI FX-1 2006; 2006.

[8] Baskaran A, Lei W, Richardson C. Dynamic evaluation of thermoplastic roofingsystems for wind performance. Journal of Architectural Engineering1999;5(5):16–24.

[9] Baskaran A, Ko SKP. Optimizing the wind uplift resistance of mechanicallyattached roof systems. Journal of Architectural Engineering 2008;14(3):65–76.

[10] CSA Number A123.21-04. Standard test method for the dynamic wind upliftresistance of mechanically attached membrane-roofing systems, CanadianStandards Association, Canada; 2004.

[11] National Research Council Canada. National building code of Canada user’sguide – structural commentaries (part 4), Ottawa, Ontario, Canada, K1A 0R6;2005.

[12] Baskaran A, Lei W. A new facility for dynamic wind performance evaluation ofroofing systems. Proceedings of the fourth international symposium onroofing technology, NRCA/NIST, Washington, D.C., U.S.A.; 1997. p. 168–79.

[13] Baskaran A, Chen Y, Vilaipornsawai U. A new dynamic wind load cycle to evaluateflexible membrane roofs. Journal of Testing and Evaluation 1999;27(4):249–65.

[14] Baskaran A, Borujerdi J. Application of numerical models to determine winduplift ratings of roofs. Journal of Wind and Structures 2001;4(3):213–26.

[15] Baskaran A, Molleti S. Application of numerical models to determine wind upliftratings of roofs – part II. Journal of Wind and Structures 2005;8(3):213–33.

[16] Prevatt David O, Schiff D Scott, Malpezzi Joseph A. Investigation of chambersize for uplift performance testing of single-ply roof systems. Proceedings ofthe fourth international symposium on roofing technology, NRCA/NIST,Washington, D.C., U.S.A.; 1997. p. 163–7.

[17] Prevatt David O, Scott Schiff, D, Stamm S Joshua, Kulkarni S Amolprasad. Winduplift behaviour of mechanically attached single-ply roofing systems: the needfor correction factors in standardized test. Journal of Structural Engineering2008;134(3):489–98.

[18] Gere M. James. Mechanics of materials – fifth edition, chapter 1, section 1.3,Books/Cole Thomson Learning Academic Resource Center; 2001.

A Novel Approach to Estimate the Wind Uplift Resistance of Roofing Systems

Documents