Which is the Weakest Link? Wind Performance of Mechanically-attached Systems

Proceeedings of the RCI 21st International Convention Baskaran and Ko – 29
Which is the Weakest Link?
Wind Performance of Mechanicallyattached
Systems
Roof Consultants Institute
Bas A. Baskaran PhD, P.Eng
Steven Kee Pink Ko, M.Eng
M24 National Research Council Canada
ABSTRACT
Wind dynamics, on a mechanically attached single ply roofing assembly, lift
the membrane and cause fluttering, introducing stresses at the attachment
locations. To identify the component of the system that has the weakest
resistance against wind uplift forces, a dynamic method of evaluating roofing
systems is beneficial. Each component in the assembly offers resistance to
wind uplift. Failure occurs when the wind uplift force is greater than the
resistance of any one or more of these links. This study presents data that
will help roof system designers maximize wind uplift ratings by choosing the
appropriate roof components at the early design stage or by replacing/adding
components to improve wind uplift resistance during reroofing. By diagnosing
the weakest link from the failure mode, one can improve the wind resistance
of a system. Therefore, a series of Mechanically Attached Systems
(MAS) were constructed and exposed to simulated dynamic wind uplift forces
at the Dynamic Roofing Facility of the National Research Council Canada
(NRC-IRC). The goal of the paper is to present the ongoing experimental program
and answer the following questions:
• By modifying the wind uplift resistance for one of the components,
how much does the wind uplift rating of the assembly improve?
• Does the presence of a vapor barrier (retrader) in a system modify
the force resistance chain? If so, to what degree does the wind
uplift rating change?
• Does a difference in steel deck thickness affect the force resistance
chain, and how much?
Through a case study, this paper concludes with a procedure for system optimization.
SPEAKER
DR. BAS A. BASKARAN is a group leader and senior research officer at the National Research
Council of Canada, Institute for Research in Construction (NRC/IRC). For 25 years, he has been
immersed in researching the wind effects on building envelopes through wind tunnel experiments
and computer modeling and has received national and international recognition for that work. He
also acts as adjunct professor at the University of Ottawa. He is a member of RCI, ASCE, SPRI,
RICOWI, ICBEST, and CIB technical committees. Baskaran has an outstanding research record
with more than 150 publications in refereed journals and conference proceedings. Baskara
received his master’s degree in engineering and PhD from Concordia University, Montreal,
Canada. Both research topics focused on the wind effects on buildings and earned best dissertation
awards from the Canadian Society of Civil Engineers.
This paper has been dedicated to the memory of James Sheahan, who represented RCI in SIGDERS meetings.
Jim was instrumental in the development of the “weakest link” concept.
Baskaran and Ko – 30 Proceeedings of the RCI 21st International Convention
PREAMBLE
More than 50% of North
American low-slope buildings are
roofed with single-ply assemblies
(NRCA, 2004). A Mechanicallyattached
System (MAS) is one in
which the waterproofing single ply
sheets are integrated with the
other components using fasteners.
This is becoming a popular
installation method due to its
lower installation cost.
Wind effects on the MAS are
shown in Figure 1. Wind dynamics
cause membrane elongation and
billowing. The magnitude of the
wind-induced suction and the
membrane’s elastic property
determine the billowing shape
and fluttering magnitude. The
components of a
MAS are: deck, fasteners,
plates, barriers
(retarders), insulation,
and membrane.
Each component
offers a certain
resistance to the
wind uplift force.
This can be illustrated
through a
force resistance link
diagram, which is
also shown in Figure
1. All resistance
links should remain
connected for the
system to be durable
and to keep the
roof properly in
place. Failure occurs
when the wind
uplift force is greater
than the resistance
of any one or
more of these links.
For example, the roof assembly is
considered failed when a fastener
(Link 6) pulls out from the deck
even if the membrane and its
seams are in good condition.
Similarly, a system is considered
failed when a seam (Link 2) tears
under gusting wind while other
components remain intact.
(Baskaran, 2002).
For a roofing designer, this
paper presents a procedure to
optimize a mechanically attached
system’s wind uplift resistance.
For a roofing component manufacturer,
this paper emphasizes
the need to scrutinize experimental
data and work to improve the
wind uplift resistance through
failure mode diagnosis. The paper
presents a cost benefit analysis
case study for a thermoplastic
system. However, the concepts are
equally valid for other MAS.
Experimental investigations for
this study were carried out at the
National Research Council Canada
(NRC/IRC) Dynamic Roofing
Facility (DRF). Details of the
DRF’s operations were documented
in Baskaran and Lei (1997).
SIGDERS dynamic test protocol
(CSA, 2004) was used and procedures
involved in the development
of the SIGDERS load cycle were
presented at a previous RCI conference
[Baskaran et. al. (1999)].
Proceeedings of the RCI 21st International Convention Baskaran and Ko – 31
Which is the Weakest Link?
Wind Performance of Mechanicallyattached
Systems
Figure 1: wind effects on mechanically-attached systems.
EXPERIMENTAL SETUP AND
RESULT DISCUSSION
System responses were measured
as time histories by equipping
the system with pressure
sensors and load cells. All the
measured time history records
were analyzed and presented as
pressure difference across the
membrane or insulation and tensile
forces on the membrane or
insulation fasteners. Throughout
the paper, data from system-sustained
conditions are reported in
three sections as follows:
• Effect of vapor barrier
(retarder) on wind uplift
• Effect of seaming technique
on wind uplift
• Effect of deck enhancement
on wind uplift
In each section, failure mode
photographs are included to identify
the weakest link of the tested
assembly.
Effect of Vapor Barrier
(Retarder) on Wind Uplift
Four full-scale system mockups
(Systems 1 to 4) were constructed
and tested. System components
were as follows:
• System 1: 22 Ga Type E
(80 ksi) steel deck. Note
that the measured tensile
strength of the deck
coupons was 101 ksi (696
Mpa). A layer of 2″ (51 mm)
thick ISO 4 by 8 boards
was attached with 8 fasteners
per board. A 45 mil
(1.2 mm) thick and 78”
wide (1981 mm) TPO
membrane was used.
• System 2 had the same
components as those of
System 1. In addition,
sheets of #15 building
papers were laid over the
steel deck. The sheets
were overlapped [6″ (152
mm)] and adhered with
cold adhesive.
• System 3: 6-mil poly
sheet. A poly sheet was
loose laid over the steel
deck.
• System 4: Self-adhered
membrane of 36″ wide
(914 mm) sheets fully
adhered directly over the
steel decks.
For discussion purpose,
System 1 is labelled as system
with no vapor barrier, (NOVB)
whereas Systems 2, 3, and 4 are
labelled as systems with a vapor
barrier (VB).
The geometric details of a typical
tested system are depicted in
Figure 2. Membrane sheets were
fastened at 12″ (305 mm) spacing
(Fs) along the seam, resulting in
seven fasteners per seam. Based
on this arrangement, the distance
between fastener rows (Fr) was
70.5″ (1791 mm). Fasteners were
5″ (127 mm) long with metal
plates of 2″ (60 mm) diameter. A
Baskaran and Ko – 32 Proceeedings of the RCI 21st International Convention
Figure 2: Typical system setup at the DRF.
Figure 3: Wind uplift ratings of thermoplastic systems with and
without vapor barrier (retarders).
typical one-side weld (OSW)
seam was used for membrane
overlaps. Each seam
had an overlap of 5.5″ (140
mm), with fasteners placed
1.5″ (38 mm) from the edge of
the under sheet, and 2″ (73
mm) from the edge of the
overlapping sheet. The portion
of the seam beyond the
fastener row was welded with
hot air such that a waterproof
top surface was obtained.
The width of the
welded portion varied between
1.5″ and 1.75″ (38 mm
and 45 mm).
When subjected to dynamic
winn.d loading cycles
the NOVB system, system 1,
sustained 90 psf (4.3 kPa) of
wind uplift rating where as
VB system sustained a higher
wind uplift rating of 135
psf (6.5 kPa). In other words,
VB systems performed about
50% better than NOVB systems.
This was found to be
true regardless of the vapor
barrier type used. Similar
observations were also
reported by Smith (1995)
under static testing.
After the wind test, failed systems
were examined for the weakest
resistance link. Figure 4 shows
a typical example of a failure
mode. Both systems failed due to
membrane tear around the fastener
plates. Closer examination
of the failure showed that the
membrane had been stretched
around the plates, and then was
torn completely away from the
plate, while the fasteners
remained engaged with the deck.
There was also an instance of
delamination failure, as the membrane
peeled slightly at one fastener
location. Examination of the
system’s seams after the test
revealed that the membrane had
experienced some stretching, and
bore teeth marks from the metal
fastener plates. The clamping
force of the membrane fastener
plate is higher than the tear resistance
of the membrane, and this
resulted in the “cookie cut” on the
membrane at the membrane center
fastener location. This might
have resulted in tearing along the
machine direction of the membrane
at the adjacent fastener
plate locations. In all of the scenarios,
the fasteners were
engaged well with the deck, and
there was no bending of the
plates. These conditions areas
lead to the conclusion that TPO
membrane resistance was the
weakest link in the assembly.
If there is no change in the
installation process and the weakest
link, then how can the wind
resistance of a VB system be better
than that of a NOVB system?
To understand this further, partial
time history data analysis,
shown in Figure 5, was collected
and analyzed. The partial time
history data analysis represents a
segment from the measured pressure
and load time history, that
respectively represents the loading
sequence 4 of Level A (90 psf
4.3 kPa) of the SIGDERS load
cycle. In the case of the NOVB
system, the membrane fastener
force varies from 0 to 350 Ibf (0 to
1557N) with 90 psf (4.3 kPa) wind
gusts. For the same 90 psf (4.3
kPa) wind gusts, on the VB system,
the membrane fastener force
ranges between 100 and 250 lbf
(445 to 1112 N). Measured loads
on the insulation fasteners were
about 150 lbf (667 N).
Proceeedings of the RCI 21st International Convention Baskaran and Ko – 33
Figure 4: Failure mode investigation identifies membrane as the
weakest link.
In other words, the insulation
shares about 40% of the tensile
forces of the membrane fasteners.
It is interesting to note that in
case of NOVB systems, there is no
pressure building up across the
insulation, and the insulation
acts as a simple spacer between
the membrane and the deck. The
membrane is the only component
resisting the wind uplift. However,
in the case of the VB system, the
insulation boards share about
40% of the applied suction. This is
due to the fact that these vapor
barrier types offer a certain resistance
to airflow in addition to
their preliminary function of limiting
vapor diffusion. A research
paper by Cook (1992) made the
following conclusion about the
presence of a vapor barrier in a
roof system. “The vapor barrier
shows a single-sided respond,
resisting suctions by pressing up
against the underside of the insulation
boards, but falling slack
into the deck troughs under pressure.
The deck takes the remainder
of the loads, since the sum of
all three components exactly balances
the applied suctions.”
Effect of Seaming Technique
on Wind Uplift
The systems failure modes
with and without vapor barriers
were membrane-related. This
means the weakest link for this
roof system assembly is the TPO
membrane. Failure mode investigation
identified developments of
asymmetrical forces at the membrane
attachment location assembly
as the cause of failure. This
asymmetrical force led to pulling
of membranes below the fastener
plate and the metal plates profile
configuration, cutting the membrane.
In order to improve the
Baskaran and Ko – 34 Proceeedings of the RCI 21st International Convention
Figure 5: System load sharing performance with and without vapor barrier (retarder).
system’s resistance, a double-side
weld (DSW) technique was used to
replace the one-side weld (OSW)
method. Figure 6 compares a force
diagram of OSW systems with
DSW systems. In OSW systems
(Figure 6a), tensile forces cause
tear forces along the direction of
the weld. With a weld on each side
of the fastener, tensile forces cause
tear forces that are developed
along the two opposite directions of
the welds.
During wind gusting, at a given
time, these forces can be expressed
as follows:
System with one-side weld:
(1)
(2)
System with double-side weld:
(3)
(4)
Where and are the respective
horizontal and vertical forces,
and , , and are the membrane
orientation angles due to
wind uplift.
It is evident from Equation 3
above that part of the forces developed
on the fasteners in the horizontal
direction for DSW systems
(Figure 6b) can cancel each other
because of their opposing equal
symmetrical stress distribution
pattern. On the other hand, the
stress distribution is asymmetrical
in the OSW system. This means
that the induced force along the
horizontal direction would be higher
for the OSW system than for the
DSW system. Consequently, the
degree to which the fasteners are
rocked sideways and cause membrane
tear would be reduced by
the DSW technology.
A new system mock-up, System 5, was constructed
and the DSW technique was used for seaming.
Polymer batten strips 1″ (25 mm) wide and 57
mil (1.5 mm) thick was used instead of round metal
plates. The DSW seam width was 4 3/8″ (111 mm)
and the hot air process was used to form the seams
with 1.5″ (38 mm) wide welds on each side of the
batten strip. System 5 also had a 6-mil poly vapor
barrier similar to that of System 3.
Proceeedings of the RCI 21st International Convention Baskaran and Ko – 35
Figure 6: Force diagram of OSW vs. DSW systems.
System 5 with DSW sustained
a maximum pressure of 158 psf
(7.56 kPa). Comparing it to a similar
system with one side weld, the
double-side weld system had
higher wind uplift resistance,
even though both systems have
the same layout and components.
After the wind test failed, the system
was examined to identify the
weakest link. Figure 7 shows the
failure mode of the DSW system.
Instead of the membrane tearing
around the fastener plate as in
the one side weld system, the
membrane center fastener was
pulled out of the seam in the double-
side weld system. This indicates
that the weakest link of this
system was fastener engagement
with the deck. Not only was this
mode of failure very abrupt, the
examination of the deck condition
showed extensive deck cracking
at the engagement location. There
was no noticeable damage at
either the polymer battens or the
membrane seams.
To understand the symmetrical
force transfer behavior of the
DSW system, time history data
was further analyzed for induced
membrane fastener force in the
horizontal direction (Fx). As shown
in Figure 8, the presented partial
time history data analysis is from
loading sequence 2 of Level A (45
psf/ 2.15 kPa) and 4 of Level A (90
psf/ 4.3 kPa) of the SIGDERS load
cycle. To quantify the relative performance
of the DSW with respect
to OSW, the Y-axis presents the
normalized data. This was
obtained by divining the DSW’s Fx
with that of the OSW. Both pressure
levels and the horizontal
forces of the DSW were significantly
reduced to less than 10% of
the relative value. The partial time
history data analysis supports the
DSW theory in reducing the
asymmetrical forces.
Effect of Deck Enhancement
on Wind Uplift
In a mechanically attached
roof system, the wind-induced
load is transmitted through the
membrane fasteners to the structural
deck. A thicker deck should
have a higher wind uplift resistance
as compared to a thinner
deck. This is due to the fact that
the pullout resistance of a given
fastener is higher in a 20 Ga deck
as compared to a 22 Ga deck. For
example, the pullout resistance of
a number 15 fastener was 700 lbf
(3114 N) and 900 lbf (4003 N) with
22 Ga and 20 Ga decks respectively.
In addition, tensile tests on
deck coupons revealed that the
tensile strength of 22 Ga deck was
101 ksi (696 Mpa), whereas 20
Ga, had tensile strength of 107
ksi (738 Mpa). System 5, with
double-side weld, failed due to
fastener pull out. An attempt was
made to find out how much the
Baskaran and Ko – 36 Proceeedings of the RCI 21st International Convention
Figure 7: Failure mode identifies deck as the weakest link.
deck thickness would affect wind
uplift forces on the roof system by
using a thicker (20 Ga) deck on
System 5. A new mock-up,
System 6, was constructed with
the same components used in
System 5 with the exception of the
deck. The new work-up was subjected
to the SIGDERS load cycle
with a test pressure of 90 psf (4.3
kPa), similar to the other systems.
System 6 passed the maximum
pressure 180 psf (8.62 kPa)
of the load cycle. It should be
noted that when testing under the
SIGDERS load cycle, one can
incrementally run the test up to a
maximum pressure level equal to
two times the selected test pressure
(Baskaran, 1999). In other
words, for System 6, there was no
failure observed when the system
passed Level E (180 psf). This
proved that thicker steel decks
could resist higher wind uplift
resistance.
Figure 9 compares the measured
fastener force at various
pressure levels for System 5 and
System 6. Only data corresponding
to the Group one of the
SIGDERS load sequences were
used. The graph indicates that the
thicker deck had higher membrane
sustained loads, as compared
to a thinner steel deck.
Other observations that are evident
from Figure 9 are as follows:
• For both deck types, the
maximum measured fastener
load is less than the
fastener pullout resistance.
For the 22-Ga deck,
the pullout was 700 lbf
(3114 N), whereas the
measured fastener force
was about 600 lbf (2669
N). This indicated that the
common practice of static
testing, i.e., dividing the
fastener pullout value by
the tributary area (fastener
spacing x fastener row
spacing) to obtain the system
resistance is not valid
under dynamic testing.
• An increase in the tensile
strength from 101 ksi (696
Mpa) to 107 ksi (738 Mpa)
increases the deck stiffness.
As such, the
responses to wind uplift
are different between the
two deck types. This is
more evident at pressure
levels above 90 psf (4.3
kPa) when the measured
fasteners loads are higher
for decks thicker deck
compared to 22 Ga.
System Optimization
There is a myth that an
increase in the wind uplift rating
can significantly increase the
overall roof system cost. Figure 10
shows a flow chart of the system
optimization process. The basic
system with deck, insulation, and
membrane can sustain a design
pressure of 90 psf and is labelled
as Roof 100. The cost of Roof 100
over 10,000 square feet is assumed
as $ X. By applying a VB to
Roof 100, the wind rating increased
by 50% and it sustained a
pressure of 135 psf (labelled as
Roof 150; i.e., 90 x 1.5 = 135). The
additional VB material cost for
10,000 square feet would be $500
plus labor. By applying the double-
side weld technique, Roof 100
increased its rating to 158 psf as
Roof 175 (90 x 1.75). The only
additional cost is labor to apply
the DSW technique instead of the
OSW technique. By enhancing
Roof 175 with a thicker steel deck,
Roof 200 sustained a pressure of
more than 180 psf (2 x 90). For
Roof 200, the additional material
cost of replacing 10,000 square
feet of 22-Ga steel decks with 20-
Ga is estimated as $1,000. By
systematically improving the
weakest links of Roof 100, Roof
200 achieved double the wind
uplift resistance. However, the
Proceeedings of the RCI 21st International Convention Baskaran and Ko – 37
Figure 8: Reduction in asymmetrical forces in DSW systems.
additional material cost is only
$1,500 plus labor. In other words,
for an increase of only 15 cents
per square foot, the wind uplift
rating can be doubled. Although a
thermoplastic system was used in
this study, the concept of weakest
link identification and the cost
benefit comparison can be equally
valid for other mechanically attached
systems.
CONCLUSION
This paper presented ways
and means to improve the wind
uplift performance of mechanically
attached systems. It also
demonstrated that experimental
data together with failure modes
investigation should be systematically
used to identify the weakest
links. This diagnostic practice can
offer design alternatives for system
enhancement. Through a
case study, this paper presented
arguments for ways to double the
wind uplift rating with minimum
increase in the cost, as long as the
weakest links were appropriately
identified.
REFERENCES
Baskaran, A., “Dynamic Wind
Uplift Performance of
Thermoplastic Roofing
System with New Seam
Technology,” ASCE, Journal
of Architectural Engineering,
Vol. 8, No. 4,
December 1, 2002.
Baskaran, A., Chen, Y. and
Vilaipornsawai, U., “A New
Dynamic Wind Load Cycle
to Evaluate Flexible Membrane
Roofs,” ASTM, Journal
of Testing and Evaluation
27(4), pp. 249 -265,
1999.
Baskaran, A. and Lei, W., “A
New Facility for Dynamic
Wind Performance Evaluation
of Roofing Systems,”
Proceedings of the
Fourth International Symposium
on Roofing Technology,
NRCA/NIST,
Washington, D.C., U.S.A.,
pp. 168 -179, 1997.
Good, C., “Surveying the Roofing
Market,” Professional
Roofing, April 2005.
“Standard Test Method for the
Dynamic Wind Uplift Resistance
of Mechanically
Attached Membrane Roofing
Systems – CSA
A123.21-04,” Canadian
Standards Association,
Toronto, 2004
Thomas, Smith, “Air Retarders
Improve Wind Performance,”
Professional Roofing,
March 1995, pp. 66.
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
who were interested in roofing
design. These include:
Manufacturers: Atlas Roofing
Corporation, Canadian General
Tower Ltd., GAF Materials Corporation,
Genflex Roofing Systems,
Firestone Building Products Co.,
IKO Industries Canada, Johns
Manville Corporation, Stevens
Roofing Systems, and Soprema
Canada
Building Owners: Canada
Post Corporation, Department of
National Defence, Public Works
and Government Services Canada.
Industry Associations: Canadian
Roofing Contractors’ Association,
Canadian Sheet Steel
Building Institute, National Roofing
Contractors Association and
Roof Consultants Institute.
Baskaran and Ko – 38 Proceeedings of the RCI 21st International Convention
Figure 9: Measured fastener resistance force
with variation in deck thickness.
Proceeedings of the RCI 21st International Convention Baskaran and Ko – 39
Figure 10: System optimization by weakest link identification.