Professional Documents
Culture Documents
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A R T I C L E I N F O A B S T R A C T
Keywords: Flexible roofing systems (e.g., membrane roofs) are widely used in low-rise commercial and industrial buildings,
Flexible membrane roof accounting for over 60% of low-sloped building roofs in North America. Despite their wide usage, the effects of
Full-scale wind tunnel testing roof flexibility are neither thoroughly studied, nor accounted for in current building design codes and standards.
Wind-induced pressure coefficients
To investigate such roof flexibility effects on wind-induced pressure, full-scale testing was conducted at the
Wall of Wind
Partial Turbulence Simulation (PTS)
NHERI Wall of Wind (WOW) Experimental Facility (EF). The mechanically attached roof system (MARS) and
partially adhered roof system (PARS), two of the most commonly used commercial roofing systems with different
flexibility characteristics, were considered in this study. “U-shaped” pressure taps were used to measure the
wind-induced pressure on the membrane roof, and the pressure coefficients were compared to those measured on
a plywood roof. The effects of wind direction and wind speed on the wind-induced pressure coefficients of
membrane roofs were studied. The results showed that the membrane deformation due to the membrane roof
flexibility resulted in a reduction in the peak pressure coefficients as compared to those of rigid plywood roof.
This reduction in wind loads is more significant for the MARS (27%) which has more flexibility than the PARS
(19%). The membrane roof flexibility also modifies the non-Gaussian characteristic which results in a lower peak
factor than the plywood roof. It was also observed that peak pressure coefficients on the flexible roof increase at
higher wind speeds. This study provides an improved understanding of the effect of roof flexibility on wind-
induced pressure coefficients. Further research is needed by testing more types and configurations of flexible
roofs to formulate new code provisions for wind effects on flexible membrane roofs.
1. Introduction North America [5]. Such roofs are covered with a flexible membrane
used as the waterproofing component. However, their relative light
Roofs of low-rise buildings are vulnerable to damage during extreme weight and flexible characteristics render them more susceptible to wind
wind events such as hurricanes and thunderstorms. Damage to roof loads. Therefore, a proper estimation of the wind-induced suctions is
components often results in a breach in the building envelope which essential to have a more robust and safer roofing system design.
subsequently creates a path for water intrusion. Such damage is mainly Over the last few decades, field investigations have been conducted
caused by the formation of vortices due to flow separation resulting in to investigate the wind effects on low-slope flexible roofing systems
high uplifts and suctions (i.e., negative pressures) on the roof surface [6–8]. Baskaran et al. [6] investigated three parameters that could in
[1]. The negative pressures, occurring at the windward edge and corners fluence the wind uplift performance of the fully-bonded single-ply
of a roof, are responsible for most of the wind-induced damages [2–4]. roofing assembly through field investigation, i.e., the curing time of the
Flexible roofing systems such as membrane roofs are widely used in low- assembly, bonding strength between the membrane and insulation
rise commercial and industrial buildings due to their competitive cost fastener plate, and insulation thickness. Bartko et al. [7] collected wind
and durability, accounting for over 60% of low-slope building roofs in speed, wind direction, and wind-induced pressure of low-slope
* Corresponding author.
E-mail address: changda.f@gmail.com (C. Feng).
https://doi.org/10.1016/j.engstruct.2021.112101
Received 21 October 2020; Received in revised form 24 January 2021; Accepted 18 February 2021
Available online 8 March 2021
0141-0296/Crown Copyright © 2021 Published by Elsevier Ltd. All rights reserved.
J. Estephan et al. Engineering Structures 235 (2021) 112101
2
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 1. 12-fan NHERI Wall of Wind Experimental Facility, Florida International University.
(a) (b)
Fig. 2. Full-scale roof model for testing. (a) Flexible roof model (MARS and PARS). (b) Rigid roof model (Plywood).
(a) (b)
Fig. 3. Pressure taps layout. (a) Flexible roof model (MARS and PARS). (b) Rigid roof model (Plywood).
3
J. Estephan et al. Engineering Structures 235 (2021) 112101
The three roof configurations were tested at 40% and 80% throttle of
the WOW full wind speed capacity, corresponding to 19 m/s (42 mph)
and 48 m/s (108 mph) mean wind speeds at a mean roof height of 1.83
m (6 ft), respectively. The wind speed data were recorded using the
Turbulent Flow Instrumentation Cobra Probes located at the mean roof
height. Also, model specimens were placed on the WOW turntable, and
tests were conducted for three wind directions: west (W), southwest
(SW), and south (S). Pressure time histories were recorded for one
minute at each tested wind direction. Table 1 summarizes the testing
protocol, where a total of 16 configurations were investigated.
The wind-induced pressure time history for each pressure tap was
collected during the experimental testing. The instantaneous pressure
coefficient CPi for each pressure tap is defined as
Pi
Cpi = 1 (1)
2
ρUmean
2
Table 1
Testing protocol.
Roof configurations Test # Mean wind speed, m/s (mph) Wind directions
4
J. Estephan et al. Engineering Structures 235 (2021) 112101
5
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 6. Pressure coefficient time history for wind from the west at 42 mph. (a) PT1. (b) PT2. (c) PT4.
Fig. 7. Membrane deformation of the MARS. (a) 40% throttle, west. (b) 100% throttle, west. (c) 40% throttle, south. (d) 100% throttle, south.
6
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 8. Mean and STD of pressure coefficients for wind from the west at 42 mph. (a) Mean. (b) STD.
those of PT2 and PT5 are very similar. a vital role in alleviating the pressure coefficient STD near the separation
The mean pressure coefficient slightly decreases on the MARS near edge. The MARS has the lowest STD, followed by the PARS, for PT1,
the separation edge (PT1, PT3, and PT4) due to membrane deformation PT3, and PT4. The reduction in terms of pressure coefficients STD is
compared to the PARS and plywood roof, with the PARS having the more than 30% and 16% for the MARS and PARS relative to the plywood
highest mean pressure coefficient for these pressure taps. The MARS and roof, respectively. This reduction is mainly due to the deformation of the
PARS have higher mean pressure coefficient magnitudes than the rigid upwind portion of the membrane, which was not observed for pressure
plywood roof when pressure taps are close to the leeward side. There taps located farther from the separation edge.
fore, the reduction of mean pressure coefficients on the flexible roof is The skewness and kurtosis are used to examine the shape of the
not observed, which is consistent with the findings of Baskaran and pressure probability distribution and determine whether it follows a
Savage [23]. For the MARS, being the most flexible roof configuration, Gaussian distribution or not. The two parameters can be used to calcu
negative mean pressure coefficients (i.e., suction) were translated into late the peak factor of the process with consideration of the non-
vertical wind-induced displacements on the flexible membrane, as seen Gaussian feature, where the peak factor is defined as (peak-mean)/
in Fig. 7. Such displacements, in their parabolic shape, can change the standard deviation. Fig. 9 illustrates the skewness and kurtosis to eval
roof aerodynamics leading to a reduction in the wind loads on the uate the non-Gaussian characteristics of the wind-induced pressure co
flexible membrane because they reduce the flow detachment, as efficients on the MARS, PARS, and plywood for the west wind direction
demonstrated in [24]. at 42 mph. Results indicate that the non-Gaussian feature is more pro
Fig. 8b portrays the STDs of the wind-induced pressure coefficients nounced for the PARS and plywood roof, especially in the reattachment
on the three roof assemblies for the west wind direction at 42 mph. The region, with the skewness and kurtosis values being higher than 0.5 and
flow separation near the separation edge (PT1, PT3, and PT4) resulted in 3, respectively. The fluctuating wind-induced pressure coefficients show
higher pressure coefficient STDs, especially for the plywood roof. It was different degrees of non-Gaussian properties depending on the PT
noted that the membrane deformation of the flexible roof assembly plays location. For instance, the plywood roof exhibits high skewness values
Fig. 9. Skewness and kurtosis of pressure coefficients for wind from the west at 42 mph. (a) Skewness. (b) Kurtosis.
7
J. Estephan et al. Engineering Structures 235 (2021) 112101
(above 0.70) in the reattachment region, while it shows the lowest turbulence intensity is IuL = 15.37% and the probability of non-
values (less than 0.34) in the flow separation region (see Fig. 9a). exceedance per subinterval is G = 1 − Ptar 1/N’ . The target probability
Moreover, the plywood roof exhibits softening non-Gaussian charac of non-exceedance Ptar = 0.78, usually used in wind tunnel studies, is
teristics (kurtosis greater than 3), as shown in Fig. 9b, which is consis adopted in this study and 1/N’ is the ratio of the time of each sub-
tent with the findings of Jiang et al. [25]. This indicates that the wind- interval over the target time interval. In the case of a 1-min time in
induced pressures on the plywood roof show softening non-Gaussian terval with N’=20, the value of 1/N’ is 0.05, yielding G = 0.0123.
properties, resulting in a higher peak factor than the Gaussian peak Fig. 10b illustrates the corrected peak pressure coefficients for the west
factor [26]. In the case of MARS, the membrane deformation mitigates wind direction at 42 mph. It should be noted that the PTS correction
the softening non-Gaussian properties, with the kurtosis being around 3 does not affect the overall peak pressure coefficient distribution.
at all PT locations. This would result in lower peak pressure coefficients To further investigate the effect of roof flexibility on wind-induced
on the MARS compared to those on the PARS and plywood. It should be pressure coefficients, a spectral analysis was performed on the MARS,
noted that the peak factors depend on the sampling length of the process. PARS, and plywood for wind from the west at 42 mph. Fig. 11 portrays
Rizzo et al. [27] investigated the time-dependence feature of the peak the PSDs of pressure fluctuations at PT1, PT2, and PT4. At lower fre
factor for a hyperbolic paraboloid roof by using different sampling quency, say the non-dimensional frequency (fH/U) less than 0.2, the
lengths. energy in the PSD is lower for the case of plywood compared with that of
The peak pressure coefficients were estimated using extreme value MARS and PARS. This indicates that the membrane deformation and
analysis. The pressure time histories were divided into 20 independent fluttering of MARS and PARS in the low-frequency range may generate
subintervals and the peaks were estimated using the Lieblein’s best an additional low-frequency vortex which increases the energy of the
linear unbiased estimation (BLUE) method [18]. Fig. 10a summarizes wind-induced pressure on MARS and PARS. On the other hand, at a
the estimated peak pressure coefficients on the MARS, PARS, and higher frequency where the non-dimensional frequency (fH/U) is
plywood for the west wind direction at 42 mph. The symmetric property greater than 0.2, the PSD of plywood is higher than that of MARS and
in terms of peak pressure coefficients is well observed on the three roof PARS. This indicates the wind-induced pressure on the plywood roof has
configurations. The magnitude of the peak pressure coefficients is the larger components vibration at higher frequencies, which leads to a
highest at the separation edge and decreases significantly when the higher crossing rate at a certain threshold and consequently, a larger
pressure taps are located close to the leeward edge. It can be seen that peak pressure coefficient. This also partially explains the higher peak
the roof flexibility reduces the peak pressure coefficient magnitude pressure coefficient on the plywood roof, as shown in Fig. 10. Fig. 11
when pressure taps are located near the separation edge (PT1, PT3, and demonstrates the effect of roof flexibility on wind-induced fluctuating
PT4). More specifically, the reductions in terms of the estimated peak pressure coefficients, where the membrane deformation mainly miti
pressure coefficients of PT1, PT3, and PT4 relative to the plywood are gates the wind loads in the high-frequency range. The difference in the
more than 22% and 14% for the MARS and PARS, respectively. Although wind-induced pressures on flexible and rigid roofs can be translated into
the mean pressure coefficients of PT1, PT3, and PT4 are slightly higher differences in the wind uplift resistance. These observations are in line
on the PARS compared to the plywood roof, the reduction in STDs of with those found in [28].
pressure coefficients on the PARS has a larger contribution to the
reduction of peak pressure coefficients. In the case of MARS, the
reduction of the mean and STD of pressure coefficients near the sepa 3.3. Effect of wind direction on the wind-induced pressure coefficients
ration edge leads to the reduction of the peak pressure coefficients.
Moreover, the Gaussian characteristics on the MARS reduce the peak To better understand the wind effects on flexible and rigid roof as
factor and consequently reduce the peak pressure coefficients. semblies, south and southwest wind directions were considered. It
To account for the missing low-frequency turbulence in the WOW should be noted that the membrane seams are parallel to the wind flow
simulations, the estimated peak pressure coefficients were corrected in the south wind direction while being perpendicular to the wind flow
using the PTS approach. The pressure time histories were divided into 20 in the west direction. Therefore, the pressure coefficient distribution for
independent subintervals and the peaks were fitted using Type I extreme the MARS and PARS for the south wind direction may be different from
value distribution. In the PTS approach, the missing low-frequency the corresponding distribution for the west direction. The estimated and
corrected peak pressure coefficients, recorded on the MARS, PARS, and
Fig. 10. Peak pressure coefficients for wind from the west at 42 mph. (a) Estimated. (b) Corrected.
8
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 11. PSD for wind from the west at 42 mph. (a) PT1. (b) PT2. (c) PT4.
Fig. 12. Peak pressure coefficients for wind from the south at 42 mph. (a) Estimated. (b) Corrected.
9
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 13. Peak pressures for wind from the southwest at 42 mph.
plywood, are illustrated in Fig. 12 pertaining to the south wind direction future studies.
at 42 mph. The MARS experienced the lowest peak pressure coefficients, To further evaluate the effect of cornering wind on pressures expe
with a reduction of more than 31% relative to the plywood. Although rienced by the plywood roof, 113 pressure taps were installed, as shown
the deformation patterns of the flexible roof are not similar for the west in Fig. 3b. Fig. 14 shows a contour plot of the corrected peak pressure
and south wind directions, the pressure coefficient reduction on the coefficients Cp,peak on the plywood roof for cornering wind. It can be seen
MARS and PARS, in the flow separation region, shows a similarity be that there is a significant drop in the Cp,peak values along the roof diagonal
tween the two wind directions. (PT1, PT4, and PT5). The highest Cp,peak values (− 2.1) were detected near
Fig. 13 shows the estimated and corrected peak pressure coefficients the roof edge due to the formation of conical vortices.
on the different roof configurations for the southwest wind direction at The highest estimated and corrected peak pressure coefficients (most
42 mph. The symmetric property along the roof diagonal was observed. critical) among the west, south, and southwest directions for the MARS,
The peak pressure coefficient magnitude decreased along the roof di PARS, and plywood are displayed in Fig. 15. It is clear that the flexible
agonal as pressure taps are located further away from the separation roof helps in the reduction of peak pressure coefficient magnitude,
corner. It was observed that the peak pressure coefficients on the MARS especially for the pressure taps near the edge. For instance, the MARS
and PARS are more than 60% higher than those on the plywood, espe and PARS reduce more than 27% and 19% of the peak pressures,
cially at PT1, PT4, and PT5. It was noted that a cornering wind can have respectively, compared to the plywood roof at PT1, PT2, and PT3. For
a less significant effect on the plywood at the roof diagonal, while the PT5, it is expected to experience similar peak pressure coefficients to
opposite can be seen on the MARS and PARS. Therefore, the roof flexi PT1, PT2, or PT3 when the wind is approaching from the east, north, or
bility does not help in the reduction of peak pressure coefficients for the northeast directions.
southwest direction. This may be due to the complex interaction be
tween the ballooning effect from the flexible roof and the conical
vortices from cornering winds, which requires further investigations in 3.4. Effect of wind speed on the wind-induced pressure coefficients of
flexible roof
10
J. Estephan et al. Engineering Structures 235 (2021) 112101
Fig. 15. Highest peak pressure coefficients (in magnitude) for all wind directions at 42 mph. (a) Estimated. (b) Corrected.
(c)
Fig. 16. Statistics of pressure coefficients on the MARS for wind from the west at 42 mph and 108 mph. (a) Mean. (b) STD. (c) Peak.
11
J. Estephan et al. Engineering Structures 235 (2021) 112101
12
J. Estephan et al. Engineering Structures 235 (2021) 112101
[19] ESDU. Characteristics of the Atmospheric Boundary Layer, Part II: Single Point [24] Rizzo F, D’Asdia P, Ricciardelli F, Bartoli G. Characterisation of pressure
Data for Strong Winds (Neutral Atmosphere). London: Engineering Sciences Data coefficients on hyperbolic paraboloid roofs. J Wind Eng Ind Aerodyn 2012;102:
Unit; 2001. 61–71. https://doi.org/10.1016/j.jweia.2012.01.003.
[20] Richards PJ, Hoxey RP. Pressures on a cubic building—Part 1: Full-scale results. [25] Jiang L, Li J-H, Li C. Comparative Study on Non-Gaussian Characteristics of Wind
J Wind Eng Ind Aerodyn 2012;102:72–86. https://doi.org/10.1016/j. Pressure for Rigid and Flexible Structures 2018. https://doi.org/10.1155/2018/
jweia.2011.11.004. 9213503.
[21] Levitan ML, Mehta KC. Texas Tech field experiments for wind loads part 1: building [26] Ding J, Chen X. Assessment of methods for extreme value analysis of non-Gaussian
and pressure measuring system. J Wind Eng Ind Aerodyn 1992;43:1565–76. wind effects with short-term time history samples. Eng Struct 2014;80:75–88.
https://doi.org/10.1016/0167-6105(92)90372-H. https://doi.org/10.1016/j.engstruct.2014.08.041.
[22] Levitan ML, Mehta KC. Texas tech field experiments for wind loads part II: [27] Rizzo F, Barbato M, Sepe V. Peak factor statistics of wind effects for hyperbolic
meteorological instrumentation and terrain parameters. J Wind Eng Ind Aerodyn paraboloid roofs. Eng Struct 2018;173:313–30. https://doi.org/10.1016/j.
1992;43:1577–88. https://doi.org/10.1016/0167-6105(92)90373-I. engstruct.2018.06.106.
[23] Baskaran A, Savage MG. Wind pressure measurements on full scale flat roofs. [28] Baskaran A, Murty B, Prevatt D. Evaluating wind effects of commercial roofs: North
NRCC- 46291, National Research Council Canada; 2003. American advancements, Amsterdam, Netherlands: 2011.
13