You are on page 1of 8

The Seventh Asia-Pacific Conference on

Wind Engineering, November 8-12, 2009,


Taipei, Taiwan

WIND FORCE COEFFICIENTS FOR THE DESIGN OF A


HYPERBOLIC PARABOLOID FREE ROOF
Yasushi Uematsu1, Fumio Arakatsu2, Satoru Matsumoto3 and Fumiyoshi Takeda4
1
Professor, Department of Architecture and Building Science, Tohoku University,
Sendai 980-8579, Japan, yu@venus.str.archi.tohoku.ac.jp
2
Chiyoda Corporation, Yokohama 230-8601, Japan
3
Graduate Student, Department of Architecture, The University of Tokyo, Tokyo 113-8654,
Japan, matsumoto@rcs.arch.t.u-tokyo.ac.jp
4
Structural Engineer, Taiyo Kogyo Corp., Osaka 532-0012, Japan,
tf001163@mb.taiyokogyo.co.jp
ABSTRACT
The present paper discusses the wind force coefficients for designing the main wind force resisting system of a
hyperbolic paraboloid (H.P.) free roof, based on a wind tunnel experiment. In the experiment, the lift coefficient
(CL) and the aerodynamic moment coefficients (CMx and CMy) about two horizontal axes (x and y) were measured
by using a force balance that had been designed, built and gauged for the purpose of the present study.
Correlations between wind force and moment coefficients are investigated, and appropriate wind force
coefficients for the design of main wind force resisting systems are proposed for two wind directions
perpendicular to each other. The roof is assumed rigid and simply supported by four corner columns, whose
axial forces are regarded as the most important load effect for discussing the design wind loads. Two load cases
generating the maximum tension and compression of columns are considered. The roof is divided into the
windward and leeward halves for each wind direction, and the wind force coefficients on these halves are
provided. The proposed values are compared with the specifications of the Australia/New Zealand Standard
from the viewpoint of load effect.
KEYWORDS: HYPERBOLIC PARABOLOID FREE ROOF, WIND LOADS, MAIN WIND FORCE
RESISTING SYSTEM, WIND TUNNEL EXPERIMENT, DESIGN WIND LOAD, CODIFICATION

Introduction
Hyperbolic paraboloid (H.P.) free roofs are widely used for structures providing shade
and weather protection in public spaces, such as sports grounds and shopping areas. Pre-
tensioned membrane is often used for these roofs. The roofs are consequently very
lightweight structures that are vulnerable to wind loading. In practice, such roofs often
experience damage during windstorms. Therefore, the wind resistance is one of the greatest
concerns for structural engineers when designing these roofs. However, few specifications of
design wind loads on H.P. free roofs are provided in codes and standards. Although the
Australia/New Zealand (AS/NZ) Standard (2002) provides the wind force coefficients, the
range of roof shape for which the wind force coefficients are specified is rather limited.
Since free roofs are usually supported by columns and no walls, wind action is directly
exerted both on the top and bottom surfaces. Consequently, the wind flow around roofs is
rather complex. Furthermore, it is very difficult to make wind tunnel models of such roofs,
compared with enclosed structures. This may be one of the reasons why only a few studies
have been made on the wind loads on free roofs. Recently, Uematsu et al. (2007) proposed
the design wind force coefficients, both for the main wind force resistant systems and for the
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

cladding and components (C & C), based on a series of wind tunnel experiments. They made
simultaneous pressure measurements for gable, troughed and mono-sloped roofs in a turbulent
boundary layer. They focused on the axial forces induced in the columns as the load effect
for discussing the design wind loads, assuming that the roof was rigid and simply supported
by four corner columns.
In addition to the fundamental shapes that Uematsu et al. (2007) tested, H.P. roofs are
often used for membrane structures, as mentioned above. However, few studies have been
made on such curved free roofs. Beutler (1963) measured the mean wind pressures on H.P.
free roofs. Simultaneous pressure measurements at many points are quite difficult for H.P.
free roofs because of difficulties in making wind tunnel models. Thus, a force balance
technique is employed in the present study. The authors designed and built a force balance
for measuring the overall aerodynamic forces and moments acting on free roofs with a
reference to Altman (2001). Based on the results on these aerodynamic forces and moments,
wind force coefficients for the design of main wind force resisting systems are proposed. The
proposed values are compared with the specifications of the Australia/New Zealand Standard.
The roof deflects and oscillates under wind actions, which may affect the wind loads
on the roof due to flow-structure interaction. However, as the first step of study, this effect is
neglected, that is, the roof is assumed rigid in order to make the problem simple in this study.
Experimental Arrangement and Procedure
The experiments were carried out in the boundary layer wind tunnel with a working
section 1.4 m wide, 1.0 m high and 6.5 m long at the Department of Architecture and
Building Science, Tohoku University. A turbulent boundary layer with a power law exponent
of α = 0.18 for the mean velocity profile was generated on the wind tunnel floor. The
turbulence intensity Iu and longitudinal scale Lx of the flow at a height of z = 100 mm are 0.17
and 0.16 m, respectively. The power spectrum of wind velocity fluctuations was consistent
with the so-called Karman-type spectrum, and the turbulence scale Lx was obtained by fitting
the curve of the spectrum to the experimental data.
Three models (Models A, B and C) with different rise/span (or sag/span) ratios were
tested (Figure 1(a)). The models are made of 1 mm thick acrylic plate, and the surface is
smooth (see Figure 1(b)). The dimensions of the models and the notations used in the present
paper are shown in Figure 1(c). The mean roof height H is 80 mm in any case. The roof
model is supported by three aluminum columns 5mm in diameter on a Y-shaped force balance
made of 1.2 mm thick phosphor bronze. The column base is pin-jointed to the leaf spring.
The bending stress at the base of each leaf spring is measured by strain gauges, from which
the concentrated load at the end of each arm is computed. The lift L and the aerodynamic
moments Mx and My about the x and y axes are computed from the concentrated loads N1 to N3.
x
(Low)
h = a/2 75

Model A
b1
(High) (High)
h = a/3
y
50 b2 θ = 90o
Model B
b3
h = a/6 25
Model C (150mm)
(Low)
212.1 θ = 0o

(a) Model shape (b) Picture of Model A (c) Dimensions and notations
Figure 1: Experimental model and a force balance used in the present study
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

The force balance provides the moments Mx and My about the x and y axes, the origin
of which is located on the plane of Y-shaped leaf springs, not at the center of the roof surface.
Therefore, the computed Mx and My values include additional moments induced by the drag.
This effect can not be compensated, because we can not measure the drag with this force
balance. Hence, based on the results of simultaneous pressure measurements on troughed
roofs made by Uematsu et al. (2007), we estimated the effect of drag on the aerodynamic
moments as well as on the resultant load effects, or the axial forces in the columns supporting
the roof. Note that the roof is assumed to be supported by four corner columns and the axial
forces in the columns are regarded as the load effect for discussing the design wind loads.
The results indicate that the eccentricity of the origin overestimates the maximum load effect
by up to approximately 11% for Model A, 8% for Model B and 3 % for Model C. Because
the drag on an H.P. roof is smaller than that on a troughed roof with the same roof pitch, the
effect of the eccentricity may be smaller for H.P. roofs than for troughed roofs.
The measurements were carried out at a wind speed of UH ≈ 6 m/s at z = H. The
design wind speed is assumed 31.5 m/s, as a typical value of strong wind events. The
geometric scale of the model is assumed 1/100, which yields a time scale of approximately
1/19. The outputs of the strain meters were sampled simultaneously at a rate of 200 Hz for a
period of 32 sec, which corresponds to 10 min in full scale. The measurements were repeated
six times under the same conditions. The statistics of the aerodynamic coefficients are
evaluated by applying ensemble average to the results of these six runs. In the present paper,
focus is on the mean peak values of the aerodynamic coefficients and the load effects.
The Reynolds number Re, defined in terms of UH and the horizontal width of the roof
a, is approximately 6 × 104. Of great concern is the Reynolds number effect on the
aerodynamics, because the H.P. free roof consists of curved surface. In practice, however, the
effect seems small. The rise/span ratio of the models tested in the present study is less than
0.18. For such a small rise/span ratio, it is thought that the wind flows along the upper
surface of the roof when θ = 0o. This is inferred from the measurements of wind pressure
distribution on a circular arc roof with a rise/span ratio of 0.2 by Hamai et al. (2009), who
showed that the wind flows along the roof and the separation point is close to the leeward
edge. This is the case for spherical domes. Hongo (1995) investigated the wind pressure
distributions on spherical domes. The results of flow visualization together with those of
pressure measurements clearly indicated that the flow separation point was close to the
leeward edge when the rise/span ratio was 0.2.
Experimental Results

Verification of the Force Balance


The accuracy of measurements by the Y-shaped force balance is first investigated by
using a mono-sloped roof model. The results for the lift and aerodynamic moments measured
by the force balance are compared with the previous results that Uematsu et al. (2007)
obtained from simultaneous pressure measurements. Figure 2 shows comparisons for the lift
coefficient CL and the moment coefficient CMy on a mono-sloped roof with a pitch of 15o.
The closed and open symbols represent the present and previous results, respectively. There
is some difference between these two results, which may be due to the differences in roof
supporting system as well as in the experimental condition, such as the wind tunnel flow. The
effect of drag on CMy may be another reason for the difference, as mentioned above.
However, it can be said that the agreement is generally good, considering the accuracy of
measurements in these two experiments. This implies that the Y-shaped force balance
provides reasonable results for the overall wind force and moments acting on free roofs.
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

1.0 0.3
Maximum peak Maximum peak
0.2
0.5
Mean CL
0.1
Mean

CMy
0.0
CL

0.0
-0.5 Minimum peak
-0.1 CMy
Minimum peak
-1.0 -0.2
0 45 90 135 180 0 45 90 135 180 θ = 0o θ = 180o
w ind direction θ (deg) wind direction θ (deg)

(a) CL (b) CMy (c) Notation


Figure 2: Aerodynamics of a mono-sloped roof (roof pitch = 15o)

Characteristics of wind force and moment coefficients


Figure 3 shows the statistical values of the lift and moment coefficients as a function
of wind direction θ for Model A (h/a = 1/2); the mean and the maximum and minimum peak
values are plotted in each figure. The lift coefficient becomes the maximum (upward) when θ
≈ 0 and the minimum (downward) when θ ≈ 90 . This feature is related to the increased
o o

wind velocity along the convex surface, that is, the upper surface when θ ≈ 0o and the lower
surface when θ ≈ 90o. The magnitude of the negative peak value of CMx becomes the
maximum when θ ≈ 90o, while that of CMy becomes the maximum when θ ≈ 0o. The values
of CMx when θ ≈ 0o and the those of CMy when θ ≈ 90o are relatively small in magnitude. The
variation of CMx with θ is opposite to that of CMy.

1.0 0.25 0.10


Maximum
0.5 Maximum
Maximum
0.00 0.00
0.0
CMy

Mean
CMx
CL

Mean
-0.5 Mean
-0.25 -0.10
-1.0 Minimum Minimum
Minimum
-1.5 -0.50 -0.20
0 30 60 90 0 30 60 90 0 30 60 90
θ (deg) θ (deg) θ (deg)

(a) CL (b) CMx (c) CMy


Figure 3: Statistics of lift and moment coefficients (Model A: h/a = 1/2)

Load effects
The axial force N induced in each column is computed from the time history of CL,
CMx and CMy, assuming that the roof is rigid and simply supported by four corner columns. In
the present study, focus is on the axial forces induced in the columns as the load effect for
discussing the design wind loads in the following section.
The maximum and minimum peak values of the non-dimensional axial force N* (=
2
N/(qHa /4)) among the four columns are plotted in Figure 4. The values of the maximum and
minimum N* values generally decrease with a decrease in h/a. The variation of the maximum
tension with wind direction is relatively small in magnitude, while that of the maximum
compression is significant. The maximum tension and compression among the all wind
directions are induced when θ ≈ 30o and θ ≈ 90o, respectively.
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

1.5 1.5 1.0


Max. (tension) Max. (tension) Max. (tension)
1.0 1.0
0.5 0.5
0.5
0.0
0.0
N*

N*
0.0

N*
-0.5
-1.0 -0.5
-0.5
-1.5 -1.0
Min. (compression) Min. (compression) Min. (Compression)
-2.0 -1.5 -1.0
0 30 60 90 0 30 60 90 0 30 60 90
θ (deg) θ (deg) θ (deg)

(a)Model A (h/a = 1/2) (b)Model B (h/a = 1/3) (c)Model B (h/a = 1/6)


Figure 4: Non-dimensional axial force

In order to investigate the dynamic effect of wind turbulence on the column axial
forces, the gust effect factor Gf, defined as the ratio of the maximum or the minimum axial
force to the mean value induced in the column, is computed for all models and wind
directions. Figure 5 shows the results for Gf plotted against the mean reduced axial force
N*mean. When the value of |N*mean| is small, Gf exhibits a large value. However, as |N*mean|
increases, the values of Gf collapse into a narrow range around Gf = 2.0, which corresponds to
a peak factor of gf ≈ 2.5, based on the quasi-steady assumption, i.e., Gf ≈ (1 + 2.5 × 0.17)2.
The value of gf ≈ 2.5 is somewhat smaller than that for gable, troughed and mono-sloped
roofs, which is approximately 3.0 (see Uematsu et al. (2007)). This difference may be due to
the effect of flow separation from the leading edges of the roof on the wind loads. The
turbulence induced by the flow separation seems lower for H.P. roofs than that for the other
roofs. In the following section, the value of Gf = 2.0 is used for evaluating the design wind
force coefficients that provide the equivalent static loads together with the gust effect factor
Gf.

10
Model A
8 Model B
Model C
6
Gf

0
0 0.2 0.4 0.6 0.8 1

|N*mean|

Figure 5: Gust effect factor based on the load effect

Discussion of the Design Wind Force Coefficients

Proposal of the Design Wind Force Coefficients


The roof is divided into two areas, i.e. the windward and leeward halves, and the
design wind force coefficients CNW and CNL for these halves are specified, as was done in the
previous study (Uematsu et al. (2007)). The procedure was successfully applied to the
evaluation of design wind force coefficients for gable, troughed and mono-sloped roofs. The
design wind force coefficients are assumed constant in these areas. They are estimated
according to the following procedure:
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

Step 1: The basic values of CNW and CNL, denoted as CNW0 and CNL0, are determined from a
combination of the lift coefficient (CL) and the moment coefficient (CMy or CMx) that
produces the maximum load effect for θ = 0o or 90o.
Step 2: Considering that the axial force induced in the column may become the maximum for
oblique winds (see Figure 4), a correction factor γ, which is defined as the ratio of the
actual peak force for θ = 0o − 45o (W.D. 1) or θ = 45o − 90o (W.D. 2) to that
computed from the CNW0 and CNL0 values, is introduced.
Step 3: The design wind force coefficients C*NW and C*NL, which give equivalent static wind
loads, are provided as follows:

γC NW 0 γC NL 0
C * NW = , C * NL = (1a), (1b)
Gf Gf

Figures 6(a) and 6(b) show a phase-plane representation of the CMy−CL relation when
θ = 0 . The circles in the figure represent the maximum and minimum peak values (CLmax and
o

CLmin) during each of six runs. The envelope of the trajectory looks like an ellipse with
inclined axes, indicating a positive correlation between CMy and CL. In the case of Model C,
the CL and CMy values are well correlated with each other. The value of CMy at the instant
when CLmax or CLmin occurs is nearly equal to the maximum or minimum value of CMy (CMymax
or CMymin). The maximum load effect may be given by a combination of the two peak values.
In the case of Model B, on the other hand, the correlation between CMy and CL is relatively
low. The ‘peak + peak’ combination of CMy and CL does not necessarily give the maximum
load effect. The maximum load effect is given by a certain combination of CMy and CL. The
envelope of CMy−CL trajectory for Models A and B is approximated by a hexagon shown in
Figure 6(c). The critical condition that produces the maximum load effect may be given by
one of these six apexes. The CMx−CL relation for θ = 90o exhibits a similar feature.
From the combination of the lift and moment coefficient (CMx or CMy) obtained above,
the basic wind force coefficients (CNW0 and CNL0) are computed for two wind directions θ = 0o
and 90o. For each wind direction, two sets of the CNW0 and CNL0 values are obtained for two
load cases (A and B) that induce the maximum tension and compression in the columns.
CMy
0.05 0.10
h/a=1/3 h/a=1/6
CMy
CMy

 2  1
0.00 0.05 CMymax
-0.5 0.0 0.5 CL 1.0
-0.05 0.00 5 CMymean
-0.5 0.0 CL 0.5 3
-0.10 -0.05
CL
-0.15 -0.10
CMymin
4 6
-0.20 -0.15 CLmin CLmean CLmax

(a) CMy −CL trajectory for (b) CMy −CL trajectory for (c) Model of the envelope
Model B (h/a = 1/3) Model C (h/a = 1/6) of CMy−CL trajectory
Figure 6: Phase-plane representation of the CL−CMy relation (θ = 0o)

The correction factor γ in Eq. (1) is obtained by calculating the ratio of the actual
maximum or minimum axial force to the predicted value from CNW0 and CNL0. Figure 7
shows the results. When the h/a ratio is small, the value of γ for W.D. 1 is relatively large.
Besides these cases, the value of γ is approximately 1.0. Similar features were observed for
gable and troughed roofs (Uematsu et al. (2007)).
The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

2.0

Correction factor γ
1.5

1.0
Load case A (W.D. 1)
Load Case B (W.D. 1)
0.5
Load Case A (W.D. 2)
Load Case B (W.D. 2)
0.0
0 0.2 0.4 0.6
h/a

Figure 7: Correction factor γ

Comparison with the Specifications of the Australia/New Zealand Standard


Plotted on Figures 8 and 9 are the estimated values of C*NW and C*NL for W.D. 1 and
W.D. 2, respectively. The specifications of the AS/NZ Standard (2002) are also shown by the
dashed lines. The Standard provides two values of wind force coefficients (expressed as
‘positive’ and ‘negative’) for each of the windward and leeward halves; the range of the h/a
ratio is limited to 0.1 to 0.3. The values of C*NW for load cases A and B are close to the two
specified values of the AS/NZ Standard. Regarding the leeward half, on the other hand, the
wind force coefficients for the two load cases are similar to each other and nearly equal to one
of the specified values of the AS/NZ Standard. These features are similar to those observed
by Uematsu et al. (2007) for gable, troughed and mono-sloped roofs.

1.0 1.0
Load case A
AS/NZ (positive)
0.5 0.5
AS/NZ (positive) Load case B
C*NW

C*NL

0.0 0.0

-0.5 Load case A -0.5 AS/NZ (negative)


AS/NZ (negative)
Load case B
-1.0 -1.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
h/a h/a

(a) Windward half (b) Leeward half

Figure 8: Wind force coefficients C*NW and C*NL (W.D. 1)


1.0 1.0
Load case A
AS/NZ (positive)
0.5 Load case B 0.5

AS/NZ (positive)
C*NW

C*NL

0.0 0.0
AS/NZ (negative)
-0.5 -0.5 loadcaseA
AS/NZ (negative)
loadcaseB
-1.0 -1.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
h/a h/a

(a) Windward half (b) Leeward half

Figure 9: Wind force coefficients C*NW and C*NL (W.D. 2)


The Seventh Asia-Pacific Conference on Wind Engineering, November 8-12, 2009, Taipei, Taiwan

Axial forces induced in the columns are computed by using C*NW and C*NL and
compared with those predicted from the AS/NZ specifications. The results are shown in
Figure 10. As mentioned above, the AS/NZ Standard generally provides four combinations
of the wind force coefficients on the windward and leeward halves. The maximum and
minimum axial forces among the four values are shown in the figure. It is interesting to note
that these values are consistent with the present results for load cases A and B, respectively,
despite the difference in the wind force coefficients, as shown in Figures 8 and 9.

1.0 1
Load case A Load case A
Load case B Load case B
0.5 0.5
AS/NZ AS/NZ
(Max.) (Max.)
N*/Gf

N*/Gf
0.0 0

-0.5 AS/NZ (Min.) -0.5 AS/NZ (Min.)

-1.0 -1
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
h/a h/a

(a) Wind direction 1 (b) Wind direction 2


Figure 10: Reduced axial force

Concluding rrmarks
Overall wind forces and moments acting on an H.P. free roof have been investigated
experimentally. A Y-shaped force balance was designed, built and gauged for the purpose of
the present study. Based on a combination of the lift and moment coefficients, the design
wind force coefficients, expressed as equivalent static loads, on the windward and leeward
roof halves have been proposed for two load cases that induce the maximum tension and
compression in the columns. The proposed values are compared with the specifications of the
AS/NZ Standard. It was found that both the proposed values and the AS/NZ specifications
predict similar results on the axial forces within the limits of h/a = 0.1 − 0.3, for which the
AS/NZ Standard provides the specifications of the wind force coefficients.
References

Altman, D.R. (2001), Wind forces on roof canopies, M. Sc. Thesis, Clemson University, USA.
Beutler, J. (1963), “Beitrag zur statischen windbelastung von seilnetzwerken - ergebnisse von
windkanaluntersuchungen”, Hanging Roofs, Proceedings of the IASS Colloquium on Hanging Roofs,
Continuous Metallic Shell Roofs and Superficial Lattice Roofs, Paris 9-11 July 1962, North-Holland
Publishing Company, 76-86.
Hamai, M., Uematsu, Y. and Tanaka, S. (2009), “Unsteady aerodynamic forces on a circular arc roof”,
Summaries of Technical Papers of Annual Meeting Architectural Institute of Japan, Structure I. (to be
published)
Hongo, T. (1995), Experimental study of wind forces on spherical roofs, Ph.D. Thesis, Tohoku University. (in
Japanese)
Standards Australia (2002), Australia/New Zealand Standard, AS/NZ 1170.2.
Uematsu, Y., Iizumi, E.. and Stathopoulos, T. (2007), “Wind force coefficients for designing free-standing
canopy roofs”, Journal of Wind Engineering and Industrial Aerodynamics, 95, 1486-1510.

You might also like