Engineering Structures 45 (2012) 536558

Contents lists available at SciVerse ScienceDirect

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

Wind tunnel tests on hyperbolic paraboloid roofs with elliptical plane shapes
Fabio Rizzo
Dept. of Engineering/CRIACIV (Inter-University Centre for Building Aerodynamics and Wind Engineering), Italy
University of Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy

a r t i c l e i n f o a b s t r a c t

Article history: No parametric data exists in the main international standards to calculate wind action on buildings with
Received 13 February 2012 hyperbolic paraboloid roofs. This type of roof, however, is particularly efcient in covering medium to
Revised 11 June 2012 large spans and is highly competitive compared to traditional structures. Our research aims at parame-
Accepted 27 June 2012
terization of pressure coefcients on roofs with hyperbolic paraboloid shapes with four different foot-
Available online 18 August 2012
prints, respectively, square, rectangular, circular and elliptical.
This paper studies the elliptical shape in particular. Aerodynamic wind tunnel tests have been per-
Keywords:
formed on these shapes with the object of calculating pressure coefcients. A comparison is also made
Wind tunnel testing
Pressure coefcients
between elliptical and circular shapes in order to demonstrate the high efciency of the elliptical shape.
Hyperbolic paraboloid roofs 2012 Elsevier Ltd. All rights reserved.
Elliptical shape

1. Introduction conguration and it is impossible to use pressure coefcient data

Results for square, rectangular and circular shapes have been
presented in previous papers [20,21,19], illustrating research to
parameterize pressure coefcients on hyperbolic paraboloid roofs
with medium and large spans (hereinafter HYP). These papers de-
scribed wind tunnel tests conducted to measure pressure coef-
cients. Interesting results were obtained by comparing data
between different geometric parameters: different heights, roof
surface curvatures and footprints. Research aims at answering
the need to have reference data for hyperbolic parabolid roofs since
no parametric data about this type of geometric shape is to be
found in national and international standards [12,1417]. This
deciency is a source of difculty for designers who want to per-
form preliminary dimensioning of a hyperbolic parabolid roof, cal-
culating, for example, cables nets, a commonly used structure to
cover large spans with small thicknesses. This type of structure
has very high performance characteristics and, due to the low
structural weight, is particularly suited for buildings built in seis-
mic areas. Wind tunnel tests have been done using models of ellip-
tical shape HYP roofs to complete this study of the main, most
popular and most exible shapes for buildings that need free large
spans without intermediate supports. This shape is used to design
sports arenas and indoor swimming pools but could also be ideal
for concert and conference spaces. This footprint, combined with
a hyperbolic paraboloid roof, gives a very complex and special
Address: University of Chieti-Pescara, Viale Pindaro 42, 65127 Pescara, Italy.
Tel.: +39 3200384186.
E-mail address: fabiorizzo79@libero.it
published in technical codes for at, dome or inclined roofs to
make a preliminary evaluation of the wind action. The object of
these tests is to compare data evaluated using the elliptical model
with those evaluated on circular models. Roof deections and
spans, when determining the geometric shape of the model, have
been chosen to equal the values reported in Rizzo et al. [20].
Calculation used a preliminary numerical design procedure
described in Section 2.1 below.
It is to be hoped that the parametric data resulting from this re-
search will be incorporated in technical codes to be used by design-
ers for preliminary design of wind action on HYP roofs before they
carry out specic aerodynamic tests. This preliminary step is, in
fact, very important for choosing one type of structure rather than
another and, in particular, for selecting a tension structure rather
than a traditional structure.
2. Wind tunnel tests on elliptical footprint HYP roofs
2.1. Experimental setup
2.1.1. Geometric parameters
Lacking unlimited resources to test all possible geometric shape
combinations of parabolid hyperbolic roofs it was decided, in order
to generate a sufciently representative sample, to examine a total
of four shapes, respectively, square, rectangular, circular and ellip-
tical. Results for the rst three shapes (Fig. 1ac) were described in
a previous paper [20]. The last, elliptical shape, (Fig. 1d) is the sub-
ject of this work. Simplied geometric shapes are illustrated in
Fig. 1. The search for the representative sample of roofs to be tested

0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.06.049
 F. Rizzo / Engineering Structures 45 (2012) 536558 537

Fig. 1. Square plane (a), rectangular plane (b), circular plane (c) [20], ellipltical plane (d) shape hyperbolic parabolid roofs.

Fig. 2. Geometrical parameters analysed.

Fig. 3. Model instrumentation.

in the wind tunnel concentrated on sag-span ratios in order to nd
load bearing and stabilizing cables that optimize structural behav-
ior under maximum snow load and wind load values. The prere-
quisite for this kind of roof, under maximum snow load values, is
to avoid excessively high stress values in the load-bearing cables
(with the concave side up) and excessively low stress values in
stabilizing cables (with concavity downwards). If stabilizing cables
are not correctly designed they may be unloaded and reverse their
concavities, causing the system to be unstable. The system reverses
its hierarchy with wind load suction: in this case it is the load

Table 1
Geometrical parameters of test models.

Test () L1/L2 () H () f1/f2 () f1 (cm) f2 (cm) L1 (cm) L2 (cm) H (cm) Hb (cm) Ze (cm)
p.14 0.5 1/10 Lmax 2.0 2.67 5.33 40 80 8.00 13.33 (1/6 Lmax) 21.33
p.15 0.5 1/10 Lmax 2.0 2.67 5.33 40 80 8.00 26.66 (1/3 Lmax) 34.66
p.16 0.5 1/6 Lmax 2.0 4.44 8.89 40 80 13.33 13.33 (1/6 Lmax) 26.66
p.17 0.5 1/6 Lmax 2.0 4.44 8.89 40 80 13.33 26.66 (1/3 Lmax) 40.00
538 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 4. Pressure taps distribution, roof (a), p.14 lateral surface (b), p.15 (c), p.16 (d), p.17 (e).

bearing cables which may reverse their curvature. To achieve this
result it was necessary to perform a simplied calculation of sev-
eral geometric shapes, simulating the deformed shape under load
and checking cable stress for each load combination.
The procedure of preliminary design use to investigate the
geometries, in order to chose a simple of optimal geometries, is
based on the classical theory of cable beam and its purpose is to
derive from a three-dimensional surface a two-dimensional static
model that simplies and speeds up the calculus (The rope beam).
It is possible to use this simplied model because the hyperbolic
paraboloid function used has a bundle of parabolas in two orthog-
onal directions. The middle cables are parallel to the other cables of
net and their curvatures are equal to the one of other cables of net.
The response given by the procedure is structurally approximate
 F. Rizzo / Engineering Structures 45 (2012) 536558 539

6.88

F
E

1.88
D
C

22.09
2.40 x 1.60
B

11.00
direzione del flusso
A
(a) (b)
Fig. 5. CRIACIV wind tunnel image (a), elements (b): (A) entrance; (B) boundary layer development zone; (C) test chamber (2.40  1.60 m); (D) connection zone; (E) engine
(160 kW); (F) diffusion T-shape.

Fig. 6. Boundary layer development articial roughness (wood panels) (a), spires (b).

Table 2
Locking value.

Test Alateral/Atot /
Side 1 Side 2 Side 1 Side 2
p.14 0.029 0.044 0.98 0.98
p.15 0.057 0.072 0.97 0.96
p.16 0.031 0.049 0.98 0.97
p.17 0.058 0.077 0.97 0.96

Fig. 7. Lateral surface prospect.

540 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 8. Velocity and turbulence proles.

(a) (b)

(c) (d)

direction 2

direction 1

(e)
Fig. 9. Models images, p.14 (a), p.15 (b), p.16 (c), p.17 (d), check direction (e).
 F. Rizzo / Engineering Structures 45 (2012) 536558 541

(a)

(b)

(c)
Fig. 10. Pressure coefcients maps of mean value about p.14 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

for many reasons: because a three-dimensional structure is simpli-
ed to a two-dimensional model; because it is considered an
equivalent uniform load to simulate nodal forces; nally because
only the middle cable couple is considered. The result obtained
with the two-dimensional model has been compared with the
one given by the three-dimensional model. Considering model
approximation, the numerical error of procedure is marginal. Also
the difference between nodal forces and uniform load are negligi-
ble in this phase because the cable wheelbase is generally much
smaller than cable span [9,13,7]. Optimum shapes are those which,
under wind suction conditions, have high stress values in stabiliz-
ing cables, but not so high as to exceed limits, and at the same time
low stress values in load bearing cables, but not so low as to invert
their curvature. Behavior under snow loads is the opposite. After
dening the sagspan ratio for the two orders of cables it is
possible to dene the f1, f2 values, respectively the load bearing
542 F. Rizzo / Engineering Structures 45 (2012) 536558

(a)

(b)

(c)
Fig. 11. Pressure coefcients maps of min value about p.14 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

and the stabilizing sags, and subsequently the total roof height (H),
dened as the sum of f1 and f2. This derivative parameter is sig-
nicant for estimating the volume covered by the roof alone. An-
other geometric parameter must be added: the height of the roof
from the ground oor, indicated by Hb. Many considerations about
use, exibility and modiability of internal spaces were carried out
to select the size of this parameter. It was concluded that there are
growing needs for multifunctional spaces. Sports arenas, for exam-
ple, need covered parking or recreational or restaurant areas. The-
aters need dressing rooms, warehouses need different storage
areas and so on. If designers do not want to change the shape of
the building by adding outbuildings that often have lower quality
and give a negative global architectural balance it is better to
introduce these spaces into the building on multiple levels. I, to
conclude, have chosen to test only one (f2/f1) sag ratio value, (c,
equal to 2, two different ratios of H/Lmax(q), with Lmax equal to
the maximum length of span, respectively equal to 1/6 and 1/10
of Lmax and, nally, two different values of Hb, (b), respectively
equal to 1/3 and 1/6 of Lmax. The geometric parameters analyzed
are shown in Fig. 2. Overall, therefore, four different geometric
combinations were tested for the elliptical structure, described in
Table 1, where Ze is the sum of H and Hb.
 F. Rizzo / Engineering Structures 45 (2012) 536558 543

(a)

(b)

(c)
Fig. 12. Pressure coefcients maps of max value about p.14 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

2.1.2. Test models
Wooden models were built to perform wind tunnel tests,
each tted with from 175 to 211 pressure taps. Pressure taps
were distributed to have the highest number along the edges
of the model and along the two centerline sections, longitudinal
and transverse, of the roof. This was done to have the most
information from the areas of the greatest interest. Each pressure
tap was connected to pressure transducers with a pneumatic
connection made of Teon pipes with 1.3 mm internal diameters.
Fig. 3 illustrates the setup of one of the models tested. Pressure
tap distribution on the roof and the side surfaces of the model
had to take into account the instrumental limits of the wind tun-
nel. In this case it was possible to acquire signals from up to 128
pressure taps in a single measurement. Consequently I decided
to limit each model to two measurements by using a number
of taps inferior to 256. Fig. 4ae shows the distribution of pres-
sure taps on the models.
Correct evaluation of aerodynamic phenomena assumes that
ow characteristics be correctly simulated during wind tunnel
tests. During testing, in fact, certain laws, called similarity crite-
ria, must be complied with. These laws make it possible for the re-
sults obtained on models to effectively represent the situation at
the real scale. The model, in the rst place, should be made at
a suitable scale in order to satisfy the conditions of geometric
544 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 13. Pressure coefcients variation about p.14 geometry, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind action direction 1 (c), 90 wind action
direction 2 (d).

similarity imposed by the ow. Then special attention must be
dedicated to scaling of the Reynolds number. In long squat bodies
for example, in spite of the fact that the sharp edges cause imme-
diate ow separation independent from the Reynolds number, ow
reconnection may take place along the body and this is a phenom-
enon in function of the Reynolds number. In these conditions the
forces induced by wind action may be signicantly different. In
the case of bodies with curved surfaces, the effects of the Reynolds
number are even more signicant. It is practically impossible, in
the wind tunnel, to strictly conserve the Reynolds number with
the scale factors usually used. For example simulation with a mod-
el scale equal to 1:100 would require the use of wind tunnel speeds
greater than one hundred times those that could be encountered in
reality. To faithfully reproduce the atmospheric boundary layer,
characterized by very large Reynolds numbers, it is necessary to
introduce technical roughness. In particular, leaving the tunnel
Reynolds number necessarily less than the real one, it is possible
to change the system (for example from laminar to turbulent ow)
by applying surface discontinuities, obtained using appropriate
strips of material or granular paints [8]. The tests described here
 F. Rizzo / Engineering Structures 45 (2012) 536558 545

Fig. 14. Pressure coefcients maps of mean value about p.15 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

used a paint with a nal result similar to the surface of an ordinary
sandpaper. The nal roughness that was chosen had a thickness
equal to 4 mm, corresponding to Lmax/2000.
2.2. CRIACIV wind tunnel
Aerodynamic tests were carried out at the CRIACIV wind tun-
nel located in Prato, Fig. 5a and b. This is an open circuit tunnel,
with a propulsion in suction and a closed test chamber. It has
axial fan blades with adjustable tilt which allow constant varia-
tion of the ow rates between 0 and 35 m/s. Power absorption is
about 160 kW. The main components of the gallery are, referring
to Fig. 5b: the entrance, equipped with a 5 mm thick honeycomb
panel and a square mesh screen, the convergent section which
guarantees a sufciently uniform ow in the test chamber, the
tunnel itself, about 8 m long with a rectangular cross section
with a width equal to 2.2 m and a height of 1.6 m, slightly diver-
gent from upstream to downstream to ensure a nearly constant
pressure along the axis. Wind speed is controlled by the inclina-
tion of the drive blades and the rotation frequency. A turbulence
546 F. Rizzo / Engineering Structures 45 (2012) 536558

(a)

(b)

(c)
Fig. 15. Pressure coefcients maps of min value about p.15 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

boundary layer is developed in the tunnel by placing, along the
length of the wind tunnel, panels with surface roughness, Fig. 6a,
that simulate the roughness of the site where the building is to
be positioned. For this project I chose an average roughness in
order to simulate general conditions. The roughness used to gen-
erate the boundary layer was made by introducing wood panels
with constant roughness and uniformly distributed. Panel sizes
were 200 by 80 cm on which a series of wooden cubes were ar-
ranged. However the limited length of the wind tunnel generates
a boundary layer of modest height and turbulence with small
eddies that modify the real spectrum of the uctuating compo-
nent of wind speed. For this reason, in addition to surface rough-
ness, some spires were positioned in the mouth of the tunnel
designed to raise the boundary layer and to generate larger inte-
gral scales of turbulence. These spires are vertical as shown in
Fig. 6b. Another important aspect to keep in mind during wind
 F. Rizzo / Engineering Structures 45 (2012) 536558 547

(a)

(b)

(c)
Fig. 16. Pressure coefcients maps of max value about p.15 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

tunnel testing is the ratio between the size of the object tested this is not possible then the effects of locking are signicant
(the model) and the size of the test chamber [6,10]. It is impor- and must be corrected using the locking coefcient / expressed
tant to avoid or minimize the locking phenomenon that occurs by Eq. (1):
when a body, placed in the test section of a wind tunnel,
partially blocks the air passage, causing an acceleration of  2
Atot  Alateral
ow around the lateral body surfaces. Generally, the scale u 1
Atot
model is chosen in order to render locking effects negligible. If
548 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 17. Pressure coefcients variation about p.15 geometry, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind action direction 1 (c), 90 wind action
direction 2 (d).

where Atot is the cross sectional area of wind tunnel test chamber
and Alateral is the sectional area of the model. The Alateral/Atot ratio
is dened as locking. Table 2 gives locking values for each of the
geometric shapes tested. Locking values range between 1.5% and
7.7% with peaks for models p.15 and p.17, reasonable when I
consider these are highest above the ground referring to Fig. 7.
The correction value for the pressure coefcients mapped is
estimated in Table 2 and varies from a minimum of 2% to a
maximum of 4% [5].
2.3. Data acquisition
An anemometric test was done before performing aerodynamic
tests to calculate the law of scalar variation coefcients that trans-
lates dynamic and static pressure values measured by a pitot located
on the ceiling of the wind tunnel. Anemometric measurements were
made using a hot wire anemometer probe equipped with a monol-
ament, positioned in line with the position of the model. Measure-
ments was repeated for four different levels, respectively equal to
 F. Rizzo / Engineering Structures 45 (2012) 536558
(a)
(b)
(c)
Fig. 18. Pressure coefcients maps of mean value about p.16 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).
the heights of the models being tested. These measurements gave an
approximation, in both exponential and logarithmic form, of the
average wind speed as a function of altitude. The proles of the aver-
age test speed, its logarithmic approximation and the variation in
turbulence intensity are plotted in Fig. 8. The pressure coefcient
values (Cp) are equal to the ratio between the orthogonal pressure
to surface p(t) and the static pressure of undisturbed ow p0, divided
by mean value of undisturbed ow kinetic pressure qm where q is air
density (assumed as 1.25 kg/m3) and vm is mean value of undis-
turbed ow speed:
Pt  P0 Pt  P0
C p t 1 2
qm 2
qv m2
549
The data obtained by experimental tests are pressure coef-
cients time-histories from which mean values (Cp,m) have been
evaluated as well as maximum values (Cp,max) and minimum val-
ues (Cp,min), using a probabilistic theory according to the Gumbel
method, with a comeback period equal to 50 years. Considering
the aim of the present study, it is very important to dene a
time-independent referential value CD where PD is wind pressure,
qD is dynamic wind pressure and VD is maximum value of wind
speed:
PD PD
C pD t 3
qD 12 qv 2D
550 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 19. Pressure coefcients maps of min value about p.16 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

Mean values of Cp and the maximum and minimum pressure
coefcients values, respectively Cp,max and Cp,min, are calculated by
pressure time histories. Data for 16 different wind angles (each
at 22.5) were acquired for each model at a frequency of 252 Hz
and for a time interval equal to approximately 30 s. This gave, for
each pressure tap, a pressure time history of 7504 values for each
wind angle. Subsequently pressure coefcients maps were plotted.
These dene, for each pressure tap, an area of inuence known as a
Thiessen polygon in which pressure is considered constant. Tests
were carried out at a mean wind speed equal to 16.7 m/s at a
height of 10 cm (model scale). If I assume that the scale length of
the model, k1, is equal to 1:100 compared to the real building scale
and that the average real scale wind speed U10, at 10 m height, is
equal to 27 m/s, then we have a speed scale, kv, equal to 0.62. This
corresponds to a time scale kt equal to 0.0162. Therefore the 30 s
acquisition time in the wind tunnel at model scale corresponds
to a real scale time of about 1800 s. This value is included within
the range 6003600 s (as suggested by many references, see, for
example [11]) [21]. Minimum and maximum pressure coefcients
have been calculated using a probabilistic method according to the
Gumbel method [1] following the procedure proposed by Cook and
Mayne [24], associated with a probability of 22% that it will be ex-
ceeded, as is done by Eurocode 1.
2.4. Experimental results
It is important to clarify that the pressure coefcient values pro-
posed in this paper can only be used by designers for preliminary
 F. Rizzo / Engineering Structures 45 (2012) 536558 551

(a)

(b)

(c)
Fig. 20. Pressure coefcients maps of max value about p.16 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

design purposes. Values can differ, depending on the building
being analyzed, in function of aspects such as surrounding geo-
graphic conditions, building surfaces and many other characteris-
tics. More precise studies, such as POD or CFD analyses, are
required when studying the dynamic response to wind forces of
buildings with hyperbolic paraboloid roofs.
Pressure coefcient maps for each geometric shape tested in
the wind tunnel and described in Table 1, p.14, p.15, p.16 and
p.17, are given below. Results obtained at a 0 angle are very
similar to those obtained at 180 and 360 angles and can be
considered as equal. Likewise results obtained at a 90 angle
are very similar to those obtained at 270, and again likewise re-
sults obtained at a 45 angle are equal to those obtained at 135,
225 and 315 angles. Sixteen angles were evaluated but report-
ing all data for all of these would be extremely lengthy. Fig. 9
gives images of the models that were tested and the two princi-
pal directions used to plot pressure coefcient variations.
Fig. 9ad, respectively illustrates models p.14, p.15, p.16 and
p.17 according to Table 1. Fig. 9e illustrates two orthogonal axes,
directions 1 and 2, along which pressure coefcient variation is
plotted. Figs. 1013 show, respectively, mean, minimum and
maximum pressure coefcient map values for test p.14. Fig. 14
plots pressure coefcient variations along directions 1 and 2.
Figs. 1517 show the maps of pressure coefcients (mean, min-
imum and maximum) for test p.15. Fig. 18 shows their variation
along directions 1 and 2. Analyzing data for the 0 angle of wind
552 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 21. Pressure coefcients variation about p.16 geometry, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind action direction 1 (c), 90 wind action
direction 2 (d).

in direction 1, respectively for models p.14 and p.15, shows two
perfectly parallel curves. Pressure coefcients increase (in
absolute value) as the height above ground increases. Flow, after
an initial minimal compression, proceeds with rapid suction. It
continues to grow after the rst jump in the separation zone
up until the middle of the roof, where it reaches the maximum
suction value. After this the values of pressure coefcients
decrease (in absolute value). In direction 2 pressure coefcients
have an almost constant trend. Behavior at 90 is different. This
time direction 2 is parallel to the ow. Note that ow in direc-
tion 2 detaches and reaches its highest level of suction in the
separation zone without compression. In this case the roof seems
to be divided into two zones, the rst with high suction and the
second with mediumlow suction. Figs. 1921 give pressure
coefcient maps (mean, minimum and maximum) for test p.16.
Fig. 22 plots variations along directions 1 and 2. Figs. 2325
show pressure coefcient maps (mean, minimum and maximum)
for test p.17. Fig. 26 reports variations along directions 1 and 2.
As with the previous pair of roofs (p.14 and p.15) the only
difference between the models, p.16 and p.17, is the height of
the roof. Pressure coefcients, in this case also, follow a parallel
trend. Note that with a 0 wind angle the rst part in compres-
sion is greater than that noted for models p.14 and p.15. This is
due to the size of the surface curvature radius of the roof. The
 F. Rizzo / Engineering Structures 45 (2012) 536558 553

(a)

(b)

(c)
Fig. 22. Pressure coefcients maps of mean value about p.17 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

radius of curvature is smaller in models p.16 and p.17, so the
curvature is greater and therefore favors the initial ow attack.
With wind angle equal to 90 in direction 2, direction is parallel
to the ow. Here, too, the roof is split into two areas, the rst
with high suction and the second with low suction. Comparison
of behavior between roofs p.14 and p.16, and consequently
between roofs p.15 and p.17, is very interesting. These roofs
have the same value of Hb but different H values and, therefore,
different curvatures. The rst shape (p.14 and p.15) has less cur-
vature than the second (p.16 and p.17). Fig. 26 compares mean
pressure coefcient values along Sections 1 and 2 for all four
models tested. Comparing data between p.14 and p.16 for a
wind angle of 0 along direction 1 (according Fig. 9e) shows
how suction values in the initial separation zone are greater
for the shape with the lesser curvature (p.14) whereas this dif-
ference is reversed in the middle of roof where, in fact, the suc-
tion value is higher for shape p.16. This pattern occurs because
the ow, in surfaces with greater curvature (p.16 and p.17),
554 F. Rizzo / Engineering Structures 45 (2012) 536558

(a)

(b)

(c)
Fig. 23. Pressure coefcients maps of min value about p.17 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

follows the curvature of the roof and uses the rst part of the
roof as a ramp for jumping. In atter roofs the ramp is shorter
and the ow jumps less. In direction 1, the difference in the
middle of roof between the two different curvatures is 15% with
the difference of geometric shape in direction 1 (the ratio be-
tween f2 of shape p.14 and f2 of shape p.16) equal to 60%. In
direction 2 (according Fig. 9e) note that the curvature does not
affect the Cp because the side is smaller (40 cm rather than
80 cm) and therefore the trend of pressure coefcients is almost
rectilinear. Comparing the data for a 90 wind angle in direction
2 shows that the values of suction in the initial separation zone
are greater for the roof with the lesser radius of curvature while
this difference is reversed in the middle zone. Note how there is,
exactly in the middle of the roof, a rapid decline of suction val-
ues for all geometric shapes whereas ow never assumes a com-
pression value at the end of the roof. Values in the end zone in
any case are very low and tending to zero.
3. Comparison between elliptical and circular shape HYP roofs
A comparison of results from aerodynamic wind tunnel testing
of a HYP roof with an elliptical shape (the elliptical model), the
 F. Rizzo / Engineering Structures 45 (2012) 536558 555

(a)

(b)

(c)
Fig. 24. Pressure coefcients maps of max value about p.17 geometry, 0 wind action (a), 45 wind action (b), 90 wind action (c).

object of the present paper, with results from testing a HYP roof
with a circular shape (the circular model), Fig. 27 [20], at the same
speed and boundary conditions (frequency, turbulence, rough-
ness), with the same curvature (H) and heights above ground
(Hb) shows that, in general, mean pressure coefcients on the ellip-
tical model are lower (in absolute values) than on circular models
[18,20]. Figs. 28 and 29 below show maps and graphics of mean
pressure coefcients on a circular model with the same curvature
and height as model p.14. Note that in mid span, direction 1 with
a wind angle of 0, suction values with average pressure coef-
cients, considering the different locking coefcient, are three times
(overall) those of the elliptical model. The difference is greater in
the separation zone where the mean value of Cp for elliptical
shapes is about 1/25th the value obtained in the same zone for a
circular shape. Results are similar in the middle of the roof. A rep-
resentative mean Cp value for elliptical models is 0.3 (absolute va-
556 F. Rizzo / Engineering Structures 45 (2012) 536558

Fig. 25. Pressure coefcients variation about p.17 geometry, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind action direction 1 (c), 90 wind action
direction 2 (d).

lue). The same value is 0.4 (absolute value) for circular models.
Pressure coefcients are numerically similar with a 90 wind angle
in direction 2 as can be noted by comparing Figs. 29c, 12d and 13.
However the trend is different: for roofs with a circular shape the
rapid decrease in pressure coefcients does not happen in the mid-
dle of the roof, as it does in roofs with elliptical shapes, but much
sooner, at about 1/8th of the length (in direction 2) of side 2
(according to Fig. 7). This behavior is more similar to a Cp trend
for hyperbolic paraboloid roofs with a rectangular footprint (F. Riz-
zo et al. (2011) Structures and Engineering) and is due to the short
dimension of side 2. We conclude that the elliptical shape has con-
siderably higher performance and better aerodynamic behavior be-
cause the shape minimizes pressure on the roof. This shows that it
is not possible to study only two dimensional cross sections for this
type of roof but that it is very important to have three-dimensional
data as well.
4. Conclusions
Aerodynamic wind tunnel tests were carried out on hyperbolic
paraboloid roofs with elliptical footprints. Different geometric
shapes and heights were take into account in order to parameterize
the pressure coefcients of the hyperbolic paraboloid roof. Re-
search in this paper follows research obtained with aerodynamic
tests performed on hyperbolic paraboloid roofs with square, rect-
angular and circular footprints, published previously. Results show
that roofs with an elliptical shape have the best aerodynamic
behavior among the four shapes considered and tested: the shape
of the footprint, combined with the complex surface curvature,
permits the building to be particularly efcient especially in a
direction parallel to the wind. Interesting differences in behavior
were observed between the different shapes. In particular it was
shown that it is not possible to use data for a hyperbolic paraboloid
 F. Rizzo / Engineering Structures 45 (2012) 536558 557

Fig. 26. Comparison between p.14, p.15, p.16 and p.17 pressure coefcients
variation, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind
action direction 1 (c), 90 wind action direction 2 (d).

Fig. 28. HP roof with circular plane, mean value of pressure coefcients with wind
action equal to 0 (a), 90 (b), 45 (c); original graphics reported in F. Rizzo et al.
(2011) Structures and Engineering.

roof with a circular shape to study a roof with an elliptical shape.

Overall this research shows that particular shapes require specic
pressure coefcients and that these should be included in codes
and legislation to facilitate design of these particular types of
Fig. 27. Circular plane test model. structure.
558 F. Rizzo / Engineering Structures 45 (2012) 536558
(a)
(c)
Fig. 29. Pressure coefcients variation about HP roof with circular plane, 0 wind action direction 1 (a), 0 wind action direction 2 (b), 90 wind action direction 1 (c), 90 wind
action direction 2 (d).
Acknowledgements
Special thanks to Professor Piero DAsdia, research coordinator,
Engineer Massimiliano Lazzari co-author of the paper [20], Profes-
sors Francesco Ricciardelli and Gianni Bartoli, co-authors of the pa-
per [21], and, nally, Engineers Ettore Masciulli and Donato
Forgione for their collaboration during wind tunnel testing [18].
References
[1] Gumbel EJ. Statistic of extremes. NY: Columbia University Press; 1958.
[2] Cook NJ, Mayne JR. On design procedures for wind loading. Garston: Building
Research Establishment; 1978.
[3] Cook NJ, Mayne JR. A novel working approach to the assessment of wind loads
for equivalent static design. J Wind Eng Ind Aerodyn 1979;4:14964.
[4] Cook NJ, Mayne JR. A rened working approach to the assessment of wind
loads for equivalent static design. J Wind Eng Ind Aerodyn 1980;6:12537.
[5] Elashkar I, Novak M. Wind tunnel studies of cable roofs. J Wind Eng Ind
Aerodyn 1983;13(13):40719.
[6] Melchers RE. Structural reliability. Elley Horwood Ltd.; 1987.
[7] Forster B. Cable and membrane roofs, a historical survey. Struct Eng Rev
1994;6(35).
[8] Simiu E, Scanlan RH. Wind effects on structures. 3rd ed. New York: John Wiley
& Sons; 1996.
[9] Majiowiecki M. Snow and wind experimental analysis in the design of long-
span sub-horizontal structures. J Wind Eng Ind Aerodyn 1998;7476:795807.
(b)
(d)
[10] Shen S, Yang Q. Wind-induced response analysis and wind-resistant design of
hyperbolic paraboloid cable net structures. Int J Space Struct 1999;
14(1):5765.
[11] ASCE (American Society of Civil Engineering). Wind tunnel studies of buildings
and structures. In: Isyumov N, editor. Manuals of practice (MOP), vol. 67; 1999.
[12] Australian/New Zealand Standard. Structural design actions Part 2: Wind
actions. AS/NZS 1170.2; 2002.
[13] Majiowiecki M. Tensostrutture: Progetto e Verica. Milano: Edizioni Crea;
2004 [in Italian].
[14] Lewis WJ. Tension structures: form and behaviour. ASCE standard; 2004.
[15] ASCE (American Society of Civil Engineers). Minimum design loads for
buildings and other structures. ASCE 7-05; 2005.
[16] CEN (Comit Europen de Normalisation). EN 1991-1-4: Eurocode 1: actions
on structures Parts 14: General actions wind actions; 2005.
[17] CNR (National Research Council of Italy). CNR-DT 207/2008 Guide for the
assessment of wind actions and effects on structures; 2011.
[18] Rizzo F. Studio Aerodinamico e Analisi non Lineare di coperture di medie e
grandi luci: Progettazione ottimizzata di tensostrutture in reti di funi. Ph.D.
thesis, University G. DAnnunzio of Chieti/Pescara; 2009 [in Italian].
[19] Masciulli E, Forgione D. Studio Aerodinamico di coperture a paraboloide
iperbolico a pianta ellittica. Graduated thesis, University G. DAnnunzio of
Chieti/Pescara; 2011 [in Italian].
[20] Rizzo F, DAsdia P, Lazzari M, Procino L. Wind action evaluation on tension
roofs of hyperbolic paraboloid shape. Eng Struct 2011;33(2):44561.
[21] Rizzo F, DAsdia P, Ricciardelli F, Bartoli G. Characterisation of pressure
coefcients on hyperbolic paraboloid roofs. J Wind Eng Ind Aerodyn, 102C, 61
71.