Professional Documents
Culture Documents
Engineering Structures: Fabio Rizzo, Piero D'Asdia, Massimiliano Lazzari, Lorenzo Procino
Engineering Structures: Fabio Rizzo, Piero D'Asdia, Massimiliano Lazzari, Lorenzo Procino
Engineering Structures: Fabio Rizzo, Piero D'Asdia, Massimiliano Lazzari, Lorenzo Procino
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
0. Introduction hand they are extremely complex in the design phase. In fact, such
structures have a ratio between live loads and dead loads which is
A dynamic approach is now required while designing such the opposite of traditional heavy structures and this determines
large free spaces where a structure’s flexibility and modifiability a different approach in structural analysis. Therefore, wind and
are the most important prerequisites; reference goes to spaces snow action analysis has an instrumental role in the design and
that act like stages for activities performed all around, such as optimization of such structures, which can easily buckle and so
sports complexes, swimming pools, sport arenas and shopping are particularly susceptible to dynamic wind action or to snow or
malls. The commonly used structures for these areas are spaces water loads, especially if they are not-symmetrical (eg. [1–3]). For
without mid-span pillar or headroom; as a consequence, they this reason a non linear static and dynamic analysis – carried out by
very often exploit a long-established covering solution such specific software and specialised technicians – is needed (eg. [4,5]).
as the tension structure, improved by new technologies that In addition to their design complexity, another constraint
make it more reliable and comfortable than before. The tension
that prevents the diffusion of tension structures is the lack in
structure meets the current market’s requirements in terms of
international codes of mandatory standards for structural design
lightness, innovation of materials, cost-effectiveness and lack of
and action distribution on complex shapes (such codes exists
intermediate support while traditional structures (as, for example,
only for temporary constructions). In fact, in Italy, in Europe
steel or concrete plates or plywood timbers) often result in being
as well as in the United States, this constructive typology has
so expensive and massive that construction becomes difficult.
been neglected so far and even ignored by building regulations.
Among the tension structures, the most frequently chosen
shape to cover large spans is the hyperbolic paraboloid roof. This is particularly true concerning the hyperbolic paraboloid
If on one hand tension structures realise a perfect compromise shape or other particular shapes such as hyperboloid or ellipsoid:
between lightness and high structural performance, on the other they are completely ignored by national & international building
regulations and thus specialized experimental tests (such as wind
tunnel tests) are the only way to study them and to investigate
∗ wind–structure interaction [4].
Corresponding author. Tel.: +39 3200384186.
E-mail addresses: fabiorizzo79@libero.it (F. Rizzo), pidasdia@tin.it (P. D’Asdia),
With the aim of compensating for lawlessness in the prelim-
massimiliano.lazzari@unipd.it (M. Lazzari), lorenzo.procino@pin.unifi.it inary design of tension structures, experimental tests have been
(L. Procino). performed in the CRIACIV’s (Interuniversity research Centre of
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2010.11.001
446 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
Fig. 1. Roof Structure: Load bearing cable (C1) and Stabilizing cable (C2) (a), Roof structure: 2D - structural system (b).
Aerodynamic and Wind Engineering) wind tunnel in Prato (Italy). The response given by the procedure is structurally approxi-
The tests have analysed in particular the wind pressure distribu- mate for many reasons: because a three-dimensional structure is
tion on roofs with a hyperbolic paraboloid shape. simplified to a two-dimensional model; because it’s considered an
The hyperbolic paraboloid is a doubly ruled saddle shaped equivalent uniform load to simulate nodal forces; finally because
surface; it belongs to open quadric surface family and thanks to only the middle cable couple is considered. The result obtained
its generating parables it allows to obtain only traction stress for with the two-dimensional model has been compared with the one
all loading conditions. In a suitable coordinate system X , Y and Z given by the three-dimensional model and the most interesting re-
(X , Y horizontal plane and Z vertical axes), it can be represented sults are listed in Table 3(while many other relevant results will be
by the equation: shown in a future paper). In Table 3 it’s possible to note that the
f1 f2 differences between the simplified procedure and Finite Element
z= x2 − y2 (1) Method Analyses are very small. Considering model approxima-
L21 L22
tion, the numerical error of procedure is marginal. Also the differ-
where C1 and C2 are bearing and stabilizing cables, f1 , f2 , are the ence between nodal forces and uniform load are negligible in this
cable sags and L1 , L2 are the cable spans Fig. 1. phase because the cable wheelbase is generally much smaller than
This is a hyperbolic paraboloid that opens up along the x-axis cable span. Finally if (1) is used, the middle cables are parallel to
and down along the y-axis. the other cables of net and their curvatures are equal to the one of
A set of surfaces is obtained exchanging some geometrical other cables of net.
parameters; among them it is necessary to choose the one with Starting from fixed geometry parameters (spans and sags),
the best geometrical configuration and, in order to obtain it,
loads conditions (live and dead), minimum and maximum limit
a procedure of structural optimization has been arranged. This
stresses (σmax , σmin ) of cable for the extreme load conditions
procedure considers wind action (pressure and suction), snow
(maximum snow action and maximum wind depression), the
action and dead loads. Optimal geometries are those with the best
procedure of preliminary design evaluates with an iterative
balance between used material and high performances: they have
process the ideal value of the cable areas in order to obtain
an optimal ratio between sags and spans. These are the shapes
which have been tested in the CRIACIV wind tunnel. the balance between the fixed geometrical configuration and the
aimed stress limits.
1. Preliminary design Four key-conditions of load configuration are considered, as
listed in Tables 1 and 2:
In designing cable structures and particularly roofs with a
hyperbolic paraboloid shape, the primary difficulty is represented
• ‘‘0 condition’’: is the starting balance state where only cable
by the setting of the main structures’ geometries for preliminary weight acts. The cable stress is an initial value calculated in
design. order to maintain, for given geometry, displacements equal to
The present work defines an automatic procedure to obtain a zero.
sample of optimal geometries starting from many different shapes. • ‘‘1 condition’’: is the permanent action state, where – in
Such a procedure is based on the classical theory of the cable addition to cable one – other elements’ weight operates:
beam and its purpose is to derive from a three-dimensional surface systems, membranes and roof panels’ weight. The difference
described by Eq. (1) a two-dimensional static model that simplifies of action between 0 condition and this condition brings
and speeds up the calculus. The three-dimensional surface is a geometrical variation. The ‘‘1 condition’’ geometry is the
represented in Fig. 1(a) and the two-dimensional model in Fig. 1(b). final and the permanent roof configuration in case of no
The steps of this process were the following: environmental intervention.
– identification of a bearing zone (load bearing cables with a • ‘‘2 condition’’: is the maximum snow action state. In this
down concave) and a stabilizing zone (stabilizing cable with condition the load bearing cables (C1) reach the maximum
an up concave) with an opposite flexure. The geometrical extension and the highest values of stress, while the stabilizing
parameters are represented in Fig. 1(a). cables (C2) reach the minimum value of stress; if the structure is
– assumption of a two-dimensional model (rope beam) to badly designed, then the stabilizing cables reach the curvature
simulate the real three-dimensional structure. reversal and therefore the structure collapses.
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 447
Fig. 2. Different geometric shapes evaluated with the procedure of preliminary design.
Fig. 3. Cables area variation with different values of γ (f2 /f1 ), ρ(H /Lmax ), α = 1, Lmax = 20 m (a). Variation of displacements with snow and wind action with different
values of γ (f2 /f1 ), ρ(H /Lmax ), α = 1, Lmax = 20 m (b).
Table 3
Comparison between F.E.M. analysis and the procedure of preliminary design with a 2D static model.
Cable Procedure of preliminary design stress (MPa) F.E.M. stress (MPa) Stress variance (MPa) (%)
Table 4
The tests programme and an outline of the geometrical parameters used.
Lable H Hb L1 (D1 ) L2 (D2 ) Lmax f1 f2 γ ρ α δ
(m) (m) (m) (m) (m) (m) (m)
– the cable geometrical length under maximum snow action 1.1. Samples of different geometrical shapes
(‘‘2’’) (L2,1,geometrical ) is evaluated by:
L1,ind Using the previously explained procedure, for a set of different
L2,1,geometrical = (6) geometries a sample of different roof shapes has been defined (it’s
1 + ε 2 ,1
– length variations (∆L0,1 and ∆L2,1 ) in ‘‘0’’ and ‘‘2’’ load presented schematically in Fig. 2). In order to obtain such a sample
conditions about the cable geometrical length without stress three different geometrical parameter ratios are varied:
(L1,ind ) are: • γ ratio between cable’s sags: γ = f2 /f1 ;
∆L0,1 = L0,1,geometrical −L1,ind • r ratio between the roof height (H) (H is equal to the sum of
(7) cable’s sags H = f1 + f2 ) and the maximum value of spans of
∆L2,1 = L2,1,geometrical −L1,ind
cables (C1 and C2) Lmax = max(L1 , L2 ) : ρ = H /Lmax ;
– the vertical displacement values (∆f2,1 ) of the C1 middle
node under snow action are therefore evaluated by:
• α ratio between cable spans: α = L1 /L2 .
∆L2,1 −∆L0,1 Some parametric curves – that can be used in the preliminary
∆f 2,1 = 16 f
. (8) design – can be drawn following the variation of the above-
1
3 L1 mentioned ratios. For example, considering a cable net with cable
The analysis for stabilizing cable and for all load configurations is spans equal to 20 m (L1 equal to L2 ), with the traction force value
carried out in the same way. equal to 150 kN, the maximum stress value of C1 cable (under snow
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 449
Fig. 5. Cables stress with an optimized geometry: example α = 1, ρ = 1/10, γ = 2.33 (a), Cables stress with a no optimized geometry: example α = 1, ρ = 1/3,
γ = 0.43 (b).
450 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
Fig. 7. Pressure take-offs: square model shape height 1 (a), square model height 2 (b), circular model shape (c), rectangular model shape (d).
Fig. 8. Wind tunnel velocity profile (a), wind directions and direction of sections plotted (b).
considered. This example shows that cables nets with a γ ratio less Two values of Hb have been added: 1/3 of Lmax and 1/6 of Lmax and
than 1 have little variations of cables stress values for each load another geometrical parameter ratio is introduced: δ ratio between
condition; on the contrary, cables nets with a γ ratio greater than Hb and Lmax : δ = Hb /Lmax .
1 have greater variations of cables stress values, greater value of Table 4 shows the tests program and the outline of the used
vertical displacements, Fig. 3(b), but lower values of cable areas geometrical parameters.
and consequently lower value of structural weight than the other The aerodynamic global forces (drag force in particular) are
configuration of Fig. 3(a). a function of Reynolds number in relevant measure only for
The selected sample has three different α ratios (and so three structures with a circular plan, while structures with a square plan
different plan shapes): the first equal to 1 (square plan), the second don’t show a big variation. In the case of structures with a circular
equal to 0.5 (rectangular plan with C1 span value less than C2 plan, drag force is influenced by side surfaces roughness; this
span value) and the last one equal to variable value (circular plan), influence is particularly evident with Reynolds numbers included
Fig. 6; it has also one value of γ ratio (equal to 2) and two different between 105 and 106 [7]. Since the structures analysed in the
values of ρ ratio equal to 1/6 Lmax and 1/10 Lmax . Having in present paper show a Reynolds number included in the above-
mind the design of sport arenas, swimming-pools, congress centres mentioned range, two tests on circular-plan models with two
with shopping malls and common inner spaces, it has been taken different degrees of side roughness have been carried out. One test
into account another geometrical parameter (in the following Hb ) has the k/D ratio equal to 0 (with k = dimension of roughness and
measuring the distance between the roof and the ground floor. D = diameter of structure circumference) and the second one has
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 451
a b
120
100
80
60
40
20
c d
120 120
100 100
80 80
60 60
40 40
20 20
0 0
e f g
h i j
Fig. 9. P1 — test (H = 1/10L, Hb = 1/6L) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
the k/D ratio equal to 0.0025. The results show that the maximum, (reported in [7]). The above-mentioned slight differences come
minimum and average values of CP don’t differ considerably in from the different geometric configuration of the separation fluid
local terms. In overall terms the global forces reflect the literature due to the different degrees of roughness.
452 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
a b 120
100
80
60
40
20
c d
120 120
100 100
80 80
60 60
40 40
20 20
0 0
e f g
h i j
Fig. 10. P.7 — test (H = 1/6L, Hb = 1/6L) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
2.1. Wind tunnel tests their height is between 18.00 and 40.00 cm. All models have about
155–233 pressure take-offs with a pipe size equal to 1.3 mm, that
The tested models are made of wood with 1:100 scale. The plan are distributed on the roofs and on the four lateral surfaces as it’s
size is between 40.00 and 80.00 cm (real size 40.00–80.00 m), while presented in Fig. 7(a)–(d).
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 453
a b
c d
Fig. 11. Square shape tests results comparison: direction 1—0° (a), direction 2—0° (b), direction 1—90° (c) direction 2—90° (d).
Experimental tests have been performed in the CRIACIV’s 2.2. Wind tunnel data
boundary layer wind tunnel (e.g. [8–10]) in Prato (Italy). The
boundary layer in the wind tunnel is developed with a ‘‘city’’ Pressure coefficients variation of experimental tests is reported
configuration. Speed profile (mean value of speed) is presented in Figs. 9–19 For each test the maps of pressure coefficients of
in Fig. 8; the maximum speed value is equal to 16.00 m/s three of the sixteen wind angles (0°, 90° & 45°) are presented
and ground elevation corresponds to 80.00 cm. The sampling below. Fig. 8 describes the velocity profile simulated in wind tunnel
frequency is equal to 252 Hz and the acquisition time is equal tests. It’s important to clarify that errors of the registered data
to 30 s. The anemometric measures have been carried out with have been replaced by symmetric configuration data and by mean
four different ground elevation values, 10, 20, 30 and 40 cm. The values of surrounding pressure take-offs data. For each wind angle,
pressure coefficients (Cp (t )) are non-dimensionalized with respect mean (Cp,m ), maximum (Cp,max ), and minimum (Cp,min ) pressure
to ground elevation of 20.00 cm and data for sixteen different wind coefficients values are plotted. The plotted values refer to two cross
angles have been obtained. The pressure coefficient values (Cp ) are sections in the middle of the roof: the first (direction 1) is parallel to
equal to the ratio between the orthogonal pressure to surface p(t ) C1 cable direction, the second (direction 2) is perpendicular to C2
and the static pressure of undisturbed flow p0 , divided by the mean cable direction. Negative values of pressure coefficients represent
value of undisturbed flow kinetic pressure qm where ρ is air density wind suction values.
(assumed as 1.25 kg/m3 ) and Vm is mean value of undisturbed flow
speed: 2.3. Square shape tests comparisons
P (t ) − P 0 P (t ) − P 0
CP (t ) = = . (9) Fig. 9 (P.1 tests) and Fig. 10 (P.7 tests) illustrate the experimental
qm 1
2
ρv 2m tests results of structures with square plan with two different
The pressure coefficients have been normalized with respect values of sag measurements. P.2 and P.8 tests results are not
to the mean value of speed evaluated at a referential ground described fully for the sake of brevity. Fig. 11 shows the comparison
elevation (ze ) equal to 0.05 m. The data obtained by experimental of the behaviour of four different geometric configurations of
tests are pressure coefficient time-histories from which mean square plan models (P.1, P.2, P.7 and P.8 as described by Table 4 for
values (Cp,m ) have been evaluated as well as maximum values square roof), reproducing also two different directions of sections
(Cp,max ) and minimum values (Cp,min ), using a probabilistic theory (direction 1 and direction 2) and two different wind angles.
according to the Gumbel method [11,12], with a comeback period Some meaningful results are obtained by analysing the interaction
equal to 50 years. Considering the aim of the present study, it’s between aerodynamic behaviour and geometry of structures. With
very important to define a time-independent referential value CD a wind angle of 0°, higher values of suction are shown more in the
where PD is wind pressure, qD is dynamic wind pressure and VD is separation zone (surrounding leading edge) than elsewhere. On the
maximum value of wind speed [13]: contrary, an even pace of pressure coefficients is obtained along
the direction 2, where the vortex shedding influence around border
PD PD zones is lower. With a wind angle of 90° along direction 2, pressure
CPD (t ) = = . (10)
qD 1
2
ρv 2D coefficient variation is slow, the separation zone is a wide area and
454 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
a b 120
100
80
60
40
20
c d
120 120
100 100
80 80
60 60
40 40
20 20
0 0
e f g
h i j
Fig. 12. P.3 — test (H = 1/6L, Hb = 1/6L) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
accordingly to this, along direction 1 pressure coefficients have a This behaviour is due to the fact that maximum values of
constant trend. Cp (Cp,max ) become negative values. In the exit zone these
Analysing Cp variation with a wind angle of 0° (Fig. 9(e), differences disappear and Cp values are similar for each geometry.
Fig. 10(e)) along direction 1, it results that around separation Along direction 2, the width of the separation zone increases
zone mean values of Cp (Cp,m ) tail off according to H increment. according to the H increment and minimum values of Cp (Cp,min )
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 455
a b 140
120
100
80
60
40
20
c d
140 140
120 120
100 100
80 80
60 60
40 40
20 20
e f g
h i j
Fig. 13. P.5 — test (H = 1/10L, Hb = 1/6L) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
increase too. Instead, maximum values of Cp (Cp,max ) tail off and The comparison of mean values of pressure coefficients (Cp,m )
therefore mean values (Cp ,m) as well. This behaviour is more for each geometry (Fig. 11) shows that when Hb increases, Cp,m
evident for wind angles of 90° along direction 1 & 2. increases too; this behaviour is more evident when H increases.
456 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
a b
c d
Fig. 14. Rectangular shape tests results comparison: direction 1—0° (a), direction 2—0° (b), direction 1—90° (c) direction 2—90° (d).
It’s interesting to note that there is a similarity of pressure consequence, flow suction increases because of the large curvature
coefficients variation between P.7 test (H = 1/6Lmax and Hb = of the roof surface and of high values of roof height (H).
1/6Lmax ) and P.2 test (H = 1/10Lmax and Hb = 1/3Lmax ), in Wind angles of 90° determine even greater differences. In fact,
particular around the flow attachment zone for direction 1 and 2 in direction 2 there isn’t any attachment of flow in the roof surface.
and wind angles of 0° and 90°. So, the flow separation trend when H This is the consequence of the small dimensions of the side parallel
increases (and so curvature of roof surface increases) is comparable to wind direction; around the middle of the roof, suction values
to the flow separation trend when Hb increases. have a more or less even pace. This effect decreases moving away
from the middle zone of roofs. However, also in this case there
2.4. Rectangular shape tests comparisons are many differences among different geometries and in particular
among different curvatures of the roof surface. So, roof surfaces
Fig. 12 (P.3 tests) and Fig. 13 (P.5 tests) present the results that have smaller curvatures have greater values of pressure on the
of experimental tests performed on a structure with rectangular opposite side of the impact area. In fact, after the first separation,
plan with variable high and sags measure. Fig. 14 shows the wind flow comes down and hits the highest structure borders near
comparison of behaviour of the different geometric configurations this zone.
of rectangular plan models P.3, P.4, P.5 and P.6 as described by
Table 4 for rectangular roof, considering two different directions 2.5. Circular shape tests comparisons
of sections (direction 1 and direction 2) and two different wind
angles. Experimental tests data of structure with circular plan present a
The analysis of Cp variation for a wind angle of 0° shows, as different behaviour of these shapes in comparison with the one of
seen before, that, in direction 1 and around the separation zone, the square and rectangular plan structures. First of all, relevant data
suction pressure coefficients tail off when H increases. In direction have been obtained by the roughness variation of lateral surface
2, the Cp,min value increases in the middle of roof (Fig. 12(f), which has been determined by the use of two different particle-
Fig. 13(f)), a fact that reveals a new flow separation caused by the sizes. In figures Fig. 15 (P.9 tests) and Fig. 16 (P.11 tests) are shown
attachment after the first flow separation around the impact area. the results of experimental tests performed with variable high and
This flow trend disappears when H reduces and so there is no flow sags measure. P.9 w.r., P.10, P.10 w.r., and P.12 tests results are
attachment. Along directions where flow is parallel to the shorter not described fully for the sake of brevity. The pressure coefficients
side (direction 2) there aren’t attachments after the first separation, (Cp,m , Cp,max and Cp,min ) decrease with a greater lateral roughness.
while along directions where flow is parallel to the longer side As happens for square and rectangular plan structures, also for
(direction 1) there are attachments near structure borders. structures with a circular plan, Cp values of suction increase with
The comparison of structures with rectangular plans with a greater values of H, even if in this case the difference is smaller.
wind angle of 0° shows that in the middle of roofs suction values Moreover, in direction 1 and with a wind angle of 0°, it shows that
increase when H and Hb increase in direction 1, (Fig. 14). As a Cp,max values in the separation zone act as pressure but assume
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 457
110
a b
100
90
80
70
60
50
40
30
20
0 20 40 60 80 100
c 110 d 110
100 100
90 90
80 80
70 70
60 60
50 50
40 40
30 30
20 20
0 20 40 60 80 100 0 20 40 60 80 100
e f g
h i j
Fig. 15. P.9 — test (H = 1/6L, Hb = 1/6) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
458 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
a b 110
100
90
80
70
60
50
40
30
20
0 20 40 60 80 100
c 110 d 110
100
100
90 90
80 80
70 70
60 60
50 50
40
40
30
30
20
20
0 20 40 60 80 100
0 20 40 60 80 100
e f g
h i j
Fig. 16. P.11 — test (H = 1/10, Hb = 1/6) (a), Cp,m 0° (b), Cp,m 90° (c), Cp,m 45° (d), direction 1 – Cp,m , Cp,max , Cp,min – 0° (e), direction 2 – Cp,m , Cp,max , Cp,min – 0° (f), direction
1 – Cp,m , Cp,max , Cp,min – 90° (g), direction 2 – Cp,m , Cp,max , Cp,min – 90° (h), direction 1 – Cp,m , Cp,max , Cp,min – 45° (i), direction 2 – Cp,m , Cp,max , Cp,min – 45° (j).
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 459
Fig. 17. Circular shapes tests results comparison: direction 1—0° (a), direction 2—0° (b), direction 1—90° (c), direction 2—90°, (d) direction 2—0 (tests with a different
roughness) (e).
a suction behaviour (Fig. 15(e), Fig. 16(e)), on the contrary, in It’s interesting to compare Cp values for a wind angle of 0°
direction 2, Cp variation has an even pace (Fig. 15(f), Fig. 16 (f)). and 90° in direction 1 (Fig. 17): the variation trend is reversed in
It’s very interesting to note that in direction 1, with a wind angle of the middle of roofs. In fact, with a wind angle of 0° the flow is
90°, Cp variation has a parabolic behaviour (Fig. 15(g), Fig. 16(g)). parallel to direction 1, it’s detached around the separation zone
Analysing the data, it appears that the size of separation zone and attached just near it; then, with suction, the flow follows roof
is smaller in these structures than in structures with square and curvature; with a wind angle of 90°, it’s orthogonal to direction 1
rectangular shapes, in particular with a wind angle of 0°. As a (parallel to direction 2), it’s detached around the separation zone
result, pressure coefficient distribution on the roof is more regular. and it’s attached in the middle of the roof. It causes also a constant
This good aerodynamic behaviour is caused by the absence of variation of pressure on the left and on the right side of structure
borders.
corners and makes this kind of roof very different from square or
rectangular plans. With a wind angle of 90°, the separation zone is
2.6. Comparisons
greater and it involves the majority of the roof’s surface. Because of
the absence of a side orthogonal to flow direction, lateral surfaces A more meaningful result is obtained by comparing Cp variation
are more subject to flow impact than lateral surfaces of square or of different shapes of structures. The analysis of data of structures
rectangular plan structures. with the same roof curvature but different plan shape shows
The minor effect of vortex shedding induces smaller difference that the square plan structures have a minor value of Cp,m in
of behaviour among different geometries, as for example minor the detached regions in direction 1 and with a wind angle of
or major curvatures or different height. Comparing different 0° (Fig. 18(a), (b)); in direction 2, structures behaviour is similar
geometry data, it appears that ground elevation is more important for each shape and Cp,m variation is constant. However, this
than roof curvature. comparison shows that square or rectangular plan models have a
460 F. Rizzo et al. / Engineering Structures 33 (2011) 445–461
Fig. 18. Test results comparison square, rectangular and circular shape models: test (H = 1/10, Hb = 1/6) – direction 1 – 0° (a), test (H = 1/6, Hb = 1/6) – direction 1 – 0°
(b), test (H = 1/10, Hb = 1/6) – direction 2 – 0° (b), test (H = 1/6, Hb = 1/6) – direction 2 – 0° (d).
a b
c d
Fig. 19. Test results comparison square, rectangular and circular shape models: test (H = 1/10, Hb = 1/6) – direction 1 – 90° (a), test (H = 1/6, Hb = 1/6) – direction
1 – 90° (b), test (H = 1/10, Hb = 1/6) – direction 2 – 90° (b), test (H = 1/6, Hb = 1/6) – direction 2 – 90° (d).
similar behaviour, which is quite different from the one of circular have a different variation than the other shapes and the rectangular
plan structures (Fig. 18(c), (d)). The differences among the three plan structures have a more or less regular variation. Analysing Cp
shapes are more in evidence with a wind angle of 90°: square plan variation in direction 2 with a wind angle of 90° (Fig. 19(c),(d)),
structures have the greatest Cp,m values (Fig. 19), a reverse of the it can be noted that the Cp,m values in the detached regions of
behaviour with a wind angle of 0°. The circular plan structures rectangular plan structures are lower than in the other geometries.
F. Rizzo et al. / Engineering Structures 33 (2011) 445–461 461
In conclusion, in the separation zone, with wind angle of 0° and angles. These coefficients will represent a help guide to design
90°, rectangular plan structures have Cp,m values lower than the roofs with hyperbolic paraboloid shape.
other geometries, while, in the orthogonal regions, this behaviour The results show that the double curvature structure of roofs
is opposite. causes a complex aerodynamic behaviour and it’s impossible to
Wind vortex shedding effects determine local actions of ignore the three-dimensional effects of vortex shedding for wind
positive and negative pressure which result particularity high, angle as 45°. This aspect has been proved also by three dimensional
close to roof edge in the direction of incident wind. The results Computational Fluid Dynamic analyses carried out for all the
described in the paper allow to define some almost constant areas geometries tested in wind tunnel; these results will be shown in
of pressure (positive/negative) which are similar to those reported a future paper.
in the regulations (for other type of roofs) to be used in the check of
secondary elements. Local pressure effects have a great relevance
References
in the applied-engineering field for the sub-horizontal roof
design.
[1] Melchers RE. Structural reliability. Elley Horwood Ltd; 1987.
[2] Majowiecki M. Snow and Wind experimental analysis in the design of long-
3. Conclusions span sub-horizontal structures. J Wind Eng Ind Aerodyn 1998;74–76:795–807.
[3] Lazzari M, Majowiecki M, Vitaliani V, Saetta A. Nonlinear F.E. analysis of
The purpose of this research project was the study of Montreal Olympic Stadium roof under natural loading conditions. Eng Struct
aerodynamic behaviour of hyperbolic paraboloid shaped roofs and 2008.
[4] Lazzari M, Saetta A, Vitaliani R. Non-linear dynamic analysis of cable-
their structural response. This geometrical shape is specific to
suspended structures subjected to wind actions. J Comput Struct 2001;79(9):
design tension structures of cables nets which are particularly 953–69.
adapt to cover large spans because they are light and characterized [5] Lazzari M, Vitaliani RV, Majowiecki M, Saetta AV. Dynamic behaviour of a
by innovative technical systems and have high performances. tensegrity system subjected to follower wind loading. Comput & Structures
2003;81(22–23):2199–217.
For about twenty years, this structural typology hasn’t been [6] Majiowiecki M. Tensostructure: Design and control 1994 [in Italian].
widespread and has been even ignored by building regulations. [7] Simiu E, Scanlan RH. Wind Effects on Structures 1996.
Even now in Italian or European codes as well as in the most [8] D’Asdia P, Rizzo F, Sepe V. Wind action and optimized shape of a tension roof:
important international codes (as, for example, the USA codes) Pescara’s roof stadium. In: Proceeding of INVENTO 2006, 2006 [in Italian].
[9] Biagini P, Borri C, Majowiecki M, Orlando M, Procino L. BLWT tests and design
there isn’t information about preliminary design or about wind loads on the roof of the New Olympic Stadium in Piraeus. J Wind Eng Ind
loads. Aerodyn 2006;94:293–307.
In order to reach the above-mentioned goal, wind tunnel test [10] Bartoli G, Costanzo C, Ricciardelli F. Characterisation of pressure on build’s
have been done on structure with an optimal geometry obtained surface with a separate flow. In: Proceeding of INVENTO 2006, 2006 [in Italian].
[11] Gumbel EJ. Statistic of extremes. N.Y.: Columbia University Press; 1958.
from a sample of the different analysed geometries. Three different Cook & Mayne, On design procedures for wind loading, Building Research
shapes of plans (rectangular, square and circular) have been chosen Establishment, Garston 1978.
in order to take into account the most important market demands. [12] Gumbel EJ. Statistic of extremes. Columbia University Press; 1974. Lieblein
J, Efficient methods of extreme value methodology, Report 74-602, National
Experimental tests have been performed in the CRIACIV’s wind
Bureau of Standards: Washington.
tunnel in Prato (Italy) on twelve different geometries and the [13] Rizzo F. Tenso-structure for Pescara’s roof stadium. 2005, Pescara, Italy, April;
pressure coefficients were obtained for sixteen different wind 2005 [in Italian].