You are on page 1of 17

A new roof grid to cover a large area:

the Wierendeel solution.


F. ESCRIG
Prof. of the School of Architecture of Sevilla.
J. SÁNCHEZ
Prof. of the School of Architecture of Sevilla.

ABSTRACT: When we planed to solve a great roof with a tubular grid by means of
rigid joints instead of pinned connections we thought that the main problem could be local
tension concentrations and then proposed detailed analysis as well as model test to
determine if the plastic concentration around the welded joints was dangerous. The
model never was constructed but the considerations achieved by means of the analysis
resulted in a new design that advances in the state of the art.
In this paper we discuss the problems found in the construction, welding and stress
distribution studied by different methods, mainly by Finite Element Method and full scale
tests.
Finally the constructed grid structure confirmed our suppositions and we arrived at
the conclusion of the good behaviour of the selected mesh. In addition we show the
complete process of works from the beginning to the final finished roof.

1. DESIGNING THE ROOF. The problem of the oval that, as a masterfully


materialized in the Roman Coliseum, is that it is
complex. While spherical caps develop equally
In 1999 we received the commission of covering
distributed stresses, the elliptic ones suffer big
an existing open-air sports area.
variations in the stress, leading to uneven use of
The first consideration was to find the most materials.
adequate shape that included beauty, economy,
The inflated solutions in commonly seen in the
safety and adaptation to the existing building without
projects of Geiger-Berguer were not possible here
interfering it. The site consists of a 3000 seater oval
since the roof was a cover and not a enclosure.
velodrome perimeter with 145 and 114 m axes as
Many of those solutions have also need a high level
shown in Fig 1.
of maintenance.
The original alternative considerations were the
The solutions with tensile membrane are always
shown in the Fig. 2, were all discarded because
spectacular and we preferred a “quiet” and “soft”
their complexity or lack of “image”.
image. Furthermore the client discarted tensile
The preliminary design presented by this team membrane solutions.
was the shown in the Fig. 3 and under this brief,
The mesh of metallic tubes was the only
we began the development.
solution to be considered. Our team has always

Fig. 1. Comparative dimensions between the Roman


Coliseum (right) and Velodrome of Dos Hermanas (left). Fig.2. Different shapes considered.
1
Fig. 4. State of the site before the new construction.

Fig. 3. Preliminary design.

avoided the employment of triangular space


frames because of the expensive of the joints. In
the relationship of final price/m2 of structure we have
demonstrated that the triangular meshes with
prefabricated articulated joints are more expensive
than those with rigid welded joints. The problem is Fig.5. Geometric layout of the contour.
the difficulty of assembling pieces toghether when
we use rigid joints for large roofs.
Another aspect of the design that interested us
was the total independence between the roof and
the existing building. We did not want to involve
ourselves in a flimsy construction, and as the area
around the building was very scarce, we could not
extend in the perimeter (See Fig. 4). It also
interested us to give a floating impression for the
roof. We decide to place four big external supports,
very thin at the top, located in the vertexes of a
square of 91x91 m2, and to support upon them an
autonomous roof (Fig. 5). With this choice we
discarded the possibility of a ellipsoidal cover and Fig.6. Cutting for the projection in plant of the intersection
we decided for an intersection of two cylindrical of the two cylinders.
surfaces, the larger with a radius of 99.6 meters
and the smaller with a radius of 173 m This would
provide us a geometry sufficient for the plan
projection (Fig. 6), and that it would be mapped onto
the existing perimeter. Those cantilevers would not
suffer big forces since, due to the effect of the
bending, the stress transmission would follow paths
that are intuitively marked in the Fig. 7.
Fig.7. Transmission of the lines of force.
2
Because the low rise profile of the cylindrical The foundations also had to be well designed,
surfaces, the horizontal stresses are very big and even though the ground was good, and better in
therefore the necessary supports are heavily loaded the reality that predicted in the geotechnical reports.
in the horizontal direction at the top. These the piles could not relied upon to generate traction
horizontal forces in the top give a very but overall weighttrust was sufficient to balance the
importantmoment in the base of the order of 20 uplift. Furthermore friction along the pile provided
thousants Tonxm, that demands the supports to be additional safety ( See Fig. 10). As general rule we
big in plan. They should also be heavily weighted to would say the foundation was generously designed
diminish horizontal ractions at the base. However due tothe large safety coefficients used. The result
these supports need only a small cross-sectional is that each platform is tied by eight piles of Ø150
area in the top region since the vertical loads and cm and 20 m of depth.
the shear forces can be easily carried. Therefore
we decided to use some big pyramidal piers (Fig. 8
and 9).

Fig.8 and 9.One of the four supports. Fig. 10. Position of the piles below the platforms.

Fig.12. Transverse section of the internal box girders (a), and of the boundary box girders (b).
3
As for the roof, the concentration of stresses in
the head of supports comes from the main arch,
the interior intersection of the two cylinders. The
peripherals arches transport efforts smaller forces
than these (Fig. 11). These arches should be
dimensioned in accordance to the forces, but they
are of great size in a warped curve of 91 m span.
The boundary arches are also warped curves but
they could be in smaller sections. Nevertheless we
decided, for aesthetic and construction reasons,
to use the same section for all the arches, a box
girder of 2.5x.2.0 m (Fig.12)

Fig. 13. Connection of the mesh to the box girders.

2. DESIGNING THE MESH.


Another consideration was to decide what kind
of mesh to use (Fig.15). Rigid joints were preferable
for us than pinned ones and, un-triangulated beams
were better as we cited before. If we compare in
Fig.16 a plane beam with a span similar to the
central beam of our design, the difference is
obvious. The triangulated beam is better and the
Vierendeel beam has poor results in terms of
weight. Nevertheless if we span the same distance
with an arch, the results are different. The
triangulated beam is worst if we considere the
Fig.11. View of mesh. weight. In all cases we have considered the joints
to be rigid and the tubes dimensioned with the same
It is posible that the original concept in the size for the maximum load at the pesimum section
preliminary design could have been achieved using the elastic limit (Fig. 17).
without the boundary arches. However the eventual We had an other question to study. Could the
construction solution to finish the cover with a single depth influence the results of the arch design? The
line and the necessity of a gutter would have Fig.18 shows how the single layer is the optimum if
changed the appearance of the edge anyway. we do not consider the overall buckling. But to check
Furthermore, wind effects were more considerable this behaviour is complicated and not included in
for a thin edge. Therefore, the use of a boundary the usual analysis computer programs. We thus
arch was beneficial. tested the general buckling shell behaviour. See the
For the mesh we chose a double layer of References 1 and 4.
continuous tubular profile and a depth between the
two layers of 250 cmas an previous design, without
any triangulation type. To build for minimal modules
we designed appropiate constructive devices:
stiffeners, telescopic connections, half pipes, etc
(Fig.13).
The intersection of two cylinders finally led to
several advantages:
a. We can support the roof on only four piers.
b. We can achieve an optimum mesh because
all their components are circular.
c. All the forces are conducted to the borders
that can be especially designed for them. Fig. 15. Internal and border box girders.
4
Fig. 16. Comparison between results of Vierendeel and triangulated girder.

Fig. 17. Comparison between a Vierendeel and a triangulated arch.


3 9

h 8h4 −1 3. ANALYSIS OF LOCAL BEHAVIOUR.
σ cr = E m b R 2 To analyse the node behaviour of the mesh
L
selected we planned a test that never was carried
being with hm the equivalent depth and scr the
out but that was studied by FEM. It was difficult to
tension which causes buckling
reproduce a considerable part of the mesh to be
A (pipe _ section) 69,13 × 2 tested and we proposed to build a model of a
hm = = = 0,552cm
d (whide_ of _ the _ mesh) 500 module of 5 m wide by 2.5 m depth with diagonall
hb = Equivalent_ Inertia_ Modulus. loading (Fig.19).
1 1
⎛ 12I ⎞ ⎛ 12 × 2 × 69,13 × 1252 ⎞ 3
3
hb = ⎜ ⎟ =⎜ ⎟ = 37.29cm
⎝ d ⎠ ⎜ 500 ⎟ The pipes used were the as those in the real
⎝ ⎠ structure and the process was the described below.
L = Lenght_ of _ the _ cylinder =10000cm
1. -To obtain, by means of finite element analysis
R = CylinderRadius = 10200cm
the structural behavior of this element.
E = 2,1 × 106 kg / cm2 PE ( Elastic load), the one that makes reach the
σ cr = 11.687kg / cm2 elastic limit in some significant sections where we
attach the strain gauges.
Ιf we had used a real depth of 200 cm:
PR (Breaking load), the load reaching the ultimate
σcr =264.50 Kg/cm2, wich is less than the Elastic
limit in some of the previous sections.
Limit of the steel.
5
Fig. 18. Comparison between different depths of arches.

2. - To load and then unload from, PE in gradual


steps while taking intermediate readings.
3. - To load and unload in cicles, with target loads
following the sequence:
Intermediate readings are taking throughout the
cycles.
4. - The deformed geometry of the tubes is
recorded.

⎛ P − PE ⎞ ⎛ P − PE ⎞
PE + ⎜ R ⎟ , PR , PE , PE + ⎜ R ⎟,
⎝ 4 ⎠ ⎝ 2 ⎠
Fig.19. Model to be tested.
⎛ P − PE ⎞
PE , PE + 3⎜ R ⎟ , PE , PR
⎝ 4 ⎠ The model in horizontal was placed in the
horizontal position in such a way that it can not
The model to carry out in the compression test buckle out of plane. If this is guaranteed, the vertical
machine is shown in the Fig. 19. position was also admitted.
6
Loads are applied in steps of five tons till the
arrival of the elastic limit. This was obtained by the
SAP2000 FEM. The diagonal load necessary to
achieve this elastic limit was evaluated as 25 Tons.
Figs. 20 and 21 show the result obtained for local
tension and displacements.

Fig.22. Mathematical model by FEM

Fig.20. Local stresses (Von Misses criterium)

Fig.23a. Detail of joint A .

Fig.21. General displacements.

After this we carried out a more complete


analysis by means of the ANSYS FEM (Fig.22).
Fig.23 shows discretitation in the joints A and B. Fig.23b. Detail of joint B.
Fig.24 shows the stress-strain curve used for
the element.
With these parameters we checked that the
curve load-displacement is linear till 20 Ton were
breaking of a joint is obtained (Fig.24).
Really is at 15 Tons that the maximum stress
permitted is produced, as we can se in the Fig.26
and 27.
Resuming the limit load of 20 Tons have been
accepted as the linear limit and is the used for
dimensioning the structure by SAP2000 FEM.
Figure 30 shows an imposible stress state if we
arrive at 30 Ton. An iterative analysis would be done
by eliminating elements broked. Fig.24 Bilinear curve stress-strain ó in K/cm2 å in %
7
Fig. 28. Relationship between actual stress and elastic
Fig.25. Stress-strain Curve in the worst element.
limit stress unit P=15 Ton. in the joint B.

Fig. 29. Relationship between actual stress and elastic


Fig.26. Maximum Von Misses stress (N/cm2) in the joint limit stress unit P=20 Ton. in the joint B.
B (20T)

Fig.30. Relationship between actual stress and elastic


Fig.27. Maximum Von Misses stress (N/cm2) in the joint limit stress unit P=30 Ton. in the joint B in a not iteralive
A(20T) analysis..
8
4. LOADS TO BE CONSIDERED. 5. ANALYSIS OF THE SPATIAL GRID.

Load Case 1. Self weight. Done automatically As a result of the analysis carried out we have
calculated by the analysis Programme (SAP2000). arrived at the stresses contained in the Figs.32 and
Load Case 2. The weight of the steel sheet and 33.
insulation material. 20 K/m2 The tubes used are of three types:
Load Case 3. Snow load and temporary loads. a. Principal directions φ 355x6.3 mm.
60 K/m2. b. Secondary directions φ 273x6.3 mm.
Load Case 4. Machinery. 40 K/m2 c. Special tubes φ 355x7 mm.
Load Case 5. Wind loads. Tested in wind tunnel d. Vertical pipes between two layers φ 355x6.3
(Reference 8) according with the expression mm.
q=150 (cpi - cpe). In the Fig. 34 we show the complete grid from
Figure 31 shows the coefficients cpe. different points of view.
Cpi is the internal pressure and we considered
a value of 1.00.
Load Case 6. Dynamic Analysis. We considered
only the first five modes.
Load Case 7. Thermal changes of +/- 30ºC.
We will combined these cases according the
codes of practice. The SAP-2000 Programme also
checked for local buckling.
Fig. 31 shows the wind tunel test.

Fig. 32. Wind tunel test.

Fig. 31. Wind load coefficients.

9
Fig.33. Stresses in Kgm. in the top and bottom layers in the directions shown.

Fig. 34. Stresses in Kgm in the top and bottom layers in the directions shown.
10
Fig. 35. General internal view of the mesh.
11
7. ELABORATION OF THE CONSTRUCTIVE
SOLUTIONS.
Up to now all the details have been
schematically described. We are now going to
explain the details more extensively. On the cover
mesh we placed a panel sandwich of steel sheet
with the profile that is shown in the Fig. 35.

Fig. 37. Encounter of gutters and supports.

The section of the gutter of the girders continues


now by the external face of the supports, so that
Fig. 36. Profile of the sheet to finish the roof. the water slide by them until the ground where the
The devastating effect that even light winds collectors will be located. From this point until the
could cause on the cover sheet when it ends in a basis the supports would be massive; massive in
thin border was a worry because it causes forces fact; but we want to give them certain volatile
up to ten times the forces of wind in the rest of the aspect. The warped surfaces are ideal for it. The
surface, and we decided to put separated their gutter of drainage seems to define two leaves that
anchorage of the external line of the structure the open up as two wings. The final result is shown in
most possible. On the other hand the quantity of the Fig. 36.
rain water to drain is such that we needed gutters In all the faces of the girders that faced to the
of big dimensions. We have opted to integrate the interior of the pavilion, we have located some circular
gutter in the structural form of the section and to openings that lighten their appearance and make
give it, starting from this idea, the form that we have less oppressive the walk through those long tunnels
seen in the Fig.12. for the maintenance and installation of services. It
The advantages are the following ones: is necessary to keep in mind that these boxes are
prepared with interior suitablefor the transit of people
a. We increase the area to distribute forces.
(Fig.37). From these perforations one would have
b. We diminish the buckled surfaces that we will to the whole mesh without need of special elevators
analyze now keeping in mind the form. or additional walkways (Fig. 38).
c. We give a dynamic appearanceto the roof that
otherwise would seems very heavy.
d. We have a good gutter.
In the lower points of the roof are the top of the
supports. In them, besides the concentration all
the forces will converge and all the water due to
rain and the accesses to the corridors inside the
girders. The head of the support that responds to
those necessities is designed as it is shown in the
Fig. 36.
To facilitate the insertion of the three girders in
the faces of the support they have been welded to
20 mm. thick sheets together to the general
reinforcement of the support and that have served
as framework. (Fig. 9). Fig. 38. Passage through the interior of the box girders
with windows and gusset plates.
12
9. DEFINITION OF THE WORK.

In such a modular construction, the pieces have


to connect together with an absolute precision.
They are permited only a very small percentage
tolerances. Nevertheless it is necessary to keep in
mind that, although it was possible to manufacture
accurately, the thermal changes, the handling in
work, the deformations for internal tensions and
many other factors, limit the precision to values
Fig. 39. Aspect of the built boxes seen from the field.
impossible to improve. For lengths of 100 m
8. ASSEMBLY OF THE STRUCTURE. tolerances of 1/1000 implies 10 cm of error. So that
these tolerances do not overall affect to the
In our project we include an assembly sequence accuracy of the structural geometry and all the
that was accepted by the builder company. It unions of the tubes had collars that absorbed the
consisted of a series of lifting by steps and tolerances (Fig. 41). The same was also applied
temporary supports to connect the big modules of for the joints of box girders to tubes (Fig. 42).
the structure. The systematic confirmation of positioning on
To eliminate the need for provisional supports site was essential from the geometry of the head
we had foreseen the placement of hydraulic jacks of the supports until the placement of each one of
that would allow a gradual descent of all the the modules and the bend with which they settle,
supports at the same time, controlling the descents since during the lifting this it can change.
and checking if they coincided with those calculated.
This is the decisive stage and the most critical.
The provisional towers were fixed and the hidraulic 10. ASSEMBLY OF THE BOX GIRDER
jacks would be placed at the bottom in the lower ARCHES.
part to permit the descent begin (Fig 40).
The builder also proposed to substitute the The arches arrived to site coming from 800 km
flanges of connection with high strength bolts by in in sections of approximately 25 m (Fig.43) and they
situ welding (Fig. 40 and 41). were joined in groups of two on the ground. They
were raised and supported on the temporary towers
(Fig 44). They were positioned correctly before
being welded and united in the continuous arch (Fig
45).
We began with the cantilever arches because
of the convenience of clearing the area of works
(Fig.46). Later in the same manner the interior
pieces were joined and mounted in two sections.
(Fig 47). In the Fig 48 we can see looseness and
Fig. 40. Jacks placed at the basement of temporary lack of welding that now it was necessary to review,
supports. unite and to fix the arches rigidly to the supports.
(Fig 49). A view from the installed arches can it be
seen in the Fig.50.
We repeated the operation on the south side.
The border arches of the main faces were
mounted last, even later than the mesh, so that they
did not interfere in the lifting of the mesh; in fact not
Figs. 41 and 42. Collars of union of the tubes and box
even they were taken to work until this moment for
girders. not representing a nuisance on the ground.
13
Fig. 43. Transport.

Fig. 47. Assembling the interior box girders

Fig. 44. Provisional towers.

Fig. 48. Clearance to be completed with welding.

Fig. 45. Union of the arches.


Fig. 49 .Dificulties of adjusting the girders to the piers.

Fig. 46. Assembling the cantilever box girders. Fig. 50. View from the corridors during works.
14
11. ASSEMBLY OF THE MESH.

The mesh assembling is illustrated in the


sequence of the Fig. 50, 51 and 52. The final result
of the mesh can it turns in the Fig. 53.

12. STEEL ROOF SHEET .

To finish the roof a sandwich sheet injected with


glass fibre was placed upon the mesh , and the
interior face was perforated to improve the acoustic
conditions. Also placed was a poly-carbonate panel Fig.54. View of the finished mesh.
in the central part to provide natural illumination. An
interior and an aerial view of the finished roof can
be seen in Figs. 55 and 56.

Fig. 51 Placing the first module of de mesh. Fig. 55. Image of the finished interior.

Fig.52. Intermediate phase of assenbling.

Fig. 56. Image of the air view of the finished roof.

Fig. 57 shows a sequence of assembling the


roof.
Fig. 58 shows the actual appearance ready to
be used.
In the future a lateral enclosure will be provided.
Fig.53. Secuence at the final phase of a ssenbling.
15
Fig. 57. Construction sequence.
16
Fig. 58 Actual appearance with the instalation ready to be used.

13. REFERENCES.

1.- Buchert, Kenneth P. “Buckling of Shell & Shell 9.- Han,Q,H; Yang,Z; Liu,X.L. “Numerical Model,
like Structures” K. P. Buchert & Associates. 1973. Buckling Analysis and Ultimate Load Capacity of
the Single-Double-Layer Reticulated Dome” Space
2.- Chilton, John “Space Grid Structures”
Structures 5. Parke and Disney Ed. Thomas Telford
Architectural Press. 2000.
2002. pp. 1445-1454.
3.- Eekhout, M. “Las estructuras tubulares en la
10.- Ishii, Kazuo. “Structural Design of
arquitectura” CIDECT. Geneve.
Retractable Roof Structures” WIT Press. 2000.
4.- Escrig, Felix “Pandeo de Estructuras”
11.- Ishii, Kazuo. “Membrane Structures in
Publicaciones de la Universidad de Sevilla. 1986.
Japan” SPS Publishing Company. 1995.
5-Escrig,F. Sánchez,J. “Great Space Curved
12.- Meseguer, J. “Aerodinámica de
Structures with rigid joints”. Theory, Design and
Instalaciones Aeroportuarias”. Fundación Aena.
Realization of Shell and Spatial Structures. IASS.
2000.
Nagoya 2000.
13.- Ramaswamy, G.S; Eekhout,M; Suresh,G.R.
6-Escrig,F. Sánchez,J. “Cubiertas para
“Analysis, design and construction of steel space
instalaciones deportivas.Velódromo de Dos
frames” Thomas Telford. 2002.
Hermanas”. STAR Books. Sevilla 2003.
7.- Escrig,F. Sánchez,J. Valcarce,J.P.”The
Roman Oval”. Fifth International Conference on
Space Structures. Univ of Surrey. UK. Thomas
Telford. 2002.
8.- EUROCODE 1. “Basis of Design and Actions
on Structures. Part 2-4:Action on Structures: Wind
Actions” 1995.

17

You might also like