EXTENDED ABSTRACT

EXTENDED ABSTRACT

REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

by
Francisco Manuel Henriques de Sena Cardoso

INTRODUCTION

The purpose of this study is to present reinforced concrete shells as a relevant and valuable structural
solution.
In designing any structure, designers always aim at achieving economy by minimizing costs within the
constraints of functional and aesthetic requirements. They thus endeavour to find newer construction
materials witch are cheaper and stronger, or with the available materials, try to evolve new forms that
resist the loads more efficiently that when the structure is designed in a conventional form.
One of the most efficient known structural forms is shells and they are a perfect example of achieving
strength through form as opposed to strength through mass.
Reinforced concrete shells have many applications in several civil engineering fields and may be used at
roofs, water tanks, pressure vessels, tunnels, cooling towers, canals, foundations, military/sports/airport
facilities and dams among others.
In this abstract, it is presented their main characteristics, advantages and disadvantages; it is described
their structural behaviour and load resistant mechanism; it is also presented the design methodology
proposed by Eurocode 2 to ultimate and service limit states with a short reference to ACI Building Code
and to IASS Recommendations. Finally, it is presented an example of a shell roof structure developed to
illustrate the previous concepts.

GENERAL

Shell may be defined as the bidimensional curved element that has tendency to carry loads mainly in
direct compression or tension, rather than in bending and in shear. This definition implies that the
thickness is small compared with its other dimensions, but it does not require that the smallness be
extreme.
The main advantages of reinforced concrete shell structures are the following: their efficient resistant
mechanism allows that the maximum structural advantage, in terms of both internal forces and
displacements, is obtained with a minimum of materials; their architectonic value is characterized by
being able to cover large columns-free areas with an aesthetical and functional structure.
The figure 1-1 is a good example of how efficient shells can be. It consists of two sheets of thin plastic
with the same span, one sheet is flat and the other formed into a double-curved shell arch. Despite the
same amount of material was used in both cases the shell arch carries a weight 30 times heavier than that
on the flat sheet with a minimum of deflection (opposed to the large deflection of the flat sheet).

1
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

Figure 1 – Model demonstrating the efficiency of a double-curved shell compared to a similar flat sheet.

Many of the shells disadvantages are related with the construction stage. Traditionally, shell construction
is expensive because it is labour intensive (and consequently with high wages costs) and it has high
formwork costs due to its complex nature and low reuse capacity. Another important disadvantage is the
complexity of shell analysis increased by the multiplicity of shell geometries. Nevertheless, the
development of structural analysis computer programs based on the finite element method is softening
this difficulty although some designers still avoid, for lack of confidence, shells structures.
Historically, in civil enginnering construction, shells in the form of dome roofs have been used from as
early the Pantheon in Rome (AD 120-124) to the domes of such monuments as Hagia Sofia in Istambul
(AD 532-537) or Santa Maria del Fiore in Florence (1420-1434). These domes were constructed either
from masonry or some form of unreinforced concrete, materials strong in compression but relatively weak
in tension.
With the advent of reinforced concrete, a mouldable material strong in both compression and tension it
has encouraged, from around the beginning of the twentieth century, the design and construction of
remarkable aesthetical shell structures throughout the world. Among the greatest concrete shell builders
are, indeed, names as Pier Luigi Nervi, Ove Arup, Eduardo Torroja, Nicolas Esquillan, Félix Candela,
Heinz Isler and works like CNIT in Paris (1956-1958), TWA Terminal in New York (1962), and Opera
House in Sidney (1958-1973) are examples of the highest expression of this structural solution

Figure 2 – Hagia Sofia in Istambul (Span: 32,5 m; Figure 3– CNIT in Paris (Designer: Nicolas Esquillan;
Thickness: 0,7 m). Span: 218 m).

Figure 4 – TWA Terminal in New York (Designers: Eero Figure 5 – Sidney Opera House (Designer: Ove Arup;
Saarinen and Boyd Anderson; Span: 96 m). Thickness: 0,15 m).

2
 EXTENDED ABSTRACT

A shell is bounded by two curved surfaces, the faces and the surface which passes midway between the
two faces is called middle surface. So, to fully describe the geometry of a shell it is enough to know the
shape of the middle surface and its thickness at all points. This is the key reason why shells are usually
classified for the geometric properties of its middle surface. The two main classification systems are
based on the following criteria: Gaussian curvature signal (Product between the two principal curvatures)
or surface generation method.
According to the first classification system, shells may be classified as developable, synclastic or
anticlastic.
- The developable shells are singly curved shells having zero Gaussian curvature. Cylinders and cones
belong to this group.
- The synclastic shells are doubly curved shells having positive Gaussian curvature. Domes, elliptic
paraboloids and circular paraboloids belong to this group.
- The anticlastic shells are doubly curved shells having negative Gaussian curvature. Hyperbolic
paraboloids and conoids belong to this group.
- Partly synclastic and anticlastic shells are doubly curved shells having variable signal Gaussian
curvature. Funicular and corrugated shells belong to this group.
According to the second classification system, shells may be classified as shells of revolution, shells of
translation and ruled shells.
- In revolution shells, the middle surface is obtained by revolving a plane curve about an axis. Circular
cylinders, cones, most domes and hyperboloids of revolution belong to this group.
- In translation shells, the middle surface is obtained by moving a plane curve along another plane curve
(the two planes containing the two curves are at right angles to each other). Cylinders, cones, elliptic
paraboloids and hyperbolic paraboloids belong to this group.
- In ruled shells, the middle surface is obtained by the translation of straight lines over two end curves.
Hyperbolic paraboloids and conoids belong to this group.

ANALYSIS OF SHELLS

As previously stated, the shells tend to resist any loading by development of in-plane internal forces –
membrane forces – associated with low relevant bending moments and shearing forces that, in general,
only gain importance near the edges. This is the reason why shell structures are so efficient and require a
minimum amount of material, membrane forces are equivalent to stresses uniformly distributed over the
entire cross section and hence the section is fully utilized opposed to bending moments that are associated
with stresses varying linearly over the thickness and hence the portions near the middle surface remain
under-utilized.
This resistant mechanism is justified by the fact that shell elements have considerable in-plane stiffness
but they are quite flexible for resisting bending and shear, i.e., if by supposition the shell resisted external
loads by developing, mainly, bending moments and shearing forces, considerable deformations would
arise contradicting the Castigliano’s theorem which states that the exact solution of a stress problem
corresponds to the equilibrium solution that minimizes the deformation energy.

3
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

The way as a doubly curved shell resists loads may be visualised as a doubly hoop effect along the
directions of principal curvatures and, in fact, if bending moments and shear forces are neglected the
equilibrium in the direction of the normal to the shell may be written as:

Nφ Nθ
+ = −p r [1]
r1 r2

Where φ and θ are the lines of principal curvature.

In the major part of cases, bending moments and shearing forces gain some importance near the edges
due to the fact that the compatibility between the displacements of the shell edges and the displacements
of the boundary elements only can be made through these kinds of forces.
There are two possible methodologies for solving shell problems: bending theory and membrane theory.
The bending theory is the only one through the exact solution can be obtained. It considers ten unknown
internal stresses resultants acting on the shell element: membrane forces - nxx, nyy,, nxy and nyx;
bending/twisting moments - mxx, myy, mxy and myx; shear forces - vxz and vyz (Figure 6).

Figure 6 – Internal stresses resultants in a shell element.

The problem can only be solved using the equations of equilibrium, the kinematic relations and the stress-
strain relations deriving into a complex problem with little pratical interest (its governing expressions
present eight-degree differential equations). Despite the fact that for a few analysis cases the analytical
solution is already found it is recommended to employ numerical and computational methods to obtain it.
The membrane theory may be faced as a particularization of the general bending theory with a key
difference: it is assumed that the bending/twisting moments and shear forces are zero or negligible. This
hypothesis rather simplifies the problem and with exception for the zones near the edges and for a few
shell geometries, it shall be considered a good approximation of the exact solution. Membrane theory may
be used at the conceptual stage and in some cases even at the design stage.
Despite the importance and value of the mathematical approach to a shell problem some authors like
Heinz Isler, Félix Candela, Lobo Fialho developed some amazing experimental techniques characterized
by the purity of the background concept to find the ideal shell form.

4
 EXTENDED ABSTRACT

Heinz Isler is a Swiss engineer who created a rich repertory of shell shapes inspired by his peculiar
methods. He developed three main methods:
- the pneumatic method that consists on inflating a rubber plastic membrane over a wooden basis with a
opening similar to the real structure horizontal plan. The resulting surface is a shell shape that resist in
pure tension to a internal pressure loading and presents a optimum buckling stability.
- the hanging membrane method that consists on suspending a cloth covered with a plastic resin between
supports in the same relative position as in the real structure and leave it to deform his own weight. Under
this load the membrane will take a shape that only has tensile forces. After the resin gets frozen the
inverted shape is the funicular form (pure compression state) for the dead weight loading.
- the expansion form method that consists in passing a expanding foam through a tube creating a doubly
curved shape. This method isn’t characterized by a particular stress state but is has been used in several
aesthetical shells.

Figure 7 – Pneumatic method model. Figure 8– Hanging membrane method model.

Figure 9 –Pneumatic inspired shell. Figure 10 – Hanging membrane inspired shell.

Félix Candela was a Spanish engineer known by his investigation about the hyperbolic paraboloid shell
type. He discovered hundreds of exciting and versatile configurations using this structural form
demonstrating that its exposed edges can be completely freed of normal and tangential stresses and hence
freed from beams.
Lobo Fialho was a Portuguese dam researcher who developed several methods to find the structure ideal
shape. One of them consisted in adding on a topographic terrain model an elastic membrane at the
implantation site. After he filled the model with water and let the membrane deform under the hydrostatic
pressure. The inverted deformed shape of the membrane corresponds to the funicular geometry of a shell
acted by the hydrostatic pressure. The other important method was similar to the last one but the
hydrostatic pressure was reproduced by a weight system.

5
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

Figure 11 – Hyperbolic paraboloid shell designed by

Figure 12 – Lobo Fialho physic model.
Félix Candela.

REINFORCED CONCRETE SHELL DESIGN

According to Eurocode 2, the design of a reinforced concrete shell is based on a sandwich model
(considering the element to be cracked). In this sandwich model three layers are identified: the two outer
layers resist the membrane actions arising from nEdx, nEdy, nEdxy = nEdyx, mEdx, mEdy and mEdxy = mEdyx while
the inner layer carries the shear forces vEdx and vEdy leading to a design solution similar with the
represented in figure 13.

Figure 13 – Sandwich model design solution.

The thickness of the outer layers is defined as the minimum of twice the edge distance of reinforcement
and twice the distance from the centre of gravity of the reinforcement to the mid-plane of the element.
The thickness of the inner layer is obtained subtracting the outer layers’ thicknesses to the element total
thickness.
The first analysis step is to convert the membrane forces and bending/twisting moments of the element in
statically membrane forces acting just in the outer layers using the following expressions:
1] In cases for which no shear reinforcement is required to resist vEdx and vEdy:
z x − y xs m Edx
n Edxs = n Edx ⋅ + [2]
zx zx
z x − y xi m Edx
n Edxi = n Edx ⋅ − [3]
zx zx

6
 EXTENDED ABSTRACT

z y − y ys m Edy
n Edys = n Edy ⋅ + [4]
zy zy
z y − y yi m Edy
n Edyi = n Edy ⋅ − [5]
zy zy
z xy − y xys m Edxy
n Edxys = n Edyxs = n Edxy ⋅ − [6]
z xy z xy
z xy − y xyi m Edxy
n Edxyi = n Edyxi = n Edxy ⋅ + [7]
z xy z xy

2] In cases for which shear reinforcement is required to resist vEdx and vEdy:

z x − y xs m Edx 1 v 2Edx
n Edxs = n Edx ⋅ + + ⋅ ⋅ cotg (θ') [8]
zx zx 2 v Ed0
z x − y xi m Edx 1 v 2Edx
n Edxi = n Edx ⋅ − + ⋅ ⋅ cotg (θ') [9]
zx zx 2 v Ed0
2
z y − y ys m Edy 1 v Edy
n Edys = n Edy ⋅ + + ⋅ ⋅ cotg(θ') [10]
zy zy 2 v Ed0
2
z y − y ys m Edy 1 v Edy
n Edys = n Edy ⋅ + + ⋅ ⋅ cotg(θ') [11]
zy zy 2 v Ed0
z xy − y xys m Edxy 1 v Edx ⋅ v Edy
n Edxys = n Edyxs = n Edxy ⋅ − + ⋅ ⋅ cotg (θ') [12]
z xy z xy 2 v Ed0
z xy − y xyi m Edxy 1 v Edx ⋅ v Edy
n Edxyi = n Edyxi = n Edxy ⋅ + + ⋅ ⋅ cotg (θ') [13]
z xy z xy 2 v Ed0

Where,
yxs, yxi, yys, yyi are the distances from the centre of gravity of the reinforcement to the mid-plane
of the element in the x and y directions, in relation to bending and axial membrane forces;
yyxs, yyxi, yxys, yxyi are the distances from the centre of gravity of the reinforcement to the mid-
plane of the element in the x and y directions, in relation to twisting moment and shear membrane forces;
zx e zy are the lever arms for bending moments in the x and y directions determined by: zx = yxs +
yxi e zy = yys + yyi;
zxy e zyx are the lever arms for twisting moments in the x and y directions determined by: zxy =
yyxs + yyxi e zyx = yxys + yxyi.

Then the outer layers are designed as membrane elements and the ultimate limit states consists in
providing the necessary reinforcement to resist the tension stresses and checking that the maximum
compressive stress does not exceed the maximum permissible compressive stress.
The required reinforcement area is given by the following expressions:

A sx F
= x [14]
sx f syd
A sy Fy
= [15]
sy f syd

7
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

Where Fx and Fy are the tensile forces obtained from figure 14 for each in-plane forces state.

The maximum permissible compressive stress is given by:

1] If the principal stresses are both compressive,

1 + 3,80α
σ cd, max = 0,85f cd ⋅ where α = nII/nI [16]
(1 + α )2
2] If a limit analysis has been carried out with at least one principal stress is in tension and no
reinforcement yields,

⎡ σ ⎤
σ cd, max = f cd ⋅ ⎢0,85 − s ⋅ (0,85 − ν )⎥ [17]
⎣⎢ f yd ⎦⎥

⎡ f ⎤
Where σs – Maximum tensile stress in the reinforcement and ν = 0,6 ⋅ ⎢1 − ck ⎥ .
⎣ 250 ⎦

3] If a limit analysis has been carried out and there is yielding of any reinforcement,

σ cd, max = ν ⋅ f cd ⋅ (1 − 0,0032 θ − θ el ) [18]

Where θel is the inclination to the x axis of the principal compression stress in elastic analysis and θ is the
angle of the limit compression field (principal compressive state) at ultimate limit state to the x axis.
This value must be compared to the quotient between the maximum compressive force (obtained from
figure 14 for each in-plane forces state) and the layer thickness.

Figure 14 – Tensile and Compressive forces for each membrane state.

8
 EXTENDED ABSTRACT

For the design of the inner layer the principal shear – vEd0 – and its direction ϕ0 should be evaluated with
the following expressions:

v Edo = v 2Edx + v 2Edy [19]

v Edy
tan (ϕ0 ) = [20]
v Edx

In the direction of principal shear, the shell element behaves like a beam and the appropriate design rules
should therefore be applied and ρl should be taken as:

ρl = ρ x ⋅ cos 2 (ϕ0 ) + ρ y ⋅ sen 2 (ϕ0 ) ≤ 0,02 [21]

Oddly, Eurocode 2 does not present any methodology to verify the ultimate limit state of buckling which
constitutes a serious omission since it is a very limitative check with special importance at shells due to
the fact of carrying loads through in-plane forces (compressive) making them very susceptible to the
buckling phenomena.
Following the methodology described by IASS, for shells in which the post buckling behaviour does not
govern (anticlastic shells) the linear critical load must be reduced exclusively to take into account the
effects of creep, reinforcement, cracking status and plasticity of concrete. In the end, the relation between
the real load and the effective critical load must be equal or greater than 1,75.
For shells in which the post buckling behaviour governs (synclastic shells, cylinders) besides the
previously referred effects, the critical load must be reduced to take into account the effect of great
displacements and geometry imperfections. In the end, the relation between the real load and the effective
critical load must be equal or greater than 3,5.
Besides the ultimate limit states, the designer should check the deformation and cracking service limit
states but the Eurocode 2 does not establish any special requirement to shell structures. However, it is
suggested to the designer to carefully study those limits because in some cases it may be important to use
more restrictive limits.
Several constructive requirements as the minimum concrete cover, minimum and maximum bar spacing,
minimum and maximum reinforcement areas are of great importance because in many situations the shell
thickness is controlled not by resistant or buckling issues but for durability requirements and
reinforcement itself may be controlled not by resistant reasons but to control shrinkage and temperature
actions.
Side by side with this evaluation of the Eurocode 2 requirements for the design of reinforced concrete
shells, it was analysed the principles existent in the ACI – Building Code Requirements and in the IASS –
Recommendation for Reinforced Concrete Shells and Folded Plates. The main conclusion is that these
documents concede designer a larger freedom choosing the security checking methodology, although,
they present more criteria and restraints not even mentioned by the Eurocode 2.
Finally, it was analysed a few matters related with the concrete shell construction stage. This stage
includes: 1] Formwork and cladding; 2] Reinforcement placing; 3] Concrete and concrete cast and 4]
Concrete finishes.

9
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

The formwork and cladding phase is characterized by the intensive labour and the complexity of the
temporary structure solution. The traditional method consists in making a structure of curved timber
beams or flexible metal trusses supported by light scaffolding. The curved beams define the shape and
support the light timber strips over which cladding panels are placed.
Less traditional methods are for instance, earth forms that consists in erecting a massive hill with the
shape of the future concrete shell or even inflatable forms.
Normally, doubly curved shells cannot be clad with rectangular or square panels and for instance, to
cover a circular dome, every panel has to be cut to different angles. Associated with all difficulties is the
fact that these elements have low reuse aptitude and it helps understand how economical relevant this
phase becomes.

Figure 15 – Traditional formwork method. Figure 16 – Earth form.

Although reinforcement small bars are preferred to large, it should be noted that bars of small diameter
are quite flexible and are subject to becoming permanently bent and displaced during the process of
construction. For this reason, small bars should be more often supported on the forms than would be the
case with bars of larger diameter. For constructions simplicity reinforcement should be placed along two
constant orthogonal directions.
About the concrete, it shall preferably be of wet-earth consistency. If it is too fluid, it will tend to slump
down the steep slopes, if it is too dry, proper compaction by vibration becomes difficult. The concrete
should have aggregate size controlled and it may be added some additives like plasticizers to improve
workability and retarders to delay the cure process. It is good practice to start concreting from the low
points and move upward toward the high points following a symmetrical pattern. The concrete
compaction cannot be made by the usual needle vibrators but, instead, with flat vibrating plates or
rotating vibrating plates.
Concrete placing may be also be made by guniting (preferably dry process) allowing a wider single form
slope. In this process sand and cement are mixed dry and pumped through a hose to a nozzle where water
is added under pressure from another hose. To avoid constructive shrinkage cracking, it is recommended
the application of fog spraying followed by the application of wet burlap mats.
By aesthetical or functional reasons the concrete surface may be coated. Interior coating elements should
be attached with hooks to the roof structure.
The prefabricated shells have several advantages, standing out the followings: 1] Assembly-line
techniques can be employed. This results in higher productivity and closer quality control. 2]
Construction is speeded up. 3] Less skilled labour in the site labour. 4] Only a small fraction of the in situ
case scaffolding is required.

10
 EXTENDED ABSTRACT

Among their main disadvantages stands out: 1] The transportation and erection limitations to size and
weight. 2] Joints introduced to keep the shells transportable dimensions. 3] It is not economically viable
unless a normalization and repetition factor is added.

a) b)

c) d)

Figure 17 – Several stages of concrete cast: a) placement, b) compaction, c) completed corner and d) final
appearance.

APPLICATION EXAMPLE

In the last stage of this work, it was developed a reinforced concrete shell roof for a sports facility with
the purpose of applying the reported concepts.
The structure should cover a column-free area with a span of 50 metres and after the conceptual stage the
adopted solution was a conoid shell with a thickness of 15 centimetres supported on the foundations and
on two edge beams/arches with 1metre and 1,2 metres heights.
The next step was to create a model of the solution in the computer software SAP2000 where besides the
structure characteristics, they were also added the load cases prescribed by Regulamento de Segurança e
Acções code.

Figure 18 – Reinforced concrete conoid model.

11
REINFORCED CONCRETE ROOFS – SHELL STRUCTURES

After running the program, it was made an analysis to stresses and displacements diagram and it was also
obtained the linear critical load.
The ultimate limit states analysis followed the determinations of the Eurocode 2, i.e., it was based on a
sandwich model and with the equivalent membrane stresses acting on the outer layers it was determined
the needed reinforcement and it was checked the maximum compression stress. In both cases, the
structure resistant capacity was much larger than the required which led to adoption of the minimum
reinforcement area in both directions. The maximum compressive stress did not reach 1/3 of the
maximum admissible compressive stress.
The inner layer shear resistance was also considerably larger than the acting one.
The service limit states analysis verifications did not present any problem and while the cracking control
was indirectly checked the maximum deformation did not reach half of the admissible one.
At last, it was analysed the buckling ultimate limit state. As predicted in the conceptual stage it was the
limitative verification and despite the high resistance showed to other limit states, this one is checked with
low tolerance levels. The first step of this verification was to get the critical linear load from SAP2000.
Then it was necessary to follow the exposed by IASS Recommendations and the effect of creep, cracking,
reinforcement position and concrete plasticity reduced the critical load to less than a half of the linear
critical load (it was not necessary to include the large deflections and geometry imperfection because it is
an anticlastic shell).
Comparing the acting load with the effective critical load, the safety factor turn 1,8 which is greater than
the minimum 1,75, so it verified the buckling limit state.

12